Advances in Human Factors of Transportation: Proceedings of the AHFE 2019 International Conference on Human Factors in Transportation, July 24-28, 2019, Washington D.C., USA [1st ed.] 978-3-030-20502-7;978-3-030-20503-4

Table of contents :
Front Matter ....Pages i-xviii
Front Matter ....Pages 1-1
Empirical Validation of a Checklist for Heuristic Evaluation of Automated Vehicle HMIs (Yannick Forster, Sebastian Hergeth, Frederik Naujoks, Josef F. Krems, Andreas Keinath)....Pages 3-14
A Novel Method for Designing Metaphor-Based Driver-Vehicle Interaction Concepts in Automated Vehicles (Jan Bavendiek, Emily Oliveira, Lutz Eckstein)....Pages 15-26
Vocal Guidance of Visual Gaze During an Automated Vehicle Handover Task (Jediah R. Clark, Neville A. Stanton, Kirsten M. A. Revell)....Pages 27-35
How Do You Want to be Driven? Investigation of Different Highly-Automated Driving Styles on a Highway Scenario (Patrick Rossner, Angelika C. Bullinger)....Pages 36-43
Using Technology Acceptance Model to Explain Driver Acceptance of Advanced Driver Assistance Systems (Md Mahmudur Rahman, Shuchisnigdha Deb, Daniel Carruth, Lesley Strawderman)....Pages 44-56
Bayesian Artificial Intelligence-Based Driver for Fully Automated Vehicle with Cognitive Capabilities (Ata Khan)....Pages 57-66
A Survey Study to Explore Comprehension of Autonomous Vehicle’s Communication Features (Shuchisnigdha Deb, Daniel W. Carruth, Lesley J. Strawderman)....Pages 67-78
How Should Automated Vehicles Communicate? – Effects of a Light-Based Communication Approach in a Wizard-of-Oz Study (Ann-Christin Hensch, Isabel Neumann, Matthias Beggiato, Josephine Halama, Josef F. Krems)....Pages 79-91
Front Matter ....Pages 93-93
Designing Adaptation in Cars: An Exploratory Survey on Drivers’ Usage of ADAS and Car Adaptations (Nermin Caber, Patrick Langdon, P. John Clarkson)....Pages 95-106
Supporting Older Drivers’ Visual Processing of Intersections - Effects of Providing Prior Information (Matthias Beggiato, Franziska Hartwich, Tibor Petzoldt, Josef Krems)....Pages 107-119
The Impact of Different Human-Machine Interface Feedback Modalities on Older Participants’ User Experience of CAVs in a Simulator Environment (Iveta Eimontaite, Alexandra Voinescu, Chris Alford, Praminda Caleb-Solly, Phillip Morgan)....Pages 120-132
User Experience in Immersive VR-Based Serious Game: An Application in Highly Automated Driving Training (Mahdi Ebnali, Cyrus Kian, Majid Ebnali-Heidari, Adel Mazloumi)....Pages 133-144
Comparison of Child and Adult Pedestrian Perspectives of External Features on Autonomous Vehicles Using Virtual Reality Experiment (Shuchisnigdha Deb, Daniel W. Carruth, Muztaba Fuad, Laura M. Stanley, Darren Frey)....Pages 145-156
An Inclusive, Fully Autonomous Vehicle Simulator for the Introduction of Human-Robot Interaction Technologies (Theocharis Amanatidis, Patrick Langdon, P. John Clarkson)....Pages 157-165
Front Matter ....Pages 167-167
Investigating Drivers’ Behaviour During Diverging Maneuvers Using an Instrumented Vehicle (Fabrizio D’Amico, Alessandro Calvi, Chiara Ferrante, Luca Bianchini Ciampoli, Fabio Tosti)....Pages 169-178
Model of Driving Skills Decrease in the Context of Autonomous Vehicles (Darina Havlíčková, Petr Zámečník, Eva Adamovská, Adam Gregorovič, Václav Linkov, Aleš Zaoral)....Pages 179-189
The User and the Automated Driving: A State-of-the-Art (Anabela Simões, Liliana Cunha, Sara Ferreira, José Carvalhais, José Pedro Tavares, António Lobo et al.)....Pages 190-201
Front Matter ....Pages 203-203
Explicit Forward Glance Duration Hidden Markov Model for Inference of Spillover Detection (John (Hyoshin) Park, Nigel Pugh, Justice Darko, Larkin Folsom, Siby Samuel)....Pages 205-213
Proposal for Graduated Driver Licensing Program: Age vs. Experience, Abu Dhabi Case Study (Yousif Al Thabahi, Marzouq Al Zaabi, Mohammed Al Eisaei, Abdulla Al Ghafli)....Pages 214-223
Impact of Mind Wandering on Driving (Minerva Rajendran, Venkatesh Balasubramanian)....Pages 224-232
Assessing the Relation Between Emotional Intelligence and Driving Behavior: An Online Survey (Swathy Parameswaran, Venkatesh Balasubramanian)....Pages 233-239
Front Matter ....Pages 241-241
The Effect of Tram Driver’s Cab Design on Posture and Physical Strain (Tobias Heine, Marco Käppler, Barbara Deml)....Pages 243-249
Engineering the Right Change Culture in a Complex (GB) Rail Industry (Michelle Nolan-McSweeney, Brendan Ryan, Sue Cobb)....Pages 250-260
Application of Cognitive Work Analysis to Explore Passenger Behaviour Change Through Provision of Information to Help Relieve Train Overcrowding (Jisun Kim, Kirsten Revell, John Preston)....Pages 261-271
Decrease Driver’s Workload and Increase Vigilance (Denis Miglianico, Vincent Pargade)....Pages 272-281
Analysis of Driving Performance Data to Evaluate Brake Manipulation by Railway Drivers (Daisuke Suzuki, Naoki Mizukami, Yutaka Kakizaki, Nobuyuki Tsuyuki)....Pages 282-288
Front Matter ....Pages 289-289
Sharing the Road: Experienced Cyclist and Motorist Knowledge and Perceptions (Mary L. Still, Jeremiah D. Still)....Pages 291-300
Examination on Corner Shape for Reducing Mental Stress by Pedestrian Appearing from Blind Spot of Intersection (Wataru Kobayashi, Yohsuke Yoshioka)....Pages 301-306
Pedestrian Attitudes to Shared-Space Interactions with Autonomous Vehicles – A Virtual Reality Study (Christopher G. Burns, Luis Oliveira, Vivien Hung, Peter Thomas, Stewart Birrell)....Pages 307-316
Front Matter ....Pages 317-317
Speed Behavior in a Suburban School Zone: A Driving Simulation Study with Familiar and Unfamiliar Drivers from Puerto Rico and Massachusetts (Didier Valdés, Michael Knodler, Benjamín Colucci, Alberto Figueroa, Maria Rojas, Enid Colón et al.)....Pages 319-329
Applying Perceptual Treatments for Reducing Operating Speeds on Curves: A Driving Simulator Study for Investigating Driver’s Speed Behavior (Alessandro Calvi, Fabrizio D‘Amico, Chiara Ferrante, Luca Bianchini Ciampoli, Fabio Tosti)....Pages 330-340
Learning Drivers’ Behavior Using Social Networking Service (Yueqing Li, Acyut Kaneria, Xiang Zhao, Vinaya Manchaiah)....Pages 341-350
Comparing the Differences of EEG Signals Based on Collision and Non-collision Cases (Xinran Zhang, Xuedong Yan)....Pages 351-360
Driving at Night: The Effects of Various Colored Windshield Tints on Visual Acuity, Glare Discomfort, and Color Perception (Ma. Gilean Fria Badilla, Elijah Gabalda, Jeonne Joseph Ramoso, Keneth Sedilla)....Pages 361-373
Front Matter ....Pages 375-375
Database Driven Ergonomic Vehicle Development via a Fully Parametric Seating Buck (Johannes Tiefnig, Mario Hirz, Wilhelm Dietrich)....Pages 377-386
Are You Sitting Comfortably? How Current Self-driving Car Concepts Overlook Motion Sickness, and the Impact It Has on Comfort and Productivity (Joseph Smyth, Paul Jennings, Stewart Birrell)....Pages 387-399
Experimental Investigation of the Relationship Between Human Discomfort and Involuntary Movements in Vehicle Seat (Junya Tatsuno, Koki Suyama, Hiroki Mitani, Hitomi Nakamura, Setsuo Maeda)....Pages 400-411
An Ergonomic Assessment of Mass Rapid Transport Trains in Metro Manila, Philippines (Anna Patricia F. Martinez, Angela Jasmin B. Caingat, Raine Alexandra S. Robielos, Benette P. Custodio)....Pages 412-424
Front Matter ....Pages 425-425
The Analysis of UK Road Traffic Accident Data and its Use in the Development of a Direct Vision Standard for Trucks in London (Russell Marshall, Steve Summerskill, James Lenard)....Pages 427-439
The Development of a Direct Vision Standard for Trucks in London Using a Volumetric Approach (Stephen Summerskill, Russell Marshall, Abby Paterson, Anthony Eland)....Pages 440-452
A Scenario-Based Investigation of Truck Platooning Acceptance (Matthias Neubauer, Oliver Schauer, Wolfgang Schildorfer)....Pages 453-461
Conceptual Testing of Visual HMIs for Merging of Trucks (Felix A. Dreger, Joost C. F. de Winter, Barys Shyrokau, Riender Happee)....Pages 462-474
“Should We Allow Him to Pass?” Increasing Cooperation Between Truck Drivers Using Anthropomorphism (Jana Fank, Leon Santen, Christian Knies, Frank Diermeyer)....Pages 475-484
Front Matter ....Pages 485-485
Gear Shifter Design – Lack of Dedicated Positions and the Contribution to Cognitive Load and Inattention (Sanna Lohilahti Bladfält, Camilla Grane, Peter Bengtsson)....Pages 487-498
Forensic Analyses of Rumble Strips and Truck Conspicuity (Jack L. Auflick, James K. Sprague, Joseph T. Eganhouse, Julius M. Roberts)....Pages 499-509
Investigation of Dubai Tram Safety Challenges and Road User Behavior Through Tram Driver’s Opinion Survey (Shahid Tanvir, Noor Zainab Habib, Guy H. Walker)....Pages 510-520
Analysis of Driving Safety and Cellphone Use Based on Social Media (Chao Qian, Yueqing Li, Wenchao Zuo, Yuhong Wang)....Pages 521-530
Trends of Crash Mitigations at High Crash Intersections in Nevada, US Based on Highway Safety Improvement Program (Wanmin Ge, Haiyuan Li)....Pages 531-541
Front Matter ....Pages 543-543
User-Centered Development of a Public Transportation Vehicle Operated in a Demand-Responsive Environment (Alexander Mueller, Stefanie Beyer, Gerhard Kopp, Oliver Deisser)....Pages 545-555
Human Factors Concerns: Drivers’ Perception on Electronic Sideview System in 21st Century Cars (Bankole K. Fasanya, Yashwant Avula, Swetha Keshavula, Supraja Aragattu, Sivaramakrishna Kurra, Bharath Kummari)....Pages 556-563
Development of a Prototype Steering Wheel for Simulator-Based Usability Assessment (James Brown, Neville Stanton, Kirsten Revell)....Pages 564-572
Should I Stay or Should I Go? - Influencing Context Factors for Users’ Decisions to Charge or Refuel Their Vehicles (Ralf Philipsen, Teresa Brell, Hannah Biermann, Teresa Eickels, Waldemar Brost, Martina Ziefle)....Pages 573-584
Driving Segway: A Musculoskeletal Investigation (Zavier Berti, Peter Rasche, Robert Chauvet, Matthias Wille, Vera Rick, Laura Barton et al.)....Pages 585-595
Using the Lane Change Test to Investigate In-Vehicle Display Placements (Sabrina N. Moran, Thomas Z. Strybel, Gabriella M. Hancock, Kim-Phuong L. Vu)....Pages 596-607
Investigation on the Effectiveness of Autostereoscopic 3D Displays for Parking Maneuver Tasks with Passenger Cars (André Dettmann, Angelika C. Bullinger)....Pages 608-617
Transport Realities and Challenges for Low Income Peripheral Located Settlements in Gauteng Province: Are We Witnessing the Genesis of a New Transport Order or Consolidation of the Old Transport Order? (James Chakwizira, Peter Bikam, Thompson A. Adeboyejo)....Pages 618-630
Front Matter ....Pages 631-631
Towards Autonomous Shipping – Exploring Potential Threats and Opportunities in Future Maritime Operations (Gesa Praetorius, Carl Hult, Carl Sandberg)....Pages 633-644
Evaluating the Impact of Increased Volume of Data Transmission on Teleoperated Vehicles (Kiome A. Pope, Aaron P. J. Roberts, Christopher J. Fenton, Neville A. Stanton)....Pages 645-655
Design of a Sustainable and Accessible Royal Rig Maxy Clipper for Single-Handed (Massimo Di Nicolantonio)....Pages 656-668
Interfaces with Legs? Documenting the Design Sprint of Prototype Future Submarine Control Room User Interfaces (Daniel Fay, Aaron P. J. Roberts, Neville A. Stanton)....Pages 669-680
Front Matter ....Pages 681-681
Considering Single-Piloted Airliners for Different Flight Durations: An Issue of Fatigue Management (Daniela Schmid, Neville A. Stanton)....Pages 683-694
An Eye in the Sky: Developing a Novel Framework for Visual Airport Traffic Control Tower Tasks (Amelia Kinsella, Lori Smith, Rebecca Collins, Katherine Berry)....Pages 695-701
Overwritten or Unrecorded: A Study of Accidents & Incidents in Which CVR Data Were not Available (Simon Cookson)....Pages 702-714
Human Factors Evaluation of ATC Operational Procedures in Relation to Use of 3D Display (Yisi Liu, Fitri Trapsilawati, Zirui Lan, Olga Sourina, Henry Johan, Fan Li et al.)....Pages 715-726
Monitoring Performance Measures for Radar Air Traffic Controllers Using Eye Tracking Techniques (Hong Jie Wee, Sun Woh Lye, Jean-Philippe Pinheiro)....Pages 727-738
Flight Eye Tracking Assistant (FETA): Proof of Concept (Christophe Lounis, Vsevolod Peysakhovich, Mickaël Causse)....Pages 739-751
How Does National Culture Help Pilots in Navigating in Different Environment? (Xiaoyu O. Wu)....Pages 752-761
Human Reliability Quantification in Flight Through a Simplified CREAM Method (Yundong Guo, Youchao Sun)....Pages 762-773
The Human Element in Performance Based Navigation: Air Traffic Controller Acceptance of Established on Required Navigation Performance Procedures (Lauren Thomas, Alicia Serrato)....Pages 774-782
Ergonomic Assessment of Instructors’ Capability to Conduct Personality-Oriented Training for Air Traffic Control (ATC) Personnel (Oleksii Reva, Sergii Borsuk, Valeriy Shulgin, Serhiy Nedbay)....Pages 783-793
Impact of Plants in Isolation: The EDEN-ISS Human Factors Investigation in Antarctica (Irene Lia Schlacht, Harald Kolrep, Schubert Daniel, Giorgio Musso)....Pages 794-806
Considerations for Passenger Experience in Space Tourism (Tiziano Bernard, Yash Mehta, Brandon Cuffie, Yassine Rayad, Sebastien Boulnois, Lucas Stephane)....Pages 807-818
Cognitive Architecture Based Mental Workload Evaluation for Spatial Fine Manual Control Task (Yanfei Liu, Zhiqiang Tian, Yuzhou Liu, Jusong Li, Feng Fu)....Pages 819-829
Back Matter ....Pages 831-834

Citation preview

Advances in Intelligent Systems and Computing 964

Neville Stanton Editor

Advances in Human Factors of Transportation Proceedings of the AHFE 2019 International Conference on Human Factors in Transportation, July 24–28, 2019, Washington D.C., USA

Advances in Intelligent Systems and Computing Volume 964

Series Editor Janusz Kacprzyk, Systems Research Institute, Polish Academy of Sciences, Warsaw, Poland Advisory Editors Nikhil R. Pal, Indian Statistical Institute, Kolkata, India Rafael Bello Perez, Faculty of Mathematics, Physics and Computing, Universidad Central de Las Villas, Santa Clara, Cuba Emilio S. Corchado, University of Salamanca, Salamanca, Spain Hani Hagras, School of Computer Science & Electronic Engineering, University of Essex, Colchester, UK László T. Kóczy, Department of Automation, Széchenyi István University, Gyor, Hungary Vladik Kreinovich, Department of Computer Science, University of Texas at El Paso, El Paso, TX, USA Chin-Teng Lin, Department of Electrical Engineering, National Chiao Tung University, Hsinchu, Taiwan Jie Lu, Faculty of Engineering and Information Technology, University of Technology Sydney, Sydney, NSW, Australia Patricia Melin, Graduate Program of Computer Science, Tijuana Institute of Technology, Tijuana, Mexico Nadia Nedjah, Department of Electronics Engineering, University of Rio de Janeiro, Rio de Janeiro, Brazil Ngoc Thanh Nguyen, Faculty of Computer Science and Management, Wrocław University of Technology, Wrocław, Poland Jun Wang, Department of Mechanical and Automation Engineering, The Chinese University of Hong Kong, Shatin, Hong Kong

The series “Advances in Intelligent Systems and Computing” contains publications on theory, applications, and design methods of Intelligent Systems and Intelligent Computing. Virtually all disciplines such as engineering, natural sciences, computer and information science, ICT, economics, business, e-commerce, environment, healthcare, life science are covered. The list of topics spans all the areas of modern intelligent systems and computing such as: computational intelligence, soft computing including neural networks, fuzzy systems, evolutionary computing and the fusion of these paradigms, social intelligence, ambient intelligence, computational neuroscience, artificial life, virtual worlds and society, cognitive science and systems, Perception and Vision, DNA and immune based systems, self-organizing and adaptive systems, e-Learning and teaching, human-centered and human-centric computing, recommender systems, intelligent control, robotics and mechatronics including human-machine teaming, knowledge-based paradigms, learning paradigms, machine ethics, intelligent data analysis, knowledge management, intelligent agents, intelligent decision making and support, intelligent network security, trust management, interactive entertainment, Web intelligence and multimedia. The publications within “Advances in Intelligent Systems and Computing” are primarily proceedings of important conferences, symposia and congresses. They cover significant recent developments in the field, both of a foundational and applicable character. An important characteristic feature of the series is the short publication time and world-wide distribution. This permits a rapid and broad dissemination of research results. ** Indexing: The books of this series are submitted to ISI Proceedings, EI-Compendex, DBLP, SCOPUS, Google Scholar and Springerlink ** More information about this series at http://www.springer.com/series/11156

Neville Stanton Editor

Advances in Human Factors of Transportation Proceedings of the AHFE 2019 International Conference on Human Factors in Transportation, July 24–28, 2019, Washington D.C., USA

123

Editor Neville Stanton Boldrewood Innovation Campus University of Southampton, TRG Southampton, UK

ISSN 2194-5357 ISSN 2194-5365 (electronic) Advances in Intelligent Systems and Computing ISBN 978-3-030-20502-7 ISBN 978-3-030-20503-4 (eBook) https://doi.org/10.1007/978-3-030-20503-4 © Springer Nature Switzerland AG 2020 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Advances in Human Factors and Ergonomics 2019

AHFE 2019 Series Editors Tareq Ahram, Florida, USA Waldemar Karwowski, Florida, USA

10th International Conference on Applied Human Factors and Ergonomics and the Affiliated Conferences Proceedings of the AHFE 2019 International Conferences on Human Factors in Transportation, held on July 24–28, 2019, in Washington D.C., USA

Advances in Affective and Pleasurable Design Advances in Neuroergonomics and Cognitive Engineering Advances in Design for Inclusion Advances in Ergonomics in Design Advances in Human Error, Reliability, Resilience, and Performance Advances in Human Factors and Ergonomics in Healthcare and Medical Devices Advances in Human Factors and Simulation Advances in Human Factors and Systems Interaction Advances in Human Factors in Cybersecurity Advances in Human Factors, Business Management and Leadership Advances in Human Factors in Robots and Unmanned Systems Advances in Human Factors in Training, Education, and Learning Sciences Advances in Human Factors of Transportation

Shuichi Fukuda Hasan Ayaz Giuseppe Di Bucchianico Francisco Rebelo and Marcelo M. Soares Ronald L. Boring Nancy J. Lightner and Jay Kalra Daniel N. Cassenti Isabel L. Nunes Tareq Ahram and Waldemar Karwowski Jussi Ilari Kantola and Salman Nazir Jessie Chen Waldemar Karwowski, Tareq Ahram and Salman Nazir Neville Stanton (continued)

v

vi

Advances in Human Factors and Ergonomics 2019

(continued) Advances in Artificial Intelligence, Software and Systems Engineering Advances in Human Factors in Architecture, Sustainable Urban Planning and Infrastructure Advances in Physical Ergonomics and Human Factors Advances in Interdisciplinary Practice in Industrial Design Advances in Safety Management and Human Factors Advances in Social and Occupational Ergonomics Advances in Manufacturing, Production Management and Process Control Advances in Usability and User Experience Advances in Human Factors in Wearable Technologies and Game Design Advances in Human Factors in Communication of Design Advances in Additive Manufacturing, Modeling Systems and 3D Prototyping

Tareq Ahram Jerzy Charytonowicz and Christianne Falcão Ravindra S. Goonetilleke and Waldemar Karwowski Cliff Sungsoo Shin Pedro M. Arezes Richard H. M. Goossens and Atsuo Murata Waldemar Karwowski, Stefan Trzcielinski and Beata Mrugalska Tareq Ahram and Christianne Falcão Tareq Ahram Amic G. Ho Massimo Di Nicolantonio, Emilio Rossi and Thomas Alexander

Preface

Human factors and ergonomics have made a considerable contribution to the research, design, development, operation, and analysis of transportation systems. This includes road, rail, aviation, aerospace, and maritime vehicles as well as their complementary infrastructure. This book presents recent advances in the human factors aspects of transportation. These advances include accident analysis, automation of vehicles, comfort, distraction of drivers (including how to avoid it), environmental concerns, in-vehicle systems design, intelligent transport systems, methodological developments, new systems and technology, observational and case studies, safety, situation awareness, skill development and training, warnings and workload. This book brings together the most recent human factors which work in the transportation domain, including empirical research, human performance and other types of modeling, analysis, and development. The issues facing engineers, scientists, and other practitioners of human factors in transportation research are becoming more challenging and more critical. The common theme across these sections is that they deal with the interactions of humans with systems in the environment. Moreover, many of the chapter topics cross domain and discipline boundaries. This is in keeping with the systemic nature of the problems facing human factors experts in rail and road, aviation and aerospace, and maritime research—it is becoming increasingly important to view problems not as isolated factors that can be extracted from the system environment, but as embedded issues that can only be understood as a part of an overall system. In keeping with a system that is vast in its scope and reach, the chapters in this book cover a wide range of topics. The chapters are organized into thirteen sections: Part 1 Human Factors in Transportation: Road and Rail Section Section Section Section

1 2 3 4

Vehicle Automation Designing Autonomy in Transportation Age and Inclusion Driving Behavior Autonomous and Automated Vehicles Driver Training and Education

vii

viii

Section Section Section Section Section Section Section

Preface

5 6 7 8 9 10 11

Human Factors in Transportation Rail Vulnerable Road Users Driving Behavior Safety and Simulation Road and Rail Comfort Trucks Safety and Hazards Road and Rail Usability

Part 2 Human Factors in Transportation: Maritime Section 12

Transportation Maritime

Part 3 Human Factors in Transportation: Aviation and Space Section 13

Human Factors in Aviation and Space

This book will be of interest and use to transportation professionals who work in the road and rail, aviation and aerospace, and maritime domains as it reflects some of the latest human factors and ergonomics thinking and practice. It should also be of interest to students and researchers in these fields, to help stimulate research questions and ideas. It is my hope that the ideas and studies reported within this book will help to produce safer, more efficient, and effective transportation systems in the future. We are grateful to the Scientific Advisory Board which has helped elicit the contributions and develop the themes in the book. These people are experts and academic leaders in their respective fields, and their help is very much appreciated, especially as they gave their time to the project. Special thanks to Giorgio Musso and Nancy Currie-Gregg for their contribution to the Space program.

Road and Rail C. Allison, UK Giles Balbinotti, Brazil Klaus Bengler, Germany Stewart Birrell, UK Gary Burnett, UK Peter Chapman, UK Fang Chen, Sweden Denis Coelho, Portugal Benjamin Colucci Rios, Puerto Rico Guillaume Craveur, France Laurel Dickson-Bull, USA L. Dorn, UK Ian Glendon, Australia Iwona Grabarek, Poland Rachel Grice, USA

Preface

R. Happee, Netherlands S. Jamson, UK Dave Kaber, USA Josef Krems, Germany Mike Lenné, Australia Elżbieta Macioszek, Poland Franck Mars, France Deborah McAvoy, USA Ann Mills, UK Ralf Philipsen, Germany K. Revell, UK Ralf Risser, Austria Paul Salmon, Australia Grzegorz Sierpiński, Poland Shafiq ur Rehman, Sweden Didier Valdes Diaz, Puerto Rico Guy Walker, Scotland Kristie Young, Australia

Aviation V. Banks, UK Marcus Biella, Germany Clark Borst, The Netherlands Tamsyn Edwards, USA Michael Feary, USA Andreas Haslbeck, Germany Becky Hooey, USA John Huddlestone, UK David Kaber, USA Kara Latorella, USA Arnab Majumdar, UK Lynne Martin, USA Joey Mercer, USA Max Mulder, The Netherlands K. Plant, UK Jon Lars Syversen, Norway Savvy Verma, USA David Yacht, USA Kim Vu, USA

Space Daniele Bedini, Italy Jason Beierle, USA Tiziano Bernard, USA

ix

x

Preface

Roberta Capra, Italy Marinella Ferrino, Italy Enrico Gaia, Italy Sandra Hauplick, Austria Kees Nieuwenhuis, The Netherlands Raffaella Ricci, Italy Adriana Salatino, Italy Irene Schlacht, Italy Domenico Tedone, Italy

Maritime Ahmet Dursun Alkan, Turkey David Andrews, UK Giuseppe Di Bucchianico, Italy Dawn Gray, USA Marc Grootjen, The Netherlands Thomas Koester, Denmark Scott Netson MacKinnon, Canada Massimo Musio Sale, Italy Stella (Styliani) Parisi, Greece Gesa Praetorius, Sweden Andrea Ratti, Italy A. Roberts, UK Andrea Vallicelli, Italy July 2019

Neville Stanton

Contents

Vehicle Automation Empirical Validation of a Checklist for Heuristic Evaluation of Automated Vehicle HMIs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Yannick Forster, Sebastian Hergeth, Frederik Naujoks, Josef F. Krems, and Andreas Keinath

3

A Novel Method for Designing Metaphor-Based Driver-Vehicle Interaction Concepts in Automated Vehicles . . . . . . . . . . . . . . . . . . . . . Jan Bavendiek, Emily Oliveira, and Lutz Eckstein

15

Vocal Guidance of Visual Gaze During an Automated Vehicle Handover Task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jediah R. Clark, Neville A. Stanton, and Kirsten M. A. Revell

27

How Do You Want to be Driven? Investigation of Different Highly-Automated Driving Styles on a Highway Scenario . . . . . . . . . . . Patrick Rossner and Angelika C. Bullinger

36

Using Technology Acceptance Model to Explain Driver Acceptance of Advanced Driver Assistance Systems . . . . . . . . . . . . . . . . . . . . . . . . . Md Mahmudur Rahman, Shuchisnigdha Deb, Daniel Carruth, and Lesley Strawderman

44

Bayesian Artificial Intelligence-Based Driver for Fully Automated Vehicle with Cognitive Capabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ata Khan

57

A Survey Study to Explore Comprehension of Autonomous Vehicle’s Communication Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Shuchisnigdha Deb, Daniel W. Carruth, and Lesley J. Strawderman

67

xi

xii

Contents

How Should Automated Vehicles Communicate? – Effects of a Light-Based Communication Approach in a Wizard-of-Oz Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ann-Christin Hensch, Isabel Neumann, Matthias Beggiato, Josephine Halama, and Josef F. Krems

79

Designing Autonomy in Transportation: Age and Inclusion Designing Adaptation in Cars: An Exploratory Survey on Drivers’ Usage of ADAS and Car Adaptations . . . . . . . . . . . . . . . . . Nermin Caber, Patrick Langdon, and P. John Clarkson

95

Supporting Older Drivers’ Visual Processing of Intersections Effects of Providing Prior Information . . . . . . . . . . . . . . . . . . . . . . . . . . 107 Matthias Beggiato, Franziska Hartwich, Tibor Petzoldt, and Josef Krems The Impact of Different Human-Machine Interface Feedback Modalities on Older Participants’ User Experience of CAVs in a Simulator Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 Iveta Eimontaite, Alexandra Voinescu, Chris Alford, Praminda Caleb-Solly, and Phillip Morgan User Experience in Immersive VR-Based Serious Game: An Application in Highly Automated Driving Training . . . . . . . . . . . . . 133 Mahdi Ebnali, Cyrus Kian, Majid Ebnali-Heidari, and Adel Mazloumi Comparison of Child and Adult Pedestrian Perspectives of External Features on Autonomous Vehicles Using Virtual Reality Experiment . . . 145 Shuchisnigdha Deb, Daniel W. Carruth, Muztaba Fuad, Laura M. Stanley, and Darren Frey An Inclusive, Fully Autonomous Vehicle Simulator for the Introduction of Human-Robot Interaction Technologies . . . . . . . 157 Theocharis Amanatidis, Patrick Langdon, and P. John Clarkson Driving Behavior: Autonomous and Automated Vehicles Investigating Drivers’ Behaviour During Diverging Maneuvers Using an Instrumented Vehicle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 Fabrizio D’Amico, Alessandro Calvi, Chiara Ferrante, Luca Bianchini Ciampoli, and Fabio Tosti Model of Driving Skills Decrease in the Context of Autonomous Vehicles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179 Darina Havlíčková, Petr Zámečník, Eva Adamovská, Adam Gregorovič, Václav Linkov, and Aleš Zaoral

Contents

xiii

The User and the Automated Driving: A State-of-the-Art . . . . . . . . . . . 190 Anabela Simões, Liliana Cunha, Sara Ferreira, José Carvalhais, José Pedro Tavares, António Lobo, António Couto, and Daniel Silva Driver Training and Education Explicit Forward Glance Duration Hidden Markov Model for Inference of Spillover Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 John (Hyoshin) Park, Nigel Pugh, Justice Darko, Larkin Folsom, and Siby Samuel Proposal for Graduated Driver Licensing Program: Age vs. Experience, Abu Dhabi Case Study . . . . . . . . . . . . . . . . . . . . . . 214 Yousif Al Thabahi, Marzouq Al Zaabi, Mohammed Al Eisaei, and Abdulla Al Ghafli Impact of Mind Wandering on Driving . . . . . . . . . . . . . . . . . . . . . . . . . 224 Minerva Rajendran and Venkatesh Balasubramanian Assessing the Relation Between Emotional Intelligence and Driving Behavior: An Online Survey . . . . . . . . . . . . . . . . . . . . . . . 233 Swathy Parameswaran and Venkatesh Balasubramanian Human Factors in Transportation: Rail The Effect of Tram Driver’s Cab Design on Posture and Physical Strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243 Tobias Heine, Marco Käppler, and Barbara Deml Engineering the Right Change Culture in a Complex (GB) Rail Industry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250 Michelle Nolan-McSweeney, Brendan Ryan, and Sue Cobb Application of Cognitive Work Analysis to Explore Passenger Behaviour Change Through Provision of Information to Help Relieve Train Overcrowding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261 Jisun Kim, Kirsten Revell, and John Preston Decrease Driver’s Workload and Increase Vigilance . . . . . . . . . . . . . . . 272 Denis Miglianico and Vincent Pargade Analysis of Driving Performance Data to Evaluate Brake Manipulation by Railway Drivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282 Daisuke Suzuki, Naoki Mizukami, Yutaka Kakizaki, and Nobuyuki Tsuyuki

xiv

Contents

Vulnerable Road Users Sharing the Road: Experienced Cyclist and Motorist Knowledge and Perceptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291 Mary L. Still and Jeremiah D. Still Examination on Corner Shape for Reducing Mental Stress by Pedestrian Appearing from Blind Spot of Intersection . . . . . . . . . . . 301 Wataru Kobayashi and Yohsuke Yoshioka Pedestrian Attitudes to Shared-Space Interactions with Autonomous Vehicles – A Virtual Reality Study . . . . . . . . . . . . . . 307 Christopher G. Burns, Luis Oliveira, Vivien Hung, Peter Thomas, and Stewart Birrell Driving Behavior: Safety and Simulation Speed Behavior in a Suburban School Zone: A Driving Simulation Study with Familiar and Unfamiliar Drivers from Puerto Rico and Massachusetts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319 Didier Valdés, Michael Knodler, Benjamín Colucci, Alberto Figueroa, Maria Rojas, Enid Colón, Nicholas Campbell, and Francis Tainter Applying Perceptual Treatments for Reducing Operating Speeds on Curves: A Driving Simulator Study for Investigating Driver’s Speed Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330 Alessandro Calvi, Fabrizio D‘Amico, Chiara Ferrante, Luca Bianchini Ciampoli, and Fabio Tosti Learning Drivers’ Behavior Using Social Networking Service . . . . . . . . 341 Yueqing Li, Acyut Kaneria, Xiang Zhao, and Vinaya Manchaiah Comparing the Differences of EEG Signals Based on Collision and Non-collision Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351 Xinran Zhang and Xuedong Yan Driving at Night: The Effects of Various Colored Windshield Tints on Visual Acuity, Glare Discomfort, and Color Perception . . . . . . . . . . 361 Ma. Gilean Fria Badilla, Elijah Gabalda, Jeonne Joseph Ramoso, and Keneth Sedilla Road and Rail: Comfort Database Driven Ergonomic Vehicle Development via a Fully Parametric Seating Buck . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377 Johannes Tiefnig, Mario Hirz, and Wilhelm Dietrich

Contents

xv

Are You Sitting Comfortably? How Current Self-driving Car Concepts Overlook Motion Sickness, and the Impact It Has on Comfort and Productivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387 Joseph Smyth, Paul Jennings, and Stewart Birrell Experimental Investigation of the Relationship Between Human Discomfort and Involuntary Movements in Vehicle Seat . . . . . . . . . . . . 400 Junya Tatsuno, Koki Suyama, Hiroki Mitani, Hitomi Nakamura, and Setsuo Maeda An Ergonomic Assessment of Mass Rapid Transport Trains in Metro Manila, Philippines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 412 Anna Patricia F. Martinez, Angela Jasmin B. Caingat, Raine Alexandra S. Robielos, and Benette P. Custodio Trucks The Analysis of UK Road Traffic Accident Data and its Use in the Development of a Direct Vision Standard for Trucks in London . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427 Russell Marshall, Steve Summerskill, and James Lenard The Development of a Direct Vision Standard for Trucks in London Using a Volumetric Approach . . . . . . . . . . . . . . . . . . . . . . . . 440 Stephen Summerskill, Russell Marshall, Abby Paterson, and Anthony Eland A Scenario-Based Investigation of Truck Platooning Acceptance . . . . . . 453 Matthias Neubauer, Oliver Schauer, and Wolfgang Schildorfer Conceptual Testing of Visual HMIs for Merging of Trucks . . . . . . . . . . 462 Felix A. Dreger, Joost C. F. de Winter, Barys Shyrokau, and Riender Happee “Should We Allow Him to Pass?” Increasing Cooperation Between Truck Drivers Using Anthropomorphism . . . . . . . . . . . . . . . . . 475 Jana Fank, Leon Santen, Christian Knies, and Frank Diermeyer Safety and Hazards Gear Shifter Design – Lack of Dedicated Positions and the Contribution to Cognitive Load and Inattention . . . . . . . . . . . . 487 Sanna Lohilahti Bladfält, Camilla Grane, and Peter Bengtsson Forensic Analyses of Rumble Strips and Truck Conspicuity . . . . . . . . . 499 Jack L. Auflick, James K. Sprague, Joseph T. Eganhouse, and Julius M. Roberts

xvi

Contents

Investigation of Dubai Tram Safety Challenges and Road User Behavior Through Tram Driver’s Opinion Survey . . . . . . . . . . . . . . . . 510 Shahid Tanvir, Noor Zainab Habib, and Guy H. Walker Analysis of Driving Safety and Cellphone Use Based on Social Media . . . 521 Chao Qian, Yueqing Li, Wenchao Zuo, and Yuhong Wang Trends of Crash Mitigations at High Crash Intersections in Nevada, US Based on Highway Safety Improvement Program . . . . . . . . . . . . . . 531 Wanmin Ge and Haiyuan Li Road and Rail: Usability User-Centered Development of a Public Transportation Vehicle Operated in a Demand-Responsive Environment . . . . . . . . . . . . . . . . . . 545 Alexander Mueller, Stefanie Beyer, Gerhard Kopp, and Oliver Deisser Human Factors Concerns: Drivers’ Perception on Electronic Sideview System in 21st Century Cars . . . . . . . . . . . . . . . . . . . . . . . . . . 556 Bankole K. Fasanya, Yashwant Avula, Swetha Keshavula, Supraja Aragattu, Sivaramakrishna Kurra, and Bharath Kummari Development of a Prototype Steering Wheel for Simulator-Based Usability Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 564 James Brown, Neville Stanton, and Kirsten Revell Should I Stay or Should I Go? - Influencing Context Factors for Users’ Decisions to Charge or Refuel Their Vehicles . . . . . . . . . . . . 573 Ralf Philipsen, Teresa Brell, Hannah Biermann, Teresa Eickels, Waldemar Brost, and Martina Ziefle Driving Segway: A Musculoskeletal Investigation . . . . . . . . . . . . . . . . . 585 Zavier Berti, Peter Rasche, Robert Chauvet, Matthias Wille, Vera Rick, Laura Barton, Tobias Hellig, Katharina Schäfer, Christina Bröhl, Sabine Theis, Christopher Brandl, Verena Nitsch, and Alexander Mertens Using the Lane Change Test to Investigate In-Vehicle Display Placements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 596 Sabrina N. Moran, Thomas Z. Strybel, Gabriella M. Hancock, and Kim-Phuong L. Vu Investigation on the Effectiveness of Autostereoscopic 3D Displays for Parking Maneuver Tasks with Passenger Cars . . . . . . . . . . . . . . . . . 608 André Dettmann and Angelika C. Bullinger

Contents

xvii

Transport Realities and Challenges for Low Income Peripheral Located Settlements in Gauteng Province: Are We Witnessing the Genesis of a New Transport Order or Consolidation of the Old Transport Order? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 618 James Chakwizira, Peter Bikam, and Thompson A. Adeboyejo Transportation: Maritime Towards Autonomous Shipping – Exploring Potential Threats and Opportunities in Future Maritime Operations . . . . . . . . . . . . . . . . . 633 Gesa Praetorius, Carl Hult, and Carl Sandberg Evaluating the Impact of Increased Volume of Data Transmission on Teleoperated Vehicles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645 Kiome A. Pope, Aaron P. J. Roberts, Christopher J. Fenton, and Neville A. Stanton Design of a Sustainable and Accessible Royal Rig Maxy Clipper for Single-Handed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 656 Massimo Di Nicolantonio Interfaces with Legs? Documenting the Design Sprint of Prototype Future Submarine Control Room User Interfaces . . . . . . . . . . . . . . . . . 669 Daniel Fay, Aaron P. J. Roberts, and Neville A. Stanton Human Factors in Aviation and Space Considering Single-Piloted Airliners for Different Flight Durations: An Issue of Fatigue Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 683 Daniela Schmid and Neville A. Stanton An Eye in the Sky: Developing a Novel Framework for Visual Airport Traffic Control Tower Tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . 695 Amelia Kinsella, Lori Smith, Rebecca Collins, and Katherine Berry Overwritten or Unrecorded: A Study of Accidents & Incidents in Which CVR Data Were not Available . . . . . . . . . . . . . . . . . . . . . . . . 702 Simon Cookson Human Factors Evaluation of ATC Operational Procedures in Relation to Use of 3D Display . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 715 Yisi Liu, Fitri Trapsilawati, Zirui Lan, Olga Sourina, Henry Johan, Fan Li, Chun-Hsien Chen, and Wolfgang Mueller-Wittig Monitoring Performance Measures for Radar Air Traffic Controllers Using Eye Tracking Techniques . . . . . . . . . . . . . . . . . . . . . 727 Hong Jie Wee, Sun Woh Lye, and Jean-Philippe Pinheiro

xviii

Contents

Flight Eye Tracking Assistant (FETA): Proof of Concept . . . . . . . . . . . 739 Christophe Lounis, Vsevolod Peysakhovich, and Mickaël Causse How Does National Culture Help Pilots in Navigating in Different Environment? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 752 Xiaoyu O. Wu Human Reliability Quantification in Flight Through a Simplified CREAM Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 762 Yundong Guo and Youchao Sun The Human Element in Performance Based Navigation: Air Traffic Controller Acceptance of Established on Required Navigation Performance Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 774 Lauren Thomas and Alicia Serrato Ergonomic Assessment of Instructors’ Capability to Conduct Personality-Oriented Training for Air Traffic Control (ATC) Personnel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 783 Oleksii Reva, Sergii Borsuk, Valeriy Shulgin, and Serhiy Nedbay Impact of Plants in Isolation: The EDEN-ISS Human Factors Investigation in Antarctica . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 794 Irene Lia Schlacht, Harald Kolrep, Schubert Daniel, and Giorgio Musso Considerations for Passenger Experience in Space Tourism . . . . . . . . . 807 Tiziano Bernard, Yash Mehta, Brandon Cuffie, Yassine Rayad, Sebastien Boulnois, and Lucas Stephane Cognitive Architecture Based Mental Workload Evaluation for Spatial Fine Manual Control Task . . . . . . . . . . . . . . . . . . . . . . . . . . 819 Yanfei Liu, Zhiqiang Tian, Yuzhou Liu, Jusong Li, and Feng Fu Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 831

Vehicle Automation

Empirical Validation of a Checklist for Heuristic Evaluation of Automated Vehicle HMIs Yannick Forster1,2(&), Sebastian Hergeth1, Frederik Naujoks1, Josef F. Krems2, and Andreas Keinath1 1

BMW Group, Usability & User Interaction, Knorrstr. 147, 80937 Munich, Germany {yannick.forster,sebastian.hergeth,frederik.naujoks, andreas.keinath}@bmw.de 2 Institute for Psychology, Chemnitz University of Technology, Wilhelm-Raabe Str. 43, 09120 Chemnitz, Germany [email protected]

Abstract. For a successful market introduction of Level 3 Automated Driving Systems (L3 ADS), a careful evaluation of Human-Machine Interfaces (HMIs) is necessary. This paper describes an empirical evaluation of a checklist that has been previously developed for the use in heuristic expert assessments, demonstrating that an ADS HMI that meets the guidelines as defined in the checklist scores higher in several measures of usability and acceptance than an HMI that does not meet the checklist requirements. Therefore, N = 57 participants completed two 15-min drives with an L3 ADS in a driving simulator. They experienced two variations of a L3 ADS HMI that differed in the degree of complying with the checklist. Inferential statistics showed that acceptance and usability measures differed substantially between the two experimental HMIs. Behavioral observations of interaction performance also demonstrate that noncompliance with the checklist items lowers the usability of the L3 system. Keywords: Automated driving
Human-Machine Interface Heuristic assessment
Driving simulator study




1 Introduction Conditionally automated driving functions will soon be available on the consumer market. L3 ADS are characterized by taking over longitudinal and lateral vehicle control, while the driver does not have to constantly monitor correct system functioning and the driving environment [1]. Instead, he/she has the possibility to engage in non-driving related tasks (NDRTs, e.g., reading, watching movies, etc.). The driver has to be ready as fallback performer if the system function fails or the operational design domain of the function ends. Potential benefits of automating the driving task are increased comfort, safety and traffic efficiency [2]. Ensuring the safe use of ADS is of primary importance so that this vision can become reality. This requires development of a methodology on how to evaluate Human-Machine-Interfaces (HMIs) in the context of automated driving. © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 3–14, 2020. https://doi.org/10.1007/978-3-030-20503-4_1

4

Y. Forster et al.

Human factors evaluation methods can be roughly divided in expert-based and empirical approaches. For example, the RESPONSE Code of Practice [3] contains “confirmation tests” involving test participants or evaluation by an interdisciplinary expert team to assess Advanced Driver Assistance Systems (ADAS). In the area of automated driving, much empirical research has been directed on the assessment of transfer of control events (so-called “take-over situations”) from the automated vehicle (AV) to the human operator. It revealed a variety of human factors issues such as fatigue [4], trust [5, 6], mode awareness [7] or controllability [8]. Usability of L3 ADS HMIs, especially when coupled with driver assistance systems, is also an emerging topic [9, 10]. On the other side, expert-based assessments have mainly been a non-issue in the scientific literature, although test institutes that aim at guiding consumer decisions such as EURO NCAP or Consumer Reports commonly use these types of assessments. We developed a easy to use checklist procedure, consisting of a twenty-item questionnaire that is to be used by human factors experts inspecting an AV HMI [11] while completing a set of standardized use cases [12]. The heuristic assessment results in a judgement whether each of the guidelines are fully met (“no concerns”), partially met, but some aspects of the HMI are non-compliant (“minor concerns”) or definitely not met (“major concerns”). The checklist itself was based on a collection of most important standards, guidelines and best practices that can be applied to the design and evaluation of AV HMIs (for further information, see [11]). Heuristic assessments have several benefits, such as being time efficient and applicable in early stages of the product development stage, however, their validity is tightly related to the principle underlying the evaluation. Therefore, a set of evaluation principles (the heuristics) is commonly used that form the basis of the expert assessment. This paper presents a simulator study that is aimed as a first evaluation of the AV checklist heuristics by varying whether an AV HMI either meets the checklist items or not. Thus, two different HMI conditions were created and their impact on the usability when interacting with a simulated L3 AV was rated using common self-report and behavioral usability measures.

2 Method 2.1

Sample

In total, N = 57 (9 female, 48 male) participants took part in study (mean age = 40.56 years, SD = 9.32 MAX = 60, MIN = 25). All participants were BMW Group employees, held a German driver’s license, had normal or corrected to normal vision and had not previously partaken in a simulator study on L3 ADS. 2.2

Driving Simulation and Automated Driving Function

The study was conducted in a fixed-based driving simulator. The integrated vehicle’s console contained all necessary instrumentation and was identical to a BMW 5 series with automatic transmission. The front channels were displayed through three LED

Empirical Validation of a Checklist for Heuristic Evaluation

5

screens (each 1920  1080 pixels, 50″ size) providing a combined field of view of 120°. Three LED screens behind the vehicle displayed the rear-view for the mirrors. Driving simulation was rendered with a frequency of 60 Hz. Once activated, the L3 ADS executed both longitudinal and lateral vehicle control. When the L3 ADS encountered a scenario that exceeded its operational design domain (ODD, see section Use cases), a 20-s take-over cascade was initiated and displayed to the driver (see section Human-Machine Interface). 2.3

Study Design and Procedure

There were two different HMIs in the present study. The study employed a one-factor within subject design with two levels of HMI guideline compliance based on Naujoks and colleagues [11]. Participants were randomly assigned to either the (1) high compliance HMI or the (2) low compliance HMI condition in the first drive and experienced the respective other condition in the second drive. Upon arrival, participants were welcomed and gave informed consent. The experimenter explained that the study purpose was to examine two HMIs for automated driving and outlined the study procedure. To accustom themselves with the driving simulation, participants completed a five-minute manual familiarization drive. Prior to each experimental drive, the experimenter explained that, once activated, the L3 ADS would execute lateral and longitudinal vehicle guidance. Furthermore, the experimenter pointed out, that in case of exceedance of the system’s limits, it would inform them with sufficient notice to take over manual control. After each drive, participants completed a set of questionnaires. 2.4

Human-Machine Interface

The HMI was depicted in the instrument cluster. When activated, the blue color of the lane symbols, the text and the steering wheel indicated that the system function carried out longitudinal and lateral vehicle guidance. This HMI (see Fig. 1) resembles that of existing HMI solutions for ACC with additional steering assistance [13]. The high compliance HMI (see Fig. 1, left) communicated information redundantly by means of pictograms and a textbox above [14]. Textual information was displayed in German language. During the approach of the system limits, the HMI announced system limitations through a take-over cascade in form of an announcement, a cautionary takeover request (“cautionary TOR”) and an imminent take-over request (“imminent TOR”) [15] The stages lasted for 7 s (announcement and cautionary TOR) and 6 s (imminent TOR), respectively. 20 s before reaching the limitation, a generic warning tone announced the upcoming limit. Additionally, the textbox displayed messages. The cautionary TOR followed this announcement after six seconds and the HMI color switched from blue to yellow. The HMI showed hands that grab the steering wheel and additional information in the text. After seven more seconds, the imminent TOR appeared with the HMI colored in red and hands grabbing the steering wheel. The imminent TOR was accompanied by a more critical auditory warning. Drivers could activate the L3 ADS by pressing a button on the left side of the steering wheel with the label “AUTO”. Deactivation was possible through either braking/accelerating, active

6

Y. Forster et al.

steering input or pressing the “AUTO”-button with subsequently putting hands on the steering wheel. During the imminent TOR, a hands-on signal immediately deactivated the L3 ADS. Out of 20 HMI guidelines, six guidelines were varied to generate a difference in visual presentation and operation component interface quality. These specific guidelines were chosen on an expert assessment to create a noticeable difference between two HMI versions. The varied items are the following: • The system mode should be displayed continuously (item2) • System state changes should be effectively communicated (item 3) • The visual interface should have a sufficient contrast in luminance and/or color between foreground and background (item 7) • Texts (e.g., font types and size of characters) and symbols should be easily readable from the permitted seating position (item 8) • Commonly accepted or standardized symbols should be used to communicate the automation mode. Use of non-standard symbols should be supplemented by additional text explanations or vocal phrase/s (item 9) • The colors used to communicate system states should be in accordance with common conventions and stereotypes (item 14) Table 1 provides an overview of specific variations in the HMI, guideline and according reference. Figure 1 depicts the high compliance HMI (left) and the low compliance HMI (right) when active, TOR and not available.

Table 1. Variations for low compliance HMI for the two components with respective criterion and reference Component

Variation

Guideline number

Operation component Display

Activation and deactivation through long-press (i.e., 0.8 s) (1) Pictograms are 60% of the original size (2) No text information except for L3 ADS availability

Item 3

Reference in guideline of Naujoks and colleagues [11] [16]

Item 8

[17]

Item 2, Item 3, item 9 Item 3, Item 7, Item 14 Item 3, Item 7, Item 14

[14, 16, 18]

(3) no color coding for cautionary and imminent TOR (4) no blue color coding for active L3 ADS

[14, 16, 18]

[14, 16, 19]

Empirical Validation of a Checklist for Heuristic Evaluation high compliance HMI

7

low compliance HMI

L3 ADS active

cautionary TOR

imminent TOR

L3 ADS not available for use

Fig. 1. HMI for high compliance (left) and low compliance (right) during normal functioning (top) cautionary TOR (2nd row), imminent TOR (3rd row) and L3 ADS not available (bottom). Numbers indicate HMI variations described in Table 1 column 2.

2.5

Use Cases

Use-cases of the present study were chosen based on the HMI testing scenario catalogue for L3 ADS proposed by Naujoks and colleagues [12]. Use-cases included driver initiated activations (UCs 1, 3,6) and deactivations (UC 2), two TORs due to road works (UC5) and the end of L3 ADS availability (UC 8) as well as independently executed system maneuvers (UCs 2 and 7) [20]. One drive lasted approximately 15 min. Table 3 gives an overview of the sequential order of use cases and information on whether the interaction was initiated by the driver or the system. 2.6

Dependent Variables

The present study investigated HMI differences by means of self-report measures of two important constructs for the evaluation of automated driving HMIs, which are usability and acceptance [21]. Satisfaction was operationalized through the System Usability Scale [22] and acceptance was operationalized through the van-der-Laan acceptance scale [23]. The SUS is a 10-item questionnaire with a two-factorial underlying structure of learnability and usability [24] which has been applied in prior research on automated driving [25]. The van-der-Laan scale is a 9-item acceptance measure with an underlying two factorial structure of usefulness and satisfaction. The scale has been used in the automated driving context [26] among others. Items are

8

Y. Forster et al.

represented through semantic differentials on a scale from –2 to 2. Participants completed the SUS and van-der-Laan scale after each experimental block. Observational measures of interaction performance was an experimenter rating and reaction times in the two TOR scenarios. Reaction times were calculated from the onset of the announcement to the inactivity of the L3 ADS. Two experimenters rated the interaction behavior in the first and third activation use-case on a 11-point scale from 0 (not at all) to 10 (perfect) [27]. Table 2 shows the experimenter rating with the observable behavior and according rating category.

Table 2. Experimenter rating with observed behavior and category. Behavior Activation not possible Serious problems, assistance of experimenter Problems, failures, hesitation Correct/targeted action, no failures Immediate activation, no failure

Category Not at all [0] poor [1–3] average [4–6] good [7–9] Perfect [10]

3 Results 3.1

Missing Data

All participants completed the entire course with all UCs. In total, 10 items of the vander-Laan scale were missing as one participant did not complete the questionnaire in one condition. Missing data was estimated by an EM Algorithm [28]. There was one incomplete data set for reaction time data. 3.2

Self-report Measures

Subscales were averaged into a composite as described in the original source. Descriptive statistics (i.e., M, SD) for the SUS and the van-der-Laan scale are shown in Figs. 2 and 3. The SUS ratings of the high compliance HMI revealed higher ratings for positively framed items and lower ratings for negatively framed items compared to the low compliance HMI. Thus, overall SUS scores of the high compliance HMI was higher than the overall SUS score of the low compliance HMI (see Table 3). On average, the difference between the high and low compliance was one scale step on most items besides items 1 (“I think that I would like to use this system frequently”) and 4 (“I think that I would need the support of a technical person to be able to use this system”). Descriptive statistics of the van-der-Laan sale showed differences on each single item with more favorable ratings for the high compliance HMI. Independent from the experimental conditions, mean ratings tended towards the positive end of the semantic differential. Higher mean ratings were observed for both the satisfaction and usefulness subscale (see Table 3). Differences were largest on the “nice-annoying” and “good-bad” items being close to one. The smallest difference was observed for the “raising alertness-sleep inducing” item.

Empirical Validation of a Checklist for Heuristic Evaluation

9

Fig. 2. Means and standard deviations for the 10 items of the SUS by HMI compliance. Note that even numbered items are reverse scored and thus lower values indicate higher satisfaction.

Fig. 3. Means and standard deviations for the 9 items of the van-der-Laan scale by HMI compliance. Note that a negative score refers to the first word of the semantic differential.

To determine whether the adherence of the HMIs to the checklist is reflected in selfreport measures of perceived usability and acceptance, a paired samples t-test was calculated for the SUS scores and a 2-factorial repeated measures ANOVA was

10

Y. Forster et al.

calculated for the van-der-Laan scale. The within-subject factors were HMI compliance (high vs. low) and number of subscales. Table 3 shows descriptive (i.e., M, SD) and inferential results for the main effect of the HMI and the interaction between HMI and subscale (i.e., Wilk’s k). Statistically significant results are colored in grey. Results revealed that the experimental manipulation produced statistically significant differences in perceived usability and acceptance.

Table 3. Descriptive (i.e., M, SD) and inferential (i.e., main effect for HMI and Interaction HMI  Subscale) statistics for each scale.

Scale

high Comp. M (SD)

low Comp. M (SD)

SUS

82.45 (14.01)

67.11 (19.14)

Usefulness Sa sfac on

1.24 (0.52) 1.24 (0.59)

0.73 (0.66) 0.57 (0.80)

3.3

Main effect HMI t(56)=5.959, p .99) (Table 3). Interestingly, comparing post-test trust scores with all three scenario trust scores, revealed a significant main effect of Scenario (F(1.37, 17.85) = 3.86, p = .016, η2p = .23), Table 3. Mean PANAS and trust scores after each of three Scenarios and post-test scores overall and as a function of Group (Day/Night)

Day

PANAS affect PANAS affect Trust Night PANAS affect PANAS affect Trust

Audio/No PickUp 36.0 (9.0)

No Audio/PickUp 36.6 (8.5)

Post test

positive

Audio/PickUp 34.0 (6.8)

negative

11.3 (2.5)

10.1 (0.4)

10.4 (1.1)

N/A

positive

3.3 (0.8) 36.3 (11.5)

2.9 (1.4) 10.1 (0.4)

3.8 (0.5) 39.7 (9.1)

3.4 (0.6) N/A

negative

7.6 (4.5)

7.4 (5.1)

8.1 (3.9)

N/A

3.0 (0.5)

3.7 (0.7)

3.9 (0.5)

3.9 (0.5)

N/A

a non-significant main effect of Group, and a non-significant Scenario x Group interaction (all Fs(1.37, 17.85)  2.48, all ps  .126). Pairwise comparisons indicated that trust post-test scores significantly increased compared to Audio/Pick-Up (p = .032) and No-Audio/Pick-Up Scenario scores (p = .022), but there was no significant difference with Audio/No Pick-Up Scenario Trust scores (p < .999) (Table 3). PANAS Scores. A mixed ANOVA with a between-subject measure of Group (day/night) and within-subject measures of Scenario and PANAS Factor (positive affect/negative affect) revealed a significant main effect of PANAS Factor (F(1, 13) = 108.83 p  .001, η2p = .893). The main effect of Scenario as well as was as main effects of Group, interactions Scenario x Group, Scenario x Factor, Factor x Group, Scenario x Factor x Group were non-significant (all Fs(2, 26)  2.54, all ps  .099). An investigation of the main effect of PANAS Factor revealed participants overall felt significantly more positive affect compared to negative affect (t(14) = 10.41, p  .001; positive affect mean = 36.60, SD = 9.01, negative affect mean = 9.24, SD = 3.35). Other comparisons were not significant (all ts(14)  1.45, all ps  .169, Table 3).

The Impact of Different Human-Machine Interface Feedback Modalities

3.4

129

Cognitive Demand as a Function of Scenario

Situation Awareness. A mixed ANOVA with a between-subject measure of Group (Day/Night) and within-subject measure of Scenario and dependent variable of SART score revealed a non-significant main effect of Scenario and Group (F(1, 13) = 1.25, p = .25) and a marginally non-significant Scenario x Group interaction with a clear trend (F(2, 26) = 3.24, p = .056, η2p = .20)(Fig. 2b). A further investigation of marginally non-significant interaction of Scenario x Group revealed that the day participant Group scored higher than night Group participants in the Audio/Pick-Up Scenario, although it was a non-significant trend. Other t-test comparisons were also non-significant (all ts(13)  1.37, all ps  .194). Task Load. A mixed ANOVA with a between-subject measure of Group (day/night) and within-subject measures of Scenario and Factor (mental demand, physical demand, temporal demand, effort, performance, frustration) revealed significant main effects of Scenario (F(2, 26) = 5.46, p = .011, η2p = .296) and Factor (F(2.19, 28.42) = 5.58, p  .001, η2p = .300). Other main effects and interactions were not significant (all Fs (2.19, 28.42)  1.19, all ps  .325). Exploration of the main effect of Scenario showed that overall cognitive load was significantly higher in No-Audio/Pick-Up scenario compared to Audio/Pick-Up scenario (p = .011). The mean differences between Audio/Pick-Up vs no pick Audio as well as Audio/Pick-Up vs. No-Audio/Pick-Up were not significant, p = .34 and p = .41, respectively (Fig. 2c). Furthermore, looking at the six component factors of this measure, there were significant differences between mental demand vs. physical demand and physical demand vs. performance (ps = .041 and p = .012, respectively). Other mean differences were not significant (all ps  .126; Fig. 2d).

4 Discussion The current study investigated the effect of different feedback modalities (audio/noaudio/with text) within a CAV human-machine interface (HMI) on user experience of older participants who underwent a series of level 5 autonomy simulated journeys. Subjective ratings of affect included trust and positive and negative mood state. Situational awareness and workload scales provided measures of task load and engagement, and physiological measures including heart rate provided an index of physiological arousal for comparison with the subjective measures. Skin temperature, heart rate and EDA indicated increased response, reflecting greater arousal, in the No-Audio/Pick-Up Scenario. Conversely, the lowest response was in the Audio/Pick-Up Scenario. The highest arousal scores were experienced in the No-Audio/Pick-Up condition, and this is consistent with literature showing that not knowing what is happening is related to the feeling of being out of control, while on the other hand, a feeling of being in control can decreased experienced stress and anxiety [8–10]. Furthermore, higher scores on EDA, heart rate and skin temperature measures were associated with higher experienced negative emotions and anxiety [20, 21].

130

I. Eimontaite et al.

Affect measures indicated overall higher positive affect scores than negative affect scores. This is perhaps not surprising as the participants were a self-selecting group. In terms of trust, the highest ratings were in the No-Audio/Pick-Up scenario. Although at first sight this result seems counterintuitive, it could be explained by considering the performance of the CAV in the simulation. In this no Audio condition participants did not know what actions the CAV simulator was going to take (No Audio), therefore successful completion of the journey could have led towards increased trust compared to scenarios where participants were informed of CAV behaviour. Furthermore, and consistent with the past literature, trust and attitudes towards automation increases with greater exposure to it [22, 23] and depends on the automation behavioural characteristics – such as working without errors [24]. Therefore, it is not surprising that participants’ post-scenario measures of trust show higher trust scores compared to Audio/No Pick-Up and Audio/Pick-Up scenarios. Furthermore, cognitive measures indicated that scenario type affected participants’ cognitive response to the CAV. With situation awareness, the interaction between scenario type, and whether participants experienced the journey in a day or night environment, resulted in a trending difference in situation awareness scores. Participants in the Night condition showed increased situation awareness in the Audio/No Pick-Up and Audio/Pick-Up scenarios compared to participants in the Day, yet in the Audio/Pick-Up scenario, both groups of participants had very similar scores. Decreased situation awareness suggests that individuals relax and trust the technology, in this case the CAV. On the other hand, high situation awareness indicates that individuals are in fight or flight readiness and suggests feelings of tension towards the environment [25]. Results reveal situational awareness and workload ratings in the no-Audio, pick up a friend condition, compared to the only condition. This was supported by the greatest increase in heart rate suggesting higher levels of both physiological and subjective arousal in the No Audio condition with similar levels in the two Audio conditions. Trust scores increased significantly post-test after the pick-up conditions, with positive affect higher than negative affect throughout. Taken together, the findings indicate that older adult participants found the simulated CAV journey a positive experience with increasing trust, based on their HMI interaction and journey experience. However, the greatest concentration was required in the no-Audio notifications condition, suggesting that sound/multimodal feedback improved ease of operation and journey experience. The findings are important to inform future CAV HMI design guidelines for this user group. In particular, clear communication of vehicle behaviour to increase trust and user experience. Furthermore, information communication is recommended in both modalities (text and audio feedback) to help the user feel more relaxed during the journeys and trusting that the vehicle will cope with user specified decisions. Future work will explore aspects such as how the user is affected by different levels of explainability of vehicle behaviour and together with user control of the level and modality of the feedback from the vehicle.

The Impact of Different Human-Machine Interface Feedback Modalities

131

References 1. Jee, C., Mercer, C.: Driverless car news: the great driverless car race: where will the UK place? https://www.techworld.com/apps-wearables/great-driverless-car-race-where-will-ukplace-3598209/ (2017) 2. SAE International: U.S. Department of transportation’s new policy on automated vehicles adopts SAE International’s levels of automation for defining driving automation in on-road motor vehicles (2016) 3. Musselwhite, C., Haddad, H.: Mobility, accessibility and quality of later life. Qual. Ageing Older Adults 11, 25–37 (2010) 4. Abraham, H., Lee, C., Brady, S., Mehler, B., Reimer, B., Coughlin, J..: Autonomous vehicles and alternatives to driving: trust, preferences, and effects of age. Presented at the Transportation Research Board 96th Annual Meeting, Washington DC, United States (2017) 5. Moreno-Jiménez, B., Rodríguez-Carvajal, R., Garrosa Hernández, E., Morante Benadero, M.A., et al.: Terminal versus non-terminal care in physician burnout: the role of decision-making processes and attitudes to death. Salud Ment. 31, 93–101 (2008) 6. Mills, M.E., Sullivan, K.: The importance of information giving for patients newly diagnosed with cancer: a review of the literature. J. Clin. Nurs. 8, 631–642 (1999) 7. Ussher, J., Kirsten, L., Butow, P., Sandoval, M.: What do cancer support groups provide which other supportive relationships do not? The experience of peer support groups for people with cancer. Soc. Sci. Med. 62, 2565–2576 (2006) 8. Lautizi, M., Laschinger, H.K.S., Ravazzolo, S.: Workplace empowerment, job satisfaction and job stress among Italian mental health nurses: an exploratory study. J. Nurs. Manag. 17, 446–452 (2009) 9. Ozer, E.M., Bandura, A.: Mechanisms governing empowerment effects: a self-efficacy analysis. J. Pers. Soc. Psychol. 58, 472 (1990) 10. Pearson, L.C., Moomaw, W.: The relationship between teacher autonomy and stress, work satisfaction, empowerment, and professionalism. Educ. Res. Q. 29, 37 (2005) 11. Morgan, P., Caleb-Solly, P., Voinescu, A., Williams, C.: Literature review: human-machine interface. Project report, UWE Bristol, Bristol (2016) 12. Morgan, P.L., Voinescu, A., Williams, C., Caleb-Solly, P., Alford, C., Shergold, I., Parkhurst, G., Pipe, A.: An emerging framework to inform effective design of humanmachine interfaces for older adults using connected autonomous vehicles. In: Stanton, N.A. (ed.) Advances in Human Aspects of Transportation, pp. 325–334. Springer (2018) 13. Gable, T.M., Walker, B.N., Gable, T.: Georgia tech simulator sickness screening protocol, 16 (2013) 14. Watson, D., Anna, L., Tellegen, A.: Development and validation of brief measures of positive and negative affect: the PANAS Scales. J. Pers. Soc. Psychol. 54, 1063–1070 (1988) 15. Hart, S.G., Staveland, L.E.: Development of NASA-TLX (task load index): results of empirical and theoretical research. In: Advances in Psychology, pp. 139–183. Elsevier (1988) 16. Buysse, D.J., Reynolds III, C.F., Monk, T., Berman, S.R., Kupfer, D.J.: The Pittsburgh sleep quality index: a new instrument for psychiatric practice and research. Psychiatry Res. 28, 193–213 (1989) 17. Taylor, R.M., Selcon, S.J.: Cognitive quality and situational awareness with advanced aircraft attitude displays. In: Proceedings of the Human Factors Society Annual Meeting, vol. 34, pp. 26–30 (1990)

132

I. Eimontaite et al.

18. Gold, C., Körber, M., Hohenberger, C., Lechner, D., Bengler, K.: Trust in automation – before and after the experience of take-over scenarios in a highly automated vehicle. Procedia Manuf. 3, 3025–3032 (2015) 19. Ben-Shakhar, G.: Standardization within individuals: a simple method to neutralize individual differences in skin conductance. Psychophysiology 22, 292–299 (1985) 20. Bechara, A., Damasio, H., Tranel, D., Damasio, A.R.: The Iowa Gambling Task and the somatic marker hypothesis: some questions and answers. Trends Cogn. Sci. 9, 159–162 (2005) 21. Ben-Shakhar, G., Bornstein, G., Hopfensitz, A., Van Winden, F.: Reciprocity and emotions: arousal, self-reports, and expectations. CESifo Working Paper 1298, pp. 1–16 (2004) 22. Bartneck, C., Suzuki, T., Kanda, T., Nomura, T.: The influence of people’s culture and prior experiences with Aibo on their attitude towards robots. AI Soc. 21, 217–230 (2006) 23. Stafford, R.Q., Broadbent, E., Jayawardena, C., Unger, U., Kuo, I.H., Igic, A., Wong, R., Kerse, N., Watson, C., MacDonald, B.A.: Improved robot attitudes and emotions at a retirement home after meeting a robot. In: 2010 IEEE RO-MAN, pp. 82–87. IEEE (2010) 24. Nomura, T., Shintani, T., Fujii, K., Hokabe, K.: Experimental investigation of relationships between anxiety, negative attitudes, and allowable distance of robots. In: Proceedings of the 2nd IASTED International Conference on Human Computer Interaction, Chamonix, France, pp. 13–18. ACTA Press (2007) 25. McLucas, A.C.: Decision Making: Risk Management, Systems Thinking and Situation Awareness. Argos Press P/L, Canberra (2003)

User Experience in Immersive VR-Based Serious Game: An Application in Highly Automated Driving Training Mahdi Ebnali1(&), Cyrus Kian2, Majid Ebnali-Heidari3, and Adel Mazloumi4 1 Applied Cognitive Engineering Lab, Industrial and System Engineering Department, University at Buffalo, Buffalo, USA [email protected] 2 Department of Information Sciences, Cornell University, Ithaca, USA 3 Electrical Engineering Department, Shahrekord University, Shahrekord, Iran 4 Department of Occupational Health Engineering, School of Public Health, Tehran University of Medical Sciences, Tehran, Iran

Abstract. In the way of smoothing driver interaction with highly automated vehicles, we designed an immersive (VR + serious game) training program with a focus on improving drivers’ mental model. Then, we tackled the usability flaws and upgraded the preliminary serious game (PSG) to usability-improved serious game (USG). Three groups of participants-no-training, PSG and USGwere tested to explore the effects of immersive training on drivers performance and experiences in highly automated driving. The results showed that both training programs significantly improved driving performance and resulted in faster takeover time (TOT), longer time-to-collision (TTC), and fewer number of the collision. Moreover, the participants in training groups reported less erratic acceptance and more calibrated trust compared to the control group. Although improving usability in USG led to better flow experience (enjoyment and engagement) and lower cognitive load during the learning process, it did not contribute significantly to training transferability. Keywords: Highly automated driving
Immersive training Virtual Reality
Usability
Mental model


Serious game


1 Introduction Autonomous vehicles can potentially improve transportation safety by assisting drivers in several ways depending on the level of automation. According to the taxonomy proposed by Society of Automotive Engineers (SAE) International J3016, automation is provided in five levels expanded from level 0, no automation, to level 5 or fully autonomous vehicle. The more functions are supported in more driving modes, the more automation is provided in a car. Fully autonomous cars promise that system is able to perform driving tasks alone in all driving modes. However, because of limitations in technology and infrastructures, most of the autonomous vehicles are highly dependent on drives. For example, automation level 2 contributes in both of primary © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 133–144, 2020. https://doi.org/10.1007/978-3-030-20503-4_12

134

M. Ebnali et al.

driving functions, speed control, and lane keeping, however, the driver is still responsible for fully monitoring the road, detecting the hazards, and reacting appropriately. More importantly, the driver needs to retake the control of car in critical situations, some cases that automation is not able to control. Shifting from regular driving to autonomous driving accompanies with several benefits like higher safety in many cases, lower physical fatigue, and engaging in some secondary and enjoyable tasks; however, it exposes drivers to the new difficulties, referred as the irony of automation. Since the role and responsibilities of drivers change from manual controlling to autonomous driving, they need to contribute in several transitions from autonomous driving mode to manual driving, and reversely. Studies in autonomous driving have identified several challenges for drivers such as un-calibrated trust [1, 2], inadequate mental model [3], erratic workload [4], loss of situational awareness (SA) [4], behavioral adaptation [5], and skill degradation [6]. Several studies have proposed training approach as a potential solution to improve human-automation teaming and minimize human factors challenges [5, 7, 8]. Drivers need to specifically be trained and acquire appropriate knowledge and skills in all aspects of partnership with automated vehicles, especially in emergency cases. Still, a very limited number of studies have evaluated if the training system could improve drivers’ interaction with highly automated vehicles. This study aims to explore how training alleviates human factors challenges such as un-calibrated trust and acceptance, erratic workload, and impaired takeover performance in highly automated driving. In order to improve the learning transferability and trainees’ experiences during the learning process, we used Virtual Reality (VR) and serious game features in designing a training program [9, 10]. These two elements promise to trigger users’ attention and deliver the learning contents in more immersive ways. Growing scientific interests in the application of game-based training discussed the quality of serious game as an effective approach in transferring educational contents and skills [10, 11]. Serious game is an entertainment medium designed to mix learning principles with game mechanics and logic, and brings about behavioral changes in its players. Furthermore, using VR in training process has been suggested to lead more learning and changes in intended behaviors compared to traditional approaches. In autonomous driving context, a recent study reported that VR could significantly improve training efficiency in fully autonomous driving [9]. VR-based training can simulate the learning scenarios in more realistic ways and transfer additional sensory inputs as images, sounds, and 3d objects. In this study, we evaluated how VR-based serious game effects on driving performance and experience in an autonomous vehicle. In addition, we tested how improving the usability of the serious game reflexes in its user experiences and training outcomes.

2 Designing Phase While a significant number of models already exist in gamification field, DPE (design, play, experience) model, proposed by [12], specifically covers the most important elements in serious game definition. DPE (Fig. 1) depicts the relationship of designer and player in an iterative process where designer constructs the serious game and

User Experience in Immersive VR-Based Serious Game

135

player engages on it, and provides the designer with feedbacks. This framework decomposes the design process into four main elements of serious game: learning content, storytelling, game-play, and user experience. To follow a standard and verified model, we used this framework and its subcomponents in the designing phase.

Fig. 1. The Design, Play, and Experience (DPE) framework by (Winn 2008)

2.1

Learning Content

The first layer in DPE model covers the learning materials and educational contents that are determined based on the learning outcomes. To understand what should be the focus of training program, we reviewed the most important challenges and concerns in highly automated driving. Well-supported results show that driving in highly automated cars leads to several challenges and concerns which impair driver-automation cooperation. Some of these challenges such as erratic workload, loss of SA, and impaired performance originated from un-calibrated trust, un-balanced acceptance, and inaccurate mental model [13]. Un-calibrated trust such as over-trust increases drivers’ tendency to be out of the loop and consequently react inappropriately in takeover requests. This challenge is more problematic when drivers need to take control of the car immediately, and but actually, it takes seconds for an out-of-the-loop driver. Low trust also damages drivers’ willingness to use the features or sometimes it causes inappropriate intervention on the system. Experimental results showed that the driver’s mental model [14] and experience [3] have important contribution in shaping the trust and acceptance. The more experience and correct mental model people have, they build more calibrated trust and acceptance. Either of these inappropriate consequents could jeopardize driver’s safety [15]. The knowledge of how a system works is technically discussed as people mental model of the work. Mental models play a decisive role in cognitive information processing system as it effects on how human collects specific information, interprets, anticipates, and consequently performs an action [16]. User’s previous experience and attitude are crucial contributors to building a preliminary mental model. These preliminary mental models will be updated across the course of events where environmental sampling modifies the internal representation of the world. Wrong or inaccurate mental model hurts drivers-automation interaction and causes misuse and abuse.

136

M. Ebnali et al.

For example, drivers may overestimate the capability of an automated features such as lane keeping assistant because of an inaccurate mental model, and it may upset performance or cause crash in critical situations. As discussed by [6], the training could help people to gain appropriate knowledge, acquire necessary skills, and improve their mental model. This improvement leads to build a balanced trust and acceptance, in the course of the time. Accordingly, this study defines three main learning outcomes for the serious game based on KSA (knowledge, skill, and attitude) taxonomy: (a) Establishing correct mental model by providing enough declarative and procedural knowledge of what is the automated driving, its capabilities, and limitations, (b) practicing automated driving skills and rules in normal and critical situations (c) improving drivers attitude toward the system and avoiding over and under trust. 2.2

Storytelling and Game-Play

The second and third layers in DPE framework, storytelling and game-play, specify how the serious game provides an interesting story with a meaningful mission and rewards. As shown in Fig. 2, we considered three levels with different training concentrations. The first level focuses on knowledge, the second level enriches users’ skills and understanding of the rules, and the last level covers both of the previous learning outcomes plus focusing on constructive feedback regarding trust and attitude. Since this study focused on partially automated driving, the serious game encompasses a combination of lane keeping assistant (LKA) and adaptive cruise control (ACC) supports in all of three levels. The training starts with an introduction in which a coach (a user-selected fun character) appears and, in a narrative way, defines the mission (passing the icy/rocky roads) in three levels, rewards (coins), and the goal (buying a super-powerful automated car). Two types of feedbacks are provided in a visual format, one is “instructive” and guides the users through the levels, and the other one is “encouraging” and tries to establish a correct behavior or action. Rewards are in the form of funny coins and can be collected when users succeed in some small challenges in each level. The difficulty of levels increases from level 1 to 3, and at the end of each level, users need to take a test and be qualified to enter to the next level. 2.3

User Experiences

The last and deepest layer in the serious game design process, user experiences, represents how the game immerses players, what they are experiencing, and how interaction happens. Prior research has discussed the importance of usability of serious game in transferring the learning contents [17, 18]. The design team constructed a preliminary serious game (PSG) prototype, then using a heuristic usability testing, we improved the issues, and developed usability-improved serious game (USG).

User Experience in Immersive VR-Based Serious Game

137

3 Usability Improvement Phase Expert Usability Testing: In order to understand to what extent a user is able to comprehend, learn and fulfill the task, we run a heuristic usability testing. Similarly, [19, 20] used heuristic usability testing to evaluate the playability of serious games. This testing does not focus on Game-Play parts like entertainment, storyline and engagement, and instead, it assesses the prototype based on user-centered design principles. Three independent evaluators performed heuristic evaluation using Nielsen’s 10 scales [21]. This test involved the evaluation of PSG against 10 usability heuristics to identify the issues. This method systematically assesses an application to understand users’ needs and limitations based on a list of user-centered design principles. Usability testing revealed several critical issues, especially in feedback, visibility, flexibility, and consistency. Based on these issues, we identified eight main improvements in the PSG and returned them to the design team. They applied five proposed changes and renovated the design in USG. Other three items were ignored because those changes were too costly and time-consuming.

Fig. 2. User journey mapping- three levels of serious game with three different concentrations

User Usability Testing: Studies in serious games have discussed that when an appropriate amount of learning content is embedded in a serious game, users experience lower cognitive load in the learning process. We used “cognitive load scale” [22] to evaluate how improving usability in USG effects on cognitive load compared to PSG. This method assesses the serious game in three dimensions: intrinsic (difficulty and easiness of comprehension), extraneous (negative load), and germane (amount of effort for learning). Moreover, the “flow experience scale” was used to evaluate how optimal is the users’ experience. Flow, as a concept proposed by [23], refers to how

138

M. Ebnali et al.

players perceiving a balance between skills and challenges, and concentrate fully on a pleasant tasks. This scale involves 12 items in three constructs: enjoyment, engagement, and control.

4 Method 4.1

Participants

Fifteen university students, 9 males and 6 females with valid driver’s licenses were recruited to participate in the study. They were aged between 19 and 29 years old (Mean = 24.1 years, SD = 3.82) and had between 3 and 9 years (M = 4.99, SD = 3.85) of driving experience. To compare the effects of training systems and usability improvement on drivers’ experiences and performance in highly automated driving, we formed three groups of 5-people (randomly selected): no training, PSG, and USG. All participants gave informed consent and were debriefed after the experiment. 4.2

Variables

Trainees’ experiences variables: Heuristic usability dimensions, cognitive load, and flow experiences were measured to assess the usability and playability of PSG and USG. Driving behavioral variables: To analysis driving performance in different modes, we collected takeover time (TOT), time to collision (TTC), the total number of collisions during take-over, and the subjective rating of the participants. Takeover time was measured from the moment the takeover request was issued until the moment that participants disengaged take the control of the car-using steering wheel, gas pedal or brake pedal. Since no automation provided in regular driving, TOT did not apply to the manual driving condition. In all of the take over request events, drivers faced with situations having objects such as lead vehicles, parked cars, heavy traffic, and crossing passengers ahead of the car. TTC was measured from the moment of issuing the warning until the moment the driver takes control of the car. In addition, subjective variables such trust and acceptance were collected using a simple questionnaire with six questions in a Likert scale, and workload was measured by NASA TLX. 4.3

Experimental Design

This study used a two-factor mixed-design, with the type of training as the betweengroups variable and the driving mode as the within-subjects variable. The driving mode includes 1-partial automation driving engaged in secondary tasks, 2-partial automation driving eyes on the road, and 3-manual driving. Each experimental session consisted of driving in the similar roads, having equally distributed sections for three driving modes, each part lasted around 10 min, and the order was counterbalanced. In order to simulate non-driving task engagement, the participants were asked to engage in instructed secondary tasks of watching a video on a smart phone. After issuing the takeover request, the subjects had to resume control immediately and continue manual driving on a normal highway traffic while they keep the range of instructed speed (90–110 km/h). Manual driving condition served as a baseline.

User Experience in Immersive VR-Based Serious Game

4.4

139

Procedure

After playtesting with three persons with no or low experience in game playing, the average time for completing all levels and achieving the final prize was 145 min for PSG and 110 min for USG. To achieve a higher training transferability, we asked the participants in the game groups to finish the game from start to the end for three times within two weeks. Before starting the experiments, the participants were shown a demonstration of the driving and details of the tasks they needed to do during the experiments. It was followed by a 5-min familiarization with the simulator, road, and how to interact with the partially automated car. The participants received a clear explanation that they are always responsible for supervising and monitoring the autonomous driving. To assure that they clearly understand the concept of takeover request, experimenters showed two examples of retake-control request scenarios and instructed them to resume the control immediately after hearing the takeover request. 4.5

Apparatus

We used Unity game engine (v5.4, non-commercial version) to simulate driving environment and tasks, in which objects and scenes were projected to two monitors (Dell E Series 23-in. Screen LED-lit Monitor). To increase the degree of reality, instead of playing the engine’s sound from the front speaker, two speakers were placed parallel to subjects’ head. A partial automation driving or level 2 [24] was designed such that handles lateral and longitudinal controls and requires driver’s continuous monitoring. The simulated system was operational on a highway, at the speeds 80–120 km/h. We used a visual message to inform the participants about the driving mode: automated driving or regular driving. There was no option for drives to active or de-active the automation intentionally, and the driving modes, events, and scenarios were determined by road pad, time, and expression triggers. The roads entities and vehicle model also were imported from EasyRoads3D package and included mostly straight with three-way highway with a speed limit of 110 km/h. All of driving parameters, including TTC, TOC, and the number of collisions were logged by the Unity cells, and extracted by unity analytics package. Because of technical limitation, lateral deviation, pedal force, and steering wheel variance parameters were not collected.

5 Results 5.1

Usability Analysis

The results of heuristic usability testing for PSG and USG is shown in Fig. 3. Based on paired t-test, the number of usability violations of Nielson 10 scale testing is significantly lower in USG compared to PSG in error prevention (p < 0.05), flexibility (p < 0.05), feedback(p < 0.01), memory (p < 0.01), and consistency (p < 0.01). This finding is consistent with the result of the user usability testing. Comparing the usability and playability of PSG and USG using “cognitive load scale” showed that the participants reported lower total cognitive load (p < 0.01) in USG compared to PSG.

140

M. Ebnali et al.

It seems that improving the usability- informed from heuristic testing- effectively reduced the cognitive load that users experience while learning in the USG. Figure 4a shows the results separated for each dimension of the cognitive load scale. Details in this figure demonstrate that the participants reported lower comprehension difficulty and negative load in USG, however, the amount of effort they exerted in the learning process did not differ meaningfully.

Fig. 3. Comparing heuristic usability violation before and after improving usability

The result of flow experience scale shows a notable difference in enjoyment (p < 0.05) and engagement (p < 0.01) between two groups of training versions. This finding reveals that the participants rated more interest, curiosity, and concentration in USG. This may imply that improving usability positively effect on serious game immersion and triggered trainee’s interest and attention. The last part in flow scale, control effect, was not improved in USG compared to PSG. This dimension involves questions related to how learning process was frustrating, how learning concepts are presented, and how things are in the control of users. It seems that usability improvement in USG has enhanced the participants’ flow experiences in terms of enjoyment and engagement, but it was not effective in control element. 5.2

Driving Analysis

We used ANOVA in Python 3.7 to explore the effect of independent variables (the type of training) on driving parameters. A remarkable portion of PSG trained participants (60%) and USG trained participants (80%) rated trust questions in middle ranges (3 and 4) of scale compared to the participants in the control group (40%). The analysis discloses a statistically meaningful effect of type of training on users’ trust (F = 8.32, p < 0.05). This finding implies that training and improving usability in USG significantly shift the drivers’ trust from unbalanced rating (very trustworthy and untrustworthy) to more balanced rating (trustworthy). Similarly, the participants in training groups reported less erratic acceptance compared to the people in the control group (p < 0.05). However, the acceptance was not affected by improving usability in USG

User Experience in Immersive VR-Based Serious Game

141

compared to PSG. Moreover, the analysis of self-reported workload did not indicate a remarkable change in workload score among the three groups of the experiment. Neither training nor improving usability effects on the users ‘workload in highly automated driving scenarios.

Fig. 4. (a) Effects of improving usability on users’ cognitive load in three dimensions (b) mean of takeover time (TOT) in different groups (in seconds)

A repeated measures ANOVA was run to test the effect of type of training (between-groups factor) and driving condition (within-subject factor) on driving behavioral data. As shown in Fig. 4b, the main effect of training on TOT was significant when comparing training groups with the control group (P < 0.01, M training = 2.16 s, SD training = 0.88, M control = 2.98 s, SD control = 0.69). Likewise, the analysis displays notable effects of training on TTC (P < 0.01) and the number of crashes (P < 0.05). The participants in training groups took control of the car while showing longer TTC, and they contributed in a fewer number of crashes. Interestingly, exploring the difference of driving behavior parameters such as TOT, TTC, and the number of crashes did not illustrate any meaningful variation as a result of type of training (PSG and USG). This may imply that usability upgrading did not contribute to driving performance improvement. Within-subject analysis shows that involving in secondary tasks while driving in partially automated mode resulted in longer TOT (P = 0.01, M engaged = 2.91 s, SD engaged = 0.26, M not-engaged = 2.11 s, SD engaged = 0.38), shorter TTC (P = 0.02, M engaged = 5.12 s, SD engaged = 1.19, M not-engaged = 8.71 s, SD engaged = 2.8), and increased number of crashes (P = 0.01, M engaged = 4.3, SD engaged = 0.98, M not-engaged = 1.01 s, SD engaged = 0.18).

6 Discussion Training has been proposed as an effective approach to improve driver-automation interaction, however, very limited studies evaluated how educating people effects on their performance in highly automated driving contexts [6, 9, 25]. In this study, we designed an immersive training program using VR and serious game elements. In the next step,

142

M. Ebnali et al.

we improved the usability flaws and upgraded the preliminary serious game (PSG) to usability-improved serious game (PSG). Then, three groups of participants-no-training, PSG and USG- were tested to explore the effects of immersive training and usabilityimproved training on drivers performance and experiences in highly automated driving. Our results showed that both training programs significantly improved driving performance. The participants in PSG and USG groups showed faster TOT, longer TTC, and fewer number of the collision. Moreover, training programs improved trust and acceptance. In line with our results, [3] investigated how a training program effects on drivers’ trust, acceptance and mental model while driving with adaptive cruise control (ACC). Their results revealed a relatively calibrated trust and acceptant after a few training session. Recently, in a highly automated driving experiment, [6] showed that elaborated trained drivers performed better to handle manual control recovery compared to other groups. These findings imply that the training could effectively explain the underlying logics of automated driving to drivers, and helps them to build an appropriate mental model. This improved mental model refines drivers’ expectation of what automated car is able to do, and consequently may lead to optimized trust and acceptance. Previous studies have discussed that immersive features such as VR and serious game elements in training systems successfully improved the training impacts. [9], for example, used head-mounted display (HMD)-based VR for training people in highly automated driving conditions. Similar to this finding, our analysis revealed that integrating VR and serious game features into training program brings about improved performance among trained participants compared to control group. However, since we incorporated both of immersive features, VR and serious game elements, in training programs, it is unclear that which one of these immersive add-ons was more effective in terms of learning outcome transferability and user experiences. Further studies are required to explore VR and serious game features on driving training system, independently. Our results acknowledge that immersive training program combining VR and serious game features effectively deliver learning outcome in highly automated driving training. These learning outcomes facilitate driver interaction with highly automated vehicles by faster takeover time, and longer time to collision. Moreover, this training system leads to more balanced trust and acceptance. However, the level of workload the participants experienced in the experiment was not affected by training interventions. One possible explanation of this result is that because of the simulated nature of the experiments the participants did not experience high fluctuation in workload. Likewise, [26] found that driving task load decreases in a driving simulator, while it remains high in a real driving condition. Trainees reported superior experiences while learning with USG compared to PSG. This can be explained in view of the fact that usability violations identified in USG were significantly lower than PSG. Similarly, user usability testing revealed upper flow score such as higher enjoyment and enjoyment for USG. Furthermore, the participants in USG group reported lower comprehension difficulty and negative load. These findings confirmed that heuristic usability testing successfully acknowledged noteworthy pitfalls that directly effects on users’ experiences and amount of cognitive load. Although refining the usability issues clues users to bettered experiences during the learning process, the analysis revealed that PSG and USG similarly influence on driving performance in highly automated driving. In fact, since the same learning

User Experience in Immersive VR-Based Serious Game

143

content and information was embedded in both of training groups, PSG and USG had similar transferability and improvement in driving performance. Improving usability in USG did not promise more opportunities to increase training transferability. This may imply that in an immersive serious game, educational content plays more influential role in learning process compared to usability aspects.

7 Limitations Despite this study’s contribution to the literature, some limitations are worth mentioning. A small sample of participants and the unbalanced ratio of gender decrease the power of generalizability of the results. Moreover, both of training groups received the same educational content and used the same medium (VR-based serious game) which are not different in many aspects but in usability. In a future study, we will consider one more training group with a focus on conventional training such as video training and will compare learning outcome across training program that is different in terms of content, medium, and usability. We also did not explore other important driving performance measures such as lane-keeping task. Because of the technical limitation in our simulator, we were not able to collect lane deviation data accurately. Furthermore, in future studies, more driving scenarios and critical events with longer driving time need to be investigated.

References 1. Miller, D., et al.: Behavioral measurement of trust in automation: the trust fall. In: Proceedings of the Human Factors and Ergonomics Society Annual Meeting. SAGE Publications Sage, Los Angeles (2016) 2. Hergeth, S., et al.: Keep your scanners peeled: gaze behavior as a measure of automation trust during highly automated driving. Hum. Factors 58(3), 509–519 (2016) 3. Beggiato, M., et al.: Learning and development of trust, acceptance and the mental model of ACC. A longitudinal on-road study. Transp. Res. Part F Traffic Psychol. Behav. 35, 75–84 (2015) 4. De Winter, J.C., et al.: Effects of adaptive cruise control and highly automated driving on workload and situation awareness: a review of the empirical evidence. Transp. Res. Part F Traffic Psychol. Behav. 27, 196–217 (2014) 5. Stanton, N.A., Young, M.S.: Driver behaviour with adaptive cruise control. Ergonomics 48 (10), 1294–1313 (2005) 6. Payre, W., et al.: Impact of training and in-vehicle task performance on manual control recovery in an automated car. Transp. Res. Part F Traffic Psychol. Behav. 46, 216–227 (2017) 7. Rudin-Brown, C.M., Parker, H.A.: Behavioural adaptation to adaptive cruise control (ACC): implications for preventive strategies. Transp. Res. Part F Traffic Psychol. Behav. 7(2), 59– 76 (2004) 8. Saffarian, M., de Winter, J.C., Happee, R.: Automated driving: human-factors issues and design solutions. In: Proceedings of the Human Factors and Ergonomics Society Annual Meeting. Sage Publications Sage, Los Angeles (2012)

144

M. Ebnali et al.

9. Sportillo, D., Paljic, A., Ojeda, L.: Get ready for automated driving using virtual reality. Accid. Anal. Prev. 118, 102–113 (2018) 10. Mettler, T., Pinto, R.: Serious games as a means for scientific knowledge transfer—a case from engineering management education. IEEE Trans. Eng. Manage. 62(2), 256–265 (2015) 11. van der Kuil, M.N., et al.: A usability study of a serious game in cognitive rehabilitation: a compensatory navigation training in acquired brain injury patients. Front. Psychol. 9, 846 (2018) 12. Winn, B.M.: The design, play, and experience framework. In: Handbook of Research on Effective Electronic Gaming in Education, pp. 1010–1024. IGI Global (2009) 13. Hoogendoorn, R., van Arerm, B., Hoogendoom, S.: Automated driving, traffic flow efficiency, and human factors: literature review. Transp. Res. Rec. 2422(1), 113–120 (2014) 14. Goodrich, M.A., Boer, E.R.: Model-based human-centered task automation: a case study in ACC system design. IEEE Trans. Syst. Man Cybern. Part A Syst. Hum. 33(3), 325–336 (2003) 15. Lee, J.D., Hoffman, J.D., Hayes, E.: Collision warning design to mitigate driver distraction. In: Proceedings of the SIGCHI Conference on Human factors in Computing Systems. ACM (2004) 16. Neisser, U., Cognition and Reality: Principles and Implications of Cognitive Psychology. WH Freeman/Times Books/Henry Holt & Co. (1976) 17. Olsen, T., Procci, K., Bowers, C.: Serious games usability testing: how to ensure proper usability, playability, and effectiveness. In: International Conference of Design, User Experience, and Usability. Springer (2011) 18. Tolentino, G.P., et al.: Usability of serious games for health. In: Third International Conference on Games and Virtual Worlds for Serious Applications (VS-GAMES). IEEE (2011) 19. Pinelle, D., Wong, N., Stach, T.: Heuristic evaluation for games: usability principles for video game design. In: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. ACM (2008) 20. Federoff, M.A.: Heuristics and Usability Guidelines for the Creation and Evaluation of Fun in Video Games. Citeseer (2002) 21. Nielsen, J.: 10 usability heuristics for user interface design. Nielsen Norman Group 1(1) (1995) 22. Chang, C.-C., et al.: Is game-based learning better in flow experience and various types of cognitive load than non-game-based learning? Perspective from multimedia and media richness. Comput. Hum. Behav. 71, 218–227 (2017) 23. Csikszentmihalyi, M.: Flow: The Psychology of Optimal Performance. Harper and Row, New York (1990) 24. SAE: SAE levels of driving automation (2016). https://www.sae.org/standards/content/ j3016_201609/ 25. Abraham, H., Reimer, B., Mehler, B.: Learning to use in-vehicle technologies: consumer preferences and effects on understanding. In: Proceedings of the Human Factors and Ergonomics Society Annual Meeting. SAGE Publications, Los Angeles (2018) 26. Alicandri, E., Roberts, K., Walker, J.: A validation study of the DOT/FHWA highway simulator (HYSIM). Federal Highway Administration (1986)

Comparison of Child and Adult Pedestrian Perspectives of External Features on Autonomous Vehicles Using Virtual Reality Experiment Shuchisnigdha Deb1(&), Daniel W. Carruth1, Muztaba Fuad2, Laura M. Stanley3, and Darren Frey1 1

2

Mississippi State University, Starkville, MS, USA {deb,dwc2,daf190}@cavs.msstate.edu Department of Computer Science, Winston-Salem State University, Winston-Salem, NC, USA [email protected] 3 Department of Industrial Engineering, Clemson University, Clemson, SC, USA [email protected]

Abstract. In the United States, pedestrians aged  14 suffer the highest percentage of motor vehicle collisions leading to injuries and fatalities. In part to reduce motor-vehicle related crashes, transportation researchers are pursuing the implementation of automated vehicles. Vehicle automation will eventually remove human control from the vehicle, but this may also remove interpersonal communication between pedestrians and human drivers. Therefore, many studies have investigated pedestrians’ choice of features on autonomous vehicles (AVs) to facilitate communication of vehicle intention. The inclusion of child populations in these studies has been rare, however. This study investigated pedestrians’ understandability of the external features on autonomous vehicles, considering different vehicle sizes and physics (speed and distance), and included both children and adults. The results revealed that children relied entirely on the communicating features of AVs to make their judgement on their safety, thereby adopted a higher risk strategy than their adult counterparts. Keywords: Autonomous vehicles Child pedestrian
Communication


Pedestrian
Virtual reality

Features

1 Introduction Pedestrians, as a category of vulnerable road-users, have the greatest risk of being involved in a traffic collision which seriously injures or kills them. In order to reduce traffic collisions, autonomous vehicle (AV) technology is being developed to assist human operators. Vehicle automation includes different types of safety features that are intended to improve overall traffic safety. Previous research has revealed that pedestrians tend to over-rely on these safety features which causes them to take higher risks © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 145–156, 2020. https://doi.org/10.1007/978-3-030-20503-4_13

146

S. Deb et al.

in front of AVs as compared to manually driven vehicles [1]. Therefore, further research is necessary to address issues with the technology as well as to facilitate the transition to using and interacting with AVs. AVs may remove interpersonal communication among road-users. For pedestrians, this will create difficulty in making a crossing decision in front of these vehicles. Many past studies have reported that pedestrians depend on cues (eye contact, head and hand gestures, etc.) from human drivers to predict safe crossing conditions [2–4]. In the context of interacting with AVs at a crosswalk, lack of communication between pedestrians and AVs can result in confusing traffic situations and increased traffic collisions [5, 6]. Extensive research is underway using survey research, field studies, and lab experiments to recommend potential communicating features on AVs to facilitate AV-pedestrian interaction [7–20]. Some recent designs include text messages [7–10], LED lights in different patterns [11–16], LED displays on the vehicle [8, 9, 13, 14, 17], projections on the road [9, 13, 17–19], audible sounds and messages [8, 13, 15], and physical cues [13, 20]. Based on these previous research, the most common recommendations which were found to be both feasible to implement and understandable without training include: text displays, such as ‘walk’, ‘do not cross’; symbol displays, such as ‘pedestrian silhouette’, ‘upraised hand’, ‘stop sign’; audible messages, saying ‘safe to cross’; and physical cues, such as eyes on the vehicle, actuated hand, etc. While much research have considered AV-to-pedestrian interaction and pedestrians’ choice of external features on these vehicles, only one study included child participants. This study was conducted using a focus group and a survey with pictures and drawings of the proposed interfaces [17]. In the United States, pedestrians aged  14 suffer the highest percentage of pedestrian-motor vehicle collisions that lead to injuries (7%) and fatalities (22%), as compared to the group aged 18–24 who suffer the lowest percentage of injuries (2%) and fatalities (10%) [21]. Child pedestrians are susceptible to high traffic risks due to their physical vulnerability (hard to see due to height), their lack of knowledge about traffic rules, and their limited cognitive and perceptual capability. Therefore, developers and researchers should consider children in the design process for new technologies [22]. The current research used a virtual reality (VR) experiment to compare child pedestrians, aged 7–14, (highest injury and fatality rates) with young adult pedestrians, aged 18–24, (lowest injury and fatality rates) to investigate their response to and preference for external feature designs on AVs. These features would assist pedestrians comprehend the vehicle’s intended action at crosswalks. The inclusion of children in the study allowed a focus on the understandability for this at-risk age group of the features which had previously been evaluated by studies performed with an adult population only. In addition, the comparison between the two age groups shed light on the design features that could confuse child populations.

Comparison of Child and Adult Pedestrian Perspectives

147

2 Method 2.1

Participants

The Institutional Review Board at Mississippi State University approved this research study. Twelve children and ten adults were recruited for participation; each of them received $10 for their time. The adult participants (aged 18–24) signed informed consent forms. For the child participants (aged 7–14), at least one parent or legal guardian signed a parental permission form and the child signed an assent form. The children were aged between 7 and 14 with a mean age of 10.41 (SD = 2.78) years; eight of them were females, four were males. The adults were aged between 18 and 24 with a mean age of 21.2 (SD = 2.35) years; five of them were females, five were males. All participants were fluent English speakers and were able to read English clearly. They reported normal or corrected-to-normal visual acuity and no hearing difficulties. Each of the adult participants had a valid driver license. 2.2

Experimental Design

A virtual reality (VR) experiment was designed to expose pedestrians to the traffic environments. The virtual traffic environment used in this study was developed with Unity and the participants wore an HTC Vive VR headset to be immersed into that environment. This experiment considered two types of AVs, four levels of vehicle physics defined based on combinations of vehicle speeds and gaps between the lead (the vehicle in front of the trial vehicle) and trial vehicles, and four levels of communicating features on the AVs. Most of the pedestrian fatalities involve cars and light trucks (SUV, pickup, van) [21, 23]. Therefore, the vehicles used in this study were an artist’s model of the Waymo Firefly autonomous vehicle and an artist’s model of the Local Motors Olli self-driving shuttle. The levels of vehicle physics involved four different speed and distance combinations: 30 mph–155 ft, 30 mph–26 5ft, 45 mph– 230 ft, and 45 mph–400 ft. The two distances at the different speed limits were selected to present one risky situation (not enough time for the vehicle to stop without hard braking) and safe situation (enough time for the vehicle to stop without hard braking). The external communicating features were selected based on previous study outcomes published by this research group [7, 8]: “walk” in text, three pedestrian silhouette icons, a voice message saying “safe to cross”, and three stop sign icons. The three visual displays blinked one and off at 500 ms intervals. Trials were presented in two blocks, one for each of the two vehicles. The blocks were randomly assigned so that half of the participants from each age group experienced the car in the first block and the other half saw the shuttle in the first block, and then they were switched in the second block. In each block, the participants experienced 16 trials (4  4) which were randomly assigned to them with different vehicle physics and feature types. Participants had a short break of five minutes after the first 16 trials. Figure 1 presents the shuttle and the car used in this study along with three visual features on them.

148

S. Deb et al.

Fig. 1. Three visual features are presented on shuttle and car.

2.3

Measures Collected

Several measures were collected to compare pedestrians’ understanding of the message conveyed by each of the proposed features and their ease in making a crossing decision after acknowledging the message. In order to confirm safe exposure to the virtual environments, the participants responded to the Simulation Sickness Questionnaire (SSQ) [24] multiple times during the experiment: at the beginning of the study, after their first exposure to the virtual reality, during the break between two blocks, and at the end of the study. For each trial condition, pedestrians’ assessment for the understandability of the meaning of a feature and of the difficulty of crossing safely were surveyed within the VR after the completion of each trial. The survey items were: (1) it was easy for me to understand the feature and (2) it was difficult for me to cross the road. The responses were collected on a 5-point Likert scale from ‘strongly disagree’ to ‘strongly agree’ where ‘strongly disagree’ was scored as 1 and ‘strongly agree’ was scored as 5. The VR system collected walking behavior data and recorded their head position and orientation at approximately 60 Hz. A mounted webcam recorded each participant’s movement in the lab area, while the participant’s view of the virtual environment was recorded from a monitor. At the end of the experiment, along with the SSQ, the participants completed a demographic survey and rated the features again, outside the VR environment.

3 Results 3.1

Pedestrian General Behavior

The analysis of the video recording showed that most of the child pedestrians (85%), as compared to the adults (47%), did not look in both directions once they had seen the signal from the autonomous vehicle. They made their decisions for crossing based solely on the communicating feature on the autonomous vehicle. 3.2

Understanding of Message Conveyed by a Feature

Participants’ assessment of their ability to understand the message conveyed by a feature was collected on a 5-point Likert scale from ‘strongly disagree (1)’ to ‘strongly

Comparison of Child and Adult Pedestrian Perspectives

149

agree (5)’, for all trial conditions based on vehicle type, feature type, and vehicle physics. A higher number indicated an agreement that it was easy to understand the message conveyed by a feature. Analysis of variance (ANOVA) was performed to find the differences in responses for the age groups and the different factors, as well as the interaction effects between the age groups and the different factors. The results show that, at a 95% significance level, both vehicle type and feature type showed significant main effects on the understanding of communicating features. Table 1 shows results from ANOVA and Figs. 2 and 3 display descriptive statistics (mean) of pedestrians’ ratings for different types of vehicles and features, respectively. Table 1. ANOVA results for feature rating based on the clarity of message.

Average Rating for Message Comprehension

Factor Age group Vehicle type Age group  Vehicle type Feature type Age group  Feature type Vehicle physics Age group  Vehicle physics

4.5 4.4 4.3 4.2 4.1 4 3.9 3.8

F [df] 2.809 [1, 4.900 [1, 0.344 [1, 4.534 [3, 1.482 [3, 1.246 [3, 0.587 [3,

700] 700] 700] 696] 696] 696] 696]

p value 0.094 0.027 0.558 0.004 0.218 0.292 0.624

Eta Squared 0.004 0.007 0.000 0.018 0.006 0.005 0.004

4.4

4.3 4.13

Car

Shuttle

Overall

4.22

4.18

4.1

Car

Shuttle

Children (aged 7-14)

Car

Shuttle

Adults (aged 18-24)

Vehicle Type

Fig. 2. Effect of vehicle type on ease of understanding vehicle’s intended action from features.

Although the display on both vehicles had the same size and position, participants’ rating of their ease in understanding the messages conveyed by the features was significantly influenced by the vehicle size. For the larger vehicle (shuttle), they were less likely to agree that the message was easy to understand as compared to the smaller vehicle (car). This difference was consistent for both age groups. The feature type exhibited significantly different ratings for pedestrians’ ease of understanding the intent of the features. Although all the features were rated with high scores on average, the most preferred features were ‘walk’ in text and voice message saying ‘safe to cross.’ The participants did not favor the stop sign, which was a negative feature as compared to the other three features. Children always expected the

150

S. Deb et al.

4.26

4.35 4.11

4.02

Overall

Children (aged 7-14)

Voice Message

Pedestrian Silhouette

Walk in Text

Stop Sign

Voice Message

Pedestrian Silhouette

3.87

Walk in Text

Voice Message

4.02

4.49

4.36 4.16

Stop Sign

4.36 4.14

Stop Sign

4.36

Pedestrian Silhouette

4.6 4.4 4.2 4 3.8 3.6 3.4

Walk in Text

Average Rating for Message Comprehension

vehicles to stop for them and were impatient when presented with the stop sign on the vehicle. They did not understand whether the stop message was for them or for the vehicle. Although a pedestrian silhouette is a common and familiar signal at pedestrian crossings, participants preferred the features that contained an unambiguous message about the safe crossing conditions.

Adults (aged 18-24)

Feature Type

Fig. 3. Effect of feature type on ease of understanding vehicle’s intended action from features.

3.3

Crossing Decision Making

Participants’ self-reported difficulty in making a crossing decision in front of the AVs was collected on a 5-point Likert scale from ‘strongly disagree (1)’ to ‘strongly agree (5)’. Analysis of variance (ANOVA, see Table 2) for the two age groups and three factors (vehicle type, feature type, and vehicle physics), as well as interaction effects between the age groups and the three factors, show a significant influence for age group and feature type. Table 2. ANOVA results for feature rating based on the difficulty in crossing. Factor Age group Vehicle type Age group  Vehicle type Feature type Age group  Feature type Vehicle physics Age group  Vehicle physics

F [df] 5.083 [1, 1.634 [1, 1.189 [1, 5.322 [3, 0.529 [3, 1.016 [3, 0.545 [3,

700] 700] 700] 696] 696] 696] 696]

p value 0.024 0.202 0.276 0.001 0.662 0.385 0.652

Eta Squared 0.007 0.002 0.002 0.021 0.002 0.004 0.002

151

It was difficult for you to cross the road 62%

Children (aged 7-14)

2%

0% Strongly Agree

4%

Agree

4%

Neutral

10%

Disagree

9%

Strongly Disagree

Neutral

30%

Strongly Agree

39%

Agree

45%

Disagree

80% 60% 40% 20% 0%

Strongly Disagree

Percentages of Responses

Comparison of Child and Adult Pedestrian Perspectives

Adults (aged 18-24)

Age Group

Fig. 4. Effect of age groups on making crossing decisions.

2

1.54 1.68 1.67

1.91

1.61 1.69 1.75

2.01 1.44

1.5

1.68 1.56 1.78

1 0.5

Overall

Children (aged 7-14)

Stop Sign

Voice Message

Pedestrian Silhouette

Walk in Text

Stop Sign

Voice Message

Pedestrian Silhouette

Walk in Text

Stop Sign

Voice Message

Pedestrian Silhouette

0 Walk in Text

Average Rating for Crossing Difficulty

Figure 4 shows the proportions of the responses for the different factors for the two age groups. Children were found to be less comfortable than adults with crossing roads in front of an autonomous vehicle in general, even when they are equipped with communicating features. Feature type significantly influenced pedestrians’ ease in making a crossing decision. A higher score indicates greater difficulty in making a crossing decision (see Fig. 5). Overall, people found all the signals satisfactory (with scores below the neutral point of 3) for making a crossing decision. Among all the features, the stop sign was the most confusing one and made the crossing decision more difficult. Based on their concept of the safety features on autonomous vehicles, pedestrians expected the vehicles to stop for them at the crosswalk, not to stop them from crossing the street.

Adults (aged 18-24)

Feature Type

Fig. 5. Effect of feature type on making crossing decisions.

152

S. Deb et al.

‘Walk’ in text was the easiest feature for them to understand, communicating safe crossing conditions and allowing them to make an easier crossing decision. The voice message and pedestrian silhouette were also better than the stop sign, but not as clear as the ‘Walk’ signal. 3.4

Street Crossing Time

Along with the self-reported subjective ratings, pedestrians’ crossing behavior data was collected from the headset tracking. The position data were analyzed to calculate crossing time after the participants initiated crossing for each trial. Feature type and vehicle physics significantly influenced crossing time. In addition, age group and vehicle physics showed significant interaction effect on crossing time. The ANOVA results for crossing time is displayed in Table 3. Table 3. ANOVA results for pedestrians’ street crossing time

5.62

6.26

5.77

5.45

5.55

6.24

Pedestrian Silhouette

Voice Message

Stop Sign

Stop Sign

5.45

Walk in Text

Voice Message

Overall

5.21

Stop Sign

6.25

Eta Squared 0.000 0.000 0.001 0.013 0.002 0.006 0.010

Voice Message

5.59

p value 0.554 0.814 0.508 0.025 0.697 0.210 0.044

700] 700] 700] 696] 696] 696] 696]

Pedestrian Silhouette

5.47

F [df] 0.350 [1, 0.055 [1, 0.439 [1, 3.126 [3, 0.479 [3, 1.513 [3, 3.545 [3,

Walk in Text

5.45

Pedestrian Silhouette

7 6 5 4 3 2 1 0

Walk in Text

Average Crossing Time

Factor Age group Vehicle type Age group  Vehicle type Feature type Age group  Feature type Vehicle physics Age group  Vehicle physics

Children (aged 7-14)

Adults (aged 18-24)

Feature Type

Fig. 6. Effect of feature type on crossing time

Comparison of Child and Adult Pedestrian Perspectives

153

Figure 6 shows the effect of feature type on pedestrians’ street crossing time. Participants took significantly more time with the stop sign. As the pedestrians were expecting the vehicle to stop for them, they had started walking. Afterwards, they realized that the vehicle was not going to stop and they had to stop to let the vehicle pass. This may have increased their crossing time for this sign. This also indicates that pedestrians did not understand a vehicle’s intended action at crosswalk from this stop sign.

Fig. 7. Interaction effect of age group and vehicle physics on crossing time

The vehicle physics, its speed and distance combination, interacted significantly with the age groups to influence crossing time. The risk associated with child pedestrians’ crossing behavior as compared to the adults can be observed from this analysis (see Fig. 7). At the higher speed and narrower gap (45 mph–230 ft), the children started crossing the road and completed crossing by running to avoid a collision (in the case of the stop sign feature). During the trials with larger gaps (30 mph–265 ft and 45 mph–400 ft), the children took a longer time to make sure the vehicle slowed down to a stop at the intersection. In most of the cases, the children managed to cross the road properly for the vehicle with a slower speed and a narrow gap (30 mph–155 ft). The adults on the other hand, showed reasonable behavior with an appropriate crossing time for the vehicle with a higher speed and a narrow gap.

4 Discussion This study was designed to identify risk factors associated with child pedestrians’ interaction with autonomous vehicles (AV). The investigation was also made to compare communicating features on AVs for children and adults and to report the

154

S. Deb et al.

findings to transportation researchers and vehicle manufacturers for the on-going feature design process. The results showed that participants were less certain of their understanding of the intent of the features on a (larger) autonomous shuttle than on an autonomous car. For the larger vehicle, they had issues understanding the message of the features. The large size of the vehicle may have required more effort for them to understand the vehicle’s intended action. Consistent with the previous study conducted by this research group [8], participants rated ‘Walk’ in text and the voice message to be the feature which made the vehicle’s intended action the easiest to understand. The analysis also revealed that child pedestrians showed risky behaviors in front of all the automated vehicles. They depended on the message of the feature only and did not rely on other environmental factors related to traffic behavior, such as looking in both directions or determining the vehicle speed or the gap between the two vehicles (lead and trial). This behavior can be considered very unsafe for real-world traffic interaction. This identifies a risk factor for child pedestrians while interacting with an autonomous vehicle with communicating features. In addition, children wanted these vehicles to always stop for them at the crosswalk; they did not like the stop sign on the vehicle indicating a prohibited crossing condition. Most of the children were surprised that a vehicle showing a stop sign did not recognize them and did not stop for them. Children are taught to cross at crosswalks and near schools where aids stop oncoming traffic with a stop sign that they carry into the street. For children, a stop sign near a crosswalk means that traffic will stop for them, not that they should stop for traffic. For children, making a crossing decision was rated more difficult compared to the adults. They found the ‘Walk’ in text to be the best option for conveying the message of a safe crossing condition, thus making the crossing decision easier. Based on their subjective ratings, they understood the verbal message clearly, even better than the text; however, it was not as easy as with the text sign for them to make a crossing decision with the voice message. Most of the children were confused about the source of the voice message at the beginning and found it difficult to make the judgement on a crossing. On the other hand, the adults’ responses to understanding the message and making a crossing decision were consistent; if they understood the message, it was easy to make the decision to cross. Adults preferred ‘Walk’ in text as easiest feature to understand vehicle’s intended action and found it to be the message that made the decision to cross easiest. Contrary to the subjective measures, the crossing-time related analysis showed that participants acted similarly for all the positive features. They understood the text, image, and voice message clearly enough to act consistently in front of the vehicles equipped with those features. With the stop sign, the participants were confused by the message and took more time to finish crossing. Video data showed that pedestrians started crossing, thinking the car would stop for them and then had to stop in the middle to let the car pass by. The factor “vehicle physics” is a very useful criterion for measuring how pedestrians interact with a traditional vehicle. Most of the pedestrians depend on vehicles’ legacy behaviors: speed and distance [4, 9, 10]. However, previous studies analyzing the thresholds for distance-gap judgements in street crossing have shown that children are not able to make safe decisions for vehicles approaching at high speeds (45 mph– 230 ft and 45 mph–400 ft) [25, 26]. The current study also found that children make hasty and unsafe decisions to cross the street in front of an AV with high speeds and

Comparison of Child and Adult Pedestrian Perspectives

155

narrow gaps. Their expectation that vehicles will stop at the crosswalk was not addressed by using a stop sign on the display to communicate restricted crossing. Children need to be familiarized and trained with general pedestrian behavior as well as with communicating with autonomous vehicle technology.

5 Conclusion This study exposed an additional concern in designing communicating features on autonomous vehicles. Compared to adults, child pedestrian behavior was hasty and risky in front of an approaching autonomous vehicle. Children showed overreliance on the features and made crossing decisions based only on the visual or audible features presented by the vehicle, even if they misinterpreted the display. They expected the autonomous vehicle to always stop for them at the crosswalk. These behaviors highlight the necessity of proper promotion and training to prepare children to interact with vehicle automation technology on the road. In order to implement these vehicles in traffic properly, it is essential to familiarize road users with their advantages and limitations. Removing drivers from vehicle control will not itself reduce traffic accidents. Education in how to use them and how to interact with them, recognizing their limitations, will be necessary to ensure successful integration. The best way to do this is to target children for training and education, as they will be the users of this future technology.

References 1. Deb, S., Strawderman, L., Carruth, D.W., DuBien, J., Smith, B., Garrison, T.M.: Development and validation of a questionnaire to assess pedestrian receptivity toward fully autonomous vehicles. Transp. Res. Part C Emerg. Technol. 84, 178–195 (2017) 2. Guéguen, N., Meineri, S., Eyssartier, C.: A pedestrian’s stare and drivers’ stopping behavior: a field experiment at the pedestrian crossing. Saf. Sci. 75, 87–89 (2015) 3. Ren, Z., Jiang, X., Wang, W.: Analysis of the influence of pedestrians’ eye contact on drivers’ comfort boundary during the crossing conflict. Procedia Eng. 137, 399–406 (2016) 4. Šucha, M.: Road users’ strategies and communication: driver-pedestrian interaction. In: Transport Research Arena (TRA) 2014 Proceedings (2014) 5. Anthony, S.: The trollable self-driving car. https://slate.com/technology/2016/03/googleself-driving-cars-lack-a-humans-intuition-for-what-other-drivers-will-do.html 6. Richtel, M., Dougherty, C.: Google’s driverless cars run into problem: cars with drivers. https://www.nytimes.com/2015/09/02/technology/personaltech/google-says-its-not-thedriverless-cars-fault-its-other-drivers.html 7. Deb, S., Warner, B., Poudel, S., Bhandari, S.: Identification of external design preferences in autonomous vehicles. In: Proceedings of the 2016 Industrial and Systems Engineering Research Conference, pp. 44–69 (2016) 8. Deb, S., Strawderman, L.J., Carruth, D.W.: Investigating pedestrian suggestions for external features on fully autonomous vehicles: a virtual reality experiment. Transp. Res. Part F Traffic Psychol. Behav. 59, 135–149 (2018) 9. Clamann, M., Aubert, M., Cummings, M.L.: Evaluation of vehicle-to-pedestrian communication displays for autonomous vehicles (2017)

156

S. Deb et al.

10. Matthews, M., Chowdhary, G., Kieson, E.: Intent communication between autonomous vehicles and pedestrians (2017) 11. Lagstrom, T., Lundgren, M.V.: AVIP-autonomous vehicles interaction with pedestrians (2015). http://www.tekniskdesign.se/download/AVIP_MasterThesis_Lagstrom_MalmstenLundgren.pdf 12. Zhang, S., Klein, D.A., Bauckhage, C., Cremers, A.B.: Fast moving pedestrian detection based on motion segmentation and new motion features. Multimed. Tools Appl. 75, 6263– 6282 (2016) 13. Mahadevan, K., Somanath, S., Sharlin, E.: Communicating awareness and intent in autonomous vehicle-pedestrian interaction. In: Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems - CHI 2018, pp. 1–12 (2018) 14. Fridman, L., Mehler, B., Xia, L., Yang, Y., Facusse, L.Y., Reimer, B.: To walk or not to walk: crowdsourced assessment of external vehicle-to-pedestrian displays. arXiv.org (2017) 15. Florentine, E., Ang, M.A., Pendleton, S.D., Andersen, H., Ang, M.H.: Pedestrian notification methods in autonomous vehicles for multi-class mobility-on-demand service. In: Proceedings of the Fourth International Conference on Human Agent Interaction - HAI 2016, pp. 387–392 (2016) 16. Siripanich, S.: Crossing the road in the world of autonomous cars. https://medium.com/ teague-labs/crossing-the-road-in-the-world-of-autonomous-cars-e14827bfa301 17. Charisi, V., Habibovic, A., Andersson, J., Li, J., Evers, V.: Children’s views on identification and intention communication of self-driving vehicles. In: Proceedings of the 2017 Conference on Interaction Design and Children - IDC 2017, pp. 399–404 (2017) 18. Habibovic, A., Lundgren, V.M., Andersson, J., Klingegård, M., Lagström, T., Sirkka, A., Fagerlönn, J., Edgren, C., Fredriksson, R., Krupenia, S., Saluäär, D., Larsson, P.: Communicating intent of automated vehicles to pedestrians. Front. Psychol. 9, 1336 (2018) 19. Mitsubishi Electric Corporation: Mitsubishi Electric Introduces Road-illuminating Directional Indicators. http://www.mitsubishielectric.com/news/2015/1023.html 20. Chang, C.-M., Toda, K., Sakamoto, D., Igarashi, T.: Eyes on a car: an interface design for communication between an autonomous car and a pedestrian. In: Proceedings of the 9th International Conference on Automotive User Interfaces and Interactive Vehicular Applications - AutomotiveUI 2017, pp. 65–73 (2017) 21. National Center for Statistics and Analysis: Pedestrians: 2016 data (Traffic Safety Facts. Report No. DOT HS 812 493), National Highway Traffic Safety Administration, Washington, D.C. (2018) 22. Druin, A.: The role of children in the design of new technology. Behav. Inf. Technol. 21, 1– 25 (2002) 23. National Center for Statistics and Analysis: Pedestrians: 2015 data (Traffic Safety Facts. Report No. DOT HS 812 375), National Highway Traffic Safety Administration, Washington, D.C. (2017) 24. Kennedy, R.S., Lane, N.E., Berbaum, K.S., Lilienthal, M.G.: Simulator sickness questionnaire: an enhanced method for quantifying simulator sickness. Int. J. Aviat. Psychol. 3, 203– 220 (1993) 25. Congiu, M., Whelan, M., Oxley, J., D’Elia, A., Charlton, J.: Crossing roads safely: an experimental study of age and gender differences in gap selection by child pedestrians. In: Proceedings of the Australasian Road Safety Research, Policing and Education Conference (2006) 26. Connelly, M.L., Conaglen, H.M., Parsonson, B.S., Isler, R.B.: Child pedestrians’ crossing gap thresholds. Accid. Anal. Prev. 30, 443–453 (1998)

An Inclusive, Fully Autonomous Vehicle Simulator for the Introduction of Human-Robot Interaction Technologies Theocharis Amanatidis(&), Patrick Langdon, and P. John Clarkson Engineering Design Centre, Engineering Department, University of Cambridge, Trumpington Street, Cambridge CB2 1PZ, UK {ta323,pml24,pjc10}@cam.ac.uk

Abstract. As fully autonomous - SAE level 5 - vehicles approach commercialisation, there is need to design and test user interfaces specifically for such use. However, testing in real environments is currently limited or, in many cases, impossible. In this paper, as a solution, we present a dedicated simulator for fully autonomous vehicles. First, we outline the design requirements for this simulator. Inclusive design principles were used to accommodate a large range of diversity in our population. It can thus be used by individuals with visual, auditory and certain physical and cognitive impairments. Second, we describe the capabilities of the simulator in terms of human-robot interaction technologies. We aim to assess both performance and non-performance characteristics of the resulting systems and how they can be integrated in the transportation experience. Third, we collate the knowledge obtained during this project to provide a deeper understanding of human factors in operating fully autonomous vehicles. We hope these results provide a basis for further research and improve the experience of users. Keywords: Autonomous vehicles
Human-robot interaction
Automotive simulators
Inclusive design
Human factors and ergonomics

1 Introduction Fully autonomous vehicles – also known as SAE level 5 vehicles [1] - are currently being developed by a number of automotive manufacturers and technology companies, such as Tesla [2] or Waymo [3], and are steadily getting closer to commercialisation. However, while the computer vision and control algorithms necessary for driving are being developed at a rapid pace, currently only one of these companies, Cruise Automation, have made public details about the user interface of their SAE Level 5 vehicles [4]. If the promised start dates for on-road trials and commercial introduction are to be met, there is an increasingly urgent need to design and test appropriate user interfaces specifically for SAE Level 5 use, both for commercial and regulatory purposes. The user interface of SAE Level 5 vehicles can be notionally divided in 3 User Interface (UI) elements: 1. Smart Device UI, 2. Interior UI, and 3. Exterior UI. The first element, Smart Device UI, connects the user with the autonomous vehicle network © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 157–165, 2020. https://doi.org/10.1007/978-3-030-20503-4_14

158

T. Amanatidis et al.

remotely, for instance when outside the vehicle, perhaps through a smartphone or tablet app. It consists of finding nearby vehicles, ordering, unlocking, identification, payment, monitoring, and other remote functionality. In other words, similar functionality to certain ride-sharing apps, such as Uber and Lyft. The second element, Interior UI, gives operational control to the user once inside the vehicle. It includes navigation, monitoring, lock/unlock, emergency control, communication, entertainment, productivity, and other interior functionality. The third element, Exterior UI, connects the fully autonomous vehicle with other road users including other vehicles, pedestrians, cyclists, road infrastructure, etc. It includes recognition, accelerating, stopping, turning, emergency warning, and communication functionality. This paper focuses exclusively on the second element: Interior UI. A well designed UI, has the potential to transform the user experience of autonomous vehicles by increasing trust, reducing errors, and making the journey more pleasant, among others. Furthermore, a potential introduction of human-robot interaction technologies in these vehicles presents a number of opportunities [5]. These technologies need to be evaluated and compared. As the fatal Uber accident [6] has shown, potential dangers lie when increased complexity or uneven development between multiple of its UI elements leads to failure. In this case there was an identification and communication failure between the vehicle, the operator and the pedestrian crossing the road. Legislation is a driver too for the development of SAE Level 5 user interfaces. For example, there is potential to harmonise the exterior UI between the different brands of vehicles to facilitate recognition of their movements by pedestrians, cyclists and other users (External UI). Furthermore, certain governments might want to promote autonomous public transport or shared private systems in order to reduce congestion and emissions/energy use [5]. In this later case, the Internal UI will need to be an enabler of the shared experience between multiple users inside a vehicle. As a result, due to the current state of legislation over most of the global road network, and the associated lack of prototype vehicles, testing in real environments is currently limited or, in many cases, impossible. As a solution, we propose in this paper the development of a dedicated automotive simulator for fully autonomous, SAE Level 5 vehicles, to circumvent the issues described above.

2 Previous Work A number of studies have developed and tested lower autonomy level simulators such as [7, 8] from which some useful information could also be applicable to SAE level 5 vehicles such as secondary task activity. Furthermore, some studies have investigated user reaction to being driven in a real vehicle whilst hiding the driver, to emulate being in a fully autonomous vehicle [9, 10]. These could also be helpful, in this case to investigate elements of the interaction that are not related to the interface itself, such as trust. Oliveira et al. did investigate how interfaces influence user interaction with autonomous vehicles, and found users prefer on-board screens to using their own devices for navigation [11]. However, there is currently very little work in integrating Human-Robot Interaction technologies in vehicles. Braun et al. [12] are the exception, investigating

An Inclusive, Fully Autonomous Vehicle Simulator

159

emotion elicitation methods while driving, and some of their work could be applied to driverless vehicles. Overall though, more work remains to be done in this field.

3 Design Requirements and Simulator Set-up In this section we outline the design requirements for our SAE Level 5 vehicle simulator, including their origin, general requirements, inclusivity aspects, and the current set-up, as built. As this was a first take at developing such a simulator and given that interfaces for fully autonomous vehicles are currently at a relatively early development stage, this was designed as a low-fidelity simulator. Furthermore, some features were excluded due to time, budget and other restrictions. 3.1

Origin of Design Requirements

The Design Requirements, as outlined in this paper, originate from a number of sources. The first and most pertinent two are a couple of formative [13] studies on the needs and expectations of fully autonomous vehicle user interfaces. The first was a qualitative interview study which investigated the main areas of interest and concern regarding such interfaces [13]. Participants formed a carefully selected sample of road users to meet all five categories of traveler-journeys from the UK Traveler Needs Survey [14]. The second was a quantitative survey study, which aimed to be more representative of the travelling population and to quantify the results obtained in the first study. While the results of the second study are to be published at a further date, Sects. 3.2 and 3.3 below outline their effects on the design requirements of the simulator. Finally, other sources of requirements include user familiarity with current vehicles and homogenisation with Cruise Automation and Waymo prototypes. 3.2

General Requirements

A considerable number of requirements for the interface involve the main control screen of the vehicle. While a large screen was not specifically requested by the study participants, it was necessary to cover all the functions the users want to be able to complete: the two main drivers were entertainment and inclusivity. The transition of the main screen from mainly control to entertainment drives the size up. This is to increase immersion while maintaining enough screen space to keep critical controls on-screen at all times. This transition also dictates the orientation and aspect ratio: landscape and at least 16:9 wide for video content. On-screen requirements originating from the user studies were extensive. This requirement group is also driven by familiarity to current automotive interfaces. Information is to be provided to users in a pertinent and context-specific way, while maintaining critical elements permanently on-screen. Information to be provided includes journey information (such as location, time of arrival, expected arrival time, etc.), information about the state of the vehicle if it was owned by themselves (such as battery/fuel levels, maintenance, etc.), as well as productivity and entertainment

160

T. Amanatidis et al.

features, and climate and ride controls. Figure 1 shows a diagrammatic representation presenting some of the features described above.

Fig. 1. Diagrammatic representation of the information and controls present on the baseline touchscreen interface as described in Sect. 3.4.

Other requirements on the visual modality included achieving good environmental validity, collecting video data over both the users and the rest of the experiment, and providing a live feed of the simulator’s interior to the experimenter. Audio requirements in comparison, are best described as equal to current automotive applications. Requirements for the interior design of the simulator were mostly driven by the two user studies and is where the difference with existing SAE 0-4 level simulators is most visible. The simulator is not to have any traditional controls such as a steering wheel or pedals. This is to test human response when not in control. In this application, the cabin is to be designed for two experimental participants; seats with armrests and the ability to recline satisfy the demand to nap or sleep. Furthermore, the seats should be able to rotate inwards to facilitate discussion between the two users and give experimenters the possibility to test such interaction between users. Optionally, given adequate space and layout of the simulator, a table, charging points and Wi-Fi could be provided. Finally, the survey study indicated that the interface technology most preferred by users is physical buttons. As such, the simulator should have a number of programmable buttons to test different features (Fig. 2).

An Inclusive, Fully Autonomous Vehicle Simulator

161

Fig. 2. 5-point Likert scales of interaction technologies, from left to right: touchscreen, voice control, physical buttons, gestures.

3.3

Inclusivity Requirements

This simulator was developed following the principles of inclusive design in order to accommodate a large range of diversity in our population [15–18]. It can thus be used by individuals with visual, auditory and certain physical and cognitive impairments. In terms of inclusivity of the screen, it needs to be large enough to host oversized, easy to read fonts and icons, and provide adequate contrast. Audio cues for visually impaired users can be provided through the audio system via Wizard of Oz. Inclusivity also drove one of the most important design requirements, to place the screen on an easily moveable arm. This affords users to change the distance and angle they receive information and caters for some visually impaired users’ requirements. However, it provides additional benefits to all users such as the freedom to position the screen at their optimal height, to pass the screen between users or to ease entry and exit for example. 3.4

Simulator Set-up

Given the above requirements, the simulator design was the following. Even though it was designed as a low fidelity simulator, in order to provide good visual environmental validity, it was decided to film immersive 360-degree videos of a real drives. The 360-degree video is the split on to 3 feeds, projected in front, left and right side of the simulator emulating the windshield and side windows. Sound consists of 2.0 stereo audio which could be upgraded to a 5.1 system if necessary. A 24 in., 16:9, full HD touch monitor was selected as the main screen. While substantially larger than what found in most automotive applications it meets the requirements for increased immersion in entertainment tasks and provides increased screen space for inclusivity features. The screen is mounted on an easily moveable arm extending 51 cm in height, 84 cm in depth and 157 cm side to side. The cabin was designed around 2 partially rotating and reclining seats with armrests, facing forwards and with easy reach to the touchscreen. No steering wheel or pedals are provided. Two cameras, one facing towards the participants and one towards the screen provide a live feed to the experimenter and record the experiment.

162

T. Amanatidis et al.

An array of 6 fully configurable LCD buttons was implemented in a 2 row, 3 column arrangement. The screen on the buttons not only can be reconfigured to test different button configurations but can also be context-dependent if necessary. Finally, the simulator is not fully wheelchair accessible. However, wheelchair users can be accommodated if they are able to either shift themselves to one of the forward facing seats or to perform the experiments while facing to the right hand side of the simulated vehicle, as there is limited room for maneuvering within the simulator area (Fig. 3). A summary of the design choices is outlined on Table 1 below. Table 1. Summary of current simulator configuration. Feature Touchscreen dimensions Arm movement range Physical buttons Audio Environment projection Video recording Cabin Inclusivity

Description 24 in., 1920  1080 pixels, 16:9 aspect ratio 51 cm in height, 84 cm in depth, 157 cm side-to-side 6 fully configurable screen buttons 2.0 stereo sound system 360-degree video, split and projected front, left and right 2 high-definition cameras; 1 facing forwards to the screen and 1 backwards to the users 2 partially rotating and reclining seats, facing forward, with armrests and easy reach to touchscreen Inclusive font and icon size, wheelchair access

Fig. 3. Photographs of Stage 1 of the experimental set-up. From left to right, (a) shows the experimenter’s control area. (b) shows the touchscreen attached to a moveable arm below the front view out of the vehicle, in this case showing a dummy 360-degree drive through New York city. (c) shows a side view of the simulator where the twin seats and moveable arm can be seen in its retracted configuration.

An Inclusive, Fully Autonomous Vehicle Simulator

163

4 Simulator HRI Capabilities and Measurement In this section we describe the capabilities of the simulator in terms of novel humanrobot interaction technologies. The goal of our experimental set-up is to assess both performance and non-performance characteristics of the resulting systems and how they can be integrated in the transportation experience. 4.1

Human-Robot Interaction Capabilities

The simulator was designed to integrate Human-Robot Interaction (HRI) technologies in multiple stages. In the first stage, the baseline consists of the touchscreen interface. The baseline interface is important as it is the only interface consisting exclusively of commercially available technologies, and especially of technologies already being used in the automotive industry. However, in SAE level 5 vehicles, the screen plays an outsize role due to the shift away from driving. More non-driving content is available and accessible on it and users are less reliant on dials and buttons which provide necessary feedback while driving. Increasingly complex voice commands are becoming available in premium vehicle. As a result, added on top of baseline are voice commands through Wizard of Oz. However, the main benefit of voice commands – not getting hands off the wheel – is not as pertinent as in SAE 0-4 level vehicles. Similarly, both due to their widespread use in vehicles and due to popular demand physical buttons as described in Sect. 3.4 were added too. In this configuration however they are configurable and context dependent. As a result, they could provide a wider range of functionality. In future stages of the simulator there is capability to recognize gestures using a Kinect device; however, given that survey users as mentioned in Sect. 3.1 ranked it last in terms of preference it has not been a priority for development. Future stages of development could also include image processing with affective recognition (for example Microsoft Project Oxford) to investigate users emotive state while operating the simulator, a 3D on-screen rendering of a virtual pilot and an augmented reality display to shift information away from and declutter the main screen. 4.2

Performance and Non-performance Characteristics

The goal of our experimental set up as described in Sect. 3.4 above is to assess both performance and non-performance characteristics of the HRI technologies and how they can be integrated in the transportation experience. As a result, the simulator was set up so that both performance and other metrics can be measured. Performance metrics that can be measured include time to completion of the tasks given, screen movement, score in certain tasks, number of errors, and other behavioral data collected through recording a video of the experiment. However, non-performance data is equally important. This is due to the observation that a lot of the additional HRI technologies did not necessarily improve participant performance but increased satisfaction. Similar investigations are needed for other measures such as acceptance or trust. Finally, workload is measured too, through subjective measures, specifically NASA TLX.

164

T. Amanatidis et al.

5 Discussion The main purpose of this work is to establish an environment by which to evaluate different interaction technologies and support the development of fully autonomous vehicles. This paper has collated and analysed all the information accumulated over the design phase of this project in order to provide deeper understanding of human factors and human behaviours in operating fully autonomous vehicles. Given this work we believe that a Level 5 vehicle simulator is capable of delivering insights in the usage of autonomous vehicle interfaces. It can provide data during the absence of Level 5 prototypes and development vehicles, provide a means of rapid and safe testing and thus cut on the development time and risk. However, there is also potential for improving in the system after more extensive experimental use. We hope to present such results at the AHFE 2019 conference. Finally, it is worth mentioning that given the cost and rarity of autonomous vehicle prototypes, working on interfaces for such vehicles has thus far been mostly limited to proprietary research by automotive groups. The approach presented in this paper will hopefully allow more research, by a wider variety of researchers and a more open discussion on potential solutions to the fully autonomous vehicle interface questions.

6 Conclusions and Future Work In conclusion, this paper described the design and development of a SAE Level 5 Autonomous vehicle simulator. The design requirements were presented, including their origin and the design decisions that determined the final design were described. Finally, the purpose and potential of the simulator was discussed. This project is however still in its infancy, and further work is required to bring it to conclusion. For instance, experimental work needs to be done to evaluate areas of improvement and a way to split the experience for two independent users in a pooled use scenario needs to be developed. We hope these initial results will not only provide a basis for further research but also improve the experience of users in fully autonomous vehicles.

References 1. SAE International: SAE J3016: Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles (2016). http://standards.sae.org/j3016_ 201609/ 2. Dikmen, M., Burns, C.M.: Autonomous driving in the real world: experiences with tesla autopilot and summon. In: Proceedings of the 8th International Conference on Automotive User Interfaces and Interactive Vehicular Applications, pp. 225–228. ACM, New York (2016). https://doi.org/10.1145/3003715.3005465 3. Waymo. https://waymo.com/ 4. GM will make an autonomous car without steering wheel or pedals by 2019. https://www. theverge.com/2018/1/12/16880978/gm-autonomous-car-2019-detroit-auto-show-2018 5. Mitchell, W.J., Borroni-Bird, C.E., Burns, L.D.: Reinventing the Automobile: Personal Urban Mobility for the 21st Century. MIT Press, Cambridge (2010)

An Inclusive, Fully Autonomous Vehicle Simulator

165

6. Death of Elaine Herzberg (2019). https://en.wikipedia.org/w/index.php?title=Death_of_ Elaine_Herzberg&oldid=888636615 7. Politis, I., Langdon, P., Adebayo, D., Bradley, M., Clarkson, P.J., Skrypchuk, L., Mouzakitis, A., Eriksson, A., Brown, J.W.H., Revell, K., Stanton, N.: An evaluation of inclusive dialogue-based interfaces for the takeover of control in autonomous cars. In: 23rd International Conference on Intelligent User Interfaces, pp. 601–606. ACM, New York (2018). https://doi.org/10.1145/3172944.3172990 8. Wintersberger, P., Riener, A., Frison, A.-K.: Automated driving system, male, or female driver: who’d you prefer? Comparative analysis of passengers’ mental conditions, emotional states & qualitative feedback. In: Proceedings of the 8th International Conference on Automotive User Interfaces and Interactive Vehicular Applications, pp. 51–58. ACM, New York (2016). https://doi.org/10.1145/3003715.3005410 9. Wang, P., Sibi, S., Mok, B., Ju, W.: Marionette: enabling on-road wizard-of-oz autonomous driving studies. In: Proceedings of the 2017 ACM/IEEE International Conference on Human-Robot Interaction, pp. 234–243. ACM, New York (2017). https://doi.org/10.1145/ 2909824.3020256 10. Baltodano, S., Sibi, S., Martelaro, N., Gowda, N., Ju, W.: The RRADS platform: a real road autonomous driving simulator. In: Proceedings of the 7th International Conference on Automotive User Interfaces and Interactive Vehicular Applications, pp. 281–288. ACM, New York (2015). https://doi.org/10.1145/2799250.2799288 11. Oliveira, L., Luton, J., Iyer, S., Burns, C., Mouzakitis, A., Jennings, P., Birrell, S.: Evaluating how interfaces influence the user interaction with fully autonomous vehicles. In: Proceedings of the 10th International Conference on Automotive User Interfaces and Interactive Vehicular Applications, pp. 320–331. ACM, New York (2018). https://doi.org/ 10.1145/3239060.3239065 12. Braun, M., Weiser, S., Pfleging, B., Alt, F.: A comparison of emotion elicitation methods for affective driving studies. In: Adjunct Proceedings of the 10th International Conference on Automotive User Interfaces and Interactive Vehicular Applications, pp. 77–81. ACM, New York (2018). https://doi.org/10.1145/3239092.3265945 13. Amanatidis, T., Langdon, P., Clarkson, P.J.: Needs and expectations for fully autonomous vehicle interfaces. In: Companion of the 2018 ACM/IEEE International Conference on Human-Robot Interaction, pp. 51–52. ACM, New York (2018). https://doi.org/10.1145/ 3173386.3177054 14. Transportation Systems Catapult: Intelligent Mobility: Traveller Needs and U.K. Capability Study (2015) 15. Amanatidis, T., Langdon, P.M., Clarkson, P.J.: Inclusivity considerations for fully autonomous vehicle user interfaces. In: Breaking Down Barriers, pp. 207–214. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-75028-6_18 16. Clarkson, P.J., Coleman, R., Keates, S., Lebbon, C.: Inclusive Design: Design for the Whole Population. Springer, London (2013) 17. Langdon, P., Johnson, D., Huppert, F., Clarkson, P.J.: A framework for collecting inclusive design data for the UK population. Appl. Ergon. 46(Part B), 318–324 (2015). https://doi.org/ 10.1016/j.apergo.2013.03.011 18. Politis, I., Langdon, P., Bradley, M., Skrypchuk, L., Mouzakitis, A., Clarkson, P.J.: Designing autonomy in cars: a survey and two focus groups on driving habits of an inclusive user group, and group attitudes towards autonomous cars. In: Advances in Design for Inclusion, pp. 161–173. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-605975_15

Driving Behavior: Autonomous and Automated Vehicles

Investigating Drivers’ Behaviour During Diverging Maneuvers Using an Instrumented Vehicle Fabrizio D’Amico1(&), Alessandro Calvi1, Chiara Ferrante1, Luca Bianchini Ciampoli1, and Fabio Tosti2 1

2

Department of Engineering, Roma Tre University, Via vito volterra 60, 00146 Rome, Italy {fabrizio.damico,alessandro.calvi,chiara.ferrante, luca.bianchiniciampoli}@uniroma3.it School of Computing and Engineering, University of West London (UWL), St Mary’s Road, London W5 5RF, UK [email protected]

Abstract. Deceleration lanes are designed to improve traffic operations of interchanges to ensure safety conditions during diverging maneuvers. Nevertheless, highway diverge areas are often characterized by high crash rates and poor operations, demonstrating that their efficiency and safety need further research. The main objective of this study is to analyze driving performance on deceleration lanes; therefore, two deceleration lanes of an existing Italian highway have been studied, collecting data from drivers who drove an instrumented vehicle along a selected route. The results demonstrated a substantial difference between drivers’ maneuvers from those adopted by most guidelines and confirmed findings of previous research developed with different modes. Such driving performance caused significant interference with through traffic that, in turn, caused subsequent issues in operating and traffic safety conditions. The results of this study are quite promising, especially since they corroborate findings of previous studies developed by the same authors as part of a wider, long-term research. Keywords: Driver’s behavior Naturalistic driving


Road safety
Human factors


1 Introduction The design of geometric elements of roads is one of the main challenges for international professionals and road engineers for both complying with design standards and ensuring adequate levels of road safety. In particular, interchange zones and auxiliary lanes are reported to be critical parts of highways and other roads (i.e. [1, 2]). To meet traffic safety and operation requirements, it is important for ramps and speed-change lanes to be appropriately © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 169–178, 2020. https://doi.org/10.1007/978-3-030-20503-4_15

170

F. D’Amico et al.

designed so that vehicles may perform sequential maneuvers. At the same time, scientific literature on the risks typically associated with these elements appears to be quite limited and has often demonstrated contradictory results when observing potential conflicts between vehicles [3–5]. In Italian highways, where the accidents rates are generally low compared to the high volume of traffic, the interchanges zones design consists in sizing both the ramps from and for the connection with the roads, and the auxiliary lanes to reach the service areas. Currently, the road geometric Italian design standard [6] presents a single approach for the design of both the exit and the auxiliary lanes. The standard is based on geometrical and functional considerations rather than an empirical approach or driver’s behavior analysis. Therefore, further analyses might be useful to verify the actual drivers’ behavior and to support designers in selecting the best solution, according to the specific conditions (kinematics, traffic flow, etc.). To this effect, the driving performance along the exiting maneuver to a service area were recorded on a real highway, by means of an instrumented vehicle. The overall aim is to provide the practitioners with effective guidelines for designing deceleration lanes, which take into account drivers’ performances and human factors in general. A procedure to extract speed and trajectory data, vehicle deceleration value and other advanced measures respectively, was carried out. These elements have already been studied in previous driving simulator studies (i.e. [7–10]) to describe driving performance along deceleration and acceleration lanes. In view of that, the on-site investigations and the related results were useful for both calibrating and validating the instrument. Indeed, the wider research purpose is to effectively analyse, in simulation, the safety and operation effects of different design choices related to existing highway service areas on driving performance, and therefore to assist designers to select the most effective solution before any implementation on the existing road. In fact, several findings of previous studies highlight the needs for evaluating the safety and operation performance of deceleration (i.e. [1, 11]) and acceleration (i.e. [12]) lanes with regard to traffic volumes, as the traffic variables were observed to significantly affect driving performance on diverge areas. This paper presents the overall findings of the analysis of the driving performance along exiting manoeuvres from the highway service areas performed by on-site investigations.

2 Case Study An important highway connecting the city of Rome to Fiumicino International Airport was selected as a study case, according to the actual demands of highway managers with specific regards to two speed-change lanes to and from service areas where additional facilities are planned to be included. In fact, due to the expected increase of traffic, the safety and operation compatibility of the current deceleration (and acceleration) lanes connecting the highway to the service areas required to be verified. Specifically, the experiment took place on a 20 km road circuit characterized by the presence of two service areas. In particular, 15 km of this circuit were driven on the highway; approximatively 5 km to reach the northbound service area and further 5 km

Investigating Drivers’ Behaviour During Diverging Maneuvers

171

to the southbound service area. The first 5 km of the circuit were used as a training route to give the participants the possibility to familiarize with the vehicle by driving along the urban part of the itinerary. The cross section of the highway is composed by: – – – –

a dual carriageway of 22 m wide, consisted of two lanes for each driving direction; each lane was 3.75 m; each shoulder of 2.00 m wide; median of 3.00 m wide.

The speed limit is fixed at 110 km/h. Along the selected route the first approached service area is located in the northbound direction (north service area) while, just in front of it on the opposite carriageway (southbound direction), it is placed the second service area investigated in this study (south service area). Both the service areas are accessible from the highway through deceleration lanes. The width of each auxiliary lane is 3.50 m. Figure 1 shows an overview of the deceleration lanes characterized by different longitudinal geometries. The first deceleration lane, to reach the north service area from the highway (defined as Deceleration Lane Northbound, DLN) is composed by a 40 meters-long taper and a deceleration length of 110 m. The second deceleration lane, used for exiting the highway to the south service area (defined as Deceleration Lane Southbound, DLS) consists in 50 meters-long taper and 65 meters-long deceleration length. Both geometries result quite short considering the Italian guidelines [6].

Fig. 1. Geometrical configuration of the deceleration lanes and measurement points

172

2.1

F. D’Amico et al.

Participants and Procedure

The participants to the field study were selected among volunteers from the Engineering Department at Roma Tre University. In particular, a sample of 13 women and 27 men, for a total of forty licensed drivers with mean age of 27 years and a standard deviation of 7 years, took part to the tests. Furthermore, all of them reported to drive an average of 5000 km on highways per year, while the 85% of the sample reported to drive a vehicle as main mode of transportation. Regarding to the organization of the tests, all the driving sessions were conducted under the same boundary conditions. Specifically, the weather was dry for all the tests, which were also conducted during two pre-defined time-slots, to assure homogeneous environmental and traffic conditions for all the participants. The first slot was fixed between 10:00 am and 1:00 pm and the other one between 2:00 pm and 3:30 pm. Under these time-slots similar low traffic volumes were previously recorded, corresponding to 1500 vehicles/hour with an average speed of 110 km/h on the left lane and 85 km/h on the right lane in both directions (northbound and southbound). All the participants were asked to drive along the highway route and to reach the two service areas travelling along the two deceleration lanes. More specifically (see Fig. 2), after a training route of 5 km, the participants were asked to exit from the highway to step into the Northbound service area along the first deceleration lane, identified by DLN; they had a rest of 5 min in the service area and then they re-joined the highway through the acceleration lane. Afterward, they made a reverse manoeuvre taking the first available interchange along the highway, and repeated the same procedure towards the south service area (located in the opposite direction) by driving

Fig. 2. Field-test circuit and training route selected

Investigating Drivers’ Behaviour During Diverging Maneuvers

173

through the second deceleration lane, identified by DLS; then, they had again 5 min of rest after which they entered the highway and finally they reached the end of the itinerary corresponding to the end point of the route. 2.2

The Instrumented Vehicle

All the driving tests were developed using a common utility car properly equipped with low-cost tools consisting of: • four action cameras set on a resolution of 720  25 pixels; • a common GPS device with a 1 Hz refresh rate; • a digital stopwatch of sensitivity 0.01 s. To collect information related to the vehicle’s trajectory and speed, all the instruments have been placed into the vehicle at four designated stations (see Fig. 3): 1. at the driver side window, a camera is placed with the overall function of giving a reference of the longitudinal position by recording the hectometer signals along the road; 2. set on the windshield midpoint, consisted of another camera that filmed the front view of the road to provide the vehicle lateral position; 3. made up of another camera, the GPS device, and the stopwatch. The camera’s function was to record the vehicle speed showed on the GPS and the time reference showed on the stopwatch, which was necessary to calculate the travelled distance and speed; 4. on the right-side window, a camera to get an overview of the driving conditions.

Fig. 3. The instrumented vehicle configuration

2.3

Data Collection

To study the driver behaviour along the exiting manoeuvres (from the highway to the service areas), the driving speeds and trajectories were collected whilst approaching the deceleration lanes, during the exiting manoeuvres and along the deceleration lanes.

174

F. D’Amico et al.

For studying the exiting manoeuvre through the deceleration lanes, the driver’s speed was recorded at seven measurement points (see Fig. 4): • two points were fixed respectively at 500 m and 100 m before the beginning of the taper (defined as site −500 and −100, respectively); • one point at the beginning of the taper (defined as site 0); • three points which depended on the driver’s lateral position or trajectory (considering the center of gravity of the driver’s vehicle), specifically defined as: – site A, where the vehicle began to change the lane, that is where the vehicle’s trajectory crossed a line parallel to the line between the right through lane and the deceleration lane at 0.85 m from its left-hand side (the vehicle was 1.70 m wide); – site B, where the driver’s trajectory crossed the line between the right through lane and the deceleration lane; – site C, where the vehicle was completely within the deceleration lane, that is where the vehicle’s trajectory crossed a line parallel to the line between the right through lane and the deceleration lane at 0.85 m from its right-hand side. • the last measurement site was fixed at the end of the deceleration lane, where the exit ramp curve began (defined as site DR). The longitudinal distances of the three points (A, B and C) from the beginning of the taper (site 0) were recorded (defined dA, dB and dC, respectively) for providing information on the trajectory of the change-lane manoeuvre. To obtain all these data, a video analysis and data post-processing was required to retrieve the information about vehicle trajectory and speed. From the action cameras more than 150 videos (four videos for each participant), were collected and then exactly synchronized with respect to the beginning of the test. A subsequent discretization and synchronization allowed to select, for each driver, four different frames extracted by the synchronized cameras every 0.1 s, providing all the information needed for depicting speeds and trajectories of the drivers. In such a way, all the

Fig. 4. Average values of speeds and positions measurement sites

Investigating Drivers’ Behaviour During Diverging Maneuvers

175

approximations on speeds and trajectories data caused by the low level of accuracy of the GPS device used in this experiment have been overcome.

3 Analysis and Results Respect to the original sample of forty drivers, three drivers were excluded from data post-processing because during exiting the highway for the service areas, their manoeuvres were biased by other vehicles in front of them. Moreover, the Chauvenet criterion was applied but no drivers were excluded; hence, a final sample of thirty-seven drivers were used for the study. From the field test, for each deceleration lane were therefore considered three samples of distances (dA, dB and dC respectively) and seven samples of speeds (corresponding to the seven measurement sites described in the previous paragraph). The average values of positions and speeds for each deceleration lane were calculated, and a standard deviation analysis for each measurement site was performed, as shown in Tables 1 and 2. The wide research program in which this study is framed allows the on-site speeds and trajectories values to be compared with those coming from both other on-field studies and driving simulation tests, in order to enlarge the database and make a more comprehensive evaluation of driving performances. For instance, with different samples of data from both field tests and driving simulations, it might be possible to perform comparative analysis and elaboration to validate the simulator and, consequently, to evaluate different performances of drivers in different auxiliary lanes redesigned in virtual reality. In this study, Fig. 4 and Tables 1–2 show all the average values of speeds and distances recorded in the field-test and in the next paragraphs some preliminary impressions in terms of drivers’ behavior are highlighted. 3.1

Speed

In the Fig. 4 the speed at −500 (V−500) can be considered as the driver’s speed that is not influenced by the diverge area, while the speed at site C (VC) corresponds at which drivers complete the change-lane manoeuvre without more interfering with through vehicles. The comparison between V−500 and VC, both in DLN and in DLS, allows a preliminary analysis of the exiting manoeuvres by different deceleration lanes: the outcomes show that the drivers decelerate before entering the deceleration lanes, both for DLN and DLS. This demonstrates the assumption of the Italian Guidelines [6] regarding the beginning of the vehicle’s deceleration after stepping into the auxiliary lane, to be in this case substantially incorrect. Moreover, along the entire deceleration maneuver the deceleration value is found not to be constant, which is also in contrast with the guidelines’ assumption on the length of the lane depending only on the speed differential that the driver has to reach from the beginning to the end of the parallel lane rather than on the deceleration pattern too. Hence, even in this case, the hypothesis behind the Italian Guidelines [6] founds no confirms from the field data.

176

F. D’Amico et al.

More in detail, the speed values show that drivers have significantly reduced their speeds before diverging from the highway and entering the deceleration lanes (in DLN from 84.42 km/h, in site −500, to 62.45 km/h, in site C, −26% approximately; in DLS, from 82.12 km/h, in site −500, to 58.14 km/h, in site C, −29% approximately). Similar results have been found in previous driving simulation studies (i.e. [7–9, 11, 12]) and highlight crucial traffic operation issues caused by the high interferences between the exiting drivers and the vehicles that intend to maintain their speeds.

Table 1. Average speed and standard deviation values for both DLN and DLS Speeds Measurement site Average speed V [km/h] Standard Deviation SD [km/h] DLN −500 84.42 10.18 −100 79.17 7.06 0 73.38 5.97 A 73.34 6.17 B 69.40 6.62 C 62.45 8.75 DR 43.18 5.11 DLS −500 82.12 9.54 −100 79.93 7.71 0 72.69 7.79 A 73.13 7.93 B 66.04 7.54 C 58.14 7.34 DR 46.66 5.23

3.2

Position and Trajectory

The instrumented vehicle and the data processing procedure allowed to collect significant information on the position and trajectory of the vehicles. The values reported in Table 2 and the trajectories deduced from the field data show a peculiar approach of the driver representing a “regular” exiting manouvre. The “regular exiting manouvre” is composed of the following steps: (1) change of the lane, (2) travel on the deceleration lane to reduce the speed, and (3) exiting at the end of the deceleration lane, where the exit ramp curve began. Both in DLN and in DLS the dA values show that the changing of the lane began before of the taper (site “0”) and it means the will of the drivers to change the lane as soon as possible probably due to the reduced length of the deceleration lanes. Moreover, this action was made at variable speed, in contrast with the assumptions of Italian guidelines.

Investigating Drivers’ Behaviour During Diverging Maneuvers

177

Evaluating the position values (A, B, C) as well as the related trajectories, it is possible to note a different approach, here defined “direct exiting manouvre” and consisting in (1) anticipating the change of lane in deceleration mode, (2) travelling on the deceleration lane with “transversal” trajectory and variable speed, (3) exiting the highway at the end of the deceleration lane. This trajectory’s configuration, according to the international literature is normally preferable along tapered and not parallel deceleration lanes. The approach of the drivers may have been due to the shortness of the deceleration lanes in presence of a highway characterized by high speed differential between running speed at right lane and exiting speed at the end of deceleration lane. This hypothesis might be verified by future comparisons with new on-field data from different deceleration lanes as well as simulation data coming from virtual scenarios created in same geometrical configurations.

Table 2. Average distance and standard deviation values for both DLN and DLS Distances Measurement site Average distance d [m] Standard Deviation SD [m] DLN A −2.25 47.86 B 39.78 26.20 C 79.93 30.87 DLS A −17.26 52.70 B 37.72 23.69 C 58.14 21.33

4 Conclusions and Future Perspectives This paper presented only a part of a wider research program that the authors have been working on for several years and consists in verifying the driver’s behavior in auxiliary lanes and the effects of different design variables on driving performance. Specifically, the deceleration lanes of two service areas on an existing highway were selected for field tests. An instrumented vehicle was used by a sample of drivers along a fixed highway circuit involving the stop into two service areas. A set of synchronized HD cameras allowed to record the front, left, right views of the travel and a navigation device that displayed the vehicle’s speed and position respectively and, finally, a GPS equipment. The authors here present a procedure based on the post-processing of a large amount of data obtained from an instrumented vehicle. Furthermore, the authors identified new advanced indicators capable of illustrating driving performance along deceleration lanes, thus highlighting potentially unsafe driving conditions. This study confirms the potential of data from naturalistic for safety analysis and assessments of the driver’s real behavior.

178

F. D’Amico et al.

Nowadays, the data might be collected more easily thanks to the increasing vehicle equipment, expanding the number and variety of samples and allowing to carry out more extensive statistical analysis. The next step of the research, currently ongoing, consists in developing simulation tests for both validating the simulator and changing geometrical and environmental variables to compare different solutions and, thereby, selecting the most effective one in terms of operating and safety performance before any implementation on the existing highway, with impressive benefits in terms of cost reduction and safety.

References 1. Chen, H., Liu, P., Lu, J.J., Behzadi, B.: Evaluating the safety impacts of the number and arrangement of lanes on freeway exit ramps. Accid. Anal. Prev. 41(3), 543–551 (2009) 2. McCartt, A.T., Northrup, V.S., Retting, R.A.: Types and characteristics of ramp related motor vehicle crashes on urban interstate roadways in Northern Virginia. J. Saf. Res. 35, 107–114 (2004) 3. World Health Organization (WHO): Global status report on road safety 2018 (2018). https:// www.who.int/violence_injury_prevention/road_safety_status/2018/en/. Accessed 2 Feb 2019 4. Khorashadi, A.: Effect of ramp type and geometry on accidents. FHWA/CA/TE-98/13. FHWA, Washington, D.C. (1998) 5. Padimitriou, E., Theofilatos, A.: Meta-analysis of crash-risk factors in freeway entrance and exit areas. J. Transp. Eng. Part A Syst. 143(10), 1–10 (2017) 6. Ministero delle Infrastrutture e dei Trasporti (Italian Ministry of Road Infrastructure and Transport). Norme funzionali e geometriche per la costruzione delle intersezioni stradali. Gazzetta Ufficiale, No. 170 (2006) 7. Calvi, A., Bella, F., D’Amico, F.: Diverging driver performance along deceleration lanes: driving simulator study. Transp. Res. Rec. J. Transp. Res. Board 2518, 95–103 (2015) 8. Bella, F., Garcia, A., Solves, F., Romero, M.: Driving simulator validation for deceleration lane design. In: Proceedings of 86th Annual Meeting of the Transportation Research Board, Washington, D.C. (2007) 9. Bella, F., Calvi, A., D’Amico, F.: Evaluating the effects of the number of exit lanes on the diverging driver performance. J. Transp. Saf. Sec. 10(1–2), 105–123 (2018) 10. Livneh, M., Polus, A., Factor, J.: Vehicle behavior on deceleration lanes. J. Transp. Eng. 114, 706–717 (1988) 11. Calvi, A., Benedetto, A., De Blasiis, M.R.: A driving simulator study of driver performance on deceleration lanes. Accid. Anal. Prev. 45, 195–203 (2012) 12. Calvi, A., De Blasiis, M.R.: Driver behavior on acceleration lanes: driving simulator study. Transp. Res. Rec. J. Transp. Res. Board 2248, 96–103 (2011)

Model of Driving Skills Decrease in the Context of Autonomous Vehicles Darina Havlíčková(&), Petr Zámečník, Eva Adamovská, Adam Gregorovič, Václav Linkov, and Aleš Zaoral Transport Research Centre, Lísenská, Brno, Czech Republic [email protected]

Abstract. The aim of this presentation is (1) to define the skills necessary to control the driving of an autonomous vehicle; (2) skills needed to tackle the errors and failures of an autonomous vehicle and (3) to propose the operationalization of these skills. The view on driving skills decrease is built on the theoretical hierarchical model of driving behavior “GDE” – Goals for Driver Education model”. This can be used as the theoretical basis for measuring the decline in driving skills. The model is then put together with knowledge about human behavior and its changes in the context of automation and autonomous mobility. This definition and measurement suggestion is the first step in the long run of tackling the issue of reducing driving skills in the context of autonomous driving. Increase in automation promises a lot of benefits but on the other side, it also brings a decline in human ability to drive. It is a well-known finding of cognitive psychology that not using skills may cause forgetting and gradual loss of that ability or skill. Therefore, there have been some concerns connected with excessive automation in various areas of human lives for some time. But the topic of the automation and the pitfalls associated with it is not a new issue. For example, Bainbridge a long time ago drew attention to possible problems. Automation limits gaining experiences that can be needed when the control is needed to be passed back to the human operator. Even autopilot monitoring itself is based on the skills acquired by operators from experience with manual control, and future generations of operators who only gain experience from overseeing automats and autopilots will no longer have such. The model, which will be presented takes into account all above-mentioned aspects of driving in the automation era. Keywords: Automated vehicles


Driving skills decrease
GDE

1 Model of Driving Skills Decrease in the Context of Autonomous Vehicles The aim of this paper is (1) to define the skills necessary to control the driving of an autonomous vehicle; (2) skills needed to tackle the errors and failures of an autonomous vehicle and (3) to propose the operationalization of these skills. We build on the theoretical hierarchical model of driving behavior “GDE” – Goals for Driver Education model” [1–3]. This can be used as the theoretical basis for measuring the decline in driving skills. The model is then put together with knowledge about human behavior © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 179–189, 2020. https://doi.org/10.1007/978-3-030-20503-4_16

180

D. Havlíčková et al.

and its changes in the context of automation and autonomous mobility. This definition and measurement proposal is the first step in long run of tackling the issue of reducing driving skills in the context of autonomous driving. Increase in automation promises a lot of benefits but on the other side it also brings a decline in human ability to drive. It is a well-known finding of cognitive psychology that not using skills may cause forgetting and gradual loss of that ability or skill [e.g. 4]. Therefore, there have been some concerns connected with excessive automation in various areas of human lives for some time. But the topic of the automation and the pitfalls associated with it is not a new issue. For example Bainbridge [5] long time ago drew attention to possible problems. Automation limits gaining experiences that can be needed when the control is needed to be passed back to human operator. Even autopilot monitoring itself is based on the skills acquired by operators from experience with manual control, and future generations of operators who only gain experience from overseeing automats and autopilots will no longer have such.

2 Key Competencies in Increasing Automation In the time of autopilots the key role of the operator is no longer the manual skills itself, but above all the control of the automated system, the ability to detect errors and the ability to adequately react in the event of failure of automated system. For the adequate ability of takeover or correct assessment of automated system performance is necessary maintenance of the driving skills – you have to understand what should system do and how good it has to be. However, the absence of manual activities leads to the above mentioned decrease of skills. By understanding to this the paradox of automation becomes apparent [5]. On an empirical level, this is supported by Kessel and Wickens [6], who have been able to detect the failures of the system in manual and automatic mode. The authors found that participants with the previous experience with manual mode perform better in controlling automatic mode than participants who only had experience with automatic system control. This finding points to the importance of manual experience to build the necessary skills to detect automated system errors. Similarly, Kaber, Onal, and Endsley [7] showed that when automated system failed, participants who used automation systems with more share of manual control showed better performance, shorter reaction times and a better overview of the situation than those who used automation with less use of manual control. Proving these experimental results in practice was made thanks to an increase in automation in air transport. It can be tracked to late eighties, when for example Weiner [8] discovered, that there is concern about the loss of pilots’ ability to fly a plane in case of over-automatization, so pilots occasionally have to drive in manual mode to avoid decreasing their ability to drive. Pilots themselves state that in addition to reducing the workload, automation has also had a negative effect on their pilot skills. That is why they consider manual driving as an important part of each flight to maintain these skills [9]. Automation failure, however, not only represent a technical error of the system, on which is needed to respond to. Automation failure may be also due to misuse of automation. One of the first studies which summed up automation problems was made by Parasuraman and Riley [10]. They divided possible automation problems into three

Model of Driving Skills Decrease in the Context of Autonomous Vehicles

181

categories: misuse, disuse and abuse of automation. Misuse refers to overreliance on automation, which can result in failures of monitoring or decision biases. As the factors affecting the monitoring of automation they identified workload, automation reliability and consistency, and the saliency of automation state indicators. Disuse refers to neglect or underutilization and abuse refers to use without due regard for the consequences for human performance. The problem with automation errors lies not only in the absence of the necessary skills to intervene when manual practice is not possible, but also in the doubts that people will be suited to this tedious monitoring role of constantly watching to detect and correct technological failures [11]. Therefore, the problem lies also in the ability of humans to register or to recognize errors of automated system in time and adequately assess it. The skills needed in the context of automated driving are therefore multi-layered. This paper is not dealing with new skills or capabilities needed for example for monitoring the performance of an autonomous vehicle. The study primarily deals with the skills that are needed when the autopilot in an automated vehicle (AV) fails to drive and requests takeover from human driver. The failure situation puts for example demands on the ability to quickly find orientation and overview in a new situation. This and other driving skills necessary for successful and safe driving when human driver is taking control (which occurs especially in critical situations) are primary interest of this article.

3 Important Driving Behavior and Driving Skills Spulber and Wallace [11] point out that in the field of automated vehicles exists much uncertainty about the impact of the deployment and adoption of AV both at the societal and individual level and in-depth human factors research is still needed in this topics: (1) Drivers’ willingness to use the automation; (2) Drivers’ support with the appropriate level of automation; (3) Transitions between manual and higher levels of automated driving; (4) Possible loss of skill; (5) Drivers’ reactions to errors of AV technology. In the field of possible skill decrease, we have just poor empirical evidence, because we do not yet have drivers who are affected by driving skills decrease due to drive an autonomous vehicle. Empirical evidence are usually connected with ADAS – Adaptive cruise control, Lane Departure [12–14], Forward Collision Warning [15] and they demonstrate a change in driver behavior. E.g. Vollrath, Schleicher and Gelau [12] document delayed driver reactions in critical situations, e.g., in a narrow curve or a fog bank. But there was a lot of done in the field of transition between manual and automated driving (manual takeover/fallback performance) and reactions to errors of AV technology. Both areas are closely related to driving skills decrease and we can be inspired. In the real environment, there are data from California motor vehicle department summed up in annual disengagement reports [16–19] or in surveys from these data [20, 21]. In most cases of fatal traffic accidents, the autonomous mode was disconnected by the driver just before the collision, bud without avoiding collision. It can be related to the topic of trust and reliance on autonomous technology. Trust in autonomous technology influence the chose to use automation [10, 22]. The field of transition between manual and automated driving also has information about reaction times when manual take-over is required. This interval varies from 0.82 s (on average)

182

D. Havlíčková et al.

[23] or 0, 83 s (on average) [20] to 3, 08 ± 1, 16 s (on average) [24]. Full lateral control requires 6–9 s interval [25] and “comfortable transition time” defined by Merat et al. [26] requires on average 40 s. Moreover performing a non-driving-related task resulted in a longer reaction time [25]. Unlike these reaction times and above mentioned willingness to use the automation which we can measure in practice, driving skill decrease is difficult to measure in practice. Because we do not yet have drivers who are affected by driving skills decrease due to drive an autonomous vehicle. Therefore there is no empirical research available on this topic and there is a question how to define and measure driving skills decrease in the context of autonomous mobility at all. In general, Strand et al. [27] results indicate that driving performance degrades when the level of automation increases. In specific Spulber and Wallace [11] define maneuvering skills that could be negatively affected by automation include: (1) maintaining longitudinal and lateral control; (2) parking; (3) respecting traffic signs, reacting to different traffic situations (e.g., speedway, inner city); (4) handling weather conditions (e.g., rain, fog, snow, ice, nighttime); (5) reacting to unexpected situations (e.g., vehicle failure, crash avoidance); (6) interacting with other vehicles or participants in traffic. It can be said that these capabilities pass through the first two levels of the GDE model [1–3], which can be used as a theoretical basis for measuring skill decrease. This model assumes four levels of driving competence: (1) Vehicle maneuvering controlling direction and position, controlling speed; (2) Mastering traffic situations – adapting to demands of present traffic situation; (3) Goals and context of driving – purpose of the trip, environment, social context, company; (4) Goals for life and skills for living – skills for self-control, importance of cars and driving on personal development. Skills for vehicle maneuvering and mastery of traffic situations are the basis for successful operation in traffic and these aspects should be learned well during driver training. Psychomotor and physiological aspects are important as basic requirements for operations at the lowest levels of the hierarchy of driver behavior [2]. Therefore, in this research, we want to focus on these first two levels. The first level include following knowledge and skills: distance to others/safety margins etc., control of direction and position, tire grip and friction, vehicle properties, physical phenomena etc. Risk-increasing factors include, for example insufficient automatism/skills, unsuitable speed adjustment. Self-evaluation include, for example strong and weak points of basic maneuvering skills. In the context of AV, the first level of the GDE model needs to be supplemented by later and more specific work. E.g. Trösterer et al. [28] suggest that mechanical driving skills are retained after longer period of inactivity. Because “driving resembles riding a bike – and as the saying goes: “You never forget how to ride a bike.” The challenge that comes along with this is that if you have not learned it properly once, you won’t master it later on. And future work will focus on what skill level is required for a driver to be sufficiently skilled so that they can react appropriately when they receive the feared “Please take over!” message” [28]. In general, there is a consensus, that in order to know how to react when faced with a limitation or malfunction, drivers must maintain a skill level allowing them to manually perform all tasks that are normally done by automation (e.g., longitudinal control, lateral control), as well as emergency maneuvers, e.g., crash avoidance [11].

Model of Driving Skills Decrease in the Context of Autonomous Vehicles

183

The second level include following knowledge and skills: traffic rules, observation of signals, anticipation of course of situation, speed adjustment, communication, driving path and driving order. Risk-increasing factors include, for example wrong expectations or vulnerable road-user. Self-evaluation include, for example strong and weak points for hazard situations. Also, at the second level, the GDE model needs to be supplemented by a study reflecting AV issues. Beside simple maneuvering skills, there is the key role of situational awareness [22, 30], especially when driving errors occur [31]. Matthews et al. [32] defined elements of situational awareness that need to be considered from a driving perspective: (1) Spatial awareness—an appreciation of the location of all relevant features of the environment; (2) Identity awareness—knowledge of salient items; (3) Temporal awareness—knowledge of the changing spatial picture over time; (4) Goal awareness—the highest goal may be the navigation plan to the destination; at a lower level, the maintenance of speed and direction to conform to the navigation plan; and at a still lower level, the need to maneuver and place the vehicle in an appropriate manner within the surrounding traffic stream; and (5) System awareness—relevant information within the larger driving environment as a system. It is doubtful whether situational awareness is retained as mechanical driving skills [28] however, in the takeover situation is crucial. Additionally, the response is slower for drivers who carry out other activities (magazine reading/smartphone handling) [25] and it can be assumed that these activities will become a common part of driving. In spite of all that has been mentioned, no detailed empirical measurement of the driving skill decrease in the context of autonomous driving was done until now. For this purpose was developed a multilevel model of driving skills required in the context of autonomous driving (Fig. 1).

Fig. 1. An overview of autonomous mobility relevant driving skills with risk of decrease

This is basic theoretical model for defining the driving skill decrease in the context of autonomous mobility. This model will be further enhanced and expanded based on empirical data from piloting and subsequent measurement. Now the model was completed with a background of age and lifestyle. There is general agreement that age is an important factor that has impact on driving skills and that older adults are at higher risk for fatal crashes [33] due to decline in cognitive, psychomotor, visual and hearing

184

D. Havlíčková et al.

abilities in traffic [e.g. 34–37] or medication [e.g. 38]. Older people are usually aware of those deficits and benefits from experience [39] and driving habits. However, autonomous mobility will not allow these experiences and driving habits to be achieved and will not allow to maintain them via practice. In addition, older drivers cannot be expected to develop self-regulatory driving behavior. This usually include driving more slowly that younger drivers, avoiding driving in poor weather condition, at night or in heavy traffic [e.g. 34]. There is no consensus on when exactly age-related changes in driving skills start to occur. The differences on individual level are enormous and factors as gender, education, life experience, biological age etc. have considerable impact as well [40]. But there is definitely a consensus that experienced drivers who drive often can keep their driving skills longer [41]. Our goal is, among other things, to measure these capabilities to determine whether and how they are decreasing. For this purpose, we are preparing a modification (for the context of autonomous mobility) of The Wiener Fahrprobe test – WF [42–44]. It is a standardized driving tests (standardized route of between 25 and 50 km in length that takes 40 to 60 min with two persons inside the car for observation) as an alternative measure tools. It was developed to analyze driving behavior in order to make sure whether a person is apt for driving a car or not [45]. In addition, we are preparing a modification (again for the context of autonomous mobility) of standardized traffic psychology assessment, focused on situational awareness. For traffic rules knowledges is available standardized test, used for final exam in driving school. A detailed overview of described skills, with their probability of decrease, relevance for context of autonomous mobility and possible measurement methods brings the following Table 1.

Table 1. Driving skills overview - with the relevance for the context of autonomous mobility Driving skills mechanical

Probability of decrease

Traffic rules knowledge

Low – possibility of training

Low – expectation of retaining *1, activation by regular training Gear shift operating Low – expectation of retaining *1, activation by regular training Mirrors usage Low – expectation of retaining *1, activation by regular training Safety distance to Middle - documented others effect of ADAS use *2 Gas brake clutch coordination

Relevance for the context of autonomous mobility High - the basic assumption of orientation Middle - when assuming regular training Middle - when assuming regular training Middle - when assuming regular training Middle - when assuming regular training

Possible measurement methods

Traffic rules test

WF - modified

WF - modified

WF - modified

WF - modified

(continued)

Model of Driving Skills Decrease in the Context of Autonomous Vehicles

185

Table 1. (continued) Driving skills mechanical

Probability of decrease

Lane keeping

Middle - documented effect of ADAS use *3

Fluency

Middle - documented effect of ADAS use *3

Driving skills mental Probability of decrease

Relevance for the context of autonomous mobility Middle - when assuming regular training Middle - when assuming regular training Relevance for the context of autonomous mobility High - especially at manual takeover *4

Reactions: other vehicles, VRU, obstacles

High - training is difficult

Interaction: achieving the necessary degree of orientation in the new situation Traffic signs awareness

High - training is difficult

High - especially at manual takeover *5

High - training is difficult

High - crucial in orientation in a new situation

Low - Speed perception can be maintained during autonomous driving

Middle prevention of collision in a critical situation

Speed awareness

Possible measurement methods

WF - modified

WF - modified

Possible measurement methods

WF – modified; Modified traffic; psychology assessment focused on traffic awareness WF – modified; Modified traffic psychology assessment focused on traffic awareness WF – modified; Modified traffic psychology assessment focused on traffic awareness WF – modified; Modified traffic psychology assessment focused on traffic awareness

*1 [28, 29] *2 [15] *3 [12–14] *4 E.g. [20, 23, 24] *5 E.g. [25, 26]

4 Conclusions and Recommendations In general, there is a consensus, that in order to know how to react when faced with a limitation or malfunction, drivers must therefore maintain a skill level allowing them to manually perform all tasks that are normally done by automation (e.g., longitudinal

186

D. Havlíčková et al.

control, lateral control), as well as emergency maneuvers (e.g., crash avoidance) [11]. Our considerations point to the critical importance of maintaining regular training. However, maintenance training requires a basic skill level at the beginning. However, skills such as reactions and interactions in traffic that cannot be trained remain, and their decline appears to be critical in the future. Our key issue for future practice and policymakers is to find the critical threshold for which an increased level of danger occurs. This critical threshold may be due to a lack of initial practice or may be due to the inadequate maintenance (practice) of skills over time. Setting these thresholds can then lead to recommendations for drivers or even to new policies like a the ability to drive an autonomous vehicle level 3 and level 4 [46] after a certain number of years of manual driving experience or inspired by air transportation [28] mandatory hours/kilometers of manual driving and different types of compulsory training or examinations. But this is a distant goal. Our goal now is to create the resources, definition and methods of measurement driving skills decrease in the context of autonomous mobility. Acknowledgements. This article was created with the financial support of the Czech Republic’s Technology Agency under the ÉTA program called Reduced Capacity to Drive (TL02000191) on Research Infrastructure acquired from the Operational Program Research and Development for Innovation (CZ.1.05/2.1.00/03.0064).

References 1. Hatakka, M., Keskinen, E., Gregersen, N.P., et al.: Theories and aims of education and training measures. In: Siegrist, S. (ed.) Driver Training, Testing and Licensing—Towards Theory-Based Management of Young Drivers’ Injury Risk in Road Traffic. Results of EUProject GADGET, Work Package 3. Bfu–Report 40 (1999) 2. Hatakka, M., Keskinen, E., Gregersen, N.P., Glad, A., Hernetkoski, K.: From control of the vehicle to personal self-control; broadening the perspectives to driver education. Transp. Res. Part F 5(3), 201–215 (2002) 3. Peraaho, M., Keskinen, E., Hatakka, M.: Driver competence in a hierarchical perspective; implications for driver education. University of Turku. Report to Swedish Road Administration (2003) 4. Rose, A.M.: Acquisition and retention of skills. In: McMillan, G.R., Beevis, D., Salas, E., Strub, M.H., Sutton, R., Van Breda, L. (eds.) Applications of Human Performance Models to System Design, pp. 419–426. Springer, Boston (1989) 5. Bainbridge, L.: Ironies of automation. In: Analysis, Design and Evaluation of Man-Machine Systems 1982, pp. 129–135 (1983) 6. Kessel, C.J., Wickens, C.D.: The transfer of failure-detection skills between monitoring and controlling dynamic systems. Hum. Factors 24(1), 49–60 (1982) 7. Kaber, D.B., Onal, E., Endsley, M.R.: Design of automation for telerobots and the effect on performance, operator situation awareness, and subjective workload. Hum. Factors Ergon. Manuf. Serv. Ind. 10(4), 409–430 (2000) 8. Wiener, E.L.: Cockpit automation. In: Wiener, E.L., Nagel, D.C. (eds.) Human Factors in Aviation, pp. 433–461. Academic, San Diego (1988)

Model of Driving Skills Decrease in the Context of Autonomous Vehicles

187

9. McClumpha, A.J., James, M., Green, R.G., Belyavin, A.J.: Pilots’ attitudes to cockpit automation. In: Proceedings of the Human Factors Society Annual Meeting, vol. 35, no. 2, pp. 107–111. SAGE Publications, Los Angeles, September 1991 10. Parasuraman, R., Riley, V.: Humans and automation: use, misuse, disuse, abuse. Hum. Factors 39, 230–253 (1997) 11. Spulber, A., Wallace, R.: Impact of automated vehicle technologies on driver skills. Transportation Systems Analysis Group. MDOT REQ. NO. 1259, Connected and Automated Vehicle Industry Coordination Task C.2. Impact of automated vehicle technologies on driver skills (2016) 12. Vollrath, M., Schleicher, S., Gelau, C.: The influence of Cruise Control and Adaptive Cruise Control on driving behavior – a driving simulator study. Accid. Anal. Prev. 43, 1134–1139 (2011) 13. Stanton, N.A., Dunoyer, A., Leatherland, A.: Detection of new in-path targets by drivers using Stop & Go Adaptive Cruise Control. Appl. Ergon. 42(2011), 592–601 (2011) 14. Jones, S.: Cooperative adaptive cruise control: human factors analysis. Federal Highway Administration (2013) 15. Muhrer, E., Reinprecht, K., Vollrat, M.: Driving with a partially autonomous forward collision warning system: how do drivers react? Hum. Factors 54(5), 698–708 (2012) 16. The California Department of Motor Vehicles. Autonomous Vehicle Disengagement Reports 2015. https://www.dmv.ca.gov/portal/dmv/detail/vr/autonomous/disengagement_ report_2015 17. The California Department of Motor Vehicles. Autonomous Vehicle Disengagement Reports 2016. https://www.dmv.ca.gov/portal/dmv/detail/vr/autonomous/disengagement_ report_2016 18. The California Department of Motor Vehicles. Autonomous Vehicle Disengagement Reports 2017. https://www.dmv.ca.gov/portal/dmv/detail/vr/autonomous/disengagement_ report_2017 19. The California Department of Motor Vehicles. Autonomous Vehicle Disengagement Reports 2018. https://www.dmv.ca.gov/portal/dmv/detail/vr/autonomous/disengagement_ report_2018 20. Dixit, V.V., Chand, S., Nair, D.J.: Autonomous vehicles: disengagements, accidents and reaction times. PLoS ONE 11(12), 1–14 (2016) 21. Favarò, F.M., Nader, N., Eurich, S.O., Tripp, M., Varadaraju, N.: Examining accident reports involving autonomous vehicles in California. PLoS ONE 12(9), 1–20 (2017) 22. Parasuraman, R., Sheridan, T.B., Wickens, C.D.: Situation awareness, mental workload, and trust in automation: viable, empirically supported cognitive engineering constructs. J. Cogn. Eng. Decis. Making 2(2), 140–160 (2008) 23. Fambro, D.B., Fitzpatrick, K., Koppa, R.: A new stopping sight distance model for use in highway geometric design. Transportation Research Circular Transportation Research Board, National Research Council, Washington, D.C. (1997) 24. Eriksson, A., Banks, V.A., Stanton, N.A.: Transition to manual: comparing simulator with on-road control transitions. Accid. Anal. Prev. 102, 227–234 (2017) 25. Dogan, E., Rahal, M.-Ch., Deborne, R., Delhomme, P., Kemeny, A., Perrin, J.: Transition of control in a partially automated vehicle: effects of anticipation and non-driving-related task involvement. Transp. Res. Part F Part A 46, 205–215 (2017) 26. Merat, N., Jamson, A.H., Lai, F.C.H., Daly, M., Carsten, O.M.J.: Transition to manual: driver behavior when resuming control from a highly automated vehicle. Transp. Res. Part F Part B 27, 274–282 (2014)

188

D. Havlíčková et al.

27. Strand, N., Nilsson, J., Karlsson, I.C.M., Nilsson, L.: Semi-automated versus highly automated driving in critical situations caused by automation failures. Transp. Res. Part F Traffic Psychol. Behav. 27, 218–228 (2014) 28. Trösterer, S., Meschtscherjakov, A., Mirnig, A.G., Lupp, A., Gärtner, M., McGee, F., Engel, T.: What we can learn from pilots for handovers and (de) skilling in semi-autonomous driving. In: Proceedings of the 9th International Conference on Automotive User Interfaces and Interactive Vehicular Applications - AutomotiveUI 2017 (2017) 29. Trösterer, S., Gärtner, M., Mirnig, A.G., Meschtscherjakov, A., McCall, R., Louveton, N., Tscheligi, M., Engel, T.: You never forget how to drive: driver skilling and deskilling in the advent of autonomous vehicles. Automotive UI (2016) 30. McCall, R., McGee, F., Mirnig, A., Meschtscherjakov, A., Louveton, N., Engel, T., Tscheligi, M.: A taxonomy of autonomous vehicle handover situations. Transp. Res. Part A Policy Pract. (2018) 31. Gugerty, L.J.: Situation awareness during driving: explicit and implicit knowledge in dynamic spatial memory. J. Exp. Psychol. Appl. 3(1), 42–66 (1997) 32. Matthews, M.L., Bryant, D.J., Webb, R.D.G., Harbluk, J.L.: Model for situation awareness and driving: application to analysis and research for intelligent transportation systems. Transp. Res. Rec. 1779, 26–32 (2001) 33. Eby, D.W., Molnar, L.J., Zhang, L., St. Louis, R.M., Zanier, N., Kostyniuk, L.P.: Keeping older adults driving safely: a research synthesis of advanced in-vehicle technologies. University of Michigan Transportation Research Institute and the ATLAS Center (2015) 34. Meyer, J.: Personal vehicle transportation. In: Pew, R.V., Van Hemel, S.B. (eds.) Technology for Adaptive Ageing, pp. 253–281. National Academies Press, Washington, D.C. (2004) 35. Yang, J., Coughlin, J.F.: In-vehicle technology for self-driving cars: advantages and challenges for aging drivers. Int. J. Automot. Technol. 15(2), 333–340 (2014) 36. Schaie, K.W.: Cognitive aging. In: Pew, R.V., Van Hemel, S.B. (eds.) Technology for Adaptive Ageing, pp. 43–63. National Academies Press, Washington, D.C. (2004) 37. Vichitvanichphong, S., Talaei-Khoei, A., Kerr, D., Hossein Ghapanchi, A.: What does happen to our driving when we get older? Transp. Rev. 35(1), 56–81 (2015) 38. Dickerson, A.E., Molnar, L.J., Eby, D.W., Adler, G., Bédard, M., Berg-Weger, M., Classen, S., Foley, D., Horowitz, A., Kerschner, H., Page, O., Silverstein, N.M., Staplin, L., Trujillo, L.: Transportation and aging: a research agenda for advancing safe mobility. Gerontologist 47(5), 578–590 (2007) 39. ERSO, European Commission: Older Drivers, European Commission, Directorate General for Transport (2015) 40. Haustein, S., Siren, A.K., Framke, E., Bell, D., Pokriefke, E., Alauzet, A., Marin-Lamellet, C., Armoogum, J., O’Neill, D.: Demographic Change and Transport. European Commission, Brussels (2013) 41. OECD: Ageing and Transport: Mobility Needs and Safety Issues. OECD Publication Service. Paris (2001) 42. Chaloupka, Ch., Risser, R.: Don’t wait for accidents - possibilities to assess risk in traffic by applying the Wiener Fahrprobe. Saf. Sci. 19, 137–147 (1995) 43. Risser, R., Lehner, U.: The Wiener Fahrprobe. Internal paper. Training for the application of the Wiener Fahrprobe at the University of Leeds. Institute for Transport Studies ITS used for training. Leeds (1998)

Model of Driving Skills Decrease in the Context of Autonomous Vehicles

189

44. Turetschek, Ch., Risser, R.: Analysis of impact of IST on driving behavior with the help of the Wiener Fahrprobe. Presentation during the Humanist Internal meeting in Toulouse, France (2004) 45. Risser, R.: Behaviour in traffic conflict situations. Accid. Anal. Prev. 17, 2 (1985) 46. SAE International: Taxonomy and definitions for terms related to on-road motor vehicle automated driving systems. SAE Int. J3016, 1–12 (2016)

The User and the Automated Driving: A State-of-the-Art Anabela Simões1, Liliana Cunha2, Sara Ferreira3(&), José Carvalhais4,5, José Pedro Tavares3, António Lobo3, António Couto3, and Daniel Silva6 1

4

5

DREAMS Research Unit, Universidade Lusófona de Humanidades e Tecnologias, Lisbon, Portugal 2 Centre for Psychology, Faculty of Psychology and Education Sciences, University of Porto, R. Alfredo Allen, 4200-135 Porto, Portugal 3 Research Centre for Territory, Transports and Environment, University of Porto, Rua Dr Roberto Frias, 4200-465 Porto, Portugal [email protected] CIAUD, Faculdade de Arquitetura, Universidade de Lisboa, Rua Sá Nogueira, Pólo Universitário, Alto da Ajuda, 1349-055 Lisbon, Portugal FMH, Laboratório de Ergonomia, Universidade de Lisboa, Estrada da Costa, 1499-002 Cruz Quebrada, Portugal 6 Centre for Psychology, University of Porto, R. Alfredo Allen, 4200-135 Porto, Portugal

Abstract. Automation in the road transport system is coming faster than expected being influencing and shaping the future of mobility. However, very few is known about the impact of automatic driving on traffic and how drivers will accept, use, trust and interact in traffic when driving a vehicle with a certain level of automation. Additionally, most of the potential users have unrealistic representations of autonomous vehicles, the driver’s role in automation or the impacts of full automation on the road transport system. Aiming at better understanding the drivers’ behavior when dealing with automated driving, this paper addresses the following issues based on a state of the art on automated driving: drivers’ preferences for the automation levels across different categories of drivers; limits of the technology; needs for changes in traffic laws, as well as licensing and training; driver’s promptness to resume the vehicle control following a long period of autonomous driving. Keywords: Automated driving
Human factors Takeover
Situation awareness
Public awareness




Trust




Overreliance




1 Introduction Automation represents a major technological advancement influencing and shaping the future of mobility. However, automation won’t replace human activity but instead it will impose new demands to the human driver or user. This requires continuous research on human factors issues towards the prevention of risky behaviors and avoidance of misuse and disuse. © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 190–201, 2020. https://doi.org/10.1007/978-3-030-20503-4_17

The User and the Automated Driving: A State-of-the-Art

191

Automation in the road transport system is coming faster than expected but most of the potential users have unrealistic representations of autonomous vehicles, the driver’s role in automation or the impacts of full automation on the road transport system. It will be necessary that users will appropriate the current innovations in this field, develop trust on their use and the required willing to use and pay for it. There is still much to research and a long run towards a fully connected and automated road system. Comparing to the technological development in aviation, where the automation of several components aims at assisting the pilot or assuming a certain level of control over the aircraft, automation in the road transport system introduces interesting perspectives in terms of human-automation interaction. Whilst aviation is a very closed system operating under very strict international regulations and being controlled and operated by highly skilled and experienced professionals, the road transport system is totally open to a great diversity of users (pedestrians, riders, drivers, etc.) just controlled by traffic laws under a poor supervision. Thus, the introduction of automation into the road transport system, requires new regulations and intensive public awareness under the required human- and technology-based supervision.

2 Learning from the History of Automation in Aviation Learning from the history of automation in Aviation is a starting point to understand the risks and costs of automation in road vehicles. In the Aviation sector, the increased automation came with some cost. On one hand, it has not been easy for pilots to understand what the automated systems were doing, but they have been taught to remain responsible for taking over when the automated systems reached their functional limits or malfunctioned. On the other hand, pilots were encouraged to use automation towards the exclusion of manual flight controls, which were leading to a potential risk of losing their manual flight skills. Systems that alert pilots to hazardous conditions (e.g., proximity to the ground or to other aircraft) have contributed significantly to aviation safety despite those initial challenges. These systems had initially a high number of false alarms, which led pilots to develop a low level of trust on them. Nevertheless, great improvements were made in terms of better sensors, as well as improved and standardized interfaces allowing for a better understanding and enhancing awareness. These improvements led to more reliable and robust systems that increased the pilot’s trust and the willing to using them. With the nowadays development of computer technology, automation in aviation increased the complexity in the cockpit with gains in safety. Thus, for the operations safety and efficiency, modern aircrafts are increasingly dependent on automation, which have some advantages and safety challenges [1]. On the one hand, automation relieves pilots from repetitive and non-rewarding tasks, for which humans are less suited; on the other hand, these conditions change the pilots’ active involvement in operating the aircraft into a monitoring role, which humans are particularly poor at doing effectively or for long periods. As a consequence, there is a workload decrease, changing the pilots’ active involvement in operating the aircraft into a monitoring role, which humans are particularly poor at doing effectively or for long periods. These situations have a potential to decrease the pilots’ situation awareness and, in consequence, to compromise their promptness to

192

A. Simões et al.

takeover [2]. Actually, as above-referred, the pilot is trained to remain always responsible to takeover when the automated systems reach their operating limits or malfunctioned. Compared to the road context, automated vehicles are running out of specific or updated regulations; on the contrary, several commercials show the driver working, reading or sleeping at the wheel of an automated vehicles.

3 Driving Automation Being the road transport system totally open to a huge range of users sharing the road environment, just submitted to traffic laws, but under poor supervision and enforcement, additional research and testing needs are emerging in terms of safety and security issues and public awareness about such new challenges on the road. The limits and risks of the available technology are known but research on human-automation interaction is still required and the reasons are twofold: on one hand, many systems have problems with implementation, human-system integration and performance when used in the real world; on the other hand, higher expectations have been created on such technology giving rise to unexpected behaviors, risky situations and even accidents. This means that the life cycle of the system development was not complete and the system maturity and readiness to use was not accomplished. Even if any incident or accident could be directly caused by unacceptable user’s behavior, the lack of a system maturity assessment will compromise the system readiness for use and the overall system capability in its expected operational environment. The use of any automated system requires more knowledge than the use of a mechanical system for the same use. This means that driving an automated vehicle or riding a self-driving vehicle will require a different level of knowledge and understanding of the system functioning. This is similar to being working at a high technology system context, which targets high educated and digital skilled employees. An automated vehicle requires understanding of the technology limits and the driver’s promptness to takeover when requested. Thus, the idea of a person at the wheel being working, watching a movie or sleeping is totally wrong and requires an urgent and serious public awareness about the limits and related risks of driving automation. 3.1

The Risks and Limits of the Technology

Recent literature discusses the main hurdles to wide adoption of fully autonomous driving, among which the vehicle technology’s level of maturity [3] and its constraints and limits, mainly at level of the physical and social world perception [4]. These issues pose new challenges in terms of accuracy, reliability and human (driver) trust in advanced technology vehicles [5]. Despite this fact, autonomous driving and its technology have been attracting economic and industrial interest for years and, for instance, commercial cars include increasing levels of driver-assisting systems year after year [6]. In light of these developments, optimistic estimations predict that by 2030 AV will be entirely reliable to replace most human driving [5]; but until then there are some technical challenges and limits to face at present [4, 5, 7]. Therefore, the debate on these technical limits is paramount, and it should be as broad as possible, since it is expected that AV will be present in all spheres of life that demand mobility services.

The User and the Automated Driving: A State-of-the-Art

193

The body of literature on the limits of AV tends to report different kinds of current constraints, which we may define in two complementary dimensions: (1) the technology design and its status in relation to the human activity in terms of the role assigned to the human in the presence of the automation technology; (2) and the implementation of technology in the real world, as a dynamic open system characterized by obstacles and unexpected events. Technology Design. The exponential progress in new and intelligent technologies of automation has already produced impacts on the labor market and employment or even on the social protections [8]. Furthermore, in some cases, recent debates about the impacts of technology in the work activity have shown that the progress of technology and the growth of workloads may go hand-in-hand [9]. Past and current trends on the technology’s design, design and development seem to understate the role of human activity, as a “second-class” [10] component of the system. As automation gets better and better (i.e., the “first-class” component), people are asked to come into play only when the technology fails; but in these situations, it is expected that human activity offsets the flaws in according to the requirements and dictates of the technology [9]. Below two situations are highlighted in the literature that can contribute to the limits issued from technology-centered design approaches: 1. Whenever the separation between the design of technological system and its implementation/execution is reinforced. If from a technological point of view, AV are practically ready to be used, the human factor seems to be the “adjustable variable” in order to assure the system reliability. In this view, the notion of human “resistance to change” is thus modified [11], not as an intrinsic trait of the human factor but as a condition determined by the way the system is designed and developed. 2. Whenever the technology is seen as something that is accepted vs. refused by the user [12]. In this case, the technology is a resource that people will accept to use if the internal conditions (attitudes, cognition, mental models and perceptions) and external conditions (level of satisfaction, context) are favorable. Unlike, in the symbiotic approach the technology is seen as an extension of the human factor. Technological design and development have underlying the notion of humantechnology interface as a continuum. The technology is not thus an end in itself but acquires a sense of a constitutive element of human activity. From these two positions we can see that automation does not mean a direct reduction or demise of human activity. On the contrary, it raises other fundamental questions about how technology changes the nature of activity, by transforming it. Hence, the human factor is now asked to perform new regulations in the face of those problems and difficulties that the technological system was unable to foresee. Technology in the Real World. Other authors identify a set of technical limits at present concerning the attempt to replace the presence of the driver and his human regulation. These limits illustrate the second aforementioned dimension. In the future, it is expected that vehicles autonomously drive in various situations (e.g. in complex urban settings), triggering target-oriented answers according to the unexpected events occurring in the natural environment [13]. In these situations, the automation technology

194

A. Simões et al.

should be able to make complex decisions based on the information obtained through automotive sensors, including the perception of the environment [6, 14, 15]. From the technological standpoint, here is precisely one of the main hurdles for the implementation of higher levels of driving automation, i.e., the understanding of the spatialtemporal relationship between the vehicle and its environment [14, 16]. During selfdriving, the ability to understand the surrounding is crucial for AV may shape their answers plans and, to a certain extent, even predict likely behaviors from other road users (e.g. non-automated vehicles; pedestrians). According to some authors [15, 16], both technology promoters and designers have been making efforts to attribute to technology (laser, radar and visual sensors; and path planning algorithms) the human ability of making sense of the world, that is, to sense the information from the environment1, reacting in accordance to unexpected or uncertain events. Nowadays, this aspect is considered as an extant limitation in terms of technological issues and therefore it should guide further researches [5]. Bearing this in mind, some authors are less optimistic about the implementation of AV fully operating without any human interaction; preferring to outlook a future scenario of collaboration between human driver and automation technology, rather than the complete replacement of the human factor in driving [10, 16]. At present, in the European and Canadian contexts [17, 18], it is estimated that AV have already operated under level 2 of automation and, in some cases, on level 3 (conditional automation)2. Levels 4 and 5 concern the highly automated and fully autonomous vehicles, able to operate in any situation without human intervention. If the current levels 2 and 3 may be considered as infancy stages of automation development [3], how far will level 5 be? Advancements in the abilities of the vehicles perceive the surrounding environment and decision-making enable more and more to adjust the behavior of the AV to different situations liable to occur in the road [14]; but will AV ever be in conditions to match (or overcome) the human ability of perception and decision-making under best conditions [16]? Could technology be as good at making sense of the information collected from the environment as it is at collecting it? Driving situations take place in an open system, entailing static as well as dynamic elements, and several environmental and meteorological factors, such as different levels of light or dense fog. As a multi-sensory adaptive system, human being uses these functions to make decisions enabling the recognition of patterns, dealing with unexpected events that the system is not programmed to handle, and reacting adequately to changed environmental conditions. High levels of Situation Awareness allow the driver to be permanently projecting ahead being proactive in avoiding hazardous situations instead of being just reactive. According to Endsley [19], as far as software for driving autonomy can demonstrate an ability to project and deal with the unexpected, the need for human drivers to stay engaged and able to act will remain.

1

2

AV obtain the perception of external environment through laser navigation (e.g. LiDAR sensors “Light Detection And Ranging”), visual navigation (e.g. for traffic sign recognition) and radar navigation (e.g. for distances perception) [17]. According to the levels of automation defined by Society of Automotive Engineers (SAE) [4].

The User and the Automated Driving: A State-of-the-Art

195

So far, driving software is created to deliver appropriate responses to a learned set of situations and conditions, which is not enough to deal with the unexpected. According to Pearl [20], a pioneer in the field of Artificial Intelligence, such systems are extremely limited because they cannot project new adaptations for changing situations. The ability to project future events will require much more capable software, built with models of the environment that can understand current and projected future situations upon which proactive decision making relies [21]. An additional risk related to the systematic use of automated systems that should be studied in order to be anticipated and prevented, is skill loss. It was previewed in aviation with recommendations for using manual controls once in a while. Additionally, pilots are subject to periodic trainings aiming at keeping intact their manual skills.

4 Acceptance, Trust and Reliance on the Automated Systems Among other topics, there is a need for research on public acceptance and trust in automation. Schoettle and Sivak [22] carried out a survey aiming at getting the public opinion regarding self-driving-vehicle technology in three major English-speaking countries: the USA, UK and Australia, having had useable responses from 1,533 persons aged of 18 and older. The main findings of this survey were the following: (1) the majority of respondents having previously heard of self-driving vehicles, had a positive initial opinion of the technology, and had high expectations about the benefits of the technology; (2) but the majority of respondents expressed high levels of concern about riding in selfdriving vehicles, security issues related to self-driving vehicles, and self-driving vehicle not performing as well as actual drivers; (3) respondents also expressed high levels of concern about vehicles without driver controls, as well as self-driving vehicles moving while unoccupied and self-driving commercial vehicles, busses, and taxis; (4) most respondents expressed a desire to have this technology in their vehicle, but they were also unwilling to pay extra for the technology offering similar amounts in each country; (5) females expressed higher levels of concern with self-driving vehicles than males and were more cautious about their expectations concerning benefits from using selfdriving vehicles. In comparison to the respondents in the U.K. and Australia, respondents in the U.S. expressed greater concern about riding in self-driving vehicles, data privacy, interacting with non-self-driving vehicles, self-driving vehicles not driving, human drivers in general, and riding in a self-driving vehicle with no available driver controls. The main implications of these results are that drivers and the general public in the three surveyed countries, while expressing high levels of concern about riding in vehicles equipped with this technology, feel positive about self-driving vehicles, have optimistic expectations of the benefits, and generally desire self-driving-vehicle technology when it becomes available, although the majority was not willing to pay extra for such technology. However, at the time the survey was applied there was not a perfect awareness about the limits of the technology and related risks. Thus, it’s now time to survey again people’s attitudes, concerns, trust and willing to use and pay for these technologies.

196

A. Simões et al.

More recently, a paper based on interviews conducted with twelve expert researchers in the field of Human Factors (HF) and automated driving aimed at identifying commonalities and distinctive perspectives regarding HF challenges in the development of AVs, has been published [23]. In this paper, Kyriakidis et al. pointed out that “many challenges pertaining to the interaction between human drivers and automated systems are yet to be resolved”. Between these, are “the human drivers’ levels of acceptance, trust, and reliance on the automated systems” [23]. Giving some examples of the experts’ considerations about this subject, the opinion of Brookhuis [23] is that as system failures cannot be excluded, additional research should focus on public acceptance and trust in automated vehicles”. In the same line, Bengler [23] says that acceptance of automated vehicles by the public is a big topic and the first of main tasks of Human Factors research is to define the acceptance criteria of human drivers regarding the automated driving functionalities. van Arem [23], considering that while the human drivers will be supervising the system and intervene, if required, they will not be allowed to be engaged in a large variety of non‐driving tasks, conclude that the benefits for the consumers, as well as their acceptance and willingness to buy such automated vehicles, are limited. Also, on this point and about the SAE automation levels 2, 3 and 4, Andersson [23] raises the question: Who would like to use automation if they remain liable at all times for a system that they partially cannot control? Merat [23] says that, within the next 10 to 15 years, it is rather likely that the cost and maintenance of vehicles with automated functionalities will be quite high, which will be a major barrier towards their deployment and acceptance by the majority of the public. Finally, Flament [23] points out that “the same vehicle, depending on its environment and its access to reliable information, could allow more than one level of automation. The HF challenge in this case will be to clearly inform the driver about the possible levels of automation at any given time and place, and why this is so. This will lead to trust and acceptance of automation, but too much trust may cause overreliance together with unintended use, misuse, and even abuse”. The problem is that, so far, there is not a concerted, cohesive and cross-cutting policy on public awareness focused on automated driving or automated cars, the automation levels and the practical meaning of each one under a safety and secure umbrella. On the contrary, there is a massive advertising on self-driving cars using unrealistic images, which together lead to misunderstanding, overreliance, negative risk-taking and, sometimes, misuse. That’s why there is an urgent need for updated traffic regulations and public awareness about this new era in the road transport system. 4.1

Overreliance and Complacency in Automated Driving

The first effect of being at the wheel of an automated car and riding in automated mode is a mental underload that can lead to drowsiness after a while; reading or watching a movie leads to a switch on the driver’s attention from the road to a different object; sleeping at the wheel of an automated car following a driver’s decision to do it; all these three conditions impair the driver’s promptness to resume control under request. As far as a vehicle arrives into the market with a cockpit, it means that a licensed driver is required behind the wheel in order to resume the control under the system request or his/her own decision to do it, once the human is assumed to have the final authority

The User and the Automated Driving: A State-of-the-Art

197

over the automation [24]. Thus, once again, it is necessary to better know and understand the system in order to make appropriate decisions when activating or deactivating an automated mode. This requires knowledge about the system and understanding of its functioning, giving rise over time to trust in the system under predefined boundaries. The lack of knowledge or understanding of the system functioning and its limits can give rise to an overreliance on the system, which is a risky attitude underlying further risky behaviors. It is also frequent to consider that the system is always running as supposed to do and that there is nothing to concern about. This attitude risks to create a path to complacency, accepting anything as normal. In the field of Aviation, it has been reported that automation-related complacency was among the top five contributing factors for accidents [25]. Experiments carried out by Parasuraman et al. [26] indicate that the operator’s attention allocation strategy appears to favor his or her manual tasks as opposed to the automated task. This strategy may itself result from an initial orientation of trust on automation, which is then reinforced when the automation performs at the same, constant level of reliability. Therefore, automation in the context of the road transport system is being perceived as highly reliable, which leads to an increasing trust on the technology that is expressed in less system monitoring once no failures were expected. This is leading to the driver’s overreliance and complacency in automation. At the same time, this is underlying some disseminated images of drivers reading or watching a movie at the wheel table, lying down and sleeping, among other images that are not realistic, inducing wrong representations of automated driving and lead to unsafe behaviors. This compromises the driver’s promptness to resume the vehicle control following a long period of autonomous driving, reinforcing the needs for updated regulations and a serious public awareness on driving automation and drivers’ behavior.

5 The Needs for Updated Regulations and Public Awareness Despite the good intentions presiding at the development of automated vehicles, their increasing number sharing roads and the urban environment with a great variety of vehicles from different generations and different categories of road users, are creating new driving conditions giving rise to behavior adaptations. However, such intuitive behavior adaptations developed out of any update of the existing traffic regulations or new ones, is highly risky and open a window to compromise the initial good intentions of this fast change. The document issued by the U.S. Department of Transportation in October 2018 “Preparing for the Future of Transportation: Automated Vehicles 3.0” [2] provides a clear and consistent approach for automated vehicles related policy, based on six principles: (1) prioritize safety, using the potential of automation to improve safety for vehicle operators and every road user being aware that new safety risks appear and must be identified and managed in order to create trust on the technology and willing to use it; (2) remain technology neutral supporting the fast development of automated vehicles and giving rise to competition and innovation as a means to achieve safety, mobility solutions and economic goals; (3) Modernize traffic regulations, eliminating

198

A. Simões et al.

outdated ones impeding the development of automated vehicles or that do not address critical safety needs; (4) encourage a consistent regulated and operational environment, building consensus among policy makers, industry and stakeholders; (5) prepare proactively the society for automation through the provision of guidance, best practices, pilot programs, and other assistance towards a dynamic and flexible automated future; (6) protect and enhance the freedom of driving each one’s vehicle and sharing the road with conventional, manually-driven vehicles and other road users. In Australia, the National Transport Commission has issued the following policy recommendations to the Transport and Infrastructure Council towards a uniform approach to driving laws for automated vehicles [27]: (1) an automated driving system that has been approved under and continues to comply with the safety assurance system will be allowed to perform the dynamic driving task when it is engaged; (2) It ensures that there is always a legal entity responsible for the dynamic driving task when the automated driving system is engaged; (3) it clarifies who is the responsible entity at various levels of automation when the automated driving system is engaged; (4) it sets out any obligations on relevant entities, including the automated driving system entity, and users of automated vehicles; (5) it provides a regulatory framework with flexible compliance and enforcement options. Such type of regulations is missing in Europe in order to avoid compromising the good intentions of zero accidents with such technological development with an increase in road accidents resulting from missing regulations and public awareness, together with the need of a road environment compatible with that new reality.

6 The AUTODRING Project and the Research Methods and Tools Driving simulators are a powerful tool to support the research focused on driver behavior [28]. The advantages of using driving simulators in this type of researches are mostly the elimination of safety and ethical issues, avoiding unexpected events. Moreover, the well-controlled environment allows the design of scenarios and experimental conditions that cannot be easily implemented in real-world. In the context of automated driving, several studies were developed based on driving simulator experiments. Regarding the study of perception and intended use of automated vehicles, and despite most of the studies were based on surveys [29, 30], Buckley et al. [31] conducted a simulator experiment in which participants experienced periods of automated driving and manual control, followed by a survey task. Nevertheless, most of the studies underlining automated vehicles that use driving simulators, are focused on aspects such as takeover of vehicle control, secondary task engagement and workload [32]. These aspects are of most relevance since until the level of full vehicle automation is reached, users of vehicle automation systems will be required to takeover manual control of the vehicle occasionally and stay fallback-ready to some extent during the drive [33]. Considering that: (1) automated driving is not yet enough disseminated or experienced by common and professional drivers in nowadays societies, particularly in Portugal; (2) and the industrial technologic development is fast and requires deep

The User and the Automated Driving: A State-of-the-Art

199

research on the use of such technology allowing for road safety improvement, trust on the technology, appropriate regulations and prevention of misuse, are addressed in this new National funded project – AUTODRIVING – together with its main purpose to contribute to the study of the driver’s activity and behavior during the autonomous driving, addressing: (1) The driver’s promptness to resume the vehicle control following a long period of autonomous driving; (2) the research of the takeover of vehicle control task under different circumstances, which is expected to be the riskiest driving task in autonomous vehicles; and (3) the identification of the driver’s understanding of the system functioning, which will allow for the formation of trust on automation, both required for a later safe behavioral adaptation. These objectives will contribute to improve knowledge about the driver’s level of promptness to switch the vehicle control levels between the driver and the system and the influence of driver characteristics (e.g. age, sex, health state, risk perception) on his/her driver behavior and performance. This knowledge will support the development of advanced driver assistance systems (e.g. the notification interval to the takeover), which are foreseen to be tailored to the driver characteristics, by automotive and software industry and R&D agents in the field of information systems and vehicle automation.

7 Final Remarks and Next Steps The state of the art on driving automation highlights the needs for research on Human Factors and driver’s activity and behavior in the context of driving a vehicle with different levels of automation. Thus, in the frame of the AUTODRING Project, it is being prepared a National Survey addressing the following issues: (1) drivers’ preferences for the automation levels across gender, age, education level, user group, etc.; (2) the perceived limits of the technology; (3) perceived needs for changes in traffic laws, as well as licensing and training. Following the survey, tests with users from different groups on a Driving Simulator will be carried out. These tests will integrate appropriate scenarios for the research purposes. The research focuses on Level 3 and Level 4 of the five automation levels defined by the SAE. Considering that the driving behavior is the visible output of an internal activity, the experiments to be carried out will explain the driver’s behavior through the analysis of the driver’s activity using complementary methods and tools. To support the experimental design, a novel approach is proposed to be applied based on a taxonomy developed by Save and Feuerberg [34] in the context of automation in air traffic management (ATM): Level of Automation Taxonomy. The taxonomy is organized according to the functions defined by Parasuraman et al. [35] and Endsley [36]. These functions are based on a four-stage model of human information processing translated into equivalent system functions: information acquisition, information analysis, decision and action selection, and action implementation. Although being developed with ATM automation in mind, this taxonomy has been considered applicable to other contexts under automation processes as it represents a human-centered approach to automation based on the definition of generic human functions that can provide an initial categorization for types of tasks in which

200

A. Simões et al.

automation can support the human. This taxonomy is useful to guide the analysis of the driver’s activity, which is decomposed for each of the four functions in driver and system activities. Acknowledgments. This research was developed under Project No. POCI-01-0145-FEDER02852, co-financed by COMPETE 2020, Portugal 2020 and the European Union through the ERDF, and by FCT through national funds.

References 1. SKYbrary, I.F.S. Cockpit Automation - Advantages and Safety Challenges. https://www. skybrary.aero/index.php?title=Cockpit_Automatio_-_Advantages_and_Safety_Challenges&oldid=133844 2. U.S. DOT. Preparing for the Future of Transportation: Automated Vehicles 3.0. U.S. https:// www.transportation.gov/av/3 3. Bagloee, S., Tavana, M., Asadi, M., Oliver, T.: Autonomous vehicles: challenges, opportunities, and future implications for transportation policies. J. Mod. Transp. 24(4), 284–303 (2016) 4. Färber, B.: Communication and communication problems between autonomous vehicles and human drivers. In: Autonomous Driving, pp. 125–144. Springer, Heidelberg (2016) 5. Khan, A.: Autonomous Vehicles: Reliability of Their Perception of the World Around Them and the Role of Human Driver. Springer, Cham (2018) 6. Trommer, S., Kolarova, K., Fraedrich, E., Kroger, L., Kickhofer, B., Kuhnimhof, T., Lenz, B.: Autonomous driving-the impact of vehicle automation on mobility behaviour (2016) 7. Kim, T.J.: Automated autonomous vehicles: prospects and impacts on society. J. Transp. Technol. 8(03), 137 (2018) 8. Degryse, C.: Digitalisation of the economy and its impact on labour markets. ETUI Research Paper-Working Paper (2016) 9. Spencer, D.A.: Fear and hope in an age of mass automation: debating the future of work. New Technol. Work Employ. 33(1), 1–12 (2018) 10. Norman, D.A.: The human side of automation. In: Road Vehicle Automation, vol. 2, pp. 73– 79. Springer, Heidelberg (2015) 11. Leduc, S., Ponge, L.: La Evolución Digital y los cambios organizativos: Qué respuestas de la Ergonomia? Laboreal. 14(2), 31–44 (2018) 12. Bastien, J.C.: Usability testing: a review of some methodological and technical aspects of the method. Int. J. Med. Inform. 79(4), e18–e23 (2010) 13. Wachenfeld, W., Winner, H.: Do autonomous vehicles learn? In: Autonomous Driving, pp. 451–471. Springer, Heidelberg (2016) 14. Hussain, R., Zeadally, S.: Autonomous cars: research results, issues and future challenges. IEEE Commun. Surv. Tutorials 50, 1–37 (2018) 15. Zhao, J., Liang, B., Chen, Q.: The key technology toward the self-driving car. Int. J. Intell. Unmanned Syst. 6(1), 2–20 (2018) 16. Anderson, J., Kalra, N., Stanley, K., Sorensen, P., Samaras, C., Oluwatola, O.: Autonomous Vehicle Technology: A Guide for Policymakers. RAND Corporation, Santa Monica (2016) 17. Will, D., Gronerth, P., von Bargen, S., Levrin, F., Larini, G.: Report on the state of the art of connected and automated driving in Europe. (2017). https://connectedautomateddriving.eu/ publication/scout-deliverable-3-2-report-on-the-state-of-the-art-of-connected-andautomated-driving-in-europe-final/

The User and the Automated Driving: A State-of-the-Art

201

18. Cutean, A.: Autonomous vehicles and the future of work in Canada. Information and Communications Technology Council, Ottawa (2017) 19. Endsley, M.R.: Situation Awareness in Future Autonomous Vehicles: Beware of the Unexpected. Springer, Cham (2019) 20. Pearl, J., Mackenzie, D.: The Book of Why: The New Science of Cause and Effect. Basic Books (2018) 21. Endsley, M.R.: Autonomous driving systems: a preliminary naturalistic study of the tesla model S. J. Cognit. Eng. Decis. Making 11(3), 225–238 (2017) 22. Schoettle, B., Sivak, M.: A Survey of Public Opinion about Autonomous and Self-Driving Vehicles in the U.S., the U.K., and Australia. The University of Michigan, Transportation Research Institute, Michigan 48109–2150 U.S.A (2014) 23. Kyriakidis, M., de Winter, J.C.F., Stanton, N., Bellet, T., van Arem, B., Brookhuis, K., Martens, M.H., Bengler, K., Andersson, J., Merat, N., Reed, N., Flament, M., Hagenzieker, M., Happee, R.: A human factors perspective on automated driving. Theoret. Issues Ergon. Sci. 20(3), 223–249 (2017) 24. Inagaki, T., Itoh, M.: Human’s overtrust in and overreliance on advanced driver assistance systems: a theoretical framework. Int. J. Veh. Technol. 2013, 8 (2013) 25. Parasuraman, R., Manzey, D.H.: Complacency and bias in human use of automation: an attentional integration. Hum. Factors 52(3), 381–410 (2010) 26. Parasuraman, R., Sheridan, T.B., Wickens, C.D.: Situation awareness, mental workload, and trust in automation: viable, empirically supported cognitive engineering constructs. J. Cognit. Eng. Decis. Making 2(2), 140–160 (2008) 27. National Transport Commission. Changing driving laws to support automated vehicles. Policy Paper. NTC, Melbourne. VIC 3000 (2018) 28. Boyle, L., Lee, J.: Using driving simulators to assess driving safety (Prologue to special issue). Accid. Anal. Prev. 42, 785–787 (2010) 29. Kyriakidis, M., Happee, R., de Winter, J.C.F.: Public opinion on automated driving: results of an international questionnaire among 5000 respondents. Transp. Res. Part F: Traffic Psychol. Behav. 32, 127–140 (2015) 30. Madigan, R., Louwa, T., Wilbrink, M., Schieben, A., Merata, N.: What influences the decision to use automated public transport? using UTAUT to understand public acceptance of automated road transport systems. Transp. Res. Part F: Traffic Psychol. Behav. 50, 55–64 (2017) 31. Buckley, L., Kaye, S.A., Pradhan, A.K.: Psychosocial factors associated with intended use of automated vehicles: a simulated driving study. Accid. Anal. Prev. 115, 202–208 (2018) 32. Clark, H., Feng, J.: Age differences in the takeover of vehicle control and engagement in non-driving-related activities in simulated driving with conditional automation. Accid. Anal. Prev. 106, 468–479 (2017) 33. Naujoks, F., Höfling, S., Purucker, C., Zeeb, K.: From partial and high automation to manual driving: relationship between non-driving related tasks, drowsiness and take-over performance. Accid. Anal. Prev. 121, 28–42 (2018) 34. Save, L., Feuerberg, B., Avia, E.: Designing human-automation interaction: a new level of automation taxonomy. In: Proceedings Human Factors of Systems and Technology (2012) 35. Parasuraman, R., Sheridan, T.B., Wickens, C.D.: A model for types and levels of human interaction with automation. IEEE Trans. Syst. Man Cybern.-Part A: Syst. Hum. 30(3), 286– 297 (2000) 36. Endsley, M.R.: Level of automation effects on performance, situation awareness and workload in a dynamic control task. Ergonomics 42(3), 462–492 (1999)

Driver Training and Education

Explicit Forward Glance Duration Hidden Markov Model for Inference of Spillover Detection John (Hyoshin) Park1(&), Nigel Pugh1, Justice Darko1, Larkin Folsom1, and Siby Samuel2 1

North Carolina A&T State University, Greensboro, USA [email protected], {nrpugh,jdarko,ldfolsom}@aggies.ncat.edu 2 University of Waterloo, Waterloo, Canada [email protected]

Abstract. To better understand the effects of distracted driving on crash causation, forward roadway glance durations need to be carefully examined. Secondary tasks that impose high cognitive load lead to spillover effects that are moderated by the duration of the forward roadway glance within an alternation sequence involving both, in-vehicle and on-road glances. Spillover effects diminish the hazard anticipation ability of drivers. When alternating glances in a time series, the probability of detecting a spillover is invisible and the hidden state depends on the amount of time that has elapsed since the secondary task was initiated in the current state which is in contrast with the hidden Markov theory, where there is a constant probability of changing state given spillover detection in the state up to that time. No research estimates the probability of spillover detection in a time series with an explicit glance duration. In the current effort, we apply a semi-hidden Markov model where secondary task severity is used as an observation to infer hidden state and relax the assumption of constant state duration. Based on the reliable accuracy of the task itself, and the proposed model, different sequences of secondary task during various time window were tested for spillover detection. With a threshold of 50%, different forward roadway glance durations are required in each sequence associated with different types of secondary tasks. Keywords: Spillover detection
Semi-hidden Markov Explicit glance
Secondary task


Time series


1 Introduction The rapid development in communication technology has induced distracting behaviors that may increase the potential near crash or crash risk due to limited attention while driving. Along with many educational outreach projects [1], the technology advancement in driving simulator and naturalistic driving studies have helped us understanding risks associated with different inattention-related activities [2]. Compared to traditional studies (e.g., [3]) that had focused on different lengths of glance durations away from © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 205–213, 2020. https://doi.org/10.1007/978-3-030-20503-4_18

206

J. (Hyoshin) Park et al.

the roadway, [4] analyzed glance durations on the roadway to estimate the probability of risk anticipation. With simulated, in-vehicle tasks limited to 2s, the glance outside was varied between 1s and 4s. In their study, when a driver is engaged in a low-load, in-vehicle task, at least 7s of minimum glance duration on the roadway was required to detect a hazard. Despite significant efforts on analyzing the impact of off-road or on-road glances on discrete risky events, the literature lacks a time-series framework to explain a sequence of several risky events. Recently, [5] proposed a modeling approach to estimate parameters of the latent hazard anticipation probability between discrete time stages conditioned on glance duration from a driving simulator experiment. The detection probability in the next stage could be predicted under various sequences of forward durations that was not practically possible to obtain from a controlled, experimental setup. Hidden Markov models were used to capture the invisible detection states inferred by observation of processing type, location of threat, and cognitive difficulty. The sojourn times in each stage is assumed to be geometrically distributed. In this study, we explicitly model the forward duration to predict the time to spillover detection in a time series. In contrast to the hidden Markov models [5] that treat the forward duration as one of the observation categories, this paper considers the duration spent in a state before transition to the next state, that can be called an explicit duration (semi) hidden Markov model (HSMM). The main benefit of treating the duration of each state as a variable rather than a geometrical distribution, is the twostage driving risk assessment involved in different factors such as, information processing type, location of threat, difficulty of secondary tasks within a sequence. Various probabilities of spillover anticipation on time series are presented with associated distraction conditions.

2 Data Methodology 2.1

Participants

In the current experiment, the driving performance of 40 young drivers between the ages of 18–21, with an average age of 20.1 (SD = 0.819) and average driving experience of 2.8 years (SD = 0.351), was evaluated on a driving simulator. The participants were remunerated for their time and were recruited from the University of Massachusetts Amherst and surrounding areas. The study had IRB approval. 2.2

Equipment and Apparatus

We used a full cab, fixed base Realtime Technologies Inc. simulator, placed in front of three screens subtending 135° horizontally. The virtual environment is projected on each screen (1024  768 pixels, 60 Hz). The participant sits in the car and operates the controls, just like he or she would in a normal car. The system provides realistic environment sounds with appropriate direction, intensity and Doppler Shift. A portable lightweight eye tracker (Mobile Eye developed by Applied Science Labs) was used to collect the eye-movement data for each driver. It consists of a pair of

Explicit Forward Glance Duration Hidden Markov Model

207

goggles including several cameras that allow the eye camera to record an image of the eye. The images from these cameras are then interleaved, and converted into a crosshair, representing the driver’s point of gaze, which is superimposed upon the scene recorded during the drive. The eye and scene information are captured at 60 Hz in a single trial and eye tracker has an accuracy of 0.5° of visual angle. 2.3

Scenarios

The virtual driving scenarios were designed to vary the type of processing (top down or bottom up) and the location of the threat (central or peripheral). There are thus four combinations: central/top down, central/bottom up, peripheral/top down, and peripheral/bottom up. There was a total of 8 scenarios, two each for the four combinations. The scenarios are discussed, described and illustrated in detail in [4]. 2.4

Procedure

All participants provided written consent to participate in the experiment as per the Institutional Review Board norms. Participants then completed a demographic questionnaire that collects participants’ driving history and some demographic information like age, following which the participants were outfitted with an eye tracker which was calibrated within the simulator. After calibration, participants were given a practice drive to familiarize them with the driving simulator and its controls. Subsequent instructions were provided to participants at the onset of each experimental drive sequence. 2.5

Experimental Design

In a between-subject design, all participants navigated the 8 driving scenarios once, on a driving simulator. Participants were assigned either to the control, alternating baseline, low load or high load conditions. The drivers’ view of the forward roadway was always visible for the control condition. For the alternating low/high load groups, the driver performed a secondary task. The driver’s view of the forward roadway on the screens outside the window of the cabin of the simulator was alternated (either 1s, 2s or 4s), with a view of the secondary task. When the view of the forward roadway was displayed, the secondary task was not visible. Conversely, when the view of the secondary task was displayed, the view of the forward roadway was not visible. The alternation baseline drives involved a similar switching between views of a black screen and the forward roadway, with a fixation task however, where the participants were asked to fixate on a ‘+’ sign that appears on the black screen during the periodic alternations. The location of the ‘+’ target changed to ensure that drivers had a reason to glance at the blank screen. The ordering of the scenarios was counterbalanced across all participants. The eye tracker was used to record eye behaviors including fixation and glance data from the participants. Glance durations on the forward roadway and within the vehicle were collected over the course of the 8 scenarios. Fixation locations were scored 0 or 1 to determine participants’ ability to anticipate potential hazards when performing

208

J. (Hyoshin) Park et al.

secondary tasks. The participants were scored 1 only if they glanced at the target zone while within the appropriate launch zone. Target zones are the threat locations where participants should scan, and the launch zone is defined as that area of the roadway from whence to begin scanning [4]. 2.6

Secondary Task

There are two secondary tasks (low load and high load) used in the experiment. The low load task was chosen with the criterion that when the forward roadway is not visible, it (the task) places no load on the driver during the period in which the task is visible. The high load task, however, places a load on the driver even when he or she is performing the primary task of driving. We used a chosen a cognitive arithmetic task that adds to the load of the low load secondary task (Fig. 1). In both tasks, drivers search the number of times, the letter ‘t’ appears on the visual search display and report the count (e.g. 0, 1, 2, 3, 4, 5, 6). In the high load condition however, after the number of targets has been reported the driver needs to count forward by 3 until the next visual search display appears. For example, if the target count is 3, then the participant would verbally count aloud 3, 6, 9, until the next search display appears where the count may be a 2 leading to a verbal count of 2, 5, 8, 11… and so on. In this task, participants are loaded even when they look at the forward since they need to mentally compute arithmetic sums.

Fig. 1. Secondary tasks

Explicit Forward Glance Duration Hidden Markov Model

209

3 Modeling Framework The driving simulator experiment provides us observations of duration and type of secondary task to estimate the hidden transition probability between stages. 3.1

Hidden Markov Theory

A Markov chain consists of discrete stage k represented by a state Sk that is equal to 1 if a participant successfully detects a spillover, and 0 otherwise. With a sequence of process, for any s0 ; s1 ; . . .; st þ 1 , conditional probability has the Markov property PðSt þ 1 ¼ st þ 1 jSt ¼ st ; St1 ¼ st1 ; . . .; S0 ¼ s0 Þ ¼ PðSt þ 1 ¼ st þ 1 jSt ¼ st Þ

ð1Þ

With a known probability distribution pi of the initial state i, the transition matrix Aðk Þ contains k-step transition probabilities of detection, pij ðkÞ ¼ PðSt þ k ¼ jjSt ¼ iÞ, between consecutive stages: from state i at stage k to state j at stage k þ 1. Let sojourn time be the number of time steps (seconds) participants spent looking at forward in a state, then we have the sojourn density di ðgÞ following a geometrical distribution.

  di ðgÞ ¼ P St þ g þ 1 6¼ i
St þ u ¼ i; St þ g1 ¼ i; . . .; St þ 2 ¼ i
St þ 1 ¼ i; St 6¼ i ¼ pg1 ij ð1  pi Þ

ð2Þ

In a stochastic process, states in the intermediate stages are hidden (detect or not detect) within an alternation sequence, inside and outside of the vehicle depending on the current state. We can only observe a sequence of Y observations O ¼ o1 ; o2 ; . . .; oY ([6]. We can derive B ¼ bi ðot Þ, the emission probability of observation ot , generated from a state i. The hidden Markov model determines the likelihood of spillover detection using task severity as an observation to infer hidden state. 3.2

Hidden-Semi Markov Model

Since the probability of a state change depends on the time spent in the current state, we explicitly model the sojourn density d ðgÞ instead of using P to define d ðgÞ in hidden Markov mode. The set of parameters in the model is denoted by k ¼ ðp; A; B; d Þ. Let se0 ; se1 ; . . .; seR , denote the R þ 1 states visited [7], the contribution of the state sequence to the complete-data likelihood is given as following: PðO ¼ o; S ¼ s; kÞ ¼ ps1 ds1 ðg1 Þ

nYR

o YT  p d ð g Þ psR1 sR DsR ðgR Þ t¼2 bst ðot Þ   r r¼2 sr1 sr sr ð3Þ

To predict the likelihood of spillover detection in the next stage, we need to estimate parameters in incomplete data. For maximum likelihood estimation, the expectation maximization algorithm [8] is used. In the expectation step, we estimate the probability of being in state i at time t given the observed sequence, the probability that the process left state i at time t and entered state j at t þ 1 given the observed sequence,

210

J. (Hyoshin) Park et al.

and the expected number of times a process spends g time steps in state j. In the maximization step, we estimate the initial transition probabilities, the emission distribution, and the state duration density.

4 Results In this section, we interpret the result of the stochastic process in spillover detection with two potential sequences of alternating tasks inside and outside of vehicle. First, each experiment was set to have an identical secondary task during the fixed time window (w = 15s). For example, baseline, secondary in-vehicle tasks with an off-road duration of 2s involving a forward glance duration of 4s in an alternation sequence with two alternations back and forth (on road and off road) are represented as ½2s  BS4 2s  BS4. Second, with estimated parameters, we present scenarios with different secondary tasks during the various time windows. Statistical analyses using logistics regression on the binary-coded, binomially distributed glance data yielded highly significant effects for alternation sequence (Wald x21 ¼ 9:468; p = 0.009) and the interaction between location of threat and both task accuracy and task attempted respectively. (Wald x21 ¼ 4:712; p = 0.030), (Wald x21 ¼ 4:459; p = 0.035). 4.1

Identical Secondary Tasks with Fixed Time Window

With a fixed time window, three explicit glance duration models are trained using 240 samples (75%), and is tested with 80 samples (25%): HSMM1, type of processing (Top-down, Bottom-up); HSMM2, location of spillover (Peripheral, Central); HSMM3, difficulty of tasks (Base, Low-load, High-load). We set the maximum iteration to be 100, and once the transition, emission, duration matrices converge to optimal solution, the iteration was terminated. Among 10 different runs with different initial random parameter values, the best model is chosen and used to calculate the different secondary tasks with various time windows. The performance measures of the models are reported by K-fold cross validation (i.e., k = 4). The original sample is randomly partitioned into four equal sized subsamples across the different secondary tasks: For each HSMM1*3, average F-values were respectively 0.91, 0.92, and 0.94, as the harmonic mean of the probability of correctly labeling the detection (recall) and probability that a positive prediction is correct (precision), and the balanced trade-off between recall and precision. Similar to the work on driver state prediction (distracted or not) by [9] Schwarz et al. (2016), the proposed model in this study presents comparably good performance. Average performance of the model was respectively 87%, 85%, and 91%. This is within 10% from other modeling efforts (mean accuracy of 95%) in a driver behavior capture situation involving different variables in similar domains [10]. Furthermore, by having three observations in the HSMM3, the spillover detection can be conducted more consistently.

Explicit Forward Glance Duration Hidden Markov Model

4.2

211

Different Secondary Task with Various Time Window

Time window is varied on the last two stages of spillover detection because a long history of forward glance duration is relatively less relevant [11]. For example, 2s of different tasks with baseline and high-load swapping followed by a forward glance (2s or 4s) is presented as ½2s  BS2 2s  HL4. Even though an alternation sequence starts with a lower detection likelihood, less demanding secondary task with at least 2s forward roadway duration leads to spillover likelihood higher than 50%. Simulation experiments are not feasible to conduct extensive analyses using all possible combinations and permutations of scenarios.

5 Discussion The explicit forward glance duration modeling in this paper allows us to extend a previously developed hidden Markov model with different secondary tasks within a time window, both fixed and variable. Figure 2 describes various glance duration scenarios with different secondary tasks. For example, in the threat location case, after 3 s of the first alternation, the driver glances to the left and right of the crossroad to determine if any spillover risk exists, measured by drivers’ anticipation ability. The driver’s attention level starts to decrease when the driver looks away from the forward roadway and starts increasing after the driver returns his/her glance towards the forward roadway. While the fixed window with identical tasks [2s – BU1 2s – BU2] increased the driver’s spillover detection likelihood (62%), the fixed window with different tasks [2s – BU1 2s – TD2], decreased the anticipation likelihood (57%). With the same glance duration, [2s – H1 2s – H1] and [2s – H1 2s – B1] presents highest difference in spillover likelihoods while [2s – C1 2s – C1] and [2s – C1 2s – P1] present almost no difference. Even though stage 1 starts with [2s – H1] or [2s – P1], baseline load of secondary task, or scenarios with centrally appearing threats with at least 2s of forward glance lead to more likely detection spillover (>50%) 5.1

Summary

There has been no research conducted to estimate the temporal probability of spillover detection with an explicit glance duration. The proposed model provides a potential way to observe patterns and trends across combinatorial factorial sequences thereby allowing for the development and evaluation of countermeasures such as, where a warning is triggered when low (e.g., a threshold less than 50%) spillover detection probability is observed. Controlled, simulator studies may be designed and conducted to validate the scenarios inferred by the HSMMs. The proposed framework can be applied to transfer of control behaviors in automation systems that promote distraction [12]. Considering that it will be several years before fully autonomous vehicles wind up on the open road, the current approach detailed here allows us to generate a set of human inputs both, for the near and immediate long-term. If humans are going to partially monitor the roadway, and emergency instances occur where they need to

212

J. (Hyoshin) Park et al.

Fig. 2. The predicted likelihood of spillover detection with different secondary tasks and time windows

respond to the automation and situations on the roadway, a lack of situation awareness will cause immense issues that are critical to safety. Until automation becomes ‘full’, there will be transfer of control related issues that cause poor perception and response to threats from the driver. Use of HSMMs allow for the understanding of the effects of complex cognitive load on driver behavior in emergency and non-emergency situations in an automated, connected roadway system. Acknowledgments. This paper is partially supported by start-up fund provided by North Carolina A&T State University, USDOT University Transportation Centers Contract 69A3551747125, VDOT Project 114591, NCDOT Project 2019-09, and NASA JPL Project 2019.

References 1. David, S.H., Miller, E., Jannat, M., Boyle, L.N., Brown, S., Abdel-Rahim, A., Wang, H.: Improving teenage driver perceptions regarding the impact of distracted driving in the Pacific Northwest. J. Transp. Saf. Secur. 8(2), 148–163 (2016) 2. Klauer, S., Guo, F., Sudweeks, J., Dingus, T.A.: An analysis of driver inattention using a casecrossover approach on 100-Car Data: Final Report. DOT HS 811 334. National Highway Traffic Safety Administration, US Department of Transportation, Washington, DC (2010)

Explicit Forward Glance Duration Hidden Markov Model

213

3. Horrey, W.J., Wickens, C.D.: In-vehicle glance duration: distributions, tails and model of crash risk. Transp. Res. Rec. J. Transp. Res. Board 22–28 (2007). No. 2018, Transportation Research Board of the National Academies, Washington, D.C. 4. Samuel, S., Fisher, D.L.: Evaluation of the minimum forward roadway glance duration. Transp. Res. Rec. J. Transp. Res. Board 9–17 (2015). No. 2518, Transportation Research Board, Washington, D.C. 5. Park, H., Gao, S., Samuel, S.: Modelling effects of forward glance durations on latent hazard detection. Transp. Res. Rec. J. Transp. Res. Board 2663, 90–98 (2017) 6. MacDonald, I., Zucchini, W.: Hidden Markov and Other Models for Discrete-Valued Time Series, Monographs on Statistics and Applied Probability. Chapman and Hall, London (1997) 7. Bulla, J.: Application of hidden Markov models and hidden semi-Markov models to financial time series, Dissertation. Technical report (2006) 8. Guedon, Y.: Estimating hidden semi-Markov chains from discrete sequences. J. Comput. Graph. Stat. 12(3), 604–639 (2003) 9. Schwarz, C., Brown, T., Lee, J., Gaspar, J.: The detection of visual distraction using vehicle and driver-based sensors. SAE Technical Paper 2016-01-0114, pp. 229–242 (2016) 10. Pentland, A., Liu, A.: Modeling and prediction of human behavior. Neural Comput. 11, 229– 242 (1999) 11. Liang, Y., Lee, J.D., Yekhshatyan, L.: How dangerous is looking away from the road? Algorithms predict crash risk from glance patterns in naturalistic driving. Hum. Factors J. Hum. Factors Ergon. Soc. 54(6), 1104–1116 (2012) 12. Samuel, S., Borowsky, A., Zilberstein, S., Fisher, D.L.: Minimum time to situation awareness in scenarios involving transfer of control from the automation. Transp. Res. Rec. J. Transp. Res. Board (2016, in press). Transportation Research Board, Washington, D.C. https://doi.org/10.3141/2602-14

Proposal for Graduated Driver Licensing Program: Age vs. Experience, Abu Dhabi Case Study Yousif Al Thabahi(&), Marzouq Al Zaabi, Mohammed Al Eisaei, and Abdulla Al Ghafli Traffic Engineering and Road Safety Department, Abu Dhabi Police GHQ, Abu Dhabi, United Arab Emirates [email protected], [email protected]

Abstract. This paper aims to present a new approach to the GDL system to be applied in Abu Dhabi. On contrary to existing GDL systems, the proposed system will feature driving experience as the main factor to new drivers in the emirate; which will determine their eligibility to attain a full driving license. The paper will present descriptive statistical analysis which helped in concluding that drivers experience has a higher significant relation to the probability of being involved in accidents, when compared Keywords: Graduated driver licensing Driving experience


GDL
Age group


1 Introduction Graduated Driver Licensing (GDL) is a driver licensing system which aims to provide drivers with licenses through progressing stages, with each stage having its own time periods and restrictions. The GDL comprises of three stages [1]. These are the following: • Supervised learning period. This stage includes restrictions in the form of adult supervision for a set number of hours. (Average ages: 16–17) • Intermediate stage. This stage has fewer restrictions. Such as number of passengers, no cell phone use, and a curfew. (Average ages: 18–21) • Final graduation. This is the final stage and has no restrictions. (Over 21 years) Since several countries use the GDL system; certain variances apply especially in the case of the United States of America, which has different rules and laws specific to each state. The age range can also vary depending on the country [2]. Traffic accidents and its resulting deaths and injuries are considered the most challenging obstacles that face the Emirate of Abu Dhabi. Despite the continuous decline in the number of accidents and fatalities, these numbers are still high compared to developed countries and resulted in the growing demand for a sustainable transportation system in the Emirate of Abu Dhabi that will improve road safety, involve safety related issues and analyze safety aspects by focusing on Abu Dhabi’s drivers’ behaviors, especially new © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 214–223, 2020. https://doi.org/10.1007/978-3-030-20503-4_19

Proposal for Graduated Driver Licensing Program

215

drivers. Graduated driver licensing (GDL) is a system designed to provide newlylicensed drivers with driving experience and skills gradually over time in low-risk environments to protect them during their learning period. GDL allows newly-licensed drivers to gain experience incrementally through close supervision in the initial phases of driving, and limited exposure to risky driving situations such as driving late at night and carrying young passengers. Drivers go through two initial phases of driving before gaining full driving privileges. The first phase of GDL is a learner’s period of several months, during which only supervised training and minimum numbers of practice hours are required. This is followed by an intermediate period where unsupervised driving is limited in higher risk situations. These situations have been restricted to night time driving and driving with young passengers. In addition to that, traffic violations and penalties are higher when it comes to drivers who are in their intermediate period of attaining their driver license. After Reviewing the GDL systems from different countries which are available in the literature review of this paper. It was found that the restrictions of these systems are more rigorous for drivers who are between the ages of 16 to 21. However, throughout this study, it was noted after analyzing the data of drivers who caused traffic accidents in Abu Dhabi during the period (2013–2017); that most drivers causing serious traffic accidents are “newly-licensed drivers”. Therefore, the results presented in this paper will demonstrate that serious traffic accidents are more related to the driver’s experience than the driver’s age.

2 Literature Review 2.1

GDL on Road Safety

According to a previous research, GDL has an overall positive impact on road safety. Benefits have been measured through the GDL if it addresses: fairness, effectiveness, reaction, and its implementation [2, 3]. Furthermore, based on historic crash data; accidents were less likely with higher age groups. Based on the research, it showed that higher ages applying for licenses have had a positive correlation in terms of road safety. Also, according to [4] when GDL has taken a non-traditional approach it can increase its benefits thus its effectiveness. Examples of such non-traditional approaches are: educational courses follow up practice, exams and the collection of data [4]. In particular, the State of Nevada initiated this nontraditional approach and it showed an increase in benefits and an increase in road safety. Paz showed that other distractions such as texting and calling still exist, as well as driving while being intoxicated. However, it’s not as frequent as before with younger age groups [4]. Recent evidence has shown that hazard perception training and situational awareness development using tablets can have significant short and long-term benefits [5]. Overall, it can be argued that the GDL has had overwhelmingly positive impacts on road safety where used, and as it is an overall system of training with numerous elements, combining its basic functions with novel educational mediums can multiply its positive impacts.

216

2.2

Y. Al Thabahi et al.

GDL for Novice Drivers and Its Impacts

According to UK statistics, 25% of accidents are caused by people aged between 15 and 19 years old [6]. This is alarming considering the fact that total UK road accidents are only at 5% per annum. A feasibility study conducted in the UK showed that the GDL system reduces the number of accidents among 17 to 19 years old can save 4471 lives and also save 224 million pounds in damage claims per year [7]. This is especially true for males. Applying a zero-tolerance policy, which is the suspension of a permit during the intermediate stage has had a massive increase in the positive impact of GDL. Applying GDL basics with educational, and follow ups is the reason for this positive result [8]. 2.3

Advantages and Disadvantages of GDL on Novice Drivers

There are several advantages and disadvantages of GDL on novice drivers. For instance, people living in rural areas are more likely to be dependent on adults as compared to people living in developed urban areas. This reduces the independence of teens and they must rely on their elders to get around. Whereas in developed areas where GDL exists, the teens have more freedom and can get around easier [9]. Also, in more populated areas there are many options such as riding, sharing and other modes of transport also exist. It should be noted that the GDL has less restrictions if it’s being used for commuting to work or school. However, due to less restriction; studies show that the effectiveness of GDL is reduced when restrictions are removed. It is also important to note that restriction on mobility has not shown GDL effectiveness. The reason for having curfew restrictions such as night time driving are due to drivers aged 15 to 20; result in fatal crashes and these restrictions are in place since most drink driving crashes occur at night time [7]. Thus, if restrictions exist, young drivers will not be allowed to drive past their curfew, which in turn shall reduce the number of serious accidents caused by younger drivers [10]. It should be noted that the research presented evidence of the negative social impacts that curfew restrictions will impose on the daily lives of young drivers. It is also important to note that there are several other psychosocial elements involved in the success of GDL on novice drivers. These are accurate learners log entries, level of supervision, their age, parental learning background, also if the driver is in a relationship [11]. Also thrill seeking; peer pressure, depression, and anxiety also play a role. Another factor is that of punishment if GDL rules are broken. These are all factors that can impact the positive results of GDL on novice drivers [12, 13] Therefore, it is imperative that the GDL must incorporate the above elements to ensure that GDL results are positive and its negatives are reduced to a minimal. An important factor in determining the success of a GDL program is having a zerotolerance policy when it comes to applying stricter penalties and fines associated to each GDL level. This also shows that when GDL is modeled with these it increases the psychosocial impact and reduces the disadvantages linked to development related traits within individuals. Overall GDL is effective but must be modeled and tweaked according to each specific region when implemented. This is the only way to keep relevant, as well since

Proposal for Graduated Driver Licensing Program

217

driving dynamics are constantly changing as are the novice drivers or learners who take GDL. In addition, modelling the GDL for a certain group of people based on their age (young drivers) won’t be efficient as their driving experience which has been found through the accident and severity analysis in this paper.

3 Problem Description Based on the literature review and historic crash data, statistical analysis of the data has been done to find out the relation of driving experience and the number of accidents. Driving experience and GDL are related because both are based on the experience of driving. It is important to mention that the driving experience mentioned in the below analysis is measured by counting the difference of days between the license issue date and the date of the accident.

4 Data Analysis The analysis below is derived from data after a process of data cleaning. The data cleaning process has eliminated around 2% of the original data due to missing important information to the research. However, since the data eliminated is less than 2% of the whole source, there will be minimal effect on the outcomes. The data has been taken from the “Unified Traffic System” which Abu Dhabi Police uses when inputting traffic incident information. The process of data entry might in rare occasions be affected by human error in the fields that are entered manually. However, it is not the case of most data used in this research since it is uploaded automatically from the Unified Traffic Number which gives the exact Age and License Issue Date. Based on the results of the data analysis, it shall support the claim of this research which states that drivers experience is statistically more significant to accident probability than the age of the driver. Data analysis was done on traffic accidents which occurred in the Emirate of Abu Dhabi only. The analysis of data focused on several factors that are important to look at which were the number of accidents and their intensity. Moreover, the analysis included analysis of age groups in order to study the dependency of age groups and driving experience. 4.1

Age Groups and Driving Experience

Figure 1 summarizes the issuances of the driving license per age group in the emirate of Abu Dhabi between 2013 and 2018. Remarkably, the percentage of age group from 18 to 24 years old is increasing as well as from 25 to 30 years old. However, the percentage of novice drivers who are above 24 years old, still form most drivers who apply for their driving licenses. This can be explained by the socioeconomic state of the United Arab Emirates, which attracts foreign laborers and expats. Furthermore, the GDL mechanism always focuses on young drivers which cannot be valuable factor to consider when applying for driver license in UAE.

218

Y. Al Thabahi et al.

Percentage of licenses Issued

Driving Licenses Issued per Age Group in the Emirate of Abu Dhabi 50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% 2013

2014

2015

2016

2017

from 18 to 24 yearsold

Age group from 25 to 30

from 31 to 40

from 41 to 50

from 51 to 60

from 61 to 70

2018

Fig. 1. Driving licenses issued per age group

4.2

Driving Experience and Crash Data

Figure 2 displays the percentage of accidents in respect to the driver’s experience. It indicates the comparison between drivers with less than 5 years of experience and more than 5 years of experience (red bar indicates 5 years of driving experience). As it shown in the figure, the drivers with less experience are more likely to be involved in an accident than the drivers with more driving experience. For instance, drivers within their first year of driving are likely to be involved in an accident twice the times of a driver who has been driving for 5 years.

Accidents Percentage

Accidents Percentage per Years of Driving Experience 15.00% 10.00% 5.00% 0.00% 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 Fig. 2. Accidents percentage per years of driving experience

Proposal for Graduated Driver Licensing Program

219

Number of Injuries & Fatalities

By going through the number of fatalities and injuries with respect to driving experience that is shown in Fig. 3, it illustrates that as driving experience increases, the number of injuries and fatalities decreases. Drivers with lesser experience are more likely to cause accident related injuries as shown in Fig. 3. On the other hand, people with greater driving experience caused lesser numbers of injuries and fatalities.

Number of Injuries and Fatalities per Driving Experience

1000 800 600 400 200 0 0

2

4

6

8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50

Year of Experience Number of Fatalities

Number of Medium Injuries

Number of Minor Injureis

Number of Sever

Fig. 3. Number of injuries and fatalities per driving experience

Figure 4 shows the percentage of the first 6 years of driving experience. There is a reduction of 6% in the number of accidents that have been caused by driver with low driving experience (0 to 6 years of driving experience). The continual decrease indicates that the number of accidents is contingent on the years of experience as the more years of experience the drivers have, less accidents to be involved in.

Percentage of Accidents

Percentage of Accidents per Years of Driving Experience (0-5 years) 14.00% 12.00% 10.00% 8.00% 6.00% 4.00% 2.00% 0.00% 0

1

2

3

4

5

Years of Experience

Fig. 4. Percentage of accidents per years of driving experience (first 5 years)

220

Y. Al Thabahi et al.

Percentage of Injuries & Fatalities

Figure 5 shows the percentage of fatally injured drivers by the years of experience. It indicates that there are higher numbers of fatalities and injuries among the drivers who have driving experience between (0–5) years. However, the drivers with more than 5 years of experience are less likely to be involved in sever accident. Overall, the higher years of experience in driving, the lesser being involved in an accident of severity in all the degrees of injuries.

30.00%

Percentage of Injuries and Fatalities per Years of Driving Experience

25.00% More than 5 Years of Driving Experience

20.00% 15.00% 10.00% 5.00% 0.00% Number of Fatalities

Number of Number of Medium Injuries Minor Injureis

Sever injuries

(0-5) Yearsof Driving Experience

Scale of Injury

Fig. 5. Percentage of injuries and fatalities per years of driving experience

Figure 6 shows the relation between the accident’s percentage and the driver’s age of group with driving experience less than 5 years. It illustrates a correlation between the numbers of accidents with respect to age group; however, this correlation is due to the less numbers of licenses issued in the groups of (41 to 50, 51 to 60, and 61 to 70). Also, it shows the higher number of accidents in all group ages and it is not strongly related to a specific age group, but more to the driving experience.

Fig. 6. Percentage of accidents vs age groups (0–5) years

Proposal for Graduated Driver Licensing Program

221

Figure 7 shows the relation between the accident involvement probability and the driver’s age group with driving experience more than 5 years. The weak relationship between the age group and driving experience with more than 5 years can be interpreted by the low R2 value of 0.0096. It indicated no correlation and the age factor has no affection on drivers being involved in a sever accidents if their driving experience exceeds the 5 years.

Percentage of Accidents

Percentage of Accidents for Drivers Have More Than 5 Years of Driving Experience 40.00% 30.00%

R² = 0.0096

20.00%

Percentage of Accidents for Driver HaveMore Than 5 Years of Experience

10.00% 0.00% from 18 to 24

from 25 to 30

from 31 to from 41 to 40 50 Age Group

from 51 to 60

from 61 to 70

Fig. 7. Percentage of more than 5 years of experience

4.3

Accident Intensity

Accident Intensity

Figure 8 shows the relation between the drivers experience and the intensity of accidents. The figure gives a clear presentation of the minimal effect that the drivers experience has on the severity of traffic accidents. The weak relationship between the drivers experience and accident severity, can be interpreted by the low R2 value of 0.0025. Figure 9 shows the relation between the driver’s age and the intensity of traffic accidents. The weak relationship between the drivers experience and accident severity, can be interpreted by the low R2 value of 0.0008. It was found that age has a minimal effect on the severity of traffic accidents. There was no correlation found between the accident intensity and the drivers experience or age. In other words, the driver’s age or experience has no relation to the severity of traffic accidents.

Accident Intensity vs Years of Driving Experience

60

R² = 0.0181

40

Accident Intensity

20 0 0

5

10 15 20 25 30 35 40 Years of Driving Years of Driving Experience

45

Fig. 8. Accident intensity vs driving experience

50

222

Y. Al Thabahi et al.

Accident Intensity

32

Accident Intensity vs Age Group

30 28 26

R² = 0.0008

24

Accident Intensity

22 from 18 from 25 from 31 from 41 from 51 from 61 More to 24 to 30 to 40 to 50 to 60 to 70 than 70 Age Group

Fig. 9. Accident intensity vs age group

5 Conclusion and Future Research The GDL is a system designed for young drivers to minimize the risks that would face them in the beginning of their driving experience that would result in causing them to be involved in severe accidents. However, this type of system is based on restrictions that are applied to specific age groups of drivers who belong to an age group of 17 to 21 years old (young drivers). Additionally, this system neglects other age groups of novice drivers which in return renders it inefficient in the UAE society according to the high percentage of driver licenses issued by people belong to those groups of novice drivers. Drivers that the GDL doesn’t include have a higher risk of being involved in an accident, as shown in the previous figures. The data analysis which was based on the accidents in the emirate of Abu Dhabi indicates the driving experience as a new factor to be considered while implementing the GDL system. According to the analysis, more than 60% of the accidents that has been caused by drivers who have less than 5 years’ experience in the period (2013–2017) belongs to the age group (25 years and above). As a result, the GDL program must address experience rather than age, due to the significant relationship between experience and the risk of being involved in an accident. In order to have the most effective results in Abu Dhabi and the UAE, a GDL program which addresses driving experience is deemed necessary. Furthermore, an interesting conclusion was found that accident intensity was not affected by neither age or experience. Moreover, the percentages of drivers being involved in serious accidents have reduced by approximately 50% in the first 5 years of their driving experience more than any other period. As a result, focusing on drivers in their first 5 years of driving experience to minimize the risks that may result in those drivers being involved in serious accidents. In order to impose the GDL program, it is important to have the system incorporated within the existing driving schools in the UAE. Further studies should be done on the quality of graduates of those schools, and the number and type of accidents they cause. In addition, it will be more accurate to impose a measurement of driving experience for each driver, which is not only based on the age of the driving license, but also on the traveled distance per vehicle per driver registered.

Proposal for Graduated Driver Licensing Program

223

References 1. RMIAA: Graduated Driver Licensing (GDL) (2015). http://www.rmiia.org/auto/teens/ Graduated_Drivers_Licensing.asp. Accessed 17 Dec 2018 2. Williams, A.F.: Graduated driver licensing (GDL) in the United States in 2016: 2Q2 A literature review and commentary. J. Saf. Res. (2017) 3. Foss, R.D.: An assessment of graduated driver licensing: pros & cons. Transp. Res. Circ. Issue 458, 44–48 (1996) 4. Paz, A., et al.: The effectiveness of driver education and information programs in the state of Nevada. Open J. Appl. Sci. 5(1) (2015) 5. Ahmadi, N., Katrahmani, A., Romoser, M.R.: Short and long-term transfer of training in a tablet-based teen driver hazard perception training program. In: Proceedings of the Human Factors and Ergonomics Society Annual Meeting, vol. 62, no. 1, pp. 1965–1969 (2018) 6. Miller, B.: Views on Graduated Driver Licensing (2014). https://www.racfoundation.org/ assets/rac_foundation/content/downloadables/views_on_graduated_licensing_ben_miller_ 020714.pdf. Accessed 17 Dec 2018 7. Department of Transport: Novice drivers: Evidence review and Evaluation Pre-driver training, Graduated Driver Licensing (2013). https://assets.publishing.service.gov.uk/ government/uploads/system/uploads/attachment_data/file/249282/novice-driver-researchfindings.pdf. Accessed 20 Dec 2018 8. McKnight, A.J., Peck, R.C.: Graduated driver licensing: what works? Inj. Prev. 8, 32–38 (2002) 9. Williams, A.F., Mccartt, A.T., Mayhew, D.R., Watson, B.: Licensing age issues: deliberations from a workshop devoted to this topic. Traffic Inj. Prev. 14(3), 237 (2012) 10. AAAM: Graduated Drivers Licensing Systems for Novice Drivers (2017). https://www. aaam.org/education-resource-center/public-policy/graduated-drivers-licensing-systemsnovice-drivers/. Accessed 20 Dec 2018 11. Scott-Parker, B., Hyde, M.K., Watson, B., Kingac, M.J.: Speeding by young novice drivers: what can personal characteristics and psychosocial theory add to our understanding? Accid. Anal. Prev. 50, 242–250 (2013) 12. Scott-Parker, B.J.: A comprehensive investigation of the risky driving behaviour of young novice drivers. Queensland University of Technology (2012) 13. Scott-Parker, B., Watson, B., King, M., Hyde, M.: Young, inexperienced, and on the road do novice drivers comply with road rules? Transp. Res. Rec. 2318, 98–106 (2012)

Impact of Mind Wandering on Driving Minerva Rajendran and Venkatesh Balasubramanian(&) Rehabilitation Bio-Engineering Lab, Engineering Design Department, Indian Institute of Technology, Chennai, TamilNadu, India [email protected], [email protected]

Abstract. According to Ministry of road transport and highways (MoRTH), Government of India, 84% of road accidents in 2016 occurred due to driver’s fault. Mind wandering, an internal source of driver distraction has been a latent and less researched factor for road accidents. This necessitates the requirement to study the influence of mind wandering during driving in greater detail to improve driver and road safety. Among various vehicle parameters, vehicle speed is considered a significant parameter in determining the probability of accidents. This study investigated the influence of vehicular speed on mind wandering in high perceptual simulated driving. This study assessed the variability of speed and frequency of mind wandering during two-speed conditions: 40 mph and 70 mph. Results indicated greater variability in speed during mind wandering for both speed conditions. The frequency of mind wandering at 70 mph and 40 mph condition was 22.5% and 26% respectively. Participants response to questionnaire revealed driving environment and personal matters to be major contributors for mind wandering. Keywords: Mind wandering


Vehicular speed
Transportation safety

1 Introduction Humans often find themselves contemplating thoughts about past or future events which are completely irrelevant to the present task. Especially during driving, thinking about thoughts unrelated to driving is a common phenomenon experienced by all drivers. This shift of attention from current external task to internal thoughts which are task unrelated is referred to as mind wandering. Individuals spent nearly 50% of their time mind wandering irrespective of the nature of current work [1]. Many studies had reported accidents, errors and decline in performance due to mind wandering during driving, lectures, continuous monitoring of patient’s critical signs and supervisory tasks in industries etc. Extrinsic and intrinsic distractions are types of distractions which can shift attention away from driving. Responding to phone calls and text messages are some examples of extrinsic distraction, while mind wandering and fatigue are intrinsic distractions. Literature about the negative impact of intrinsic distractions namely mind wandering was limited when compared to reports about extrinsic distractions. However, due to the contribution and collaboration of many researchers, literature about the impacts of mind wandering on various sustained attention tasks started expanding. Driving and its © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 224–232, 2020. https://doi.org/10.1007/978-3-030-20503-4_20

Impact of Mind Wandering on Driving

225

association with mind wandering can be considered as one of the key areas to study the negative impacts of mind wandering. Vehicle parameters, behavioral parameters such as level of attention, boredom, practice effect, driving style, etc., and driving environment parameters namely perceptual load, city driving, highway driving, etc., have been identified as causal/effect factors of mind wandering. Variations in vehicle parameters such as lane control, speed, and steering position due to mind wandering were reported in the literature [2–6]. Behavioral parameters namely boredom proneness had been associated with inattention [7], decrement in performance [8]. Cognitive studies have also reported the influence of perceptual load experienced by drivers on mind wandering. Traffic conditions, building densities, billboards etc., contribute to perceptual load. These findings show a correlation between behavioral parameters and inattention towards the task. Investigating the link between mind wandering and these parameters may provide insights about the occurrence of mind wandering during sustained attention tasks like driving. In order to know the occurrence of mind wandering, capturing mind wandering episodes become crucial. Literatures reported retrospective questionnaires [9], probe-caught [9–11]) and selfcaught methods [9, 12] to measure frequency of mind wandering. Retrospective methods could provide detailed information about the contents of mind wandering, but people might forget the frequency of mind wandering as it was retrospective. The selfcaught method required people to register their mind wandering period by constantly monitoring their thoughts. This has the advantage of not missing any mind wandering episodes, but constant monitoring of thoughts itself might reduce the occurrence of mind wandering. Probe-caught method overcome that disadvantage of self-caught by probing subjects at regular or random intervals. This prevented people from constantly monitoring their thoughts. However, the accuracy of a probe-caught method to accurately determine the frequency of mind wandering lied in determining inter-tone interval [10]. This study utilized probe-caught method to measure mind wandering frequency and retrospective questionnaire. Combining two methods (probe-caught & retrospective questionnaire) could provide more details about mind wandering. This study investigated associations between vehicle, behavioral and environmental parameters on mind wandering. The purpose of this study was of two-fold, one was to investigate the impact of vehicle parameter namely vehicle speed and high perceptual driving environment on mind wandering frequency; second, was to study variability in speed during mind wandering.

2 Methods and Materials 2.1

Materials

Twenty-one healthy volunteers (4 females; age = 22.6 ± 3.38) recruited from Indian Institute of Technology, Madras participated in this study with an average driving experience of 11 months (± 19 months). All participants had normal and corrected to normal vision and did not have any history of mental disorders which would directly affect their participation in the task. Participants confirmed no intake of any medication

226

M. Rajendran and V. Balasubramanian

which would affect their alertness. All participants were briefed about the experiment and they gave informed consent for their participation. Participants drove in two different speed limits, i.e., 40 mph (here on referred as SL 1) and 70 mph (here referred as SL 2) in a Logitech G25 static driving simulator for a distance of 22.36 Kms (13.9 miles). The simulated driving environment had high perceptual loading on subjects. Buildings, bridges, billboards, traffic density in the driving environment contributed to perceptual loading. Participants were instructed to drive on the right side of the road; to have lane control and to avoid collisions. The self-reportable questionnaire was administered to participants to record their demographics, health conditions, state of mind before the start of experiment. The questionnaire was again administered to record their state of mind after the experiment, level of attention required during driving, sources of mind wandering and perceived effects of mind wandering. 2.2

Methods

All participants took practice session for 15–20 min to familiarize with driving controls and drove for two conditions at two different days of the week. The order of speed conditions (40 mph and 70 mph) were mixed to avoid the conditional effect on results. Minimum and maximum speed for SL 1 was 15 mph and 40 mph respectively and 45 mph and 70 mph respectively for SL 2. This speed range was set to increase the utilization of attentional resources and to refrain subjects from driving similar speeds for both conditions. Participants were instructed to maintain the respective speed range for the given condition and were re-instructed whenever their speed changed ±10 mph the speed range. Mind wandering during driving was captured by the probe-caught method. Arduino controlled push-button system was mounted on the simulator steering wheel. Probe tone of 2250 Hz was played for 500 ms duration to capture mind wandering episodes. Inter-tone intervals were kept in the order of 30, 60, 90 and 60 s and the order repeated till the completion of task. This randomization was needed to remove the habituation effect of probe tone on participants. Subjects were instructed to press left pushbutton if they were focused on driving just before tone or to press right pushbutton if they were mind wandering just before the tone. Vehicle speeds were collected every 15 s throughout the drive for both speed conditions. In addition, speeds at 5 s before the probe tone and response time (latency between probe tone and push button press) were collected. Data were separated into mind wandering and on-task dataset based on the response to probe tone. Speeds and response time recorded when the left push button was pressed was referred as on-task data set. Similarly, speeds and response time corresponding right pushbutton was referred as mind wandering (referred as MW) dataset. On-task referred to the period when subjects focused only on driving without mind wandering. Speeds collected at 15 s and 5 s prior to probe tone for on-task dataset and mind wandering dataset were averaged across subjects and analyzed for a statistically significant difference between speeds recorded during on-task and mind wandering episodes.

Impact of Mind Wandering on Driving

227

3 Results This study hypothesized that driving in high perceptual environment at higher speed reduces the frequency of mind wandering. This hypothesis was in accordance with the load theory of attention [13]. According to the theory, individuals have a fixed capacity of cognitive resources. When task demand increases, in order to maintain performance, utilization of cognitive resource increases. This situation leaves very less residual cognitive capacity for the mind to wander. When the demands are low, the remaining capacity called spared capacity is used for mind wandering. The result in Fig. 1 depicts the load theory of attention. As vehicle speed and speed range in SL 2 increased, subjects utilized more of their attentional resources to maintain control of the vehicle, which resulted in less mind wandering at SL 2. The frequency of mind wandering was 26% at SL 1 and reduced to 22.5% at SL 2.

Fig. 1. Mind Wandering Probe Count

Wilcoxon Signed Rank test for 40 mph conditions showed that vehicle speed collected at 15 s prior to probe tone was slightly greater during on-task periods (mean rank = 11.25) than during mind wandering periods (mean rank = 10.20) (Z = –2.242, p = 0.025). Wilcoxon test did not show any statistical significance for 70 mph conditions for the speeds collected at 15 s prior to probe tone. These results were different to [6], who had reported greater speeds during mind wandering than during on-task state. No statistically significant results were found for speeds collected 5 s prior to probe tone for SL 1 and SL 2 conditions. The second purpose of this study was to investigate the impact of mind wandering on the variability of speed. High variability in speeds was noted during mind

228

M. Rajendran and V. Balasubramanian

wandering periods (Fig. 2). F-test two sample for variance test revealed statistically significant speed variability between on-task and mind wandering episodes (F (19, 19) = 4.77, p < 0.00, one-tailed) and (F (19,19) = 11.76, p < 0.00, one-tailed) during SL 1 and SL 2 respectively for speeds collected 15 s prior to probe tone. These results showed that whenever mind wandered, the driver’s control over the speed of the vehicle reduced and resulted in higher variability in speed.

Fig. 2. Speed Variability for 40 mph and 70 mph Conditions: The legends placed at the bottom of the figures are explained as follows: 15 s and 5 s speed referred to speed collected at 15 s and 5 s prior to probe tone. Ontask and MW correspond to on-task dataset and mind wandering dataset respectively. Errors bars show the standard deviation of datasets.

Impact of Mind Wandering on Driving

229

Linear regression was conducted to check whether vehicle speed could be a predictor for response time. No significant regression was found for mind wandering dataset for 40 mph speed condition. A significant regression was found for 70 mph speed condition (F (1,19) = 15.97, p < 0.05), with R2 of .457. Thus, response time increased 24 ms for each mph of speed during mind wandering. Speed at 15 s prior to probe tone was a significant predictor of response time during mind wandering. Linear regression equation with speed sampled at 15 s as predictor: Response time of the participant = 59.717 + 24.067*(15sp_mw#) during mind wandering # Speed sampled at 15 s prior to probe tone Participants were administered retrospective questionnaire after the experiment to record detailed responses about sources and effects of mind wandering during the experiment and outside the laboratory conditions. Simulated driving environment in which the experiment was performed was referred as driving environment. Driving environment, boredom, and personal matters were reported as major contributors to mind wandering (Fig. 3). Due to high perceptual driving environment, it became a source of mind wandering. This option might not be a source when driving in highway, which is usually considered as a monotonous driving environment with low perceptual load. Effects of mind wandering were shown in Fig. 4. Participants answered this question on how they felt when they realized they were mind wandering in their daily life and could choose more than one option. Participants reported feeling guilty about mind wandering in real-life. These responses gave detailed insight about how mind wandering – a natural phenomenon, was perceived as a hindrance when it occurred during sustained attention tasks. Nevertheless, it was also perceived as a source of creative thinking and problem-solving sketch pad by participants. 85% of participants reported the requirement of moderate to high-level attention for this task. This indicated that the perceptual load of driving environment was felt by participants during simulator driving.

4 Discussions Mind wandering is an inevitable natural phenomenon occurring for everyone. Mind wandering during a task results in poor performance, while its occurrence when not doing work might result in creative thinking. The results presented showed its negative impact when doing a high attention task like driving. The high perceptual driving environment did load the subjects for two-speed conditions. However, the frequency of mind wandering was reduced during 70 mph speed condition than in 40 mph, this could be attributed availability of less residual capacity during 70 mph than during 40 mph conditions. At higher speeds, subjects used more attentional resources to drive safely and to avoid collisions.

230

M. Rajendran and V. Balasubramanian

Fig. 3. Sources of Mind Wandering

Fig. 4. Effects of Mind Wandering

This is true in real-life driving. Nevertheless, even few mind wanderings at higher speed pose an equal, otherwise greater challenge to road safety as compared more mind wanderings at a lower speed. Greater variability in speed during mind wandering than during on-task state indicated a reduced focus on driving. This reduced focus could be due to perceptual decoupling, which shifts the attention from external to an internal

Impact of Mind Wandering on Driving

231

state. Perceptual decoupling refers to a mental state, in which an individual would be completely unaware of the external environment [14]. This nature of mind wandering had been reported by studies as a causal factor for the decline in performance whenever a mind wandered away from the current external task. Regression analysis showed an increase in response time by 24 ms during mind wandering. In real-time driving, increase in response time for an external event increases the probability of crashes at greater speed. In this study, vehicle speeds collected at 15 and 5 s correlated. This correlation gave insight on the time spent by drivers on mind wandering. Vehicle speeds could be collected at more time intervals to have in-depth understandings about the time spent on mind wandering. Further research with more samples and collecting physiological signals are needed to verify the occurrence of perceptual decoupling during mind wandering. Guilt due to mind wandering, reported as an effect may reduce people from mind wandering during their work, but due to its inevitable nature, rather than curbing its occurrence, managing its frequency might be a more practical solution. Thus, this study showed, mind wandering as an influencer on vehicle speed variations and speed conditions as an influencer on the frequency of mind wandering. Therefore, mind wandering could act as a causal and effector. This study indicated that mind wandering is indeed a critical factor which has to be addressed to improve the safety of drivers.

References 1. Killingsworth, M.A., Gilbert, D.T.: A wandering mind is an unhappy mind. Science 330(6006), 932–932 (2010) 2. Baldwin, C.L., Roberts, D.M., Barragan, D., Lee, J.D., Lerner, N., Higgins, J.S.: Detecting and quantifying mind wandering during simulated driving. Front. Hum. Neurosci. 11, 406 (2017) 3. He, J., Becic, E., Lee, Y.C., McCarley, J.S.: Mind wandering behind the wheel: performance and oculomotor correlates. Hum. Factors 53(1), 13–21 (2011) 4. He, J., McCarley, J.S., Kramer, A.F.: Lane keeping under cognitive load: performance changes and mechanisms. Hum. Factors 56(2), 414–426 (2014) 5. Bencich, E., Gamboz, N., Coluccia, E., Brandimonte, M.A.: When the Mind “Flies”: The Effects of Mind Wandering on Driving. EUT Edizioni Università di Trieste (2014) 6. Yanko, M.R., Spalek, T.M.: Driving with the wandering mind: the effect that mind wandering has on driving performance. Hum. Factors 56(2), 260–269 (2014) 7. Farmer, R., Sundberg, N.D.: Boredom proneness–the development and correlates of a new scale. J. Pers. Assess. 50(1), 4–17 (1986) 8. Bakan, P.: An analysis of retrospective reports following an auditory vigilance task. In: Vigilance: A symposium, pp. 88–101. McGraw Hill, New York (1963) 9. Mrazek, M.D., Smallwood, J., Schooler, J.W.: Mindfulness and mind wandering: finding convergence through opposing constructs. Emotion 12(3), 442 (2012) 10. Giambra, L.M.: A laboratory method for investigating influences on switching attention to task-unrelated imagery and thought. Conscious. Cogn. 4(1), 1–21 (1995) 11. Barron, E., Riby, L.M., Greer, J., Smallwood, J.: Absorbed in thought: the effect of mind wandering on the processing of relevant and irrelevant events. Psychol. Sci. 22(5), 596–601 (2011)

232

M. Rajendran and V. Balasubramanian

12. Hasenkamp, W., Wilson-Mendenhall, C.D., Duncan, E., Barsalou, L.W.: Mind wandering and attention during focused meditation: a fine-grained temporal analysis of fluctuating cognitive states. Neuroimage. 59(1), 750–760 (2012) 13. Lavie, N., Hirst, A., de Fockert, J.W., Viding, E.: Load theory of selective attention and cognitive control. J. Exp. Psychol. Gen. 133(3), 339–354 (2004) 14. Antrobus, J.S., Singer, J.L., Goldstein, S., Fortgang, M.: Mind wandering and cognitive structure. Trans. NY Acad. Sci. 32, 242–252 (1970)

Assessing the Relation Between Emotional Intelligence and Driving Behavior: An Online Survey Swathy Parameswaran and Venkatesh Balasubramanian(&) Rehabilitation Bio-Engineering Lab, Engineering Design Department, Indian Institute of Technology, Chennai, Tamil Nadu, India [email protected], [email protected]

Abstract. Risky driving has been one among the major causes of accidents on road. Motivation behind the study was to identify if emotional intelligence influences impetuous driving in highly congested roads. Volunteers answered a 4-point online survey of questions sampled from Indian license test, driving skill and behavior questionnaires. Cronbach alpha was acceptably high validating the questions. Spearman correlation values indicated a strong correlation between driving hours and driving skill and behavior. K-means clustering was used to cluster the subjects into 4 categories based on driving skill and driving behavior scores. The cluster with highest number of subjects consisted of people who drive every day with high risky driving scores. The results suggest that people who drive everyday have poor emotional intelligence which impedes a safe driving. The study proposes that educating drivers with emotional regulation could help in safer roads. Keywords: Driving behavior


Emotional intelligence
Clustering

1 Introduction A recent report on road accidents from Ministry of Road Transport and Highways in India reports 4,64,910 accidents causing injuries to 4,70,975 persons and claiming 1,47,913 lives in the country during the year 2017 [1]. According to the report, on an average, 1274 accidents and 405 accident deaths take place on Indian roads every day; which translates to 53 accidents and 17 deaths every hour. Hence, a necessity for studies regarding road safety still persist. To analyze the factors that contribute to the mishaps happening on road can be grouped at organizational, vehicle and human levels. While constructional and interventional solutions are aimed at addressing organizational level, design solutions are being proposed at vehicular level to prevent accidents. Human factors in combination with the either of other two factors are considered to contribute to most of the accidents. Hence, accommodating human factors and errors in building a safe road will further reduce the accidents statistics aforementioned. Risky and aggressive driving remain as key causes of accidents in India [2]. Literature suggests that driving style and skill are crucial determinants of risky driving [3]. © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 233–239, 2020. https://doi.org/10.1007/978-3-030-20503-4_21

234

S. Parameswaran and V. Balasubramanian

While driving skill could refer to abilities of a person to handle road and vehicle, driving style could refer to abilities of a person to handle road, vehicle and emotions. Also, driving skill indicates driver’s awareness of driving rules while driving style indicates a person’s choice to be abide by them. Several studies have analyzed causes for poor emotional regulation that lead the drivers not abide by the rules causing accidents. Frequently studied human factors causing a poor emotional regulation and thus leading to a risky driving include age, gender, personality, attitude, risk perception, anger, temperament. Such studies typically focus on the role of overt traits or transient emotional states as indicators of risky driving. Very limited studies have studied the possible mediating role of Emotion Intelligence (EI) on engaging in risky driving [4]. EI is defined as set of skills that enables monitoring personal and other people’s emotions, discriminate between different emotions, label emotions, and use emotions to guide behavior or thinking [5]. Driver’s response to the transient states of moods like impulsiveness, narcissism, sensation seeking and anxiety depends on their EI quotient [6]. Apart from individualistic factors, social desirability like time pressure and work nature also play a major role in regulating EI [7]. In this paper, we define EI to include understanding about the vehicle functioning, knowledge about rules and behavioral aspects like empathy, patience and resilience. To assess the understanding about vehicle and driving, questions from Driving Skill Inventory (DSI) were exploited. Questions from Driver Behavior Questionnaire (DBQ) were used to assess behavioral aspects. Questions from Indian learner license test were used to test the knowledge about road rules in India. The DBQ [8] which studies the violations, errors and lapses during driving has been extensively used in literature to observe driver behavior pattern. The DSI [9] measures self-reported perceptual -motor skills and safety skills studying the opposite dimensions of DBQ. Questions about violations were sampled from DBQ and combined with questions from DSI to access the assumed qualities of EI in drivers. Indian learner license test comprises of questions that test the knowledge of driver about Indian driving rules, traffic signals and road signs. Apart from explicit transient emotional states, covert factors such as everyday driving, traffic congestion, accident history and years of driving licensure are also studied to have effect on EI. Hence this study encompasses the aforementioned aspects in the questionnaire. To study the variability of EI, two different population who drive in a state with highest accident [2] were considered. One group comprised of nonprofessional drivers who drive on a daily basis for at least 1.5 h and the other group of people who seldom drive two hours per week. The criterion was fixed at 1.5 h with the average commutation time to work from home in mind. A medium-heavy traffic congested road was considered because it is prone to high and frequent emotional regulations [10]. It is to be noted that people belonging to age group 18–35 years are accounted for 46.3 per cent (69,851 persons) of total number of accidents [1]. Traffic injuries are also a leading cause of death among people aged between 15 and 29 years globally. Studies in the past concluded that younger drivers are prone to traffic violations while older drivers are prone to crashes at intersections [11, 12]. Hence this study aims at drivers within age group of 20 and 30 years.

Assessing the Relation Between Emotional Intelligence and Driving Behavior

235

2 Methodology 2.1

Participants

60 subjects aged between 20 and 30 years (M = 26.85, SD = 6.46) volunteered to participate in the study. Inclusion criteria for the current study required participants to be aged between 20 and 30 years currently holding a valid license within India and having a minimum of 2 years driving experience. The sample constituted of a control group of 30 people (18 male, 12 female) who drive personal cars in congested city roads on a daily basis and experiment group of 30 people (25 male, 5 female) who drive seldom. The sample consisted of people from the same city to keep the traffic congestion condition homogeneous. 2.2

Procedure

Participants had to complete an online survey which consisted of 50 questions about demographic details, driving skill and driving behavior on 4-point scale (0 – never, 1 – sometimes, 2-most of the times, 3 – always). The driving skill questions were sampled from the DSI and also questions from Indian learner license test that consisted questions about roads rules, traffic signals and road signs. Questions about driving behavior were sampled from DBQ. Questions were graded based on intentional mistakes and unintentional mistakes that could either cause no harm or inconvenience for the perpetrator or some possibility of risk to others or cause a definite risk to other road users. Data was collected through an online surveying platform. A medium to high traffic congestion condition was asked to be assumed to answer the questions. 2.3

Statistics

Cronbach alpha was calculated to determine the internal consistency of questions. To test the normality of data, Shapiro–Wilk tests were performed. Spearman correlation was also calculated between driving behavior, driving skill factors, age, gender, accidents, number of years of driving and hours of driving per week. K-means clustering techniques was used to classify the driver group into 4 clusters based on the EI. Finally, Mann–Whitney non-parametric tests were performed to see whether the clusters were statistically significant.

3 Results 3.1

Reliability and Correlation

Table 1 shows Spearman correlation values between all seven factors, the driving behavior score, driver skill score, number of years of driving age, gender, accidents and hours of driving with acceptable high Cronbach’s alpha values. Results show that number of driving hours per week is highly correlated to driving skill score and is negatively correlated to driving behavior.

236

S. Parameswaran and V. Balasubramanian Table 1. Alpha values and Spearman’s correlations obtained

Cronbach alpha (a)

Driving skill score a = 0.76 1

Driving behavior score a = 0.81

No. of years of driving

Age

Driving skill score Driving .143* 1 behavior score .189** .017* 1 No. of years of driving Age .181** .446** .295* 1 Gender .077* .135* .043** −.148* Number −.212* .124** .084* −.003** of accidents Hours of .865* −.742* .231* −.03* driving *- Correlation is significant at the 0.05 level (2-tailed) **- Correlation is significant at the 0.01 level (2-tailed)

3.2

Gender

Number of accidents

1 .244*

1

.341*

−.018*

Hours of driving

1

K Means Clustering

The skill scores were obtained as an additive score of Indian learner license test and DSI scores while behavioral scores were computed from DBQ. The 2 groups were further grouped into 2 predetermined clusters based on only skill and behavior scores. Hence, we obtain daily driving safe drivers (cluster 1), daily driving unsafe drivers (cluster 2), seldom driving safe drivers (cluster 3), seldom driving unsafe drivers (cluster 4). Least mean square distance between the cluster centers was chosen to classify the data points. Clusters converged at 4th iteration indicating reliability of clusters obtained. The mean and standard deviation of the skill scores and behavior scores of the 4 groups are shown in Fig. 1. Shapiro–Wilk tests showed that the data is non-normally distributed. Hence Non-parametric tests were conducted to check for significant differences between the clusters. From the Fig. 1, it can be inferred that the group 1 classified as daily driving unsafe drivers clearly outlie in skill and behavior scores. The Mann Whitney U test returned a statistical significance of (95% significance) across the groups thus validating the four different clusters.

Assessing the Relation Between Emotional Intelligence and Driving Behavior

237

Fig. 1. Comparison of skill and behavior scores between the 4 clusters obtained. cluster 1- daily driving safe drivers, cluster 2 - daily driving unsafe drivers, cluster 3- seldom driving safe driver and cluster 4 - seldom driving unsafe drivers.

4 Discussion The study was designed to verify the relationship between EI and driving behavior in congested traffic conditions. The EI was calculated based on questions from DSI, DBQ and Indian learner license test. A high EI was considered to be a combination of high skill score and low risky driving behavior score. The acceptably high Cronbach alpha value indicates high internal consistency of the survey. There have been discussions in recent time if DBQ can be used to predict accidents [13]. This study supports the concept of self-report measures having an importance in road safety. This study has

238

S. Parameswaran and V. Balasubramanian

helped in identifying potentially risky drivers by cluster analyzing the subjects based on their driving skill and behavior scores. The four clusters differed in scores obtained for EI while their individual driving factors like age, gender and number of accidents did not show a significant difference. Groups that are considered to be unsafe are observed to have unskilled and violating qualities. Unskilled drivers reported low confidence, high driving stress, and low perceptual-motor skills. Violating drivers reported high intentional mistakes and low safety skills. The groups that are considered to be safe are observed to have high perceptual-motor skills, low intentional mistakes and high safety skills. Notably, the unsafe group consisted of higher number of persons with one or more accidents. The self-reported study thus shows that the drivers are aware of their driving discourtesy which raises the question of why would they do harm and cause morbidity and mortality. A possible explanation stems from the observation made by a study [14] that suggests driver’s learning from exposure to driving without accidents decrease their risk perception and safety concern. A study indicated that young drivers undergoing driver education were less likely to be involved in traffic violations during their first two years of driving [15, 16]. If such potential risky drivers could undergo training to improve the awareness of their own driving skills, it could prevent the false safety sense and also reduce their overconfidence. This study will be further explored by including more socio-demographic factors that contribute to EI. One another limitation of current study is the number of responses and hence the results are to be interpreted with caution. The survey could be extended to people with different driving purposes to explore the relationship of job nature and driving pattern. The study utilizes only self-report measures to draw conclusions. The results could be validated with actual driving skill and driving behavior.

5 Conclusion This study successfully assesses the relationship between emotional intelligence and driving behavior in young adults in highly congested driving conditions. Set of 50 questions sampled from DSI, DBQ and Indian learner license test were rated on a 4point scale by 60 selected people. The subjects were clustered into 4 groups namely daily driving unsafe, daily driving safe, seldom driving unsafe and seldom driving safe driver based on scores calculated using K-means algorithm. Statistical testing results suggest that people who drive everyday are prone to have less EI suggesting for a driver education system. The proposed solution could bring down risky driving.

References 1. Ministry of road transport & highways transport research wing, Road accidents in india – 2017, (Report to Government of India, 2017), New Delhi (2017) 2. Singh, S.K.: Road traffic accidents in India: issues and challenges. Transp. Res. Procedia 25, 4708–4719 (2017)

Assessing the Relation Between Emotional Intelligence and Driving Behavior

239

3. Martinussen, L.M., Møller, M., Prato, C.G.: Assessing the relationship between the driver behavior questionnaire and the driver skill inventory: revealing sub-groups of drivers. Transp. Res. Part F Traffic Psychol. Behav. 26, 82–91 (2014) 4. Hayley, A.C., de Ridder, B., Stough, C., Ford, T.C., Downey, L.A.: Emotional intelligence and risky driving behaviour in adults. Transp. Res. Part F Traffic Psychol. Behav. 49, 124– 131 (2017) 5. Salovey, P., Mayer, J.D.: Emotional intelligence. Imagin. Cogn. Pers. 9(3), 185–211 (1989) 6. Gohm, C.L.: Mood regulation and emotional intelligence: individual differences. J. Pers. Soc. Psychol. 84(3), 594–607 (2003) 7. Cœugnet, S., Naveteur, J., Antoine, P., Anceaux, F.: Time pressure and driving: work, emotions and risks. Transp. Res. Part F Traffic Psychol. Behav. 20, 39–51 (2013) 8. Reason, J., Manstead, A., Stradling, S., Baxter, J., Campbell, K.: Errors and violations on the roads: a real distinction? Ergonomics 33(10–11), 1315–1332 (1990) 9. Lajunen, T., Summala, H.: Driving experience, personality, and skill and safety-motive dimensions in drivers’ self-assessments. Pers. Individ. Differ. 19(3), 307–318 (1995) 10. Wickens, C.M., Wiesenthal, D.L.: State driver stress as a function of occupational stress, traffic congestion, and trait stress susceptibility. J. Appl. Biobehav. Res. 10(2), 83–97 (2005) 11. Jessor, R., Turbin, M.S., Costa, F.M.: Predicting developmental change in risky driving: the transition to young adulthood. Appl. Dev. Sci. 1(1), 4–16 (1997) 12. McGwin Jr., G., Brown, D.B.: Characteristics of traffic crashes among young, middle-aged, and older drivers. Accid. Anal. Prev. 31(3), 181–198 (1999) 13. De Winter, J.C.F., Dodou, D.: The driver behaviour questionnaire as a predictor of accidents: a meta-analysis. J. Saf. Res. 41(6), 463–470 (2010) 14. Bandura, A.: Self-efficacy: toward a unifying theory of behavioral change. Psychol. Rev. 84 (2), 191–215 (1977) 15. Hatakka, M., Keskinen, E., Gregersen, N.P., Glad, A., Hernetkoski, K.: From control of the vehicle to personal self-control; broadening the perspectives to driver education. Transp. Res. Part F Traffic Psychol. Behav. 5(3), 201–215 (2002) 16. Wozniak, R.L.: Risky sexual behaviors in adolescence: their relationship to social-emotional intelligence. Doctoral dissertation, Alfred University, Alfred, NY (2013)

Human Factors in Transportation: Rail

The Effect of Tram Driver’s Cab Design on Posture and Physical Strain Tobias Heine(&), Marco Käppler, and Barbara Deml Institute of Human and Industrial Engineering, Karlsruhe Institute of Technology, Engler-Bunte-Ring 4, 76131 Karlsruhe, Germany {tobias.heine,marco.kaeppler,barbara.deml}@kit.edu

Abstract. The ergonomic quality of a tram driver’s cab is essential to ensure the physical well-being of the drivers and the general attractiveness of the workplace. We investigated the ergonomic quality of the cabs of two different trams in a field study during real operation. The results show that the experienced physical strain differs significantly between the two trams. A video analysis relates this to different posture and movement patterns. The main factor for these differences is the position of the main control panel, which needs to comply with visibility requirements according to DIN 5566. However, our study shows that an ergonomic workplace cannot be accomplished by only pursuing isolated factors, instead the interaction of all relevant factors has to be considered. Keywords: Tram driver’s cab Ergonomic design


Field study
Physical strain


1 Introduction Trams play an important role in the public transportation sector in Germany, for many years the number of passengers has constantly been increasing [1]. In comparison to other European countries, Germany has by far the most light rail and tram systems [2]. However, public transportation companies in Germany are more and more confronted with the challenge of demographic change. On average, the workforce is getting older and older and it is becoming increasingly difficult to recruit junior staff [3]. In order to keep older drivers healthy at work for a longer period of time and to recruit new people for the profession of tram driver, it is essential to provide an attractive workplace. A central element of the attractiveness of the workplace of a tram driver is the ergonomic quality of the driver’s cab. Nowadays, trams (US: streetcars) can easily achieve service lives of 25–30 years. For tram drivers, the design of the cab determines their workplace, often for decades. This makes ergonomic considerations in the design process of a cab an even more important topic. The central standard for the design of a driver’s cabs for railway vehicles in Germany is DIN 5566, which consists of three parts (for standards used in other European countries see [4]). Part 1 [5] deals with general requirements for railway vehicles, part 2 [6] formulates additional requirements for standard gauge railway vehicles. Part 3 deals with additional requirements for urban and suburban rolling stock, which comprises also trams [7]. © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 243–249, 2020. https://doi.org/10.1007/978-3-030-20503-4_22

244

T. Heine et al.

With respect to ergonomic questions, the standards provide requirements regarding anthropometry, noise, seat design, visibility and the design of the control panel (design, reachability and positioning of control elements). However, the requirements are sometimes only “minimum requirements” and/or have a recommendatory nature. This gives degrees of freedom for the concrete design of the tram cab. While European railways have made efforts to standardize the design of the whole driver’s cab (DBEinheitsführerstand, European Driver’s Desk), tram driver’s cabs nowadays are all designed differently. Therefore, it is quite interesting to compare differently designed cabs with regard to their ergonomic quality. The tram operation company of Karlsruhe (Verkehrsbetriebe Karlsruhe, VBK) approached our institute with the aim of learning more about specific factors which determine the ergonomic quality of a tram driver’s cab and which therefore should be taken into account for the development of future trams.

2 Methods 2.1

Tram Types

In our study, we compared the two different tram types that are currently in operation at VBK: 1. Type: NET2012 Manufacturer: Vossloh Rail Services, Stadler Rail Number of trams at VBK: 75 (2018) Start of operation: 2014 2. Type: GT6-70D/N Manufacturer: DUEAWAG, Siemens Manufacturer: Stadler Rail (until 2015: Vossloh Rail Services) Number of trams at VBK: 45 + 25 (longer version) Start of operation: 1995 2.2

Measurement Devices

For the analysis of the work processes and the accompanying body postures, the driver’s cab was equipped with the following measurement devices (see Fig. 1 left): – Two video cameras (GoPro Hero 3+ , GoPro Hero 5 black). One camera was positioned on the control panel to record the frontal view, the other camera was positioned on the left window to record the sagittal view of the driver. – Two pressure distribution mats (novel pliance©). The mats were placed on the seat and on the left armrest (the results of the pressure distribution mats are not part of this article).

The Effect of Tram Driver’s Cab Design on Posture and Physical Strain

2.3

245

Materials

In order to assess driver’s subjective physical strain, we used a map of the human body (according to [8]). The map divides the body into five regions: left forearm, left shoulder, neck, upper back, lower back (see Fig. 1 right). The drivers were asked to give an estimation of the experienced strain for each body region on a ten-point rating scale (1 = no strain; 10 = extreme strain).

3. neck

2. left shoulder 3. 2. 4.

4. upper back 5.

1.

5. lower back

1. left forearm

no strain

1

extreme strain

2

3

4

5

6

7

8

9

10

Fig. 1. (left) Location of the video cameras in the NET2012. (right) Body map with five highlighted body regions that were used for the measurement of the experienced physical strain.

2.4

Route

The route started in a northwestern district of Karlsruhe (Neureut) with a rather low passenger volume and with dedicated tram tracks (i.e. the tracks are separated from other vehicles/cars). After about 15 min, the route entered Karlsruhe City. Here, the passenger volume is generally much higher and the tram runs on tracks that are partly shared with other vehicles. After about 45 min, the route ended in a northeastern district (Waldstadt). Here, the passenger volume is again rather low and the tram operates on dedicated tracks. After a scheduled break, the tram returned to Neureut on the reverse route. The first part of the route is called outward ride, the second part is called inward ride. 2.5

Subjects

A total of six people participated in the study. All participants were regular tram drivers of the VBK and were familiar with both, the two tram types and the route.

246

2.6

T. Heine et al.

Procedure

The tram rides took place on four successive days. Every participant performed two rides: one with the NET2012 and one with the GT6-70D/N. The rides on the two tram types took place on different days, the route, however, was always the same (see above). We tried to keep relevant environmental influences such as visibility conditions (day/night) and passenger volume comparable by driving in the same time slots on each day of the measurements (09:15–14:15). At the beginning of each ride, participants were welcomed at the starting station in Neureut and asked to read and sign a consent form. Afterwards, they had the possibility to adjust the seat and assume their regular driving position. The examiner showed the drivers the body map and the rating scale and asked for a first estimation of their subjective physical strain in each body region, which served as a baseline measurement. The subjective physical strain was assessed approx. every 15 min at predefined stations. For this purpose, the examiner entered the cab while passengers were exiting or entering the tram and asked for a quick oral assessment. Except of these short interruptions, the drivers could operate undisturbed in their cab. After the return to the starting station in Neureut, the driver was asked to give a last subjective assessment. Finally, some body dimensions (e.g. body height) were measured. 2.7

Data Analysis

The subjective data were averaged across all drivers. Due to the small sample size, no statistical test was calculated. Instead, we performed a descriptive analysis of the data with different graphs. The video data were analyzed by two independent raters. With the help of a specially designed Excel-Sheet, the raters counted the frequency of the occurrence of specific body postures.

3 Results 3.1

Subjective Physical Strain

First, we compared the subjective physical strain that drivers experience in the two different trams. In the GT6-70D/N, drivers show almost no physical strain in any body region during the entire ride (see Fig. 2 right). In contrast, drivers show signs of physical strain after about 30 min in the NET2012. The affected body regions are the neck and the left shoulder. Both body regions show a slow but constantly increasing trend until the end of the ride (see Fig. 2 left).

The Effect of Tram Driver’s Cab Design on Posture and Physical Strain Subjective physical strain – NET2012

Subjective physical strain – GT6-70D/N

(mean value of all participants)

(mean value of all participants)

8,0

Time slot for video analysis inward ride

Time slot for video analysis outward ride

7,0

247

Time slot for video analysis inward ride

Time slot for video analysis outward ride

6,0 5,0 4,0 3,0 2,0 1,0 0

15

30

45

60

75

90

105

120

0

15

left forearm

left shoulder

30

45

60

75

90

105

120

Time of questioning (min.)

Time of questioning (min.) neck

upper back

lower back

Fig. 2. Mean subjective physical strain for every body part across all N = 6 tram drivers for the NET2012 (left) and the GT6-70D/N (right). The dotted lines mark the time slots for the video analysis.

3.2

Video Analysis

In order to investigate the reason for the increase of physical strain in the NET2012, we performed a thorough video analysis of the body movements and postures. Due to the fact that an analysis of the entire ride would be too time consuming, we chose a total recording time of 30 min per ride. This time consists of two separate time slots of 15 min each. The start of the first time slot corresponds to the first increase of physical strain (about 30 min after the start of the outward ride, see Fig. 2). The second time slot comprises the same part of the route on the inward ride (about 75 min after the start, see Fig. 2). We identified two movements which differ considerably between the two tram types and which are therefore very likely to cause the physical strain: neck flexion and trunk flexion. Tables 1 and 2 show the results of the neck and trunk flexions, respectively. In the GT6-70D/N, drivers perform on average 74.2 neck flexions and 0.2 trunk flexions during the observed 30 min time slot. In the NET2012, drivers perform 129.3 neck flexions and 39.7 trunk flexions. Both movements occur much more frequently in the NET2012 than in the GT6-70D/N (neck flexion: +74.3%, trunk flexion: +19750%). Particularly, while the GT6-70D/N requires nearly no trunk flexions during normal operation, the tram drivers perform this movement regularly in the NET2012.

248

T. Heine et al.

Table 1. Observed frequency of neck flexions during the 2  15 min video for the NET2012 (NET) and the GT6-70D/N (GT). The table shows the mean frequency across all N = 6 tram drivers. Minimum and maximum show the frequency of the drivers with the most and with the least neck flexions. % cf. shows percentage difference of neck flexions between the two tram types. The interpolated data show the estimated frequency of neck flexions for 1 h, 8 h (work day) and 40 h (work week). Neck flexion

Overall % cf. (2  15 min.)

NET GT Mean 129.3 74.2 Maximum 152 117 Minimum 95 17

Interpolation 1h 8h 40 h NET to GT NET GT NET GT NET GT +74.3% 259 148 2069 1187 10344 5936

Table 2. Observed frequency of trunk flexions during the 2  15 min video for the NET2012 (NET) and the GT6-70D/N (GT). The table shows the mean frequency across all N = 6 tram drivers. Minimum and maximum show the frequency of the drivers with the most and with the least trunk flexions. % cf. shows percentage difference of trunk flexions between the two tram types. The interpolated data show the estimated frequency of trunk flexions for 1 h, 8 h (work day) and 40 h (work week). Trunk flexion Overall % cf. (2  15 min.) Mean Maximum Minimum

NET 39.7 61 7

GT 0.2 1 0

Interpolation 1h 8h 40 h NET to GT NET GT NET GT NET GT +19750% 79 0.4 635 3 3176 16

4 Discussion The subjective data show that drivers experienced physical strain when interacting with the NET2012. The strain experience arose about 30 min after the start of the ride and exhibited a constantly increasing trend until the end of the ride. The affected body regions were primarily the neck and the left shoulder. The strain values reached a level of about 3 and were thus still located on the lower end of the 10-point rating scale. However, this value has to be seen in the context that we only observed a 1.5 h ride at the beginning of the workday. It can be expected that a longer ride would provoke higher strain levels. Over the years, this strain may accumulate and likely lead to health issues. The comparison with the GT6-70D/N reveals that the reason for the higher strain values in the NET2012 lies in its ergonomic design, as drivers in the GT6-70D/N showed almost no signs of physical strain. The video analysis shows that there is a huge difference in the way drivers physically interact with the two tram models. The operation of the NET2012 requires significantly more neck and trunk flexions. These differences seem to be a plausible explanation for the higher strain values in the NET2012.

The Effect of Tram Driver’s Cab Design on Posture and Physical Strain

249

A closer look at the design of the two cabs revealed one major difference: the placement of the main control panel. In the NET2012, the control panel is located further down than in the GT6-70D/N. This requires trunk flexions to reach the control elements. On top of that, even the mere look at the control panel in the NET2012 requires a flexion of the neck. As mentioned in the introduction, DIN 5566 formulates visibility requirements. To comply with the standard, a tram driver has to be able to see a parallel contour of 1.20 m height located 30 cm in front of the tram. The visibility of the parallel contour depends on three factors: (1) the distance between the driver and the front end of the tram, (2) the position of the eyes of the drivers (seat height) and (3) the location and shape of the control panel. One main difference between the two trams is the seating position. While the minimum seat height in the GT6-70D/N is located 122.5 cm above track level, the minimum seat height in the NET2012 is located 146.5 cm above track level. In order to comply with the visibility requirements, the control panel in the NET2012 has been installed further down which comes at the cost of a reduced usability.

5 Conclusion Our results show that it is possible to design a cab in a way such that the driver experiences nearly no physical strain during regular operation. For the NET2012, this could be accomplished by the design of a more sophisticated shape for the control panel, which can be located further up without impairing the visibility. In general, a mere focus on isolated ergonomic factors has to be avoided but instead the interaction of different factors has to be considered. Only such a systems approach ensures the ergonomic quality of a tram driver’s cab which contributes to both, the attractiveness of the workplace and the physical well-being of the drivers.

References 1. Verband Deutscher Verkehrsunternehmen (VDV). VDV-Statistik 2017, Köln (2018) 2. The European Rail Research Advisory Council. Metro, light rail and tram systems in Europe (2012) 3. Neubert, A.: Verkehrsbetrieben gehen Fahrer aus. Mitteldeutscher Rundfunk (2018) 4. Orthner, C.: Ergonomische Gestaltung von Fahrerständen in Schienenfahrzeugen. AUVA Forum Prävention (2008) 5. DIN Deutsches Institut für Normung e.V. Schienenfahrzeuge – Führerräume – Teil 1: Allgemeine Anforderunge (5566-1) (2006) 6. DIN Deutsches Institut für Normung e.V. Schienenfahrzeuge – Führerräume – Teil 2: Zusatzanforderungen an Eisenbahnfahrzeuge (5566-2) (2006) 7. DIN Deutsches Institut für Normung e.V. Schienenfahrzeuge – Führerräume – Teil 3: Zusatzanforderungen an Nahverkehrs-Schienenfahrzeuge (5566-3) (2006) 8. Corlett, E.N., Bishop, R.P.: A technique for assessing postural discomfort. Ergonomics 19(2), 175–182 (1976). https://doi.org/10.1080/00140137608931530

Engineering the Right Change Culture in a Complex (GB) Rail Industry Michelle Nolan-McSweeney1(&), Brendan Ryan2, and Sue Cobb2 1

Network Rail, London, UK [email protected] 2 University of Nottingham, Nottingham, UK

Abstract. This paper focuses on an interview and observational study of two major change programmes, designed to transform workforce safety across Great Britain’s railways. The implications of the pace of change and the challenges of user-influenced design are considered in the context of a railway system where there are rapidly evolving technologies and need to consider the impact of cooperative work systems and the skills workers will need to engage with them. The study shows how things have changed over time since the programmes were first introduced, identifying the factors that have influenced this, such as a focus on a continuous improvement culture. Further research directions are proposed, including the need to identify the tools to help predict how future interventions in the change programmes might manifest themselves, e.g. the effects of new technology introduction, or factors outside of the organisation’s control such as Government policy change. Keywords: Socio-technical system User-influenced design


Culture
Complex
Change


1 Introduction When change is to be deployed, such us a new system or process, this may fall short of expectations through poor design or unworkable procedures. Mumford [1] showed that in the pursuit of delivering change the focus is often only on the technological solution, neglecting social aspects and thus failing to achieve a successful outcome. The ‘socio-technical systems’ approach [2] can help to blend the social and technical sub-systems when introducing new technology or processes. Reiman [3] has shown how maintenance often focuses on human errors and issues at an individual-level. There are few studies of normal work and the practices and cultures that influence maintenance, especially in railway settings. More research is needed to understand how major change programmes are implemented and the effect these can have on railway workers, especially their behaviours and decision-making. Network Rail is a large rail engineering organisation within a complex industrial setting. As the pace of change continues to accelerate, the challenges of introducing new technologies and developing a high-performance culture, are getting tougher. To achieve its safety performance targets the organisation has identified two major change programmes to help support the transformation of workforce safety. The first involves © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 250–260, 2020. https://doi.org/10.1007/978-3-030-20503-4_23

Engineering the Right Change Culture in a Complex (GB) Rail Industry

251

simplification of rules (known as the Business-Critical Rules (BCR) framework), providing a clear understanding of the controls necessary to manage risk, and the performance measures for those controls. The second requires the implementation of new processes for planning and delivering safe work (PDSW). This ensures that a person is accountable for managing task and operational safety risk on each worksite, including planning of the work using new planning tools and materials to develop safe work packages for onward briefing. Both of these change programmes focus on workforce safety. Implementation has been tracked over time, highlighting differences in how end-users and subject matter experts are involved, and how the changes have been managed and evaluated. The challenge for the organisation is whether there can be a clearer understanding of how human actions, decisions and technological factors are connected – particularly postimplementation, and the extent to which the company can truly embed changes to its systems and processes – especially when the organisation faces huge political pressures to reduce costs, deliver more work with less funding, and improve operational performance. This paper focuses on research to understand the extent to which a systems approach is evident in the implementation of the two major change initiatives, and the impact of these on frontline staff behaviour; engineering the right change culture. It builds on research work previously published [4], but also offers new insights following additional observations, interviews and survey work. This wider study seeks to understand the human factors associated with the people who are in the best position to appreciate what is needed to implement change within this type of complex work system, particularly the frontline. The work has helped to explore the gaps between the ‘work as imagined’ described in the formal rules and procedures, and referenced routinely in the interviews undertaken, and the ‘work as done’ that emerged from the longitudinal study, and the surveys of the workforce Hollnagel et al. [5].

2 Method of Investigation 2.1

Research Activities

A longitudinal study to track progress has been developed to observe change programme board reviews, and gather and analyse safety data and metrics, incident report recommendations, training evaluation and feedback. Interviews have also been undertaken with a number of senior and middle managers and practitioners across Network Rail and its supply chain, as well as a survey of frontline staff, to develop a picture of the socio-technical system and specific aspects of the change culture. 2.2

Observation of Change Programme Boards

Changes in programme-specific change boards have been observed over a three-year period (e.g. looking at milestones, programme risks and issues, system and process change requirements and funding implications). The researcher has also reviewed

252

M. Nolan-McSweeney et al.

internal organisational reports on the lessons that have been learned in relation to these change programmes. Changes in the programme boards have also been examined (e.g. their constitution, changes in personnel, programme sponsors, and required governance arrangements. 2.3

Interviews with Senior and Middle Managers, and Practitioners

In-depth interviews have been conducted with 16 selected individuals – senior and middle managers in the rail industry as well as practitioners involved with, or impacted by, the change programmes1. The aim of each interview was to develop understanding of the two change programmes, from the perspectives of those having knowledge and experience of the content and effectiveness of their implementation during a lengthy period of roll-out. Interview questions included the following: • Examples of organisational goals/objectives and priorities, and views on accountability and responsibility, and any boundaries; • Examples of their understanding of key interfaces, complexity, work flows, capability and risk management, and the need for any trade-offs; and • Views and perceptions of how change is managed within a complex sector such as rail and the impact on the employees, structure, funding (e.g. organisational change, culture change, programme change) and relevant learning from this. 2.4

Survey of Frontline Staff

The interviews undertaken helped identify the need for survey work, to further understand how effective the implementation of the change programmes was in engineering the right type of change to culture, so that the safety systems could work in a complex GB Rail industry. A survey was undertaken over a one-month period with 4 groups of frontline staff: (1) responsible managers involved in ensuring the delivery of safe work, (2) the workforce involved in the execution of tasks on the rail infrastructure, (3) persons in charge of activity (individuals responsible for overseeing work on the worksite(s)), and (4) those individuals involved in the planning of the work. The number of survey responses totalled 1355 across the 4 groups and reflects c. 2.5% of the workforce involved in planning and delivering safe work in a ‘normal’ week (noting very high peaks in workload come over bank holidays and weekends, which the researcher chose to avoid).

1

Ten interviews were undertaken in earlier research [4]; an additional six have been concluded in this later study as implementation of the changes continued over a 12-month period.

Engineering the Right Change Culture in a Complex (GB) Rail Industry

253

The survey participants were approached directly using an online survey, and the responses were collated from those willing to participate and give feedback. Their responses were not identifiable to them as individuals, only in their work group domain, e.g. responsible manager, or planner. A total of forty-four questions were used in the survey; nine identical questions were asked of the four different frontline staff groups, and included questions around: • Whether there had been challenges with team workloads since the changes in processes for planning safe work had been introduced; • Specific examples, if any, of areas of non-compliance with planning safe work; • If individuals felt safe when working on/near the (railway) line since the safe work packs were introduced, and the extent to which briefing(s) were undertaken. Answer choices ranged from simple yes/no responses, to several choices to some questions, e.g. strongly agree through to strongly disagree, or to give a scoring range when asked about a frequency of an activity in a typical week. The benefits of the survey meant some elements of the data could be triangulated with the interview responses around safe working practices, and briefings to give a fuller picture of how effectively the major change initiatives have been implemented. 2.5

Method of Conducting and Analysing the Data from the Research

This longitudinal study through data analysis and programme board observations initially, has allowed the researcher to understand the primary aims of the change initiatives, which also helped inform the interviews and survey work. Thus, the interviews used a number of common topic areas to guide the discussion, e.g. safety leadership, change management processes, decision-making authorities, technology introduction, and change programme goals. The different functions and responsibilities – across track maintenance and construction project teams, and organisational levels – were used to explore different perspectives of the complexity of the rail industry. Five themes from Rasmussen [6] (see Fig. 1) were used to classify responses to the interview questions [4]. These themes were used in the previous study to help examine safety leadership from a systems perspective, and the impact on attitudes, behaviours and organisational performance. They are used again here to help capture how people cooperate in their use of tools, technology, and processes to complete what is needed during real work contexts to achieve organisational goals. The themes also reflect business and regulatory constraints for Network Rail, such as those relating to the work system, complex communication structures and information flows, and differing boundaries of acceptable performance.

254

M. Nolan-McSweeney et al.

Objectives: are objectives and values with respect to operational as well as safety issues properly communicated within the system? Status information: are individual decision makers (staff, management, regulators) properly informed about the system status in terms comparable to the objectives? Are the boundaries of acceptable performance around the target state 'visible' to them? Capability: are decision makers competent with respect to the functional properties of the organisation, of the technical core and the basic safety design philosophy? Awareness: are decision makers prompted to consider risk in the dynamic flow of work? Are they - during normal work - made aware of the safety implications of their business decisions? Priorities: are decision makers committed to safety? Is management, for instance, prepared to allocate adequate resources to maintenance of defences? Does regulatory effort serve to control management priorities?

Fig. 1. Themes related to information available to decision makers and capability of control [Rasmussen]

3 Results 3.1

Findings from the Research Activities

The early observations of the two change programme boards established that they focus on workforce safety, especially planning and removing error. The subsequent interviews and survey work highlighted that visibility of the objectives of the change programmes was important to those likely to be affected. Many were clear that the intended ‘vision’ needed to be communicated to afford people the opportunity to commit to the changes (examples included “wanting to feel part of the change”, and “having a sense of purpose”). A recurring theme across the interviews – both in the earlier study, and this more recent research – was that people seem to resist change and fear what is unknown. This echoes with conclusions within organisational reports, specifically for the PDSW change programme [7], where it was concluded that changes need to be relevant for local requirements. Table 1 reflects some of the emerging research themes and the socio-technical aspects that have been found to influence the design and implementation of the new systems and processes in Network Rail. Views have been expressed regarding overly ambitious plans, insufficient attention to what people do in practice, and the extent to which change is possible within a complex system. The table is a brief summation of some of the observations undertaken, and interview and survey outcomes, and reflects on what may be the implications for engineering the right change culture in a complex industry, such as that found in Great Britain’s railways.

Engineering the Right Change Culture in a Complex (GB) Rail Industry

255

Table 1. Summary of emerging research themes and main findings Emerging research themes 1. Social system: personal imperatives, engagement, job change, interactions, autonomy, an acceptance of change, and compliance

2. User-Influenced Design, Usability and Capability: scope and involvement, incentives, competence, levels of responsibility, trust in the system

3. Technical System: Technology (Infrastructure), work processes, workflows, information flows, system access

4. System Integration and Systems Development: multiple users, accessibility, security

5. Evolving Technologies: artificial intelligence, cloud computing, computer supported cooperative working

Main findings 1. Interviews and survey results noted the need for the right kind of culture and behaviours in the social system (organisation) Interviewees mentioned the need for more inclusive engagement and communications; survey feedback echoed this, and the need for a greater understanding on users when change occurs 2. The issues emerging from the research highlighted complicated work processes because of multiple users and interfaces, an initial distrust in new planning tools, and userinfluenced design being considered belatedly 3. The surveys and interviews reinforced that new technologies for planning safe work were complex, and in computerising processes the human experience was overlooked, particularly regarding workflows, system access and workload 4. The research shows that Network Rail did not allow the system requirements to evolve; and were too prescriptive. Allowing users to be involved to shape/influence the design concept was only acknowledged at a later stage of system roll-out 5. Programme boards highlighted that technology design should balance human factors and technology; users need new skills to make best use of the tech

Examples quoted “Making changes is hard; we are poor at communicating our vision. There is a deeply embedded fear of change we have to overcome if we are to be more dynamic in delivering our goals”

“We need to develop mutual trust across the organisation and industry as the basis for a positive safety [performance] culture”

“We expect too much of some of our people. Put the change programmes together and it exposes the enormity of the task for our frontline staff”

“We should have delayed decisions on how parts of the system were implemented, rather than leap to a design we found we couldn’t work with or staff trust”

“No-one asked how we plan, so the new tools don’t reflect the realities of complex work”

256

M. Nolan-McSweeney et al.

3.2

Views on the Social System

An earlier study [4] found that leaders and employees need better skills and abilities to manage change in this type of complex socio-technical railway system. Interviewees, both in the previous study and this current research, found that regular engagement is needed with those affected by change, in order to ensure a smooth transformation and so that those affected by change will know that they will be supported. There is also a need for personal accountability and developments in competence. Most interviewees believed there were genuine efforts to improve safety, but these were often attempted too quickly. Four of the six more recent interviews emphasised individual’s concerns about the scale of the changes required to underlying processes, the organisation (social system), and new technology introduction, noting that the organisation did not have the resources or capability that was needed for successful change. Many of the sixteen interviewees thought that each of the programmes should have a clearer specification of work packages; delivered locally in a targeted area. Following successful review of lessons from this local implementation, elements of the programme could then be implemented more widely in the organisation. This was mentioned even more so in the later interviews, with middle managers stating that job design factors meant a number of structural changes were needed but these took too long to implement and were then not given time to ‘bed in’. Several senior managers had concerns about the ‘big bang approach’ particularly related to the scope of the programmes and the job changes this entails. This phrase had also been identified by the researcher in observations of programme boards in a preliminary study for the research. Later interviews of some programme board members also identified that there were fears regarding compliance to rules and procedures, and thus the potential for a significant risk exposure, because of the introduction of new requirements and systems on such a large scale. This concern was also recognised by respondents in the survey work, where they indicated that the lack of communications and/or engagement in the programmes made it difficult to understand their part in delivering the planned changes, and so accountabilities and responsibilities could become blurred. Some expressed worries about ‘inadvertent non-compliance’. 3.3

Views on User-Influenced Design, Usability and Capability

Survey results show that the majority of respondents believe the revised safe work processes introduced make them ‘feel safe’ when working on or near the railway line. However, they also report in high numbers their concerns with the workload increase for planning work, little involvement in the design of the system or the usability of this, and lack of autonomy in decision making (even for routine, cyclical, work). Many of the interviewees – across the research to-date – said that people should be placed more centrally during programme changes. The senior managers recognised that the current operating model – how staff are motivated and carry out tasks – may be different to what is anticipated in the change programmes. Therefore, some people reported through the survey work that they can be marginalised, disengaged and do not want to be involved in change.

Engineering the Right Change Culture in a Complex (GB) Rail Industry

257

When a new system has been introduced in the past, Network Rail workers admitted to often feeling rather daunted by it and/or reluctant to use it. This seemed to be the case when the Planning and Delivering Safe Work planning tool was initially introduced – identified through the observations, in interviews, and the survey feedback. Extra training in the use of the system seems to have obviated some of this, and the problem of user take-up has started to shift more to areas of concern around user expectations of both the system and its outputs (e.g. responsiveness/speed of processing data). Workers are increasingly familiar with smart phones and apps, touch screens, and gamification, therefore their user experience and expectations of how they interact with technology has changed in the years since the planning tool was first introduced. This presents a challenge for technology and innovation designers to try and address system user requirements, especially as simple customisation is not always possible on bigger software applications used in the workplace, already recognised by Maguire [8], often because of security restrictions or capabilities of the operating environment/platform. In Network Rail, there is an emerging belief that user capability may present an issue of system usability as the workforce is becoming more tech/system ‘savvy’ and expectations grow around accessibility for all. 3.4

Views on the Technical System

It is worth noting that the ‘technical’ system does not necessarily mean physical technology but things like software, procedures, work flows, and data structures/information flows. In the case of Network Rail, when implementing its two major change initiatives, the programme boards identified that there were a number of sub-system factors to consider within the technical infrastructure (i.e. a new planning tool, ‘bow tie’ risk modelling etc.). However, they didn’t plan for subsequent problems with the software and delays in release of the systems, resulting in knowledge fade within the workforce (after training and documentation was rolled out in anticipation of system delivery). Many interviewees shared their frustration at what they saw as ‘wasted effort’, and some frontline staff said they were reluctant to repeat training. Thus, the lesson learnt was that system practice facilities need to be available online to maintain skill levels, rather than be purely training session-driven. Integration with other systems also needs tighter control – for example, access to track diagrams, so that dependencies are understood at the design and development stage(s). 3.5

Views on System Integration and Systems Development

The interviews and survey work show that integration of current systems with new systems proved problematic for Network Rail in its two major change programmes. For example, established work processes could be disrupted by the new systems introduction (i.e. the planning tool replacing a traditional paper-based systems), but also as a result of user-system access requirements (multiple passwords etc.), complex information flow(s), multiple user roles within the work flow, and attempts at taking off-theshelf technologies and turning them into bespoke systems for rail.

258

M. Nolan-McSweeney et al.

Survey feedback highlighted that managing large amounts of information is a major issue for workers. This can include information that they receive through email, text messages, file transfers, safe work planning packs, standards, rules and procedures, and can have an effect on their decision making because of ‘information overload’ (a term popularised by Toffler [9]). A suggested approach will be for the organisation, in future, to standardise information storage and retrieval activity, but until it does this, many of the interviewees believe there will continue to be an unwieldy array of information as the amount and variety grows. Observation of the programme boards also revealed that fixed requirements written into contracts and a change programme to be implemented at pace, meant that there was not much time to incorporate feedback from early parts of the programme, nor rescope work without incurring financial penalties. 3.6

Views on Evolving Technologies

Whilst developing technologies such as artificial intelligence, information integration, and cloud computing are evolving and becoming part of ‘business as usual’ (BAU) they are still relatively new and not commonplace in everyday tasks associated with frontline rail infrastructure activity. The two major change programme boards were keen to use technological solutions to aid decision-making, plan safe work, provide e-diagrams for site access to rail locations etc. However, developing such systems and tools requires an understanding of ‘work as done’ and although the concept of computer supported co-operative work (CSCW) system is well established in office environments (shared document creation, whiteboards and workspaces) [8], it is unfamiliar territory to trackside workers. The early developments of the technologies for planning and delivering safe work did not, unfortunately, consider that group working requires a critical mass of people to participate (responsible managers, planners, persons in charge, authorisers etc.) with a shared sense of purpose and understanding of job roles. Similarly, the need to have users feeling empowered to seek technical support when faced with system problems (rather than find work-arounds) and recognising that individuals could find themselves restricted by organisational boundaries/hierarchies, thus impinging on their decision making and/or feeling of autonomy.

4 Discussion and Conclusion Building on the earlier research [4], comparing two major change programmes with additional interviews and survey work, has helped to highlight important issues for implementation, especially in relation to integration of human, technical, information, social, political, economic and organisational components. It is apparent from the survey work that PDSW relies on how planning tools are introduced to frontline staff, but also the roles of third parties play in developing these new technologies and deliver the training that is required. The BCR programme initially relied on organisational processes used for undertaking business change, but consultants were brought in to support the ‘Bow Tie’ Analysis work without necessarily understanding Network

Engineering the Right Change Culture in a Complex (GB) Rail Industry

259

Rail’s business. Findings from observation of the programme boards indicated that expertise within the organisation (internal ‘know how’) was needed to effectively analyse and communicate scenarios arising from the risk assessment work, which in turn could support the rationalisation of Standards and the development of appropriate ‘means of control’ documentation. This evolved to also consider role-based capabilities and the use of technology to aid decision making as part of the evaluation method for high risk scenarios. Involving users in systems (and process) design is well established [10, 11], and some key principles are outlined in an ISO ergonomics-related publication [12]. However, Harris and Weistroffer [13] report that design decisions are often made at a strategic level, and in practice users may have little influence on fundamental aspects of design and technical functions. Socio-technical systems theory can help with the design of new technologies and technology-led change [14], producing systems that are better for end users. Certainly, from the longitudinal study and from the extensive range of interviews and survey work carried out, more participation is needed from frontline staff and subject matter experts to support changes that are well designed and can be implemented. The survey work has also reinforced the importance of effective change management programmes, including the need to communicate plans, reasons for change, and develop processes with clear accountabilities. The results indicate that Network Rail has attempted to consider many different components of the system whilst implementing change programmes (for example, how people and their actions can impact on others, the tools and equipment needed for work, technology introduction, systems integration, and information flow). The findings also show that as individuals have evolved their experiences in using technology, their confidence in and adoption of new systems is growing rapidly. The rail industry will continue to change. The magnitude of the change and the pace of change is evident in these two programmes. The organisation will need appropriate methods to monitor the timings and progress of the implementation of change, in order to build the assurance that is needed in demonstrating that change programmes have realised their intended benefits. Further research is proposed, to identify the tools to help predict how future interventions in the change programmes might manifest themselves, e.g. the effects of new technology, or factors outside of the organisation’s control. One way will be to explore the research findings to-date using existing system modelling methods (e.g. FRAM or STAMP) [5, 15]. This approach could be used to consider how safety can be designed into systems from the beginning of the concept development process and using analysis to derive the functional safety requirements and design process. The selected model could be linked to specific events, in this case the two business change programmes, and would afford the opportunity to reflect on how things have changed over time and the factors that influenced this, being able to predict interventions and ways of depicting this, e.g. the effects of new technology introduction, or external influences such as Government policy. The findings from this present study are valuable in understanding the range of attitudes, aspirations and perceived constraints towards change, from the perspective of different levels of management and practitioners within the GB Rail sector. The further work that is proposed will contribute to a

260

M. Nolan-McSweeney et al.

socio-technical based description of rail engineering, helping to understand how human actions, decisions and technological factors can interact in achieving successful work. This will assist Network Rail in developing strategies to engineer the right change culture in an already complex industry.

References 1. Mumford, E.: Socio-technical systems design: evolving theory and practice. In: Bjerknes, G., Ehn, P., Kyng, M. (eds.) Computers and Democracy, pp. 59–77. Avebury, Aldershot (1987) 2. Trist, E.: The evolution of socio-technical systems – a conceptual framework and an action research program. In: Van De Ven, A., Joyce, W. (eds.) Perspectives on Organisational Design and Behaviour. Wiley Interscience, August 1981 3. Reiman, T.: Understanding maintenance work in safety-critical organisations – managing the performance variability. Theoret. Issues Ergon. Sci. 12(4), 339–366 (2010) 4. Nolan-McSweeney, M., Ryan. B., Cobb, S.: The challenges and strategies for an effective organisational structure in a complex rail socio-technical system. In: Human Factors Rail Conference, London, UK, 6–9 November 2017 5. Hollnagel, E.: FRAM: The Functional Resonance Analysis Method: Modelling Complex Socio-Technical Systems. Ashgate Publishing Ltd., England (2012) 6. Rasmussen, J.: Risk management in a dynamic society: a modelling problem. Saf. Sci. 27, 183–213 (1997) 7. Network Rail: PDSW Tranche 1 Stakeholder Lessons (2017) 8. Maguire, M.: Socio-technical systems and interaction design – 21st century relevance. Appl. Ergon. 45(2), 162–170 (2014) 9. Toffler, A.: Future Shock. Bantam Books, New York (1971) 10. Eason, K.: The process of introducing information technology. Behav. Inform. Tech. 1(2), 197–213 (1982) 11. Damodaran, L.: User involvement in the systems design process – a practical guide for users. Behav. Inform. Technol. 15(6), 363–377 (1996) 12. ISO 9241-210. Ergonomics of Human-System Interaction – Part 210: Human-Centred Design for Interactive Systems. International Organisation for Standardisation, Geneva (2010) 13. Harris, M.A., Roland Weistroffer, H.: A new look at the relationship between user involvement in systems development and systems success. Commun. Assoc. Inform. Syst. 24 (42), 739–756 (2009) 14. Baxter, G., Sommerville, I.: Socio-technical systems: from design methods to systems engineering. Interact. Comput. 23, 4–17 (2011). https://doi.org/10.1016/j.intcom.2010.07.003 15. Leveson, N.: Engineering a Safer World: Systems Thinking Applied to Safety. MIT Press, Cambridge (2012)

Application of Cognitive Work Analysis to Explore Passenger Behaviour Change Through Provision of Information to Help Relieve Train Overcrowding Jisun Kim(&), Kirsten Revell, and John Preston Transportation Research Group, Faculty of Engineering and the Environment, University of Southampton, Southampton, UK {J.Kim,K.M.Revell,J.M.Preston}@soton.ac.uk

Abstract. It is unrealistic to expect rail passengers to experience a comfortable journey while travelling in crowded trains. Given that passenger behaviour is one of the contributing factors of crowding, understanding and promoting changes in their behaviour would help moderate overcrowding. Therefore, this study aims to develop strategies to encourage passenger behaviour change. Focusing particularly on the provision of train occupancy information, Cognitive Work Analysis (CWA) is applied to gain a systematic understanding about constraints of the behaviour in the rail system environment. Participant observations, staff interview, and online survey data were used to develop an Abstraction Hierarchy (AH), which was validated with two rail subject-matter experts. The output enhance our understanding about passenger behaviour while travelling in crowded conditions, and provide insights about how rail service providers could better assist passengers’ decision making to inform real behaviour change. The AH provides the foundation for how to reduce crowding by supporting passengers’ decision making so they can select less crowded trains or carriages. Keywords: Human factors
Cognitive work analysis Decision making
Information




Behaviour change




1 Introduction It is well known that rail passengers’ travel satisfaction and comfort could be negatively influenced by overcrowding [1–4]. Crowding discomfort is explained by not being able to sit, and not having enough space when travelling on train [5]. Crowding is caused by not only growing rail demand and insufficient supply, but passenger behaviours [6, 7]. This shows the need to expand related infrastructure, and also to encourage changes in passengers’ behaviour to alleviate overcrowding issues more effectively. Focus on both hard and soft methods is therefore required to maximize capacity of the current rail system capacity [8]. Given that it is lengthy and costly to expand and implement track capacity and rolling stock [5, 9], this study focuses on passenger behaviour change to develop possible solutions. © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 261–271, 2020. https://doi.org/10.1007/978-3-030-20503-4_24

262

J. Kim et al.

A number of studies have shown that passengers are likely to change their behaviour to be able to travel on less crowded trains by referring to occupancy information. Preston et al.’s stated preference survey result demonstrates that when passengers were informed about availability of seats and levels of crowding of arriving trains, they intended to wait longer for the train [10]. Information about predicted in-vehicle occupancy and waiting passengers at stops could be beneficial to distribute passengers more evenly across vehicles [11]. In Nuzzolo et al.’s study, crowding information helps passengers make better travel decisions on departure times, boarding stops, and vehicles especially when travelling in crowded conditions. Additionally, “fail-toboard” events were decreased [12]. Drabicki et al.’s simulation study explains the effects of provision of real-time (RT) occupancy information in public transport (PT) systems. The result presented that en-route path choices are linked to penetration rates of the information: how accessible the information is, and types of the information: smoothed over recent runs (shows a better effect on passenger distribution) or captured from the most recent run [13]. Ahn et al.’s simulation study presents carriage occupancy information has a positive impact on spreading passengers more uniformly on the platform. When the information was provided, more equal numbers of passengers per door were recorded, however when it was not, patterns of passengers waiting near the platform entrance were seen. Furthermore, the information resulted in more balanced loads among carriages [14]. Fukasawa et al.’s study discusses that passengers’ tendency in selection of train to board was affected by information on estimated train arrival times and crowding. It was verified by the different tendencies to choose specific carriages. For example, they tended to select one train before receiving the information, however they chose several trains which could better suit their various needs after being exposed to the information [15]. Therefore, this work provides the foundation to help answer the question: how can the provision of information be used to encourage changes in passengers’ behaviour to reduce overcrowding. CWA was used to gain an understanding about constraints of passengers’ behaviour in the rail systems because it enables identifying and representing various “goal-relevant constraints” with the purpose to achieve an ultimate goal of the system [16]. This systematic approach is useful to suit the study purpose for three reasons. First, it allows analysing and modeling complex and dynamic sociotechnical systems, such as rail system environment [17]. Second, it is applicable to open systems in which unpredictable interruptions can affect the performances [16]. Those of rail systems can often be negatively affected by unexpected circumstances such as disruptions, crowding that cannot be easily predicted [18]. Third, it can be applied to develop systems which do not exist because the technic focuses on possible behaviour rather than existing behaviour [16]. This motivates the consideration of features under development in the analysis, such as RT occupancy information on a mobile App. In this study, Work Domain Analysis (WDA), the first phase of CWA was applied because it is the most commonly chosen phase. Owing to its usefulness, it has been extensively used in the previous studies: assessing design proposals [19], designing interfaces [20], and developing decision-aids [21]. Abstraction Hierarchy (AH) is applied to conduct WDA because it is an essential tool to practice and represent WDA [20]. This paper describes the robust development and validation of the first phase of CWA. The output is described in detail with reference to literature, and provides a

Application of Cognitive Work Analysis

263

strategic approach to promoting passenger behaviour change through provision of occupancy information. In addition, strategies focusing on two types of behaviour change are also discussed: (1) to select a less busy carriage by moving along the platform, and (2) to select a less busy train departing earlier or later as actions to have better experience with seated travel. This provides the foundation for ongoing work to device strategies to reduce overcrowding on trains.

2 Methodology 2.1

Introduction and Development of Abstraction Hierarchy (AH)

AH is employed because it is an essential tool to practice and represent WDA [20]. It is comprised of five different levels from highest to lowest: (1) Functional purpose: indicates goal-related purpose of the system affecting all the lower levels; (2) Values & priority measures: explain how well the system works to achieve the functional purpose; (3) Purpose-related functions: link up physical elements and their affordances positioned at the lower levels to more abstract functional components at the higher levels; (4) Object-related processes: show what physical elements listed below afford and function; and (5) Physical objects: present physical elements (man-made or natural) constituting the system. Nodes at each level are connected with means-ends-links between adjacent levels, and the links and linked nodes enable the evaluation of “what the system is capable of, and what the system purpose is”. The connected nodes at the level above account for why a certain node needs to be accomplished, and the linked nodes at the level below tell how a specific node is fulfilled [16]. AH was developed with the Vicente’s guidance of an AH representation [22]. Various secondary and primary resources were used to identify nodes and links. First, literature on train overcrowding and its influences on passengers’ experience was reviewed. Insights were gained to define nodes at Level 2 which enable to accomplish the goal. Primary data was reviewed thoroughly, and relevant segments or words representing nodes and links were colour-coded, and positioned at appropriate levels. First, participant observation data gathered on the 28th March, 2018 at London Victoria, Gatwick Airport, and Brighton stations, and on the 9th May, 2018 at Gatwick Airport station was used. The purpose was to explore passengers’ travel behaviours and the constraints by being immersed in the real settings. The observation notes were reviewed to identify physical objects, the processes, and effects on passengers’ activities to organise nodes and links at Levels 3, 4, and 5. Second, questionnaire data discussing the effect of occupancy information on passengers’ decision-making and intention to change behaviour related to travelling on crowded trains was used to determine nodes and links throughout the AH. The findings gave insights to find out what and how physical objects assist passengers to make best use of the service and created values of seated travel. The questionnaire was administered in June and July, 2018. Third, staff interview data (with on-board supervisors, platform staff, area station manager) collected on the 27th September, 2017 furnished information to identify nodes and links through the AH. They served as supporting information that would be difficult to gain through observation or literature alone, as they related to interactions

264

J. Kim et al.

between staff through internal channels. Fourth, two Subject Matter Experts (SMEs) in the rail domain took part in the AH validation process. The interviewees reviewed: nodes at all levels; links between levels 1 & 2, and 2 & 3; nodes connected to behaviour changes in relation to selection of an empty carriage by moving along the platform, and of a less busy train departing earlier or later. The first interviewee supplied information based on experience in managing rail service staff, and customer service from a service provider’s, and a customer’s standpoints. The second interviewee provided comments based on experience and expertise in IT industry in the rail domain, and from a rail user’s point of view.

3 Results In this section, developed AH will be proposed (See Fig. 1). Explanations will be given about each level, and justification for selection of nodes at individual levels in sections from 3.1 to 3.5. Further analysis will be conducted around ‘to enable passenger behaviour changes’ to develop strategies in the following section (See Fig. 1, Level 3, Node 5).

1 Functional Purpose 2 Values & priority measures 3 Purposerelated functions 4 Objectrelated processes

5 Physical objects

Potential behaviour changes Nodes linked to the behaviour changes Nodes not selected in discussion on strategy development in section 4

Fig. 1. Nodes in AH (Simplified)

3.1

Functional Purpose (Level 1)

The ultimate goal of the systems is set as to enhance passengers’ travel experience within the scope of satisfaction and on-board comfort through seated travel (See Level 1, Nodes 1 and 2). Space availability is one of the most significant factors linked to attractiveness of PT [1]. Understandably, crowded rail services seem less appealing to

Application of Cognitive Work Analysis

265

customers [23]. Not surprisingly, passenger overcrowding is one of the major concerns to them. In the dense condition, insufficient space for sitting and standing is one of the main causes of discomfort [1, 3, 5]. In other words, mitigation of passenger crowding is needed to improve of passenger experience. Regarding passengers’ valuation of seat availability, most passengers would decide to sit when many seats are available, and they prefer to sit although some are happy to stand when making a short train journey [3, 24]. In some cases, passengers travel backwards several stations to avoid overcrowding and find a seat on the Underground train [25]. As reviewed, being able to sit has an effect on passengers decision making and subsequent behaviours, this study places emphasis on having a seat as a facilitating factor, and identifies satisfaction and comfort with seated travel as functional purpose in AH. 3.2

Values and Priority Measures (Level 2)

Nodes at this level have been identified by reviewing literature pertaining to customer experience about travelling on crowded trains. Node 1, on-board stress on crowded trains is elicited by insufficient room due to high passenger density [26]. The difference between crowding and density is elucidated as crowding is a perception results from “interplay of cognitive, social, and environmental factors” in dense environment, and density is “objective physical characteristics” [5]. Crowding, in this sense, is related to experience of exhaustion and stress on board [27]. Node 2, perceived risk to personal safety and security is considered as consequences of travelling on crowded trains [5, 28, 29]. The environment is perceived less safe and comfortable while travelling with other passengers due to high density [28]. Additionally, crowding is a potential predictor of perceived risk of personal safety, for instance, against accidents and crimes [5]. Node 3, value for money is clarified that the degree of perceived value in comparison to the money spent on purchasing the ticket. They perceive that standing on overly crowded trains, and standing for longer than 10 min are of reduced value [9]. If that is the case, seated travel could be linked to improvement in perceived value for money. Node 4, passenger boarding and alighting times linked to dwell time are partly determined by the level of passenger density. As observed at Gatwick Airport station where passenger density was higher near the platform entrance than other areas of the platform, boarding and alighting were slower in the crowded area. Therefore, more uniformly distributed passengers along the platform and carriages might contribute to reducing vehicle dwell time, and in turn, enhancing timeliness and punctuality ultimately at the network level. Node 5, although train is considered as potentially the most productive mode of transport because users can do secondary tasks: reading, working with laptop computers, or using a mobile phone [30], productivity and utility of invehicle time could be decreased by train crowding [1, 24]. Further, productivity and efficiency are negatively associated with high passenger density [5]. Node 6, a sense of control is explained as individuals’ perception that they have the power to have an influence on decision-making to see “desirable consequences and a sense of personal competence” in the given situation [31]. It is particularly important to have the sense in crowded conditions because it might have a stress buffering effect. Also, it is more likely for those who lack the perception to find the high-density environment more stressful and crowded. Additionally, availability and quality of options to choose also

266

J. Kim et al.

matter to passengers’ sense of control [5]. Node 7, higher density could be represented as a deficiency of space, and this increases the likelihood of unwanted social and physical interactions with fellow passengers [26]. Although the same number of passengers travel at the same time of day, if passengers choose a less busy service: other carriage, train, mode, biased passenger distributions across them may be reduced, and personal space may be expanded. Node 8, it is presumed that seat availability is one of the considerations to choose a departure time and plan a route. This was inferred from data presenting that some travellers are likely to wait longer to board the next train in order to get a seat rather than using the first departing train. Thus, it is suggested to offer information about the probability of getting a seat at boarding points that influences passengers’ decision making [32]. Node 9, passengers often experience invasion of privacy while travelling on crowded trains. Privacy is particularly valued by private mode users [33, 34], and its invasion is mostly related to having to sit close to fellow passengers [2, 35]. 3.3

Purpose-Related Functions (Level 3)

Identified nodes at this level are enablers to accomplish the values described above. They were mainly defined by reviewing the interview transcripts and observation notes. A prompt, ‘what do these object-related processes afforded by relevant physical objects influence or benefit users?’ was used to find out words, phrases, sentences, or paragraphs that could provide information for answering the question were highlighted. At this level, key nodes are presented as to enable passengers to: Node 1, react to risks/hazards: assess the probability of occurrence of an accident during alighting, boarding, or waiting such as slips, falls, assault, train accident, and risk of injury [36]; Node 2, access to discount: special offers, and cheaper off-peak tickets [37]; Node 3, weigh up opportunity cost: in relation to optimal choice of travel time and comfort (slower seated and quicker standing journeys), alternative route or mode [33]; Node 4, use and access to necessary information easily (e.g. occupancy, route and disruptions); easiness of use for passengers to fulfill their journey goal including getting a seat [38]; Nodes 5, change behaviour: taking a less crowded alternative mode [5, 39], earlier or later train by arriving earlier or waiting longer at the station [10], carriage by moving on the platform [14], and by moving through the train as behavioural reactions to crowding; Node 6, increase likelihood of getting a seat: by reserving seats, or helping passengers find, move to, and acquire a seat more easily in case of turn-up-and-go service; Node 7, help improve passenger flow on the platform and train: including boarding and alighting (unidirectional, bidirectional) [40]; Node 8, get assistance for physically less able users: transporting luggage, accessing to facilities or necessary information [41]. 3.4

Object-Related Processes (Level 4)

Identified nodes at this level explain how the functions presented above can be enabled. They were discovered through reviewing the observation notes, interview transcripts, and survey data. A prompt used to identify relevant concepts was ‘what do physical objects in rail systems perform and afford?’. Additionally, passengers’ reactions to or

Application of Cognitive Work Analysis

267

interactions with the physical objects were paid attention to. Key nodes are presented as to: Node 1, provide and communicate information: including general or train specific information about route, departures/arrivals, disruptions, occupancy; Node 2, help estimate seat availability: appraisal involved in the process to estimate where passengers can find available seats more easily; Node 3, support navigation: for passengers to locate themselves on the platform or in the station in relation to the carriage to board; Node 4, enable financial transaction: for such as buying and modifying a ticket; Node 5, provide offerings and discounts; Node 6, mitigate passenger density on the platform: areas around platform entry are likely to be more crowded; Node 7, support psychological needs: to feel more relaxed when handling luggage, moving or waiting on the platform; Node 8, support physiological needs: to feel more comfortable when waiting or moving in the station and on the environment. 3.5

Physical Objects (Level 5)

Nodes at this level account for what physical components afford the selected objectrelated processes presented above. They were determined by reviewing the observation notes and interview transcripts, as well as online resources [42]. Key nodes are classified as: Node 1, information systems: information displays scattered in the station/train, information desks, ticket offices/gates, and users’ devices: mobile phones; Node 2, mediums of information: websites, signage, adverts, announcements, and printed materials; Node 3, physical facilities/environments; Node 4, personnel; Node 5, other passengers in the station/train; Node 6, manageable luggage. These physical components will enable object-related processes that will promote changes in passenger behaviour. This will in turn enable passengers to enhance passengers’ perceived value of rail services that will lead to improvement of passenger experience of travel, and alleviation of overcrowding. This means development and enablement of the physical elements in the system will need to be accomplished to achieve the ultimate goal. Also they will provide opportunities to apply strategies to promote passenger behaviour change through provision of information described in the following section.

4 Discussion and Conclusion This study aims to develop strategies to promote passenger behaviour change to use less busy rail services with provision of occupancy information as a facilitator. CWA was used as a method to investigate behaviour-shaping constraints in the rail service environment. In AH development process, a variety of secondary and primary resources was used to identify nodes relevant to help achieve an ultimate goal of the system from a user’s point of view: improvement of travel experience including satisfaction and comfort. Strategies are proposed through the analysis conducted centring around: (1) selection of a less busy carriage on the platform (See Fig. 2, Level 3, Node 1), and (2) selection of a less busy train departing earlier or later (See Fig. 3, Level 3, Node 1).

268

J. Kim et al.

1 Functional Purpose

2 Values & priority measures

3 Purposerelated functions

4 Objectrelated processes

5 Physical objects

: Potential behaviour change : Nodes not linked to the behaviour change

: Nodes linked to the behaviour change

Fig. 2. Nodes linked to ‘Choose a less busy carriage by moving along the platform’

1 Functional Purpose

2 Values & priority measures

3 Purposerelated functions

4 Objectrelated processes

5 Physical objects

: Potential behaviour change : Nodes not linked to the behaviour change

: Nodes linked to the behaviour change

Fig. 3. Nodes linked to ‘Choose a less busy earlier/later train’

Application of Cognitive Work Analysis

269

For both cases, effective communication of information will be helpful for passengers to find out, and choose which train to board that better suits their requirements during the travel. Provision of related information about RT carriage/train occupancy, and service: train departure/arrival, and disruptions is requried. It will be even better if passengers can ask for, and get customised information. This seems useful for passengers in the decision-making involving planning and modification stages. In the first case (See Fig. 2. Level 3. Node 1), support of physical needs: locomotion at the platform and station enabled by easier transport of luggage, movement seems necessary. This type of help is vital for physically less able passengers. In addition, assistance in navigation at the platform and station, and mitigation of passenger density at the platform should be performed for them to easily move. In the second case (See Fig. 3. Level 3. Node 1), promoting offerings and discounts could be motivating to change their travel plan to take less busy train services especially for time-flexible passengers. If this decision is made en-route, financial transaction needs to be allowed to purchase and modify a ticket preferably online. Facilities at the station and platform should be able to accommodate waiting passengers’ physical and psychological needs (for safety, com fort, and reduced boredom). This approach is meaningful because few studies have attempted to understand changes in passengers’ behaviour about seated travel in the frame of customer experience in crowded conditions with considering behaviour-shaping constraints. It is hoped that findings from this analysis will give insights to rail service providers to enhance the environment to better support passenger behaviour changes which will lead to improvement of travel experience, and alleviation of overcrowding. Acknowledgements. The authors would like to thank all the participants for their valuable opinions and collaboration. We also acknowledge the funder of this project, the Rail Safety and Standards Board for their support.

References 1. Haywood, L., Koning, M., Monchambert, G.: Crowding in public transport: Who cares and why? (2017) 2. Pel, A.J., Bel, N.H., Pieters, M.: Including passengers’ response to crowding in the Dutch national train passenger assignment model (2014) 3. Passenger Demand Forecasting Council (PDFC), Passenger demand forecasting handbook (2017) 4. Cantwell, M., Caulfield, B., O’Mahony, M.: Examining the factors that impact public transport commuting satisfaction. J. Public Transp. 12(2), 1 (2009) 5. Cox, T., Houdmont, J., Griffiths, A.: Rail passenger crowding, stress, health and safety in Britain. Transp. Res. Part A: Policy Pract. 40(3), 244–258 (2006) 6. Network Rail 2017: Railway Upgrade Plan 2017/18, London (2017) 7. Rail Safety and Standards Board (RSSB) 2018, Crowd management on trains: a good practice guide (2018) 8. Toriumi, S., Taguchi, A., Matsumoto, T.: A model to simulate delay in train schedule caused by crowded passengers: using a time-space network. Int. Reg. Sci. Rev. 37(2), 225–244 (2014)

270

J. Kim et al.

9. Douglas, N., Karpouzis, G.: Estimating the passenger cost of train overcrowding. In: 29th Australian Transport Research Forum (2006) 10. Preston, J., Pritchard, J., Waterson, B.: Train overcrowding: investigation of the provision of better information to mitigate the issues. Transp. Res. Rec.: J. Transp. Res. Board 2649, 1–8 (2017) 11. Cats, O., Koutsopoulos, H.N., Burghout, W., Toledo, T.: Effect of real-time transit information on dynamic path choice of passengers. Transp. Res. Rec. 2217(1), 46–54 (2011) 12. Nuzzolo, A., Crisalli, U., Comi, A., Rosati, L.: A mesoscopic transit assignment model including real-time predictive information on crowding. J. Intell. Transp. Syst. 20(4), 316– 333 (2016) 13. Drabicki, A., Kucharski, R., Cats, O., Fonzone, A.: Simulating the effects of real-time crowding information in public transport networks. In: 2017 5th IEEE International Conference on Models and Technologies for Intelligent Transportation Systems (MT-ITS), p. 675. IEEE (2017) 14. Ahn, S., Kim, J., Bekti, A., Cheng, L., Clark, E., Robertson, M., Salita, R.: Real-time information system for spreading rail passengers across train carriages: agent-based simulation study. In: 38th Australasian Transport Research Forum (ATRF) 2016, Melbourne, Victoria, Australia (2016) 15. Fukasawa, N., Yamauchi, K., Murakoshi, A., Fujinami, K., Tatsui, D.: Provision of forecast train information and consequential impact on decision making for train-choice. Q. Rep. RTRI 53(3), 141–147 (2012) 16. Jenkins, D.P.: Cognitive Work Analysis: Coping with Complexity. Ashgate Publishing, Ltd., Farnham (2009) 17. Wilson, J.R.: Fundamentals of systems ergonomics/human factors. Appl. Ergon. 45(1), 5–13 (2014) 18. Cats, O., Jenelius, E.: Beyond a complete failure: the impact of partial capacity degradation on public transport network vulnerability. Transportmetrica B: Transp. Dyn. 6(2), 77–96 (2018) 19. Naikar, N., Sanderson, P.M.: Evaluating design proposals for complex systems with work domain analysis. Hum. Factors 43(4), 529–542 (2001) 20. Burns, C.M., Hajdukiewicz, J.: Ecological Interface Design. CRC Press, Boca Raton (2004) 21. Gualtieri, J.W., Elm, W.C., Potter, S.S., Roth, E.: Analysis with a purpose: narrowing the gap with a pragmatic approach. In: Proceedings of the Human Factors and Ergonomics Society Annual Meeting, p. 444. SAGE Publications, Los Angeles (2001) 22. Vicente, K.J.: Cognitive Work Analysis: Toward Safe, Productive, and Healthy ComputerBased Work. CRC Press, Boca Raton (1999) 23. Lyons, G., Jain, J., Weir, I.: Changing times–a decade of empirical insight into the experience of rail passengers in Great Britain. J. Transp. Geogr. 57, 94–104 (2016) 24. Cats, O., West, J., Eliasson, J.: A dynamic stochastic model for evaluating congestion and crowding effects in transit systems. Transp. Res. Part B: Methodol. 89, 43–57 (2016) 25. Hörcher, D., Graham, D.J., Anderson, R.J.: Crowding cost estimation with large scale smart card and vehicle location data. Transp. Res. Part B: Methodol. 95, 105–125 (2017) 26. Evans, G.W., Wener, R.E.: Crowding and personal space invasion on the train: please don’t make me sit in the middle. J. Environ. Psychol. 27(1), 90–94 (2007) 27. Mahudin, N.D.M., Cox, T., Griffiths, A.: Measuring rail passenger crowding: scale development and psychometric properties. Transp. Res. Part F: Traffic Psychol. Behav. 15 (1), 38–51 (2012) 28. Katz, D., Rahman, M.M.: Levels of overcrowding in bus system of Dhaka, Bangladesh. Transp. Res. Rec. 2143(1), 85–91 (2010)

Application of Cognitive Work Analysis

271

29. Cheng, Y.: Exploring passenger anxiety associated with train travel. Transportation 37(6), 875–896 (2010) 30. Lyons, G., Urry, J.: Travel time use in the information age. Transp. Res. Part A: Policy Pract. 39(2–3), 257–276 (2005) 31. Rodin, J., Schooler, C., Schaie, K.W.: Self Directedness: Cause and Effects Throughout the Life Course. Psychology Press, Routledge (2013) 32. Schmöcker, J., Fonzone, A., Shimamoto, H., Kurauchi, F., Bell, M.G.: Frequency-based transit assignment considering seat capacities. Transp. Res. Part B: Methodol. 45(2), 392– 408 (2011) 33. Wardman, M., Whelan, G.: Twenty years of rail crowding valuation studies: evidence and lessons from British experience. Transp. Rev. 31(3), 379–398 (2011) 34. Tirachini, A., Hensher, D.A., Rose, J.M.: Crowding in public transport systems: effects on users, operation and implications for the estimation of demand. Transp. Res. Part A: Policy Pract. 53, 36–52 (2013) 35. MVA Consultancy 2008, “Understanding the passenger: valuation of overcrowding on rail services”, Report for Department of Transport, London, United Kingdom 36. Rail Safety and Standards Board (RSSB): Risk at the platform-train interface 2013 37. National Rail: Discounts available on National Rail services (2018). http://www.nationalrail. co.uk/times_fares/46506.aspx. Accessed 25 Jan 2018 38. Interaction Design Foundation: What is usability? (2018). https://www.interaction-design. org/literature/topics/usability. Accessed 29 Jan 2019 39. Tirachini, A., Hurtubia, R., Dekker, T., Daziano, R.A.: Estimation of crowding discomfort in public transport: results from Santiago de Chile. Transp. Res. Part A: Policy Pract. 103, 311– 326 (2017) 40. Seriani, S., Fernandez, R.: Pedestrian traffic management of boarding and alighting in metro stations. Transp. Res. Part C: Emerg. Technol. 53, 76–92 (2015) 41. Cardell, J., Idris, S., Wilks, P.: Investigating Assistance Provision to Disabled Rail Users: Rail Human Factors Around the World: Impacts on and of People for Successful Rail Operations, p. 57 (2012) 42. National Rail: Station services and facilities (2018). http://www.nationalrailco.uk/stations_ destinations/. Accessed 25 Jan 2018

Decrease Driver’s Workload and Increase Vigilance Denis Miglianico(&) and Vincent Pargade Alstom, 48 rue Albert Dhalenne, 93482 Saint-Ouen, France {Denis.Miglianico,Vincent.Pargade}@alstomgroup.com

Abstract. Light Rail Vehicle (LRV) driver ability to drive is checked through a device called a deadman. It requires the driver to activate the device periodically to monitor the level of vigilance and to prevent the safety system from triggering the emergency brake. Different tests have been done to evaluate how tram drivers use the device in more or less demanding conditions. The relation between the environment and the frequency of button push is investigated. The current level of workload is analyzed. After the introduction of a new approach including the activation of other controls as an input for the deadman device, the impact on workload and safety are analyzed. Introduction of contactless technology is finally investigated as future improvement of the system. Keywords: Workload


Safety
Deadman

1 Introduction The Deadman device was introduced in the 1920s to ensure that a train would stop automatically in case of incapacitation of a driver (losing consciousness or heart attack for example). Before the emergency brake is applied, the system sounds an alarm to alert the driver in the case of fatigue or drowsiness. Later on, it was a condition to allow the removal of an assistant or a second driver. For example, in France, the decree of 22nd of March 1942 stated that in case of a failure of the driver, the assistant shall be able to stop the train. The assistant can be optional if equipment agreed by French safety authority can stop the train in case of the failure of the driver. The implementation of the first generation of deadman was a button that had to be pressed all the time. The assumption was that in case of drowsiness, the driver will release the button and the braking system is triggered. However the system could be bypassed because some drivers found that the effort to maintain the button pushed could be avoided by fixing the button in the pressed position with a piece of string or a rubber band. To avoid the misuse of the deadman, the system was improved to require the driver to release the button on a periodic basis. Three timers characterize the deadman: T1 is the duration for which the button shall be pressed and T2 the duration of the release and T3 is the duration after which the emergency brake is triggered. If the timer T1 or T2 is elapsed without further action a buzzer is sounded, and T3 s later an emergency brake is triggered. Typical durations are presented hereafter (Table 1):

© Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 272–281, 2020. https://doi.org/10.1007/978-3-030-20503-4_25

Decrease Driver’s Workload and Increase Vigilance

273

Table 1. Three timers characterize the deadman Type of train High speed train Locomotive Metro Tram train LRV LRV

T1 (s) 30/55/60 55 30 30 12 10

T2 (s) 2,5 5 2 2 2 3

T3 (s) 2,5 2 2 2 2 2

2 Physical and Cognitive Workload Assessment 2.1

Task Analysis

We have explained in [1] how task analysis [2] is valuable to introduce innovation. The starting point of the evolution of the deadman function started also with the re-use of the task analysis results. The analysis shows that drivers have to achieve several different objectives (see [3]) which are related to the driver (safety, comfort, respect of the timetable) and to the prescription (respect of the rules). All these tasks require sustained attention. We provided a way to minimize the workload when introducing a new task (saving energy). The objective of the innovation presented in this paper is to decrease the workload of the driver especially in the situation of high workload (i.e. in city center). 2.2

Deadman Implementation in LRV

Different ways of implementing the deadman function are possible (at the driver’s hand level or at the foot level). In an LRV, the usual location of the deadman is on the master controller As the master controller has different shapes, the deadman can be located at different locations (Fig. 1):

Linear,

«T»

« Joystick »

Fig. 1. On a linear master controller, the deadman is above the control. It’s on the side for the “T” shape or the “Joystick” shape.

274

2.3

D. Miglianico and V. Pargade

Physical Workload

The deadman is a mechanical button on the linear type. On the “T” master controller, the deadman was initially a push button. Due to the frequency of the press hold, it appears that the use of such controls combined with other psychosocial aspects may lead to Musculoskeletal disorder. To decrease the physical workload, a new type of deadman has been introduced: a capacitive sensor button. It is placed at the side of the master controller (see red arrow on the “T” and “Joystick”) and decreases dramatically the physical effort required. Another improvement has been brought to the master controller with a new shape for the deadman sensitive button. The circular button has been replaced with an oblong one. It provides way to minimize the finger displacement taking into account the different size of hand breadth across Finger Knuckles (typically the dimension range is 60–90 mm of the 5th percentile female to the 95th percentile male) (Fig. 2).

Fig. 2. Type of the deadman (from circular to oblong)

2.4

Cognitive Workload

According to interviews done with LRV drivers, the activities are split between areas where there are few external stimulations and areas where a lot of information have to be taken into account and several actions carried out. The first type of driving is like a regional train and the second one is close to the city bus. The workload can vary from low (with a risk of drowsiness) to high. An example of these two areas is presented hereafter (Fig. 3). In particular in the City Centre, the driver has to manage different information such as: signaling, pedestrians and their behavior, people with reduced mobility, people with headsets, bicycles, electric scooters, motorbikes, cars. Therefore, drivers report that they perceive that the deadman alarm is ringing more often (T1 and T2 time elapsed). The reason is that drivers focus more on what they consider as more important information (for example people crossing the track). It happens in particular in rush hours (12:00–14:00 and 16:00–19:00).

Decrease Driver’s Workload and Increase Vigilance

275

Fig. 3. Examples of environment

2.5

Observation in Real Environment

In addition, with the management of the traction/brake controller, the driver uses the bell/gong and the horn. The bell/gong is the most used control and in general they leave a right-hand finger on it. Due to rules, it shall be used in the following cases: • • • •

Each time the LRV is starting, When the LRV arrives or leaves a station, When the LRV is crossing a street, When the LRV is crossing another LRV

Therefore, the bell/gong and horn are good candidate to replace the deadman as an indicator of the state of alertness of the driver.

3 The New System 3.1

Description

In order to tackle the main physical and cognitive issues described above, Alstom developed and tested a new implementation of the vigilance function for LRV. This system, called Smart Vigilance (SV) is intended to decrease driver’s workload while increasing their vigilance for safety related tasks. SV is embedded in Alstom next generation tachymetric central for LRV and is an evolution of the logic of deadman function. The principle behinds SV function is to distribute deadman’s function actuation over several actuators at driver’s desk, each of them testifying of driver’s activity, in order to release the obligation of actuating deadman through a single actuator on master controller. SV functioning is a logical “OR” between the following actuators: – Deadman actuator on master controller, – Master controller inclination of more than 5° angle, towards traction, brake or neutral position,

276

D. Miglianico and V. Pargade

– Gong actuation, – Horn, – Deadman’s pedal. Such a distribution is intended to satisfy the need for automatized vigilance monitoring and, at the same time, to reduce the use of the deadman itself. The hypothesis is that the driver will be less obliged to focus on deadman’s actuation through a repetitive gesture on the master controller button, and will therefore experience a workload decrease. This hypothesis has been the subject of the user tests described in the next section. 3.2

The User-Tests

Organization. In order to evaluate the potential benefits in terms of ergonomics of the SV technology, Alstom organized, in close relation with the Montpellier city Operator TAM, a session of user tests of the function on a LRV simulator, supervised by Alstom ergonomists. Hypotheses. First and main hypothesis was: Hypothesis 1: Drivers will be less obliged to focus on deadman’s actuation through a repetitive gesture on the master controller button, and will therefore experience a workload decrease. It was the main mission of Alstom ergonomists to objectify this aspect, in order for the company carry out SV development with confidence, based on evidence collected with the future users, in collaboration with the customer. This hypothesis can be broken into two: Hypothesis 1.1: Drivers will actuate deadman’s button or foot pedal less often, Hypothesis 1.2: Drivers will experience a smaller workload in SV modality than in classical modality. Following hypotheses were also explored: Hypothesis 2: Driving experience and overall satisfaction will be enhanced. Hypothesis 3: SV may induce that driver will use master controller for traction and braking more often, in the sole intention of deadman actuation. Hypothesis 3 is introduced to evaluate the potential side-effects of SV on mechanical solicitation of the rolling stocks. Indeed, a measure that enhances ergonomics and driver’s experience may also penalize rolling stock performance, life cycle and operation costs. Test Environment and Methodology. The tests took place in Alstom facilities in La Rochelle (France), on an LRV simulator representative of Alstom technologies and networks they are operated in. The user test had the following design: Six LRV drivers in exercise and a training instructor participated to the test session over 3 days for a total of 51 runs. Participants were asked to drive an LRV vehicle as per commercial operation in different conditions (urban and peri-urban). Weather and

Decrease Driver’s Workload and Increase Vigilance

277

Tramway adherence parameters were kept constant all over the session in order not to increase the number of independent variables of the test. A set of scenario-based trials were implemented in the simulator: – – – – – – –

Free driving (to get use to the simulator and deadman modality), Potential accident with a car, driving fast and burning a red light, Potential accident with a pedestrian, falling from a LRV stop, pedestrian on the trackway, Future Nice T2 line, from Airport to port, Future Nice T2 line, from port to port, Free driving with pedestrian having hazardous behaviour (crossing the tracks at the last moments, suicidal behavior…). Simulator engineer at training instructor workstation controlled the behavior of the pedestrian via a 3D-avatar function, thus inducing some randomness in the tests’ perturbations.

Each scenario was played in Classical and SV modality in equal number, in a randomized order. Participants began either by C modality scenarios, either by SV ones, evenly distributed. Simulator workstation was coupled with a training instructor workstation recording the following run parameters in real time: – – – –

Deadman button activation on master controller and foot pedal, Gong activation, Horn activation, Master controller’s actuation, in traction or brake, by an angle superior or equal to 5° (SV actuation threshold), – Emergency brake. Each scenario was executed in two modalities: – Classical deadman modality, say, with the deadman function actuated only on master controller’s sensitive button actuation, – SV modality, with deadman function actuated as described in Sect. 3.1. After each modality, each participant was asked – To complete a full NASA-TLX workload assessment on tablet-based application. Participants were given a paper-based translation of the texts. – To complete a Usability Scale Questionnaire (see [4]) translated in French by Alstom ergonomist. At the end of his session, each participant carried a semi-structured interview with Alstom ergonomist, in order to collect immediate feedback, sensations, and testimony on user experience. All debriefing sessions were audio-recorded for further analysis. All participants where mounted with an heart-rate monitoring device for investigation purposes. Heart- rate data were not intended to be treated as part of this experiment and will not be discussed in this article. Results. The analysis of all the data collected during this user test conducts to the following results:

278

D. Miglianico and V. Pargade

Hypothesis 1.1: Actuation of deadman button or pedal. It has been measured that: There is a significant difference showing that drivers use less master controller’s deadman button or deadman pedal in SV modality (S) than in classical one (C). (MEAN (C) = 18, SD = 11, MEAN (S) = 6, SD = 3, double-sided t-test t = 2.89, p value = 0.02). Computation of confidence intervals shows that there is a 95% probability that SV modality reduces deadman button or pedal actuation by 2 to 22 actuations per minute, representing a 10% to 90% reduction of manual activation of the controls dedicated to deadman. Participant have accounted for the following facts in debriefing: – “I almost do not have to press the button anymore” – “I have to do it only in coasting mode1, – “It’s simpler”. Conclusion on number of manual actuation of deadman: Both objective and subjective data shows that SV technology induces a reduction of the use of the deadman actuation button or pedal. Although this result seems straightforward as SV functionality was designed to reach that purpose, it is far from trivial form an ergonomics perspective. It releases a constraint on the driver’s activity, especially related to a repetitive and frequent hand/thumb activity. Literature on ergonomics related to deadman function has extensively argued for the link between the mechanical and technological constraint of repetitive actuation and upper-limbs musculoskeletal disorders in tramway drivers population (see [5]). This fact has also been reported by the TAM drivers who participated in the experiment, although it has been said that this painfulness was reduced by the introduction of the sensitive button technology (as described in Sect. 2.3). Smart Vigilance technology introduces a further progress towards MusculoSkeletal Disorder (MSD) prevention for two reasons. First, it lowers the frequency of actuation of the function, thus reducing mechanical solicitations on drivers’ hand and thumb. Second, the fact that driver can actuate deadman with diversified means allows them to position their hand freely on the master controller. They are not obliged anymore to adopt a stereotyped position, that they think efficient because it is the position they feel the less constraining – but as this position is fixed, it may still induce MSD potentials in the long run. With SV technology, it has been observed by Alstom ergonomist that drivers quickly adopt a relaxed and changing position of the left hand on master controller. After few minutes of adaptation, this new freedom of movement appears, for all participants. It seems reasonable to think that this freedom of hand movement is a good thing towards upper-limb MSD prevention. Indeed it is compliant with the ergonomics good practice stating that “a good position is a position one can change”.

1

Free running of a train with no traction current and no brakes applied (see [6]).

Decrease Driver’s Workload and Increase Vigilance

279

Hypothesis 1.2: Workload decrease. It has been measured that: There is no significant effect in workload assessments detected per NASA-TLX, between the two modalities. (MEAN (C): 54, SD: 14, MEAN (S): 56, SD: 11, t = 0.36, p value = 0.73). Null hypothesis cannot be rejected, therefore the differences in NASATLX scores is not significant. Nevertheless, participant have accounted for the following facts in debriefing: – “It frees my attention, I can concentrate more on driving, on peripheral vision”, “it frees some mental energy”, – “it brings comfort, it’s undeniable”, – “it’s easier, you don’t think about it anymore. Sometimes you realize that you’ve rearmed deadman and it’s enjoyable, you have done two things at the same time (…) it is less constraining” Conclusion on workload: Although NASA-TLX score do not show significant difference between the two vigilance modalities, participants have testified of an overall sensation of reduced cognitive workload. It seems reasonable to think that SV, by removing the need for a background repetitive tasks frees some attentional resources that can be redistributed to more complex, safety-related tasks (according to a model of attention such as Wickens and Holland’s (see [7]). This is particularly true in urban, dense traffic context where: – Driver’s attention is required on safety related tasks, as traffic and people behavior necessitates increased vigilance from the LRV driver, – SV is almost always rearmed by master controller movement or gong actuation, as these actuators are solicited in the driving task itself. In this condition, it is a bit as if the deadman function “disappeared” while still being here. Indeed, for the driver, he doesn’t have to manage it and can focus on driving safely, but for the safety/functional perspective the function is still here, as the tachymetric calculator still monitors driver’s vigilance and presence via his activations of gong and master controller. Only in speed constant, periurban context deadman still remain to be actuated “classically” with the master controller button, as there is no traction/brake or gong control from the driver at this phases. Hypothesis 2: Satisfaction and driver experience. It has been measured that: There is no significant effect on satisfaction ratings as per SUS scores. (MEAN (C) = : 85.5, SD: 14, MEAN (S): 82.8, SD: 11, t value = 0.42, p value = 0.69). Null hypothesis cannot be rejected, therefore the differences in SUS scores is not significant. Nevertheless, participant in debriefing have accounted for increased comfort and surprisingly quick adaptation. Overall satisfaction was high among the participants, seeing this technology as “a good thing”, “especially for newcomers, that will be less focused on deadman and more on safety”. Conclusion on satisfaction: SV was positively apprehended by most of the drivers (6 out of 7 without limits, 1 driver being more reserved on the benefits it could bring to him but nevertheless saying that it would be good for newcomers). As adaptation time

280

D. Miglianico and V. Pargade

is short and learning curve almost immediate, it is reasonable to thinks that SV technology will be quickly and largely accepted by the future population of users. Hypothesis 3: increase in mechanical solicitation induced by SV. Simulator logs and video recording analysis of the driving activity have shown no significance difference in master controller use for traction, braking & shift to neutral position between the two vigilance monitoring modalities. (MEAN (C) = 21, SD = 5.54, MEAN (S) = 21, SD = 3.78, t = 0, p value = 1). Subjectively, 5 out of 7 drivers accounted for no intentional use of master controller in the purpose of rearming the vigilance function, one driver has accounted that “to be honest, I will not say that I never did it, but it was very rare” and one has reported to have solicited master controller more often in this intention, because he thought that was expected from Alstom ergonomist and was part of the test’s instruction. Conclusion on mechanical solicitation: Hypothesis 3 can be rejected as it has been shown both objectively and subjectively that LRV drivers do not traction or brake more frequently in SV modality with the sole intention of rearming deadman function. Conclusions of the User Test. Formal conclusions on efficiency (operators’ workload) and satisfaction cannot be drawn from statistical analysis of the results due to small population sample size and disparity in certain results. A better and more comprehensive approach is therefore to take note of drivers’ testimonies and subjective accounts of Smart Vigilance driving experience, including: • General sensation of increased comfort, • Cognitive workload decrease with more attentional resources available for the performance of safety-related tasks, • Stress decrease, • Very short adaptation time required, • General positive feeling, improved user experience. Additionally it has been noticed by ergonomic observations that Smart Vigilance solution goes in the sense of upper limb MSD prevention. Indeed it allows the driver’s hand to be positioned freely on the manipulator, this new freedom of movement going in the sense of MSD prevention and better physical ergonomics. The function does not induce an increase in rolling stock mechanical solicitation. The function could be enriched to be more transparent while driving in coasting mode where deadman still remains to be activated “classically”. Biometrics-based solution could be studied to further complement the solution. General conclusion from this ergonomic evaluation is therefore positive. It gives confidence that Smart Vigilance provides improvements in tramways’ driving in terms of physical and cognitive ergonomics, user experience, human factors and safety. Confidence can also be put on future acceptance of the smart vigilance technology by tramway drivers.

Decrease Driver’s Workload and Increase Vigilance

281

4 Future Technologies and Discussion One of the requests of LRV drivers is to replace the deadman system. New technologies are under development and some of them are already integrated in cars or in trucks. Some solutions are based on physiological measurement using camera that analyses the blink of the eye to calculate the rate of closure, from open to closed. Another method is based on an evaluation of the pupil dilatation. In order to integrate one of them in an LRV safety system, safety demonstration needs to establish that the system detects the incapacitation of the driver. A solution would be to integrate both systems (Smart Vigilance and physiological measures) to evaluate the validity of the indicator. Another requirement is to get the acceptance of drivers to integrate a camera inside the cab.

References 1. Miglianico, D., Mouchel, M., Moyart, L.: From Task Analysis to Innovation. Springer, Cham (2017) 2. Annett, J., Duncan, K.D., Stammers, R.B., Gray, M.: Task Analysis. Her Majesty’s Stationery Office, London (1971) 3. Mouchel, M., Naveteur, J., Miglianico, D., Anceaux, F.: Contrôle cognitif et prise de décision en environnement dynamique. Cas particulier de la traction et du freinage chez les conducteurs de tramway. In: Bonnardel, N., Pellegrin, L., Chaudet, H., (Eds.) Actes du Huitième colloque de psychologie ergonomique EPIQUE 2015, pp. 201–210, Arpege Science Publishing, Aix en Provence, Juillet(2015) 4. Brooke, J.: SUS: a “quick and dirty” usability scale. In: Jordan, P.W., Thomas, B., Weerdmeester, B.A., McClelland, A.L. (eds.) Usability Evaluation in Industry. Taylor and Francis, London (1996) 5. Foot, R., Garrigou, A.: Homme mort et conditions de travail des conducteurs de tramway. HAL open archives (2015). https://halshs.archives-ouvertes.fr/halshs-01105145 6. Electropedia: The World’s Online Electrotechnical Vocabulary. http://www.electropedia.org 7. Wickens, C.D., Hollands, J.C.: Engineering Psychology and Human Performance. Prentice Hall Inc, New Jersey (2000)

Analysis of Driving Performance Data to Evaluate Brake Manipulation by Railway Drivers Daisuke Suzuki1(&), Naoki Mizukami1, Yutaka Kakizaki2, and Nobuyuki Tsuyuki2 1

2

Ergonomics Laboratory, Railway Technical Research Institute, 2-8-38 Hikari-cho, Kokubunji-shi, Tokyo, Japan [email protected] Transportation and Marketing Department, Central Japan Railway Company, 1-3-4 Meieki, Nakamura-ku, Nagoya, Japan

Abstract. Here, we aim to investigate the relationship between the braking operations used to stop a train at a station and the errors in the train’s stopping position. Hence, using driving performance data, a logistic regression analysis was conducted. This analysis revealed that the train stopping-position errors at stations were associated with the standard deviation of the sum of brake notches, the mean of the additional brake notches, and the duration of driving experience. Drivers with a larger dispersion of brake notches in the individual were more prone to cause stopping-position errors at stations. Further, drivers who frequently used additional brake notches were more likely to cause stoppingposition errors at stations. Furthermore, operators with more driving experience were less likely to incur stopping-position errors. Keywords: Driving performance data
Brake manipulation
Railway driver
Human error
Driving skill

1 Introduction Driving performance data from railway vehicles are often saved in data-recording devices. These data can be used to analyze accidents, prevent human error, and improve crew skills. There have been several studies of data describing the driving performance of railway drivers. Sakashita et al. [1] analyzed more than 20 elements of driving performance data from approximately 100 drivers recorded over six months. The following evaluation indices related to braking operations for stopping a train at a station were used: (1) duration of strong braking (seven notches) used when stopping, (2) number of additional times brakes stronger than 3 notches used when stopping, (3) amount of brake notch changes used when stopping, (4) number of times coasting position used when stopping, and (5) train speed at a fixed point (five seconds before train stops). It was demonstrated that the use of strong braking (1) and many brake notch changes (3) strongly influenced whether the drivers experienced train stoppingposition error at stations. Moreover, it was found that the train speed at a fixed point © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 282–288, 2020. https://doi.org/10.1007/978-3-030-20503-4_26

Analysis of Driving Performance Data to Evaluate Brake Manipulation

283

five seconds before the train stopped varied widely before an error occurred. Marumo et al. [2] analyzed the braking behavior of train drivers in a driving simulator to estimate drivers’ mental conditions. The results showed that the relationship between the velocity deviation and the braking operation could be used as an indicator of abnormal driving behavior. Further, Marumo et al. [3] analyzed the braking operation when stopping at a station and found that simultaneously performing secondary tasks was associated with significantly greater variance in the brake handle operation. Although several examples of the factors influencing driver error have been studied, the relationship between brake manipulations and train stopping-position errors at stations has not been studied in depth. Therefore, we aim to investigate how the brake manipulation is related to the occurrence of stopping-position errors at a station by analyzing driving performance data.

2 Methods 2.1

Analysis Data

The data concerning driving operations (i.e., distance, velocity, and braking notches) when stopping at one train crew depot were extracted from the driving performance data recorded by railway vehicles in Japan. The sampling rate was 1 Hz. The data were collected over a period of two months beginning in August 2016. 2.2

Participant Group

118 drivers aged 25–61 years (mean age: 33 years, standard deviation: 9 years) participated in the study. The driving experience of the participants ranged from 0 to 29 years (mean: 7 years, standard deviation: 7 years). The “expert” group comprised 12 drivers, who were instructors of trainees, and selected by a manager of the train crew depot. These drivers were aged 30–57 years (mean age: 36 years, standard deviation: 7 years), and their driving experience ranged from 2 to 29 years (mean: 8 years, standard deviation: 7 years). The “error” group included 22 drivers who had made a train stopping-position error at a station in the previous nine months. These drivers were aged 25–61 years (mean age: 33 years, standard deviation: 11 years), and their driving experience ranged from 0 to 29 years (mean: 6 years, standard deviation: 9 years). Table 1 summarizes the age and driving experience of each group. Table 1. Age and driving experience of each group All drivers (118) Expert group (12) Error group (22) Age (years)

Range Mean SD Driving experience (years) Range Mean SD SD: Standard Deviation

25–61 33 9 0–29 7 7

30–57 36 7 2–29 8 7

25–61 33 11 0–29 6 9

284

2.3

D. Suzuki et al.

Evaluation Indices and Analysis Method

As described in the previous study [1] and summarized in Table 2, five evaluation indices regarding velocity and brake manipulation for stopping a train at a station were utilized. Figure 1 shows a schematic of the brake manipulations used when stopping. Table 2. Evaluation indices Evaluation indices (1) Train speed at a fixed point (2) Number of brake notch changes used when stopping (3) Additional brake notches when stopping (4) Sum of brake notches when stopping (5) Maximum brake notch when stopping “when stopping” implies “during the

Explanation Velocity five seconds before the train stops Number of times the brake notch was changed during the five second period before stopping Brake notches added during the five second period before stopping per 10 stops Sum of brake notches at each second during the five second period before stopping Maximum brake notch during the five second period before stopping five seconds immediately before the stop”

Fig. 1. Schematic of an example of brake manipulations when stopping

Each of the evaluation indices in Table 2 was evaluated for each stop, and the means and standard deviations of all the stops were evaluated for each driver. The means were used to compare the evaluation indices among the drivers, and the standard deviations were used to investigate the variability of each individual. A logistic regression analysis was conducted to identify the influence of the brake manipulations that are used to stop the train at the station on the occurrence of a stopping-position error at a station. The objective variable was the participant group, and the explanatory variables were the means and the standard deviations of the five

Analysis of Driving Performance Data to Evaluate Brake Manipulation

285

evaluation indices and the duration of driving experience. The step-wise method was used to select the variables. The statistical level of significance was 0.05. BellCurve for Excel, which is statistical software made by Social Survey Research Information Co., Ltd., was utilized to perform the analysis.

3 Result Table 3 shows a list of objective and explanatory variables used in logistic regression analysis. Table 3. List of objective and explanatory variables Objective variable

Explanatory variables

Mean of the train speed at a fixed point (km/h)

Standard deviation of the train speed at a fixed point (km/h)

Mean of the number of brake notch changes (times)

Standard Mean deviation of the of the additional number of brake brake notch notches changes (notches) (times)

Standard deviation of the additional brake notches (notches)

Mean of the sum of brake notches (notches)

Standard deviation of the sum of brake notches (notches)

Mean of the maximum brake notch (notches)

Standard deviation of the maximum brake notch (notches)

Driving experience (years)

No.

Participant group

1

Expert

8

2

1

1

5

7

8

3

2

1

29

2

Error

9

2

2

1

5

7

10

4

3

1

29

3

Error

11

3

2

1

8

16

14

5

4

2

7

4

Expert

7

2

1

1

3

6

7

3

2

1

3

5

Error

8

2

1

1

3

5

7

3

2

1

10

6

Expert

7

2

1

1

3

6

7

3

2

1

10

7

Expert

7

2

1

1

3

5

7

3

2

1

9

8

Expert

9

2

1

1

2

4

9

3

3

1

6 20

9

Error

8

3

1

1

6

8

9

4

2

1

10

Expert

8

2

1

1

3

5

8

3

2

1

2

11

Error

8

2

1

1

4

7

8

4

2

1

6

12

Expert

8

1

1

1

2

5

8

2

2

1

9

13

Expert

9

2

1

1

4

7

9

3

3

1

6

14

Expert

8

2

1

1

3

5

7

3

2

1

7

15

Error

8

2

1

1

3

6

7

3

2

1

6

16

Expert

9

2

1

1

1

4

10

4

3

1

6

17

Expert

9

2

1

1

1

4

9

3

3

1

8

18

Error

7

2

1

1

5

7

7

3

2

1

5

19

Expert

7

2

1

1

4

6

7

3

2

1

5

20

Error

8

2

1

1

3

5

8

3

2

1

3

21

Error

7

2

1

1

3

5

7

3

2

1

0

22

Error

8

2

1

1

2

5

9

4

2

1

1

23

Error

8

2

1

1

4

6

8

4

2

1

3

24

Error

8

2

1

1

5

6

8

3

2

1

3

25

Error

7

2

1

1

3

5

6

2

2

1

1

26

Error

8

2

1

1

4

5

9

4

2

1

2

27

Error

7

2

1

1

7

7

7

4

2

1

0

28

Error

8

2

1

1

3

5

8

4

2

1

0

29

Error

8

2

1

1

6

9

8

4

2

1

1 0

30

Error

7

2

1

1

4

6

6

3

2

1

31

Error

8

2

1

1

6

8

9

4

2

1

0

32

Error

10

3

2

1

11

11

11

5

3

1

29

33

Error

8

2

1

1

3

5

8

3

2

1

0

34

Error

7

2

1

1

6

9

7

4

2

1

0

286

D. Suzuki et al.

The variance-information factors (VIFs) were calculated to evaluate the multicollinearity. Generally, it is said that there is a possibility of multicollinearity if the maximum VIF is above 10 or the average VIF is significantly more than 1 [4]. In this case, the maximum VIF among the explanatory variables used in this study was 1.78; thus, it was concluded that multicollinearity was not an issue for this analysis. Table 4 shows the results of the logistic regression analysis. In case that the p value is lower than 0.05, the occurrence of a train stopping-position error is significantly associated with the explanatory variable. The p values revealed that the occurrence of a train stopping-position error was significantly associated with the mean of additional brake notches used, the standard deviation of the sum of brake notches used, and the driving experience. Table 4. Results of logistic regression analysis Partial regression Coefficient

Standardized partial regression coefficient

Odds ratio

Mean of additional brake notches (notches)

2.58

5.04

Standard deviation of sum of brake notches (notches)

2.43

Driving experience (years)

-0.21

Constant

-7.43

Explanatory variables

95% confidence interval of the odds ratio

p value

*: p 10 and ps < .001) as well as an interaction between participant type and speed limit, F(2, 256) = 4.954, p = 0.008. The data show that cyclists reported higher “reasonable” speeds than motorists and the difference between the groups increased with higher speed limits (see Fig. 1).

70 60 50 40 30 20 10 0

Cyclist Motorist

25

45 Speed Limit

60

Fig. 1. Cyclist and motorist estimation of “reasonable” speed to pass a cyclist in the context of various posted speed limits. Error bars represent standard deviation.

3.2

Perceived Hazards

In order to determine if motorists and cyclists are sensitive to the same perceived hazards, both groups rated perceived hazardousness of specific situations. Cyclists rated how hazardous the situation would be for themselves; motorists rated how hazardous the situation would be for cyclists. In general, cyclists rated all situations as less hazardous (M = 3.82, SE = 0.06) than motorists (M = 4.16, SE = 0.09) who were perspective-taking rated the same situations, t(129) = –3.368, p < .001, Cohen’s d = –0.635. There was a notable departure from this trend that is clear when the data are split into subcategories (road and weather conditions, motorist interactions, non-motorist interactions). When considering hazardous road and weather conditions, cyclists rate them as significantly less hazardous (M = 3.51, SE = 0.07) than motorists (M = 3.89, SE = 0.11), t(129) = –3.024, p = .003, Cohen’s d = –0.570. Similarly, cyclists rate interactions with non-motorists (e.g., cyclists, pedestrians, animals) as being less hazardous (M = 3.54, SE = 0.11) than motorists rate those interactions (M = 4.09, SE = 0.13), t(129) = –3.120, p = .002, Cohen’s d = –0.588. In contrast, no significant difference in ratings was found for interactions with motorists: Cyclists rated them similarly (M = 4.58, SE = 0.05) to motorists (M = 4.693, SE = 0.09), t(129) = -1.167, p = .246, Cohen’s d = –0.220. These results suggest that while motorists have a bias to

Sharing the Road: Experienced Cyclist and Motorist Knowledge and Perceptions

297

ProporƟon of Respondents

see cyclists as vulnerable road users, cyclists are comparatively more sensitive to the threat of motorists than the other hazards. Another approach to understanding perceived hazards is to examine how behavior might change in the context of a hazard. For instance, if a pothole is seen as extremely hazardous, a cyclist or motorist might move toward the center of the lane to avoid it. As is seen in Fig. 2, cyclists indicate that they actively avoid some hazards that motorist do not. Cyclists report moving toward the center of the lane to avoid significantly more (M = 8.044, SE = 0.24) of the 12 hazards than motorists (M = 4.927, SE = 0.43), t(129) = 6.823, p < .001, Cohen’s d = 1.286. This difference is notable because although motorists may know that cyclists are vulnerable road users, motorists may not actively monitor for the same hazards as cyclists. This lack of experience for motorists could decrease their ability to detect cycling-specific hazards and, therefore, impede their ability to anticipate cyclist responses to those hazards.

1 0.8 0.6 0.4 0.2 0

Cyclist

Motorist

Fig. 2. Proportion of cyclists and motorists reporting active avoidance of specific hazards

4 Conclusion Many factors contribute to the challenge of safely sharing the road. This study focused on two factors that could impact cyclist and motorist expectations for road behavior. First, we examined cycling-related law in Virginia and the corresponding safety recommendations. Second, we asked cyclists and motorists to consider specific road hazards and how those hazards may or may not impact their riding or driving behavior. The results indicate that cyclists had more accurate knowledge of bicycle-related laws and recommendations. Motorists in this sample had an acceptable understanding of the laws governing motorists’ responsibilities but were much less aware of laws and recommendations for cyclists. With that in mind, the results of a few test items were

298

M. L. Still and J. D. Still

particularly concerning for safe roadway interactions. Most motorists reported that cyclists were required to use the bike lane or multi-use path if they were available; they indicated that vehicles have the absolute right-of-way if no bike lane is present or no shared lane markings appear on the road; and that cyclists were required to ride within three feet of the right side of the road. These results demonstrate a clear need for more proactive approaches to expose both motorists and cyclists to the law and recommended best practices. Beyond knowledge of the law and the associated recommendations, there does seem to be common ground between cyclists and motorists. For instance, cyclists have reasonable expectations for how fast they should be passed by vehicles (i.e., they allow for faster passing speeds than the motorists). Further, motorists seem to perceive cyclists as vulnerable road users. When asked to imagine how hazardous specific situations would be to a cyclist, motorists consistently rate those hazards as being more hazardous than actual cyclists rate the same hazards. One point of concern with the designation of vulnerable road user, is that the term could also imply that cyclists are irresponsible road users because they have chosen to take part in a risky activity (c.f., [14]). In a similar vein, Haworth et al. (2018) suggest that the increased awareness of cyclists on the roads (and their vulnerability) could be driven by frustration, not necessarily by concern for safety or a desire to share the road [4]. Finally, even though motorists may be able to identify situations that are hazardous to a cyclist, that does not mean they have experience monitoring the roads for the same hazards as cyclists. In these cases, motorists would not have the same quality of hazard perception as a cyclist (e.g., [8]). This could mean that motorists may still be surprised if a cyclist moves to avoid a hazard, like a gutter on the side of the road, because motorists do not have experience looking for and anticipating cyclist responses to those hazards. This missing link could be contributing to the regular “close calls” cyclists report having with motorists. These findings demonstrate some clear differences in cyclist and motorist expectations for sharing the road. Critically, it appears that motorists lack knowledge about bicycle-related laws and recommendations. A lack of understanding is likely contributing to motorists’ frustration with cyclists on the roadway. Motorists do view cyclists as vulnerable and are aware of their own limited knowledge of bicycle-related laws. The next step in protecting cyclist and decreasing motorist frustration is to provide basic education and assessment. Some of this educational information could be tested during motorist licensing procedures. If the law and associated recommendations are followed, it should create a safer and more comfortable road experience for everyone. Acknowledgments. We thank Hampton Roads Cyclists, Swamp Stomp, Rustbucket Races, East Coast Bicycles, Rogue Velo Racing and Tradewinds Racing – Mermaid Winery for their assistance collecting data.

Sharing the Road: Experienced Cyclist and Motorist Knowledge and Perceptions

Appendix: Bicycle-Related Law and Recommendation Items Cyclist-specific questions (proportion correctly answered) Cyclists Motorists T/F When a cyclist is riding on a road with parked cars on 0.57 0.36 the street, it is safest to move out of the main lane of traffic when the parking lane is clear. This provides more space for passing motorists. T/F When riding on an ordinary width lane, it is safest for a 0.76 0.60 bicyclist to ride near the center of the lane when approaching intersections. T/F It is legal to travel between two lanes of traffic moving in 0.88 0.69 the same direction. T/F A shared lane marking indicates to cyclists they should 0.47 0.33 be taking the lane. T/F It is safest to pass parked cars within five feet to allow 0.61 0.26 motorist more space to pass. T/F Bicyclists should ride on the sidewalk when a sidewalk 1.00 0.57 is available. T/F Bicyclists should avoid alternating between the sidewalk 0.79 0.76 and road in order to more predictable to motorists. T/F Bicyclists are required to ride within 3 feet of the curb. 0.86 0.24 T/F Cyclists need to hug the curb to stay as far right as is 0.81 0.50 safely practicable. T/F It is okay for a bicyclist to ride in the turn lane even if 0.91 0.79 they do not plan to turn. T/F Riding towards the flow of motorists can be safer than 0.97 0.69 riding with the flow of traffic. T/F Before turning bicyclists should always use hand signals. 0.98 0.91 If a bicyclist is waiting at a traffic signal and it does not 0.63 0.48 respond to their presence, they can proceed through the signal, with caution, after waiting _2__ minutes or cycles. T/F It is against the law for bicyclists to wear earphones in 0.88 0.81 both ears while riding on the road. T/F Bicyclists should not pass on the right of motorists at 0.76 0.74 intersections. T/F The law requires cyclists riding on the road to always use 0.11 0.12 a front light and rear light between sunset and sunrise. T/F Bicyclists should always carry their identification and 0.96 0.88 medical insurance information. T/F Bicyclists should cross railroad tracks at a perpendicular 0.99 0.83 angle to avoid getting their wheel caught in the tracks. T/F Bicyclists are required to use the bike lane or separate 0.66 0.07 multi-use path when they are available. Motorist-specific questions (proportion correctly answered) Cyclists Motorists T/F Motorists should not drive in a bike lane expect when 0.81 0.64 turning. When turning into a bike lane, motorists need to check: 0.93 0.76 The rear The left and rear The right and rear Not necessary bikes must yield T/F If no bike lanes or shared lane markings are present the 0.99 0.55 motorist has complete right-of-way. T/F Motorists must look for bicycles and pedestrians when 0.97 0.98 turning across sidewalks into driveways. T/F The law requires that motorists pass bicyclists at a 0.99 0.93 reasonable speed and allow at least three feet of space. T/F When turning across a bike lane you should signal and 0.90 0.93 scan the lane before turning. T/F When entering a bike lane, you should attempt to overtake 0.94 0.88 an oncoming bicycle to create a safe position within the intersection. T/F Bike lanes can be used as loading zones where motor 0.81 0.76 vehicles can temporarily stop as long as the lane is clear.

299

300

M. L. Still and J. D. Still

References 1. National Highway Traffic and Safety Administration’s National Center for Statistics and Analysis. Traffic Safety Facts 2016 Data: Bicyclists and Other Cyclists. (DOT Publication No. HS 812507). Government Printing Office, Washington, DC (2018) 2. Amoros, E., Martin, J.L., Laumon, B.: Under-reporting of road crash casualties in France. Accid. Anal. Prev. 38, 627–635 (2006) 3. Winters, M., Branion-Calles, M.: Cycling safety: quantifying the under reporting of cycling incidents in Vancouver, British Columbia. J. Transp. Health Part A 7, 48–53 (2017) 4. Haworth, N., Heesch, K.C., Schramm, A.: Drivers who don’t comply with a minimum passing distance rule when passing bicycle riders. J. Saf. Res. 67, 183–188 (2018) 5. Aldred, R., Crosweller, S.: Investigating the rates and impacts of near misses and related incidents among UK cyclists. J. Transp. Health 2, 379–393 (2015) 6. Winters, M., Weddell, A., Teschke, K.: Is evidence in practice? Review of driver and cyclist education materials with respect to cycling safety evidence. Transp. Res. Rec. 2387, 34–45 (2013) 7. Chaurand, N., Delhomme, P.: Cyclists and drivers in road interactions: a comparison of perceived crash risk. Accid. Anal. Prev. 50, 1176–1184 (2013) 8. Ventsislavova, P., Gugliotta, A., Peña-Suarez, E., Garcia-Fernandez, P., Eisman, E., Crundall, D., Castro, C.: What happens when drivers face hazards on the road? Accid. Anal. Prev. 91, 43–54 (2016) 9. Crundall, D.: Hazard prediction discriminates between novice and experienced drivers. Accid. Anal. Prev. 86, 47–58 (2016) 10. Lehtonen, E., Havia, V., Kovanen, A., Leminen, M., Saure, E.: Evaluating bicyclists’ risk perception using video clips: comparison of frequent and infrequent city cyclists. Transp. Res. Part F 41, 195–203 (2016) 11. Horswill, M., Kemala, C.N., Wetton, M., Scialfa, C., Pachana, N.A.: Improving older drivers’ hazard perception ability. Psychol. Aging 25, 464–469 (2010) 12. McKenna, F.P., Horswill, M.S., Alexander, J.L.: Does anticipation training affect drivers’ risk taking? J. Exp. Psychol. Appl. 12, 1–10 (2006) 13. Zeuwts, L., Cardon, G., Deconinck, F.J.A., Lenoir, M.: The efficacy of a brief hazard perception interventional program for child bicyclists to improve perceptive standards. Accid. Anal. Prev. 117, 449–456 (2018) 14. Bonham, J., Johnson, M., Haworth, N.: Cycling related content in the driver licensing process. Transp. Res. Part A 117, 117–126 (2018)

Examination on Corner Shape for Reducing Mental Stress by Pedestrian Appearing from Blind Spot of Intersection Wataru Kobayashi1(&) and Yohsuke Yoshioka2 1 Department of Architecture, Division of Creative Engineering, Graduate School of Science and Engineering, Chiba University, 1-33, Yayoi-cho, Inage, Chiba, Japan [email protected] 2 Graduate School of Engineering, Chiba University, 1-33, Yayoi-cho, Inage, Chiba, Japan [email protected]

Abstract. This paper investigates the effective minimum size for cutting out the edge of the intersection for reducing the mental stress of the pedestrian appearing from an intersection, by the experiment using the electrodermal activity measuring method and the virtual reality technology. 30 college students as the participants were experienced the unlimited long virtual passage where the intersections would be appearing regularly and continuously, through a head mounted display. The width of the passage was set at two sizes, 1,600 mm and 2,000 mm. The result is that the mental stress for the crossing pedestrian could be reduce by applying a cutting out of more than 1,000 mm to the corner of the intersection. According to these results, it was suggested that cutting out the corner of the intersection of about 1,000 mm would give a certain effect on the worry-free passage design. Keywords: Virtual reality system
Intersection
Psychological and physiological stress
Electrodermal activity

1 Introduction Corners and intersections are dangerous places for pedestrians who are likely to collide with each other because of blind spots. Everyone has experienced a situation where they realized the danger just before colliding with a pedestrian coming from the blind spot at an intersection. One only needs to be able to notice the pedestrian before a collision, but if one crashes at a dangerous speed without noticing the other pedestrian, it will lead to a serious accident. Regarding corner lots, each municipality establishes “corner cutting” for the purpose of securing the view of the intersection1. This aids in reducing the risk of collision of automobiles, bicycles, and pedestrians, and allows vehicles and pedestrians to turn easily.

1

In New Jersey It is stipulated that the corner should be cut into a triangle with 25 feet on each side.

© Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 301–306, 2020. https://doi.org/10.1007/978-3-030-20503-4_28

302

W. Kobayashi and Y. Yoshioka

If “corner cutting” would also be applied to indoor passages, sufficient view of the area could be secured and the danger of collision would be reduced. However, since structure columns are often hidden behind the edge of the corner, it is impossible to increase cutting out indiscriminately. Therefore, this study aimed to investigate the effective minimum size for cutting out the edge of the intersection for the early detection of a pedestrian appearing from an intersection. Various studies have been conducted on the relationship between the shape and visibility of intersections. In the study by Senda, M., the corner shape was elucidated by investigating the pedestrian’s flow lines at the corner of the school corridor [1]. In the study by Ono, H., the corners in a building were analyzed from the viewpoint of easiness in walking, visibility, and wheelchair mobility [2]. The study by Chibana, K., investigated the variation in visibility due to the presence or absence of corner cutting at the intersection from the range of visibility between pedestrians and bicycles [3]. However, no experiments have been conducted using a physiological reaction such as “surprise”. This investigation aimed to examine the effective minimum size of cutting out an edge of a corner at an intersection for reducing the psychological stress caused by the pedestrian appearing from behind the corner. An experiment using the electrodermal activity measuring method and virtual reality technology was conducted [4].

2 Methods We designed and modelled an unlimited number of long virtual passages with intersections appearing regularly to be displayed via a VR software (Vizard 5.0). At an intersection of the passage, avatars were programed to cross in front of the subject randomly while the subject experienced the virtual environment through a head mounted display (Oculus Rift). Subjects could look around the virtual environment by detecting the inclination and rotation of the head with Sensors (Oculus Sensors). To measure the mental stress level during crossing intersections, the subjects were made to wear an electrodermal activity meter (TS02 SPL/R-AD) attached to three skin patches on their left hand [5–7] (Fig. 1).

Fig. 1. The condition of the experiment showing the unlimited virtual passages to a subject

Examination on Corner Shape for Reducing Mental Stress

2.1

303

Electrodermal Activity

When humans experience emotional stress, like fear, excitement, surprise, and other such emotions, physiological responses, such as an increase in the heart rate, perspiration, and deterioration of the immune system response, are noted. The physiological response measured in the study is mental perspiration. Mental perspiration is said to be closely related to the fight or flight response that humans use to deal with emergencies. Palm sweating at the time of fighting aids in assuring a tight grasp, while sole sweating at the time of escaping increases the friction when kicking the ground. Mental perspiration is a function performed by humans since ancient times. It is not effective when assessing whether or not the function is effective; however, mental perspiration is performed effectively if the simulated condition closely matches the natural conditions. Electrodermal activity (EDA) is the sum of electrical characteristics detected on human skin. While other physiological indices have a time lag between receiving a stimulus and the appearance of a reaction, electrodermal activity has only a time lag of about 1 to 2 s for a reaction to appear after stimulation. Therefore, the measurement of electrodermal activity is considered appropriate. When EDA is measured, a transient reaction (skin potential response) and a slow reaction (skin potential level) are observed. In this experiment, since we wanted to measure the instantaneous mental burden on pedestrians appearing from the blind spot at the intersection, we only analyzed the skin potential response. The magnitude, which is the sum of the amplitudes of positive and negative waves, was measured (Fig. 2). There are individual differences in the reactions of the electrodermal activity. Therefore, if the magnitude of all the subjects is analyzed on the same basis, the effect of the subjects who easily respond to electrodermal activity on the analysis result will be large, and the analysis will be incorrect. Therefore, we analyzed the data using a “reaction ratio”. First, the average value of the reaction ratio when passing through the intersection of the “reference condition”, which is the condition when the corner cutting depth is 0 mm and the avatar is not present, is calculated for each subject and set as R0. Subsequently, in the condition group where the avatar appears, the average value of the reaction amounts of “conditions appearing from the left” and “conditions appearing from the right”. These are calculated for each depth of corner cutting for each subject and are taken as RX. In the group of conditions where no avatar appears, the magnitude was RX. RX/R0 is defined as “the reaction ratio” under each experimental condition [6].

Fig. 2. The schematic graph of skin potential response

304

2.2

W. Kobayashi and Y. Yoshioka

Experimental Space

The height of the passage was set at 2,700 mm, and the width was set at 1,600 mm and 2,000 mm. A stucco was used for the wall material and tiles were used as flooring materials. A virtual walking avatar appeared randomly from the blind spot of the intersections set at 35,000 mm intervals and crossed in front of the subjects. Figure 3(A) shows the cross section of the experimental space and Fig. 3(B) shows a plan of the experimental space. The avatar walking at 4 km/h emerged from the blind spot of the intersection and crossed in front of the subject. The avatar was a general adult male body with a height of 1750 mm [8].

Fig. 3. (A) A cross section. (B) A plan.

2.3

Experimental Conditions

There were 5 levels of depths of cutting out the edge of the corner from 0 m to 2 m. Figure 4 shows the image of each corner cutting depth viewed from the position of 2,000 mm from the corner. There were 3 conditions for the appearance of the avatar, as follows: first, no avatar appeared; second, avatar appeared from the left side of the intersection; third, the avatar appeared from the right side of the intersection. Therefore, there were 15 types of experimental conditions obtained by multiplying the 5 types of cutting edge depths and the 3 types of avatar appearance conditions. The condition where the cutting depth was 0 mm and the avatar did not appear was referred to as the “reference condition” and the subjects experienced it 7 times [6]. The other 14 conditions where experienced by the subjects only once. The subjects randomly experienced these 21 intersections continuously with two breaks in the middle.

Fig. 4. Image of each corner cutting depth viewed from the position of 2,000 mm from the corner

Examination on Corner Shape for Reducing Mental Stress

305

3 Results As a result of factorial ANOVA performed on the profiles of the maximum amplitude of skin potential level, an interaction effect was detected between the depths of cutting out the corner and the avatar’s appearance. By a multiple comparison performed on the profile of the maximum amplitude of skin potential responses on each experimental condition when the avatar appeared, a significant difference was not found at the conditions of 0 mm depth and 500 mm depth. By a multiple comparison performed on the movement of the maximum amplitude of skin potential levels on each experimental condition when the avatar appeared, a significant difference was detected between the data with less than 500 m and the data with more than 1000 m (Fig. 5). Multiple comparison performed on the movement of the maximum amplitude of skin potential levels, for each experimental condition when the avatar does not appear, revealed no significant difference.

Fig. 5. (A) The averages of the reaction ratio of the electrodermal activity when the avatar appear. (B) The averages of the reaction ratio of the electrodermal activity when the avatar does not appear.

4 Discussion The depth of the edge of the corner required to reduce the mental burden of a person crossing associated with a blind spot is clearly of 1,000 mm or more. A significant difference was not found in conditions of 0 mm depth and 500 mm depth. This result suggests this depth made the subjects feel a potential fear by the chance of encountering other pedestrians from the blind spot. On the other hand, it was discovered that the reaction ratio under the condition that the avatar does not appear is not changed depending on the depth of the corner cutting. In other words, it was shown that it was impossible to dispel the feeling of anxiety that “pedestrians may come out from the blind spot” when passing through the intersection only by operating the depth of corner cutting.

306

W. Kobayashi and Y. Yoshioka

5 Conclusions In this study, we clarified the effect of cutting out the edge of the corner of the intersection on human mental activity. As a result, we found that the skin potential level caused by a crossing pedestrian was significantly reduced by applying a cutting out of more than 1,000 mm. Although the analysis aimed at applying the results to general dimensions of corner cutting, the result of this research is limited. For example, various information such as footsteps and shadows of the pedestrian are present in actual intersections. In order to make the knowledge of this research even more reliable, it is necessary to verify such supplementary information. Also, all subjects in this experiment are college students in their twenties. Therefore, additional experiments for children and elderly people will be necessary to generalize the results. Acknowledgments. This work was supported by JSPS KAKENHI Grant Number JP17H03359.

References 1. Senda, M., Yata, T., Ohkoshi, H.: Children’s flow lines in the corridor and design guidelines for rounding the corner: video research at a primary school and flow experiment in full-sized models. J. Archit. Plann. Environ. Eng. (Trans. AIJ) 59(455), 109–118 (1994) 2. Ono, H., Kudo, R.: Study on corners in building from a viewpoint of easiness for walking, visibility and wheelchair mobility. J. Archit. Plann. Environ. Eng. (Trans. AIJ) 70(597), 25–31 (2005) 3. Chibana, K., Kajimot, H., Kubota, M.: The recognizable distance of the bicycle by the pedestrian at intersection. J. Archit. Plann. Environ. Eng. (Trans. AIJ) 67(558), 145–150 (2002) 4. Seinfeld, S., Bergstrom, I., Pomes, A., Arroyo-Palacios, J., Vico, F., Slater, M., Sanchez-Vives, M.V.: Influence of music on anxiety induced by fear of heights in virtual reality. Front. Psychol. 6, 1–12 (2016) 5. Natsuko, N., Daiu, M., Hitoshi, W.: Relation between psychological stress and window side design in height: research of architectural plan of psychological stress by height: Part 1. J. Archit. Plann. Environ. Eng. (Trans. AIJ) 76(662), 741–746 (2011) 6. Daiu, M., Natsuko, N., Jumpei, S., Hitoshi, W.: Relationship between psychological stress and postural stability of sense as design capacity in altitude: research of architectural plan of psychological stress by height: Part 2. J. Archit. Plann. Environ. Eng. (Trans. AIJ) 77(676), 1319–1324 (2012) 7. Friedman, D., Suji, K., Slater, M.: SuperDreamCity: an immersive virtual reality experience that responds to electrodermal activity. In: International Conference on Affective Computing and Intelligent Interaction, vol. 4738, pp. 570–581 (2007) 8. Garau, M., Slater, M., Pertaub, D.-P., Razzaque, S.: The responses of people to virtual humans in an immersive virtual environment. Teleoperators Virtual Environ. 14(1), 104–116 (2005)

Pedestrian Attitudes to Shared-Space Interactions with Autonomous Vehicles – A Virtual Reality Study Christopher G. Burns1(&), Luis Oliveira1, Vivien Hung1, Peter Thomas2, and Stewart Birrell1 1 WMG, University of Warwick, Coventry CV4 7AL, UK {c.burns.2,l.oliveira,w.hung,s.birrell}@warwick.ac.uk 2 Jaguar Land Rover, Coventry CV3 4LF, UK [email protected]

Abstract. The automotive industry is steadily moving towards fully autonomous vehicles, and it is becoming important to understand attitudes towards them. This study is an aspect of the www.ukautodrive.com project with JaguarLand Rover, RDM Automotive, and The University of Warwick’s Warwick Manufacturing Group (WMG). Uniquely, we used a prototype fully autonomous vehicle, and were interested in pedestrian attitudes towards this vehicle manoeuvring in close proximity. Using virtual reality (VR) cameras, we filmed 18 manoeuvring scenarios and presented them using VR equipment. Participants answered four short rating-scale questions after each exposure, and self-reported less trust and safety when the vehicle was faster and closer. This work has implications both for real-world autonomous vehicles, and for further use of VR technology. That the VR environments seemed sufficiently convincing to evoke consistent responses from volunteers represents a considerable opportunity across a variety of experimental domains, and can improve further with advances in this technology. Keywords: Trust


Safety
Autonomous vehicles
Human factors

1 Introduction There is a growing body of research among the Automotive User Interface community to understand aspects of user interaction with autonomous vehicles [1]. A number of challenges and questions are frequently discussed but still need to be addressed, including those surrounding the ergonomics of users’ interaction with a vehicle, situational awareness, acceptance, trust and ethical issues [2]. Early tests have been performed with autonomous vehicles (AVs) to transport passengers at low speed and short distances [3, 4], and on how these vehicle should communicate intention to pedestrians and cyclists via external human-machine interaction [5]. Given the potential physical risks surrounding automotive research with human participants (i.e. experimental vehicles can suffer various malfunctions which could be physically unsafe), safer methods to study the interaction between people and vehicles are of interest [6, 7]. Virtual Reality (VR) is an immersive technology where users experience simulated © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 307–316, 2020. https://doi.org/10.1007/978-3-030-20503-4_29

308

C. G. Burns et al.

environments through visual and auditory inputs. The immersion is created via a headmounted display which tracks the user’s head orientation to update a rendered viewpoint within a video-recorded or simulated environment with stereo sound to improve sensory immersion. VR environments are often used to test aspects of the user interaction with technology in a variety of scenarios, especially in situations which could be physically dangerous [8]. VR can induce a stronger sense of “presence” compared to traditional nonimmersive virtual environments (VEs) such as a standard computer with a monitor screen. Presence is defined as the subjective experience of being immersed in a virtual space and environs [9, 10]. VR simulations can better control most variables and ensure that participants have an objectively similar experience, which is typically harder to do in real-world conditions. VR can also simulate situations which could be too hazardous or infeasible to implement in real life. Software-rendered 3D VEs have also been used to understand pedestrian behaviours with autonomous vehicles; e.g. investigating how people negotiate the space with autonomous cars during road crossings and the feedback given by an autonomous vehicle when it indicated that it was about to stop or to move off [11]. Another study using a VR environment had vehicles driving past a pedestrian crossing and evaluated the impact of vehicle’s external lights on the user experience [12], while VR has also been used to simulate AVs with ‘eyes’ on the headlights that ‘see’ pedestrians and indicate intention to stop [13]. Very few studies, however, investigate the perceptions of acceptance or trust when people have to share the same areas as vehicles [5], and it is unclear the distances and speeds at which autonomous vehicles should drive in these scenarios in order to improve perceptions of safety. The present study examined if the speed of an autonomous vehicle and its lateral distance from road users affected acceptance and trust, especially in a semipedestrianized area. Autonomous vehicles are characterized according to six increasing levels of automation from level 0 to 5 [14], with level 5 vehicles being fully self-driving and self-navigating across nearly all potential driving situations. However, increases in public acceptance of AVs is hindered by media coverage of incidents or even fatalities which involve AVs, which are often caused by a mixture of misuse, human error, and imperfect systems [15, 16]. Over-trust in AVs could result in higher chances of accidents as any AVs with automation up to level 3 require the driver to intervene and regain control of the vehicle when necessary [17, 18]. Some proportion of accidents involving AVs are inevitable, caused by imperfect automated systems which cannot infallibly react appropriately to e.g. poor weather or visibility conditions, as well as in scenarios where the interaction between the AV and other road users make predictive calculations prone to inaccuracy [19]. There is also evidence [20] that peoples’ acceptance and experience of AVs decreases as the level of automation increases. The reluctance in acceptance and trust in AVs is perhaps mediated by the public knowledge that AVs are still in development and there is still a long way to go until we reach a widespread adoption of level 4 or 5 AVs.

Pedestrian Attitudes to Shared-Space Interactions

309

2 Methods 2.1

Participants

13 participants (4 females) were recruited for this study, aged between 21 to 58 (M = 36.69, SD = 12.3). Participants were recruited through posters placed on common areas within the International Digital Lab building, University of Warwick. Participation was entirely voluntary, and no financial incentives were given. Participants required only normal or corrected-to-normal vision to participate. 2.2

Design

This study was a within-subjects multifactorial design. Three independent variables included camera positions (C1, C2), AV distance from camera positions (1 m, 2 m, 3 m), and AV speeds (1.75 m/s, 2.25 m/s, 3 m/s). All 18 AV passing scenarios were recorded using a Kodak PixPro 360 camera positioned 1.48 m off the ground using a tripod. We used an electric LSATS (Low-speed autonomous transport system) PodZero AV with dimensions of 1.4 m(w)  2.5 m(l)  2 m(h); images of the pod can be found at . Each scenario featured the AV travelling around a corner, driving past the camera then turning a second corner, out of the camera’s field of vision. The 18 VR videos were 13–20 s in length and presented in a randomised order, and edited such that the starting viewpoint was aimed towards the corner where the pod would emerge. Participants watched the videos using an Oculus Rift VR system connected to an Alienware 17 R5 laptop (Fig. 1).

Fig. 1. Layout of the indoor testing arena.

310

C. G. Burns et al.

Questionnaires were administered using Google Forms. After each interaction, participants were asked to rank their experience using four questions on a 7-point Likert-type scale: 1. What do you think about the speed of the vehicle? (1 Unsafe speed > Safe speed 7) 2. What do you think about the distance between you and the vehicle? (Unsafe distance > Safe distance) 3. What is your general feeling of safety at this speed and distance? (Unsafe > Safe) 4. What is your general feeling of trust in the vehicle at this speed and distance? (Distrust > Trust) The Simulator Sickness Questionnaire (SSQ) [21] was used to monitor if participants experienced any cybersickness effects during the VR simulations. Cybersickness refers to a condition which is specifically induced by exposure to VR environments, whereas simulator sickness is used in the context of flight, driving, or other training simulators. The SSQ contains 27 symptoms on the axes of nausea, oculomotor discomfort and disorientation, each rated on a nominal scale of ‘None’, ‘Slight’, ‘Moderate’, and ‘Severe’ to denote the severity of the symptom experienced. 2.3

Procedure

The scenarios were briefly explained to participants, and informed consent was obtained after questions were addressed. Participants were asked to complete the presimulation SSQ. A still image of the simulation was presented from the perspective of the specific camera position of their first allocated condition and participants were given a few minutes to familiarize themselves with the headset. They were reminded that they could adjust the headset for comfort using straps on the sides of the headset, and to check that they could see objects in the simulation reasonably clearly. Then, each of the 18 simulations were presented to the participant, in a randomized order and participants responded to the 4-item questionnaire. When the participants completed all 18 simulations and subsequent questionnaires, they completed a post-simulation SSQ. Afterwards, a semi-structured interview was conducted, but this data will be presented elsewhere and is not analysed as part of this study. Finally, the participants were debriefed and any questions addressed. The duration of the experiment was around 30–35 min per participant.

3 Results 3.1

Cybersickness (SSQ)

No participants reported any noteworthy discomfort or sensations associated with simulator sickness at any point in the study. Individual total scores ranged from zero to 76.32 points on the scale, where the maximum achievable score is at 2437.9.

Pedestrian Attitudes to Shared-Space Interactions

311

Although SSQ mean scores rose across all three intervals where it was measured (pretest; = 7.48, mid-test = 7.192, and post-test = 9.206), there were no significant differences in SSQ Total score (F(2,24) = 0.329, p = N.S.), nor any of the sub-scales for Nausea (F(2,24) = 0.618, p = N.S.); Oculomotor disturbances (F(2,24) = 0.625, p = N.S.), or Disorientation (F(2,24) = 0.178, p = N.S.). No participant reported any subjective or anecdotal discomfort. 3.2

Safety Questionnaire

Rating responses to each question in turn after each passing scenario were analysed via repeated-measures 4-way ANOVAs using IBM SPSS Statistics 24. The ANOVA comprised Camera position (2 positions)  Distance (1 m, 2 m, 3 m)  Speed (1.75 m/sec, 2.25 m/sec and 3 m/sec)  3 groups – “Prior experience with VR” (4 individuals), “Prior experience with LSATS” (4 individuals) and “Have owned a vehicle with autonomous features” (3 individuals) for a 2  3  3  3 design. All pairwise comparisons were conducted using SPSS’s Least Significant Difference (LSD) method. Only two tests of a single between-subjects factor (“Have owned a vehicle with autonomous features”) produced significant results, (for Q1, F(1,9) = 5.493, p < 0.047, and for Q4, F(1,8) = 5.603, p < 0.045), but subsequent post-hoc tests produced no further discrimination in scores. The pod’s passing distance exerted a main effect across all 4 questions - on pod speed and safety (F(2,16) = 10.175, p < 0.001), on pod passing distance (F(2,16) = 25.956, p < 0.0001), on general feelings of safety in the scenario (F(2,16) = 26.7, p < 0.0001), and on participants’ self-reported trust in the vehicle (F(2,16) = 23.668, p < 0.0001). In all cases, as pod passing distance increased, participants reported significant increases in self-reported safety (Table 1 and Figs. 2 and 3). Table 1. Summary of significant main effects Question Q1 - Pod speed and safety Q2 - Distance from participant Q3 - General feeling of safety Q4 - General trust in vehicle

Distance main effect F(2,16) = 10.175, p < 0.001 F(2,16) = 25.956, p < 0.0001 F(2,16) = 26.7, p < 0.0001 F(2,16) = 23.668, p < 0.0001

Speed main effect F(2,16) = 15.047, p < 0.0001 F(2,16) = 1.566, p = N.S. F(2,16) = 7.514, p < 0.005 F(2,16) = 25.948, p < 0.0001

312

C. G. Burns et al.

Effects of Pod Distance 7

Mean Ra ng

6 5 4 3 2 1 0 Q1 - Veh. Speed

Q2- Veh. Distance Q3 - General Safety Q4 - General Trust Ques on Items 1 metre

2 metres

3 metres

Fig. 2. Mean ratings for pod distance effects per question. In every case, as pod passing distance increased, participants self-reported greater safety and trust, with LSD significance levels indicating all variables as significantly different from each other as distance increased (Table 2).

Effects of Pod Speed 7

Mean Ra ng

6 5 4 3 2 1 0 Q1 - Veh. Speed

Q2- Veh. Distance Q3 - General Safety Q4 - General Trust Ques on Items 1.75m/s

2.25m/sec

3m/sec

Fig. 3. Mean ratings for pod speed effects. For Q1, Q3 and Q4, increasing speed showed increasingly lowered self-reported safety and trust in the vehicle. No effects were recorded for Q2. For Q3 (general safety) participants’ self-reported safety did not distinguish between speeds of 1.75 m/sec and 2.25 m/sec where they did in Q1 and Q4. Participants felt least safe when the pod was at 3 m/sec.

Pedestrian Attitudes to Shared-Space Interactions

313

Table 2. Summary of significant post-hoc contrasts (LSD pairwise) Q1 Distance Mean Q1 Speed Mean 1m 5.983 1.75 m/sec 6.719 2m 6.378 2.25 m/sec 6.278 3m 6.611 3 m/sec 5.975 Q2 Distance Mean Q2 Speed Mean 1m 4.564 1.75 m/sec 6.019 2m 6.403 2.25 m/sec 5.922 3m 6.764 3 m/sec 5.789 Q3 Distance Mean Q3 Speed Mean 1m 4.886 1.75 m/sec 6.083 2m 6.183 2.25 m/sec 5.903 3m 6.531 3 m/sec 5.614 Q4 Distance Mean Q4 Speed Mean 1m 4.867 1.75 m/sec 6.039 2m 6.083 2.25 m/sec 5.819 3m 6.333 3 m/sec 5.425 (Mean values *with no significant post-hoc contrasts* of at least p < 0.05 are underlined, after the method described in [22])

4 Discussion Humans can estimate speeds and distances from visual observation, variously referred to as tau [23], relying largely on physiological retinal effects, or the newer time-tocontact concept (e.g. [24]) incorporating cognitive as well as biological phenomena. These mental faculties operate as there is an obvious survival value in knowing whether a moving object in the world (e.g. a vehicle) poses any physical relevance to one’s person. Pedestrians have also been shown to simultaneously underestimate approaching vehicles’ speeds and stopping distances [25]; a potentially hazardous combination. Similarly, it has been found that pedestrians base their road-crossing decisions “mainly…on the distance between them and the oncoming vehicle”, as well as the perceived time-to-contact which can be “easily misjudged” [26], perhaps due to underestimations of vehicle speed. In brief, pedestrian estimations of the risk posed by oncoming vehicles are not readily modelled with great accuracy, can be highly subjective, and can be influenced by a host of variables including their age, estimations of their own speed of mobility, cognitive function, choice reaction time and more [27]. With these phenomena in mind, it is relatively unsurprising that participants generated self-reports in line with logical expectations and established theory, although importantly this also indicates that the virtual experience we presented was naturalistic and convincing. Pod passing distance alone exerted a consistent effect among participants; when the pod was relatively further away (3 m), participants reported they felt safer and more trusting of the pod than when it passed by at any speed. Main effects of

314

C. G. Burns et al.

pod speed were nuanced. Consistent with the main effects of pod passing distance, participants seemed to focus more on the distance between the pod and themselves than their estimation of the pod’s speed. When asked specifically about their safety due to pod passing distances (Q2), there were no main effects of speed. Participants reported feeling least safe when the pod passed at 3 m/sec, and did not feel more or less safe at the pod’s two other passing speeds. Participants also self-reported significantly less general trust in the AV as its passing speed increased. Although our camera positions altered participants’ visual perception of the vehicle - position C1 gave participants a slightly longer exposure time to view the pod as it passed, while position C2 gave a view of the pod as it turned the corner before proceeding along the straight - neither position exerted any effects on participant self-reports. In summary, when the pod was closer to the camera (down to 1 m separation) and travelling faster (up to 3 m/sec, or approximately 6 pmh), participants reported feeling less safe and less trusting of the pod. This is noteworthy given that the pod is capable of travelling more than twice as fast, around 15 mph. It seems clear that during the introductory phases of autonomous transport vehicles in a locale, autonomous vehicles should be segregated from pedestrians such that they are clearly not a potential collision hazard, perhaps avoiding shared spaces entirely until bystanders become accustomed to them. Our volunteers clearly did not feel completely safe when in simulated proximity to the pod, although perhaps some of this insecurity is explained by a natural caution around vehicles arising from decades of exposure to roadtraffic and pedestrian safety information. The absence of any noteworthy cybersickness is a positive sign for continued use of VR as a stimulus presentation method. Although the scenarios themselves were brief, the experiment in total lasted approximately 30 min, yet participants self-reported little in the way of motion- or simulator-related sickness. This is likely due to our use of a static recording location – although participants could turn their head to follow the path of the AV, they could not “move” within the environment, similar to a pedestrian standing at a fixed location at a roadside. In contrast, using a fully rendered virtual environment simulating road-crossing behaviours, participant drop-outs of approximately 15% have been reported [6], the majority of which occurred within minutes of the study commencing. Our results were consistent with using a randomised presentation methodology and a relatively small sample size, indicating that contemporary, mainstream VR equipment has reached a stage where it is sufficiently convincing to the (mostly visual) senses to serve as a research tool where real-world, in vivo experimentation would be problematic or impossible. This is additionally interesting given that the scenarios we presented to participants were non-interactive video clips, where participants literally viewed events “from the sidelines”. Regarding our methodology, relatively inexpensive commercially available VR technology appears to have reached a level of fidelity where convincing sensory experiences (supported by statistical evidence) can apparently evoke genuine feelings and attitudes in participants in the absence of physical risk, opening up whole new avenues of research possibilities for vehicular and other research. This study was part of the UK Autodrive project, a flagship, multi-partner project, focusing on the development of the Human Machine Interface (HMI) strategies and performing real-world trials of these technologies in low-speed AVs (http://www. ukautodrive.com).

Pedestrian Attitudes to Shared-Space Interactions

315

References 1. Kun, A.L., Boll, S., Schmidt, A.: Shifting gears: user interfaces in the age of autonomous driving. IEEE Pervasive Comput. 15, 32–38 (2016). https://doi.org/10.1109/MPRV.2016.14 2. Meschtscherjakov, A., Tscheligi, M., Fröhlich, P., McCall, R., Riener, A., Palanque, P.: Mobile interaction with and in autonomous vehicles. In: Proceedings of the 19th International Conference on Human-Computer Interaction with Mobile Devices and Services - MobileHCI 2017, pp. 1–6 (2017). https://doi.org/10.1145/3098279 3. Fu, X., Vernier, M., Kurt, A., Redmill, K., Ozguner, U.: Smooth: improved short-distance mobility for a smarter city. In: Proceedings of the 2nd International Workshop on Science of Smart City Operations and Platforms Engineering - SCOPE 2017, pp. 46–51 (2017). https:// doi.org/10.1145/3063386.3063760 4. Pendleton, S., Uthaicharoenpong, T., Chong, Z.J., Fu, G.M.J., Qin, B., Liu, W., Shen, X., Weng, Z., Kamin, C., Ang, M.A., Kuwae, L.T., Marczuk, K.A., Andersen, H., Feng, M., Butron, G., Chong, Z.Z., Ang, M.H., Frazzoli, E., Rus, D.: Autonomous golf cars for public trial of mobility-on-demand service. IEEE International Conference on Intelligent Robots and Systems, December 2015, pp. 1164–1171 (2015). https://doi.org/10.1109/iros.2015. 7353517 5. Merat, N., Louw, T., Madigan, R., Wilbrink, M., Schieben, A.: What externally presented information do VRUs require when interacting with fully Automated Road Transport Systems in shared space? Accid. Anal. Prev. 118, 244–252 (2018). https://doi.org/10.1016/j. aap.2018.03.018 6. Deb, S., Carruth, D.W., Sween, R., Strawderman, L., Garrison, T.M.: Efficacy of virtual reality in pedestrian safety research. Appl. Ergon. 65, 449–460 (2017). https://doi.org/10. 1016/j.apergo.2017.03.007 7. Simpson, G., Johnston, L., Richardson, M.: An investigation of road crossing in a virtual environment. Accid. Anal. Prev. 35, 787–796 (2003). https://doi.org/10.1016/S0001-4575 (02)00081-7 8. Chittaro, L., Zangrando, N.: The persuasive power of virtual reality: effects of simulated human distress on attitudes towards fire safety. In: Ploug, T., Hasle, P., Oinas-Kukkonen, H. (eds.) Persuasive Technology, pp. 58–69. Springer, Heidelberg (2010) https://doi.org/10. 1007/978-3-642-13226-1_8 9. Sheridan, T.B.: Musings on telepresence and virtual presence. Presence Teleoperators Virtual Environ. 1, 120–126 (1992). https://doi.org/10.1162/pres.1992.1.1.120 10. Whitelock, D., Romano, D., Jelfs, A., Brna, P.: Perfect presence: what does this mean for the design of virtual learning environments? Educ. Inf. Technol. 5, 277–289 (2000) 11. Keferböck, F., Riener, A.: Strategies for negotiation between autonomous vehicles and pedestrians. In: Mensch Und Computer 2015 Workshopband, pp. 525–532. Oldenbourg Wissenschaftsverlag, Stuttgart (2015) 12. Böckle, M.-P., Brenden, A.P., Klingegård, M., Habibovic, A., Bout, M.: SAV2P – exploring the impact of an interface for shared automated vehicles on pedestrians’ experience. In: Proceedings of the 9th International Conference on Automotive User Interfaces and Interactive Vehicular Applications Adjunct - AutomotiveUI 2017, pp. 136–140 (2017). https://doi.org/10.1145/3131726.3131765 13. Chang, C., Toda, K., Sakamoto, D., Igarashi, T.: Eyes on a car: an interface design for communication between an autonomous car and a pedestrian. In: Proceedings of the 9th International Conference on Automotive User Interfaces and Interactive Vehicular Applications (AutomotiveUI 2017), pp. 65–73 (2017). https://doi.org/10.1145/3122986. 3122989

316

C. G. Burns et al.

14. NHTSA. Automated Vehicles for Safety [WWW Document]. National Highway Traffic Safety Administration (2018). https://www.nhtsa.gov/technology-innovation/automatedvehicles-safety. Accessed 13 Apr 18) 15. BBC. Tesla in fatal California crash was on Autopilot. BBC News March 31th (2018) 16. Levin, S., Wong, J.C.: Self-driving Uber kills Arizona woman in first fatal crash involving pedestrian. Guardian (UK Edition), 19 March 2018 (2018) 17. Perelló-March, J., Burns, C., Elliott, M., Stewart, B.: Integrating trust in automation into driver state monitoring systems. In: International Conference on Human Interaction & Emerging Technologies, 22–24 August 2019. Université Côte d'Azur, Nice (2019) 18. Mirnig, A.G., Wintersberger, P., Sutter, C., Ziegler, J.: A framework for analyzing and calibrating trust in automated vehicles. In: Proceedings of the 8th International Conference on Automotive User Interfaces and Interactive Vehicular Applications Adjunct - Automotive’UI 2016, pp. 33–38. (2016). https://doi.org/10.1145/3004323.3004326 19. Fagnant, D.J., Kockelman, K.: Preparing a nation for autonomous vehicles: opportunities, barriers and policy recommendations. Transp. Res. Part A Policy Pract. 77, 167–181 (2015). https://doi.org/10.1016/j.tra.2015.04.003 20. Rödel, C., Stadler, S., Meschtscherjakov, A., Tscheligi, M.: Towards autonomous cars: the effect of autonomy levels on acceptance and user experience. In: Proceedings of the 6th International Conference on Automotive User Interfaces and Interactive Vehicular Applications, pp. 1–8 (2014). https://doi.org/10.1145/2667317.2667330 21. Kennedy, R.S., Lane, N.E., Kevin, S., Lilienthal, M.G.: Simulator sickness questionnaire: an enhanced method for quantifying simulator sickness. Int. J. Aviat. Psychol. 3, 203–220 (1993). https://doi.org/10.1207/s15327108ijap0303 22. Duncan, D.B.: Multiple range and multiple F tests. Biometrics 11, 1–42 (1955) 23. Lee, D.N., Young, D.S., Reddish, P.E., Lough, S., Clayton, T.M.: Visual timing in hitting an accelerating ball. Q. J. Exp. Psychol. A 35, 333–346 (1983) 24. Smeets, J.B., Brenner, E., Trébuchet, S., Mestre, D.R.: Is judging time-to-contact based on ‘tau’? Perception 25, 583–590 (1996) 25. Sun, R., Zhuang, X., Wu, C., Zhao, G., Zhang, K.: The estimation of vehicle speed and stopping distance by pedestrians crossing streets in a naturalistic traffic environment. Transp. Res. Part F: Traffic Psychol. Behav. 30, 97–106 (2015) 26. Liu, Y.-C., Tung, Y.-C.: Risk analysis of pedestrians’ road-crossing decisions: effects of age, time gap, time of day, and vehicle speed. Saf. Sci. 63, 77–82 (2014) 27. Butler, A.A., Lord, S.R., Fitzpatrick, R.C.: Perceptions of speed and risk: experimental studies of road crossing by older people. PLoS One 11(4), e0152617 (2016). https://doi.org/ 10.1371/journal.pone.0152617

Driving Behavior: Safety and Simulation

Speed Behavior in a Suburban School Zone: A Driving Simulation Study with Familiar and Unfamiliar Drivers from Puerto Rico and Massachusetts Didier Valdés1(&), Michael Knodler2, Benjamín Colucci1, Alberto Figueroa1, Maria Rojas1, Enid Colón1, Nicholas Campbell2, and Francis Tainter2 1

University of Puerto Rico at Mayaguez, Mayaguez, PR, USA {didier.valdes,benjamin.colucci1,alberto.figueroa3, maria.rojas7,enid.colon1}@upr.edu 2 University of Massachusetts at Amherst, Amherst, MA, USA {mknodler,nlcampbell,ftainter}@umass.edu

Abstract. Traffic crashes in suburban school zones pose a serious safety concern due to a higher presence of school-age pedestrians and cyclists as well as potential speeding issues. A study that investigated speed selection and driver behavior in school zones was carried out using two populations from different topographical and cultural settings: Puerto Rico and Massachusetts. A school zone from Puerto Rico was recreated in driver simulation scenarios and local drivers who are familiar with the environment were used as subjects. The Puerto Rico school simulation scenarios were replicated with subjects from Massachusetts to analyze the impact of drivers’ familiarity on the school-roadway environment. Twenty-four scenarios were built with pedestrians, on-street parked vehicles, and traffic flow used as simulation variables in the experiment. Results are presented in terms of speed behavior, reaction to the presence of pedestrians, speed compliance, and mean reduction in speeds for both familiar and unfamiliar drivers. Keywords: Driving simulator
Unfamiliar drivers
School zone Driving behavior
Human factors
Speed behavior




1 Introduction Pedestrian fatalities in the United States increased in the last decade. Traffic crashes in school zones are a serious safety concern. According to the NHTSA, there were an average of 128 fatalities per year in school-transportation-related crashes in the US and Puerto Rico from 2007 to 2016 [1]. School zones are areas with a high presence of vulnerable users of young age, which increases the risk of crashes. Several studies have shown that a significant number of drivers violate the posted speed limit in school zones [2–4] and approximately one-third of the traffic fatalities in 2016 involved speeding behavior [5]. © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 319–329, 2020. https://doi.org/10.1007/978-3-030-20503-4_30

320

D. Valdés et al.

School zones typically require a transition from high-speed to low-speed areas to comply with the posted speed limits needed to adequately manage pedestrians’ presence. The transition zones, in terms of driver expectancy in a suburban road, represent a safety management problem. This transition zone phenomenon is also present in suburban roads with high operating speeds where drivers tend not to adequately comply with the posted speed limit of lower speeds areas [6]. In Puerto Rico, there are school zones located in areas adjacent to major arterial streets with a posted speed limit of 40 mph or more. A recent study conducted in school zones in the western region of Puerto Rico show that drivers’ mean speeds were higher than the posted speed limit in 63% of the evaluated school zones [7]. Different countermeasures have been developed and studied to improve school zone safety. The Safe Routes to Schools (SRTS) Program was developed by the Federal Highway Administration (FHWA) to promote healthy and equitable mobility for everyone, to increase safety in the vicinity of school zones, and to raise awareness of the benefits of walking and biking. Several research studies have focused on the effects of road environment characteristics on drivers’ speed in school zones to increase compliance and improve safety and those zones with high sign saturation resulted in drivers exhibiting lower speeds and higher compliance [8]. Also, the implementation of speed display devices has had positive results in terms of reducing speed violations in school zones and playgrounds [9]. The Manual on Uniform Traffic Control Devices (MUTCD) specifies the standards by which all traffic control devices (TCD) in public roads are installed and maintained [10]. Part 7 of the MUTCD has the standards, guidance and options for TCDs applicable for school zones. In the case of Puerto Rico, traffic control devices substantially comply with the MUTCD, but with Spanish text. It is pertinent to recognize that for unfamiliar drivers (i.e., tourists and first-time users), the difference in language and general highway environment may generate additional challenges to driving tasks. Throughout the years, human factors research has identified the importance of a driver’s familiarity with the road environment. Intini et al. have shown that familiarity is an influential factor on crash risk; due to either distraction or over-confidence [11]. Therefore, new designs should consider unfamiliar as well as familiar users to improve compliance with regulations and enhance safety and mobility. Driving simulators at the University of Puerto Rico at Mayaguez (UPRM) and University of Massachusetts at Amherst (UMass) were used to conduct experiments that aimed to analyze drivers’ responses to changes in road infrastructure configuration, school zone speed limits, and roadway signage. Simulators have been used as an innovative and cost-effective research tool to evaluate the behavior in a wide diversity of research fields such as human factors, transportation, psychology, medicine, computer science, training, and other driving activities [12]. Simulators are useful for evaluating existing and emerging transportation treatments without exposing subject drivers to physical harm in scenarios where a potential crash may occur. The UPRM-UMass collaborative research includes the assessment of TCD configurations and understanding unfamiliar drivers’ behavior along suburban school zone scenarios. Familiar and unfamiliar drivers’ behavior was compared for a school zone in Puerto Rico (PR) with different roadway scenarios to determine whether familiarity influences driver performance.

Speed Behavior in a Suburban School Zone

321

2 Objective The objective of this collaborative research is to evaluate speed behavior and compliance in school zones for familiar and unfamiliar drivers using driving simulation. The research evaluated the best TCD configuration to maximize the drivers’ speed limit compliance rate in school zones and tested whether there any significant differences in behavior between familiar and unfamiliar drivers in the school zone.

3 Methodology The research methodology included a literature review on school zone safety and speeds, TCDs, and how unfamiliar environments affect drivers’ behavior. A base simulation scenario was then constructed by recreating the characteristics of a suburban school zone in Puerto Rico. The school zone was selected using a procedure that included a detailed screening process based on a Highway Safety Manual (HSM) concepts and road safety audit (RSA). The conditions at the chosen school zone were then inspected to generate scenarios and define the variables for the simulation scenarios. A survey conducted among Puerto Rican drivers was used to select the preferred combination of signage and pavement markings that most effectively conveyed the message of speed reduction in a school zone. Based on the 196 responses received, the preferred TCD combination included: SCHOOL pavement marking symbols next to the S1-1 school zone warning sign followed by an overhead sign showing the school speed limit and flashing beacons with the END OF SCHOOL ZONE sign at the end of the school area. Both existing and proposed TCD configurations were tested using simulation experiments conducted with the UPRM and UMass driving simulators with drivers who were familiar and unfamiliar with the original roadway environment. Comparisons were made between the behavior of familiar and unfamiliar drivers in the different scenarios. 3.1

Driving Simulator Equipment

The driving simulator located in the UPRM consists in a desktop simulator configured as a portable cockpit simulator with three main components: a driving cockpit, visual display, and computer system. The driving cockpit consists of a car seat, steering wheel, gear shifter, two turn signals, and the acceleration and braking pedals all mounted in a wooden base that has six wheels; making it compatible with mobile applications. The visual display consists of three overhead projectors and three screens that generates 120° of road visibility at 1080p resolution. Finally, the computer system uses a laptop and a desktop computer with the Realtime Technologies Inc. (RTI) SimCreator/SimVista simulation software and an audio system that represents the vehicle and environment noises. The driving simulator used by UMass for this study is a fixed-base simulator with full-body Ford Fusion Sedan model 2013. The visual display for this equipment consists of five main projectors with a resolution of 19020  1200 pixels, one rear

322

D. Valdés et al.

projector with a resolution of 1400  1050 pixels, and six screens that generates a field of view of approximately 330°. The sound system used to recreate the vehicle and environmental noises consists of a five-speaker surround system plus a sub-woofer for exterior noise and a two-speaker system plus a sub-woofer for interior vehicle noise. The computer system uses two desktop computers with the RTI SimCreator/SimVista simulation software. 3.2

Scenario Development

School Zone. The school zone selected for this study is the Second Unit Samuel Adams in the municipality of Aguadilla in Puerto Rico. This school is in a suburban area and provides a level of education from Pre-Kinder to 9th grade with a student enrollment of 900 children approximately. This school has direct access from the arterial highway PR-2. In this section, this highway has 2 lanes per direction, an AADT of 42,900 vpd and a posted speed limit of 25 mph in the school zone and 45 mph elsewhere [3]. TCD Configuration. A site inspection performed in this school zone showed that, currently, the speed limit and school zone signs have not been updated to the fluorescent yellow-green color indicated in the last version of the MUTCD. Also, it was found that there are yellow transversal lines used to delimit the beginning and end of the school zone, as required by Puerto Rico’s Traffic Law. There is no pavement marking with the word “School” in the existing conditions. There is no END OF SCHOOL ZONE sign at the end of the school zone as required by the MUTCD. Therefore, this study aims to study familiar and unfamiliar drivers’ behavior not only with the base configuration of signage and pavement markings of the school, but also in a recommended configuration implementing the guidelines for school zones specified in the MUTCD and the preferred enhanced sign selected in the online survey. Figure 1 shows the comparison between the signage and pavement markings that were used for the configuration of the scenarios and the location of each sign in the school zone area. There are five zones of interest. Zone 0 refers to the segment of the road where the subjects are travelling at free flow speed, before the school zone signage and pavement markings. Zone 1 is the area prior to the school zone warning sign. Zone 2 corresponds to the area between the school zone warning sign and the school speed limit sign (roadside or overhead, respectively). Zone 3 corresponds to a location in the vicinity of the school driveway where one pedestrian walks on the shoulder near the right travel lane in direction toward the oncoming traffic. Vehicles parked at an angle in the right shoulder are also present in this zone. Zone 4 represents the end of the school zone identified with the last traffic control device on each configuration. 3.3

Experimental Design

A factorial design with two blocks was used for this experiment. The factors considered in this study were: traffic, pedestrian presence, vehicles parked in shoulder, and configuration. The Traffic factor represents if ambient traffic is included or not on the scenarios, with two levels: moderate number of vehicles and no vehicles.

Speed Behavior in a Suburban School Zone

323

Fig. 1. Signage and pavement marking configuration.

The Pedestrians factor denotes the presence of pedestrians on the sidewalks near the school zone with three different levels: no pedestrians, adults and children, and only children. The Vehicles Parked in Shoulder factor represents the presence of vehicles parked on the right-side shoulder in front of the school, with two levels: parked vehicles and no parked vehicles. The Configuration factor was used for the blockage: base configuration and recommended configuration. A total of twelve scenarios for each configuration were developed to evaluate each combination of the factors. Table 1 shows a description of the experimental scenarios. A total of 72 subjects participated in the collaborative project. Two groups of 36 subjects were recruited from Puerto Rico and from Massachusetts. Each group at each university were divided in two samples of 18 subjects. Each sample drove in one configuration. The scenarios in each configuration were shown to each subject in random order.

4 Analysis of Results The analyses of the driving simulation experiments concentrated on the following variables: speed behavior, influence of pedestrian presence and speed limit compliance. The comparison of the results shown were between the behavior of familiar and

324

D. Valdés et al. Table 1. Experimental scenarios Vehicles Parked on Shoulder

Pedestrians Scenario

1 2 3 4 5 6 7 8 9 10 11 12

Adults and Children x x x x

No Pedestrians

Only children

Yes

Traffic

No

x x

x

x x

x x x x

x x x x x x

No

x x x

x x x x

Yes

x x

x x x x

x x

x x

unfamiliar drivers between Zone 0 and Zone 3. The familiar drivers correspond to the subjects recruited at the UPRM and the unfamiliar drivers correspond to the subjects recruited at UMass. The local drivers at UPRM are assumed to be familiar with the roadway-school environment and the existing TCDs of Configuration 1. The UMass drivers are assumed to be unfamiliar and treated as first-time driver in the existing TCDs of Configurations 1 and 2. A second factor to consider in the evaluation of the differences is the use of Spanish-text on the TCDs. 4.1

Speed Behavior

Table 2 shows the results from the statistical test of the difference in mean speeds between each group of subjects (familiar vs. unfamiliar) and for each configuration. The results show that familiar and unfamiliar drivers had similar mean speeds at Zone 0 (before the school zone) for all scenarios in Configuration 1. When comparing speeds at Zone 0 in Configuration 2 with the enhanced TCDs, larger differences are observed between the groups of drivers. In this case, unfamiliar drivers significantly had higher mean speeds at Zone 0 for 67% of the scenarios, in a range of 3.3 to 8.0 mph. When observing mean speeds at Zone 3, with drivers already inside the school zone, there are significant differences between familiar and unfamiliar drivers at a 5% significance level. These differences were observed in 75% of the scenarios for Configuration 1 with the existing TCDs and in 92% of the scenarios for Configuration 2 with the enhanced TCDs. The overall trend is that familiar drivers selected lower speeds than unfamiliar drivers.

Speed Behavior in a Suburban School Zone

325

Table 2. Statistical analysis of mean speed for unfamiliar vs familiar drivers.

–

–

–

– –

4.2

Reaction to the Presence of Pedestrians/Avatars

Figure 2 shows the trajectories of individual subjects along Scenario 12 for both driver samples at each configuration. Scenario 12 included only the presence of children pedestrians with no vehicles parked on the shoulder and no ambient traffic in the simulation. Besides avatars present on the sidewalks, there was an additional avatar that was walking along the shoulder in the direction toward traffic between Zones 2 and 3. The sudden speed reductions observed between Zones 2 and 3 for the individual speed trajectories in the first three of the graphs of Fig. 2 reflect that the driver reacted to the presence of the pedestrian on the shoulder. Most of these drivers were traveling at speeds well above the school zone speed limit (over 30 mph) before applying the brakes. Of all unfamiliar drivers traveling in Configuration 1 and 2, 33% and 29% of the drivers, respectively, reduced their speeds in reaction to the presence of the child

326

D. Valdés et al.

a)

b)

Unfamiliar drivers

Familiar drivers

Fig. 2. Speed profiles of individual drivers along school zones

pedestrian. The trend observed for familiar drivers showed 33% and 19% of all drivers reduced their speeds between Zones 2 and 3 for Configurations 1 and 2, respectively. 4.3

Speed Compliance

Table 3 shows the speed limit compliance in percentage between familiar and unfamiliar drivers on Zones 0 and 3. For Zone 0, familiar drivers had a higher compliance percentage in 75% of the scenarios. The overall trend in Zone 3 is that familiar drivers always have a higher compliance as opposed to with unfamiliar drivers. In terms of the enhanced TCDs effectiveness of for improving speed limit compliance, the results show

Speed Behavior in a Suburban School Zone

327

that unfamiliar drivers improve their compliance on 25% of the scenarios, whereas familiar drivers improve compliance on 67%. Improvement was defined as an increase of 1% or higher.

Scenario

Configurat ion

Table 3. Speed limit compliance between familiar and unfamiliar drivers.

1 2 3 4 5 6 7 8 9 10 11 12

Unfamiliar Drivers (%) Zone 0 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2

44.44 22.22 44.44 29.41 38.89 22.22 44.44 29.41 27.78 22.22 55.56 17.65 50.00 5.56 50.00 44.44 33.33 27.78 33.33 38.89 38.89 27.78 66.67 16.67

Zone 3 5.56 0.00 11.11 11.76 0.00 11.11 5.56 5.88 11.11 0.00 5.56 11.76 5.56 0.00 5.56 0.00 0.00 11.11 16.67 11.11 16.67 16.67 16.67 16.67

Familiar Drivers (%) Zone 0 46.67 43.75 53.33 56.25 13.33 43.75 33.33 37.50 33.33 50.00 40.00 43.75 20.00 25.00 26.67 25.00 33.33 56.25 46.67 56.25 40.00 43.75 53.33 56.25

Zone 3 60.00 62.50 66.67 62.50 40.00 43.75 46.67 68.75 46.67 31.25 40.00 68.75 46.67 62.50 66.67 62.50 60.00 62.50 66.67 62.50 53.33 68.75 73.33 81.25

Table 4 shows the mean reduction in speeds for familiar and unfamiliar drivers between Zones 0 and 3 for those scenarios without ambient traffic. The expected minimum reduction in speeds was 20 mph for this school zone (from 45 to 25 mph). None of the scenarios exhibited the expected speed reduction. In 83% of the scenarios, familiar drivers had larger reductions in mean speeds than for unfamiliar drivers. On 67% of the scenarios (4 out of 6), the enhanced TCDs had the expected effect of achieving a higher speed reduction for unfamiliar drivers, even though the speed compliance on Zone 3 was between 0% and 16.67%. The positive effect of the enhanced TCDs for the familiar drivers was only observed on 17% of the scenarios (1 out of 6). However, the range for the compliance rate was between 33% and 81% for familiar drivers.

328

D. Valdés et al.

Scenario

Table 4. Mean reduction in speeds between Zones 0 and 3.

2

Unfamiliar Drivers Configuration 1 PValues 0.05).

5 Conclusion In this paper, we proposed and implemented the experiment to study Holding Stack Management, Continuous Climb Operations, Continuous Descent Operations, and Trajectory Based Operations procedures in relation to the use of the additional 3D

Human Factors Evaluation of ATC Operational Procedures

725

display in 2D+3D setting. Twelve Air Traffic Control officers (ATCOs) were invited to take part in the experiment which used the EEG-based emotion and stress recognition algorithms to evaluate whether the additional 3D display setup can be beneficial to ATCOs when HSM, CCO, CDO, and TBO operational procedure were performed. A 30-minute scenario was implemented and performed in 2D display only and 2D+3D settings. EEG data were recorded and traditional human factors questionnaires were given to the participants. State-of-the-art algorithms of cognitive workload, emotion and stress recognition from EEG were implemented to process and analyse the data. The results of the data analyses showed that by using 2D+3D display setting, more positive emotions were experienced by ATCOs in TBO, CCO and CDO procedures than in 2D setting; however, they had higher stress and workload levels than in 2D setting in those procedures. In HSM, reduced stress and significantly lower workload were experienced by ATCOs when they were using 2D+3D setting. Thus, ATCOs benefited from use of 3D visualization in HSM operational procedure. The experiment results showed that further improvement of the 3D display implementation and analyses of operational procedures in relation to 3D visualization could reduce cognitive workload and stress of ATCOs in increasing traffic density demand. Acknowledgments. This research is supported by Civil Aviation Authority of Singapore (CAAS), Air Traffic Management Research Institute (ATMRI) Project ATMRI: 2014-R5-CHEN and by the National Research Foundation, Prime Minister’s Office, Singapore under its international Research Centres in Singapore Funding Initiative. We would like to acknowledge the final year project students of School of MAE of Nanyang Technological University for their contribution in this work.

References 1. Lapin, K., Čyras, V., Savičienė, L.: Visualization of aircraft approach and departure procedures in a decision support system for controllers. In: Proceedings of the Ninth International Baltic Conference on Databases and Information Systems (2010) 2. Prevot, T., et al.: Toward automated air traffic control—investigating a fundamental paradigm shift in human/systems interaction. Int. J. Hum.-Comput. Interact. 28(2), 77–98 (2012) 3. Erzberger, H.: Transforming the NAS: the next generation air traffic control system (2004) 4. Isaac, A.R., Ruitenberg, B.: Air Traffic Control: Human Performance Factors. Routledge, New York (2017) 5. Wong, B., et al.: 3D-in-2D displays for ATC (2007) 6. National Research Council: Flight to the Future: Human Factors in Air Traffic Control. National Academies Press, Washington, D.C. (1997) 7. Masotti, N., Persiani, F.: On the history and prospects of three-dimensional human– computer interfaces for the provision of air traffic control services. CEAS Aeronaut. J. 7(2), 149–166 (2016) 8. Hou, X., et al.: EEG-based human factors evaluation of conflict resolution aid and tactile user interface in future Air Traffic Control systems. In: Advances in Human Aspects of Transportation, pp. 885–897. Springer, Heidelberg (2017) 9. IVAO. https://www.ivao.aero/ViewDocument.aspx?Path=/training:atc:docs

726

Y. Liu et al.

10. Melby, P., Mayer, R.: Benefit potential of continuous climb and descent operations. In: The 26th Congress of ICAS and 8th AIAA ATIO (2008) 11. Clarke, J.-P., et al.: Optimized profile descent arrivals at Los Angeles international airport. J. Aircraft 50(2), 360–369 (2013) 12. Gardi, A., et al.: 4 Dimensional trajectory functionalities for air traffic management systems. In: 2015 Integrated Communication, Navigation, and Surveillance Conference (ICNS). IEEE (2015) 13. Loft, S., et al.: Modeling and predicting mental workload in en route air traffic control: critical review and broader implications. Hum. Factors 49(3), 376–399 (2007) 14. Miller, S.: Workload Measures. National Advanced Driving Simulator, Iowa City (2001) 15. Hart, S.G., Staveland, L.E.: Development of NASA-TLX (Task Load Index): results of empirical and theoretical research. Adv. Psychol. 52, 139–183 (1988) 16. Liu, Y., Sourina, O.: Real-time subject-dependent EEG-based emotion recognition algorithm. In: Transactions on Computational Science XXIII, pp. 199–223. Springer, Heidelberg (2014) 17. Lim, W.L., et al.: EEG-based mental workload recognition related to multitasking. In: Proceeding of the International Conference on Information, Communications and Signal Processing (ICICS) (2015) 18. Lim, W.L., et al.: EEG-based mental workload and stress monitoring of crew members in maritime virtual simulator. In: Gavrilova, M.L., Tan, C.J.K., Sourin, A. (eds.) Transactions on Computational Science XXXII: Special Issue on Cybersecurity and Biometrics, pp. 15– 28. Springer, Heidelberg (2018) 19. Jian, J.-Y., Bisantz, A.M., Drury, C.G.: Foundations for an empirically determined scale of trust in automated systems. Int. J. Cogn. Ergon. 4(1), 53–71 (2000) 20. Emotiv. http://www.emotiv.com

Monitoring Performance Measures for Radar Air Traffic Controllers Using Eye Tracking Techniques Hong Jie Wee1,2(&), Sun Woh Lye1, and Jean-Philippe Pinheiro3 1

3

School of Mechanical and Aerospace Engineering, Nanyang Technological University, 50 Nanyang Avenue, Block N3, Singapore 639798, Singapore [email protected] 2 Thales Solutions Asia Pte. Ltd., 21 Changi North Rise, Singapore 498788, Singapore Thales LAS France SAS, 3, Avenue Charles Lindbergh, 94628 Rungis, France

Abstract. This paper presents an approach describing how Air Traffic Controller (ATCO) fixations can be mapped to dynamic moving flight objects (track and label) on the radar screen in real-time, using a remote eye tracker. Real time simulations were conducted for 30 one-hour experimental sessions with participants from three expertise levels, using scenarios that mimic actual air traffic, consisting of both wide and medium angle crossing points. Monitoring performance metrics using fixation count and duration on an aircraft’s flight object on the radar screen were investigated in a macroscopic one-hour duration and four minutes before a crossing point for both wide and medium angle crossings. Distinct differences in the monitoring behavior of participants were found in the macroscopic one-hour duration and wide angle crossing. Four new parameters relating to the fixation counts and durations on dynamic flight objects, which could be used to distinguish the expertise level of ATCOs, were established. Keywords: Tactical air traffic control
Eye tracking Monitoring performance
Modelling and simulation




1 Introduction Dynamic changes in Air Traffic Management (ATM) in recent years have brought about major developments in Air Traffic Control (ATC). One emerging technological approach involves the adoption of 4-Dimensional Trajectory Based Operations (4DTBO), where the flight trajectory of aircraft can be predicted in a timely, accurate and consistent manner. As a result, aircrafts are freer to fly on their own desired trajectories, creating a more complex air route structure with tighter separation standards [1–4]. In tactical monitoring, radar Air Traffic Controllers (ATCOs) are required to monitor the aircraft as they fly through their airspace that were depicted on their radar screens. ATCOs will endeavor to create and maintain a mental picture of where the aircraft were on the radar screen [5]. Incident reviews have highlighted crossing point angles and lack of experience of ATCOs in tactical radar monitoring tasks as reasons © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 727–738, 2020. https://doi.org/10.1007/978-3-030-20503-4_65

728

H. J. Wee et al.

for loss of separation and ineffective monitoring. According to incident reports from the Skybrary database, findings on an incident involving AFR 989Z and RYR 1702 in 2009 highlighted the lack of experience in the newly trained ATCO in which he was unaware of where he should be looking at, on the radar screen. In addition, a complex air route design having different crossing point angles in the airspace resulted in the misjudgment of separation standards by ATCOs, with one example seen in the case of two Qantas aircraft flying through the en-route airspace in 2013. With a more complex air route design predicted in future and the presence of a variety of crossing point angles that are prevalent in the airspace today, more conflicts between aircrafts are expected. Based on present practice, this would invariably impact the ATCO’s operational performance and taskload during tactical radar monitoring sessions. Research studies have therefore been conducted that seek to better understand and deal with the tactical monitoring behavior of ATCOs, for current and future ATC operations [1, 6–8]. Eye tracking technique is being deployed increasingly in many surveillance fields like ATC [9, 10], driving [11, 12], pilot assessment [13, 14], healthcare [15], as a method for monitoring and analyzing the viewing behavior of an ATCO. It is deemed to be the most direct and objective form of measurement of an ATCO’s visual monitoring behavior [16] and it can also provide behavioral metrics reflecting cognitive activity as well as to predict human error [17, 18]. In ATC, previous studies using eye tracking techniques found that differences do exist in the visual monitoring behavior ATCOs of different expertise levels. However, such studies tend to be simplistic while others failed to capture a wide spectrum of air traffic situations with little attempt made to map the eyeball movement to the radar data in real-time. As a result, this paper presents an approach describing how ATCOs’ fixations can be mapped to the dynamic moving flight objects (track and label) on the radar screen in real-time, using a remote eye tracker. Tactical monitoring behavior and monitoring performance evaluation of different ATCO expertise level is also demonstrated in this paper.

2 Related Work In recent years, eye tracking techniques have been increasingly adopted by researchers in the field of ATC. Studies were conducted using eye tracking techniques to detect the difference in monitoring performance, mainly for distinguishing the expertise level of ATCOs [19]. A Multifractal Detrended Fluctuation Analysis method was applied to the time series eye metric data to classify novice and expert ATCOs [20]. It is found that an expert ATCO uses an efficient search strategy by fixating on core targets. A series of air traffic metrics was also examined to investigate their correlations with eye movement measures, serving as a suitable reference for future studies [21]. However, no studies were made between a specific aircraft of interest with corresponding set of eye metrics for radar monitoring. On the angle of crossing, a critical airspace factor, a study was also made to determine the effect it has on the monitoring performance of ATCOs, using an eye tracker [22]. Results from this study showed that large saccade amplitudes do change

Monitoring Performance Measures for Radar Air Traffic Controllers

729

significantly with the angle of crossing and the ATCO’s performance is positively related to the number of fixations and saccades. However, this study used a simple one crossing configuration involving two aircraft encountering a single conflict in the entire airspace, with angles variations at 30°, 60°, 90° and 120°. Hence, these results only highlight the influence of geometry had on an ATCO’s monitoring based on a specific measurement applicable in that situation. It is not robust in determining the expertise level of ATCOs in an operational airspace with many aircraft flying across different angular crossing points. Eye tracking techniques were also used to distinguish the expertise level of ATCOs by modelling their visual scanning and aircraft selection behaviors with a Maximum Transition-based Agglomerative Hierarchical Clustering (MTAHC) algorithm [23, 24]. Results from this study have successfully characterized the visual scanning behavior between expertise level of ATCOs. Similarly, these studies were made based on aircraft in conflict situation, which is a specific measure and not suitable for measuring the monitoring behavior of ATCOs in general.

3 Method 3.1

Experimental Sessions

In this paper, 30 one-hour experimental sessions of data were collected among male participants whose ages varied between 24 and 41 (M = 29, SD = 5.34). These sessions comprised of 10 novice sessions (zero ATC experience, only ATC knowledge), 10 intermediate sessions (more than 50 h of simulator ATC training and ATC knowledge, yet to handle live traffic) and 10 expert sessions (licensed ATCOs with more than three months of live traffic experience) [25]. 3.2

Experimental Setup

For the experiment, a real-time simulator, the NLR ATM Research SIMulator (NARSIM), was used to generate and simulate air traffic scenarios in real time, with every radar update at 9.8 s. Pseudo-pilots were used to facilitate the flying of aircraft, by following the same scenario script, so that events in the same scenario occurred at the same instant. A standard Controller Working Position (CWP), consisting of a 2 K radar screen (50 cm by 50 cm, equivalent to 20 inches by 20 inches) was used to display air traffic data in a given airspace sector, as seen in Fig. 1. A remote eye tracker, Tobii X2-30 was used to capture the eye data of the participants in a non-intrusive manner. Under ideal conditions, its accuracy and precision are 0:4 and 0:26 respectively [26]. In this setup, the eye tracker is positioned 18 cm in front of the screen, 55 cm from the participant’s eye, to extend the eye tracker’s field of view so that gaze data on any regions of the 2 K screen can be captured. This meant that the eye tracker is deployed in a non-ideal condition that closely mimic the ATCOs’ working environment. Hence, a study was conducted to determine the accuracy and precision of the eye tracker without a chinrest, prior to this study, for this non-ideal condition [27–29]. For a total of 3128 gaze data samples, the accuracy and precision of

730

H. J. Wee et al.

Fig. 1. A standard Controller Working Position (CWP) used to conduct real time simulations

the Tobii X2-30 was found to be 3:5 and 1:1 respectively, with good validity of 93.7%. These values of accuracy and precision of the Tobii X2-30 is quite close to that reported by Clemotte et al., which is 2:5 and 2:0 respectively, performed using for the same eye tracker model under non-ideal conditions Raw eyeball movement and radar data are extracted from the Tobii eye tracker and NARSIM respectively via an intelligent real-time post processing server that has been developed, which ensures that the two data streams are synchronous in time. The two sets of extracted data are then computed to determine the various eye tracking metrics. The metrics data were then relayed to the Thales TopSky-HF software to record the data of the different test participants [30, 31]. 3.3

Experimental Procedure and Scenario Definition

Prior to the experiment, each participant involved in a session underwent an initial 9point eye-tracker calibration exercise. Calibration was repeated until a satisfactory calibration was obtained that makes reference to Tobii Studio’s qualitative calibration report. Real time simulations were conducted for 30 one-hour experimental sessions, with 60 aircraft that mimic actual air traffic, consisting of both wide and medium angle crossing points. The airspace of interest in this real time study was based on an existing airspace sector, with information referenced from the publicly available SkyVector website [32]. A wide angle crossing point is defined as a crossing point, with an angular difference of 45 to 180° between the two intersecting airways, whereas a medium angle crossing point is one with an angular difference of 16 to 44° between the two intersecting airways [33]. A crossing point envelope depicts waypoint with up to 4 min flight time from an aircraft’s position. This crossing point envelope is chosen as it twice that of the Short

Monitoring Performance Measures for Radar Air Traffic Controllers

731

Term Conflict Alert (STCA), which is 2 min. Furthermore, this is also consistent with the study of Eyferth et al., as they found that ATCOs start to detect a conflict 4 to 5 min before it occurs [34]. This boundary would therefore be the region in which participants seek to detect and monitor the aircraft in crossing. As a result, all sessions were subjected to a total of 450 cases of wide and 450 cases of medium angle crossing point instances. Monitoring performance of participants were then investigated at a macroscopic level (full one-hour duration), and for a 4-minute period before a crossing point for both wide and medium angle crossing. It must be said that other studies involved only a single crossing of varying angles. 3.4

Acquisition of Model Measurements

Fixation count [30, 35], a common eye metric, on an aircraft was used as the mode of measurement to determine the monitoring performance of the test participants. To detect any fixation made by the test participant from the captured gaze data, a velocity threshold of 105°/s (based on the velocity threshold fixation identification (I-VT) algorithm [36]), which corresponds to the eye tracker’s accuracy of 3.5° in this experimental setup was used. Any captured gaze data that falls below this velocity threshold is classified as a fixation sample. A fixation is then computed when consecutive fixation sample adds up to more than 250 ms (ms) [37]. Its location is then computed by taking the average of Pixel X and Pixel Y values from the fixation sample. Taking note of the radar refresh frame rate at 9.8 s, a lower boundary of 250 ms for a fixation should generate enough data for further analysis. This is equivalent to a maximum of 39 fixations within a frame of 9.8 s. Typically, ATCOs make no more than 20 fixations within a frame. To map a fixation to a dynamic moving flight object (track and label), the pixel values of the fixation location are traced and normalized back to the original state, using macros that have been written and developed in the real-time post processing server [30]. Once the fixation data and radar data were aligned to the same reference frame, fixations can then be mapped onto the aircraft to determine if the test participant was looking at the aircraft. For the mapping of a fixation to an aircraft, suitable size boundaries around the aircraft’s track and label on the radar screen were to be set. The standard values of an aircraft’s track are a 3 mm diameter circle and an aircraft’s label is a 2 cm by 2 cm square on the radar screen. These values correspond to a circle of a 12 pixel diameter and 81 by 81 pixel square. However, as the precision of the eye tracker has a small tolerance resolution of 1.1°, this is equivalent to 58 pixels. Owing to this resolution, the new boundaries of the aircraft’s track and label for the mapping of a fixation would be a 70 pixel diameter and 230 pixel diameter circle respectively. Figure 2 shows an example of a radar frame image with the various fixations on the radar screen at a particular radar frame. The fixations and fixation sequence are represented by the square bits and lines joining them, while the number indicates the order of the fixations. The circles and squares in Fig. 2 represent the track and label of an aircraft, as seen by the ATCO on the radar screen along with their corresponding callsign. Other radar image information like the airways and its flight direction were also illustrated. In this radar frame, nine different aircraft were shown with a total of

732

H. J. Wee et al.

eight fixations were registered. Of these 8 fixations, 2 fixations were mapped onto 2 aircraft’s label, while another 2 were mapped onto 1 aircraft’s track, based on these new representation and pixel boundary values of the aircraft’s track and label. In total, 4 fixations were mapped onto flight objects (track or label).

Fig. 2. Example radar frame image with fixation data and radar image information

4 Results and Discussion In this section, the monitoring metrics of the 3 expertise levels of three situations were analyzed. These situations are macroscopic one-hour duration, over a 4-minute period before a crossing point for wide and medium angle crossings respectively. A one-way ANOVA test on a set of monitoring parameters is then performed in each of these situations to determine if there are any significant differences between the 3 expertise levels. For parameters with significant differences, a further analysis is performed to identify these differences. 4.1

Macroscopic One-Hour Duration

For macroscopic one-hour duration analysis, 6 monitoring parameters were investigated. They are the three fixation count ratios for aircraft track only, aircraft labels only and flight object (track and label) as well as the three fixation duration ratios for aircraft track only, aircraft labels only and flight object (track and label). The fixation count and duration ratios can be derived from the formulae stated in (1) and (2) for the entire

Monitoring Performance Measures for Radar Air Traffic Controllers

733

one-hour duration, where the object of interest refers to the track only, label only or flight object (track and label). Fixation Count Ratio ¼ Fixation Duration Ratio ¼

Total Fixation Countobjectofinterest Total Fixation Count

ð1Þ

Total Fixation Durationobjectofinterest Total Fixation Duration

ð2Þ

Table 1 shows the results of a one-way ANOVA test performed on all 30 experimental sessions of the entire radar screen. Significant differences were observed for 2 of the 6 monitoring parameters, fixation count and fixation duration ratio on flight objects and a display of their distribution for 3 expertise levels is depicted in Fig. 3. This meant that the importance of the flight objects across the 3 expertise levels are different, which is detected in a macroscopic one-hour duration consisting of various air traffic situations. Table 1. One-way ANOVA test result for macroscopic one-hour duration ANOVA test result Fixation count ratio

Aircraft track F2;28 ¼ 2:383 p ¼ 0:111 Fixation duration ratio F2;28 ¼ 2:136 p ¼ 0:137 * indicates significance at 95% level

Aircraft label F2;28 ¼ 0:618 p ¼ 0:546 F2;28 ¼ 0:564 p ¼ 0:575

Flight object F2;28 ¼ 10:452 p ¼ 0:000 * F2;28 ¼ 4:109 p ¼ 0:027 *

By comparing the mean values of monitoring metrics with significant differences, it is found that experts, intermediates and novices spent 88%, 78% and 65% of their fixation count and 83%, 79% and 63% of their fixation duration on flight objects respectively. Such indicates that macroscopically, for the full one-hour duration and the entire radar screen, experts tend to be more focused in their monitoring behavior. Though more experience and familiar with the air traffic situation, the finding suggests that experts tend to focus and update themselves more in relation to the vital information of the aircraft which are displayed on the flight objects, more regularly. These percentages could also serve as a guide relating to the expertise level of ATCOs or detect any anomaly in terms of an ATCO monitoring behavior. 4.2

4-Minute Period Before Wide Angle Crossing

As noted from the above, participants tend to look most at the aircraft’s track and label during this 4-minute period for wide angle crossing instances. A set of the fixation count and fixation duration ratios for track and labels were calculated, using (1) and (2) for a 4-minute duration, based on these 450 wide angle crossing instances. A oneway ANOVA test is performed, and its results are seen in Table 2.

734

H. J. Wee et al.

Fig. 3. Distribution of monitoring parameters on flight objects Table 2. One-way ANOVA for 4-minute period before wide angle crossing ANOVA test result Fixation count ratio

Aircraft track F2;448 ¼ 1:030 p ¼ 0:358 Fixation duration ratio F2;448 ¼ 1:527 p ¼ 0:218 * indicates significance at 95% level

Aircraft label F2;448 ¼ 3:837 p ¼ 0:022 * F2;448 ¼ 3:332 p ¼ 0:037 *

Significant differences were observed for the fixation count and fixation duration ratio on an aircraft label. Such indicates that participants of different expertise level looked at the aircraft label differently in a 4-minute period before wide angle crossing. Therefore, fixation count and fixation duration on an aircraft’s label for wide angle crossing instances were analyzed further to distinguish the three expertise levels. The distribution of the aircraft label with different levels of fixation count and fixation duration was examined and depicted in Fig. 4. For more than 10 fixations within a frame (9.8 s), it is observed that expert, intermediate and novice participants fixate 44%, 28% and 16% on the aircraft label respectively. Furthermore, for fixations of more than 2 s per aircraft label, the percentages for experts, intermediates and novices are 90%, 72% and 58%. This meant that expert participants tend to have more fixations over a longer duration while monitoring aircraft labels on wide angle crossing. This suggests that experts are looking for the aircraft’s flight information on the label more regularly and processing them for a longer duration, which is crucial in guiding aircraft across crossing. This is consistent with the earlier macroscopic findings.

Monitoring Performance Measures for Radar Air Traffic Controllers

735

Fig. 4. Distribution of fixation count and fixation duration on aircraft label 4 min before wide angle crossing

Table 3. One-way ANOVA for 4-minute period before medium angle crossing ANOVA test result Proportion of fixation count

Aircraft track F2;441 ¼ 1:810 p ¼ 0:165 Proportion of fixation duration F2;441 ¼ 1:056 p ¼ 0:349 * indicates significance at 95% level

4.3

Aircraft label F2;441 ¼ 1:691 p ¼ 0:185 F2;441 ¼ 2:493 p ¼ 0:084

4-Minute Period Before Medium Angle Crossing

Similarly, the fixation count and fixation duration ratios for track and labels were calculated for 450 of medium angle crossing instances, using (1) and (2) for a 4-minute duration. A one-way ANOVA test is also performed, and its results are seen in Table 3. No significant differences were observed for any of the monitoring parameters, indicating that medium angle crossings are not useful in determining the expertise levels of ATCOs.

736

H. J. Wee et al.

The results meant that these monitoring metrics exhibited by participants for medium angle crossing points are not suitable for distinguishing their expertise levels. This could likely be due to the shorter distance spatially between aircraft flying on medium angle crossing points, as opposed to aircraft flying on wide angle crossing points, making it easier for participants of all expertise level to identify them visually [22].

5 Conclusion In this study, distinct differences in monitoring behavior between expert, intermediate and novice participants were observed in macroscopic one-hour duration and wide angle crossing. This involves mapping ATCOs’ fixations to the dynamic moving flight objects (track and label) on the radar screen in real-time, using a remote eye tracker. Four new monitoring metrics relating to the macroscopic flight duration and wide angle have been established. These metrics include means of fixation count and duration ratios on flight objects for a macroscopic one-hour duration, proportion of aircraft label with more than 10 fixation count and proportion of aircraft label with more than 2 s fixation duration for a 4-minute duration before wide angle crossing. Based on the mean value of the fixation count and duration ratios on flight objects for the entire radar screen in a macroscopic one-hour duration, experts fixate 10% and 23% more on flight objects proportionally as compared to intermediates and novices respectively, while experts fixate 4% longer than intermediates and 20% longer than novice proportionally on flight objects. For a 4-minute duration before wide angle crossing instances, for aircraft label with more than 10 fixations, experts have 16% and 28% more than intermediates and novices respectively. In addition, for aircraft label with more than 2 s fixation duration, experts have 18% and 32% more than intermediates and novices respectively. Acknowledgments. This research is funded by Thales Solutions Asia Pte Ltd, under the Economic Development Board, Industrial Postgraduate Programme, with Nanyang Technological University, Singapore. The authors would like to acknowledge and thank the staff at Thales LAS France SAS, Thales Solutions Singapore, Air Traffic Management Research Institute of Nanyang Technological University, Singapore and the participants in this study for their contributions and support towards this work.

References 1. Wickens, C.D., et al.: The Future of Air Traffic Control: Human Operators and Automation. National Academies Press, Washington, D.C. (1998) 2. Enea, G., Porretta, M.: A comparison of 4D-trajectory operations envisioned for Nextgen and SESAR, some preliminary findings. In: 28th Congress of the International Council of the Aeronautical Sciences (2012) 3. Bowen, D.: The SESAR concept and i4D. In: Educational Workshop in ATM Global (2014) 4. Mutuel, L.H., Neri, P., Paricaud, E.: Initial 4D trajectory management concept evaluation. In: Tenth USA/Europe Air Traffic Management Research and Development Seminar (ATM2013) Airport (2013)

Monitoring Performance Measures for Radar Air Traffic Controllers

737

5. Dittmann, A., Kallus, K., Van Damme, D.: Integrated Task and Job Analysis of Air Traffic Controllers-Phase 3-Baseline Reference of Air Traffic Controller Tasks and Cognitive Processes in the ECAC Area (2000) 6. Wickens, C.D., Mavor, A.S., McGee, J.P.: Flight to the Future: Human Factors in Air Traffic Control. National Academies Press, Washington, D.C. (1997) 7. Parasuraman, R., Riley, V.: Humans and automation: use, misuse, disuse abuse. Hum. Factors 39(2), 230–253 (1997) 8. Metzger, U., Parasuraman, R.: The role of the air traffic controller in future air traffic management: an empirical study of active control versus passive monitoring. Hum. Factors: J. Hum. Factors Ergon. Soc. 43(4), 519–528 (2001) 9. Taukari, A., Pant, R.S., Garg, A.K.: A comparative study of cognitive strategies used to resolve air traffic conflict between air traffic controllers and general population using eye tracking machine. J. Psychosoc. Res. 5(2), 125 (2010) 10. Van Orden, K.F., et al.: Eye activity correlates of workload during a visuospatial memory task. Hum. Factors 43(1), 111–121 (2001) 11. Underwood, G., et al.: Visual attention while driving: sequences of eye fixations made by experienced and novice drivers. Ergonomics 46(6), 629–646 (2003) 12. Schwehr, J., Willert, V.: Driver’s gaze prediction in dynamic automotive scenes. In: 2017 IEEE 20th International Conference on Intelligent Transportation Systems (ITSC) (2017) 13. Regis, N., et al.: Ocular metrics for detecting attentional tunnelling. In: Human Factors and Ergonomics Society Europe Chapter Annual Meeting Toulouse, France (2012) 14. Biella, M., et al.: How eye tracking data can enhance human performance in tomorrow’s cockpit. In: RAeS Flight Simulation Conference Royal Aeronautical Society: Hamilton Place, London (2017) 15. Kasprowski, P., Harezlak, K., Kasprowska, S.: Development of diagnostic performance & visual processing in different types of radiological expertise. In: Proceedings of the 2018 ACM Symposium on Eye Tracking Research & Applications, pp. 1–6. ACM, Warsaw (2018) 16. Lundberg, J., et al.: The use of conflict detection tools in air traffic management: an unobtrusive eye tracking field experiment during controller competence assurance. In: Proceedings of the International Conference on Human-Computer Interaction in Aerospace, pp. 1–8. ACM, Santa Clara (2014) 17. Jacob, R., Karn, K.S.: Eye tracking in human-computer interaction and usability research: ready to deliver the promises. Mind 2(3), 4 (2003) 18. Damacharla, P., Javaid, A.Y., Devabhaktuni, V.K.: Human error prediction using eye tracking to improvise team cohesion in human-machine teams. In: Advances in Human Error, Reliability, Resilience, and Performance. Springer International Publishing, Cham (2019) 19. Imants, P., Greef, T.D.: Eye metrics for task-dependent automation. In: Proceedings of the 2014 European Conference on Cognitive Ergonomics, pp. 1–4. ACM, Vienna (2014) 20. Wang, Y., et al.: Statistical analysis of air traffic controllers’ eye movements. In: The 11th USA/Europe ATM R&D Seminar (2015) 21. Cong, W., et al.: On the correlations between air traffic and controller’s eye movements. In: 7th International Conference on Research in Air Transportation. Drexel University, Philadelphia (2016) 22. Marchitto, M., et al.: Air traffic control: ocular metrics reflect cognitive complexity. Int. J. Ind. Ergon. 54, 120–130 (2016) 23. Kang, Z., Landry, S.J.: An eye movement analysis algorithm for a multielement target tracking task: maximum transition-based agglomerative hierarchical clustering. IEEE Trans. Hum.-Mach. Syst. 45(1), 13–24 (2015)

738

H. J. Wee et al.

24. Kang, Z., Bass, E.J., Lee, D.W.: Air traffic controllers’ visual scanning, aircraft selection, and comparison strategies in support of conflict detection. Proc. Hum. Factors Ergon. Soc. Ann. Meeting 58(1), 77–81 (2014) 25. ICAO, Annex 1, Personnel Licensing, in International Standards and Recommended Practices, Montreal, Canada (2011) 26. Tobii, Tobii X2-30 Eye Tracker Accuracy and Precision Test Report, T.T. AB, Editor (2013) 27. Clemotte, A., et al.: Accuracy and precision of the Tobii X2-30 eye-tracking under non ideal conditions. In: Proceedings of the 2nd International Congress on Neurotechnology, Electronics and Informatics - Volume 1: NEUROTECHNIX, p. 2 (2014) 28. Dalrymple, K.A., et al.: An examination of recording accuracy and precision from eye tracking data from toddlerhood to adulthood. Front. Psychol. 9, 803 (2018) 29. Holmqvist, K., Nystrom, M., Mulvey, F.: Eye tracker data quality: what it is and how to measure it. In: Proceedings of the Symposium on Eye Tracking Research and Applications, pp. 45–52. ACM, Santa Barbara (2012) 30. Wee, H.J., Lye, S.W., Pinheiro, J.-P.: Real time eye tracking interface for visual monitoring of radar controllers. In: AIAA Modeling and Simulation Technologies Conference. American Institute of Aeronautics and Astronautics (2017) 31. Wee, H.J., et al.: Real time bio signal interface for visual monitoring of radar controllers. In: Transdisciplinary Engineering: A Paradigm Shift: Proceedings of the 24th ISPE Inc. International Conference on Transdisciplinary Engineering, 10–14 July 2017. IOS Press (2017) 32. SkyVector. SkyVector Aeronautical Charts 2006 2018. https://skyvector.com/ 33. Eurocontrol, A Consistent Vertical Collision Risk Model for Crossing and Parallel Tracks. Eurocontrol (1997) 34. Eyferth, K., Niessen, C., Spaeth, O.: A model of air traffic controllers’ conflict detection and conflict resolution. Aerospace Sci. Technol. 7(6), 409–416 (2003) 35. Hasse, C., Bruder, C.: Eye-tracking measurements and their link to a normative model of monitoring behaviour. Ergonomics 58(3), 355–367 (2015) 36. Salvucci, D.D., Goldberg, J.H.: Identifying fixations and saccades in eye-tracking protocols. In: Proceedings of the 2000 Symposium on Eye Tracking Research & Applications, pp. 71– 78. ACM, Palm Beach Gardens (2000) 37. Sereno, S.C., Rayner, K.: Measuring word recognition in reading: eye movements and eventrelated potentials. Trends Cogn. Sci. 7(11), 489–493 (2003)

Flight Eye Tracking Assistant (FETA): Proof of Concept Christophe Lounis(&), Vsevolod Peysakhovich, and Mickaël Causse ISAE-SUPAERO, Université de Toulouse, Toulouse, France [email protected]

Abstract. Accident investigations show that piloting errors (e.g., incorrect trajectory) often result from an inadequate monitoring of the cockpit instruments. Recent improvements of the eye tracking technology now allow a reliable and rather accurate recording of eye movements in ecological environments. The present study investigates how the integration of eye tracking in the cockpit could help pilots performing an efficient surveillance of their instruments. We developed FETA, an embedded system that evaluates online the visual monitoring of the cockpit. The system compares the current visual scan of the pilot with a database of “standard” visual circuits established thanks to eye-tracking recordings from 16 airlines pilots. If the current visual scan deviates too much from the database, e.g., the speed is not fixated during a too long period, FETA emits a vocal alarm to reorient attention. This paper presents the development of FETA and its preliminary evaluation with 5 airlines pilots. During an approach-landing phase in flight simulator; we assessed the impact of FETA on situation awareness, cognitive resources, flight performance, and visual scans. Results showed that FETA system efficiently redirected attention toward critical flight instruments. However, improvements must be performed to satisfy with operational requirements. For example, it seems important to take also into-account flight parameters in order to limit unnecessary alerts. Keywords: Eye-tracking
Aviation
Human factors
Human computer interaction
Neuroergonomics
Flying assistant Assistive technology




1 Introduction Over the past 50 years, continuous technical and technological improvements in commercial aviation made it the safest modes of transportation [1]. Progress in cockpit systems and aircraft design [2], in pilot training, in flight crew and air-traffic control procedures, are still essential to maintain a low accident rate despite an ever-increasing traffic [3]. Nevertheless, accidents always occur and a large part of them involve human error (approximatively 60 to 80% of accidents), as shown Fig. 1. One critical solution provided by the industry to reduce crew’s workload [6] and to deal with human errors [4, 5] has been the introduction of automation. However, automation also shifted the role of the crew from controllers to supervisors [7]. Unfortunately, automation is not always fully understood and nor correctly monitored [8]. It can induce complacency, © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 739–751, 2020. https://doi.org/10.1007/978-3-030-20503-4_66

740

C. Lounis et al.

overconfidence, and airline pilots sometimes rely too much on it [9]. Therefore, increasing the pilot’s ability to stay in the loop, in particular by promoting an appropriate monitoring of the cockpit instruments, is a main current safety challenge [10]. This is particularly true during the approach and landing phases, critical periods of a flight, in which safety margin and tolerance on flight parameters deviations are very low [11].

Fig. 1. Aeronautical accidents from 1969 to 2019 with human factors, technical failure and others causes as contributory factors. These data were retrieved from Bureau of Aircraft Accidents Archives (www.baaa-acro.com).

A report of the active pilot monitoring working group published by the Flight Safety Foundation [12] investigated 188 cases involving monitoring issues leading to accidents. They find out that 66% of monitoring errors occurred during dynamic phases of flights (e.g., climb, descent, approach, and landing). They also identified the failure to cross-check instruments as a major cause of those monitoring errors that resulted in excessive deviations of the flight parameters (e.g., altitude, trajectory, or speed deviation). Recently, the Federal Aviation Administration (FAA) published a final training rule that requires enhanced pilot monitoring training to be included into existing air careers training programs [13]. Furthermore, the French accident investigation bureau (Bureau Enquêtes et Analyses, BEA) suggested to analyze pilots’ monitoring with eye tracking to improve piloting procedures [14]. In this sense, a recent paper described the different ways in which eye tracking could be used to assist commercial pilots during the flight [15].

Flight Eye Tracking Assistant (FETA): Proof of Concept

741

Based on the latter studies, we propose the FETA system. It compares the current visual scan of a pilot with a database of “standard” visual circuits. If the current visual scan deviates too much from the database (e.g., the speed is not fixated during a too long period), FETA emits a vocal alarm (e.g., “check speed”). The current paper describes the development and the evaluation of the FETA (Flight Eye Tracking Assistant) system [16]. In particular, we evaluated the impact of FETA on situation awareness [17], subjective workload, flight performance, and visual scans.

2 FETA System Development The main purpose of the FETA system is to warn the pilot when he looks not sufficiently at an instrument. In order to calibrate the “not sufficiently”, the threshold beyond which visual scans become “abnormal”, we built a database of standard visual circuits in the cockpit with a sample of 16 airline pilots. They performed approachlanding phases in flight simulator while their eye movements were recorded. We also ensured that their flight performance remained in the standard safety thresholds. 2.1

Participants

Sixteen male professional airline pilots (ATPL: Airline Transport Pilot License or CPL: Commercial Pilot License) volunteered to participate in this study. Their mean age was 34 years old (range: 23–59). Their total flight experience ranged from 1,600 to 13,000 h (M = 4,321.73 h, SD = 2,911.41 h). They were not paid for their participation. They had normal or corrected-to-normal vision. The experiment was approved by the Research Ethics Committee (CER, n°2019-131). 2.2

Procedure

Each participant signed a consent form and provided demographic information, their flight qualifications (type of aircraft), and their total flight experience hours. Pilots were briefed on the study and receive instruction about the flight scenario and the goal of this experiment. They filled a fatigue questionnaire. Next, pilots were installed in the flight simulator and were submitted to the eye-tracking calibration procedure. Participants took the captain position and performed a training consisting in two approach-landings scenarios in order to familiarize themselves with the flight simulator. Then, they performed the two experimental approach-landings scenarios. 2.3

Flight Simulator

The study was conducted in the PEGASE (Platform for Experiments on Generic Aircraft Simulation Environment) flight simulator of the ISAE-SUPAERO (Toulouse, France), illustrated in Fig. 2. It simulates an Airbus A320 with a glass cockpit. The simulator includes pilots’ seats, sidestick controllers, throttles, trim wheels, and rudder pedals.

742

C. Lounis et al.

Fig. 2. The PEGASE flight simulator used during FETA development and assessment.

2.4

Eye-Tracking Measurements

Eye tracking data was collected with a Smart eye System embedded in the cockpit. The Smart eye System consists of 5 deported cameras (0°–2° of accuracy), with a sampling frequency of 60 Hz. Furthermore, the cockpit has been divided into several Areas of Interests (AOIs), as presented in Fig. 3. They correspond to the main flight displays. These AOIs are used by the FETA system to evaluate online current visual scans. We also used these AOIs during the human factor evaluation to examinate the impact of FETA on visual scans. The threshold for detecting a fixation on an AOI was set at 100 ms [18].

Fig. 3. Cockpit display with AOIs and Sub-AOIS: (1) Primary Flight Display (PFD), (2) Navigation Display (ND), (3) Electronic Centralized Aircraft Monitoring (ECAM), (4) Out of Window (OTW), (5) Flight Control Unit (FCU), (6) Flight Mode Annunciator (PFD.FMA), (7) Speed Tape (PFD.SPD), (8) Attitude Indicator (PFD.ATT), (9) Vertical Speed Tape (PFD.VS), (10) Heading Tape (PFD.HDG), (11) VOR tag reading area in ND (ND-zone).

Flight Eye Tracking Assistant (FETA): Proof of Concept

2.5

743

Experimental Conditions

The 16 pilots performed two times the same flight scenario. The flight scenario consisted of a manual approach-landing task to Toulouse-Blagnac Airport Runway LFBO 14R. Flight began at coordinates 1.2159 longitude and 43.7626 Latitude. During the scenario, the pilot had to comply with some specific instructions. In particular: maintain a vertical speed between +500 ft/min and −800 ft/min, a speed of 130 knots, and a heading of 143° (Fig. 4).

Fig. 4. The landing scenario with the flight parameter values that pilots had to maintain.

2.6

Flight Parameters

Firstly, we checked flight performance of the pilots, assuming that correct flight performance is likely correlated to an efficient cockpit monitoring. Figure 5 shows the mean flight parameters deviation for vertical speed, speed and heading during the landing task. Flight performance of each pilot was adequate. Average vertical speed was in the correct range, and average speed and average heading were very close to the target values.

Fig. 5. Violin plot of flight parameters deviations during the landing task. The red lines correspond to target values, as given by the experimenter before the flight scenarios. N = 16.

744

2.7

C. Lounis et al.

Visual Behavior Database and Notification Threshold

The Visual Behavior Database (VBD) has been established with the eye recordings made on the 16 pilots that performed the two approach-landing scenarios. Mean nondwell times were calculated for each AOI. While dwell times represent the time during which an individual gaze inside an AOI [19], non-dwell times correspond to the period of time during which an individual does not look at an AOI, see Fig. 6.

Fig. 6. Violin plot of non-dwell times during the landing task for the main AOIs. N = 16.

We used the “non-dwells times” of the 16 expert pilots as the metric for the FETA notification threshold. More precisely, the thresholds consisted of the averages of the non-dwell time for each AOI plus a standard deviation, as presented in (1). uthreshold ¼ lNDT þ rNDT :

ð1Þ

This metric indicates the maximum non-dwell time tolerance for each AOI (i.e., beyond which an insufficient monitoring is diagnosed). 2.8

FETA Interface

Besides the Visual Behavior Database, the eye tracking system, and the vocal alarms, FETA also has an application permitting to visualize the activity from outside the cockpit. Coded in C#, the FETA interface has many features shown in Fig. 7.

Flight Eye Tracking Assistant (FETA): Proof of Concept

745

Fig. 7. FETA interface with its 7 different features.

The features of FETA are: 1. AOI Monitoring Panel (on the left of Fig. 7) It shows the state of each AOI. The color turns from green to blue when the AOI is not monitored enough according to the VBD. 2. Show timer (center of Fig. 7) User can tick the tick boxes of any of the AOIs to see the timer of each AOI. This timer shows the elapsed duration since the last monitoring (in seconds). 3. AOI Heat Map Panel (on the right of Fig. 7) This heat map panel indicates the proportion of fixation times on the AOI since the beginning of the flight. 4. Timer (center of Fig. 7) This feature shows the elapsed time until the beginning of the simulation in seconds. 5. AOI Text Alert and Current Area of Interest Annunciator (at the bottom left of Fig. 7) The AOI Text Alert will show the name of the AOI that needs to be monitored. If more than one AOI needs to be monitored, this alert will only show the name of the AOI with the highest priority. The Current Area of Interest Annunciator shows the currently monitored AOI. 6. Flight Parameter Indicators (at the bottom left of Fig. 7) This feature shows the several flight parameters that affect the dynamic of the database. 7. Start/Stop Tracking Button and Show/Hide Heat Map Button (centred at the bottom of Fig. 7) The Start/Stop Tracking Button starts or stops FETA, while the Show/Hide Heat Map Button shows or hides the AOI heat map. 8. Audio Alarm (cannot be shown) FETA will emit an audio alarm that corresponds to the AOI Text Alert (e.g. “check speed”).

746

C. Lounis et al.

3 FETA System Assessment The second part of this paper focuses on the evaluation of the FETA system. In particular, its impact on mental workload, situational awareness, flight performances, and cockpit monitoring. As a preliminary assessment, five pilots were submitted to three different scenarios varying in terms of monitoring difficulty. 3.1

Participants

Five male professional pilots (ATPL, CPL) volunteered to participate in this study. They had normal or corrected-to-normal vision. Their mean age was 39 years old (range: 33–50). Their total flight experience ranged from 2,500 to 8,500 h (M = 3,176 h, SD = 2,645 h). Pilots were not paid for their participation. The experiment was approved by the Research Ethics Committee (CER, n°2019-131). 3.2

Procedure

Procedure was essentially the same than during the FETA calibration, except that the new 5 pilots performed four additional landings. During this evaluation, FETA auditory notifications (in case of abnormal monitoring) were restricted to three instruments: speed, vertical speed, and heading. These instruments were chosen because they corresponded to the flight parameter values that pilots had to maintain. Possible auditory alarms emitted by FETA were: “check speed”; “check vertical speed”, “check heading”. 3.3

Apparatus

This experiment also took place in the PEGASE flight simulator, using the same eye tracking system. 3.4

Experimental Conditions

Pilots performed two times three different randomized landing scenarios. The first scenario (Scenario 1) was identical to the one performed by the pilots for the building of the VBD. In the second and the third scenarios, we increased monitoring difficulty. During these two scenarios, pilots were asked to read aloud the distance between the aircraft and a specific radio beacon (information displayed in the ND-zone) either every 0.5 Nm (scenario 2) or every 0.2 Nm (scenario 3). The pilots had to comply with the same speed, vertical speed, and heading constraints than during the VBD building. At the end of the simulation, pilots filled out 2 subjective questionnaires: situational awareness measures using SART [20] and workload Instantaneous Self-Assessment [21], which is a subjective scale ranging from 1 to 5. The latter allows assessing overall workload. After the flight scenarios, open interviews were conducted to garner the various opinions of the pilots according to the system.

Flight Eye Tracking Assistant (FETA): Proof of Concept

3.5

747

Human Factors Assessment

Due to the low number of participants, we only present descriptive statistics for subjective assessments, and flight performance. However, eye tracking data allows us to use inferential statistics regarding the comparison with and without FETA. In particular by taking into account the difficulty of scenarios (1, 2, 3) as a covariate. Subjective Results Figure 8 shows the SART results. A higher SART score indicates a better situational awareness. On average, FETA seemed to disturb the situational awareness when flying context was easy (scenario 1 and 2), but it tended to be the opposite when flying context was more complex (scenario 3). As presented in Fig. 8 (at right), ISA workload indicator did not show marked difference with or without the FETA system. However, in an easy flying context (scenario 1), the FETA system seems to induce more workload and this trend is reversed when flying context is more difficult (scenario 2 and 3).

Fig. 8. Left: SART results (higher the values, better the situational awareness); Right: ISA results (lower is the value and lower is the subjective workload). All three scenarios with and without the FETA system are showed. N = 5.

Flight Performance Results The Fig. 9 shows flight parameters deviations. During the easy scenario (scenario 1), pilots had higher speed deviations with FETA than without. Concerning the heading in the difficult condition (scenario 3), pilots had on average lower heading deviations with the FETA system than without. Eye Tracking Results Figure 10 shows the percentage dwell times on each AOI for all scenarios with and without FETA system. The Wilcoxon-Mann-Whitney nonparametric test shows a significant effect (p < 0.05) of FETA vs. without FETA condition on the AOIs according to speed, vertical speed, heading, flight mode annunciator and out the window.

748

C. Lounis et al.

Fig. 9. Root Mean Square (RMS) of the flight parameters for each scenarios with and without FETA (the higher the value, the lower the performance). N = 5.

Fig. 10. Bar plot of percentage dwell times on each AOI. All three scenarios with and without the FETA system are showed. N = 5. (*p < 0.05, Wilcoxon Mann-Whitney test).

4 Discussion and Conclusions The purpose of this study was on the one hand to present the concept and the development of a flight eye-tracking assistant (FETA) calibrated thanks to eye-movement recordings from 16 airline pilots. On the other hand, we also proposed a user-centered evaluation (e.g., situation awareness, mental workload) of the first version of this

Flight Eye Tracking Assistant (FETA): Proof of Concept

749

assistant together with an assessment of its impact on the cockpit monitoring. This evaluation was performed with 5 other airline pilots. Overall, this first version of FETA demonstrated mixed results. First, results showed that there was no clear improvement of the flight parameters that had to be maintain during the landing (speed, vertical speed, and heading). There was an increased speed deviation during the easier landing and on the contrary an improvement of heading accuracy during the most difficult landing scenario. Consistently, subjective results tend to show that FETA was not detrimental only when flight scenario was difficult. In particular, situation awareness seemed slightly improved by FETA in the scenario 3. Eye tracking results were more favorable to FETA, with an increase of the time spent on some instruments subjected to the FETA audio notification in case of insufficient visual consultation. In presence of FETA, pilots checked more often the speed, the vertical speed, and the heading. This additional time gazing these instruments impacted the time spend on the window. Most likely, FETA was efficient to redirect attention toward the critical flight instruments thanks to the vocal alarm triggered when visual circuit deviated too much from the database. Despite this positive result, our experiment shed to light several issues that should be addressed in the future. Open interviews with the pilots allowed revealing some areas of improvements. For example, the use of the auditory modality is not necessary the best one. This channel is already used by the synthetic voice in the cockpit, and also during the exchanges between pilots and air traffic control. To overcome this problem, other notifications methods could be explored, such as visual and/or haptic modalities. Another important improvement would be to both integrate flight parameters values and eye movements in FETA. Indeed, it would be more appropriate to trigger notifications when both visual scans and flight parameters deviates too much from standards. For example, when speed decline too fast etc. This would help avoid triggering spurious notification (useless auditory notification from FETA), which was one of the main problems raised by the pilots during the debriefing. More generally, the FETA system should consider other eye tracking metrics when considering the landing task; for example, it could analyze the visual patterns (transitions between AOIs, not only the fixation on each AOI) and correct them when they deviate from established standards, using artificial intelligence. Furthermore, FETA could take into account other flight phases, automatically identified considering the flight data (e.g., altitude, speed, flight mode…). Then, this would enable to adapt eye-tracking metrics to the given flight phases. For example, cockpit monitoring is much less intense during the cruise, but this phase is more prone to drowsiness or fatigue. FETA could integrate metrics based on the percentage of eye closure [22] or considering the frequency of eye blinks [23]. Future study should consider these improvements and assessing FETA during complex flight phases with a higher number of pilots. Acknowledgments. This work was supported by a chair grant from Dassault Aviation (“CASAC”, holder: Prof. Mickaël Causse)”. The Authors thank the PEGASE simulator technical team and all the pilots who participated in this study.

750

C. Lounis et al.

References 1. Boeing. Statistical Summary of Commercial Jet Airplane Accidents Worldwide Operations, 1959–2014; Aviation Safety. Boeing Commercial Airplanes: Seattle, WA, USA (2015). http://www.boeing.com/news/techissues/pdf/statsum.pdf. Accessed 2 Nov 2017 2. Sarter, N.B., Woods, D.D.: “From tool to agent”: the evolution of (Cockpit) automation and its impact on human-machine coordination. In: Proceedings of the Human Factors and Ergonomics Society Annual Meeting, vol. 39, no. 1, pp. 79–83. SAGE Publications, Los Angeles, October 1995 3. Button, K.: Wings Across Europe: Towards an Efficient European Air Transport System. Routledge, New York (2017) 4. Chialastri, A.: Automation in aviation. In: Automation. IntechOpen (2012) 5. Lee, J.D.: Human factors and ergonomics in automation design. In: Handbook of Human Factors and Ergonomics, 3rd edn, pp. 1570–1580. Wiley. Hoboken (2006) 6. Lee, J.D.: Human factors and ergonomics in automation design. In: Handbook of Human Factors and Ergonomics, pp. 1580–1596 (2006) 7. Mumaw, R.J., Sarter, N., Wickens, C.D.: Analysis of pilots’ monitoring and performance on an automated flight deck. In: 11th International Symposium on Aviation Psychology, Columbus, OH.8, May 2001 8. Parasuraman, R., Manzey, D.H.: Complacency and bias in human use of automation: an attentional integration. Hum. Factors 52(3), 381–410 (2010) 9. Wickens, C.D., Clegg, B.A., Vieane, A.Z., Sebok, A.L.: Complacency and automation bias in the use of imperfect automation. Hum. Factors 57(5), 728–739 (2015) 10. Berberian, B., Somon, B., Sahaï, A., Gouraud, J.: The out-of-the-loop Brain: a neuroergonomic approach of the human automation interaction. Ann. Rev. Control 44, 303–315 (2017) 11. Sumwalt, R., Cross, D., Lessard, D.: Examining how breakdowns in pilot monitoring of the aircraft flight path. Int. J. Aviat. Aeronaut. Aerospace 2(3), 8 (2015) 12. Flight Safety Foundation (FSF). A Practical Guide for Improving Flight Path Monitoring: Final Report of the Active Pilot Monitoring Working Group, November 2014. https:// flightsafety.org/toolkits-resources/flight-path-monitoring/ 13. Federal Aviation Administration (FAA). 14 CFR Part 121 Qualification, Service, and Use of Crewmembers and Aircraft Dispatchers; Final Rule; Federal Register: Washington, DC, USA, vol. 78, no. 218 (2013) 14. British European Airways (BEA). Study on Aeroplane State Awareness during Go-Around (2013). https://www.bea.aero/etudes/asaga/asaga.study.pdf. Accessed 2 Nov 2017 15. Peysakhovich, V., Lefrançois, O., Dehais, F., Causse, M.: The neuroergonomics of aircraft cockpits: the four stages of eye-tracking integration to enhance flight safety. Safety 4(1), 8 (2018) 16. Lounis, C., Peysakhovich, V., Causse, M.: Intelligent cockpit: eye tracking integration to enhance the pilot-aircraft interaction (2018) 17. Endsley, M.R.: Toward a theory of situation awareness in dynamic systems. Hum. Factors 37(1), 32–64 (1995) 18. Russi-Vigoya, M.N., Patterson, P.: Analysis of pilot eye behavior during glass cockpit simulations. Procedia Manuf. 3, 5028–5035 (2015) 19. Goldberg, J.H., Stimson, M.J., Lewenstein, M., Scott, N., Wichansky, A.M.: Eye tracking in web search tasks: design implications. In: Proceedings of the 2002 Symposium on Eye Tracking Research & Applications, pp. 51–58. ACM, March 2002

Flight Eye Tracking Assistant (FETA): Proof of Concept

751

20. Taylor, R.M.: Situational awareness rating technique (SART): the development of a tool for aircrew systems design. In: Situational Awareness, pp. 111–128. Routledge (2017) 21. Tattersall, A.J., Foord, P.S.: An experimental evaluation of instantaneous self-assessment as a measure of workload. Ergonomics 39(5), 740–748 (1996) 22. Sommer, D., Golz, M.: Evaluation of PERCLOS based current fatigue monitoring technologies. In: 2010 Annual International Conference of the IEEE Engineering in Medicine and Biology, pp. 4456–4459. IEEE, August 2010 23. Caffier, P.P., Erdmann, U., Ullsperger, P.: Experimental evaluation of eye-blink parameters as a drowsiness measure. Eur. J. Appl. Physiol. 89(3–4), 319–325 (2003)

How Does National Culture Help Pilots in Navigating in Different Environment? Xiaoyu O. Wu(&) Bowling Green State University, Bowling Green, OH, USA [email protected]

Abstract. The study surveyed 919 Chinese student pilots with 20 national culture questions. The researcher uses 5 questions to measure one national culture variables including Power Distance, Individualism, Masculinity, and Uncertainty Avoidance. The study performance a principle factor analysis (PCA) to the questionnaires and found that environment setting was essential to abstracted factors from national culture survey. The Kaiser-Meyer-Oklin measure of sampling adequacy (KMO) of this study was 0.85. A Bartlett’s Test of Sphericity (Bartlett’s test) was v2(n = 919) = 4975.077 and P value 0.001. The study founding echoed with Harari and Perkins who suggested that a culture system is different within different environments [1]. The result of PCA showed the survey could extract 4 latent factors, and the cumulative variance of the PCA indicated that the survey questions only explained 50% of the variances. The abstracted factors were reflecting large group environment, cockpit environment, general society expectation, and self-esteem. Keywords: National culture Transportation




Aviation




Pilot training




Human factor




1 Introduction Culture is closely tied to safety, especially in aviation communities and other highconsequence industries. Discussions and analyses from Eurocontrol and FAA, Ford et al., Griffin and Neal, Patankar and Sabin, and Wiegmann et al. [2–6] have demonstrated the role of culture and its relationship to the development of safety. In the aviation field, Hofstede’s national culture model [7] provides a method to check closely into culture issues. He identified several culture variables and suggested that national culture could be identified using these variables: Power Distance, Individualism/Collectivism, Long Term Orientation, Masculinity/Femininity, Indulgence, and Uncertainty avoidance [7–10]. Hofstede further developed an index system to describe these cultural variables that have been found helpful in defining the national culture. Merritt used correlation to demonstrate that her subset survey questions can predict Hofstede’s culture indices in a cockpit environment [11]. Studies of Helmreich et al., Johnston shows considerable differences occurs when pilots conduct their tasks in the cockpit based on their national culture variables [12, 13], and that such cultural variable differences have clear implications for safety. However, most studies of aviation culture © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 752–761, 2020. https://doi.org/10.1007/978-3-030-20503-4_67

How Does National Culture Help Pilots

753

are usually limited in single terms such as Power Distance, Individualism, Masculinity, and Uncertainty avoidance [14]. However, the usage of single term in aviation culture has its limitations, such as lacking the understanding of the whole picture. This study uses factor analysis to investigate how do student pilots use those culture variables to navigate themselves in aviation environment. Abstracted factors from the cockpit culture survey were reflecting participants attitude in different environment settings. The environment setting was from a microcosmic level that was the self-esteem to a macro level that was the general society; the cockpit environment and the large group environment were falling somewhere in between. It had no questions to measure, or describe, a culture by using culture variables like Power Distance, Individualism, Masculinity, and Uncertainty avoidance.

2 Hofstede’s National Culture In Hofstede’s framework, Power Distance, Individualism, Masculinity, and Uncertainty Avoidance can be measured by survey tools, and Merritt [11] created an instrument suitable for the aviation field to access these national culture variables. The Power Distance variable refers to the individual’s perception of hierarchy, or the gap between the supervisors and the individual personnel within an organization [8, 15]. In Hofstede’s culture variable index system, the score of Power Distance is referred to as the Power Distance Index (PDI). The higher a culture’s PDI is, the more likely that a hierarchical structure is more deeply embedded. The Individualism variable represents the autonomy and the growth of an individual in reaching certain goals [15]. In the culture variable index system, the score of Individualism is referred as Index of Individualism (IDV). The higher a culture’s (IDV) is, the stronger that culture’s emphasis is on individual behaviors and possible lack of cooperation. The Masculinity variable refers to preferences of achievement, heroism, and success in a society [8, 15]. Within the culture variable index system, the score of Masculinity is referred as Index of Masculinity (MAS). The higher a culture’s MAS is, the more its citizens tend to be assertive, and be in denial of errors or failures. The Uncertainty Avoidance variable represents the degree of accepting uncertainty and ambiguity [8]. Within the culture variable index system, the score of Uncertainty Avoidance is referred as Index of Uncertainty Avoidance (UAI). The higher a culture’s UAI is, the more its citizens follow the rules and avoids ambiguity.

3 Method This research used a self-reporting survey as the research instrument to access Chinese student pilots’ perceptions and attitudes of national culture. A principle factor analysis (PCA) was performed to test the survey instrument.

754

3.1

X. O. Wu

Survey

Self-reporting surveys served as the main data source in this study. This study used a composite survey which contained four original designed safety performance scenarios and a preexisting instrument of accessing Power Distance, Individualism, Masculinity, and Uncertainty avoidance. Participants’ demographic was collected. The detail of questionnaires can be found in the following Table 1. Table 1. Survey questions for cockpit culture Individualism Q1 I prefer to work alone Q2 I need sufficient time for personal and family life Q3 My personal problems can adversely affect my performance Q4 If a coworker gets a prize, I would feel proud -R Q5 It is important to me that I respect the decisions made by my group -R Power distance Q6 A First Officer should never assume command of the aircraft Q7 I preferred to work for a consultative leader -R Q8 My decision-making ability is as good in emergencies as in it is daily routine tasks Q9 A Captain should encourage crew member questions –R Q10 If I perceive a problem, I will speak up –R Masculinity Q11 Competition is the law of nature Q12 How often do you feel nervous or tense during a flight Q13 How often are you afraid to disagree with your instructors Q14 A Self-reporting system is useless. Nobody would use it. -R Q15 I get a personal sense of satisfaction from challenging tasks Uncertainty Avoidance Q16 Written procedures are required for all in-flight situations Q17 I like changing my work routine with new unfamiliar tasks –R Q18 In abnormal situations, I rely on superiors to tell me what to do Q19 It is important to find the truth, the correct answer, the one solution Q20 Organization’s rules should not be broken Note: The questions with “R” in the table means that the item is measured in reverse scoring Source: National culture survey [19].

3.2

Sample

The population of this study was Chinese pilot students who were receiving flight training either within China or overseas. They have been chosen to explain the cockpit culture that could be observed and to answer the research questions in this study. According to the Civil Aviation Administration of China (CAAC), in 2017 the population of Chinese trainee pilots was 5053 [16]. This research used a convenience

How Does National Culture Help Pilots

755

sample of the population. The sample size chosen was based on the reference table generated by Krejcie and Morgan [17]. According to the sample size reference table, 350 participants should be able to generalize the study to the population. For Chinese student pilot participants, the survey was conducted online and delivered via the most popular Chinese social media app “WeChat” which has the ability to host an online survey. The “WeChat” was used by schools as a platform of managing and delivery information to their students [18]. The credentials of the “WeChat” platform had been supported by research, such as the white paper of the development of Chinese Online Travel Agency by China National Tourism Administration [19] and 2016 Chinese cellphone market report by Yang [20]. The survey hosting platform has been shown to be reliable and trustworthy. Approximately 919 Chinese student pilots responded to the cockpit culture survey. After cleaning some missing data, 919 cases were analyzed for the study. The size sample was approximately 19% of all Chinese registered student pilots. 3.3

Principle Component Analysis

The PCA is a statistical analysis method to reduce dimensions of data by clustering data into multiple factors. A set of factors is created by using a correlation matrix, and those factors are independent from each other. Factors are also able to summarize the correlation of the data empirically [21]. PCA can also keep the test sensitive to the relative scaling of the original variables after reducing the dimensions of the data [22]. However, PCA assumes that linear orthogonal relation is present against factors [23]. In research, there are some advantages and disadvantages of using PCA in dimension reducing analysis. The most significant advantage of using PCA is that the researcher could obtain stable estimates even if there are violations of certain assumptions. Meanwhile, there are certain disadvantages of PCA. One significant disadvantage lies in the dependence on sample quality, with partial dependence of the findings based on the sample. In order to test if it is appropriate to conduct a PCA, three validity tests were used to ensure the data meets the assumptions of PCA. First, a Cronbach’s Alpha test will analyze the internal consistency of the survey. The Cronbach’s Alpha test would show how closely related a set of items are as a group, and it is considered to be a measure of scale reliability. The return of Cronbach’s Alpha test is a value ranging from 0.0–1.0. A return is greater than 0.7 will be acceptable in most social science research [24]. Second, the Kaiser-Meyer-Oklin (KMO) test assists the researcher in determining the adequacy of sample size. A return of KMO is a value between 0.0–1.0. A value above 0.7 is considered acceptable. Third, Bartlett’s Test of Sphericity (Bartlett’s test) will be conducted to test the correlation within the data. The Bartlett’s test will assist the researcher in determining if the data set meets the first assumption of PCA that the data is linearly correlated. Only upon the rejection of the null hypothesis of the Bartlett’s test, can the PCA be performed with validity. According to Bagozzi and Yi [25], critical individual item factor loading has to exceed 0.6. However, in the original Hoftstede’s original work, the factoring loading cut-off value was 0.35 [8]. On the other hand, Stevens suggests using a cut-off of 0.4, irrespective of sample size, for interpretative purposes [26]. Hair et al. [27] suggest that the cut-off value of a factor loading

756

X. O. Wu

can be vary due to the sample size. For sample sizes smaller than 100, the cut-off value should not be lower than 0.6; and for sample size around 200, a 0.4 cut-off factor loading value is sufficient [27]. This study uses a cut-off factor loading value of 0.4. When multiple samples exist with similar loadings, the similar loading might indicate that there is an association of the observed item with multiple contracted factors. This situation is usually referred as cross-loading. The cross-loading of items can be a result of many factors, such as the research design, or the rotation of the data [22]. O’Rourke and Hatcher suggest that the cross-loading situation will be more frequent to see in oblique rotation-based analysis, because the oblique rotation allows for the correlation of the factors [28]. In the ideal case, items are not preferred to be loaded over a number of factors. However, many ways are available to manage a crossloading situation. In many cross-loading situations, people choose to discard these items, unless there is a strong practical or theoretical rationale for their retention [23]. Sometime, researchers usually try different types of rotation to reduce the cross-loading [28]. Gill and Shergill [29] also suggest that in the cross-loading situation, variable can be included in the factor with the highest loading. In this study, cross-loading variables were included in the factor with the highest loading values.

4 Results 919 samples were collected in this study. Tabachnick and Fidell [21] argued that a sample size of 500 is a good sample size for conducting PCA. The KMO of this study was 0.85 which indicates the sampling is adequate [25]. A Bartlett’s test was conducted. The results for Bartlett’s test was v2(n = 919) = 4975.077, and p value 0.001. Results of Bartlett’s test suggested sufficiently large correlations between items to use PCA. Following the PCA calculation, four factors which explained approximately 50% of the variance were yielded. The eigenvalue setting was 1.0. Initial eigenvalues and cumulative percentage are provided in Table 2. Based on suggestion of Stevens, factors with loadings in excess of 0.4 were extracted and identified. The survey item was discarded, if the loading did not reach 0.4. Negative loading scores indicated a negative relationship between the abstracted factor and the survey item.

Table 2. PCA - initial eigenvalues & cumulative variables Factor initial Eigenvalue 1 4.79 2 2.85 3 1.31 4 1.21

Cumulative variables 23.97% 38.19% 44.38% 50.19%

Table 3 showed a different variable structure in the survey. The abstracted factors were still surrounded the basic national culture variable framework, after examining the

How Does National Culture Help Pilots

757

survey item and the inner logic relationship between abstracted factors. Abstracted factors reflected Power Distance, Individualism, Masculinity, and Uncertainty avoidance in one way or another. The PCA results showed that environment setting was essential to the outcomes. The abstracted factors were reflecting large group environment, cockpit environment, general society expectation, and self-esteem. Survey items loading of the first factor showed the attitude of working in a group environment regarding rules and working relationships with supervisors. This factor was a mix with Power Distance and Uncertainty avoidance. The second factor emerging from the PCA showed the perception of hierarchy in the cockpit. The second factor can be treated as a factor of Power Distance. The third factor showed the survey items on perception of competition. The third factor can be seen as a factor of Masculinity. And the fourth factor showed students’ attitudes of Individualism. The factor loadings are presented in Table 3.

5 Discussion The initial PCA revealed four factors from the survey questions. Results of PCA showed that survey questions were mixed loaded into each four abstracted factors (see Table 3). The four factors explained 50% of the variance, using an oblique rotation. The survey questions showed a loading of at least 0.4 on one of these four factors. The first factor of the initial PCA revealed four factors from the survey questions. Results of PCA showed that survey questions were mixed loaded into each four abstracted factors (see Table 3). The four factors explained 50% of the variance, using an oblique rotation. The survey questions showed a loading of at least 0.4 on one of these four factors. The first factor of the PCA was mixed with three questions of Power Distance (Q7, Q9, Q10), two questions of Uncertainty avoidance (Q19, Q20), and one question of Individualism (Q5). After examining the survey questions that were loaded in the first factor, this factor actually showed participants’ perception and attitude towards the hierarchy in a large group environment. This factor tended to show how would participants’ expectation to their supervisors in the company, as well as, they need rules to solve the conflicts between their supervisors. The second factor of the PCA was formed with three questions of Masculinity (Q12, Q13, Q14), one question of Individualism (Q2), one question of Power Distance (Q6), and one question of Uncertainty avoidance (Q18). This mixed question indicated that this factor revealed the personal opinions towards the flight and work. Questions that loaded in this factor had implied the conduction of a flight. Unlike the first factor, the second factor was at a much smaller working environment, the cockpit. The third factor of the PCA was generated from two questions of Masculinity (Q11, Q15), and one question of Uncertainty avoidance (Q17). The combination of these three questions indicated that this factor revealed participants’ attitude of competition and challenge. The third factor had a strong indication of Masculinity.

758

X. O. Wu Table 3. PCA - rotated pattern matrix Factor questions 1 2 3 4 Power distance & Uncertainty avoidance Q9 0.737 Q19 −0.708 Q5 0.699 Q10 0.682 Q7 0.585 Q20 −0.526 Power distance Q13 0.77 Q12 0.75 Q18 0.68 Q2 0.63 Q14 −0.62 Q6 0.51 Masculinity Q15 0.69 Q17 −0.58 Q11 0.51 Individualism Q3 0.79 Q1 0.54 Q16 0.49 Q8 0.41 Note: PCA with direct oblique rotation Component loadings 0.4 have been suppressed Cross- loadings suppressed

The last factor of the PCA was a combination with two questions of Individualism (Q1, Q3), one question of Power Distance (Q8), and one question of Uncertainty avoidance (Q16). The mixed questions showed a factor that ex- pressed participants’ attitude of themselves, self-esteem. This factor had a strong indication of Individualism. In spite of the fact that Power Distance, Individualism, Masculinity, and Uncertainty avoidance were mix loaded on factors that were revealed by PCA, this study continued treating them as cockpit culture variables. Based on the survey data, results from PCA, and theoretical structure of national culture framework proposed by Hofstede and Merritt [8, 11]. Merritt used correlation to demonstrate that her subset survey questions can predict Hofstede’s culture indices [11]. However, in her study, she had not conducted an exploratory factor analysis to examine latent factors that were driven from her survey questions. This study further examined her survey questions and found that environment setting was essential to abstracted factors. This founding echoed with Harari and

How Does National Culture Help Pilots

759

Perkins who suggested that a culture system is different within different environments (see Fig. 1) [1].

Fig. 1. Levels of Culture. [based on 1]

Abstracted factors from the cockpit culture survey were reflecting participants’ attitude in different environment settings. The environment setting was from a microcosmic level that was the self-esteem to a macro level that was the general society; the cockpit environment and the large group environment were falling somewhere in between. It had no questions to measure, or describe, a culture by using culture variables like Power Distance, Individualism, Masculinity, and Uncertainty avoidance. This study showed that aviation individuals mixed cockpit culture variables to navigate themselves in different environment. At a microcosmic level, the last factor of the PCA, the Masculinity grouping of questions was generated from the personal point of view. Question 1 and question 3 examined what student pilots thought about the independence; these two questions were from the measure of Individualism. Question 8 identified their self-confidence; this question demonstrates what the opinion of themselves was. Question 16 demonstrated their attitude in regarding rules for an individual; this question was a measure of Uncertainty avoidance. At this level, because every opinion was surrounded the idea of individual and oneself, the lack of competition and Masculinity was reasonable. This microcosmic level showed what was Chinese student pilots’ inner relationship with themselves in a measure of Power Distance, Individualism, and Uncertainty avoidance. Individual’s opinions and attitudes are changing based on the change of the environment. When the size of the environment escalated, people’s ideas would change accordingly. Under a cockpit environment, student pilots’ main task was conduction of a flight. Flying became the focus point at the second factor that the PCA was formed. Question 12, question 13, and question 14 were designed to measure the vulnerability and the will of accepting making errors within student pilots; these three questions were from Masculinity. Question 2 from Power Distance measured the idea of balance between flight and personal life. Question 6 from Power Distance demonstrated the idea of student pilots’ position in a cockpit. Question 18 from Uncertainty avoidance indicated students’ reliable on instructors. This factor was that what did student pilots think, when they were sitting in the cockpit and flying. At this environment, students were not able to just focus on themselves, but also, they needed to consider a relationship with instructors and a flying task. When there was an introduction of instructors and the associate relationships, the strong appearance of vulnerability of Masculinity showed up. As a student pilots, they were on a road from civilians to pilots;

760

X. O. Wu

and they generated errors and mistakes in the cockpit. Ipso facto the opinion of vulnerable of Masculinity was the dominant attitude when student pilots were in a cockpit and conducting a flight, and other measurements of cockpit culture variables, such as Power Distance, Individualism, and Uncertainty avoidance were all surrounded by the dominant attitude. When the environment grew larger, the dominant cockpit culture variables changes within student pilots. The first factor of the PCA was describe students’ attitude at an organizational level. At this level, the dominant cockpit culture variable was the Power Distance; question 7, question 9, and question 10 measured the perception of hierarchy in the organization. In an organization, more relationships were expected within student pilots; there were their peers, instructors, chief instructors, dispatchers, and administrators. In order to navigate within an organization—in this study, the environment setting was a flight school—the rules and regulations became important. Thus, two questions of Uncertainty avoidance (Q19, Q20) were loaded at this factor. Question 5 from Individualism demonstrated the relationship between individual and the group of people. The PCA revealed the third factor that described student pilots’ perceptions of cockpit culture variables the macro level. This factor focused on one thing: challenge. Question 11, question 15 and question 17 all had the notion of ideology of challenge and competition. Nota bene the aviation industry was dominant by males, and it always delivered an image of explorers, expeditioners, and pioneers to the public. Thus, student pilots had this very competitive imagination of being aviators. This imagination was also the expectation to aviators from general public.

References 1. Harari, Y.N., Perkins, D.: Sapiens: A Brief History of Humankind. Harvill Secker, London (2014) 2. Eurocontrol and FAA: Safety culture in air traffic management: a white paper. Government report (2008) 3. Ford, J., Henderson, R., O’Hare, D.: The effects of crew resource management (CRM) training on flight attendants’ safety attitudes. J. Saf. Res. 48, 49–56 (2014) 4. Griffin, M.A., Neal, A.: Perceptions of safety at work: a framework for linking safety climate to safety performance, knowledge, and motivation. J. Occup. Health Psychol. 5(3), 347 (2000) 5. Patankar, M.S., Sabin, E.J.: The safety culture perspective. In: Human Factors in Aviation, p. 95 (2010) 6. Wiegmann, D.A., Zhang, H., Von Thaden, T.L., Sharma, G., Gibbons, A.M.: Safety culture: an integrative review. Int. J. Aviat. Psychol. 14(2), 117–134 (2004) 7. Hofstede, G.: Value systems in forty countries: Interpretation, validation and consequences for theory. In: Ecksenberger, L., lonner, W., Poortinga, Y.H. (eds.) Cross Cultural Contributions to Psychology (1979) 8. Hofstede, G.: Cultures Consequences. Beverly Hills, Los Angeles (1980) 9. Hofstede, G., Hofstede, G.J., Minkov, M.: Cultures and Organizations: Software of the Mind, vol. 2. Citeseer (1991)

How Does National Culture Help Pilots

761

10. Minkov, M., Hofstede, G.: Cross-Cultural Analysis: The Science and Art of Comparing the World’s Modern Societies and Their Cultures. Sage, Thousand Oaks (2012) 11. Merritt, A.: Culture in the cockpit do Hofstede’s dimensions replicate? J. Cross-cultural Psychol. 31(3), 283–301 (2000) 12. Helmreich, R.L., Wilhelm, J.A., Klinect, J.R., Merritt, A.C.: Culture, error and crew resource management. In: Improving Teamwork in Organizations: Applications of Resource Management Training, pp. 305–331 (2001) 13. Johnston, N.: CRM: cross-cultural perspectives. In: Cockpit Resource Management, vol. 13, pp. 367–398 (1993) 14. Mjøs, K.: Cultural changes (1986–96) in a Norwegian airline company. Scand. J. Psychol. 43(1), 9–18 (2002) 15. Van Oudenhoven, J.P.: Do organizations reflect national cultures? A 10-nation study. Int. J. Intercultural Relat. 25(1), 89–107 (2001) 16. Civil Aviation Administration of China: The annual report addressing the development of chinese pilots 2017 (2018). http//pilot.caac.gov.cn/jsp/portals/newsList. jsp?type=Statisticalinfo 17. Krejcie, R.V., Morgan, D.W.: Determining sample size for research activities. Educ. Psychol. Meas. 30, 607–610 (1970) 18. Jiang, S.: The application of WeChat in universities student management. Science and technology vision (2016) 19. Liu, J., Wu, M., Lu, J., Zhou, L., Yu, X., Yan, D., Cao, Y., Chen, X., Guo, C.: The white paper of the development of Chinese online travel agency. China National Tourism Administration (2012) 20. Yang, Y.: 2016 chinese cellphone market report. Online (2016) 21. Tabachnick, B.G., Fidell, L.S.: Experimental designs using ANOVA. Thomson/Brooks/Cole (2007) 22. Abdi, H., Williams, L.J.: Principal component analysis. Wiley Interdisciplinary Rev.: Comput. Stat. 2(4), 433–459 (2010) 23. Bro, R., Smilde, A.K.: Principal component analysis. Anal. Methods 6(9), 2812–2831 (2014) 24. Webb, N.M., Shavelson, R.J., Haertel, E.H.: 4 reliability coefficients and generalizability theory. In: Handbook of Statistics, vol. 26, pp. 81–124 (2006) 25. Bagozzi, R.P., Yi, Y.: On the evaluation of structural equation models. J. Acad. Market. Sci. 16(1), 74–94 (1988) 26. Stevens, J.P.: Applied Multivariate Statistics for the Social Sciences. Routledge, Abingdon (2012) 27. Hair, J.F., Black, W.C., Babin, B.J., Anderson, R.E., Tatham, R.L.: Multivariate Data Analysis, 5th edn, Upper Saddle River (1998) 28. O’Rourke, N., Hatcher, L.: A step-by-step approach to using SAS for factor analysis and structural equation modeling. Sas Institute (2013) 29. Gill, G.K., Shergill, G.S.: Perceptions of safety management and safety culture in the aviation industry in New Zealand. J. Air Transp. Manag. 10(4), 231–237 (2004)

Human Reliability Quantification in Flight Through a Simplified CREAM Method Yundong Guo and Youchao Sun(&) College of Civil Aviation, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, Jiangsu, People’s Republic of China [email protected], [email protected]

Abstract. The complex flight procedures and various operating devices form a sophisticated operational context in flight, and the aircraft may encounter multitudinous risky factors. A large number of surveys show that human error is the most important factor in aviation accidents. The flight crew needs pay more attention to operational risks in critical flight-phases, and it is a serious concern for aviation safety to conduct human reliability analysis (HRA). However, the issues of lacking data, and the complexity of human behavior have greatly reduced the applicability of well-established HRA methods in flight context. The main purpose of the study is to determine human error probability (HEP) for specific flight tasks and predict safety level of operation in flight. This paper adopts a simplified Cognitive reliability and error analysis method (CREAM) to quantify human reliability for critical flight-phases. The example of HRA of the Boeing 737–800 operation process is utilized to demonstrate the proposed model. The results provide contributions to aviation safety and realizes the effective assessment of human reliability for specific flight tasks. Keywords: Human error


Aviation safety
Human reliability
CREAM

1 Introduction Probability safety assessment (PSA) is an evaluation method of engineering safety system and provides a great contribution for organizational and operational decisionmaking [1]. It mainly focuses on the effect of hardware and software reliability on system security without considering human reliability [2, 3]. However, with the advancement of new science and technology, the equipment reliability (hardware and software) has been greatly improved, human reliability is becoming increasingly important in sophisticated man-machine system. In fact, human error has become one of the factors that have the greatest impact on system security. Especially, in the area of aviation, the statistic data show that about 70% of civil aircraft flight accidents are caused by human factors [4, 5]. It means that the flight crew could make some mistakes and the vast majority errors are unintentional, but such human errors may finally lead to aviation accidents even though the aircraft technical system is good. A series of codes and recommendations, which presented by aviation companies and the relevant airworthiness authorities, provide security assurance for the operation of flight process. Although a mass of regulations and rules have been adopted by airworthiness © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 762–773, 2020. https://doi.org/10.1007/978-3-030-20503-4_68

Human Reliability Quantification in Flight

763

authorities, aviation accidents rate do still not drop to ideal level. One of the main reasons is substandard human actions in flight. Therefore, it is necessary to conduct human reliability analysis (HRA) in aviation industry, and the various HRA methods have been generally utilized to many industries with high safety requirements, such as nuclear power plant [6, 7], marine engineering [8, 9], healthcare systems [10, 11] and petroleum industry [12, 13], etc. However, HRA methods have roughly gone through two stages after the 1960s. The typical first-generation methods mainly have Human Cognitive Reliability (HCR) [14], Operator Action Trees (OAT) [15], Technique for Human Error Rate Prediction (THERP) [16], and Human Error Assessment and Reduction Technique (HEART) [17]. HCR and OAT are the time-decision methodology and they are suitable for tasks that are closely related to time, such as the operation in the event of emergency must have a rapid response time. Whereas, THERP and HEART are the taskdecision methodology and they are appropriate for some tasks that require low time, such as normal operations. In addition, all of these methods always suppose that the human errors are similar to the failure of mechanical, electrical, and structural parts, which have intrinsic defects and failure rate. Given the deficiency of first-generation methods, some second-generation methods have gradually been developed after the 1990s, such as A Technique for Human Event Analysis (ATHEANA) [18], and Cognitive Reliability and Error Analysis Method (CREAM) [19]. These methods emphasize that human reliability is determined by the task environment instead of the task type, and it indicates that researchers have focused on the important role of environmental factors. In addition, the HRA process of secondgeneration methods is based on the human cognitive model considering the factors of contextual conditions, operator and equipment status. Although HRA for nuclear industry utilizing CREAM methodology has demonstrated to be reliable and effective, related HRA studies of flight operation process is still limited. The process of executing flight task is in a complex and dynamic context with various uncertainties, and this is very different from nuclear industry and other industries. In general, the flight crew operations during the critical flight phases are always very sensitive and hazardous. There is no doubt that if there would be any operation errors during critical flight phases, it might result in security incidents and even disastrous accidents. Hence, correlative airworthiness authorities should stipulate control organizations to supervise the aircraft and the flight crew operations. Aviation companies should take a series of measures and rules to avoid serious risk. The flight crew must pay more attention to the potential risks during the operation process of critical flight phases. In addition, it is strongly necessary to expend HRA method to important flight operation process considering the special characteristics of flight mission. Therefore, this paper concentrates on extending CREAM technique for controlling and monitoring potential human errors in various contextual conditions. The aim of the method is to accurately evaluate HEP and determine safety control levels for a specific flight task. This paper includes four sections. This section summarizes the present situation and gives the purpose and significance of this paper. The following section states the research methodology. The HRA demonstration of Boeing 737–800 flight

764

Y. Guo and Y. Sun

operation process is offered in the third section. Consequently, conclusion is provided to reinforce aircraft operation safety and prevent the aviation accidents in the final section.

2 Method 2.1

The Basic CREAM Method

The CREAM method has four control modes and each mode has corresponding failure probability intervals. The basic theory of control modes is derived from the Contextual Control Model (COCOM), and the control level is influenced by the contextual conditions, the technical system, and personal experience, etc. It has a relationship between each control mode and common performance conditions (CPCs), and the specific control mode could be selected with the combined CPCs score. Furthermore, the corresponding probability interval of each control mode is illustrated in Table 1 [19]. Table 1. Control modes and probability intervals Control mode Strategic Tactical Opportunistic Scrambled

HEP intervals 0.00005 < P < 0.01 0.001 < P < 0.1 0.01 < P < 0.5 0.1 < P < 1.0

In special context situation, the value of combined CPC score can be expressed as an array [Rimproved, Rnot significant, Rreduced]. The value of Rimproved refers to the count of CPCs improved performance reliability. The value of Rnot significant refers to the count of CPCs which has no impact on performance reliability, and it can be neglected. Similarly, the value of Rreduced refers to the count of CPCs reduced performance reliability. In this paper, context impact index (CII) is proposed to simplify and quantify the basic CREAM method. The combined CPCs score can be represented with CII value as following Eq. 1 [20]. CII ¼ X  Y ¼ Rreduced  Rimproved

ð1Þ

Where the X refers to the count of reduced CPCs and the Y refers to the count of improved CPCs. If the equation of Y = X – CII is plotted, as Fig. 1, there would be three exceptions based on the CII numerical value. Nevertheless, the continuous and gradual change of the control modes in contiguous region will inevitably lead to few potential overlaps. Hence, the CII numerical value can be utilized to confirm the control mode and it is acceptable and reasonable.

Human Reliability Quantification in Flight

765

Fig. 1. CII and control modes.

2.2

The Extended CREAM Method

The basic CREAM method can be utilized to the original screening process. However, it has no regard for specific impact of a CPC upon the performance reliability. Therefore, the value of performance impact index (PII) is introduced to indicate the specific quantitative effects of the CPCs on performance reliability instead of the linguistic category (reduced, not significant and improved). The relationship between CII and PII could be expressed as Eq. (2) [20]. CII ¼

9 X

PII

ð2Þ

i¼1

The PII value can be confirmed by the weight factors based on the CREAM extend method and expert judgment, and it has been achieved so as to appoint one weight factor for different CPC level, as shown in Table 2 [20]. In this section, the CFP represents corrected failure probability of each cognitive function type, and it will be applied to the man-machine system operation action for calculating final HEP. Given nominal cognitive failure probability, denoted as CFP0, which is available through [19], then the function of CFP can be expressed as Eq. (3) [20]. CFP ¼ CFP0  100:25CII

ð3Þ

It is necessary to consider the dependencies between sub-tasks so as to determine the total HEP. In this paper, the whole task involves parallel and serial task. We can utilize the rules listed in Table 3 to decide the task type and calculate the total HEP [20].

766

Y. Guo and Y. Sun Table 2. PII values for CPCs CPC Adequacy of organization

CPC level Very efficient Efficient Inefficient Deficient Working conditions Advantageous Compatible Incompatible Adequacy of MMI and operational support Supportive Adequate Tolerable Inappropriate Availability of procedures/plans Appropriate Acceptable Inappropriate Number of simultaneous goals Fewer than capacity Matching current capacity More than capacity Available time Adequate Temporarily inadequate Continuously inadequate Time of day Day-time (adjusted) Night-time (unadjusted) Adequacy of training and experience Adequate, high experience Adequate, limited experience Inadequate Crew collaboration quality Very efficient Efficient Inefficient Deficient

PII values −0.6 0 0.6 1.0 −0.6 0 1.0 −1.2 −0.4 0 1.4 −1.2 0 1.4 0 0 1.2 −1.4 1.0 2.4 0 0.6 −1.4 0 1.8 −1.4 0 0.4 1.4

Table 3. Description of the rules for calculating the HEP System description System sub-task dependency Parallel system High dependency Low dependency Serial system High dependency Low dependency

HEP of the task HEPtask = Min (HEPsub-task i) HEPtask = P HEPsub-task i HEPtask = Max (HEPsub-task i) HEPtask = R HEPsub-task i

Human Reliability Quantification in Flight

767

3 HRA of Flight Operation Process 3.1

Analysis of Critical Operation Procedure

Flight process can be broken up into six flight-phase including take-off, climbing, cruising, descent, approach, and landing. The Boeing 737–800 operation process in flight is selected as an example with the simplified CREAM method. After taking off, the climbing operation process plays a very important role since the aircraft is more vulnerable to influence by birds or severe weather. In addition, the aircraft will be forced to land or even crash if it is attacked by birds. The cruising process is also a critical stage since it may reduce the crew situation awareness and lead to neglect some key operation procedure. The flight crew has heavy work tasks and high pressure in the process of descent, approach and landing, which is the high-risk stages of accident happened. Furthermore, when facing with the combined effects of terminal route requirements, air traffic control restrictions, airport conditions and other risk factors such as high-density traffic area around the airport, the flight crew needs timely and accurately to complete sophisticated operating procedures. Especially in emergency circumstances, if the crew has not accurately and quickly taken corresponding emergency procedure, it might cause an unexpected accident. Hence, the operations in climbing, cruising, descent, approach, and landing are used as the examples. And the specific operation procedure (OP) are available from the [21]. 3.2

Quantification Based on the Basic Method

The nine CPCs level of whole flight task has been determined based on the basic method and CPCs questionnaires, and all of CPCs level about the five important flight phases are provided in Table 4. In order to successfully accomplish the flight task, aviation companies provide comprehensive organization structure and adopt the well-designed flight operation procedure. Therefore, adequacy of organization is considered as ‘effective’, and availability of procedures/plans is considered as ‘appropriate’ for climbing and cruising procedure (OP1). Since the great improvement of pilot’s skills and experiences in the last few decades, adequacy of training and experience is identified as ‘adequate, high experience’. Furthermore, the flight crew has enough time to perform operation procedure for climbing and cruising process so that available time is identified as ‘adequate’. A large number of investigations from the flight crew indicates that adequacy of MMI and operational support is friendly, cockpit working conditions is fine, time of day has no influence, and the number of simultaneous goals is moderate during climbing and cruising process. Hence, the working conditions is deemed as ‘compatible’, the time of day is deemed as ‘not significant’, the adequacy of MMI and operational support is deemed as ‘adequate’, and number of simultaneous goals is deemed as ‘matching current capacity’. In addition, communication and cooperation between the flight crew is very smooth and relaxed. Thus, crew collaboration quality is considered as ‘efficient’. However, the nine CPCs level of descent process (OP2) are similar with OP1 as shown in Table 4.

768

Y. Guo and Y. Sun Table 4. Description of CPCs level assessment

CPC

Adequacy of organization Working conditions Adequacy of MMI and operational support Availability of procedures/plans Number of simultaneous goals Available time Time of day Adequacy of training and experience Crew collaboration quality

Climb and cruise procedure Efficient

Descent procedure

Approach procedure

Landing procedure

Efficient

Efficient

Efficient

Compatible Adequate

Compatible Adequate

Compatible Adequate

Compatible Adequate

Appropriate

Appropriate

Appropriate

Appropriate

Matching current capacity Adequate

Matching current capacity Adequate

More than capacity

More than capacity

Day-time Adequate, high experience Efficient

Day-time Adequate, high experience Efficient

Temporarily inadequate Day-time Adequate, high experience Inefficient

Temporarily inadequate Day-time Adequate, high experience Inefficient

During approach process (OP3) in the whole flight task, the number of simultaneous goals is more than climbing, cruising and descent process, and available time is urgent, since the operation process in approach stage is more complicated and needs to be completed accurately and timely. Therefore, number of simultaneous goals in this stage is confirmed as ‘more than capacity’, and available time is confirmed as ‘temporarily inadequate’. Similarly, the flight crew communication workload is excessive in that they need not only to evaluate constantly external environment and exchange information with each other, but also communicate timely with air traffic controller. Therefore, crew collaboration quality is deemed as ‘inefficient’. In addition, the remnant CPCs have no influence on the performance reliability during the approach process, and their influences are considered as ‘not significant’. The nine CPCs level of landing process (OP4) is illustrated in Table 4 respectively. Judging by Eq. (1), we can get related results, CII (OP1) = −3, CII (OP2) = −3, CII (OP3) = −1, CII (OP4) = −1. So for OP1, OP2, OP3 and OP4, the control mode is tactical. It implies that the flight crew performance reliability is reasonable and acceptable in accordance with operation procedure in critical flight stage, but some temporary deviation is possible. Therefore, the crew error probability is inside 1.0E−3 to 1.0E−1 as shown in Table 1. In addition, preliminary screening of human error mode based on the basic method is inadequate, and the extended method is provided to determine a numerical value for specific flight task.

Human Reliability Quantification in Flight

3.3

769

Quantification Based on the Extended Method

Each flight operation procedure is first decomposed into some sub-steps for identifying corresponding cognitive activities. Then, each cognitive function and probable failure type are determined separately. Consequently, each CFP is calculated with Eqs. (2) and (3) considering the nominal CFP values and each CPC features. Based on the outcome derived from the basic method, performance impact index values of nine CPCs for the aforesaid five critical flight-phase are respectively got, as shown in Table 5. Therefore, the CII values of OP1, OP2, OP3 and OP4 are obtained as follows according to Eq. (2). CIIOP1 ¼ CIIOP3 ¼

9 X

PII¼  4:4;

CIIOP2 ¼

9 X

i¼1

i¼1

9 X

9 X

PII¼  0:4;

CIIOP4 ¼

i¼1

PII¼  4:4 ð4Þ PII¼  0:4

i¼1

So corrected CFP for each sub-task can be calculated as follows according to Eq. (3). CFP1 ¼ CFP0  100:25CII ¼CFP0  101:1 CFP2 ¼ CFP0  100:25CII ¼CFP0  101:1

ð5Þ

CFP3 ¼ CFP0  100:25CII ¼CFP0  100:1 CFP4 ¼ CFP0  100:25CII ¼CFP0  100:1

Table 5. PII values under different flight tasks CPC

Adequacy of organization Working conditions Adequacy of MMI and operational support Availability of procedures/plans Number of simultaneous goals Available time Time of day Adequacy of training and experience Crew collaboration quality

PII value Climb and cruise procedure 0 0 −0.4

Descent procedure 0 0 −0.4

Approach procedure 0 0 −0.4

Landing procedure 0 0 −0.4

−1.2

−1.2

−1.2

−1.2

0

0

1.2

1.2

−1.4 0 −1.4

−1.4 0 −1.4

1 0 −1.4

1 0 −1.4

0

0

0.4

0.4

770

Y. Guo and Y. Sun

For each flight operation process, the final analysis results of operation actions according to the extended method are listed in Table 6. Table 6. Reliability analysis based on extended version Sub-steps Cognitive activity

Cognitive function

1.1 1.2 1.3 1.4 1.5 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 3.1 3.2 3.3 3.4 3.5 3.6 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 4.10 4.11 4.12 4.13 4.14 4.15

Execution Interpretation/Planning Interpretation Execution Observation/Interpretation Observation/Interpretation Observation/Interpretation Execution Planning/Execution Observation/Interpretation Execution Execution Execution Planning/Execution Execution Execution Planning/Execution Planning/Execution Execution Observation/Interpretation Execution Observation/Interpretation Observation/Execution Observation/Interpretation Execution Execution Planning/Execution Execution Execution Execution Execution Execution Observation/Interpretation Observation/Interpretation

Execute Evaluate Identify Execute Monitor Verify Verify Execute Co-ordinate Verify Execute Execute Execute Co-ordinate Execute Execute Co-ordinate Co-ordinate Execute Verify Execute Verify Regulate Verify Execute Communicate Co-ordinate Execute Execute Execute Execute Execute Verify Monitor

Generic failure type E3 I1 I1 E2 O2 O3 O3 E1 E1 O2 E1 E4 E4 E2 E3 E3 E2 E2 E2 O1 E2 O1 E4 O1 E3 E2 E2 E3 E2 E3 E3 E2 O1 O2

CFP0

Corrected CFP

5.0E−4 2.0E−1 2.0E−1 3.0E−3 7.0E−2 7.0E−2 7.0E−2 3.0E−3 3.0E−3 7.0E−2 3.0E−3 3.0E−3 3.0E−3 3.0E−3 5.0E−4 5.0E−4 3.0E−3 3.0E−3 3.0E−3 1.0E−3 3.0E−3 1.0E−3 3.0E−3 1.0E−3 5.0E−4 3.0E−3 3.0E−3 5.0E−4 3.0E−3 5.0E−4 5.0E−4 3.0E−3 1.0E−3 7.0E−2

4.0E−5 1.6E−2 1.6E−2 2.4E−4 5.6E−3 5.6E−3 5.6E−3 2.4E−4 2.4E−4 5.6E−3 2.4E−4 2.4E−4 2.4E−4 2.4E−3 4.0E−4 4.0E−4 2.4E−3 2.4E−3 2.4E−3 8.0E−4 2.4E−3 8.0E−4 2.4E−3 8.0E−4 4.0E−4 2.4E−3 2.4E−3 4.0E−4 2.4E−3 4.0E−4 4.0E−4 2.4E−3 8.0E−4 5.6E−2

Human Reliability Quantification in Flight

771

The five sub-steps of OP1 should be completed in an orderly and correct manner in accordance with climbing and cruising procedure. It means that the flight crew will operate unsuccessfully if any of the five sub-steps is failed (serial system). According to the calculating rules of serial system with low dependency, the total error probability value of OP1 is considered as the sum of five sub-steps values and the value is 3.8E−2. Similarly, the OP2 will not be completed successfully if any of eight activities fails (serial system). Therefore, according to the calculating rules of serial system with high dependency, the total error probability value of OP2 is 5.6E−3. In addition, the OP3 is also part of serial system and the six activities have high dependency. Therefore, the total error probability value of OP3 is 2.4E−3. Likely, the OP4 also belongs to serial system and the fifteen sub-steps have high dependency. Hence, the total error probability value of OP4 is 5.6E−2. Apparently, the whole flight task will not be completed properly if any of the four critical flight sub-tasks fails, since these sub-tasks constitute a serial system with high dependency. Therefore, the final HEP value is determined as the maximum value of four critical tasks, that is 5.6E−2. 3.4

Discussions on Results

The final HEP value indicates that the outcome based on the CREAM extended method is coincident with the basic method. In addition, the HEP value based on the basic method is an interval value (1.0E−3 < P < 1.0E−1), and the HEP value based on the extended method is 5.6E−2, which shows that it belongs to the probability interval. Therefore, the calculation results manifest that the uncertainty can be decreased apparently with the proposed method. The first Boeing 737–800 was put into use in 1998. We utilize the real flight data derived from aviation safety reporting system (ASRS) between 1998 and 2018 to validate the model. The available records of aviation safety incident or accident correlated with the flight crew operation are reviewed. While concentrating on events or errors during the five critical flight-phase in the past two decades, the database provided by ASRS indicates that there are 45 recorded operation errors during the critical flight phases (OP1, OP2, OP3, and OP4). The number of flight incidents or accidents of critical flight-phase in the last two decades is approximately 978 in total. So the error frequency of critical flight-phase is about 4.6E−2, which is in close proximity to the 5.6E−2. The statistical result indicates that the error probability value obtained by the extended method is consistent and acceptable. Consequently, the final HEP value demonstrates that the pilot’s performance reliability in critical flight process is desirable in accordance with standard operation procedure, even if there is some temporary deviation.

4 Conclusion A quantified and simplified CREAM method is proposed and applied into the critical flight-phase to evaluate human performance reliability. The results indicate that the HEP of the critical flight process is 5.6E−2, which is acceptable and desirable, even

772

Y. Guo and Y. Sun

there is a little bias. Furthermore, the available statistic data from ASRS has also proved that the results obtained with modified CREAM method are reasonable. This paper could make contributions to airlines and aviation organizations for evaluating human performance reliability in flight, and prompt the international airworthiness authorities to reinforce the management of aviation safety and prevent human errors in critical flight phases. In conclusion, a simplified and practical human reliability evaluation method is developed to monitor the crew cognitive actions or situation awareness in flight. The further research should be focused on the specific performance impact index value for aviation transportation to reinforce consistency in the CREAM. Acknowledgements. This research is supported by the National Natural Science Foundation of China (U1333119) and National Defense Basic Scientific Research program of China (JCKY2013605B002).

References 1. Groth, K.M., Swiler, L.P., Stevens-Adams, S.M., Smith, C.L.: A Bayesian method for using simulator data to enhance human error probabilities assigned by existing HRA methods. Reliab. Eng. Syst. Safety 128, 32–40 (2014) 2. Lin, Y.H., Li, Y.F., Zio, E.: Integrating random shocks into multi-state physics models of degradation processes for component reliability assessment. IEEE Trans. Reliab. 64(1), 154– 166 (2015) 3. Hao, H., Su, C.: A bayesian framework for reliability assessment via wiener process and MCMC. Math. Probl. Eng. 2014(3), 1–8 (2014) 4. Friedman, M.P., Carterette, E.C.: Human Factors in Aviation. Academic Press, London (2014) 5. Boyd, D.D.: A review of general aviation safety (1984–2017). Aerosp. Med. Hum. Perform. 88, 657–664 (2017) 6. Park, J., Jung, W.: Comparing cultural profiles of MCR operators with those of non-MCR operators working in domestic nuclear power plants. Reliab. Eng. Syst. Safety 133, 146–156 (2015) 7. Zhou, X., Deng, X., Deng, Y., et al.: Dependence assessment in human reliability analysis based on D numbers and AHP. Nucl. Eng. Des. 313, 243–252 (2017) 8. Yang, Z.L., Bonsall, S., Wall, A., et al.: A modified CREAM to human reliability quantification in marine engineering. Ocean Eng. 58(1), 293–303 (2013) 9. Wu, B., Yan, X., Wang, Y., et al.: An evidential reasoning-based CREAM to human reliability analysis in maritime accident process. Risk Anal. 37, 1936–1957 (2017) 10. Paul, B., Vitaly, L., Elena, Z.: New methods for healthcare system evaluation using human reliability analysis. In: Proceedings of the Human Factors and Ergonomics Society Annual Meeting, vol. 61, no. 1, pp. 583–587 (2017) 11. Chadwick, L., Fallon, E.F.: Human reliability assessment of a critical nursing task in a radiotherapy treatment process. Appl. Ergonomics 43(1), 89–97 (2012) 12. Gould, K.S., Ringstad, A.J., van de Merwe, K.: Human reliability analysis in major accident risk analyses in the Norwegian petroleum industry. In: Proceedings of the Human Factors and Ergonomics Society Annual Meeting, vol. 56, no. 1, pp. 2016–2020. Sage, Los Angeles (2012)

Human Reliability Quantification in Flight

773

13. Van De Merwe, K., Hogenboom, S., Rasmussen, M., et al.: Human-reliability analysis for the petroleum industry: lessons learned from applying SPAR-H. SPE Econ. Manag. 6(4), 159–164 (2014) 14. Wakefield, D.J.: Application of the human cognitive reliability model and confusion matrix approach in a probabilistic risk assessment. Reliab. Eng. Syst. Safety 22(1–4), 295–312 (1988) 15. Hall, R.E., Fragola, J., Wreathall, J.: Post-event human decision errors: operator action tree/time reliability correlation. Brookhaven National Lab., Upton, NY (USA); Science Applications, Inc., New York (USA); NUS Corp., Gaithersburg, MD (USA), 1982 16. Swain, A.D.: Handbook of Human Reliability Analysis with Emphasis on Nuclear Power Plant Applications. Sandia National Laboratories, Albuquerque (1983) 17. Williams, J.C.: A data-based method for assessing and reducing human error to improve operational performance. In: Human Factors and Power Plants, Monterey, CA, USA, pp. 436–450, 5–9 June 1988 18. Hahn, H.A.: The action characterization matrix: a link between HERA (Human Events Reference for ATHEANA) and ATHEANA (a technique for human error analysis). Los Alamos National Laboratory, NM (United States) (1997) 19. Hollnagel, E.: Cognitive Reliability and Error Analysis Method (CREAM). Elsevier, Amsterdam (1998) 20. He, X., Wang, Y., Shen, Z., et al.: A simplified CREAM prospective quantification process and its application. Reliab. Eng. Syst. Safety 93(2), 298–306 (2008) 21. Boeing 737–800 Standard Operation Procedure (SOP). Japan Airlines (2013)

The Human Element in Performance Based Navigation: Air Traffic Controller Acceptance of Established on Required Navigation Performance Procedures Lauren Thomas(&) and Alicia Serrato Evans Incorporated, 3110 Fairview Park, Falls Church, VA 22024, USA {lthomas,aserrato}@evansincorporated.com

Abstract. This paper provides a summary of results from a research project commissioned by the Federal Aviation Administration NextGen Human Factors Division to explore the factors associated with air traffic controller acceptance of a Performance Based Navigation (PBN) procedure, “Established on Required Navigation Performance” (EoR). Interviews were arranged at two terminal air traffic control facilities that were “early adopters” of EoR approaches. A total of 38 interviews were conducted with facility personnel, including 24 Certified Professional Controllers. Questions focused on how air traffic controllers integrated the new procedures into their controlling style and practice, and the organizational and operational factors that either supported or hindered controller utilization of the new procedures. A framework of the results is presented, providing insights into how to support air traffic controllers in moving towards trajectory-based operations. The results could be used to increase the probability that the potential benefits of PBN procedures can be realized. Keywords: Performance Based Navigation
Trajectory based operations User acceptance
Procedures
Established on Required Navigation Performance
Air traffic control




1 Introduction Performance Based Navigation (PBN) is an advanced form of air navigation that specifies precise three-dimensional flight paths [1]. Rather than certifying specific systems (including sensor equipment and crew requirements), PBN protocols specify the navigational performance that is required to permit the proposed operations. PBN protocols may include requirements in terms of the navigational system’s accuracy, integrity, availability, continuity and functionality. A Required Navigation Performance (RNP) level is specified for each PBN procedure, route or element of airspace. The RNP is expressed as a value that represents a distance performance tolerance in nautical miles from the intended position to the actual position of an aircraft - the lower the number, the higher the performance standard required. The RNP level is conveyed to pilots through published navigation specifications. A defining feature of RNP

© Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 774–782, 2020. https://doi.org/10.1007/978-3-030-20503-4_69

The Human Element in Performance Based Navigation: Air Traffic Controller

775

operations is the ability of the aircraft to monitor the navigation performance it achieves, and to inform the pilot if the performance requirement is not met during an operation. Using PBN procedures can result in highly accurate, consistent and replicable flight paths. The benefits can include more efficient airspace management, particularly in congested areas, because separation requirements can be reduced. There are also operational benefits such as optimized descents, and the ability to fly the safest and most efficient routes near difficult terrain. The environmental benefits include reduced fuel burn and exhaust emissions, as well as the potential for improved noise management with routing around noise sensitive areas. To maximize the benefits of PBN technology within the terminal environment, the Federal Aviation Administration recently introduced a concept known as “Established on Required Navigation Performance” (EoR) at several facilities [2, 3]. EoR approach procedures capitalize on the path-keeping capabilities of RNP technology. These approaches provide a pre-defined path to an aircraft on approach and may incorporate a precise curve towards the runway. To date, EoR operations within the National Airspace System (NAS) have utilized “authorization required” RNP approaches (RNP-AR), which require appropriately certified aircraft and trained aircrew1. The sharp precision of these approaches means that aircraft can be considered “established” on the approach at an earlier point than with a straight in-approach. From an airline perspective, the procedures allow an aircraft to turn to final sooner than on a vectored approach, and the pilots do not need to receive air traffic control instructions to make the turn since it is incorporated into the procedure. From an air traffic control perspective, an aircraft on an EoR approach may be considered as “established” while still downwind of the airport, prior to turning inbound and aligning with the extended runway center line for landing. The downwind leg and the inbound turn-to-final are incorporated within the pre-defined procedure. This means that the standard separation requirements cease to apply earlier in the approach, at a greater distance from the airport. The high accuracy of PBN capabilities allows controllers to take more of a monitoring role with aircraft on EoR approaches, making fewer transmissions than would be necessary if they were tactically directing the aircraft on the entire descent. This represents a fundamental shift in how controllers manage the traffic. While the human factors issues associated with PBN on the flight deck have received significant research attention [4, 5], efforts to understand the air traffic controller perspective have been more limited. At facilities where EoR approaches are available, air traffic controllers may choose to assign and clear eligible aircraft when they consider it appropriate. Understanding how controllers integrate EoR procedures into their existing controlling practices, and the operational and organizational factors that enhance or hinder that process, is critical to maximizing air traffic controller acceptance and adoption of EoR procedures and realizing the potential benefits of PBN.

1

The current research considered only RNP-AR approaches since these were in use at the two “early adopter” facilities implementing the EoR concept. In the future, EoR may be extended to other RNP approaches, including RNAV (GPS) and Advanced RNP (A-RNP).

776

L. Thomas and A. Serrato

2 Method The Terminal Radar Approach Control Facilities (TRACONs) at Seattle and Denver were the first air traffic control facilities within the NAS to introduce EoR into their operations. This research represented the first opportunity to conduct dedicated human factors research to understand the factors associated with successful procedure implementation from the air traffic controller perspective. The research team visited each facility and conducted a total of 38 interviews with a range of facility personnel including Certified Professional Controllers (CPCs), on-the-job-training instructors (OJTIs), contract training instructors, front line managers, support managers, bargaining unit representatives, and operational support engineers. Interviews at Seattle TRACON included 10 current CPCs and 6 other management and staff. Interviews at Denver TRACON included 14 current CPCs and 8 other management and staff. A site briefing was provided to each facility management team in advance, including the Air Traffic Manager and the National Air Traffic Controllers Association (NATCA) facility representative. Interview topics and suggested interview questions were provided in advance to allow each facility an opportunity to make suggestions for changes and improvements. Since the availability of certified professional controllers is always subject to operational demands and constraints, each facility was advised that interviews would accommodate operational shift patterns and breaks as necessary. All data was provided on a non-attributable basis, and every effort was made to preserve this undertaking throughout the data analysis phase. Interviews were conducted in an office located close to the operational floor and typically lasted 20 to 40 min.

3 Results and Discussion Thematic analysis of the interview data collected from Seattle and Denver TRACONs revealed eight broad categories of factors associated with successful implementation in the context of EoR. These were the collaborative context within the air traffic control facility, geographical factors, the nature and level of leadership support, the design of EoR procedures, fleet capability, time and opportunity, operational complexity, and individual factors. Consideration and integration of these factors appeared to be necessary to achieve acceptance of EoR procedures among air traffic controllers (Fig. 1).

Fig. 1. Framework of factors associated with EoR acceptance, utilization, and benefit realization.

The Human Element in Performance Based Navigation: Air Traffic Controller

3.1

777

Collaborative Context

Collaborative context refers to the nature of the working relationships among EoR stakeholders. In this context, air traffic stakeholders could include controllers at towers and TRACONs, airspace and procedures managers and specialists, training departments, operational support facility engineers, and bargaining unit representatives. Airline stakeholders would include pilots and flight crews as well as airline officials and managers. Representatives from airport authorities and regulatory authorities are also important stakeholders, and representatives from environmental groups, lobby groups, and local resident or home owner associations may also be important. Data from interviews at Seattle and Denver TRACONs indicated that high levels of collaboration at the beginning of the EoR project were associated with successful implementation of new EoR procedures. This was particularly the case for relationships between airlines and air traffic control, because it was associated with an increased recognition of the air/ground perspective. Denver reported that several people had working relationships with external stakeholder organizations that were both positive and collaborative in nature, enabling the facility to work with external stakeholders on mutually beneficial projects. For example, one air traffic controller functioned as the point of contact for recurrent joint training with an airline. Regarding this, one manager commented: “He goes to the airline to talk about air traffic, and he shares information on our operation. It’s outreach, and it’s a very good thing to support. I think it’s one of the reasons we have been successful, and we wish more airlines would do that. We have great relationships. We’re here for safety and our customers.”

Similarly, a quote from Seattle highlighted the value of shared understanding, derived partly from collaborative working relationships and joint training activities: “Simulators help a lot, and joint training of controllers and flight deck help both parties better understand each other. I don’t know what the pilot is doing other than typing in fixes. It would really help to know what this all looks like from the pilot’s side.”

3.2

Geographical Factors

Geographical factors near the air traffic control facility, including the runway configuration, the runway use plan, and the local terrain, were also reported to impact implementation success. Many of the controllers’ interviews acknowledged that geographical factors impacted the acceptance and utilization of the EoR procedures. At Seattle, one of the most significant geographical challenges reported was a local noise abatement restriction, which some controllers felt had unduly influenced the design of the procedures and resulted in a less than optimal approach design: “These approaches would be easier to use if the arc was further away from the airport. Right now, it joins final at six miles out. Unless we are incredibly slow, we hardly turn anyone there it is usually further away from the airport. If we were using it for everyone then it would be easy, but as soon as you get someone on the East side it’s difficult. There is no arc there because of the noise rules. Aircraft there must join much further North, eleven miles instead of six. If we could get rid of the noise abatement rules, it would be great.”

778

L. Thomas and A. Serrato

Geographical considerations mentioned at Denver included the mountainous terrain, the airport altitude, and the physical space available in terms of the number of runways. At Denver, the number of runways was associated with successful EoR implementation because the facility was able to segregate EoR operations to an outboard runway initially, making it possible for controllers to experiment using EoR procedures and become more comfortable with them, before they had to blend EoR traffic with non-EoR approaches in a single sequence. 3.3

Leadership Support

Support from facility leaders, managers and supervisors was reported to be a powerful driver of acceptance and utilization of the new EoR procedures. In this context, leadership support refers to whether leaders provided the time and resources necessary to reach successful design solutions, and to plan a managed and progressive implementation. This includes ongoing reinforcement to encourage both air traffic controllers and pilots to use the new procedures. One of the main challenges reported was encouraging controllers to keep offering these approaches to pilots, when the procedures were also new to those on the flight deck: “Many times, when pilots are offered the approach, they’ve declined, either because they already set up for ILS or because in their mind the approach is too cumbersome. Maybe one out of ten times they’ll say yes. And it’s not the controller’s job to persuade pilots to take these approaches.”

Leadership support included supervisors understanding the challenges controllers face in learning how an aircraft on an EoR approach will behave. Because these approaches are “hands-off”, it can be difficult for controllers to trust that the procedure will work, and successful implementation requires that supervisors recognize that there is a learning curve for everyone: “In terms of understanding the controller perspective, they don’t like trying to use new things when they are busy. You need to give them time to “play” with the approaches, check that they are doing it, and make it easier for them to try it. We also had to get them to start sequencing with these approaches when they were mixing them into a sequence for another runway.”

3.4

Procedure Design

The design of the EoR approach procedure is an important factor in air traffic controller acceptance. There is regulatory guidance that governs the design, development and implementation of PBN procedures [6, 7]. The interview data emphasized that designing good procedures takes time and commitment and requires the collaboration of all impacted stakeholders. In addition to the intended use of the approach, “anticipated under-use” might also need to be considered. At some airports, pilots may prefer to land at runways closer to the terminal, for example, impacting requests for runways and approaches. Differences between operator business models also need to be considered – what works for the dominant carrier might not work for all operators. Successful design initiatives are likely to be those that include some short-term wins for all operators, including those not yet fully equipped to fly an RNP approach.

The Human Element in Performance Based Navigation: Air Traffic Controller

779

From the perspective of encouraging air traffic controllers to use the approach procedures, there was reported to be a need to overcome a tendency for air traffic controllers to prefer vectoring: “The design must be better than what a controller can achieve by vectoring. It cannot be as good, because that won’t be good enough. And it cannot be worse – because for them to assign it, you need to overcome their beliefs that vectoring is better.”

An approach procedure that is well-designed from the perspective of airlines and local residents may not be assigned and cleared if the air traffic controllers do not believe it to be the most efficient and effective option available to an eligible aircraft at the time. Additional considerations reported in this category of the framework were the “error-resistance” of the approach and the need to consider the potential impact of the new procedures on alarms and alerts. 3.5

Fleet Capability

Fleet capability refers to the extent to which the fleet at TRACONs and airports is equipped and able to use various approach procedures. There is likely to be a mixed fleet with varying RNP capabilities at airports and managing this mix can represent a major challenge for air traffic controllers. Terminal controllers need to sequence aircraft with different capabilities on different types of approaches, and achieving this via radar control is a complex judgement of speed, distance, and prevailing conditions: “It’s like you’re trying to work around the EoR aircraft instead of with it. It feels like you’re cutting in line. That one approach feels like an exception. It makes the workload harder. Trying to work that one approach into everything else? The wind can change and when it does, it changes everything. The challenge is making it all work together in an integrated way, instead of doing it on an individual basis.” “It’s very hard for controllers to mix different types of approaches because the continuity is not there. You may forget, and a lot can go wrong, something can slip through the cracks. There are times when an aircraft coming from South wants the EoR approach, but if there are seven aircraft on final, I’ll just say ‘unable’.” “I think it should be like a High Occupancy Vehicle lane, with no blending of the other traffic. The blending is the hardest. It doesn’t matter how veteran you are – the eye-balling is a challenge.”

The “eye-balling” referred to here is a traditional method of determining spacing and sequencing using radar. Decision support tools, such as range rings, tie-points and automation adaptations, can be used to provide support to controllers in clearing and monitoring EoR approaches, which helps them to better manage a mixed sequence. 3.6

Operational Complexity

Every air traffic control facility has a unique combination of factors that drive complexity, including the traffic volume and mix, the operational tempo of peaks and troughs in demand, and the interaction between terrain and the weather. Since the sources of complexity are different at every facility, a “one-size-fits-all” approach to

780

L. Thomas and A. Serrato

procedure implementation is unlikely to be successful. Air traffic controller acceptance is contingent on tailoring the procedure implementation to local complexity and allowing controllers the scope to adjust their plans accordingly: “We’re a complex operation! We have the weather, the winds, the mountains, the runways, the tailwinds. Really, it’s all complex.” “There’s the airspace complexity, and the interaction with Boeing Field. With traffic volume and complexity, when there’s drama, conflict or high traffic, our first reaction is to ditch the EoRs and go for vectoring. You’ve got to consider the Paine Field interactions too. We’re talking all kinds of hiccups.”

3.7

Time and Opportunity

For controllers to assign and clear an EoR approach, they need to be able to identify a space in their planned landing sequence for the eligible aircraft. It can be challenging for controllers to identify where they may be able to “insert” an EoR approach, and to have confidence that the required spacing and separation can be maintained between the EoR aircraft and the other aircraft in the sequence. Decision support tools can assist controllers with making these predictions. A local adaptation to the Converging Runway Display Aid (CRDA) was reported to be of significant value as a decision support tool to controllers at Denver TRACON. This adaptation to CRDA generates a “ghost” target that appears on an extended runway centerline, allowing controllers to see the position of an aircraft on an EoR approach “as if” it were flying a straight in approach [8]. The interview results indicated that this tool was hugely valuable to controllers in helping them to identify opportunities to integrate an EoR approach into their planned landing sequence: “CRDA helps make the assessment more consistent, now I can tell earlier if I can make it work. It’s the time window. The feeders use it too – they can see 35 miles out, so it’s much more predictable.” “CRDA takes a lot of the work out, you know in advance if they will be a fit, or if they will be a tie.”

The time and opportunity theme also included the need to provide air traffic controllers with the time and opportunity to practice using EoR under different conditions and gain familiarity and confidence with these procedures. This is important to assist air traffic controllers in integrating EoR into their preferred suite of controlling techniques. If air traffic controllers are working consistently at high capacity and high workload, they will find it challenging to try something new, and acceptance of the procedures is likely to be negatively impacted. 3.8

Individual Factors

There are also subjective factors that play a role in whether a new procedure will be accepted by air traffic controllers. Some air traffic controllers are very open to trying

The Human Element in Performance Based Navigation: Air Traffic Controller

781

new things, while others tend to prefer tried and tested techniques. A variety of individual factors including workload, motivation, teamwork, confidence and the level of training were reported to influence an individual controller’s decision to try EoR procedures: “As the feeder you don’t know what the exact sequence is, or what the final controller has in mind. It depends on which downwind they’re coming in, the base turn in. You never know. You don’t want to throw them under a bus if you can avoid it. So, I put off the decision as a feeder and let the final controller decide whether they can do it. It can affect other parts of the sequence.” “My driver, why I use them, is to reduce carbon emissions. I know that’s rare… of the 80 folks I work with, two or three might think that carbon emissions are an issue for us as a species, for our planet.” “It’s about lack of control. With this the pilots are in control, not the controllers. And we like being in control.”

4 Conclusion One quotation from a controller sums up succinctly the challenge of air traffic controller acceptance in using EoR procedures: “You know how it is with a new way of doing anything. 10% will be the champions and the trendsetters, 80% might not be into it to start with, but will give it a try, and 10% will drag their heels. It takes time to change behavior and change culture. It’s about habits, and trust, and knowledge.”

Controllers ultimately rely on their own expert judgement to make optimal decisions for every aircraft in their airspace. Controllers are keenly aware of their professional responsibility to ensure both safety and efficiency, and they do not take this responsibility lightly. Encouraging these experts to try EoR procedures requires them to accept being less directly involved in controlling an outcome that they regard as a personal obligation. Controllers are required to take a “leap of faith” to trust procedures that are new and unfamiliar. This paper summarizes some of the factors affecting controller acceptance, based on human factors interviews conducted at two EoR early adopter sites. To manage the transition to EoR, implementation teams and human factors specialists should devise a carefully planned approach to operationalizing procedures within air traffic control facilities, to ensure air traffic controller needs and concerns are identified and addressed [9]. Air traffic controller acceptance is a key determinant of the way that PBN procedures will be utilized, and controller acceptance is necessary to deliver the potential benefits into the NAS. Acknowledgements. The authors would like to record their appreciation to Bill Kaliardos, Federal Aviation Administration NextGen Human Factors Division, and to Mitchell Bernstein, Federal Aviation Administration NextGen Technology Development and Prototyping Division (Navigation Branch) for their support and guidance throughout this research project.

782

L. Thomas and A. Serrato

References 1. Federal Aviation Administration: Performance Based Navigation Strategy. U.S. Department of Transportation, Washington D.C. (2016) 2. Mayer, R.: Established on required navigation performance (EoR) concept validation. Technical report, MITRE (2016) 3. Pollock, M., Hudak, T., Spelman, J.: Methodology for established on required navigation performance (RNP) concept validation for a dependent parallel approach operation. Technical report, MITRE (2015) 4. Chandra, D.C., Markunas, R.: Line pilot perspectives on complexity of terminal instrument flight procedures. In: IEEE/AIAA 35th Digital Avionics Systems Conference, pp 1–10. IIEEE Press, Sacramento (2016) 5. Chandra, D.C., Markunas, R.: Line pilot perspectives on complexity of terminal instrument flight procedures. Technical report, Volpe National Transportation Systems Center (2017) 6. Federal Aviation Administration: Performance based navigation implementation process. Air traffic organization policy JO 7100.41. Effective April 29, 2016. U.S. Department of Transportation. Washington, D.C. (2016) 7. Federal Aviation Administration: Unites States standard for performance based navigation (PBN) instrument procedure design. JO8260.58A. Effective 03/14/2016. U.S. Department of Transportation. Washington D.C. (2016) 8. Garcia, R., Goodlin, T.: Converging runway display aid CRDA denver TRACON RNP-AR sequencing with straight-in approaches. Technical report, Denver Terminal Radar Approach Control (2016) 9. Thomas, L.J., Serrato, A., Newton, K.: Established on required navigation performance (EoR) (RNP) concept validation and implementation plans: human factors gap analysis. Technical report, Evans Incorporated (2018)

Ergonomic Assessment of Instructors’ Capability to Conduct Personality-Oriented Training for Air Traffic Control (ATC) Personnel Oleksii Reva1(&), Sergii Borsuk2, Valeriy Shulgin3, and Serhiy Nedbay4 1

3

Ukrainian Institute of Scientific and Technical Expertise and Information, Kiev, Ukraine [email protected] 2 National Aviation University, Kiev, Ukraine [email protected] Flight Academy of the National Aviation University, Kropyvnytskyi, Ukraine [email protected] 4 Flight School “Condor”, Kiev, Ukraine [email protected]

Abstract. The mutual influence of the components of the safety paradigm of ICAO is explained from the standpoint of the manifestation of the human factor - “attitudes towards safe actions or conditions”. “Attitude” is characterized by the following indicators of decision-making: the main dominants (propensity, aversion, indifference to risk), levels of aspirations, and fuzzy estimates of risk on the ICAO scale. The indicators are examined through the attitude of ATCstudents and professional ATCs to violations of the standards for the separation of aircrafts in horizontal flight. This made it possible to evaluate potential violations in well physically measured and interpreted distances between controlled aircraft. Quantitative indicators of violations were described in the algorithm for conducting personality-oriented simulator training of ATCstudents. The additional instructor’s work load associated with the implementation of the algorithm is determined by the following indicators: stereotype (Zn = 0.58, which clearly fits the established standards of 0.25
Zn
0.85) and logical complexity (Ln = 0.30), which is 1.5 times higher than capabilities of the “average” operator. The paper suggests the directions for improvement of professional training of ATC-students. Keywords: Flight safety  Human factor  Decision-making  Air traffic control  ATC students personality-oriented training algorithms Informative criteria  Operator-instructor  Ergonomic indexes of actions of logical complexity and stereotype

© Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 783–793, 2020. https://doi.org/10.1007/978-3-030-20503-4_70



784

O. Reva et al.

1 Introduction Today, it is generally accepted that well educated and professionally trained front-line aviation operators, including air traffic controllers (ATCs) can effectively address the consequences of technical failures in the complex organizational and active control system “flight crew - aircraft - environment – ATC” [1]. Improving the training of ATCs should be continuous and carried out in the framework of the ICAO’s concept of ensuring flight safety [2]. The mutual influence of the components of this concept is explained from the standpoint of the manifestation of the human factor through “the attitude of personnel to safe actions or conditions” (Fig. 1) [3]. We associate this “attitude” with decision-making, because: (1) it is the most frequently repeated type of human intellectual activity; (2) ATCs’ professional activity is usually viewed as a continuous chain of decisions made and implemented in explicit and implicit forms under the influence of various factors, especially risks; (3) statistics show that the overwhelming majority of aviation accidents are the result of erroneous decisions. Therefore, ICAO included in the qualification requirements for the ATCs “the ability to make decisions and fulfill the duties necessary to ensure safe, orderly and operational dispatching services at a level corresponding to the rights granted, including recognition and control of threats and errors” [4]. The main decision-making dominants characterize attitude to risk and are determined by constructing and analyzing estimated utility functions of the continuum separation rate in the process of solving the “closed” decision-making problem [1, 3, 5–7 and others]. Propensity to risk/risk aversion characterizes the motivation to achieve success/avoid failure. Most professional ATCs demonstrate propensity to risk (Table 1) [8]. Table 1. The manifestation of the main dominant decision-making as a function of the experience of air traffic control. Separation rate Ratio of persons with different psychological dominant of decision making ATC-students ATC profeccionals 1 2 3 L = 10 km P : I : A , 1 : 1; 7 : 25; 4 A : I : P , 1 : 2; 8 : 4 , 3; 5% : 6; 2% : 90; 3% , 12; 9% : 35; 7% : 51; 4% L = 20 km P : I : A , 1 : 3; 8 : 22 A : I : P , 1 : 4; 5 : 12 , 6; 6% : 11; 8% : 81; 6% , 5; 7% : 25; 7% : 68; 6% Note: P– propensity to risk, A- aversion to risk, I– indifference to risk.

The risks of a non-stochastic nature in aviation are illustrated by the ICAO scale of the level of danger [2], which, based on the methodology of fuzzy mathematics, can be represented as a term set of the corresponding linguistic variable [9, 10]:

 ~D ~ C þ Dangerous R T M ðLDÞ ¼ Catastrophic R


 ~ ~ ~M ; þ Significant RS þ Insignificant RI þ Miserable R

ð1Þ

Ergonomic Assessment of Instructors’ Capability

785

Fig. 1. The illustration of the mutual influence of the components of the safety concept of ICAO with the manifestation of the human factor.

where “+” is the generally accepted sign of the logical combination of terms - the names of the hazard levels assessed in the appropriate scale This scale was used to construct fuzzy qualimetric models of the attitude of ATCstudents and professional ATCs to violations of aircraft separation standards (Fig. 2) and to solve the “Risk triangle” in well-presented and interpreted indicators of distances between aircraft.

786

O. Reva et al.

Fig. 2. An example of a fuzzy proactive model of qualimetry attitudes of ATC students to violations of the separation level of aircraft L = 8 km.

The level of claims L* is the main system-forming factor of the personality of aviation operators, which determines the adequacy of their self-esteem and is necessarily revealed among the participants of aviation accidents. By the level of claims, we mean the distance between the controlled aircraft (within the norm of separation), which corresponds to the maximum jump in utility (acceptability) in the ATC’s views about the possibility of ensuring an adequate level of safety. The level of claims is determined by solving the conditionally “open” task of making decisions and constructing an estimator of the utility of the continuum separation norm for a formally unlimited number of points [1, 11, 12 and etc.].

2 Development of the Algorithm of Personality-Oriented Simulator Training for Air Traffic Control Students Let us consider an algorithm for the individualization of simulator training of ATCstudents, taking into account proactive qualimetric indicators of their attitude to violations of the separation standards (horizontal flight). We are talking about formalized instructions for the instructor of the dispatching simulator, which contribute to obtaining the target result of training. During the development of the algorithm (Fig. 3), we sought to ensure its compliance with informative criteria (without ranking): cyclicity; determinism; discreteness; mass character; finiteness; correctness; performance. We will determine the rules of modeling training exercises during the implementation of the algorithm. 1. The implementation of the algorithm is carried out taking into account the main dominants of decision-making, levels of aspirations and fuzzy estimates of risk, the number of ATC students, as well as separation standards. 2. Solving the “closed” decision-making task, successively for the continuum of each i-th separation standard, individual evaluation utility functions are constructed, from the analysis of which the attitude to violations of this norm (propensity, aversion, indifference to risk) isrevealed.

Ergonomic Assessment of Instructors’ Capability

787

Fig. 3. Algorithm of the organization of person-oriented simulator training for air traffic control students, based on human factor (fragment)

3. Solving the “open” decision-making problem, successively for the continuum of each i-th separation standard, individual estimated utility functions are constructed, from the analysis of which the level of claims is found. 4. Group evaluation utility functions are constructed. 5. The implementation of the algorithm in the general case should lead to the following change in the main dominant of the decision-making by ATC-students: aversion ! indifference ! propensity to take risks. Or, the result of training should be an increase in the level of claims of ATC-students (decrease in absolute value)

788

O. Reva et al.

within their decision-making dominant, if it remained unchanged. In this case, a synergistic effect is achieved in the trainig. 6. Formation of the skills of ATCs is carried out with the following sequence of use of separation standards: L ¼ 30 km ) L ¼ 20 km ) L ¼ 12 km ) L ¼ 10 km ) L ¼ 8 km:

ð2Þ

7. Considering the specificity of training for ATC students, the simulation of training exercises does not include indicators of distance between aircraft, which correspond to a catastrophic level of danger (see Fig. 2). That is, the condition L > LC is always satisfied. 8. If the ATC-student has the necessary psycho-physiological potential and an adequate level of knowledge and professional skills, he/she will definitely complete a training exercise, despite the possible contradiction between his/her self-esteem (level of claims) and actual training results. This will motivate the ATC-student for a more adequate assessment of personal knowledge and professional skills, and, as a result, help improve the level of flight safety he/she provides. 9. Management of training provides the opportunity to address the issue of professional fitness of ATC-students who could not master the training exercises in the simulation of the normative conditions of professional activity (Table 2). Table 2. Conditions for modelling of distances between aircraft during simulation training of air traffic control students

1 Propensity to risk Indifference to risk Aversion to risk Notice:

 g L

Values of the level of claims 2 g Lj \L

The sequence of establishing distances between aircraft 3 LD ) LS ) LI ) LM

g Lj  L g Lj \L

LS ) LI ) LM

g Lj  L g Lj \L

LI ) LM

g Lj  L

LM

average level of claims in the group

3 Ergonomic Estimation of the Correspondence of the Developed Algorithm to the Psycho-Physiological Capabilities of the Instructor The instructor of the simulator performs a variety of operator functions in professional activity, acting as an operator-technologist, operator-manager, operator-observer (controller), operator-manipulator, and operator-researcher [13]. Let us evaluate the capacity of the instructor with “average” psycho-physiological capabilities to perform additional functions related to the implementation of the proposed personality-oriented

Ergonomic Assessment of Instructors’ Capability

789

simulator training algorithm for ATC-students. We consider it expedient to use ergonomic indicators of stereotype Zn and logical complexity Ln of operator activity for this purpose [13–17 and others]. The analysis of the complete algorithm of implementation, focused on the human factor training for ATC-students (Fig. 3), shows that the entire algorithm consists of N = 42 members, of which N0 = 30 is the number of elementary operators distributed by n0 groups and Nlog = 12 logical conditions distributed by nlog groups. Further, the algorithm is divided into complex groups, which include one group of elementary operators and logical conditions. Suppose each complex group contains elements from which m0 are elementary operators and mlog - logical conditions. The stereotype of the algorithm depends on: – the number of elementary operators in the algorithm; And if N = const then the larger N0, the more stereotype component is expressed; – the number of groups of operators; if N = const and N0 = const, then with the decrease of the exponent the stereotype component of the algorithm increases; – the total number of members of the algorithm: if N = const and N0 = const, then, with the increase in the total number of members of the algorithm N (with the addition of logical conditions), the stereotype component of the algorithm decreases; – Distribution of operators by complex groups. These factors can be taken into account by the ratio N0 =N that characterizes the content of elementary operators in the algorithm, as well as by the relations m0 =N0 and m0i =mi , which characterize the distribution of operators in groups. Then the expression for the normalized coefficient of stereotype activity Zn will have the following form: Zn ¼

N0 1X m20i N i¼1 mi

ð3Þ

A normalized coefficient of the logical complexity of the activity Ln is determined as follows: log: m2log:j 1 X N j¼1 mj

n

Ln ¼

ð4Þ

The breakdown of the algorithm in Fig. 3 for complex groups when calculating Zn should be carried out from the first group of operators, and when Ln is calculated, from the first group of logical conditions, that is, the group of elementary operators facing the logical condition is not taken into account, therefore in formula (3) instead of N we write N*. Zn and Ln are normalized, so that Zí = [0,1] и Lí = [0,1].

790

O. Reva et al.

If conditions are fulfilled [14] 0; 25
Zn
0; 85:

ð5Þ

Ln
0; 2

ð6Þ

then we can assume that the algorithm under study fully takes into account the capabilities of the human operator in its implementation. If the criterion restrictions (5) are not met, especially if: Zn  0; 9:

ð7Þ

then it is considered necessary to transfer the functions of the human operator (in our case, the simulator instructor) to the computer. That is, we are talking about the need to develop a special intelligent module of the instructor’s decision support system. Having divided the algorithm into complex groups and applying formulas (3) and (4), we find that the stereotype indicator Zí = 0,58 and satisfies the criterion (5). At the same time, the indicator of logical complexity Li = 0.30, whih is 1.5 times worse than criterion (6). Thus, the question arises about the development of an online module of the decision support system for an instructor, which should be the subject of further research.

4 Conclusions 1. The paper has substantiated the interaction of components of the ICAO safety concept from the human factor perspective, namely, “the attitude of personnel to safe operations or conditions”. 2. The paper proposes an algorithm for simulator training, which considers human factor in decision-making main dominants in decision-making, levels of claims, and fuzzy risk assessments. The algorithm meets the criteria of cyclicity, determinism, accuracy, discreteness, finiteness among others. The paper also substantiates the necessity of developing ATC students’ motivation to succed. Therefore, the training exercise should consider ATC students’ attitude towards the distances between airplanes. 3. The readiness level of the simulator instructor to manage a personality-oriented training of ATC students is assessed using the ergonomic indicators of stereotype and logical complexity. It was established that for the proposed algorithm, the rationed indicator of the stereotype of the instructor’s actions is equal to Zn = 0.58 and clearly meets the criteria. At the same time, the normalized indicator of the logical complexity of actions is equal to Ln = 0.30, which is 1.5 times higher than the standard values established for the average operator. 4. Further research should focus on development of an on-line module of the instructor’s decision support system.

Ergonomic Assessment of Instructors’ Capability

791

References 1. Reva, A.N., Tumyishev, K.M., Bekmuhambetov, A.A., Reva, nauch. red. A.N., Tumyishev, K.M.: Chelovecheskiy faktor i bezopasnost poletov: (Proaktivnoe issledovanie vliyaniya) [Tekst]: monografiya. Almatyi, 242 s (2006) 2. Podgotovka letnogo ekipazha: optimizaciya raboty ekipazha v kabine (CRM) i letnaya podgotovka v usloviyax, priblizhennyx k realnym (LOFT) [tekst]. Chelovecheskij faktor: sbornik materialov № 2: Cir. ICAO 217 – AN/132. Monreal, Kanada (1989) 3. Guidelines for TRM Good Practices [Tekst], EUROCONTROL, 18 March 2015 4. Jensen, R.S., Andrien, J., Lawton, R.: Aeronautical decision making for instrumental pilot [Tekst]. DOT/ FAA/ PM-86/42 5. Brecher, B.R.: A question of judgment [Tekst]. Flying, vol. 108, № 5, pp. 48–52 (1981) 6. Reva, O.M., Mirozoev, B.M., Nasirov, Sh.Sh., Muxtarov, P.Sh.: Profesijni situativni vpravi diagnostiki i ko-rekciї nebezpechnix strategij prijnyattya rishen aviadis-petcherami [tekst]. Suchasni informacijni ta in-novacijni texnologii na transporti (MINNT - 2013): zb. m-liv P’yatoї Mizhnar. nauk.-prakt. konf. u 2-x t. – Kherson, pp. 28–30 travnya 2013 r., - T. 2. KhDMA, Kherson, S. 23–26 (2013) 7. Reva, O.M., Mirozoev, B.M., Nasirov, Sh.Sh., Muxtarov, P.Sh.: Rozrobka metodichnogo zabezpechennya procedur diagnostiki i korekciї nebezpechnix strategij prijnyattya rishen aviadispetcherami [tekst]. Naukovij visnik xersonskoї derzhavnoї morskoї akademiї: nauk. zh. vyd-vo XMDA, Kherson, № 1, pp. 90–96 (2013) 8. Reva, O.M., Borsuk, S.P.: Nechitka model stavlennya avi-adispetchera do riziku nastannya potentsiyno-kon-fliktnoyi situatsiyi [Tekst]. AvIatsIyno-kosmIchna tehnika I tehnologiya: nauk.-tehn. zh. – H. NatsIonalniy aerokosmichniy un-t im 9. Zhukovskogo, E.: « HAI » - No. 10, pp. 214–221 (2013) 10. Reva, O.M., Borsuk, S.P., Shylgin, V.A.: Viznachennya granichnix rivniv riziku pid chas porushennya normi eshelonuvannya po-vitryanogo prostoru [tekst]. Aviacijno-kosmichna texnika i texnologiya: nauk.-texn. zh. – Kh.: Nacionalnij aerokos-michnij un-t im. M.E. Zhukovskogo « XAI » , № 9, pp. 151–156 (2014) 11. Borsuk, S.P.: Viznachennya osnovnoї dominanti povedinki studentiv dispetcheriv v umovax porushennya norm eshelonuvannya [tekst]. naukoєmni texnologiї. K., № 3, S. 261–265 (2015) 12. Reva, O.M., Borsuk, S.P.: Vpliv specifiki zastosuvannya normi eshelonuvannya na osoblivosti proyavu rivniv domagan aviadispetcheriv [tekst]. Naukovij visnik Khersonskoї derzhavnoї morskoї akademiї: nauk. zh. vyd-vo KhMDA, Kherson, № 1 S. 281–288 (2015) 13. Reva, O.M., Borsuk, S.P.: Pilotnij analiz rivniv domagan aviadispetcheriv na spektri gorizontalnix norm eshelonuvannya povitryanogo prostoru [tekst]. Aviacijno-kosmichna texnika i texnologiya: nauk.-texn. zh. – Kh.: Nacionalnij aerokosmichnij un-t im. M.Є. Zhukovskogo « HAI » , № 9, S. 153–160 (2015) 14. Reva, O.M., Borsuk, S.P.: Appliance of area under air traffic controller estimate function for main decision taking dominant determination [Tekst]. Aviacijno-kosmichna texnika i texnolo-giya : nauk.-texn. zh. Kh.: Nacionalnij aerokosmichnij un-t im. M.Є. Zhukovskogo « HAI » , № 7, S. 157–163 (2016) 15. Reva, A.N., Muxtarov, P.Sh., Mir-zoev B.M., (ta in.): Dinamika osnovnoj dominanty prinyatiya rishenij aviadispetcherom pri uslozhnenii uslovij dey-atelnosti [tekst]. Suchasni informacijni ta innovacijni texnologiї na transporti (MINNT - 2014): zb. m-liv VI Mizhnar. nauk.-prakt. konf., prisvyachenoi 180-richchyu z dnya zasnuvannya Khersonskoї derzhavnoї morskoї akademiї, Kherson, 27–29 travnya 2014 r. vyd-vo XDMA, xerson, S. 86–89 (2014)

792

O. Reva et al.

16. Muhtarov, P.Sh.: Osnovnyie dominantyi v prinyatii resheniy aviadispetcherom pri otsenke polez-nosti-bezopasnosti normyi eshelonirovaniya vozdushno-go prostranstva [Tekst]. AvIatsIyno-kosmIchna tehnIka i tehnologIya : nauk.-tehn. zh. К.: NatsIonalniy aerokosmIchniy un-t im. M.E. Zhukovskogo « HAI » , № 9, S. 143–150 (2014) 17. Reva, O.M., Muhtarov, P.Sh., NasIrov, Sh.Sh. (ta in.): Stiykist osnovnoyi dominanti priynyattya rIshen avIadispetcherom v umovah riziku [Tekst]. AvIatsIyno-kosmIchna tehnIka I tehnologiya: nauk.-tehn. zh. Kh.: Natsionalniy aerokosmIchniy un-t im. M. E. Zhukovskogo « HAI » , No. 10, S. 147–153 (2014) 18. Pravila medichnogo rozsliduvannya avia-cijnix podij [tekst]. Zatv. nakazom Derzhavnoї sluzhbi Ukraїni z naglyadu za zabezpechennyam bezpeki aviaciї 05.12.2005, № 919 19. Rukovodstvo po upravleniyu bezopasnostyu poletov (RUBP) [Tekst]: Doc. ICAO 9859 – AN/474. – Izd-e 3-e. - Monreal, Kanada (2013) 20. Kamishin, V.V., Reva, O.M.: Metodi sistemnogo analizu u kvalimetriї navchalno-vixovnogo procesu [tekst]: monografiya. K.: TOV « Informacijni sistemi » , 270 p. (2012) 21. Kamyshin, V.V.: Teoretiko-metodologichni osnovi sistemno-informacijnoї kvalimetriї v upravlinni navchalno-vixovnim procesom [tekst]: avtoref. dis… doktora pedagogichnix nauk. specialnist 13.00.06 – teo-riya ta metodika upravlinnya osvitoyu. - K., 39 s (2014) 22. Organizaciya vozdushnogo dvizheniya: Pra-vila aeronavigacionnogo obsluzhivaniya [tekst]: doc. ICAO 4444 – ATM/501. – Izd-e pyatnadcatoe. Monreal, Kanada (2007) 23. Igoshin, V.I.: Matematicheskaya logika i teoriya algoritmov [tekst] 2-e izd., ster. M.: IC « Akademiya » , 448 s (2008) 24. Kormen, T.X., Lejzerson, Ch.I., Rivest, R.L., Shtajn, K.: Algoritmy: postroenie i analiz [tekst]. 3-e izdanie = Introduction to algorithms 3rd (edn.) - M.: Vilyams, 1328 s (2013) 25. Kartamyshev P.V., Ignatovich, M.V., Orkin, A.I.: Metodika letnogo obu-cheniya [tekst]. M.: Transport, 278 s (1987) 26. Gander, D.V. Professionalnaya psixope-dagogika [tekst] M.: « Voentexinizdat » , 336 s (2007) 27. Borsuk, S.P., Reva, O.M.: Nechetkaya ocenka riska ne-stykovki blokov « chelovek procedury » modeli SHELL IKAO [tekst]. Intelektualni sistemi prijnyattya rishen i problemi obchislyuvalnogo intelektu: m-li Mizhnar. nauk. konf., prisvyachenoї pam’yati profesora F.B. Rogalskogo, - s. Zaliznij Port Khersonskoї obl., 28–31 travnya 2014 r., XNTU, Kherson, S. 153–155 (2014) 28. Borsuk, S.P.: Fuzzy model of air traffic controller attitude to the risk during decision making. In: Reva, O.M., Borsuk, S.P. (eds.) Proceedings of the 5th International Conference on Applied Human Factors and Ergonomics AHFE 2014, Kraków, Poland, pp. 6229–6238 19– 23 July 2014 29. Reva, A.N., Svirkin, V.A.: Formirovanie letnyx navykov kursantov na trenazhere s variruemymi dinamicheskimi xarakteristikami. Metody ergonomicheskoj ekspluatacii i obsluhivaniya aviacionnoi texniki: sb. nauch. tr. K.: KIIGA, S. 53–59 (1986) 30. Dushkov, B.A., Lomov, B.F., Rubaxin, V.F. [i dr.]: Osnovy inzhenernoj psixologii: ucheb. dlya vuzov [tekst]. pod red. Lomova, B.F. M.: Vysshaya shkola, 448 s (1986) 31. Reva, A.N., Goryachev, V.A., Kuznecov, V.A. [i dr.]: Optimizaciya deyatelnosti instruktora aviacionnogo trenazhera: nauchno-prakticheskie re-komendacii [tekst]. pod red. Bodrova, V. A., Revy, A.N. IPAN USSR, M. 126 s (1990) 32. Reva, A.N., Komarov, A.A., Ku-z-necov, V.A. [i dr.]: Ergonomicheskie metody i sredstva trenazhernoj podgotovki letnogo sostava: nauch.-prakt. rekom. [tekst]. pod red. Revy, A.N., Rubca, M.I. GLAU, Kirovograd, 106 s (1995) 33. Shulgin, V.A.: Kompleksna ocinka efektivnosti profesijnoї diyalnosti instruktora aviacijnogo trenazhera [tekst] Naukovij visnik Khrsonskoї derzhavnoї morskoї akademiї: nauk. zh. vyd-vo KhDMA, Khrson, № 1, S. 302–308 (2013)

Ergonomic Assessment of Instructors’ Capability

793

34. Spravochnik po inzhenernoj psixologii [tekst]. pod red. Lomova, B.F., Mashinostroenie, M, 368 s (1982) 35. Shibanov, G.P. Kolichestvennaya ocenka deyatelnosti cheloveka v sistemax « chelovektexnika » [tekst], Mashinostroenie, M, 263 s (1983) 36. Nevinicyn, A.N., Starchenko, I.V.: Algoritmicheskie modeli kak sredstvo issledovaniya faktornogo rezonansa i sootvet-stviya procedur deyatelnosti psixofizio-logicheskim vozmozhnostyam aviacionnyx operatorov [tekst]. In: Problemi aeronavigaciї: tematich. zb. nauk. pr. Kirovograd: DLAU, 1997. vip. III. Udoskonalennya procesiv diyalnosti ta profesijnoї pidgotovki aviacijnix operatoriv, S. 69–75

Impact of Plants in Isolation: The EDEN-ISS Human Factors Investigation in Antarctica Irene Lia Schlacht1(&), Harald Kolrep2, Schubert Daniel3, and Giorgio Musso4 1

HMKW-DLR, Berlin, Germany [email protected] 2 HMKW, Berlin, Germany [email protected] 3 DLR, EDEN Initiative, Bremen, Germany [email protected] 4 Thales Alenia Space, Turin, Italy [email protected]

Abstract. The EDEN-ISS is a greenhouse project at the Neumayer Station III in Antarctica. For the first time, this greenhouse supplied the station with fresh food and enabled research regarding sustainable and autonomous food production from Earth to Space. To investigate the plants’ impact on the crew (biophilia), a debriefing, questionnaires, and behavioral observation were used. The results show that the crew was satisfied with the consumption of fresh vegetables, which are usually not available in Antarctica. All (9 of 9 crew members) also agreed on the positive psychological and physiological impact of the plants on their well-being. The investigation will be repeated with the next crew of the Neumayer Station III and will also be proposed for comparison at stations like Concordia. Keywords: Human Factors
Human-systems integration
Space mission Habitability
Psychology
Isolation plants
Interaction
Biophilia




1 Summary In isolation, far away from the natural human conditions and Nature itself [1, 2] people may suffer from asthenia and depression. This effect has emerged in many isolation conditions, being experienced by astronauts inside the International Space Station (ISS), prisoners, military personnel in submarines, as well as scientists at Antarctic stations, but also in more common conditions such as in nursing homes or prisons [3]. As explained in the concept of biophilia the interaction with plants appears to have a positive effect on motivation and performance; however, further research is needed to demonstrate its relationship with performance in long-duration isolation [4–6]. For these reasons, Human Factors and psychological studies of the effects of a greenhouse on the psychological well-being of crews living and working in isolation are extremely relevant. © Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 794–806, 2020. https://doi.org/10.1007/978-3-030-20503-4_71

Impact of Plants in Isolation: The EDEN-ISS Human Factors Investigation

795

From December 2017 to February 2019, a group of ten scientists has been living at the Neumayer Station III (NMIII) in Antarctica surrounded by ice and isolated from the normal variety of life forms. During the mission, the crew built and tested the EDEN ISS greenhouse of the German Aerospace Center (DLR). For the first time, this greenhouse supplied the station with fresh food and enabled research regarding sustainable and autonomous food production from Earth to Space [7–9]. Various instruments were applied during the mission to investigate whether in isolation, interaction with plants might have a positive impact from a nutritional and psychological perspective, increasing performance and safety as in the Biophilia perspective [12]: 1. The “Human Factors debriefing” as a guided group discussion 2. Questionnaires: 2:1 Robert Koch Institute Food Frequency Questionnaire 2:2 POMS (Profile of Mood States) 2:3 Dedicated questionnaire on the interaction with the plants [10] 3. Observation (e.g. interviews, recording of time spent in the greenhouse). Meaningful instruments were the analysis of individual items from the questionnaire, interviews, and group discussions. In the questionnaire specifically, 7 of 7 subjects stated that they were satisfied with the consumption of raw vegetables. Moreover, everyone rated as “quite a lot” at least one of the positive effects of the plants listed. Finally, in the group discussion, the crew unanimously (9 of 9) agreed on the positive psychological and physiological impact of the plants on their well-being. The investigation will be repeated with the next crew of the Neumayer Station III and will also be proposed at stations like Concordia in order to have a comparison crew without a greenhouse. The aim of this paper is to present the results of the EDEN ISS investigation on the impact of interaction with plants during long-term missions on the mood of the crew members, on their performance, and generally on crew cohesion from a Human Factors perspective.

2 Introduction The relevance of the connection between Nature and humans has been the topic of much research found in the literature, in particular from the 1960s when it was associated with the term ‘biophilia’ [11]. Biophilia refers to the desire for a (re)connection with natural life and natural systems1[12]. The healing power of a connection with Nature was established in the 1980s with a study comparing the recovery rates of patients with and without a view of Nature [13]. In the 1990s, experiments showed an increase in productivity when building occupants were connected through biophilic design [14].

1

We should be genetically predisposed to prefer certain types of Nature and natural scenery, specifically the savanna (6. Savanna Hypothesis, 1986)

796

I. L. Schlacht et al.

During the same period, architecture groups started to apply a biophilic approach, confirming the correlation between improved environmental quality and worker productivity [15]. While the financial gains due to productivity improvements were considered significant, productivity was identified as a placeholder for health and wellbeing, which have an even broader impact. In Space and Antarctica as well as in other extreme, dangerous, and isolated environments, where a person’s productivity and reliability may impact the safety of human lives, the concept of biophilia may be of great importance [13].

3 EDEN ISS The ability to grow food and other essential resources for humans through biological processes is a major aspect for Space missions, as it helps to: – decrease the resupply mass currently required [16]; – allows astronauts to travel further and stay longer in Space [17]; – increases human safety, performance, and well-being as a result of biophilia [8, 9, 18]. In this context, DLR with decided in 2011 to develop a project together with a consortium of partners: the EDEN ISS (Evolution and Design of Environmentallyclosed Nutrition-sources) greenhouse project [19]. EDEN ISS aims to develop and test plant cultivation systems in isolation in terms of: – technologies and processes for ISS, planetary habitats, Earth application – study of microbial behavior and countermeasures – research of the physical and psychological impact, including oxygen production; enrichment of the diet with fresh food; and psychological effects on safety, performance, and well-being. In the context of this project, the EDEN consortium has developed the Spaceanalog Mobile Test Facility (MTF), a Bio-regenerative Life Support System (BLSS) for Space and Earth application, which has been tested at the German Neumayer Station III in Antarctica [16] (Figs. 1, 2 and 3).

Fig. 1. EDEN laboratory at the DLR Institute of Space Systems in Bremen

Impact of Plants in Isolation: The EDEN-ISS Human Factors Investigation

797

Fig. 2. The MTF with in the background the Neumayer Station III in Antarctica (© DLR)

Fig. 3. The MTF 3D structure on the left and the inside of the greenhouse on the right (© DLR)

The MTF It is composed of 2 connected containers and contains: the Cold Porch (airlock entrance from the outside (−35 °C) to the inside (+20 °C)), the Service Section (laboratory for controlling, processing, and testing the plants), and the Greenhouse (aeroponic cultivation of 13 m2 in shelves and trays associated with the bio-detection and decontamination system that ensures food quality and safety) [20]. (© Liquifier). 3.1

Project Challenges

The Greenhouse MTF was first built and tested in Bremen, Germany, and then transported by truck to Bremerhaven and loaded onto the Polarstern research vessel. After four weeks of travel to Cape Town, South Africa, it was loaded with equipment and traveled for ten days to the vicinity of the Neumayer Station III on an icebreaker vessel. Finally, it was loaded onto a sled and pulled for two hours over the ice by a PistenBully to the Neumayer III station, where it was raised by a crane onto its elevated platform (Fig. 4).

Fig. 4. EDEN ISS container transportation sequence

798

I. L. Schlacht et al.

In order to be transported, the MTF had to be able to face complex system capabilities and logistics to be settled in Antarctica, such as: restricted access to Antarctic stations only during five months per year via aircraft; extremely warm (equatorial regions) and cold (Antarctic) zones traversed during the journey; fulfillment of the Container Safety Convention (CSC) requirements; shipment transportation dynamics (location of sensitive electronics in order to deal with significant vibration and shock loads, cranes, exposure to humidity, …). 3.2

Successful Achievement

Finally, the EDEN ISS was shipped to Antarctica and assembled there in January 2018, 400 m south of the Neumayer Station III. From the station it can be reached on foot along a secured trail with a railing, which assures safe passage, particularly in the dark winter or during a whiteout. After the greenhouse had been assembled and all the equipment and all the systems had been tested, the greenhouse went fully productive. After about half a year, it was producing a harvest of 7–8 kg per week (Fig. 5).

Fig. 5. EDEN ISS harvest 15 Feb.–1 Sep. 2018

4 Analysis of the Impact of Plants on the EDEN ISS Crew To analyze the impact of plants in terms of fresh food production and consumption on performance and well-being, different tools were selected to be used by the Neumayer Station III 2017–2018 crew during the year of isolation in conjunction with the test of EDEN-ISS (Fig. 6): 1. The “Human Factors debriefing” as a guided group discussion about the crew’s human factors interaction with the plants. The goal was to collect in an open manner collective ideas/opinions; weaknesses/strengths/problems experienced and solutions found; and lessons learned during the mission regarding interaction with the plants from the perspective of the crew.

Impact of Plants in Isolation: The EDEN-ISS Human Factors Investigation

799

2. Questionnaires: 2:1 Robert Koch Institute Food Frequency Questionnaire 2:2 POMS (Profile of Mood States) for assessing the changes in morale the crew might exhibit while consuming fresh food. 2:3 Dedicated questions on the interaction with the plants developed on the basis of previous investigations [10] to assess the wishes and needs of the participants regarding interaction with plants as well as the frequency, quality, and number of individual activities with the greenhouse and the plants. 3. Behavioral observation and interviews, including a record of the amount of time spent in the greenhouse by the crew, to identify the impact of the plants on the mood and performance of the people who used the greenhouse often versus those who used it less often. Final interviews were conducted after the end of the mission.

Fig. 6. Neumayer Station III team 2018

These tools aimed to assess the effects of the interaction with plants during longterm missions on the mood of the crew members, on their performance, and generally on crew cohesion from a psychological and Human Factors perspective (Fig. 7).

Fig. 7. The crew in October 2018, after ¾ of the mission, performing the Debriefing. It is interesting to note the longer beards like at the end of some long-duration Space missions.

The planning of the experiment started in 2015. In order to acquire comparison data, the investigation was also planned to be performed with the winter crew 2016–2017. The times selected for the data collection were discussed with the consortium of partners to identify the main investigation events and their interactions with the crew (Fig. 13).

800

I. L. Schlacht et al.

Of the six proposed iterations, only three sets were approved by the consortium and after the ethical review: one before the mission and two (in June and October) during the mission. The time frame for the investigation was selected to be during the period of full vegetable production and outside possible highlights or important events that might influence the data collection on interaction with the plants (Fig. 8). Crew: Crew 2016–2017 without greenhouse -> control and comparison data Crew 2018 with greenhouse Time Schedule: • Before the mission, as reference baseline data -> 10 October • Middle of the mission -> 10 June • End of the mission -> 10 October

Fig. 8. Parameters for the selection of the time frame for performing the investigation.

4.1

Problems Encountered and Solutions

Some difficulties were encountered in the data collection, but solutions were found and adopted. These included, in particular: The investigations were performed only by crew 2018 during the mission in June and October. All the crew members had contact with the plants at least as fresh vegetables for food consumption. All the crew members were free to work or interact with the plants in the greenhouse. In order to enable comparison with a crew without interaction with plants, the investigation will also be proposed to the Concordia Station Crew 2020 (Table 1).

Table 1. Problems and solutions during the investigation Problem Missing Crew 2016–2017 data collection without greenhouse -> control and comparison data

Cause The crew was not informed. Prior to the mission, no crew member had been clearly identified as the person who would coordinate the investigation

Solutions Data collection performed on: - Crew 2019 - Concordia Crew 2020 (without interaction with plants) (continued)

Impact of Plants in Isolation: The EDEN-ISS Human Factors Investigation

801

Table 1. (continued) Problem Missing Crew 2018 baseline data

Cause General difficulties collecting data before and after the mission as well as lack of a time schedule with open access to all the consortium members where they could implement the tasks In the POMS, the items were not As in each questionnaire, a few filled out completely questions had not been answered, the data could only be compared at the level of individual question items; a standardized factors analysis was not possible

4.2

Solutions Data collection on: - Crew 2018 post-mission meeting with interviews

- Explanation of the importance of filling out all the items - Sensibilization of the crew regarding the importance of data collection

Analysis of the Results

Meaningful instruments were the analysis of individual items from the questionnaire, interviews, and group discussions. On the food frequency questionnaire, in particular, 7 of 7 subjects stated that they were satisfied with the consumption of raw vegetables. Moreover, on the plants interactions questionnaire everyone rated as “quite a lot” at least one of the positive effects of the plants listed. Finally, in the group discussion, the crew unanimously (9 of 9) agreed on the positive psychological and physiological impact of the plants on their well-being. Table 2. Debriefing main results of Crew 2018 Factor

Strength/Weakness

Vote

Psychological Psychological Psychological Physiological

+ + + +

9/9 8/9 8/9 9/9

Operational

−

8/9

Operational

−

8/9

Operational

−

6/9

Operational

−

6/9

SocioCultural

−

7/9

Description of most voted matters Fresh vegetables to eat Natural colors Observing, living, growing Fresh vegetables, valuable nutrition Frequent system malfunctions Alarm sounds at NMIII too frequent and annoying Cameras make the greenhouse less comfortable to relax in Too much lettuce/leafy greens Interaction with plants limited to only a few people (except consummation)

Impact Well-being Well-being Well-being Well-being Performance Well-being Well-being

Performance Well-being

802

I. L. Schlacht et al.

Breakdown of the results: 1. The “Human Factors debriefing” was performed in October 2018. As a result of the guided discussion, the crew reported that the positive impact on their well-being was the most important element of their interaction with the plants. Moreover, also other values were described such as: “relaxing”, “living things”, nice “smell”. Also elements that need improvements were described such as: “the variety of herbs”, “variability”, “partly overripe”, “it would be great to have it closer”, “too far” (Table 2). The psychological factor, in particular, was the factor associated most with interaction with the plants compared to the physiological, operational, socio-cultural, and environmental factors. As for the impact on safety, performance, and well-being, the latter was the one most frequently selected (Fig. 9).

Fig. 9. Debriefing results showing the effect of interaction with plants.

1. Questionnaire The Robert Koch Institute Food Frequency Questionnaire showed that in July as well as in October, all the subjects were satisfied with the consumption of fresh vegetables. The consumption of fresh vegetables is normally not possible during such missions. Here, it was already possible two months after the start of the mission and until the end of the mission after 14 months (Fig. 10). The POMS factors analysis can only be performed after the collection of data from the next investigations because only 2 subjects of 10 fill in completed the questionnaire 2 times. Considering the strength of the feelings selected by the subjects (“a lot” weighted 0,5 & “extremely” weighted 1), it appears that the crew had the tendency to select lower feelings strength at the start of the mission and higher ones at the end of the mission (Fig. 11), however this tendency need to be confirmed with more subjects.

Impact of Plants in Isolation: The EDEN-ISS Human Factors Investigation

803

Fig. 10. Questionnaires results in June 2018 (left) and October 2018 (right) compare the level of satisfaction of the crew with the supply of the raw vegetables (provided only by the greenhouse system) in comparison with other food like fruits that need to be stored for long time(e.g. frozen).

Fig. 11. Weight of the feeling selected on the POMS questionnaire on June and October 2018.

The dedicated questionnaire on the interaction with the plants required the crew members to evaluate whether the plants may have a recreational and nutritive value as well whether the interaction with the plants may positively impact their well-being, their motivation as well as crew relations. The results show that both in June 2018 and October 2018, the majority of the crew considered the nutritive value as well the impact on their well-being very important (Fig. 12). This confirms both the results of the debriefing and those of the Robert Koch Institute Food Frequency Questionnaire.

804

I. L. Schlacht et al.

Fig. 12. Evaluation of the factors of impact on the interaction with plants.

Fig. 13. EDEN ISS partners

5 Conclusion The EDEN ISS project is the first European bio-regenerative system tested in Antarctica. Although the cost of designing, deploying, and running Antarctic hydroponic facilities presently outweighs their return in terms of psychological and nutritional benefit, they make an important contribution to the advancement of international research on bio-regenerative systems both for the exploration of the Universe and for Earth applications [21]. Particularly, during the Crew 2018 mission, the psychological impact of the plants on their well-being was assessed as positive by all crew members as also foreseen by the biophilia concept. The investigation will be repeated with the next crew of the Neumayer Station III and will also be proposed at stations like Concordia in order to have a comparison crew without a greenhouse.

Impact of Plants in Isolation: The EDEN-ISS Human Factors Investigation

805

Acknowledgments. A sincere thanks to the Neumayer station III crew that performed this investigation. The EDEN ISS project has received funding from the European Horizon 2020 Programme within the topic of ‘Space exploration/Life support’ under grant agreement number 636501. Partners: Deutsches Zentrum fuer Luft- und Raumfahrt (DLR), Germany (specifically Dr. Johannes for the psychology investigation); LIQUIFER Systems Group (LSG), Austria; Consiglio Nazionale Delle Ricerche (CNR), Italy; University of Guelph, Canada; AlfredWegener-Institut Helmholtz-Zentrum fuer Polar- und Meeresforschung (AWI), Germany; Enginsoft Spa (ES), Italy; Airbus Defense and Space, Germany; Thales Alenia Space Italia Spa, Italy; Aero Sekur S.p.A., Italy; Stichting Dienst Landbouwkundig Onderzoek (DLO), The Netherlands; Heliospectra AB, Sweden; Limerick Institute of Technology (LIT), Ireland; Telespazio SPA, Italy; University of Florida, USA; Extreme-Design.eu research group, Europe. The Human Factors investigation was sponsored by DAAD and performed by Dr. Irene Schlacht in cooperation with the project coordinator, Daniel Schubert, and Prof. Harald Kolrep, head of the Faculty of Psychology at Berlin HMKW University of Applied Sciences for Media, Communication and Management.

References 1. Schlacht, I.L.: Space habitability: integrating human factors into the design process to enhance habitability in long duration mission. Doctoral Dissertation, Technische Universität Berlin, Germany (2012). http://www.extreme-design.eu/publications.html 2. Schlacht, I.L.: Habitability. In: Sgobba, T., Kanki, B.K., Clervoy, J., Sandal, G. (eds.) Space Safety and Human Performance, 1st edn., pp. 655–664. Butterworth-Heinemann, Elsevier (2017). http://www.extreme-design.eu/publications.html 3. Kanas, N., Manzey, D.: Space Psychology and Psychiatry. Springer, Heidelberg (2008) 4. Bates, S.C., Marquit, J.D.: Space psychology: natural elements in habitation design. Pers. Ubiquit. Comput. 15, 519–523 (2010) 5. Ulrich, R.S., Simons, R.F., Losito, B.D., Fiorito, E., Miles, M.A., Zelson, M.: Stress recovery during exposure to natural and urban environments. J. Environ. Psychol. 11, 201– 230 (1991) 6. Bates, S.C., Gushin, V.I., Marquit, J.D., Bingham, G., Sychev, V.: Plants as countermeasures in long-duration space missions: a review of the literature and research strategy. In: Presented at the European Space Agency Workshop: Tools for Psychological Support (2007) 7. Schlacht, I.L., Bernini, J., Schubert, D., Ceppi, G., Montanari, C., Imhof, B., Waclavicek, R., Foing, B.: EDEN-ISS: human factors and sustainability for space and Earth analogue. In: 67th International Astronautical Congress 2016, Guadalajara, Mexico (2016). http://www. extreme-design.eu/doc/2016IAC-16.E5.1.7.x35557.pdf 8. Schlacht, I., Ceppi, G., Nazir, S.: Sustainable quality: from space stations to everyday contexts on Earth. In: Fostervold, K.I., Johnsen, S.Å.K., Rydstedt, L., Watten, R.G. (eds.) Creating Sustainable Work Environments, NES 2015, 1–4 of November, Lysaker, Lillehammer, pp. A1-5–A1-7. Norwegian Society for Ergonomics and Human Factors, Norway (2015). http://www.extreme-design.eu/publications.html 9. Quantius, D., Schubert, D., Maiwald, V., París Lopéz, R., Hauslage, J., Seboldt, W., Doule, O., Schlacht, I.L., Ransom, S.: Initial design of laboratories for sustainable habitation. Acta Astronautica 94, 541–562 (2014) 10. Mauerer, M., Schubert, D., Zabel, P., et al.: Initial survey on fresh fruit and vegetable preferences of Neumayer Station crew members: input to crop selection and psychological benefits of space-based plant production systems. Open Agric. 1(1) (2016). https://doi.org/ 10.1515/opag-2016-0023. Accessed 9 Sept 2018

806

I. L. Schlacht et al.

11. Fromm, E.: The Heart of Man. Harper & Row, Manhattan (1964) 12. Wilson, E.O.: Biophilia. Harvard University Press, Cambridge (1984) 13. Terrapin: 14 Patterns of Biophilic Design. Improving Health & Well-Being in the Built Environment (2012). https://www.terrapinbrightgreen.com/reports/14-patterns/ 14. Heerwagen, J.H., Hase, B.: Building Biophilia: connecting people to nature in building design. US Green Building Council. Posted March 8 2001. Web, 9 July 2013 (2001) 15. Browning, W.D., Romm, J.J.: Greening the Building and the Bottom Line. Rocky Mountain Institute, Colorado (1994) 16. Schubert, D., et al.: Status of the EDEN ISS greenhouse after on-site installation in Antarctica. In: 48th International Conference on Environmental Systems, CES-2018-140, 8–12 July 2018, Albuquerque, New Mexico (2018) 17. Zabel, P., et al.: Introducing EDEN ISS - a European project on advancing plant cultivation technologies and operation. In: 45th ICES: International Conference on Environmental Systems, 12–16 July 2015, Bellevue, Washington (2015) 18. Schlacht, I.L., Foing, B., Bannova, O., Blok, F., Mangeot, A., Nebergall, K., Ono, A., Schubert, D., Kołodziejczyk, A.M.: Space analog survey: review of existing and new proposal of space habitats with Earth applications. In: Proceeding of the International Conference on Environmental Systems, ICES 2016. Published Online on the Texas Digital Library at (2016). https://ttu-ir.tdl.org/ttu-ir/handle/2346/67692 19. Schlacht, I. L., Bernini, J., Schubert, D., Ceppi, G., Montanari, C., Imhof, B., Waclavicek, R., Foing, B.: EDEN-ISS: human factors and sustainability for space and Earth analogue. In: 67th International Astronautical Congress 2016, Guadalajara, Mexico (2016). http://www.iafastro. net/download/congress/IAC-16/DVD/full/IAC-16/E5/1/manuscripts/IAC-16,E5,1,7,x35557. pdf 20. Imhof, B., Schlacht, I.L., Waclavicek, R., Schubert, D., Zeidler, C., Vrakking, V., Hoheneder, W., Hogle, M.: EDEN ISS – a simulation testbed to an advanced exploration design concept for a greenhouse for Moon and Mars. In: IAC-18.B3.7, International Astronautical Conference, 1–5 October 2018, Bremen (2018) 21. Bamsey, M.T., et al.: Early trade-offs and top-level requirement definition for Antarctic greenhouses. In: 46th International Conference on Environmental Systems, ICES 2016, 14 July 2016, Vienna, Austria (2016)

Considerations for Passenger Experience in Space Tourism Tiziano Bernard(&), Yash Mehta, Brandon Cuffie, Yassine Rayad, Sebastien Boulnois, and Lucas Stephane Human-Centered Design Institute, Florida Institute of Technology, 150 W. University Blvd., Melbourne, FL 32901, USA {tbernard2011,ymehta2016,bcuffie2018,yrayad2008, sboulnois}@my.fit.edu, [email protected]

Abstract. Space Tourism is a topic of ever-growing discussion as commercial space providers are closer to opening opportunities for aspiring spaceflight participants. The current efforts on defining requirements for commercial space flight crews and participants, in the United States, are mainly safety-based and take into consideration the minimization of risks both from an operational and regulatory perspective. There is however a need to open discussions on space passenger experience design and create a paradigm that covers this novel role. A design approach is outlined to identify areas of study that already attempt to address human factors aspects in astronautical applications. This paper employs the PEAR model to identify high-level passenger-environment interactions that ought to be considered within the context of tourism. Using Virgin Galactic’s concept of operations as a baseline, the model is used to gather the human factors that appear influential in passenger experience as well as methods to evaluate them. Keywords: Space tourism
Commercial space
Passenger experience
Human factors in transportation
Spacecraft operations
Astronautics
Aerospace design
Human-Centered design

1 Introduction: The New Frontier of Space Tourism 1.1

Concept

Passenger/tourist experience (PX) during suborbital flights is an emerging complex topic that needs to be addressed from a human-centered perspective. Up to recently, humans in spaceflight belonged to the professional category of selected and trained astronauts. While a pre-selection is going to be set up for suborbital spaceflight tourists, current objectives of suborbital spaceflight companies are to open suborbital space tourism to a growing number of people. The current paper presents state of the art on needfinding and requirements for suborbital spaceflight tourists, as well as the Design Thinking framework and the people, environment, actions and resources (PEAR) model [1].

© Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 807–818, 2020. https://doi.org/10.1007/978-3-030-20503-4_72

808

1.2

T. Bernard et al.

Purpose

The main purpose of this early stage research is to provide a novel framework based upon aerospace expertise that will support human factors practitioners and experience designers to be actively involved for delivering the best suborbital spaceflight configurations to future tourists. 1.3

Approach

The encompassing framework for addressing the suborbital space tourist experience is Design Thinking [2]. The Design Thinking (DT) cycle is composed of five stages that are: (1) problem identification and statement; (2) need-finding, requirements or empathizing; (3) ideation; (4) prototyping; (5) testing, learning or evaluation. Each of these stages has both multidisciplinary teams as well as dedicated methods. Furthermore, the DT process can be performed iteratively and incrementally in a non-linear way, e.g. skipping some stages and coming back to those according to research needs (e.g. it is possible to go directly from problem identification to need-finding and ideation and perform some iterations on these stages before pursuing with prototyping and evaluation; it is also possible to go directly from problem identification to prototyping and use an early stage prototype or wireframe for the need-finding stage, etc.) [3]. Another main benefit of DT is that it enables multidisciplinary teams to systematically trace and integrate a variety of methods for each DT stage and therefore to better tackle complex research such as spaceflight tourist experience. This early research paper deals mainly with the need-finding stage, i.e. the state of the art of suborbital tourist spaceflight, and with the ideation stage that proposes to use the people, environment, actions and resources model (PEAR) for capturing various characteristics and features of suborbital spaceflight tourists. Using scenario-based design, the tourist PEAR model was integrated with spaceflight phases in order to identify specific tourist experiences per spaceflight phase. The paper concludes on future research related to the prototyping and evaluation stages.

2 Design Considerations 2.1

Proposed Problem Statement

In an effort to design the suborbital space tourism passenger experience holistically, one must consider the human factors that come into play. There is a need to develop a framework that takes into consideration typical phases of the suborbital journey as well as potential in-cabin passenger related incidents in the analysis of the human factor aspects experienced by the passengers. Such a framework can provide valuable insight into various design efforts focused on making space tourism a tangible reality.

Considerations for Passenger Experience in Space Tourism

2.2

809

Need-Finding

Federal Regulations. Current regulations for commercial space travel are expressed in Title 14 of the Code of Federal Regulations Part 460: Human Space Flight Requirements [4]. Subpart B specifically refers to space flights by licensed operators that include “space flight participants” – or passengers. The regulations start by defining those who are licensed to operate these flights and those who wish to participate in them. Appendix E to Part 440 provides the Agreement for Waiver of Claims and Assumption of Responsibility in which the flight participants and related family are not allowed to enter claims against the United States government. Regulations also address training requirements for spaceflight participants in 460.51. The operators are required to train passengers on emergency situations including smoke, fire, loss of cabin pressure, and emergency evacuations. Finally, 460.53 requires operators to maintain a level of security in order to safeguard the lives of the other passengers, crew, and the general population. This includes limitations on carrying knives, explosives, and so on. It is clear that the regulations are designed as a baseline for future operations. Similarly to the development of airline operations, time will dictate further needs and details, as already shown by a reserved section (460.47). 14 CFR 460 does not however target any customer experience requirements, allowing researchers and designers room for research and considerations. Infrastructure. The technological evolution that defines the progress toward a feasible tourist experience is mainly present in the private industry sector. There are, however, a number of spaceports across the world that can provide logistical and operational support to commercial space operators. The global view of all active and proposed orbital and suborbital spaceports is illustrated in Fig. 1 [5]. In order to provide considerations on passenger experience, it is important to understand the type of artifacts that are currently available and how these vehicles can provide insight into the experience of spaceflight participants.

Fig. 1. Existing and Proposed Global Spaceport Map (FAA) [5]. The proposed spaceports support vertical takeoff (rocket launch), runway take-offs, or both.

810

T. Bernard et al.

With recent advancements in suborbital operations, reusable vehicle architectures, and an ever-evolving society, space tourism has the potential to become the new frontier. Companies and manufacturers such as Blue Origin, Virgin Galactic, and the Sierra Nevada Corporation have already successfully designed solutions that would not only address proper suborbital operations, but constructed architectures that would allow spaceflight participants (commercial passengers) the opportunity to visit the upper atmosphere and space as an actual “tourist”. Blue Origin’s New Shepard Capsule. Blue Origin’s efforts toward a commercial crew vehicle is the New Shepard [6, 7]. Named after Alan Shepard, the first American in space, the capsule is installed on a single-stage rocket (called propulsion module) that launches vertically reaching an apogee of 351,000 ft. During apogee the crew members will experience weightlessness (for both leisure and scientific research). At apogee the crew module separates from the propulsion module; the two systems subsequently land individually using parachutes and thrusters (similarly to SpaceX rockets). It can carry up to six passengers and has no pilot; it is controlled autonomously by the onboard computers without the need for any flight attendants or controllers. The overall flight time (take off to landing) is expected to be around 10 min. Virgin Galactic’s SpaceShip Two. Virgin Galactic uses a completely different flight profile from Blue Origin. The spaceplane, SpaceShip Two, is attached to a “carrier” vehicle, WhiteKnight Two, that flies with airbreathing engines to an altitude of 50,000 ft. The spaceplane is then released and uses rocket propulsion to accelerate reaching supersonically an altitude of about 360,000 ft. At apogee the SpaceShip Two wings feather and the crew experiences weightlessness. As the vehicle reenters the lower atmosphere the wings return to the original position in order to safely land like a typical airplane. The total flight time (takeoff to landings) is expected to be 2.5 h. It can carry 6 passengers and operates with two crew members (i.e. pilots) [8, 9]. Sierra Nevada Corporation’s Dream Chaser. Sierra Nevada Corporation (SNC) designed and built a spacecraft for crew transportation operations between Earth, the International Space Station (ISS), and back. The vehicle was not selected (SpaceX and Boeing were chosen instead) in the Commercial Crew Development program, but the company recently received clearance to be a cargo carrier to and from ISS. The vehicle was however designed to serve commercial operations and is well-fit to carry passengers. The vehicle fits strategically within the cargo space of a rocket and is deployed in orbit. The Dream Chaser then glides gently to Earth at a maximum acceleration of 1.5 g (as compared to the 3–4 g of Virgin Galactic). It can carry up to seven passengers and the pilot is optional. It is capable of operating and landing autonomously [10]. 2.3

Ideation

Scenario-Based Design: Phases of Flight. A baseline scenario was developed based on the assumptions of using Virgin Galactic SpaceShip Two flight profile. The figure below provides details for the scenario. The scenario is divided by phases of flight as

Considerations for Passenger Experience in Space Tourism

811

well as the associated passenger tasks and environmental factors affecting the passenger experience. The scenario was designed based on the information available in the open source domain (e.g. research publications and media). The passenger experience is heavily dependent on the flight dynamics and the related environmental factors - the flight phases with the greatest impact on passenger experiences are: (1) Ascent, (2) Microgravity, and (3) Descent. Passenger Profile. While a specific passenger profile was not provided to avoid creating constraints, passengers can be expected to above the age of 18, to legally provide consent, and physiologically healthy for both hyper and micro-gravity. McDonald et al. [13] believe that the average age of the passenger will be around 50 (though a wide range of individuals is expected) and in good health to withstand the gloads as well as perform emergency evacuation. Also, in 2006 the FAA Civil Aeromedical Institute in Oklahoma provided a series of guidelines for assessing the medical status of prospective space flight participants [11]. Flight Profile and Passenger Experience. During launch and ascent phase, passengers can experience 3–4 g and similar g-load during reentry [12, 13]. During this phase, passengers are limited in terms of tasks/activities they can perform but must be prepared to mitigate any adverse impact of the g-loads. Once having reached the microgravity coasting phase, passengers can move out of their seats and float/maneuver within the cabin for approximately four minutes [14]. During the microgravity phase, passengers have much wider selection of tasks/activities they can perform; the risks too are higher due to injury from unintentional maneuvers or unexpected physiological responses to microgravity. This is described below in Fig. 2.

Fig. 2. Phases of flight in the concept of operations

During the end of the microgravity phase, passengers must return to their seats and secure themselves. That particular window opens possibilities for confusion, disorientation, and incidents. Training provided prior to the flight would mitigate such occurrences. Subsequently, during the reentry phase, passenger tasks shift to those similar during ascent phase with limited interaction with the environment. Upon landing, the passengers’ primary task would be to safely egress out of the vehicle.

812

T. Bernard et al.

The PEAR Model. To frame passenger experiences with regards to human factors, the People, Environment, Actions and Resources (PEAR) model was used. The rationale to use the PEAR model is as follows: 1. Simplicity in implementation 2. Knowledge elicitation of factors encapsulating suborbital flight passenger experience 3. Ease of mapping interactions between different factors Table 1. Applied PEAR model Pre-flight readiness People Physical, characteristics Psycho, Physiological for this particular stage Environment High-low physical Lighting conditions Ambient Noise and vibrations

Environment Ground organizational Control Flight Crew Passengers Actions Passenger’s passengers fastened into their seats Passengers wearing IVA suits Follow preflight safety instructions Monitor flight information system Interact with flight information system Interaction with other passengers Resources Embedded passengers Smart Health Station, Passenger Wearable Suit

Release and launch

Microgravity coast

Re-entry and descent

Psychological, Psychological, Psychological, Physiological Physiological Physiological Psychosocial

Landing and return Physical, Psychological, Physiological Psychosocial

Vehicle acceleration High-low Lighting conditions High Ambient Noise and vibrations High-g (3–4 g) environment Flight Crew Passengers

Low Lighting conditions Micro-g environment

High-g (3–4 g) environment High noise and vibrations

Changing vehicle attitude Vehicle Deceleration

Flight Crew Passengers

Flight Crew Passengers

Ground Control Flight Crew Passengers

Monitor flight information system Interact with flight information system Observe and Monitor the environment

Unfasten seatbelt Float and maneuver inside the cabin Interact with other passengers Observe the external environment

Return to seats Fasten seatbelt Prepare for reentry – high-g Monitor environment

Observe and Monitor the Environment Interact with other passengers Unfasten seatbelt Vacate the cabin

Passenger Wearable Suit

Passenger Passenger Wearable Suit Wearable Suit

Embedded Smart Health Station Passenger Wearable Suit

Considerations for Passenger Experience in Space Tourism

813

A New Paradigm for Passenger Experience. Even though the main concerns of industry and government focus on safety and technology, the role of the passenger will undoubtedly become increasingly important. If the major intent of these space systems is space tourism, then there needs to be a paradigm for the spaceflight participant (or passenger) experience. The International Association for the Advancement of Space Safety (IAASS) provides guidelines for Space Flight Participants (SFP) that are focused mainly on safety aspects potentially affecting the vehicle, the crew, and the passengers. In addition, the IAASS also highlights the importance of the “quality of the service provided” as it relates to “compliance with functional, reliability, robustness or security expectations.” Considering the significance of safety in the design of future passenger space flight experiences, it is key to consider it as early as possible in the design stage [15]. Historically, as aircraft became more and more commercially available to the masses, greater considerations were made on customer experience. From the luxurious Boeing StratoCruisers operated by Pan American World Airways to the modern airliners of today, customer experience has changed drastically, requiring optimization and careful attention to safety and policies. The scientific community (especially the human factors one) has learned a lot from commercial airliner development, and many lessons can be easily applied onto touristic spacecraft. For example, the International Air Transport Association (IATA) provides areas of activity to airlines, airport operators, infotainment providers, government entities, etc., aimed at improving passenger experience. Figure 3 provides the structure for addressing passenger experience [16].

Fig. 3. IATA’s structure for addressing passenger experience

814

T. Bernard et al.

For decades now, space tourism has been a highly publicized concept. Multiple companies have claimed their intention to provide a sub-orbital space travel opportunity for leisure purposes. These announcements and associated initial investments, while encouraging, have often been followed by news of these companies failing to deliver on their commitments. Furthermore, the framing of the various space tourism options has been heavily focused on a luxurious conception of human space-flight described by a prospective Virgin Galactic customer as “the ultimate joyride of all time” [17]. Considering that touristic sub-orbital flights would be cost-prohibitive for the large majority of the population of the world, it is desirable to continue designing the passenger experience with a focus on potential wealthy prospective customers. A definition of passenger experience is not as trivial as the interactions with the environment in astronautical operations are extremely diverse and particular. The setting of all interactions, if assumed to be similar to the Virgin Galactic mission profile, can be relatable to the experience of flying in a business jet up until separation from the carrier aircraft, upon which the environment takes a more extreme nature, with visual, psychological, and physical cues that are not relatable to past experiences. In order to theorize about this new paradigm for passenger experience, we must seek to understand why humans have a desire to go to space. According to market research referenced by Collins, the main reason justifying the desire to travel to space for most people in the middle class of wealthy countries is “to be able to look back at the Earth” [18]. In addition, Peeters offers other reasons collected from more recent market surveys including “experiencing weightlessness”, “experiencing astronaut training”, and “communicating from space” [19]. Moreover, these considerations point to the importance of space cabin designs that are conducive to a pleasurable experience while accommodating for the limited astronaut-style training that would be received by passengers. Ultimately, the passenger experience in sub-orbital space tourism breaks with the traditional view of human space travel conditioned by historical aspects. Since the early days of humans in space, the framing has been extensively “mission-based” however nowadays a shift towards a more “journey-based” framing is becoming evident. Further structured user research is highly desirable as more potential and prospective sub-orbital flight passengers are identified in order to ensure that the design of experience is integrated in the greater philosophy of achieving sustainable commercial space tourism. Passenger-Environment Interaction. Grounded in the scenario-based design from the CONOPS model, and directly linked to the PEAR model, a classification of human factors principles, reflecting the Joint Cognitive System (JCS) model of the Passenger Environment Interaction (PEI) was created. In the current stage, the PEI describes all the interactions that take place for the passenger in a condensed and easily readable form. In this model the primary interactions occur between the passenger, the given resources, the environment and the actions that must be done and achieved to sustain the optimal passenger experience (Fig. 4).

Considerations for Passenger Experience in Space Tourism

815

Fig. 4. Passenger experience framework

As we take into consideration the design the passenger experience, it must be recognized that in order for the passenger to have the optimal experience, all the elements that comprise of the environment must work together with the user; neither can work independently. The environment, the resources and the passenger must be able to understand each other to work to the best benefit. This is akin to a joint cognitive system [20]. The passenger environment interaction is also a sociotechnical system, because of the interaction between social and technical factors that exist (Fig. 5).

Fig. 5. Simple cyclic model

Thus, the interactions between passengers and the spacecraft environment follow two main joint cognitive systems principles: (1) the optimization of system performance cannot be achieved by the optimization of the social or technical components in isolation; and (2) safety can be neither analyzed nor managed by considering only the system components and their failure probabilities [21]. As such, as illustrated in the PEAR Table 1 above, passenger resources required onboard are dependent on passenger characteristics, the environment and their actions. Resources identified so far range from tangible, such as the passenger wearable suit (PWS) and the Smart Health

816

T. Bernard et al.

Station (SHS), to less tangible such as the variety of passenger communications with the flight crew and among themselves. Actions are identified in terms of passenger allowed activities onboard. The environment is considered from both physical and organizational perspectives. While the figure above illustrates a simple cyclic model [22], the passenger environment interactions are identified in relationship with the different flight phases. Therefore, the passenger as well as all the other PEAR entities and parameters will be taken into account as interconnected for ensuring from the very early design stages the best passenger experience.

3 Conclusions and Future Work 3.1

Prototyping

Prototyping efforts will be carried out in an incremental and iterative approach for addressing various human factors examples identified as essential for the best passenger experience. This section of the design cycle is used during the testing phase of the design thinking process for revealing solutions to problems, generate new questions, and decide which solutions should be furtherly implemented [2]. For example, cognitive tunneling is an experience that can occur during the microgravity coasting. When given the instruction to unfasten seat belts and experience the microgravity environment, passengers may become lost in internal thoughts or external stimuli, e.g. instrumentation or the outside view. It has already been shown that pilots experience cognitive tunneling on their head-up displays (HUD) resulting in decreased performance that requires continuous monitoring of information outside the HUD [23]. The result is that they may lose focus on their current actions or environment, resulting in a degraded onboard experience. As with many other topics, further research into cognitive tunneling for space tourism passengers using a HUD will be essential for contributing to the optimal design of passenger experience. 3.2

Evaluation

As mentioned in the introduction, the current paper focuses mainly on the need-finding and ideation stages of the design thinking cycle. Therefore, human factors, usability, user experience and usefulness methods and tools, as well as specific interventions by human factors teams per tourist space flight phase were not developed in this paper. However, research efforts are ongoing for consolidating previous human-centered aerospace research carried out by the authors. Indeed, the aim of the proposed Passenger Environment Interaction model and overall Design Thinking framework is to ensure continuity with a variety of topics that were already studied, such as the Enhanced Space Navigation and Orientation Suit equipped with haptic feedback actuators, and the Human Biosensing and Monitoring suite sponsored by NASA FSGC [24, 25]. Knowledge elicitation from these previous studies is in the process to being adapted specifically to tourist spaceflight. Ongoing research for the evaluation stage encompasses the following:

Considerations for Passenger Experience in Space Tourism

• • • • •

3.3

817

Identifying best candidate HF, UX and usefulness methods per spaceflight phase Identifying dedicated HF equipment transversal to all flight phases Identify feasibility of using terrestrial equipment in spaceflight Identify intrusiveness and acceptability of such HF equipment Set up an integrated environment for both sensory-enhanced space tourist suits and space simulators. General Conclusion

The state of the art performed in this paper enabled to establish a foundation necessary for future research related to tourist experience in spaceflight. At the same time, it was beneficial to identify and select the main approaches that will support future research. Indeed, the Design Thinking framework enables multidisciplinary teams to systematically trace and integrate a variety of methods for each stage (i.e. need-finding, ideation, prototyping and evaluation), and therefore helps to deal with the complexity of spaceflight tourism from the overall tourist experience. Furthermore, the people, environment, actions and resources (PEAR) model enables to systematically capture various characteristics and features to be integrated for each stakeholder category, i.e. passenger, flight crew, ground control and human factors team. While only the passenger PEAR model was presented in this paper, the final goal of this research is to refine PEAR models for each stakeholder category, and then integrate them for providing the optimal organizational configuration ensuring the output of the best experience for space tourists. Last but not least, from a design thinking process perspective, the goal of the current research effort is to go beyond a descriptive human factors and user experience approach, e.g. what methods and tools to use, toward an active engagement of human factors and experience design teams, e.g. how human-centered practitioners will be optimally involved in emerging space tourism.

References 1. Johnson, W.B., Maddox, M.E.: A model to explain human factors in aviation maintenance (2007). https://www.aea.net/AvionicsNews/ANArchives/April07HumanFactors.pdf 2. Meinel, C., Leifer, L.: Design thinking research. In: Plattner. H., Meinel, C., Leifer, L. (eds.) Design Thinking: Understand-Improve-Apply, pp. xiii–xxi. Springer, Berlin (2011) 3. Boulnois, S., Stephane, L.: Human-centered design of a 3D-augmented strategic weather management system: first design loops. In: Advances in Intelligent Systems and Computing Proceedings of the 20th Congress of the International Ergonomics Association, IEA 2018, pp. 555–575 (2018) 4. Electronic Code of Federal Regulations: 14 CFR 460. https://gov.ecfr.io/cgi-bin/text-idx? SID=bb4a10d55aee1159eda311744287c6f4&mc=true&node=pt14.4.460&rgn=div5 5. Shelton-Mur, K.: Commercial space transportation overview. https://slideplayer.com/slide/ 7339577/ 6. Wagner, E., DeForest, C.E.: Opportunities for suborbital space and atmospheric research facilities on Blue Origin’s New Shepard Crew Capsule. In: AGU Fall Meeting Abstracts (2016) 7. Blue Origin: New Shepard. https://www.blueorigin.com/new-shepard/

818

T. Bernard et al.

8. Virgin Galactic: Mission. https://www.virgingalactic.com/mission/ 9. Seedhouse, E.: Virgin Galactic: The First Ten Years. Springer, Cham (2015) 10. Sierra Nevada Corporation: Dream Chaser Spacecraft. https://www.sncorp.com/what-we-do/ dream-chaser-space-vehicle/ 11. Antuñano Melchor, J., et al.: Guidance for medical screening of commercial aerospace passengers (2006) 12. Mehta, Y.B., Doule, O.: Biomedical monitoring of spaceflight participants during suborbital flights via agile architecture. In: AIAA SPACE and Astronautics Forum and Exposition, p. 5268 (2017) 13. McDonald, P.V., et al.: AST commercial human space flight participant biomedical data collection. Final Report: DOT Volpe Center Contract DTRT57 (2007) 14. Gohd, C.: Virgin galactic has launched its SpaceShipTwo into space. http://www.astronomy. com/news/2018/12/virgin-galactic-has-launched-its-spaceshiptwo-into-space 15. Suborbital Safety Technical Committee Guidelines Manual. https://www.faa.gov/about/ office_org/headquarters_offices/ast/advisory_committee/meeting_news/media/2014/may/ 16_IAASS_SuborbitalSafetyTC_COMSTAC_5-07-2014.pdf 16. IATA: Passenger Experience. https://www.iata.org/about/sp/Documents/passengerexperience.pdf 17. Billings, L.: Exploration for the masses? Or joyrides for the ultra-rich? Prospects for space tourism. Space Policy 22, 162–164 (2006) 18. Collins, P.: Meeting the needs of the new millennium: passenger space travel and world economic growth. Space Policy 18(3), 183–197 (2002) 19. Peeters, W.: From suborbital space tourism to commercial personal spaceflight. Acta Astronaut. 66(11–12), 1625–1632 (2010) 20. Woods, D.D., Hollnagel, E.: Joint Cognitive Systems: Patterns in Cognitive Systems Engineering. CRC, Taylor & Francis, Boca Raton (2006) 21. Hollnagel, E.: The human in control: modelling what goes right versus modelling what goes wrong. In: Human Modelling in Assisted Transportation, pp. 3–7 (2011) 22. Son, C., et al.: Modeling an incident management team as a joint cognitive system. J. Loss Prev. Process Ind. 56, 231–241 (2018) 23. Jarmasz, J., et al.: Object-based attention and cognitive tunneling. J. Exp. Psychol.: Appl. 11 (1), 3–12 (2005) 24. Bernard, T., Gonzalez, A., Miale, V., Vangara, K., Stephane, L., Scott, W.E.: Haptic feedback astronaut suit for mitigating extra-vehicular activity spatial disorientation. In: AIAA SPACE and Astronautics Forum and Exposition, p. 5113 (2017) 25. Cuffie, B., et al.: Proposed architecture of a sensory enhanced suit for space applications. In: 2018 AIAA SPACE and Astronautics Forum and Exposition (2018)

Cognitive Architecture Based Mental Workload Evaluation for Spatial Fine Manual Control Task Yanfei Liu1(&), Zhiqiang Tian2, Yuzhou Liu3, Jusong Li1, and Feng Fu1 1 Department of Computer Science and Technology, Zhejiang Sci-Tech University, No. 928, 2nd Avenue, Xiasha, Hangzhou 310018, China [email protected], [email protected], [email protected] 2 China Institute of Marine Technology and Economic, No. 70, Xueyuan South Road, Haidian District, Beijing 100081, China [email protected] 3 Department of Computer Science and Software Engineering, Auburn University, Auburn, AL 36849, USA [email protected]

Abstract. Based on cognitive extent, this paper focuses on workload evaluation for spatial fine-grained tracking control tasks. Cognitive model for manual rendezvous and docking (RvD) control task is setup in the light of cognitive architecture firstly. Then, total active time for each module in cognitive architecture is calculated to represent the active time for corresponding brain region. Workload predicted by both the NASA-TLX subjective scale method and proposals are compared to verify the evaluation’s validation on a cognitive degree. Finally, mapping the corresponding activities of the cognitive model to the human brain functional related area and making the brain cortex region’s activity animation with time of model’s running, the simulation for mental workload of R&D manual task is implemented. The results show that evaluation of human brain workload from a cognitive level is more effective, objective and accurate than traditional scales and physiological measurement methods. Keywords: Cognitive architecture
Cognitive modeling evaluation
Simulation
Space fine manual control




Mental workload

1 Introduction Human performance is considered as a measure of human functions and actions under some specific conditions. According to recent research, more and more modern commercial and industrial organizations nowadays want to develop better methods for assessing human performance rather than simply using performance measures such as efficiency and effectiveness [1].

© Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 819–829, 2020. https://doi.org/10.1007/978-3-030-20503-4_73

820

Y. Liu et al.

Space exploration is rapidly developing nowadays ever than before. However, for the astronauts, they usually expose to numerous stressors during spaceflights, such as microgravity, confinement, and radiation, which impair their cognitive capabilities. While some critical operations for spaceflight, such as handling robotic arms, extravehicular activities, and driving the spacecraft, etc., fault operation of them may cause serious disasters. The crewmembers’ cognitive abilities by no means affect task’s performance; therefor, essential operating skills should be assessed as a team member in an environment of highly dynamic, fast changing, and even unpredictable [2]. To improve crewmembers’ performance, researchers conducted many studies in these decades. However, due to limitations for experimental conditions, uncertainty and poor features of experimental results in study of human cognitive behavior for space exploration mission, experimental researches are rather difficult to implement in reality. Human performance technology, also known as human performance improvement, or human performance assessment, is a field of study related to process improvement methodologies. The models of its development and use have grown steadily since the successful application of servo-theory in the 1950s. However, problems and unresolved issues still restrict the utility and application in design and development. Although many different modeling approaches have been taken and a wide variety of limited models that focus on some particular aspect of human performance has been developed, the potential would increase for utility of these models if an integrated and practical representation of human performance developed [3]. The concept of mental workload has long been recognized as an important, albeit elusive, factor in human performance in complex systems [4]. Much research has been devoted to defining workload constructs and conducting careful measurements. However, the “holy grail” has not been found, and it has been argued that approaches to workload should be reconsidered [5]. For the restriction and deficiency of the studies on human mind, using a computer modeling and simulation method to investigate human cognition to improve performance becomes a new method of study on human factors in spaceflight tasks. Especially cognitive model based human performance exploration in complex systems is generally accepted. Zhang et al. pose time-fuel optimal control model to modeling human control strategies to simulated rendezvous docking tasks [6] and Liu et al. evaluate human performance by constructing assessment platform based on architecture [7]. Adaptive Control of Thought—Rational (ACT-R) is one of the most typical and widely used cognitive architecture, it has been used to successfully model a variety of behavioral phenomena and has proven particularly successful at modeling tasks with a demanding cognitive component [8]. Dozens of ACT-R models have been developed and empirically validated through an active user community in academic, industrial, and military research laboratories. Anderson introduced the Brain Mapping Hypothesis in ACT-R, which furthermore associates the modules to specific brain regions: “We have defined these regions once and for all and use them over and over again in predicting different experiments” [9]. Different modules of ACT-R perform specific functions during cognitive processes, which the function is abstraction when compared with the human neural network [10].

Cognitive Architecture Based Mental Workload Evaluation

821

This paper proposes a cognitive model based method for mental workload assessment and simulation for spatial fine manual control task in manual rendezvous and docking (RvD) mission.

2 Manual RvD Control Task and Cognitive Model A realistic cognitive model reflects human behavior in cognitive details stage. An ACT-R model typically consists of declarative knowledge procedural knowledge. Besides, it is also necessary to model the control on when to reason about what [11]. 2.1

Manual RvD Control Task

Space RvD refers to the process of two on orbits spacecraft combine into one joint object. Manual RvD means that astronaut operates the RvD process. Camera and sensor equipped on vehicles capture relative position and posture of the vehicles and the information are display on monitor. During the RvD process, the astronaut, the chasing vehicle and target vehicle form a human-in-loop system. According to discerning RvD vehicles’ relative status that displayed on the cabin’s monitor, and by controlling the chasing vehicle’s velocity and postures, the astronaut accomplish RvD control task. Figure 1 shows the diagram of human-in-loop manual RvD control process.

Fig. 1. Diagram of manual rendezvous and docking control loop.

There are two main types of operations for manual RvD task – position control and posture control. During manual RvD, the astronaut’s mental workload changes psychologically and physiologically. In the meantime, some operations must be satisfied

822

Y. Liu et al.

with certain conditions, such as contacting speed the vehicles should be less than a specific value and the process must be save fuels as possible. 2.2

Cognitive Process Analysis

Cognition refers to a range of mental processes relating to the acquisition, storage, manipulation, and retrieval of information. The finish for a total mission can be taken as an accumulation of series completion of small tasks’. Similarly, a whole task’s cognitive process can be seen as loop fulfillment of some small cognitive segments. Someone names the segment as cognitive loop, which is a system that senses, learns, acts, and operates. A cognitive segment is also a cognitive behavioral process that includes sensation, perception, information processing, decision-making, and motor. The perceptual sources are prerequisites for information processing. Such sources are number limited and they are assigned to all stages for processing. Attention selection is the starting point of perception processing, and it determines the way of information processing. Attention selection is influenced by four main factors: the salience (S) of events that might capture attention, the effort (E) required to redirect attention from one location to, the expectancy (E) that a given location in the visual field will contain information, and the value (V) of information to be obtained at that location. For manual RvD tasks, information processing relevant cognitions mainly includes perception, memory and understanding. Attention selection for cognitive sources is mainly to pay attention to tracking drone and target vehicle, as well as the dynamic information displaying on the screen. Showing as a circular dial the salience of drone produces visual stimuli to operator and generates visual attention. Expectancy and values is how much attentions attribute to the perceptual sources. The perception process collects information from the outside environment. With prior knowledge, meaningful interpretation of the perceived information is revealed through a top-down processing mechanism. Knowledge is stored in long-term memory, sometimes reacting directly to the perception, and sometimes the perceptual information is processed in working memory as human “thinking”. This process of the perceptual information generates many psychological activities (retelling, understanding, and decisionmaking). Working memory is similar to a temporary depository to store short-term memories temporarily. Its further processing makes it possible to be a representation persistent of information, i.e. knowledge, which stored in long-term memory. These sort of knowledge can be recalled within a certain limited period. Figure 2 shows the diagram of human-in-loop manual RvD control process. ACT-R based manual RvD task cognitive modeling aims to implement information processing for astronaut’s RvD control behavior according to ACT-R cognitive architecture specification. As mentioned above, the whole task’s cognitive process is the loop of cognitive segments. Depending on cognitive development illustrated in Fig. 3, the information processing for each cognitive segment involves the following seven steps. They are observing and finding the tracking drone and target vehicles; perceiving target and chasing vehicle’s position and posture; processing perceptual

Cognitive Architecture Based Mental Workload Evaluation

823

1. Target Vehicle Searching on Screen

PosiƟon and posture saƟsfy manual RvD condiƟon 2. AƩenƟon SelecƟon

6. Motor & Control Focus on target vehicle and tracing drone

Change vehicle posiƟon and posture

CogniƟve Loop

5. Decision Making

Fire the producƟon rule

Find movement trend for posiƟon and posture

3. InformaƟon Processing

Retrieve producƟve knowledge for current situaƟon

4. Knowledge Retrieve

Fig. 2. Information processing and cognitive development diagram for manual RvD task.

information and sensing changes of vehicle’s position and posture; querying and retrieving knowledge; imaging the spatial transition of the spacecraft; information processing for visual imagination, and making decision etc. Therefore, cognitive process for ACT-R based cognitive model of manual RvD task mainly includes: visual attention, attention selection, attention tracking, perception of vehicle’s status, information processing, knowledge retrieving, firing of productions, motoring action etc. 2.3

Cognitive Model for Manual RvD Control Task

ACT-R assumes that there are two types of knowledge – declarative and procedural. Procedural knowledge specifies how to bring declarative knowledge to bear in solving problems. ACT-R represents declarative knowledge using chunks and represents procedural knowledge by adopting productions that being collections of production rules. An ACT-R model is a computer file written in the LISP programming language, and can be executed on ACT-R platform. Declarative Knowledge. As examples, showing as the below are some declarative knowledge for manual RvD cognitive model.

824

Y. Liu et al.

(chunk-type (drone-feature (:include visual-location)) vlcty-x vlcty-y) (chunk-type (drone (:include visual-object)) sides) (chunk-type drone-position x-pos yz-pos) (chunk-type drone-size distance size) (chunk-type rvd-task state vlcty-x vlcty-y vlcty-z) (add-dm (start isa chunk) (find-drone isa chunk) (drone-appear isa chunk) (drone-location isa chunk) (float-space isa chunk) (op-left isa chunk) (op-down isa chunk) (op-right isa chunk) (op-up isa chunk) (goal isa rendocking-task state start) ... )

Procedural Knowledge. According to above cognitive processing analysis of manual RvD control task, there are three main types of procedural knowledge, which they are for visual perception, for control operation and for status identification of speed, position and posture. As examples, visual perception and control operation procedural knowledge are shown as following. The following program shows the procedural knowledge that the model find the drone of the target vehicle.

(p found-drone =goal> isa rendocking-task state find-drone =visual-location> isa drone-feature ?manual> state free ==> +visual> isa move-attention screen-pos =visual-location =goal> state drone-appear )

Cognitive Architecture Based Mental Workload Evaluation

825

And the following is procedural knowledge of the model for operation moving tracing drone to the right direction.

(p right-op =goal> isa rvd-task state op-right ?manual> state free ?visual> state free =visual-location> isa drone-feature distance =di-s screen-x =sc-x screen-y =sc-y vlcty-x =vel-x vlcty-y =vel-y ==> !bind! =dis-new (- =di-s 2) ... !bind! =v-y-new (+ =vel-y 0) +manual> isa press-key key "l" =visual-location> distance =dis-new screen-x =scx-new screen-y =scy-new velocity-x =v-x-new +visual> isa move-attention screen-pos =visual-location =goal> state drone-location !bind! =real-x (- =sc-x 300) !bind! =real-y (- =sc-y 300) !output! (=dis-new =real-x =real-y =v-x-new =v-ynew) )

ACT-R model for manual RvD task has the following structure like a typical model that a text file written in LISP format.

826

Y. Liu et al.

(clear-all) ; reset ACT-R’s state to a clean state {Lisp functions for presenting an experiment, data collection or other support needs} (define-model rend-docking ; model’s name goes here (sgp {parameter value}*) ; set parameters {chunk-type definitions} ; declare chunk type {initial chunks are defined}; chunks goes here {productions are specified} ; production rules {any additional model set-up commands} ; first goal {additional model parameter settings} ; others )

There is an initial chunk (first goal) incorporated usually, and place it into goal buffer. As running the model, the first goal is retrieved, and it achieves some other sort of goal until there is no production rule for the last one goal.

3 Cognitive Workload and Verification ACT-R is a biologically inspired cognitive architecture promoting and facilitating with transdisciplinary study, and its modules explicit mapping human brain’s relevant functional region. Therefore, to some extent the model’s activities give an expression of human cognitive workload of correspondence task. 3.1

Cognitive Workload Correspondence to Active Time of ACT-R Modules

The motivation to build an ACT-R cognitive model for manual RvD task is to make clarify of the cognitive processes and then to evaluate the cognitive workload for each cognition segment. A running ACT-R model can vividly depict the cognitive details in a cognitive cycle 50 ms. While model running, the activities and active time for each events are displayed and can be recorded. It is reasonable to believe that while the module is active it is the fact that brain is working, i.e. there is workload for human. Based on this, each modules total active time for a specific task can be superimposed, and the correspondence brain region’s active time is obtained. While summing the brain active total time, this time is taken as the human workload of the task. Module’s total active time for ACT-R model represent brain’s active time of the task-related cognitive activity and can be an index for task’s cognitive workload. By comparing the total amount of time that task cognition related brain activation one can investigate workload for manual RvD tasks in different cognitive factors. Task’s intensity and difficulty is chosen as research subjects in this paper, and 8 participants, 9 ACT-T model, and 4 group task intensity and difficulty situation in experiment. Figure 3 shows the workload of manual RvD tasks to different task intensity and difficulty for nine ACT-R models.

Cognitive Architecture Based Mental Workload Evaluation

827

Fig. 3. The workload of manual RvD tasks to different task intensity and difficulty for nine ACT-R models.

We notice from Fig. 3 that workload is highly relevance with task’s intensity and difficulty. The more difficult the workload of the task will increase, and so does with task’s intensity. In addition, we find that both task’s intensity and task’s difficulty affect workload in Prefrontal brain region mostly. 3.2

The Comparison with NASA-TLX Scale Measurement

The NASA-TLX is a widely used, subjective, multidimensional assessment tool that rates perceived workload using six factors, namely the Mental demand (MD), Physical demand (PD), Temporal demand (TD), Performance (OP), Effort (EF), frustration levels (FR). Measurement of a subjective mental workload most widely used because experiences shown that it has a high degree of validity. To verify the cognitive model based workload assessment method, the average value predicate by NASA-TLX method is applied to compare with those of the proposal method. Table 1 shows the comparison of NASA-TLX method with model based workload assessment. Table 1. The comparison of NASA-TLX with model based workload assessment. Task Wact-r-ave NASA-TLX Mean workload S2 2.323 42.950 S4 3.711 50.392 D2 5.851 55.967 D4 7.281 66.277

Standard error 2.440 1.795 1.470 2.283

828

Y. Liu et al.

In above table, Wact-r-ave is the workload predicated using ACT-R model. We find it is highly relevant to the result derived by NASA-TLX Scale Measurement (F (1,11) = 87.837, p < 0.001, and R2 = 0.898). 3.3

Workload Simulation onto Human Brain Region

Each module in ACT–R is designed to associate with distinct functional cortical regions. During ACT-R model running, the activation of each module can map to an anatomical human brain. By applying animation technics, the changes of human workload can be portrayed on a 3D brain image along with human mind activities.

4 Discussion Mental workload remains an important variable with which to understand user performance. However, there is no single measure that clearly discriminates mental workload but there is a growing empirical basis with which to inform both science and practice [12]. Even we did not make a conception distinction between cognitive workload and mental workload in this paper. Therefore, there are no much efforts on quantities assessment for mental workload. In this paper, we did not mention the following issues that the consequence will be more convincing if considered. In the experiment, the degree of task’s intensity and difficulty are evaluated subjectively, and it is a sense of experiment’s conductor. Correspondence between ACT-R modules and brain regions is well-designed and corroborated by BOLD curves obtained from fMRI. However, this is a qualitative relationship between module’s active and mental workload other than quantities one.

5 Conclusions This paper presents a method to evaluate the mental workload for manual RvD control task on the base of cognitive architecture ACT-R. In this study, through the analysis of cognitive process for manual RvD control task, the declarative knowledge and procedural knowledge are constructed for inspecting the workload affected by task’s intensity and difficulty two cognitive factors. By running manual RvD cognitive model, the active time for each ACT-R module are obtained. By accumulating the total active time for each module, the brain region and whole task’s workload achieved. To validate the assessment, the result is compared to that of NASA-TLX measurement. The results show that they are consistent. This paper’s main job is two-folded. The one is we propose and implement a mental workload assessment method in a cognitive level for space fine manipulation task. The other one is we achieved simulation of human cognitive action’s mapping in human’s brain region.

Cognitive Architecture Based Mental Workload Evaluation

829

Acknowledgments. This work is supported by the Fusion Development Foundation of China Fusion Co. Ltd., the CCiS Institute of Zhejiang Sci-Tech University and Natural Science Foundation of Zhejiang Provincial (No. LY12C09005).

References 1. Razak, I.H.A., Kamaruddin, S., Azid, I.A.: Towards human performance measurement from the maintenance perspective: a review. Int. J. Eng. Manag. Econ. 2(1), 60–80 (2011) 2. Liu, Y.F., Tian, Z.Q., Liu, Y.Z., Li, J.S., Fu, F., Bian, J.: Cognitive modeling for robotic assembly/maintenance task in space exploration. In: International Conference on Applied Human Factors and Ergonomics, pp. 143–153. Springer, Cham (2017) 3. National Research Council: Quantitative Modeling of Human Performance in Complex. Dynamic Systems. National Academies Press, Washington, DC (1990) 4. Moray, N. (ed.): Mental Workload: Its Theory and Measurement, vol. 8. Springer, Heidelberg (2013) 5. Rouse, W.B., Edwards, S.L., Hammer, J.M.: Modeling the dynamics of mental workload and human performance in complex systems. IEEE Trans. Syst. Man Cybern. 23(6), 1662– 1671 (1993) 6. Zhang, S.Y., Tian, Y., Wang, C.H., Huang, S.P., Fu, Y., Chen, S.G.: Modeling human control strategies in simulated RVD tasks through the time-fuel optimal control model. In International Conference on Digital Human Modeling and Applications in Health, Safety, Ergonomics and Risk Management, pp. 661–670. Springer, Cham (2014) 7. Liu, Y.F., Tian, Z.Q., Zhang, Y., Sun, Q., Li, J.S., Sun, J., Fu, F.: COMPAss: a space cognitive behavior modeling and performance assessment platform. In: Digital Human Modeling. Applications in Health, Safety, Ergonomics and Risk Management, pp. 630–636. Springer, Switzerland (2014) 8. Anderson, J.R., Bothell, D., Byrne, M.D., Douglass, S., Lebiere, C., Qin, Y.: An integrated theory of the mind. Psychol. Rev. 111(4), 1036–1060 (2004) 9. Liu, Y., Tian, Z., Liu, Y., Li, J., Fu, F.: Cognitive architecture based platform on human performance evaluation for space manual control task. In: Advances in Neuroergonomics and Cognitive Engineering, pp. 303–314. Springer, Heidelberg (2017) 10. Möbus, C., Lenk, J.C., Özyurt, J., Thiel, C.M., Claassen, A.: Checking the ACTR/brain mapping hypothesis with a complex task: using fMRI and Bayesian identification in a multidimensional strategy space. Cogn. Syst. Res. 12(3), 321–335 (2011) 11. Muller, T.J., Heuvelink, A., Both, F.: Implementing a cognitive model in soar and ACT-R: a comparison. In Proceedings of Sixth International Workshop, From Agent Theory to Agent Implementation (2008) 12. Charles, R.L., Nixon, J.: Measuring mental workload using physiological measures: a systematic review. Appl. Ergon. 74, 221–232 (2019)

Author Index

A Adamovská, Eva, 179 Adeboyejo, Thompson A., 618 Al Eisaei, Mohammed, 214 Al Ghafli, Abdulla, 214 Al Thabahi, Yousif, 214 Al Zaabi, Marzouq, 214 Alford, Chris, 120 Amanatidis, Theocharis, 157 Aragattu, Supraja, 556 Auflick, Jack L., 499 Avula, Yashwant, 556 B Badilla, Ma. Gilean Fria, 361 Balasubramanian, Venkatesh, 224, 233 Barton, Laura, 585 Bavendiek, Jan, 15 Beggiato, Matthias, 79, 107 Bengtsson, Peter, 487 Bernard, Tiziano, 807 Berry, Katherine, 695 Berti, Zavier, 585 Beyer, Stefanie, 545 Bianchini Ciampoli, Luca, 169, 330 Biermann, Hannah, 573 Bikam, Peter, 618 Birrell, Stewart, 307, 387 Bladfält, Sanna Lohilahti, 487 Borsuk, Sergii, 783 Boulnois, Sebastien, 807 Brandl, Christopher, 585 Brell, Teresa, 573 Bröhl, Christina, 585 Brost, Waldemar, 573

Brown, James, 564 Bullinger, Angelika C., 36, 608 Burns, Christopher G., 307 C Caber, Nermin, 95 Caingat, Angela Jasmin B., 412 Caleb-Solly, Praminda, 120 Calvi, Alessandro, 169, 330 Campbell, Nicholas, 319 Carruth, Daniel W., 67, 145 Carruth, Daniel, 44 Carvalhais, José, 190 Causse, Mickaël, 739 Chakwizira, James, 618 Chauvet, Robert, 585 Chen, Chun-Hsien, 715 Clark, Jediah R., 27 Clarkson, P. John, 95, 157 Cobb, Sue, 250 Collins, Rebecca, 695 Colón, Enid, 319 Colucci, Benjamín, 319 Cookson, Simon, 702 Couto, António, 190 Cuffie, Brandon, 807 Cunha, Liliana, 190 Custodio, Benette P., 412 D D’Amico, Fabrizio, 169, 330 Daniel, Schubert, 794 Darko, Justice, 205 de Winter, Joost C. F., 462 Deb, Shuchisnigdha, 44, 67, 145

© Springer Nature Switzerland AG 2020 N. Stanton (Ed.): AHFE 2019, AISC 964, pp. 831–834, 2020. https://doi.org/10.1007/978-3-030-20503-4

832 Deisser, Oliver, 545 Deml, Barbara, 243 Dettmann, André, 608 Di Nicolantonio, Massimo, 656 Diermeyer, Frank, 475 Dietrich, Wilhelm, 377 Dreger, Felix A., 462 E Ebnali, Mahdi, 133 Ebnali-Heidari, Majid, 133 Eckstein, Lutz, 15 Eganhouse, Joseph T., 499 Eickels, Teresa, 573 Eimontaite, Iveta, 120 Eland, Anthony, 440 F Fank, Jana, 475 Fasanya, Bankole K., 556 Fay, Daniel, 669 Fenton, Christopher J., 645 Ferrante, Chiara, 169, 330 Ferreira, Sara, 190 Figueroa, Alberto, 319 Folsom, Larkin, 205 Forster, Yannick, 3 Frey, Darren, 145 Fu, Feng, 819 Fuad, Muztaba, 145 G Gabalda, Elijah, 361 Ge, Wanmin, 531 Grane, Camilla, 487 Gregorovič, Adam, 179 Guo, Yundong, 762 H Habib, Noor Zainab, 510 Halama, Josephine, 79 Hancock, Gabriella M., 596 Happee, Riender, 462 Hartwich, Franziska, 107 Havlíčková, Darina, 179 Heine, Tobias, 243 Hellig, Tobias, 585 Hensch, Ann-Christin, 79 Hergeth, Sebastian, 3 Hirz, Mario, 377 Hult, Carl, 633 Hung, Vivien, 307

Author Index J Jennings, Paul, 387 Johan, Henry, 715 K Kakizaki, Yutaka, 282 Kaneria, Acyut, 341 Käppler, Marco, 243 Keinath, Andreas, 3 Keshavula, Swetha, 556 Khan, Ata, 57 Kian, Cyrus, 133 Kim, Jisun, 261 Kinsella, Amelia, 695 Knies, Christian, 475 Knodler, Michael, 319 Kobayashi, Wataru, 301 Kolrep, Harald, 794 Kopp, Gerhard, 545 Krems, Josef F., 3, 79 Krems, Josef, 107 Kummari, Bharath, 556 Kurra, Sivaramakrishna, 556 L Lan, Zirui, 715 Langdon, Patrick, 95, 157 Lenard, James, 427 Li, Fan, 715 Li, Haiyuan, 531 Li, Jusong, 819 Li, Yueqing, 341, 521 Linkov, Václav, 179 Liu, Yanfei, 819 Liu, Yisi, 715 Liu, Yuzhou, 819 Lobo, António, 190 Lounis, Christophe, 739 Lye, Sun Woh, 727 M Maeda, Setsuo, 400 Manchaiah, Vinaya, 341 Marshall, Russell, 427, 440 Martinez, Anna Patricia F., 412 Mazloumi, Adel, 133 Mehta, Yash, 807 Mertens, Alexander, 585 Miglianico, Denis, 272 Mitani, Hiroki, 400 Mizukami, Naoki, 282 Moran, Sabrina N., 596 Morgan, Phillip, 120

Author Index Mueller, Alexander, 545 Mueller-Wittig, Wolfgang, 715 Musso, Giorgio, 794 N Nakamura, Hitomi, 400 Naujoks, Frederik, 3 Nedbay, Serhiy, 783 Neubauer, Matthias, 453 Neumann, Isabel, 79 Nitsch, Verena, 585 Nolan-McSweeney, Michelle, 250 O Oliveira, Emily, 15 Oliveira, Luis, 307 P Parameswaran, Swathy, 233 Pargade, Vincent, 272 Park, John (Hyoshin), 205 Paterson, Abby, 440 Petzoldt, Tibor, 107 Peysakhovich, Vsevolod, 739 Philipsen, Ralf, 573 Pinheiro, Jean-Philippe, 727 Pope, Kiome A., 645 Praetorius, Gesa, 633 Preston, John, 261 Pugh, Nigel, 205 Q Qian, Chao, 521 R Rahman, Md Mahmudur, 44 Rajendran, Minerva, 224 Ramoso, Jeonne Joseph, 361 Rasche, Peter, 585 Rayad, Yassine, 807 Reva, Oleksii, 783 Revell, Kirsten M. A., 27 Revell, Kirsten, 261, 564 Rick, Vera, 585 Roberts, Aaron P. J., 645, 669 Roberts, Julius M., 499 Robielos, Raine Alexandra S., 412 Rojas, Maria, 319 Rossner, Patrick, 36 Ryan, Brendan, 250 S Samuel, Siby, 205 Sandberg, Carl, 633

833 Santen, Leon, 475 Schäfer, Katharina, 585 Schauer, Oliver, 453 Schildorfer, Wolfgang, 453 Schlacht, Irene Lia, 794 Schmid, Daniela, 683 Sedilla, Keneth, 361 Serrato, Alicia, 774 Shulgin, Valeriy, 783 Shyrokau, Barys, 462 Silva, Daniel, 190 Simões, Anabela, 190 Smith, Lori, 695 Smyth, Joseph, 387 Sourina, Olga, 715 Sprague, James K., 499 Stanley, Laura M., 145 Stanton, Neville A., 27, 645, 669, 683 Stanton, Neville, 564 Stephane, Lucas, 807 Still, Jeremiah D., 291 Still, Mary L., 291 Strawderman, Lesley J., 67 Strawderman, Lesley, 44 Strybel, Thomas Z., 596 Summerskill, Stephen, 440 Summerskill, Steve, 427 Sun, Youchao, 762 Suyama, Koki, 400 Suzuki, Daisuke, 282 T Tainter, Francis, 319 Tanvir, Shahid, 510 Tatsuno, Junya, 400 Tavares, José Pedro, 190 Theis, Sabine, 585 Thomas, Lauren, 774 Thomas, Peter, 307 Tian, Zhiqiang, 819 Tiefnig, Johannes, 377 Tosti, Fabio, 169, 330 Trapsilawati, Fitri, 715 Tsuyuki, Nobuyuki, 282 V Valdés, Didier, 319 Voinescu, Alexandra, 120 Vu, Kim-Phuong L., 596 W Walker, Guy H., 510 Wang, Yuhong, 521 Wee, Hong Jie, 727

834 Wille, Matthias, 585 Wu, Xiaoyu O., 752

Y Yan, Xuedong, 351 Yoshioka, Yohsuke, 301

Author Index Z Zámečník, Petr, 179 Zaoral, Aleš, 179 Zhang, Xinran, 351 Zhao, Xiang, 341 Ziefle, Martina, 573 Zuo, Wenchao, 521