Drones in Smart-Cities: Security and Performance [1 ed.] 0128199725, 9780128199725

Table of contents :
Copyright
Dedications
Contributors
Preface
Chapter 1 - Aerial and underwater drone communication: potentials and vulnerabilities
Chapter outline
1 -Introduction
1.1 - Motivation
1.2 - Modern computing systems: cyberattacks
2 - Main communication technologies in UWC and UAV
2.1 - Seawater categorization
3 - Aerial vehicle communication
4 - Aerial vehicle control architecture
5 - Path planning
6 - Machine learning and artificial intelligence technologies
7 - Vision-based technologies
8 - UAV attacks
9 - Phases of UAV networks
10 - UAV challenges and main issues in underwater RF communication
11 - Discussions and open issues
11.1 - Most common Issues in underwater acoustic communication
11.2 - Applicability and applications
11.3 - Applications based on machine learning
12 - Conclusion and future work
References
Chapter 2 - Machine learning for cyber security frameworks: a review
Chapter outline
1 - Introduction
2 - Machine learning tasks
3 - Cyber security tasks
4 - Machine learning approaches for cyber security
5 - The effectiveness of machine learning in cyber security
6 - Conclusion
References
Chapter 3 - Emerging use of UAV’s: secure communication protocol issues and challenges
Chapter outline
1 - Introduction
2 - Unmanned aerial vehicles (UAVs)
3 - Ground control station (GCS)
4 - Types of UAVs
5 - Communication protocols for UAVs
5.1 - UranusLink protocol
5.2 - UAVCAN protocol
5.3 - MAVLink protocol
5.3.1 - MAVLink 1.0 header protocol
5.3.2 - MAVLink 2.0 header protocol
6 - Critical analysis of these protocols
7 - Discussion
8 - Conclusion
9 - Future work
References
Chapter 4 - A survey study on MAC and routing protocols to facilitate energy efficient and effective UAV-based communica...
Chapter outline
1 - Introduction
2 - Routing protocols for UAVs
2.1 - Single-hop routing
2.2 - Multihop routing
2.3 - Position-based routing
2.4 - Topology-based routing
2.4.1 - Proactive routing
2.4.2 - Reactive routing
2.4.3 - Hierarchical routing
2.4.4 - Hybrid routing
3 - Data delivery models in UAVs
4 - MAC protocols for UAVs
4.1 - Effects of MAC
4.1.1 - Schedule-based MAC protocols
4.1.2 - Prioritized frame selection MAC protocol based on CDMA (PFSC)
4.1.3 - Prioritized frame selection MAC protocol based on TDMA (PFST)
4.1.4 - Adaptive-opportunistic (AO) aloha CDMA-based MAC
4.2 - Contention-based MAC protocols
4.2.1 - Priority-based contention window adjustment scheme (PCWAS)
4.2.2 - Partnerships-based MAC protocol
4.2.3 - Cooperative sensing data collecting framework
4.2.4 - Prioritized data gathering MAC protocol
4.3 - Hybrid MAC protocols
4.3.1 - Fixed inter Beacon duration and proactive scheduling (FDPS MAC)
4.3.2 - Adaptive inter-Beacon duration and proactive scheduling (ADPS MAC)
4.3.3 - Hybrid medium access control (HP MAC)
5 - Conclusion
References
Chapter 5 - UAVs: communication aspects and cellular service provisioning
Chapter outline
1 - Introduction
1.1 - UAV communication
2 - System model
2.1 - Channel model
2.2 - User association metric
3 - Performance analysis
4 - Results
5 - Conclusion
References
Chapter 6 - AI simulations and programming environments for drones: an overview
Chapter outline
1 - Introduction
1.1 - What is a simulator
2 - Why simulators are important in drones
3 - Where it is applicable
4 - Simulation environments
5 - Artificial intelligence types and issues
5.1 - AI types in drones’ simulations
5.2 - AI Issues in drones’ simulations
5.3 - Programming languages
6 - Applications
6.1 - Performance assessment parameters
7 - Other assessment methods
8 - Open research issues
9 - Conclusion
References
Chapter 7 - Smart agriculture framework using UAVs in the Internet of Things era
Chapter outline
1 - Introduction
1.1 - What is smart agriculture?
1.2 - Why we need it?
1.3 - Where can we apply it?
2 - System model
2.1 - Multispectral imaging device
2.2 - Unmanned aerial vehicle
3 - The smart framework
3.1 - Factors affecting grain production
3.2 - Plant nutritional elements
4 - Discussions
5 - Concluding remark
References
Chapter 8 - Coastline change determination using UAV technology: a case study along the Konyaaltı coast, Antalya, Turkey
Chapter outline
1 - Introduction
2 - Study area
3 - Methodology
3.1 - Flight planning with UAV
3.2 - Cameras and systems used in UAVs
3.3 - Distortion and calibration
3.4 - Distortion error correction
3.5 - Calibration of the camera used in the present study
4 - Results and discussion
4.1 - West of the study area
4.2 - East of the study area
5 - Summary and conclusions
References
Chapter 9 - Explorative analysis of AUV-aided cluster-based routing protocols for Internet of intelligent underwater sensors
Chapter outline
1 - Introduction
2 - Internet of intelligent underwater sensor networks (IoIUSN) application: AUV-aided WQM
2.1 - Conceptual definition, framework and features of IoIUSN
2.2 - Conceptual framework and features
2.3 - Characteristics and functional requirements of AUV-aided WQM
2.4 - Implementation challenges
3 - CBR protocols for UWSN
3.1 - Technical definitions, features, and architecture
3.2 - Expository analysis
3.3 - Network performance and statistical analysis
4 - Results and discussion
5 - Conclusion
Acknowledgments
References
Chapter 10 - Physics of stabilization and control for the Drone’s quadrotors
Chapter outline
1 - Introduction
1.1 - Basic concepts
1.2 - Physical concepts
1.2.1 - Equations of motion
1.2.2 - Angular rates transformation
1.2.3 - Linear acceleration
1.2.4 - Angular acceleration
1.2.5 - Gyroscopic moments of the propellers
2 - Momentum theory of rotors
2.1 - Basic concepts
2.2 - Axial motion momentum theory
3 - Control techniques
3.1 - PID control
3.2 - LQR control
3.3 - H infinity control
4 - Hardware
4.1 - KK2 flight controller board
4.2 - Out-runner brushless DC motor
4.3 - Electronic speed controller
4.4 - Radio transmitter and receiver
4.5 - LI-PO battery
5 - Conclusion
References

Citation preview

DRONES IN SMART-CITIES SECURITY AND PERFORMANCE

Edited by

FADI AL-TURJMAN Professor, Department of Artificial Intelligence Near East University Nicosia, Mersin 10, Turkey

Elsevier Radarweg 29, PO Box 211, 1000 AE Amsterdam, Netherlands The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom 50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States Copyright © 2020 Elsevier Inc. All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www.elsevier.com/permissions. This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be noted herein). Notices Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary. Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility. To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein. Library of Congress Cataloging-in-Publication Data A catalog record for this book is available from the Library of Congress British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library ISBN: 978-0-12-819972-5 For information on all Elsevier publications visit our website at https://www.elsevier.com/books-and-journals

Publisher: Joe Hayton Acquisitions Editor: Brian Romer Editorial Project Manager: Ali Afzal-Khan Production Project Manager: Swapna Srinivasan Designer: Mark Rogers Typeset by Thomson Digital

To the best parents. To Sinem and the little beautiful stars. Love you all. Fadi Al-Turjman

Contributors Emmanuel Adewale Adedokun Department of Computer Engineering, Ahmadu Bello University, Zaria, Nigeria Hamit Altiparmak Department of Computer Engineering, Near East University, Nicosia, Mersin, Turkey Fadi Al-Turjman Department of Artificial Intelligence Engineering, Research Centre for AI and IoT, Near East University, Nicosia, Mersin, Turkey Habeeb Bello-Salau Department of Computer Engineering, Ahmadu Bello University, Zaria, Nigeria Sarfraz Nawaz Brohi School of Computing and Information Technology (SOCIT), Taylors University, Subang Jaya, Selangor, Malaysia Bakkiam David Deebak School of Computer Science and Engineering,Vellore Institute of Technology, Vellore, India Kamil Dimililer Department of Electrical and Electronic Engineering, Research Centre for AI and IoT, Near East University, Nicosia, Cyprus, Mersin, Turkey Krishna Doddapaneni Amazon Web Services, San Francisco, CA, United States Eustace M. Dogo Institute for Intelligent Systems, University of Johannesburg, Johannesburg, South Africa Enver Ever Computer Engineering, Middle East Technical University, Northern Cyprus Campus, Mersin, Turkey Syed Ali Hassan School of Electrical Engineering and Computer Science (SEECS), National University of Sciences and Technology (NUST), Islamabad, Pakistan Galip Savas Ilgi Information Systems Engineering Department, Faculty of Engineering, and Research Centre for Artificial Intelligence in IoT, Near East University, Nicosia, Mersin, Turkey Muhammad Asim Jan School of Electrical Engineering and Computer Science (SEECS), National University of Sciences and Technology (NUST), Islamabad, Pakistan

xi

xii

Contributors

Noor Zaman Jhanjhi School of Computing and Information Technology (SOCIT), Taylors University, Subang Jaya, Selangor, Malaysia Haejoon Jung Department of Information and Telecommunication Engineering, Incheon National University, Incheon, Korea Navid Ali Khan School of Computing and Information Technology (SOCIT), Taylors University, Subang Jaya, Selangor, Malaysia Yoney Kirsal Ever Software Engineering Department, Faculty of Engineering, and Research Centre for Artificial Intelligence in IoT, Near East University, Nicosia, Mersin, Turkey Yoney Kirsal-Ever Research Centre for AI and IoT, Near East University, Nicosia, Cyprus, Mersin, Turkey Ali Danandeh Mehr Civil Engineering Department, Antalya Bilim University, Antalya, Turkey Anand Nayyar School of Computing and Information Technology (SOCIT), Taylors University, Subang Jaya, Selangor, Malaysia Ezekiel T. Ogidan Research Centre for AI and IoT, Near East University, Nicosia, Cyprus, Mersin, Turkey Oluwaseun Priscilla Olawale Department of Software Engineering, Research Center for AI and IoT, Near East University, Nicosia, Mersin, Turkey Bashir Olaniyi Sadiq Department of Computer Engineering, Ahmadu Bello University, Zaria, Nigeria Abdulazeez Femi Salami Department of Computer Engineering, University of Ilorin, Ilorin, Nigeria Hanifa Teimourian Department of Electrical and Electronic Engineering; Department of Artificial Intelligence Engineering, Research Centre for AI and IoT, Near East University, Nicosia, Mersin, Turkey Rifat Tur Civil Engineering Department, Akdeniz University, Antalya, Turkey Levent Uzunsakal Civil Engineering Department, Akdeniz University, Antalya, Turkey

Preface Security and Drones have been considered as a complementary package toward realizing the emerging smart-cities paradigm. From this perspective, it is essential to understand the role of these significant components, which will provide a comprehensive vision for the worldwide smart city project in the near future. It is also essential to consider the emerging Drones-based intelligent applications for better lifestyle and more optimized solutions in our daily life. The objective of this book is to overview the existing Drone-based Smart-cities applications while focusing on security issues/challenges. The main focus is on the Drones-based intelligence aspects that can help in realizing such paradigm in a secured way. The Artificial Intelligent (AI) techniques as well as the new emerging technologies such as the Internet of Things (IoT) accompanied with critical evaluation metrics, constraints, and open research issues are included for discussion. This conceptual book, which is unique in the field, will assist researchers and professionals working in the area to better assess the proposed Drone-based Smart cities paradigms, which have already started to appear in our societies. Hope you enjoy it... Fadi Al-Turjman

xiii

CHAPTER ONE

Aerial and underwater drone communication: potentials and vulnerabilities Bakkiam David Deebaka, Fadi Al-Turjmanb

School of Computer Science and Engineering,Vellore Institute of Technology,Vellore, India Department of Artificial Intelligence Engineering, Research Centre for AI and IoT, Near East University, Nicosia, Mersin, Turkey a

b

Chapter outline 1 Introduction 1.1 Motivation 1.2 Modern computing systems: cyberattacks 2 Main communication technologies in UWC and UAV 2.1 Seawater categorization 3 Aerial vehicle communication 4 Aerial vehicle control architecture 5 Path planning 6 Machine learning and artificial intelligence technologies 7 Vision-based technologies 8 UAV attacks 9 Phases of UAV networks 10 UAV challenges and main issues in underwater RF communication 11 Discussions and open issues 11.1 Most common Issues in underwater acoustic communication 11.2 Applicability and applications 11.3 Applications based on machine learning 12 Conclusion and future work References

1 3 4 6 7 8 11 13 14 15 16 17 17 18 19 19 20 21 22

1  Introduction An autonomous-based unmanned aerial vehicles known as drones has gained more attention for boarder surveillance, disaster monitoring, and relay communication [1,2]. An aircraft without pilot on board is particularly aimed to ease deployment complexity, low acquirement and maintenance cost. The development activities address the control issues related Drones in Smart-Cities. http://dx.doi.org/10.1016/B978-0-12-819972-5.00001-X Copyright © 2020 Elsevier Inc. All rights reserved.

1

2

Bakkiam David Deebak, Fadi Al-Turjman

to robustness, stability and maneuverability. In general, the autonomous vehicles are capable of minimum user interaction to collaborate with UAV communication system. The vehicular applications connect multiple smart aerials over networks to meet the time critical missions. However, they have limited data payload and energy. They can autonomously fly through flight plans controlled by software or controlled by a remote control. Generally speaking, navigation refers to the activity or process to accurately determine one's position and plans to follow a path routing. Hence, drone navigation can be the field of research focusing on system development to measure the position and lay down an effective path plan for drones. It is the key to develop the next generation autonomous drones. They may be helpful to reach the areas beyond the reach of human being. In such situations, they are engaged in critical operations such as disaster monitoring [3], surveillance, transportation, agriculture [4], forestry, and environment protection [5]. Farmers use drones to map out their crops in order to check irrigation systems, and also determine which areas of crops are impaired to make proper treatment accordingly. Drones have also been used by many retailers like Amazon to deliver the products. They are also heavily used in movie industries to cover the beautiful cinematic shots. Nowadays, it is easily available for everyone to purchase and access at low prices to boot. Initially, the independent drones are deployed in such situations owing to the advancement in drone and communication technologies. Multiple drones perform the critical operations to experience proper synchronization. More UAV are being employed in civil applications due to high mobility and flexibility. However, in some complex environments, UAV cannot sense the environment parameters due to limited communication and traditional sensor perception capabilities. Despite many efforts to overcome these weaknesses, it is still necessary to develop more efficient and effective method in order to perform more stability, predictability, and security. Therefore, high performance independent navigation is of great importance to develop the application of UAV as shown in Fig. 1.1. The control of each drone falls on pilot to use visual tracking to determine position and orientation. More advanced drones use global positioning system (GPS) receivers to play a significant role, that is, navigation and control loop. Some smart features include drone memorization to track the position track. The trajectory of the drone can be predetermined to establish GPS waypoints. When this function is executed, the drone will use autopilot to follow this path.

Aerial and underwater drone communication: potentials and vulnerabilities

3

Figure 1.1  High-level design architecture for drone communication.

1.1 Motivation This paper aims at providing all recent developments in various communication fields of drone navigation on one platform. So as to gain knowledge about developments in the field, we are trying to contribute the state-of-art concepts. The motivation of this work is derived from vast applications to possess. A flying robot can be used to perform data collection and patrolling [3]. Multiple microaerial vehicles (M-MAVs) can be used to extend joint search and rescue after disaster situations [4]. Mini drones can be perceived through the distinct sound and can act as an HCI for navigating visually impaired persons [6]. Mid-air displays can be used in scenarios like controlling large crowds in emergency situations, navigation in crowd gatherings like sports [7]. The paper is divided into eight major domains that together comprise navigation systems of a drone. This includes ML-, AI- and visionbased technologies, control systems, communication systems, path planning, and autonomous navigation [8,9]. Shakhatreh et al. [10] identifies the application areas and its related domain functionalities to signify the key factors of public safety and civil security. From Table 1.1, a comprehensive overview is summarized to realize the deficiency factors of application domains. M-MAV provides diversity to observe and sense the environment, which improves data reliability and fault tolerance. The design challenges

4

Bakkiam David Deebak, Fadi Al-Turjman

Table 1.1  Listing application areas and its domain functionalities. Authors WAN RS RTM SAR GL

Shakhatreh et al. [10] Khuwaja et al. [11] Khawaja et al. [12] Z. Ullah et al. [13] Motlagh et al. [14] Gupta et al. [15] Hayat et al. [16]

× × × × × × ×

×

×

× ×

INT

SI

× ×

×

× ×

×

×

× ×

GD, Goods Delivery/ Logistics; INT, Surveillance; RS, Remote Sensing; RTM, Real Time Monitoring; SAR, Search and Rescue; SI, Structural Inspection; WAN, Wide Area Network

are highly extended to meet the goals of civil applications. Some research challenges are associated with on-board sensors, embedded hardware and software that exchange the data over wireless channel. Importantly, communication networks have been a source of interconnected systems for effective data sharing. Modern day networks have extended their limits for effective data transmission. UAVs have gained a lot of attention over a decade. These vehicles use UAV as network nodes to represent a real-world case study to investigate distributed and centralized approaches [17]. The applications of existing network envision to illustrate the time-critical applications, that is, for real-time support. In cooperative mode, multiple UAVs have further introduced as new network node that behaves as traditional networks to operate.

1.2  Modern computing systems: cyberattacks Modern computing applications such as Urban Computing, Internet of Things, Ubiquitous Computing, and the Internet for All have sought applications of UAVs to attain complex tasks [18–21]. However, the rapid usage and its related issues could not ensure better robustness and fault-free networks. Security is one of the major issues for all types of aerial networks. A lot of network attacks have been addressed for vulnerability assessments that may be cooperative to identify the possible threats on an aerial network [22–25]. Security has always been a concern for UAV. Trustworthy and secure communication system is the major demand of unmanned aerial vehicle systems [26,27]. In the past, the concept of attacks in UAVs has been studied for cyber-physical systems. UAVs incorporate cyber systems to analyze various possible attacks and countermeasures [28]. Cyber attacks can heavily impact the operations of regular aerial vehicles affecting process of

Aerial and underwater drone communication: potentials and vulnerabilities

5

autopilot systems [29]. In autopilot systems, vulnerability may cause severe devastation effects resulting in fatal incidents [30]. False data injection is the major defect of cyberattacks that may influence a crucial issue with drone delivery system [31]. The cyberattacks can damage the entire business model to transform the aerial vehicles as potential attackers [32].This paper introduces a new attack that proves as an intruder for the whole networks. This attack is termed as antidrone cyberattacks for target systems. The attack derives its name from both cyber and physical system properties. With an ease of reconfigurability and availability, this type of attack has full possibilities to affect the whole networks. In this paper, the concept of antidrone cyberattacks, issues in handling them, various research aspects and simulative case study have been discussed that focus on robust and fault-free connectivity over networked UAVs. Of late, global warming has become an open issue that gradually unfreezes the polar ice caps causing rise of sea level. Underwater Wireless Communication (UWC) provides a support for surveillance of boarder security especially for military purposes and commercially useful for the detection of natural resources in underwater. Moreover, it is helpful to map and discover the unobserved underwater region. Nowadays, it is using for experimental observation, data collection, and analysis, underwater navigation, disaster prevention, and early detection warning of tsunami [33]. Optical, acoustic and radio frequency (RF) wireless carriers are considered to envisage UWC in underwater applications. Deploying UWC techniques in an unexplored water medium are highly challenging as compared to terrestrial wireless communication [34]. However, quality and reliability of data transmission in shallow and deep water completely depends on physical characteristics of water medium [35]. The UWC quality of service (QoS) depends on water medium and physical characteristic of electromagnetic radio frequency, optical and acoustic waves. UWC plays a significant role in underwater application on wireless communication network. The deployment of configurations in underwater system consist of fixed and anchored sensor nodes with the sea bed, floating unmanned underwater vehicle nodes or autonomous underwater vehicle, signal receiver processing towers, floating devices (buoy) submarines, ship, and onshore base station [36]. Underwater wireless optical communication (UWOC) has less explored and rather difficult to deploy than acoustic propagation in underwater [37]. The existing underwater wireless acoustic communication (UWAC) has limited performance of low bandwidth, latency and multipath

6

Bakkiam David Deebak, Fadi Al-Turjman

propagation in water medium [38]. Electromagnetic (EM) waves, in RF 3 Hz–3 kHz ranges are capable for high data acquisition and transformation in shallow water over short distances and usually attenuated easily by seawater [38]. On the other hand, acoustic waves are affected by different propagation factors due to ambient noise and external interferences, geo-metrical expansion, attenuation, multipath effect and Doppler spreading [39]. Optical waves have high bandwidth but affected by absorption, scattering and different level of temperature in underwater. The huge demand of fifth generation (5G) wireless communication would be the future perspective of emerging technologies in wireless communication with extremely low latency, high data rates and bandwidth as compared with 4G LTE. LTE uses orthogonal frequency-division multiplexing (OFDM) technique and support 20 MHz of maximum bandwidth of communication [40]. 5G wireless networking system is an extendable cellular technology of smart devices and their applications through wireless communication which has already excessively load on existing 4G LTE wireless technology. The future aspect of the latest 5G wireless network implement to enhance the data rates, capacity, latency, and QoS are to be nostrum of current cellular technological issues [41]. The main purpose of this survey is to understand the main characteristic and existing features in UWC. This work has an overview of possible underwater communication techniques and latest updates. The remaining structure of this paper as follows: In Section 2, we discuss the main deployable technologies in UWC toward the next generation of wireless communication. Underwater wireless RF communication (UWRF) and related issues in underwater described in Section 3. UWOC has been widely discussed in Section 4. In Section 5 UWAC and issues have been discussed. Finally we conclude the above information of paper in the last section.

2  Main communication technologies in UWC and UAV Using UWOC provides much higher data rates and bandwidth as compared to acoustic waves. Acoustic waves are widely used and most popular method to achieve low latency signals in underwater over long distance and high spreading delay as compared with UWOC. UWOC, UWAC, and UWRF signal propagation techniques have been reviewed in the most recent articles [42]. A high-speed acoustic communication through OFDM techniques has been discussed in [43].The existing UWC technologies have

Aerial and underwater drone communication: potentials and vulnerabilities

7

future perspective toward the next generation (5G) to users’ requirement. In this chapter, comparison among UWC techniques, related issues and UWC emerging technologies in underwater are discussed [42]. Most recent article has an experimental discussion of generalized frequency division multiplexing technique toward underwater 5G communication system based on multibranch multicarrier filter bank [44]. Filter bank multicarrier is also a new promising technique toward for 5G application in underwater [45].

2.1  Seawater categorization The seawater has been categorized differently for UWC, clearest water, intermediate, and murkiest water [46]. The first type of seawater is clearest water and is found in Atlantic and Mid-Pacific Ocean, and second type of seawater is intermediate water which exist in Northern Pacific ocean. The third type murkiest water can be found typically in North Sea and Eastern Atlantic ocean. As in [47], seawater for optical communication has been divided into four types such as pure seawater, clear ocean water, coastal ocean water, turbid harbor, and estuary water. Scattering losses occurs due to high particle concentration in clear ocean water while the absorption and major losses occur in pure seawater. The high concentration of suspended particles that affect scattering and absorption in coastal ocean but there is the highest concentration of particles find in turbid harbor and estuary water. The method of indoor localization using low cost hardware presents in [48] to ignore other mini drones as well as static obstacles. Collision free paths can be efficiently selected while exploring the environment using a binary structure called collision disk. It can also find its own location as well as other mini drones and has some additional sensors to monitor height and sideways drift. The automatic determination of sensor's position and orientation as specified in [49] is related to a three-dimensional (3D) scene model which matches geometrically two-dimensional (2D) image structures with projected components of the related 3D model. This method is efficient and flying over urban areas to navigate the execution of a path following controlling system for a drone. It depends on 3D localization that consists of three constituents, such as drone navigation and image analysis and 3D pose estimation. It has been designed for indoor flights, which experimentally verifies on a real drone in Gazebo simulator. A new technique of observation understands the abilities of photogrammetric RPAS multisensors platform observing in terms of mapping, creation of Orthophoto, 3D model's generation, data integration into a 3D geographic information system and authentication using autonomous techniques.

8

Bakkiam David Deebak, Fadi Al-Turjman

A vision-based method uses multidimensional geometric features [50] utilizing extraction technique for monocular simultaneous localization and mapping (SLAM) [51]. It is widely used to navigate the autonomous activities of small drones. Moreover, a fast line search matching algorithm realizes the extraction and comparison of 2D lines effectively to build on monocular SLAM system. A J-Linkage algorithm generates multiplane features that develop the efficiency of map descriptor, and a technique for building point-and-line combination map. A 2D motion control of drones based on integration of Fuzzy/Lyapunov and kinetic controllers is presented to control distantly from ground, air, satellite, sea, and also during the semiimmersible operation [52]. The SI-Drone system can solve the logistic problem in shallow water that initially works on a technique of drone battery charge. It is derived using Dijkstra algorithm taking charge of queuing time at QCMs. Flight drones consider the information such as speed, current position, and destination position to report Traffic control center using cloud-based management systems. It is intended to minimize overall travel delay in drone communication at QCMs. A multiview sensing drone network provides VR associated communication session to remote user. It formulates priorityaware scene reconstruction error as a function to allocate sampling rates and calculate their best values in order to minimize drone positions and the system limitations. In addition, it designs a view sampling policy to takes actions that search new drone locations to detect the best drone network configuration over the transmission area.

3  Aerial vehicle communication Detecting techniques and location accessibility helps to improve image qualities and live video streaming [53]. Using “structure from motion” technique, orthomosaic images and 3D scene rebuilding are generated using drone images. After comparing unsupervised classification method with the direct observations, some of the problems like flying drones for longer distance are identified to overcome using HOG and SVM algorithms. The technique for self-navigation of MAVs with minimal cost quadcopter is equipped with monocular camera that works on LSD-SLAM to assess the MAV trajectory in real time. A monocular vision-based self-flying and hindrance handling system is presented. It sends the camera video and the navigation data to the ground station laptop wirelessly. ORB-SLAM computes the exact location of surrounding 3D scattered map. A Kalman Filter

Aerial and underwater drone communication: potentials and vulnerabilities

9

use sensor fusion as a proportional integral derivative (PID) controller to control 3D positions. It shows how a low-cost MAV uses experimentally to validate scientific theories and methodologies. Conclusively, MAVs demonstrates the actual world situations to present an interesting idea of learning by crashing round 11,500 times instead of using small dataset approach or simulation approach [54]. The data related to positive and negative flying is useful and powerful to set the UAV navigation. It focuses on data control characteristics to insert high-quality video feeding in drone. To analyze the flight data precisely, a video piloting (FPV/RPV) is available along with image processing. Moreover, data obtaining from Arduino combined with sensors such as Altimeter, GPS, and Inertial Measurement Unit (IMU) is helpful to provide stable flight characteristics. An analytical method for calibration of IMUs places a vital role to vary the positions physically to analyze the system errors. The model presented the digitized intermediate frequency GPS signal [55]. In C/A code, some settings are initiated to navigate noise data at the same time. After simulation, GPS IF signaling data may be collected from GPS receiver that is used to track in use of GPS spoofing signal. For global navigation satellite system (GNSS) signal reception, an active circularly polarized antenna is specified in [56] which has two bands namely GPS L1 (∼1,575 MHz) and GLONASS L1 (∼1,602 MHz). Antenna can be divided into three layers such as patch layer, feed layer and circuit layer. Wideband radio sends the signal over fixed channel information using air-to-ground/satellite connection or an aircraft or a ground station. Monterrosa et al. [57] demonstrates the use of “Drone Bosco,” which is more specific to develop the PWM driver using its rigorous coding and implementation. Also, it presents the use of FPGA with hardware descriptive language. Sato et al. [58] presents a “GPS–nonGPS integrated navigation” structure using drones of power lines. Some UTM rules are formulated for drone collision avoidance. An aerial system for landmine detection is applied to design and integrate ground penetrating radar (GPR) on autonomous drone [59]. Software defined radio (SDR) technology is introduced to develop a custom-based light GPR, which is capable of detecting landmines in different situations. A control strategy named “back-stepping + DAF” regulates the altitude of the drone. Maathuis et al. [60] studied an autonomous detection of landmines based on airborne multisensory system. This system has two high-quality IR sensors, a RGB camera (video) plus a laser illuminator, and radar.

10

Bakkiam David Deebak, Fadi Al-Turjman

Figure 1.2  Role of communication using drone in outdoor environment.

It uses a GNSS reflectometry technique to determine the properties of reflecting object on earth using reflected GNSS signals. These signals can penetrate with few centimeter of ground and its depth is depending on soil moisture. Using SDR technology, a cheap receiver (sample/prototype) can be developed to detect the metal objects. It uses UAV to assist in data collection through automation tool and relieving operators, that is, time-consuming task. Ghallabi et al. [61] presented a system architecture that allows service aerial vehicle, that is, drone to fly over an informational structured environment (ISE) to complete a particular service job on its own based on robot operating system (ROS)-TMS framework. A navigation system is provided to solve path planning and flight control system associated with ISE using finite state machine (FSM). A tracking system uses infrared camera based on ROS-TMS framework to track the optical motion that is used to navigate the aerial vehicles. Fig. 1.2 shows the communication roles of drone in outdoor environment. Sensor is used in front camera that uses vanishing point algorithm to perform navigation. Hough transform is used to detect the obstacles and SVM classifier plus HOG descriptors are used to detect an individual identity. Wang et al. [62] designed and implemented the optical navigation

Aerial and underwater drone communication: potentials and vulnerabilities

11

device to detect UAV information. The design functions can be divided into two parts such as tracking a specific object on the basis of optical image information, and perfect landing on particular target with drone stabilization. Nitsche [63] presented Monte Carlo method to exploit teach-andreplay navigation. Monte Carlo method is used to get the exact location along the learned path.The robot is localized with MCL in 1-D that generates the probabilities of loop-detection. The parallel tracking and mapping algorithm (PTAM) applied to MAV control systems [64] that show a practical possibility to work PTAM on 5 watt computer. [25]. It has destructive mind-sets to develop jamming and spoofing circuitries/devices that can alter the navigation signals. It leads to provide wrong navigation signal in order to misinform the system information. The challenges related to communication and future aerial network of UAV has a natural safety of UAS [65], which has direct inferences to control nonpayload communication (CNPC) systems. UAS-CNPC presents some new challenges such as flights to represent low elevation, requirement of wide bandwidth, etc.

4  Aerial vehicle control architecture The control architecture is based on the ROS-TMS framework [66] that allows navigation in an ISE involving drone and real-time environment. It is intended as a part of the ROS-TMS comprising of a navigation system to establish route plan in ISE and flight control system to make a performance-based FSM. Aguilar et al. [67] proposed autonomous navigation control for quadrotor MAVs. It is contained of controller design, system modeling, planning and simulations that is primarily based on two models. The former is a linear model to provide steady state whereas the latter is a nonlinear model to offer dynamic transitions. Paukovits et al. [68] presented a notion of reactive control that surpasses a time-triggered method focusing on attributes of control logic. The control decisions use navigation sensors and reactive control to meet certain objectives such more execution time, hardware utilization and provision control rates. There is a inertial navigation system (INS) that allows drone to reach back to the base without use of navigation. Fig. 1.3 shows the concept of system integration for aerial vehicular networks. Zhang et al. [69] demonstrated the accuracy of available INS on UAV with aboard video camera to improve the computer vision (CV) method. Using Google and Yandex satellite maps, wireless communication hardware has a chance to enhance accuracy factors of drone to fly several

12

Bakkiam David Deebak, Fadi Al-Turjman

Figure 1.3  A concept of system integration for aerial vehicular networks.

hours. Wong et al. [70] presented a conformal planar inverted-F antenna for mounting antennas that is designed as a good candidate for wireless communications and navigation structures such as UAV, drones, and bicycles. Moreover, it is low cost and easy to fabricate as a gradient methodology that is helpful to tune the attributes of PID, that is, proportional gain (Kp), integral gain (Ki) and derivative gain (Kd). This technique is used to develop waypoint navigation and formation controller. AR-Drone quadrotor can also act as a platform to educate the research in robotics. Drone hardware and software, drone equipment, abilities and performance are integrated to process the associative tasks such as stabilizing positions, target and autonomous navigation. Generally speaking, the drones have the abilities to navigate and control an arrangement of mobile robots is demonstrated. There are many other efforts to extend the research in this field. A small tutoring program for students has no knowledge in the field of robotics in form of a single-day competition on autonomous visual navigation of an UAV. An open-source programming library is introduced as a program to guide and accelerate the learning mechanism. It makes a vital contribution to track through nested-marker-based design. It can precisely calibrate the various parameters of AR drone. A widespread review on available open-source flight controllers, for both hardware and software has an appropriate research to develop an essential UAV system with different constituents.

Aerial and underwater drone communication: potentials and vulnerabilities

13

5  Path planning A system uses a capture motion to provide 3D coordinates for aerial vehicles, such as quadcopters through 433 MHz known as RF905 wireless communication module. To make a user interface, a visual C# programming is used that can trace the current location of quadcopter to control in real time. Finally, using human machine interface, three multipropeller flying robots are controlled for an indoor group flight. It shows the ability of Parrot AR-Drone 2.0 to direct a predefined route. To obtain the obstacle-free path, two important algorithms are used. They are optimized using path planning algorithm through simulated annealing optimization algorithm. IMU is placed on aerial vehicle, that is, drone to track location and orientation of quadcopter. Simulink model with PID is employed to control the quadcopter that uses trajectory generation and tracking algorithms. Cubic polynomials and Bezier curves are employed to create flight paths. A PID controller applies to drive the drone, which uses a navigation method to apply artificial repulsive and attractive potential fields. It is used to enforce hindrance-handling and target-achieving tasks. To fetch the global location of the drone, an extended Kalman filter (EKF) and visual landmarks are used. A navigation system to govern a nonholonomic mobile robot presented an accurate environmental data from a scanning laser rangefinder (ladar), which has a robot external stimuli perception system. This information is used to generate a 3D tessellated environmental model. Using rapidexploring random tree (R-ERT) approach, a Kino dynamic path planning algorithm is devised that allows point-to-point path traversal. A solution trajectory employs a rover to plan because there is a need to pay a higher price in terms of energy consumption and time. As a result, a physically controlled drone is attached to the rover that searches a system to identify hindrances using stereo- imaging. It is optimized the routing path to reduce time and energy consumption. It is studied to identify the drone potential factors in order to map an area. These images are processed using Pix4D software to obtain ortho-images. It uses a digital surface model (DSM) of the area to access the accuracy of differential GNSS products. It is shown that a GPS-enabled with an armature camera is an economic and powerful mapping device. A solution to real-time collision free path planning uses AR to overcome the unreachable goal problem. This approach has three components: pattern-based ground localization, potential field method for path planning,

14

Bakkiam David Deebak, Fadi Al-Turjman

and proportional differential (PD) controllers for steering commands. Flying a drone/quadcopter in an indoor environment requires precise locality and control methods, which combine microIMU and monocular vision to update the location of the element of reference lines given by monocular vision algorithm. Indoor Navigation of UAV using AscTec Firefly hardware is equipped with a RGB-D camera. The output of sensor uses to build a map of environment to generate a collision free path.The path is computed in real time and the entire computation runs on-board to solve the issue of automatic skyline extraction in digital images.The skyline is extracted using Canny filter to read the edges. The upper part is taken from the map edge to connect the breaks using shortest path algorithm. A technique for UAV is visually served to control 3D feedback of drone. A remote monocular camera can move in 3D space to track UAV image. A 3D tracking based on a cuboid model is used to collect the information of 3D spaces.

6  Machine learning and artificial intelligence technologies An observation method uses forward velocity of a mini drone to derive the approximation of velocity based on raw data from radar. First, subaperture map-drift method is used to estimate the Doppler rate variation due to flight deviation of drone. A forward velocity and radial acceleration are extracted using an adaptive filter.The researchers at Air Force Research Laboratory's TechEdge Labs have developed a scalable, physically distributable, and easily deployable for collaborative platform based on smartphone under Summer at the Edge and Year at the Edge student research programs. It is aided in detection and tracing of unknown hostile drones. It is comprised of smartphone application that uses data from on-board sensors to capture a drone's characteristics and real-time ability to collaborate a processing unit to forecast the possible flight path of an unknown drone. An approach for mapping investigates an unknown environment using image stitching technique. A search and rescue operation is crucial to perform a quick scan of disaster area, and localize the wounded people to depute a rescue team. The image map is employed to detect the human activities using CV techniques. For safe and collision-free navigation, swarms of UAV can be adopted as an effective algorithm to control the localization and formation. It investigates the effect of formations on reconstruction accuracy of a swarm of drones patrolling a 3D environment. To exploit camera-inview side information, an ineffective configuration can be addressed. To

Aerial and underwater drone communication: potentials and vulnerabilities

15

synchronize and recognize the partner drones, an accurate 3D reconstruction is applied. A SLAM for MAVs in indoor environments uses a fusion of monocular camera and sensors to fuse the visual information and measurements from IMU. It is estimated through an EKF to obtain better inference of degrees of freedom of local environment. The indoor waypoint navigation is presented to process the information of a RGB-D sensor and an IMU to examine predefined positions of an indoor location, approximates orientation and position of UAV. A PD controller is adopted to predetermine the UAV orientations. A method fuses artificial vision techniques with artificial neural networks (ANN) to achieve autonomous landing of a quadrotor drone. It is identified through reference points identifying a landing area near the finishing point of its planned path. Images from a mounted camera feed two cascaded ANNs to identify and determine the position of marker without the use of GPS. A Lucas Kanade uses optical flow interpreter to track the marker based on image characteristics. A method for autonomous navigation of UAV is presented in forest environments based on detection and evasion of trees. A deep neural network is trained using a limited database to learn an Alexnet to detect the obstacles. Due to rigorous development, autonomous navigation is more realistic goal.

7  Vision-based technologies A solution for indoor autonomy using visual marking can be determined using vision-based multirobot quadrotor. It represents the architecture level approach to navigate the obstacles like other members of multirobot systems. It also presents the software architecture, which includes multiple modules such as mission scheduler, supervisor, trajectory planner and controller, obstacle detector, localization pose estimator, and visual detector. UAV can also be steered visually in an unknown environment to estimate the set of all possible 3D motion parameters. UAV can undergo through a single perspective on-board camera to develop the estimated parameters of motion and reconstructed visuals in order to avoid the obstacle during navigation. For a Robust Autonomous Navigation, a selective technique can be used to make better explorative decisions. It applies mapping system failure using ROS to improve the quality of the maps as well as ability to explore the environment. Drones can sense and avoid obstacles using surround stereo vision and automatic sense and avoid algorithms to address safety. An instantaneous 3D occupancy grid map is constructed using disparity

16

Bakkiam David Deebak, Fadi Al-Turjman

information from multiple stereo cameras. Importantly, noise is filtered using spatial information using temporal information. Autonomous navigation drones employs uniform pattern of canopy using an embedded vision system that drive the vehicle and take decisions in new environments with nonstandard canopy pattern. This approach is comprised of two primary steps such as nonrelevant features with uniform pattern, and Scale Invariant Feature Transform algorithm. The target position on 3D Cartesian plane in small indoor environment is estimated using Unscented Kalman filter. A small UAV mounted with a single board computer and a frontal camera is used to project an oval trajectory at fixed height. Azimuth and elevation is obtained to process the images along with the drone position and orientation. Opti-track motion system and ROS retrieves the position of the drone. A framework for direct vision controls the autonomous drones flying in indoor environments. It has a velocity to measure an optical flow sensor and precision navigation using a sequence of obstacles from a mounted stereo camera.This framework is implemented on a quadrotor carrying an on-board vision-processing computer. A UAV-based aid for accidents primarily identify vehicle through visuals to investigate the real-time scene using ground control service, that is, for waypoint navigation. It has an ability to identify the victim's vehicle and determine the position. For vehicle detection, cascaded Haar classifier is used.“mapKITE”- a geodata acquisition system is presented to comprise UAV and terrestrial vehicle. UAV and ground vehicle are linked through a “virtual tether.” UAV seems like a “mapping kite.” Optical cameras, sensors, and technologies based on IMU and GNSS are used as a novel “Point-and-scale” photogrammetric measurementbased system. It is introduced to install on the top of terrestrial vehicle [69].

8  UAV attacks UAVs operations primarily affect the physical configurations and maneuvers. It is not easy to detect, and countermeasures involving proper capturing of vulnerabilities. Based on the factors, UAV attack is categorized into four major parts, namely, UAV freezing, waypoint alterations, enforced collision, and UAV hijacking. The major adversaries cause a simple attack scenario showing on a single node.The attack details are given in as follows: UAV freezing: It refers to the node failure caused due to alterations in physical configurations of the UAVs directly affecting its maneuvers. UAV freezing causes mobility loss leading to network failures. These attacks are caused from the intrusion, signal jamming, and session hijacking.

Aerial and underwater drone communication: potentials and vulnerabilities

17

Waypoint alterations: UAVs are fully functional that might be under the threat of waypoint modifications. Waypoint modifications cause overlapping of mobility patterns leading to enforced collision.This attack is fatal, and thus difficult to identify and trace their effect. Enforced clustering: Unlike waypoint alterations, this causes UAVs to form subclusters, and create the own subnetwork that operates alternatively to the existing networks. It is used to provide the significant information regarding patterns and configurations of existing UAVs. UAV hijacking: It refers to capture UAVs from a remote location not physically but connective-based. UAVs operate initially under the control of the third party to override the instructions if required.

9  Phases of UAV networks These attacks have been identified as a new set of vulnerabilities that UAV network faces more difficult to operate in cooperative mode. These attacks can either affect a single link between two nodes or can affect the whole networks.These types of attack are divided into three phases, namely, identification phase, session break/creating vulnerabilities, and attack phase. The description of each phase is given as follows: Identification phase: It is the initial phase which starts with the network to operate the entire network to fetch the network data using existing approaches such as session hijacking, and other possible vulnerabilities. Session break and creating vulnerabilities: Most of the networks are prone to network attacks to acquire a session to launch any one of the abovelisted attacks. During this phase, a new set of codes can be used to launch cyberattack. Attack Phase: Once an attack is launched, the cyberattack compromises the network to revoke the session. The main objective of a network controller is to prevent a network from undergoing the state of cyberattack during network mission.

10  UAV challenges and main issues in underwater RF communication UAVs are very expensive network that cannot afford a lot of damage in terms of cost, nation sovereignty, and life. It may cause a devastating effect to introduce the concept of cyberattack in order to counter-measures while developing and deploying UAV networks. The issues and challenges

18

Bakkiam David Deebak, Fadi Al-Turjman

faced by a network are prone to cyberattack. Network performance should not be affected while taking countermeasures against the attacks. The network system should be capable of identifying nodes and errors in case of compromised network. Network restoration should be fast and safe during recovery phase. Dynamic topology offers large benefits to the networks that may cause hindrance as identification of fault with rapid topology change. A selection of appropriate bands for data sharing is one of the key issues with these networks. Battery consumption for attack-reducing computations is one of the major challenges to affect UAV proximity and decreases the level of coordination. Increases computation complexity may lead to network failures. These attacks decrease network adaptability, which might become unresponsive. This increases the counterfeit possibilities, which can cause an attack on a ground station that controls and manages the coordination between aerial vehicles. RF waves probably affect several factors depending on water channel properties, such as different density level which can be varies with temperature, high permittivity, electrical conductivity, and salinity affected by turbidity, underwater ambient conditions. During UWRF, multipath propagation is the most influential phenomena, which has a direct impact on RF propagation from water into air.The refraction angle considers RF signaling to cross air–water boundary layer patching through an antenna [4]. Electromagnetic waves use a limited range of water communication which could improve to implement a long range communication in Deep Ocean using specific antenna features. The size of large antenna requires RF signal propagation between terrestrial and underwater communication. The magnetic types of antennas are the most compact to apply in practical solution [20]. RF technology scheme in underwater consider antenna design, transmitting power, bandwidth and noise as major factors to resolve.

11  Discussions and open issues Attacks on UAVs have already been studied by many researchers. Most of them have focused on cyberattack, which directly affects the performance of network forming between the UAVs. The simulation study presents the impact of such attacks focusing for possible elimination over UAVs. Apart from the study and analysis, there are several open issues to understand the detailed impact of cyberattack.This is included as the impact of communication standard between the UAVs. The cyberattack need to evaluate in the presence of other cyberattacks and network threats such as

Aerial and underwater drone communication: potentials and vulnerabilities

19

Sybil attack, impersonation attack, wormhole attack, node capturing, eavesdropping, sinkhole attack, etc. Apart from the impact of attacks, an optimistic approach is needed to develop the counterfeit against vulnerabilities and security concerns addressed by UAVs. Security analyses and formal evaluations are emphasized to conduct an analysis over networked UAVs. The effect of such attacks should be studied over performance metrics to evaluate the issues of security approaches in order to prevent cyberattacks.

11.1  Most common Issues in underwater acoustic communication The main issues in UWAC affect communication links through man-made noises, path and multipath losses, Doppler spread, high and variable propagation delay. The temporal and spatial variability of the acoustic channel are determined to limit the communication range, frequency, and bandwidth of acoustic communication link. Bandwidth requires few kHz to establish short-range communication. These factors lead to low bit rates, that is, for acoustic nodes in order to minimize the power consumption, huge data, and inexpensive computation [31]. Underwater sensor networks is a possible technique to improve the possible communication to create a large scale network demand, and inefficient power backup [9].

11.2  Applicability and applications The navigation systems on an autonomous quadrotor using smartphone can be designed. It is consisted of low-level navigation, control layer and a high-level human-robot communication layer. At low cost sensors, position, velocity, and height are estimated using nonlinear complementary filters. Human–Robot interface can effectively define remote waypoint navigation, and its performance is better than traditional, PX4 controller-based approach. AR-Drone can be used as a robotic platform usable for research and education [39]. It can perform tasks like position stabilization, autonomous navigation, and object tracking, which are very basic to mobilize the external localization system and its performance cane be measured in an autonomous surveillance task. A new sensor package for small UAVs and a successful application plan to work in large outdoor structures that use mapping or 3D measurement of landscape and supervising to cooperate with different drones. Sensor aided video coding technique uses aerial motion imagery to reconstruct SFM 3D. It uses homograph to start the block matching algorithm, which accepts a strong motion estimation and a tiny search window

20

Bakkiam David Deebak, Fadi Al-Turjman

to reduce the complexity using a low-complexity sensor aided H.264 encoder. A spatial exploration and obstacle detection system provides automotive functionalities as a wearable/portable device to develop INSPEX [68]. It detects and localizes in real-time static and mobile obstacles under various environmental 3D conditions. A human navigation in low visibility (e.g., for firefighters) determines the obstacle to avoid system of navigation, that is, for visually impaired people. A drone can be navigated autonomously using a cheap quadcopter [13]. It monitors the motion to locate a dense reconstruction using LSD-SLAM. It uses Parrot Bebop MAV for vision-based autonomous search and navigation. A hybrid, citywide urban navigation system for mobile agents is dedicated to support the navigation features. It combines GPS and communication system between vehicles with an ad hoc network of parked cars, and RFID as a permanent structure to enable a safe navigation in the city. It is explained to use a cooperating group of intelligent UAVs in a network that relies on communication technologies in search and rescue situation. GPS is used to locate UAVs in order to increase location estimation coverage area. A semiautonomous drone uses video conferencing to connect a desktop that makes mobile video conferencing very different in real-time practice. It elaborates about the value of bringing modular design to drone delivery system. Ann optimized method is proposed for task management of a group, that is, modular delivery drones. It compares the method with dynamic programming and a stable management strategy. An UAVs control is compromised without detection of ground station that changes the sensor value to modify the navigation. It demonstrates MAVR as a defensive technique to control the ROP attacks. Drone surveillance system can be implemented to sense the behavior of smartphone devices.The system provides the participants to record the activities of anonymous drones through partial flight path measurements physically embedded in smartphones.

11.3  Applications based on machine learning A drone in forest environment autonomously uses sense-and-avoid system. It is based on AlexNet, a skilled deep neural network.The autonomous flight extracts the instructions based on classification output comprehended by a state machine corresponding to three possible values—free-space, obstacleclose, and obstacle-very-close. The use of heterogeneous Asymmetric multi processing techniques in IoT scenario considers a smart device in which real-time acquisition and processing capabilities are necessary: a Drone. It can be highly useful to reduce the disaster risks, that is, floods. It uses DSM

Aerial and underwater drone communication: potentials and vulnerabilities

21

to generate digital elevation model which can be processed using classical photogrammetry to modern CV. The method of quantitative assessment exploits the risk of navigation information system, that is, air drones. It is possible to measure the probability of system failure without system failure. To assess the exploitation risk of navigation information system an appropriate mathematical model is suggested. An algorithm is designed to assess the information security using AR Drone 2.0. Dos attack tools are used as LOIC, Netwox, and Hping3 to exploit a drone behavior. WiDrone is a system that uses an antihijacking technique on commodity drones which controls the Wi-Fi fingerprint to validate the location. It demonstrates the effectiveness of authentication algorithm mitigate against spoofing attack. A system to search and rescue victims uses wireless UWB technology for precise localization. UAV navigation in an indoor environment avoids any contact with humans, occupying the same area, while preventing any injuries. The system relies on vanishing point algorithm, the HOG descriptors for sensing pedestrians using SVM classifier and the hough transform to find the research solution.

12  Conclusion and future work In this paper, the basics of a quadrotor UAV have been reviewed and several elements related to UAV including different sensors, applications, and their advantages have thoroughly been surveyed to signify importance of navigation. It started with the review of autonomous navigation of UAVs using various techniques related to indoor and outdoor environment. It discussed about communication aspects of drones, different algorithms like LSD-SLAM, HOG, to communicate. According to the environment, 3D images form different sensors, use of machine learning, and artificial intelligence in the working aspects of drone communication. The use of artificial intelligence and machine learning for the autonomous navigation of a drone proves the efficiency of optimal path to discuss the application of UAV in various research fields. Finally, the safety measures like information security, system failures which can occur during the flight measuring for fine navigation system. A quadcopters path planning and SLAM resulted in specific area of a UAV like take-off and hover stabilization that would aid in finding new practical applications. Moreover, the inclusion of an automatic return feature in a drone solves the complexities of energy efficiency by adding and optimizing performances of modern cyberattack causing node mobility. Multiple

22

Bakkiam David Deebak, Fadi Al-Turjman

variants of cyberattack lead to devastating effect on network operations. As such attacks raise serious issues in the deployment of aerial nodes on flying network. Efficient solutions are required to keep these networks aloof from such attacks. Protocols and architectures are required to embed properties to counteract the cyberattacks. Further investigations and studies are required to analyses all the factors that cause vulnerabilities leading to cyberattack. Implementation analysis and possible remedies have been presented for UWC technology that enables a platform to build up a network connection between underwater devices with terrestrialbased station. The main objective was to provide solution of challenges of underwater channel with particular reference to monitor the applications for water medium. The different water channels and communication links have crucial properties to challenge the possible solutions in order to deploy the UWAC and UWRF communication along with UWOC. The main outline of this chapter was to encourage research efforts and development of new advanced communication techniques. In addition, this chapter reviewed technical issues, and challenges in underwater networks and communication of entire technologies toward the next generation wireless networks.

References [1] S. Hayat, E.Yanmaz, R. Muzaffar, Survey on unmanned aerial vehicle networks for civil applications: a communications viewpoint, IEEE Commun. Surv. Tut. 18 (4) (2016) 2624–2661. [2] M. Erdelj, E. Natalizio, K.R. Chowdhury, I.F. Akyildiz, Help from the sky: leveraging UAVs for disaster management, IEEE Pervas. Comp. 16 (1) (2017) 24–32. [3] H. Huang, A.V. Savkin, Towards the Internet of flying robots: a survey, Sensors 18 (11) (2018) 4038. [4] S.M. Adams, C.J. Friedland, A survey of unmanned aerial vehicle (UAV) usage for imagery collection in disaster research and management, in: Proceedings of the 9th International Workshop on Remote Sensing for Disaster Response, Stanford, CA, USA, September 15–16, 2011, p. 8. [5] C. Zhang, J.M. Kovacs, The application of small unmanned aerial systems for precision agriculture: a review, Precis. Agric. 13 (2012) 693–712. [6] A. Bhowmick, S.M. Hazarika, An insight into assistive technology for the visually impaired and blind people: state-of-the-art and future trends, J. Multimodal User Interf. 11 (2) (2017) 149–172. [7] M. Haghani, M. Sarvi, Crowd behaviour and motion: empirical methods, Trans. Res. B: Methodol. 107 (2018) 253–294. [8] S. Rajeev, et al. Augmented reality-based vision-aid indoor navigation system in GPS denied environment, in: Mobile Multimedia/Image Processing, Security, and Applications 2019, vol. 10993, International Society for Optics and Photonics, 2019. [9] A.A. Zhilenkov, I.R. Epifantsev, Problems of a trajectory planning in autonomous navigation systems based on technical vision and AI, in: IEEE Conference of Russian Young Researchers in Electrical and Electronic Engineering, IEEE, 2018.

Aerial and underwater drone communication: potentials and vulnerabilities

23

[10] H. Shakhatreh, A.H. Sawalmeh, A.I. Al-Fuqaha, Z. Dou, E.K. Almaita, I.M. Khalil, et al., Unmanned aerial vehicles (UAVs): a survey on civil applications and key research challenges, IEEE Access 7 (2019) 48572–48634. [11] A.A. Khuwaja,Y. Chen, N. Zhao, M. Alouini, P. Dobbins, A survey of channel modeling for UAV communications, IEEE Commun. Surv. Tutor. 20 (2018) 2804–2821. [12] W. Khawaja, I. Guvenc, D.W. Matolak, U. Fiebig, N. Schneckenberger, A survey of airto-ground propagation channel modeling for unmanned aerial vehicles, IEEE Communications Surveys & Tutorials 21 (3) (2019) 2361–2391. [13] Z. Ullah, F. Al-Turjman, L. Mostarda, Cognition in UAV-aided 5G and beyond communications: a survey, IEEE Trans. Cogn. Commun. Netw., 2020, https://doi.org/10.1109/ TCCN.2020.2968311. [14] N.H. Motlagh, T. Taleb, O. Arouk, Low-altitude unmanned aerial vehicles-based Internet of Things services: comprehensive survey and future perspectives, IEEE Internet Things J. 3 (2016) 899–922. [15] L. Gupta, R. Jain, G.Vaszkun, Survey of important issues in UAV communication networks, IEEE Commun. Surv. Tutor. 18 (2016) 1123–1152. [16] S. Hayat, E. Yanmaz, R. Muzaffar, Survey on unmanned aerial vehicle networks for civil applications: a communications viewpoint, IEEE Commun. Surv. Tutor. 18 (2016) 2624–2661. [17] H. Khayyam, et al. Artificial intelligence and Internet of Things for autonomous vehicles, Nonlinear Approaches in Engineering Applications, Springer, Cham, (2020) 39-68. [18] E.F.Z. Santana, et al., Software platforms for smart cities: concepts, requirements, challenges, and a unified reference architecture, ACM Comput. Surv. 50 (6) (2018) 78. [19] E. Ismagilova, et al. Smart cities: advances in research—an information systems perspective, Int. J. Info. Manag. 47 (2019) 88–100. [20] C. Chang, S.N. Srirama, R. Buyya, Mobile cloud business process management system for the Internet of Things: a survey, ACM Comput. Surv. 49 (4) (2017) 70. [21] Y. Lu, S. Papagiannidis, E. Alamanos, Internet of Things: a systematic review of the business literature from the user and organisational perspectives, Technol. Forecast. Social Change 136 (2018) 285–297. [22] B.D. Deebak, F. Al-Turjman, A hybrid secure routing and monitoring mechanism in IoT-based Wireless Sensor Networks, Ad Hoc Netw.J. 97 (2020)doi: 10.1016/j.adhoc.2019.102022. [23] B.D. Deebak, F. Al-Turjman, M. Aloqaily, O. Alfandi, An authentic-based privacy preservation protocol for smart e-Healthcare systems in IoT, IEEE. Access 7 (2019) 135632– 135649, doi: 10.1109/ACCESS.2019.2941575. [24] B.D. Deebak, Mutual authentication scheme for multimedia medical information systems, Multimedia Tool Appl. 76 (8) (2017) 10741–10759. [25] F. Al-Turjman, et al. Seamless key agreement framework for mobile-sink in IoT based cloud-centric secured public safety sensor networks, IEEE Access 5 (2017) 24617–24631. [26] B. Li, Z. Fei,Y. Zhang, UAV communications for 5G and beyond: Recent advances and future trends, IEEE Internet Things 6 (2) (2018) 2241–2263. [27] H. Ullah, et al. 5G communication: an overview of vehicle-to-everything, drones, and healthcare use-cases, IEEE Access 7 (2019) 37251–37268. [28] G. Chopra, R.K. Jha, S. Jain, A survey on ultra-dense network and emerging technologies: security challenges and possible solutions, J. Network. Comp. Appli. 95 (2017) 54–78. [29] H. Sedjelmaci, S.M. Senouci, N. Ansari, A hierarchical detection and response system to enhance security against lethal cyber-attacks in UAV networks, IEEE Trans. Syst. Man Cybernet. Syst. 48 (9) (2017) 1594–1606. [30] F. Al-Turjman, Intelligence and security in big 5G-oriented IoNT: an overview, Future Gener, Comput. Syst 102 (1) (2020) 357–368.

24

Bakkiam David Deebak, Fadi Al-Turjman

[31] B. Li, et al. Distributed host-based collaborative detection for false data injection attacks in smart grid cyber-physical system, J. Parall. Distrib. Comp. 103 (2017) 32–41. [32] T.C. Luciani, et al., Use of drones to assist with insurance, financial and underwriting related activities, U.S. Patent Application No. 14/843,455. [33] E. Felemban, et al. Underwater sensor network applications: a comprehensive survey, Int. J. Distrib. Sensor Networks 11 (11) (2015) 896832. [34] D.N. Sandeep,V. Kumar, Review on clustering, coverage and connectivity in underwater wireless sensor networks: a communication techniques perspective, IEEE Access 5 (2017) 11176–11199. [35] M.C. Domingo, An overview of the internet of underwater things, J. Network Comput. Appl. 35 (6) (2012) 1879–1890. [36] M.M.M. Manhães, et al., UUV simulator: a gazebo-based package for underwater intervention and multi-robot simulation, in: OCEANS 2016 MTS/IEEE Monterey, IEEE, 2016. [37] Z. Zeng, et al. A survey of underwater optical wireless communications, IEEE Communications Surveys & Tutorials 19 (1) (2016) 204–238. [38] R. Kapoor, R. Gupta, R. Kumar, S. Jha, New scheme for underwater acoustically wireless transmission using direct sequence code division multiple access in MIMO systems, Wireless Networks (2018) 1–13. [39] A. Ahmed, M. Younis, Optimized beam selection for efficient long range underwater acoustic communication, in: IEEE International Conference on Communications, IEEE, 2017. [40] F. Al-Turjman, 5G-enabled devices and smart-spaces in social-IoT: an overview, Futu. Generat. Comp. Syst. 92 (2019) 732–744. [41] F. Al-Turjman, E. Ever, H. Zahmatkesh, Small cells in the forthcoming 5G/IoT: traffic modelling and deployment overview, IEEE Commun. Surv. Tutor 21 (1) (2019) 28–65. [42] M.F. Ali, D.N.K. Jayakody,Y.A. Chursin, S. Affes, S. Dmitry, Recent advances and future directions on underwater wireless communications, Archives of Computational Methods in Engineering (2019) 1–34. [43] G. Chen, et al.Wireless power and data transmission system of submarine cable-inspecting robot fish and its time-sharing multiplexing method, Electronics 8 (8) (2019) 838. [44] J. Wu, et al. Influence of pulse shaping filters on PAPR performance of underwater 5G communication system technique: GFDM,Wireless Commun. Mobile Comput. (2017) (2017). [45] X. Cheng, M. Zhang, M. Wen, L. Yang, Index modulation for 5G: striving to do more with less, IEEE Wireless Commun. 25 (2) (2018) 126–132. [46] M. Lanzagorta, Underwater communications, Synth. Lect. Commun. 5 (2) (2012) 1–129. [47] M.H. Gholizadeh, A.M. Melesse, L. Reddi, A comprehensive review on water quality parameters estimation using remote sensing techniques, Sensors 16 (8) (2016) 1298. [48] Z. Farid, R. Nordin, M. Ismail, Recent advances in wireless indoor localization techniques and system, J. Comp. Networks Commun. (2013) (2013). [49] F. Al-Turjman, A novel approach for drones positioning in mission critical applications, Trans. Emerg. Telecommun. Technol. (2019), https://doi.org/10.1002/ett.3603 [50] G. Verhoeven, et al. Mapping by matching: a computer vision-based approach to fast and accurate georeferencing of archaeological aerial photographs, J. Archaeol. Sci. 39 (7) (2012) 2060–2070. [51] X. Zhang, A.B. Rad,Y.-K. Wong, Sensor fusion of monocular cameras and laser rangefinders for line-based simultaneous localization and mapping (SLAM) tasks in autonomous mobile robots, Sensors 12 (1) (2012) 429–452. [52] R. Li, et al., LIDAR/MEMS IMU integrated navigation (SLAM) method for a small UAV in indoor environments, in: DGON Inertial Sensors and Systems (ISS), IEEE, 2014.

Aerial and underwater drone communication: potentials and vulnerabilities

25

[53] R. Cucchiara, A. Prati, R. Vezzani, A multi-camera vision system for fall detection and alarm generation, Expert Syst. 24 (5) (2007) 334–345. [54] A. Loquercio, et al. Dronet: learning to fly by driving, IEEE Robot. Automat. Lett. 3 (2) (2018) 1088–1095. [55] L. Dong, C. Ma, G. Lachapelle, Implementation and verification of a software-based IF GPS signal simulator, in: Proceedings of the US Institute of Navigation NTM, San Diego, January 26–28, 2004, 378–389. [56] Z.Wang, H. Liu, S.J. Fang,Y. Cao, A low-cost dual-wideband active GNSS antenna with low-angle multipath mitigation for vehicle applications, Prog. Electromagnet. Res. 144 (2014) 281–289. [57] N. Monterrosa, J. Montoya, F. Jarquín, C. Bran. Design, development and implementation of a UAV flight controller based on a state machine approach using a FPGA embedded system, in: IEEE/AIAA 35th Digital Avionics Systems Conference, September 2016, IEEE, pp. 1–8. [58] S. Sato, T. Anezaki, Autonomous flight drone for infrastructure (transmission line) inspection (2), in: International Conference on Intelligent Informatics and Biomedical Sciences, November 2017, IEEE, pp. 294–296. [59] J. Colorado, M. Perez, I. Mondragon, D. Mendez, C. Parra, C. Devia, et al. An integrated aerial system for landmine detection: SDR-based ground penetrating radar onboard an autonomous drone, Adv. Robot. 31 (15) (2017) 791–808. [60] B.H.P. Maathuis, J.V. Genderen, A review of satellite and airborne sensors for remote sensing based detection of minefields and landmines, Int. J. Remote Sensing 25 (23) (2004) 5201–5245. [61] F. Ghallabi, A. Kawamura,Y. Pyo, T. Tsuji, R. Kurazume, Control architecture for service drone in informationally structured environment, in: IEEE/SICE International Symposium on System Integration, December 2015, IEEE, pp. 611–618. [62] Z. Wang, R. Ding, The algorithm design and implementation of optical navigation device for UAV, in: IEEE 2nd Information Technology, Networking, Electronic and Automation Control Conference, IEEE, 2017. [63] M. Nitsche,T. Pire,T. Krajník, M. Kulich, M. Mejail, Monte carlo localization for teachand-repeat feature-based navigation, in: Conference Towards Autonomous Robotic Systems, September 2014, Springer, Cham, pp. 13–24. [64] T. Pire, T. Fischer, J. Civera, P. De Cristóforis, J.J. Berlles, Stereo parallel tracking and mapping for robot localization, in: IEEE/RSJ International Conference on Intelligent Robots and Systems, 2015, IEEE, 2015, pp. 1373–1378. [65] G. Cai, J. Dias, L. Seneviratne, A survey of small-scale unmanned aerial vehicles: recent advances and future development trends, Unmanned Systems 2 (2) (2014) 175–199. [66] F. Ghallabi, A. Kawamura,Y. Pyo, T. Tsuji, R. Kurazume, Control architecture for service drone in informationally structured environment, in: IEEE/SICE International Symposium on System Integration, 2015, IEEE, pp. 611–618. [67] W.G. Aguilar, C. Angulo, R. Costa-Castello. Autonomous navigation control for quadrotors in trajectories tracking, in: International Conference on Intelligent Robotics and Applications, 2017, Springer, Cham, pp. 287–297. [68] C. Paukovits, H. Kopetz, Concepts of switching in the time-triggered network-on-chip, in: 14th IEEE International Conference on Embedded and Real-Time Computing Systems and Applications, 2018, IEEE, pp. 120–129. [69] J. Zhang,Y.Wu,W. Liu, X. Chen, Novel approach to position and orientation estimation in vision-based UAV navigation, IEEE Trans. Aerospace Electronic Syst. 46 (2) (2010) 687–700. [70] K.-L. Wong, C.-H. Chang, Y.-C. Lin, Printed PIFA EM compatible with nearby conducting elements, IEEE Trans. Antennas Propagat. 55 (10) (2007) 2919–2922.

CHAPTER TWO

Machine learning for cyber security frameworks: a review Ezekiel T. Ogidan, Kamil Dimililer, Yoney Kirsal-Ever

Research Centre for AI and IoT, Near East University, Nicosia, Cyprus, Mersin, Turkey

Chapter outline 1 Introduction 2 Machine learning tasks 3 Cyber security tasks 4 Machine learning approaches for cyber security 5 The effectiveness of machine learning in cyber security 6 Conclusion References

27 28 29 30 32 35 35

1  Introduction The most common applications for machine learning have been image recognition and natural language processing. Extensive research has been done and is still ongoing with regards to these topics. However, there has been recent research that looks to opportunities to leverage machine learning in the development of security protocol frameworks. In recent times, considering the large amounts of data being processed and transmitted, coupled with the rising threats to the confidentiality and integrity of all this data, machine learning appears to be a useful tool in providing security online [16]. Between 2016 and 2019, there has been more concern and awareness of data breaches and hacks. These hacks are now, more than before making their way to public news and making people more aware of the importance of cyber security. The effective automation of the detection and analysis of cyberattacks has been a major goal in the world of cyber security and in light of these recent developments, the need is even more pressing. This makes the idea of implementing autonomous cyber security protocols all the more enticing. The core purpose of research on machine learning applications for cyber security is to employ the cognitive capabilities of machine learning to automate intrusion detection and forensic analysis of security breaches. Drones in Smart-Cities. http://dx.doi.org/10.1016/B978-0-12-819972-5.00002-1 Copyright © 2020 Elsevier Inc. All rights reserved.

27

28

Ezekiel T. Ogidan, Kamil Dimililer, Yoney Kirsal-Ever

Most already existing applications of machine learning to cyber security are implemented as a subsection of network operation centers (NOC) and Security Operation Centers (SOC) primarily for detection and analysis and a fully autonomous cyber security model has not yet been achieved [7]. In our study of machine learning oriented security protocol frameworks, we would first discuss the machine learning taxonomy to highlight the general tasks that employ the use of machine learning and the machine learning approaches that are used to perform these tasks. We would also categorize the tasks involved in cyber security and its points of concern.We would then show the intersection between these two fields, pointing out how machine learning tools have been applied to provide network security and current systems that already apply machine learning to cyber security. Finally, we would discuss the efficiency of the current cyber security frameworks that employ machine learning and propose improvements that could be used to increase the efficiency of the frameworks.

2  Machine learning tasks Machine learning tasks fall under a number of generalized categories for which different algorithms are applied. These categories are applied to solve different kinds of problems in different applications. Some of these tasks, which have become correlated to some cyber security tasks include: • Regression: Making deductions based on a precondition. Based on the specifics of the task, this can be seen as prediction. Some machine learning methods used to solve this task include linear regression, polynomial regression, ridge regression, decision trees, support vector regression and random forest. Deep learning can also be applied to regression tasks. • Classification: This task involves separating data unit or items into categories based on some attribute of the data. The supervised learning approach is usually used for this task with the programmer highlighting the distinguishing attribute. Some machine learning methods used to solve this task include logistic regression, K-nearest neighbors (K-NN), support vector machine, Kernel SVM, naive Bayes, decision tree classification, and random forest classification. There are also deep learning methods that can be applied the classification problems, although they are best suited to cases where you have more data and would consequently require more computational power to train. • Clustering: Just as in the case of the classification task, clustering also involves grouping of data units or items. The only difference in this task

Machine learning for cyber security frameworks: a review

29

is that the data are grouped based on similarity with no major attention paid to any specific attribute of the data. This is essentially the closest task to forensic analysis. Some machine learning methods used to solve this task include also K-NN, K-means, mixture models, Gaussian mixture model, hierarchical clustering, mean shift and self-organized maps. • Association rule learning: This task is similar to recommendation in that it makes a conclusion based on an analysis of a previous event. Depending on the application, this can be perceived as learning from experience. This is the approach that is used by websites like YouTube and Netflix to make recommendations based on your viewing history. Some machine learning methods used to solve this task include Apriori, Euclat, and FP-growth. • Generative models:While most machine learning tasks deal with already known data and their corresponding consequent decisions, generative models are used to simulate data based on previous decisions. Some machine learning methods used to solve this task include Markov Chains and Genetic algorithms.

3  Cyber security tasks The most important points of concern when Cyber security is discussed include intrusion detection, malware analysis, and spam detection. • Intrusion detection: The aim of intrusion detection is to detect and notify concerned agents about an intrusion. The entire architecture and system involved in intrusion detection is known as an intrusion detection system. Intrusion detection systems traditionally study patterns of known attacks and watch for these anomalies. The major focus in intrusion detection are botnets and domain generation algorithms (DGA) (Fig. 2.1). A botnet is a connection of computers infected with a bot script that allows attackers to remotely control the infected computers. The infected computers are often called zombies. Botnet detection involves identifying communications between infected hosts and the external command and control (C2) servers used in the network. DGA on the other hand are used to periodically generate domain names to be used by infected computers to communicate with external C2 servers which server. With DGAs, attackers can automatically change the domains that they use for malware attacks.This makes them more elusive to intrusion detection systems that are based on static blacklists of domain names.

30

Ezekiel T. Ogidan, Kamil Dimililer, Yoney Kirsal-Ever

Figure 2.1  A simple intrusion detection system (IDS).

• Malware analysis: This has become a bigger point of focus in cyber security because of the improvements on malware that give them polymorphic and metamorphic features. This has made traditional rule-based malware detectors useless. Traditional malware consisted of viruses, worms, logic bombs, and trojan horses. Improvements led to the development of advanced malware such as adwares, rootkits, bootkits, keyloggers and spyware. There has also been the development of remotely controlled malware, such as ransomwares and bots. Backdoors are also an example of remotely controlled malware. One of the methods used to analyze malware is to reverse engineer the code using a decompiler and debugger to decrypt the data stored by the malware and understand the how it works and its capabilities. Another way is to analyze its static properties like its metadata and the details of its file header. Also, its behavior could be analyzed in controlled experiments to understand how it works and note signatures that could be observed in future incidents. • Spam detection: This involves techniques used to identify spam emails which could potentially carry dangerous payloads. Attackers have been known to use spam emails to initiate phishing attacks. Spam emails emulate a trojan-like nature and can be used to spread malware or bots. Attackers have continued to improve and modify their techniques to avoid traditional spam detection filters (Fig. 2.2). The aforementioned points have a number of subcategories such as website defacement, data injection, and so on.

4  Machine learning approaches for cyber security The machine learning tasks shown in Section 2 can be related to the cyber security tasks described in Section 3. Table 2.1 shows the ways in which the machine learning approaches can be applied to particular aspects of cyber security.

Machine learning for cyber security frameworks: a review

31

Figure 2.2  A simple spam filter.

Table 2.1  Machine learning approaches and their applications. Machine learning approach Application to cyber security

Regression Classification Clustering Association rule learning Generative models

Fraud detection [17,18] Spam filters in email gateways, malware detection [10,19,20], and intrusion detection [8,9,15,21] Malware protection and behavior analysis [15] for implementing intrusion detection systems [11,13] Malware scanners Intrusion tests [6,12]

As stated in the table, regression learning algorithms can be applied in fraud detection. The probability of fraudulent actions can be deduced by a linear regression model which takes the total amount of suspicious transactions and their details as input features. Classification can be used in spam filters in email gateways to separate spam from other messages. Other aspects of machine learning such as natural language processing could also be applied in spam filters and email gateways. Spam filters were the first area of cyber security that machine learning was applied. Deep learning has also been applied to spam filtering [1]. Clustering can be used for malware protection to separate valid user files from malicious files. They could also be used in behavior analytics to identify normal user behavior and malicious activity on a system. Clustering is usually applied as part of a section of cyber security frameworks to perform certain tasks in the overall framework. Association rule learning is used to associate specific responses to different incidents in a system. They could be used to initiate a scan or raise flags

32

Ezekiel T. Ogidan, Kamil Dimililer, Yoney Kirsal-Ever

in case of suspicious activities. They could also be applied in risk management to update risk values for vulnerabilities. Generative models are used mainly in offensive cyber security. They can be used to generate possible input parameters to test vulnerabilities in intrusion tests. In the real world, companies have already started integrating machine learning into their cyber security protocol frameworks. For example, eSentire (formerly Versive) is a company that provides enterprise cyber security software to banks and financial institutions that uses machine learning approaches for anomaly detection in the identification of known and unknown network security threats. Its Versive Security Engine (VSE) is used in the analysis large datasets of transactions. The VSE takes banks NetFlow (this is a network protocol developed by Cisco for collecting IP traffic information and monitoring network traffic), DNS data and other information as input for analysis.The system can then notify human agents of any anomalies in the form of deviation from normal procedure or with similarities to previous known cyberattacks. Tessian also, an email monitoring software, uses natural language processing and anomaly detection to prevent data breaches and phishing attacks on financial firms. CylanceProtect, also claims to be redefining what antiviruses can and should do by taking a mathematical approach towards malware detection using patent-pending machine learning techniques instead of traditional methods involving the use of reactive signatures.

5  The effectiveness of machine learning in cyber security While there are a number of applications for machine learning in the field of cyber security, it is important to analyze the effectiveness of these machine learning applications in comparison to already existing traditional methods employed in cyber security. This is to justify the use of machine learning for these tasks. Most existing cyber security protocols that leverage machine learning in performing their tasks only apply machine learning to specific aspects of their frameworks. Other systems use machine learning to perform tasks to help as a part of a larger security frame work. Ultimately, the idea of a fully autonomous cyber security framework still requires more research and experiments.

33

Machine learning for cyber security frameworks: a review

In deciding whether to apply machine learning approaches to NOC and SOC, the choices should be evaluated with the following considerations [2]. 1. What machine learning algorithm is best suited for the application. Most of the research done on the selection of machine learning algorithms best suited for specific cyber security applications do not consider the training and testing environments for their analysis.Therefore the results could be very different under different training and testing environments leaving a gap in the efficiency of the work. 2. If the framework is aimed at general or specific threats. Most machine learning oriented cyber security frameworks are advertised as one-sizefits-all systems. However, experiments have shown that some machine learning algorithms might perform better with focus on specific threats. In an experiment [2] a Random Forest classifier is trained with a dataset consisting of 500,000 benign data flows and 5,000 malicious data flows. This data were collected from an organization that consisted of nearly 10,000 hosts.The data flows had been flagged by the enterprise network IDS, reviewed by an expert and labeled accordingly. IDSs were then developed based on the random forest classifier focusing on specific attacks as well as a baseline general approach. The F1 score, precision and recall are recorded. The results are shown in Table 2.2 [2]. 3. The frameworks vulnerability to adversarial attacks. Adversarial attacks provide innovative approaches to cyberattacks which pose a threat to machine learning oriented cyber security frameworks. They could attack the integrity of the system by producing seemingly legal activities.They could attack the availability of the system by producing many seemingly normal events Table 2.2  Results for classification of specific attacks and general attacks. Attack F1-score Precision Recall

DOS Attempt Overflow attempt SSH brute force login Suspicious DNS query Cache poisoning attempt Possible malware infection General approach (baseline)

0.9953 0.9939 0.9916

0.9938 0.9933 0.9941

0.9969 0.9949 0.9892

0.9753

0.9953

0.9586

0.9676

0.9872

0.9506

0.9587

0.9939

0.9337

0.7985

0.8727

0.7360

34

Ezekiel T. Ogidan, Kamil Dimililer, Yoney Kirsal-Ever

Table 2.3  Recall of the random forest classifier for different DGA methods and for a Deep DGA GAN. DGA method Recall

corebot cryptolocker dicrypt kraken_v2 lockyv2 pykspa qakbot ramdo ramnit simda DeepDGA GAN

1 1 0.99 0.96 0.97 0.85 0.99 0.99 0.98 0.96 0.48

classified as attacks and causing the system to raise a lot of false alarms.They could attack the privacy of the system by acquiring information of the cyber security algorithm and exploiting its vulnerabilities. Innovation has even seen the development of generative adversarial networks which are DNN that automatically generate adversarial samples to use in targeting systems [3,4]. The random forest classifier was used again to experiment with generative adversarial networks (GAN). A DGA using DNN is used to test the random forest classifier. The classifier is also used against traditional DGA methods and the recalls are compared.The results are shown in Table 2.3. 4. The need for continuous and regular retraining of the machine learning model. This is necessary to keep it up to date with the developments with cyberattacks. This is a major issue as there is limited supply of extensive labeled datasets available for retraining since most organizations are secretive with their internal network data.This leads to the development of networks trained on outdated datasets which perform poorly. Unlike some other machine learning applications like image classification or identification, the same datasets cannot be used over extended periods of time. In experiments [2], the same random forest classifier was trained with recent and outdated datasets and tested to show the effects of not having networks trained with new datasets frequently.The results are shown in Table 2.4. Table 2.4  Performance of classifier trained on outdated and recent datasets. Dataset F1-score Precision Recall

Outdated Recent

0.3306 0.8999

0.1984 0.9126

0.9913 0.8875

Machine learning for cyber security frameworks: a review

35

6  Conclusion Considering the nature of cyber security, machine learning oriented cyber security frameworks still require a lot of continuous and fresh research. Also, the idea of a fully autonomous cyber security framework is still a distant objective because of the development of adversarial attacks that particularly exploit vulnerabilities in machine learning oriented cyber security frameworks. Machine learning cyberattacks have also automated cyberattacks as well, making them more difficult to handle. To this effect, more useful research on the topic must take into consideration the shortcomings of machine learning approaches to cyber security. As of now, cyber security frameworks that leverage Machine learning serve better as an aid to cyber security experts for detection and for performability modeling or analysis of security systems [5] than as an autonomous unit although further research could make this possible eventually. Further research with focus on the considerations given in Section 5 would help close the gaps in current research and the effectiveness of current systems.

References [1] G. Tzortzis, A. Likas, Deep belief networks for spam filtering, in: 19th IEEE International Conference on Tools with Artificial Intelligence, Patras, IEEE, 2007, pp. 306–309. [2] G. Apruzzese, M. Colajanni, L. Ferretti, A. Guido, M. Marchetti, On the effectiveness of machine and deep learning for cyber security, in: 10th International Conference on Cyber Conflict, Tallinn, NATO CCD COE Publications, 2018, pp. 371–389. [3] H. Anderson, J. Woodbridge, B. Filar, DeepDGA: adversarially-tuned domain generation and detection, in: ACM Workshop on Artificial Intelligence and Security, Vienna, ACM, 2016, pp. 13–21. [4] I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, et al., Generative adversarial nets, Advances in neural information processing systems, NIPS, 2014. [5] K. Dimililer, Y. Ever, F. Ergun, A preliminary framework using BPNN for performability modelling and analyses of a Kerberos server, in: 6th International Conference on Advances Computing, Communications and Informatics, Karnataka, IEEE, 2017, pp. 698–700. [6] M. Alom,V. Bontupalli, T. Taha, Intrusion detection using deep belief networks in: National Aerospace and Electronics Conference, Dayton, IEEE, 2015, pp. 339–344. [7] H.J. Arevalo, J.E. Carmago, A survey on machine learning applications for software defined network security, IEEE Access 21 (2019) 95397–95417. [8] Y. Ever, B. Sekeroglu, K. Dimililer, Classification analysis of intrusion detection on NSLKDD using machine learning algorithms, in: International Conference on Mobile Web and Intelligent Information Systems, Cham, Springer, 2019, pp. 111–122. [9] A. Feizollah, N. Anuar, R. Salleh, F. Amalina, R. Ma’arof, S. Shamshirband, A study of machine learning classifiers for anomaly-based mobile botnet detection, Malaysian J. Comp. Sci. 26 (2013) 251–265. [10] W. Hardy, L. Chen, S. Hou, Y. Ye, A deep learning framework for intelligent malware detection, Knowl. Info. Syst. 54 (2018) 265–285.

36

Ezekiel T. Ogidan, Kamil Dimililer, Yoney Kirsal-Ever

[11] A. Javaid, Q. Niyaz, W. Sun, M. Alam, A deep learning approach for network intrusion detection system, in: 9th EAI International Conference on Bio-inspired Information and Communications Technologies (formerly BIONETICS), Brussels, ACM, 2016, pp. 21–26. [12] Y. Li, R. Ma, R. Jiao, A hybrid malicious code detection method based on deep learning, Int. J. Soft. Eng. Appl. 9 (2015) 205–216. [13] R. Pascanu, J. Stokes, M. Sanossian, A. Thomas, Malware classification with recurrent networks, in: IEEE International Conference on Acoustics, Speech and Signal Processing, Brisbane, IEEE, 2015, pp. 1916–1920. [14] M. Stevanovic, J. Pedersen, An efficient flow-based botnet detection using supervised machine learning, in: International Conference on Computing, Networking and Communications, Honolulu, IEEE, 2014, pp. 797–801. [15] P.Torres, C. Catania, S. Garcia, C. Garina, 2016, An analysis of recurrent neural networks fofr botnet detection behaviour, in: IEEE Biennial Congress of Argentina, Buenos Aires, Argentina, IEEE, 2016, pp. 1–6. [16] F. Al-Turjman, H. Zahmatkesh, An overview of security and privacy in smart cities’ IoT communications, Transactions on Emerging Telecommunications Technologies (2019) pp. e3677, doi: 10.1002/ett.3677. [17] S. Patil, V. Nemade, P. Soni, Predictive modelling for credit card fraud detection using data, in: International Conference on Computational Intelligence and Data Science, 2018, pp. 385–395. [18] L. Mercer, Fraud detection via regression analysis, Comp. Secur. 9 (1990) 331–338. [19] F. Ullah, H. Naeem, S. Jabbar, S. Khalid, M. Latif, F. Al-Turjman, et al., Cyber security threats detection in internet of things using deep learning approach, IEEE Access (2019) 124379–124389. [20] G. Dahl, J. Stokes, L. Deng, D.Yu, Larga-scale malware classification using random projections and neural networks, in: IEEE International Conference on Acoustics, Speech and Signal Processing,Vancouver, IEEE, 2013, pp. 3422–3426. [21] J. Kim, J. Kim, H. Thu, H. Kim, Long short term memory recurrent neural network classifier for intrusion detection, in: IEEE International Conference on Platform Technology and ServiceJeju, IEEE, 2016, pp. 1–5.

CHAPTER THREE

Emerging use of UAV’s: secure communication protocol issues and challenges Navid Ali Khan, Noor Zaman Jhanjhi, Sarfraz Nawaz Brohi, Anand Nayyar

School of Computing and Information Technology (SOCIT), Taylors University, Subang Jaya, Selangor, Malaysia Graduate School, Duy Tan University, Da Nang,Viet Nam

Chapter outline 1 Introduction 2 Unmanned aerial vehicles (UAVs) 3 Ground control station (GCS) 4 Types of UAVs 5 Communication protocols for UAVs 5.1 UranusLink protocol 5.2 UAVCAN protocol 5.3 MAVLink protocol 6 Critical analysis of these protocols 7 Discussion 8 Conclusion 9 Future work References

37 38 39 40 41 43 45 45 49 50 52 52 52

1  Introduction Unmanned aerial vehicles (UAVs) have experienced an outstanding development and gained fast-growing popularity throughout the past few decades.These are commonly used in military and defense applications such as surveillance, reconnaissance other security missions [1,2]. The sales of the military UAV market are expected to grow by over 60% by 2020 [3,4]. The UAVs are not only limited to military and defense, but they are also widely used in civilian applications as well, such as traffic surveillance, environmental monitoring law enforcement, disaster management, infrastructure observation, agriculture assessment [5], entertainment, photography, search, and rescue operations. Numerous studies such as [1,6–10] have highlighted the fact that soon, the utilization of UAVs for civilian purposes especially in Drones in Smart-Cities. http://dx.doi.org/10.1016/B978-0-12-819972-5.00003-3 Copyright © 2020 Elsevier Inc. All rights reserved.

37

38

Navid Ali Khan, Noor Zaman Jhanjhi, Sarfraz Nawaz Brohi, Anand Nayyar

smart cities and Internet of Things will be higher than the military uses, and this can eventually overcome the military demands in the future. Due to the widespread use and security weakness of the UAVs have made them an attractive target for hackers and attackers. A hacked UAV might for multiple negative reasons by the adversaries. Since this is an emerging technology, there are fewer studies available to address the security solutions for UAVs. Most of these solitons are just proposals, or they are in the early stages of their development processes [11].The security issues can lead to severe consequences such as loss of mission-critical [12] or essential data or even physical damage to infrastructure and human beings.The impact is both in the form of cost and society. Most of the issues and threats occur due to the security lapses in the communication protocols. This study tackles the issue of UAV security by designing and developing a secure communication protocol for UAVs.

2  Unmanned aerial vehicles (UAVs) A UAV is an autonomous or remotely controlled vehicle with no crew [13]. A UAV can be operated in two ways: using a control system and using the ground control station (GCS). In a remote control system, the user looks directly at the UAV or watches a camera mounted on the UAV and controls the camera using the controller. The UAV is controlled in real-time by transmitting the controller signals. Both the controller and the UAV are connected through a communication module that carries out communication between them with the help of communication protocols. Typically, telemetry, Wi-Fi, ZigBee, and many other networking devices are used for communication. On the other hand, GCS-based control uses a computer to connect the software with the UAV, which then carries out user-uploaded mission commands. By collecting information from various sensors installed on the UAV, GCS can monitor the UAV status, such as current altitude, distance, map location, and actual mission status [14]. Unmanned aircraft systems include different parts, consisting of sensor payloads and one or more ground control station [15,16], which are controlled by onboard or electronic equipment from the ground. Remotely piloted vehicles (RPVs) are the type of UAVs which are controlled from ground and require optimum wireless communication for this purpose. On the other hand, GCSs are required for large UAVs to enable close control of them to overcome the range and communication barriers. Today UAVs are used to assist crew members in scientific, tactical, environmental-based applications [17] and in the emergency response area.They are also used as a support system in other applications, such as military and

Emerging use of UAV’s: secure communication protocol issues and challenges

39

Figure 3.1  Examples of small UAVs. (Available from: https://www.aniwaa.com/wp-content/uploads/2019/02/drone-buying-guide-type-of-drones.jpg.)

commercial applications. UAVs are categorized by altitude range, weight, and flight endurance [18]. Usually, small UAVs are supported by ground controlling stations consisting of laptops or smartphones and other small devices that can easily be carried in backpacks. High-precision UAV cameras fitted on these UAVs can take images from the disaster area etc. and allow the crew to carry out object and structural analysis. Increasing research and developments in recent years have improved the use of UAVs in various applications. However, UAVs are still in their experimental stages, and shortage of trained crew members limits its use. According to [19], a minimum of three crew members are required to operate a UAV from a ground control station [20]. Examples of small UAVs can be seen in Fig. 3.1 while the example of large UAVs from (A–I) can be seen in Fig. 3.2

3  Ground control station (GCS) A GCS consists of a processing unit, a telemetry/telecommand module, a user control module, a wireless datalink subsystem, and a graphical user interface or command-line interface. For remote communication with a UAV, a wireless datalink subsystem is configured.The telemetry/telecommand module is configured to download onboard data from UAV as well as to upload instructions from the ground station to UAV.The user interface, consisting of a display module, is configured to display downloaded data from UAV [21].

40

Navid Ali Khan, Noor Zaman Jhanjhi, Sarfraz Nawaz Brohi, Anand Nayyar

Figure 3.2  Examples of large UAVs. (From M., Hassanalian, A. Abdelkefi, Classifications, applications, and design challenges of drones: A review, Progr. Aerospace Sci. 91 (2017) 99–131.)

The UAV ground control is important in military missions, especially and can monitor UAV from location close or inside the battlefield. GCS receives, processes, and transforms the data from UAV and transfer it to other users in the same network. Stationary UAVs are typically expensive, consisting of complex hardware and computer equipment with larger UAVs set up on a mobile vehicle or base station. The GCS needs a UAV pilot and a payload operator to capture and transfer data to other battlefield users using computer systems (for example, in a military or other search and rescue mission). Small unmanned aerial vehicles (SUAVs) are controlled by portable GCS consisting of hand controller, a laptop or a smartphone, radio frequency transceiver unit, and a controller box or joystick. An operator controls the movement of SUAV in operation using a hand controller, while another operator gathers information and analyzes received data using a laptop and send this information to other users [22]. A secure commutation link between UAV and GCS, and between the GCS and end-users is necessary to carry out these complex operations [23]. The actual GCS of a military UAV is shown in Fig. 3.3.

4  Types of UAVs There are commonly two types of UAVs named as (1) fixed-wing UAVs [25] and (2) rotorcraft [26,27]. The fixed-wing UAV is usually bigger and is more sophisticated in terms of technology and components such as

Emerging use of UAV’s: secure communication protocol issues and challenges

41

Figure 3.3  An actual ground control station (GCS) of a military UAV. (Ref. [24])

flight range, speed, high payload capability, and endurance. This implies that they are best intended for critical and major missions. While its countertype UAV rotorcraft is relatively small. This type of UAV has an outstanding hover capacity that enhances its flexibility in fulfilling a mission. Compared to fixed-wing UAVs, the strength and speed of such UAVs are limited. So, rotorcraft is suitable for ultra-low flight speed and/or restricted environment. To further fulfill the new demands and potential requirements, they are both further split as rotorcraft: (1) multirotor [28] (2) single rotor [29] and fixedwing: (1) fixed-wing (2) fixed-wing hybrid [30]. All these types have their pros and cons and their uses in different ways [31], as shown in Table 3.1. Figs. 3.4 and 3.5 shows fixed-wing and rotorcraft UAVs, respectively

5  Communication protocols for UAVs The UAV and GCS communications occur through communication protocols [35,36]. Unfortunately, the existing communication protocols are not suitable for this environment [37]. They are unable to utilize the resources properly [38] because of the limited nature of these resources and the dynamic environment of the unmanned systems. Moreover, this also complicates the problem because when dealing with security solutions, the resources are insufficient, such as battery life, real-time computation, and autonomous control. The energy resources, communication bandwidth, and computational capacity, makes the existing protocols, such as TLS and

42

Navid Ali Khan, Noor Zaman Jhanjhi, Sarfraz Nawaz Brohi, Anand Nayyar

Table 3.1  Pros and cons of the types of UAVs. Type Pros Cons

Fixed-wing Long survival Large space is required for UAV Covers large area Launch and recovery Speed of flight is fast They cannot take-off, hover or land vertically Not easy to fly, proper training is required More costly Fixed-wing Flight survival time is Lack of perfection at hybrid longer can take-off when they hover or at UAV and land vertically forwarding flight Under development Multirotor Accessibility The duration of flight UAV Easy to use time is short Can take-off and Can carry an only small hover vertically payload Very good for camera control use Best suited to operate in a confined area Single rotor Same as multirotor Risky UAV UAV Not easy to control Proper training is required More costly Ref. [31]

Figure 3.4  Fixed-wing UAV. (Ref. [32].)

Typical uses

Aerial Mapping, other mega infrastructure inspection Military operations Drone delivery

Used for aerial videography or photography or aerial inspection

Aerial light detection and ranging (LIDAR) laser scanning

43

Emerging use of UAV’s: secure communication protocol issues and challenges

Figure 3.5  Normal rotocraft UAV [33,34].

Kerberos, impractical for UAV networks [39]. There are other different communication protocols available for UAVs. Most common of them discussed briefly.

5.1  UranusLink protocol UranusLink is created to provide unreliable and reliable services as a packetoriented protocol [40]. The protocol determines the packet structure and the data representation transmitted. The overall working mechanism and description of UranusLink protocol is presented in Ref. [41] by Kriz et al. For this study, the UranusLink packet structure is derived from their work. The packet structure is shown in Fig. 3.6. There are six fields in every packet: 1. preamble (PRE), 2. sequence number (SQN), 3. message identification (MID), 4. data length (LEN), 5. data as such, and 6. checksum (CS). PRE

SQN

MID

LEN

DATA

CS

1B

2B

1B

1B

1–252 B

1B

Figure 3.6  UranusLink packet structure [41]. (Ref. [34].)

44

Navid Ali Khan, Noor Zaman Jhanjhi, Sarfraz Nawaz Brohi, Anand Nayyar

The UranusLink protocol is specially designed to be used in radio ways. In radio communications, normally data losses and wrong data receiving can occur. The first field of the packet is preamble (PRE). The data packet always starts with such a value (0xFD), often occurring in packet data to ensure the validity of a packet in the input buffer on the receiving side. The second field is sequence number (SQN). It is always an “even” number and at the end of the packet is checksum. The optimum preambles and checksums length has been chosen to achieve the balance in the given environment, load, and link capacity between protocol robustness and overhead. SQN allows a protocol to identify failures in packets and process the most current information only.The UAV must be able to identify any communication problems and respond properly if they continue. On the receiving side, if there is any missing number in the given SQN row, the packet has been lost. Or, if multipath information links are possible, these packets may arrive in the incorrect order. Because the control of the UAV with the latest data is always important, it will be dropped if a packet arrives that has a smaller SQN than the current one. MID determines data interpretation in the packet’s information segment. There are currently eight kinds of messages identified in the UAV direction and 16 in the base or control station direction. The important types are: (1) the connection from the ground station to the UAV and (2) the connection from the UAV to the ground station. There are two primary UAV modes: the flight mode and the config mode. The rotors operate in flight mode while when the UAV stops on the ground, the configuration mode is active. Robot mode switch message is sent to change mode and the recognition packet has to be checked by the other side.This is a security system which maintains that the mode is altered (e.g., engines are turned off) even if the message of change in the robot mode is lost. Moreover, the ground station maintains an SQN list of messages sent to alter the robot mode to be able to differentiate which mode is recognized if more than once the link was lost when mode was changed. The only acknowledged message is the change in robot mode, while the others are not recognized since overhead is much more important than an advantage. In-state of the art, if UranusLink is contrasted with existing protocols for interaction with UAV and low overheads, the MAVLink protocol is widely used. Nonetheless, it has up to 33% additional overhead and is not secure [41].

Emerging use of UAV’s: secure communication protocol issues and challenges

45

5.2  UAVCAN protocol UAVCAN is an open-source protocol designed to provide secure connectivity over robust vehicle networks such as CAN buses in aerospace and robotic applications. The UAVCan protocol works on publish-subscribe architecture. It has no master node, and all nodes have the same rights, which means that it has no single point of failure. The nodes exchange long payloads which fits into a single CAN frame (such as GNSS solutions, 3D vectors, etc.).This protocol also supports multiple nodes and multiple interfaces; this feature is normally required in safety concern applications. UAVCAN describes standard high-level services and communications, such as network discovery, node setup, firmware node upgrade, monitoring of node status, network-wide time synchronization, and adaptive node ID allocation (a.k.a. plug-and-play), etc. This protocol is lightweight and can be easily implemented and validated.The protocol is designed for resource-constrained and real-time systems which is suitable in case of UAVs. The MIT license provides for the implementation of high-quality open-source references [42]. The UAVCan protocol is based on CAN bus (controller area network), which works as a standard design to allow communication in other applications between devices and microcontrollers without a host computer. Originally the protocol was designed to save on copper within the automobiles for multiplex electrical wiring, but as due to the mentioned characteristics, it is also widely used in other domains [43]. Each UAVCAN node has a unique bus ID. It is the integer at interval {1 – 127}, where the value 1 is usually the autopilot or some other kind of central control unit, and the values 126 and 127 are usually a debugging or monitoring system.The value 1 is the most commonly used value of the UAVCAN node. Any unit that can communicate through MAVLink or UAVCAN has to use the same MAVLink Component ID (COMPID) number as well as the UAVCAN Node ID; otherwise, serious discrepancies may occur. In the normal case, the UAVCAN Node ID and the ID of the MAVLink component will be set to 1 (one) if a single nonredundant autopilot is available. Every message/command of an outgoing/incoming MAVLink about a UAVCAN node will have its COMPID field set to the same value as the UAVCAN node ID [43].

5.3  MAVLink protocol MAVLink is an open-source and lightweight protocol, primarily used for bidirectional communication between GCSs and UAVs. Lorenz Meier first

46

Navid Ali Khan, Noor Zaman Jhanjhi, Sarfraz Nawaz Brohi, Anand Nayyar

published MAVLink 1.0 under the LGPL license [35,44] in early 2009. In early 2017, the MAVLink 2.0 protocol [45] was published and is the current version recommended. The new version is also compliant backward with MAVLink 1.0 and makes some improvements over the previous version. Two types of messages are available in the MAVLink: (1) messages sent from the GCS to the UAV and (2) UAV messages sent to GCS concerning current vehicle status (such as location, altitude, heartbeats, and system status or information). As the MAVLink protocol is used for real-time communications, it is intended to be a lightweight protocol [35]. In the following section, the headers of the protocol of MAVLink 1.0 and the protocol headers of the new MAVLink 2.0 [45] are given. MAVLink refers to MAVLink 1.0 [46,47] in the remaining work except as otherwise indicated. 5.3.1  MAVLink 1.0 header protocol The survey study [44] by Koubaa et al. is a detailed and the only survey available on the MAVLink protocol tutorial and working mechanism. For this study, the header of MAVlink 1.0 and MAVlink 2.0 are presented from their contribution. The MAVLink 1.0 header structure is given as follow. There are eight essential fields, as shown in Fig. 3.7. The initial frame or byte is STX is equal to the unique 0XFE number in MAVLink 1.0, referring to the beginning of a MAVLink frame.The second field is (LEN) encoded as 1 byte. The third field SEQ is also encoded in 1 byte and requires a value of between 0 and 255. The SEQ will again be reset to 0 when it reaches 255 in every message created. The SQN enables the recipient to identify the lost messages. For each unmanned system, it is necessary to have its ID, especially when they are managed from a single ground station. For this purpose, the fourth frame or byte SYS ID is present. The SYS ID 255 is usually reserved for a ground station. As the system ID is encoded as 1 byte, it limits the MAVLink 1.0 to manage a maximum of only 254 ground stations. The components which transmit the signal for the fifth-byte COMPID is presented. The sixth byte is the payload message

Figure 3.7  The header structure of MAVLink 1.0 [44]. (Ref. [41].)

Emerging use of UAV’s: secure communication protocol issues and challenges

47

Table 3.2  The MAVLink header fields/bytes and their description. Field/Byte Data Working

STX

0XFE

LEN SEQ

0–255 0–255

SYS ID COMP ID Message ID Payload

1–255 0–255 0–255 0–255 bytes

Checksum (CRC)

Two bytes contents

This field defines the start of the frame and always will be 0xFE as in MAVLink 1.0 official documentation. The payload size represents the LEN value. This part of the message shows the sequence of the packet. The first message is 0, for example. It is used for MAVlink packets that are lost. This field is the unmanned system ID. This field shows what system element sends. The type of message is represented in this field. Actual message information is provided in this byte or field. This field regulated the checksum. Packet signature takes place from LSB (least significant bit) to MSB (most significant bit).

Ref. [44]

type, it’s named as Message-ID (MSGID). For example, the HEARTBEAT message ID of 0 shows that the device is alive and this is sent every one second. MSGID is the key data for parsing the payload and extracting information following the message type. The payload next to the message ID is up to 255 bytes. The last two bytes are eventually for the checksum. The cyclic redundancy check (CRC) with seed values A and B is computed respectively by CKA and CKB. The CRC guarantees no alteration of messages during its transmission, and that the sender and the receiver have the same message. The minimum and maximum message length of MAVLink 1.0 is 8 bytes (without payload) and 263 bytes (with full payload), respectively. Table 3.2 provides an overview and explanation of each MAVLink 1.0 header field. 5.3.2  MAVLink 2.0 header protocol In early 2017, the MAVLink 2.0 was published in early 2017 [45], and it is the latest version that is recommended. It contains several updates over MAVLink 1.0 and is backward compatible with it. The following section presents the MAVLink 2.0 protocol header, and then the difference between the two versions are also highlighted. The MAVLink 2.0 header is shown in Fig. 3.8.

48

STX

Navid Ali Khan, Noor Zaman Jhanjhi, Sarfraz Nawaz Brohi, Anand Nayyar

LEN

INC

CMP

FLAG

FLAG

SEQ

SYS

COMP

MSG

PAYLOAD

CHECKSUM

SIGNATURE

ID

ID

ID

(0 to 256)

(2 bytes)

(13 bytes)

(3 bytes)

Figure 3.8  The MAVLink 2.0 header structure [45]. (Ref. [44].)

The MAVLink 2.0 includes all fields that are identical to the MAVLink 1.0, but also adds new fields and increases the size of some existing fields. The first byte is the start of the message, and its value is usually 0xFD for MAVLink 2.0 (0xFE for MAVLink 1.0). Therefore the parser must first identify these characters before the rest of the MAVLink 2.0 signal fields can be processed. The next field is the payload length, which is identical to the preceding protocol. Before the message SEQ, MAVLink displays two new flags. Incompatibility flags are the first flags impacting the design of the messages. The flags indicate whether the packet needs to be taken into account when processing the packet. For instance, a 0x01 means the packet is signed, and at the end of the packet, a signature is added. The other flag is known as compatibility flags without affecting message design. It shows flags which can be ignored if they are not understood. Even if a parser cannot read the flag, it will not be able to use the signal.These may include flags that indicate the packet priority (such as high priority) because this does not affect the structure of the packet). SEQ, system ID, and COMPID are identical to the MAVLink 1.0 protocol header. Nevertheless, in the previous version, the MSGID is encoded in the previous version in 24 rather than 8 bit so that the number of possible MAVLink 2.0 messages can be increased by as much as 16,777,215. The reason why a huge range of possible messages is designed is not apparent because it is too large.The payload field may contain up to 255 bytes of data depending on the message type. The MAVLink 1.0 checksum is equal to its peer. In the end, a 13-byte field is used in MAVLink 2.0 to make sure that the connection is tamper-proof and manipulative. This feature considerably enhances MAVLink 1.0 safety elements as it enables the message to be authenticated and ensures that it comes from an authentic source. The message signature is appended when flags for incompatibility are set to 0x01. The following fields can be found in 13 bytes of the message signature: • LinkID: it is one byte and is used for the representation of the link (channel) when they are sending the packet. Link means any telemetry device (e.g.,Wi-Fi). For every channel that sent information, the LinkID is different.

Emerging use of UAV’s: secure communication protocol issues and challenges

49

• Timestamp: Since January 1, 2015 GMT, the encode sheme of timestamp is kept 6 bytes in 10-microsecond units. The timestamp increases with each message sent over the channel.The timestamps help avoid the replay attack. • Signature: The field is focused on the full message, the secret key, and timestamp, sent and is encrypted in 6 bytes for the message. The first 6 bytes (48 bites) of SHA-256 hash applied to the 2.0 signal are covered by the signature. A 32-byte shared symmetric key is saved on either side, that is, the autopilot and the ground or the MAVLinking API. The authenticated message is discarded if: (1) it is earlier compared to the previous packet which is received from the same tuple source (LinkiD, SystemID, componentID); (2) the reception’s calculated signature differs from the message signature; (3) in comparison with the local system time, if the time stamp was greater than one minute. Acceptance/refusal of the packet is executed unless the message is signed [35,44,45].

6  Critical analysis of these protocols Fundamentally, UranusLink was designed for radio ways in which data loss and wrong data receiving can happen. It includes checksum as a component to check whether the original message was received or not. The checksum can only validate whether the original message is modified or not. However, in UAVs, if an intruder reads this sensitive information can result in a mission failure. Therefore the confidentiality of the commands is crucial in UAVs and needs to be secure to make it hard for the intruder to read the packet and understand the message. Simple checksum does not ensure the confidentiality and integrity of data. UAVCan is designed for nonmission critical robots and aerospace. The original specification document of UAVCAN states that the protocol provides no shielding and is not recommended for mission-critical and safetycritical systems. As the MAVLink message is based on the header, it evaluates and classifies a message in the first field (frame) of the data packet. The initial frame STX value is therefore verified, and base on the value, it decides if it is a MAVLink packet or not. In order to increase the transfer speed and efficiency for communication, there is no direct encryption mechanism in MAVLink. Furthermore, if the message is encrypted, the header value changes and thus a system cannot recognize whether it is a MAVLink packet.This means, though MAVLink provides better communication but lacks security mechanism.

50

Navid Ali Khan, Noor Zaman Jhanjhi, Sarfraz Nawaz Brohi, Anand Nayyar

Huge empirical evidence is available for MAVLink regarding its reliability and efficiency. However, there is lack of enough empirical evidence for UranusLink and UAVCan. MAVLink is widely used and well-established protocol as compare to UAVCan and UranusLink. Loss of data and latency in MAVLink is reported than other protocols such as UAVCan and UranusLink. UAVCan has recently been proposed, and its first stable version is not yet available and is still under work [41,42]. UranusLink protocol is suitable for UAVs with small overhead. However, there is less empirical evidence regarding its applicability and enhancement. MAVLink is more scalable, allowing more concurrent systems and supports many programming languages. In contrast, UranusLink and UAVCan have no support for multiple languages and concurrency. The overall comparison of these protocols is presented in Table 3.3. Among these protocols, MAVLink is one of the most commonly used, and due to its distinguishing feature. The MAVLink protocol despite being widely used; however, it does not perform any encryption and provides no security to the payload/messages. This vulnerability can be exploited and can result in negative consequences. This research work recommends designing and developing a new security protocol for UAVs communication, which overcomes the stated issue.

7  Discussion The development of UAVs has been increased in the past few years. The main reason for this is their wide use and a large number of applications. First, UAVs were only used for military purposes, and the technology and applications were not very well known to people. However, now a day, the perspective has been changed. UAVs have become very familiar and yet exciting technology. That is why it has been adopted and used for many recreational activities like aerial photography, sports, theme parks and entertainment, personal amusement, etc. In addition to this, the UAVs have been emerging as the vital need in many large civilian applications such as rescue operations, environmental and disaster management, agriculture, monitoring, etc. The advantages of using UAVs for various civilian and military operations are vibrant and cannot be ignored. At the same time, this can be used for negative purposes, as well. Due to their architectural and/or communication weaknesses, this technology has attracted hackers and attackers to compromise their security through various vulnerabilities. There are

Emerging use of UAV’s: secure communication protocol issues and challenges

51

Table 3.3  Comparison among UranusLink, UAVCan, and MAVLink protocols. Protocols Pros Cons Gap

UranusLink • Open-source • Lightweight • Designed for aerospace and robotic applications • Supports dual and triple modular redundant transports UAVCan

• • • •

MAVLink

• • • • • • • •

• Less empirical evidence • No security • UranusLink has recently for payload. been proposed and its The first stable version is not checksum yet available mechanism • No support for multiple only checks programming languages if the • No support for original concurrent systems message was • Not scalable received Open-source • Less empirical evidence • No subtle Lightweight • Not widely used security Low latency • No support for multiple mechanism Ability to detect programming languages and overcome data • No support for loss concurrent systems • Not scalable • Designed for UAVs with small overhead • Limited encryption ability • Designed for only small data flow Widely accepted • No security mechanism • No Scalable encryption, Support for messages are multiple languages sent in open Support for format concurrent systems Large empirical evidence Lightweight Open-source Low latency

numerous studies available in the literature which mentions the misuse and the consequences when the security of a UAV is compromised. The security issues mostly are traced back to the insecure communication of these UAVs. The security attacks on UAVs are launched against communication protocols in the network. Numerous research work has been carried out to

52

Navid Ali Khan, Noor Zaman Jhanjhi, Sarfraz Nawaz Brohi, Anand Nayyar

overcome the security issues, but most of the work is only in the form proposed solution or is in the early stages of the development. Compared to other domains, there are limited communication protocols available for UAVs communication, which are only intended for this environment. As mentioned earlier, UAVs operate from a remote controller or a GCS with the help of different communication protocols such as MAVLink, UranusLink, UAVCan. Among these protocols, the MAVLink is a well-developed and deployed lightweight protocol used for communication between GCS and UAVs.The messages contain significant information about the state of UAV and certain control commands sent from GCS to the UAV. Though MAVLink providing better communication, there is no subtle mechanism for securing these messages and are prone to several security attacks. These attacks can result in serious consequences, for instance, crash land a military or civilian UAV, steal important data of a military operation, false injection of reports in a reconnaissance or search and rescue operation, and many more.

8  Conclusion This chapter presents the overall importance of UAVs in both military and civilian applications. In the past, UAVs were mostly used for military applications, but soon the utilization of the UAVs in civilian application is going to exceed military use. As the technology is in its early stages of development and new areas in the applications are exploring day by day, this has attracted hackers and attackers to compromise their security for various intended purposes. The security attacks are normally carried out against communication protocols. In this chapter, we presented different communication protocols which are intended specifically for this environment. Their structure, working mechanism, and their critical analysis has been discussed. It is identified that MAVLink is the most widely used protocols for UAVs communication. However, MAVLink, though providing better communication, lacks security mechanism to encrypt messages and can result in serious consequences. Therefore there is a need for new secure communication protocol which can overcome the stated issue.

9  Future work As identified in the literature, there is no subtle mechanism for securing the MAVLink messages and are prone to several security attacks. This can result in serious consequences, for instance, crash land a military

Emerging use of UAV’s: secure communication protocol issues and challenges

53

or civilian UAV, steal important data of a military operation, false injection of reports in a reconnaissance or search and rescue operation, and many more. One of the reasons for not applying the encryption/security is that it increases the overhead and complexity and affects the overall performance and efficiency. For noncritical applications such as pizza delivery, these might be compromised over security, but for critical-applications, especially military or search and rescue, security cannot be compromised. So, therefore, there is a need for a secure communication protocol which can ensure the required security standard sets for communication between UAVs and ground stations. We propose a secure communication protocol which is intelligent. The protocol will work with the help of an artificial intelligence agent, which will get the input from GCS and measure the criticality of the mission and then apply encryption/security accordingly to achieve both efficiency and security simultaneously. We will design this protocol in our future work.

References [1] A.S. Saeed, A.B. Younes, C. Cai, G. Cai, A survey of hybrid unmanned aerial vehicles, Prog. Aerosp. Sci. 98 (2018) 91–105. [2] F. Al-Turjman, J.P. Lemayian, S. Alturjman, L. Mostarda, Enhanced deployment strategy for the 5G Drone-BS using artificial intelligence, IEEE Access 7 (2019) 75999–76008. [3] J.R. Wilson, UAV roundup, Aerospace America, 2013, pp. 26–36. [4] C. Drubin, UAV market worth $8.3 B by 2018, 2013. [5] V. Puri, A. Nayyar, L. Raja, Agriculture drones: a modern breakthrough in precision agriculture, J. Stat. Manag. Syst 20 (4) (2017) 507–518. [6] J.T.K. Ping, A.E. Ling, T.J. Quan, C.Y. Dat, Generic unmanned aerial vehicle (UAV) for civilian application-A feasibility assessment and market survey on civilian application for aerial imaging, in: 2012 IEEE Conference on Sustainable Utilization and Development in Engineering and Technology, 2012, pp. 289–294. [7] T. Skrzypietz, Unmanned aircraft systems for civilian missions. BIGS, 2012. [8] M. Saleh, N.Z., Jhanjhi, A. Abdullah, Proposing a privacy protection model in case of civilian drone, in: 13th International Conference on Mathematics, Actuarial Science, Computer Science and Statisitics, 2019. [9] A. Nayyar, B.-L. Nguyen, N.G. Nguyen, The Internet of Drone Things (IoDT): future envision of smart drones, in: First International Conference on Sustainable Technologies for Computational Intelligence, 2020, pp. 563–580. [10] F. Al-Turjman, S. Alturjman, 5G/IoT-enabled UAVs for multimedia delivery in industry-oriented applications, Multimed. Tools Appl. (2018) 1–22. [11] R. Hamsavahini, S. Varun, S. Narayana, Development of light weight algorithms in a customized communication protocol for micro air vehicles, Int. J. Latest Res. Eng.Technol. (2016) 73–79. [12] F. Al-Turjman, A novel approach for drones positioning in mission critical applications, Trans. Emerg. Telecommun. Technol. (2019) e3603. [13] A.C. Watts, V.G. Ambrosia, E.A. Hinkley, Unmanned aircraft systems in remote sensing and scientific research: classification and considerations of use, Remote Sens 4 (6) (2012) 1671–1692.

54

Navid Ali Khan, Noor Zaman Jhanjhi, Sarfraz Nawaz Brohi, Anand Nayyar

[14] J.A. Marty, Vulnerability analysis of the MAVlink protocol for command and control of unmanned aircraft (No. AFIT-ENG-14-M-50). Air force institute of technology wright-patterson AFB OH graduate school of engineering and management, 2013. [15] S.G. Gupta, M.M. Ghonge, P.M. Jawandhiya, Review of unmanned aircraft system (UAS), Int. J. Adv. Res. Comput. Eng. Technol. 2 (4) (2013) 1646–1658. [16] K.P. Valavanis, G.J. Vachtsevanos, Handbook of Unmanned Aerial Vehicles, vol. 1, Springer, Dordrecht, (2015). [17] F. Al-Turjman, H. Zahmatkesh, I. Al-Oqily, R. Daboul, Optimized unmanned aerial vehicles deployment for static and mobile targets’ monitoring, Comput. Commun. 149 (2020) 27–35. [18] K. Dalamagkidis, Classification of uavs, in: Handbook of unmanned aerial vehicles, 2015, pp.83–91 [19] K. Pratt, R.R. Murphy, S. Stover, C. Griffin, Requirements for semi-autonomous flight in miniature uavs for structural inspection, AUVSI’s Unmanned Systems North America, Association for Unmanned Vehicle Systems International, Orlando, Florida, 2006. [20] R.G.L. Narayanan, O.C. Ibe, Joint network for disaster relief and search and rescue network operations, in: Wireless Public Safety Networks 1, Elsevier, 2015, pp. 163–193. [21] M. Rath, N. Shekarappa, V. Ramachandra, Ground control station for UAV, Google Patents, October 18, 2007. [22] B. Markelj, I. Bernik, Mobile devices and corporate data security, Int. J. Educ. Inf. Technol. 6 (1) (2012) 97–104. [23] K. Mansfield,T. Eveleigh,T.H. Holzer, S. Sarkani, Unmanned aerial vehicle smart device ground control station cyber security threat model, in: IEEE International Conference on Technologies for Homeland Security, 2013, pp. 722–728. [24] J. Keller, Air force asks general atomics to upgrade UAV ground-control stations for use with the Internet. Available from: https://www.militaryaerospace.com/unmanned/ article/16718607/air-force-asks-general-atomics-to-upgrade-uav-groundcontrol-stations-for-use-with-the-internet. Accessed 29.09.19. [25] K. Klausen,T.I. Fossen,T.A. Johansen, Autonomous recovery of a fixed-wing UAV using a net suspended by two multirotor UAVs, J. F. Robot. 35 (5) (2018) 717–731. [26] M.R. Mokhtari, B. Cherki, Sliding mode control for a small coaxial rotorcraft UAV, in: 2015 3rd International Conference on Control, Engineering & Information Technology, 2015, pp. 1–6. [27] J. Gimenez, D.C. Gandolfo, L.R. Salinas, C. Rosales, R. Carelli, Multi-objective control for cooperative payload transport with rotorcraft UAVs, ISA Trans. 80 (2018) 491–502. [28] J. Song, M. Zhao,Y. Liu, H. Liu, X. Guo, Multi-rotor UAVs path planning method based on improved artificial potential field method, in: 2019 Chinese Control Conference, 2019, pp. 8242–8247. [29] M. Kotwicz Herniczek, D. Jee, B. Sanders, D. Feszty, Rotor blade optimization and flight testing of a small UAV rotorcraft, J. Unmanned Veh. Syst. 7 (4) (2016) 1–20. [30] J. Kalpa Gunarathna, R. Munasinghe, Development of a quad-rotor fixed-wing hybrid unmanned aerial vehicle, in: Moratuwa Engineering Research Conference, 2018, pp. 72–77. [31] A. Chapman, “Types of drones: Multi-rotor vs fixed-wing vs single rotor vs hybrid VTOL,” Drone Magazine, 2016. [32] UAVOS fixed-wing UAV SITARIA completed flight tests. Available from: https://www. uavos.com/uavos-fixed-wing-uav-sitaria-completed-flight-tests. Accessed 05.11.19. [33] Chinese unmanned flying surveillance drones enter Washington D.C.! Available from: https://www.suasnews.com/2013/07/chinese-unmanned-flying-surveillance-dronesenter-washington-d-c/. Accessed from 05.11.19.

Emerging use of UAV’s: secure communication protocol issues and challenges

55

[34] How drones are changing the maritime industry. Available from: https://www.shiptechnology.com/features/featurehow-drones-are-changing-the-maritime-industry-4865807/. Accessed 05.11.19. [35] A. Allouch, O. Cheikhrouhou, A. Koubâa, M. Khalgui, T. Abbes, MAVSec: Securing the MAVLink protocol for Ardupilot/PX4 unmanned aerial systems, in International Wireless Communications & Mobile Computing Conference, IEEE, 2019, pp. 621–628. [36] N.A. Khan, S.N. Brohi, N.Z., Jhanjhi, UAV’s applications architecture security issues and attack scenarios: a survey, in: 1st International Conference on Technology Innovation and Data Sciences, 2019. [37] J.-P.E. Kaps, Cryptography for ultra-low power devices, 2006. [38] N. Larrieu, How can model driven development approaches improve the certification process for uas?, in: 2014 International Conference on Unmanned Aircraft Systems, 2014, pp. 253–260. [39] O. Zouhri, S. Benhadou, H. Medromi, A new adaptative security protocol for UAV network, in: International Symposium on Ubiquitous Networking, 2016, pp. 649–657. [40] P. Gabrlik, V. Kriz, L. Zalud, Reconnaissance micro UAV system, Acta Polytech. CTU Proc. 2 (2) (2015) 15–21. [41] V. Kriz, P. Gabrlik, Uranuslink-communication protocol for UAV with small overhead and encryption ability, IFAC-PapersOnLine 48 (4) (2015) 474–479. [42] U. Development Team, UAVCAN Communication Protocol. Available from: https:// uavcan.org/Specification/1._Introduction/. Accessed 28.08.19. [43] L. Foundation, UAVCAN interaction. Available from: https://mavlink.io/en/guide/ uavcan_interaction.html. Accessed 28.08.19. [44] A. Koubâa, A. Allouch, M. Alajlan, Y. Javed, A. Belghith, M. Khalgui. Micro Air Vehicle Link (MAVLink) in a Nutshell: a survey, IEEE Access 7 (2019), 87658–87680. [45] A. Tridgell, L. Meier, MAVLink 2.0 packet signing proposal. October, 2015. [46] Y.-M. Kwon, J.Yu, B.-M. Cho,Y. Eun, K.-J. Park, Empirical analysis of mavlink protocol vulnerability for attacking unmanned aerial vehicles, IEEE Access 6 (2018) 43203– 43212. [47] Y.M. Kwon. Vulnerability analysis of the Mavlink protocol for Unmanned Aerial Vehicles, Doctoral dissertation, DGIST, 2018.

CHAPTER FOUR

A survey study on MAC and routing protocols to facilitate energy efficient and effective UAV-based communication systems Krishna Doddapanenia, Enver Everb

Amazon Web Services, San Francisco, CA, United States Computer Engineering, Middle East Technical University, Northern Cyprus Campus, Mersin, Turkey

a

b

Chapter outline 1 Introduction 2 Routing protocols for UAVs 2.1 Single-hop routing 2.2 Multihop routing 2.3 Position-based routing 2.4 Topology-based routing 3 Data delivery models in UAVs 4 MAC protocols for UAVs 4.1 Effects of MAC 4.2 Contention-based MAC protocols 4.3 Hybrid MAC protocols 5 Conclusion References

57 58 59 60 61 62 64 66 68 71 73 74 75

1  Introduction Unmanned aerial vehicles (UAVs) have recently introduced an important area of research for communications world. Because of their autonomy, flexibility, mobility, dynamic nature, and the broad range of application domains, they have enabled alternative perspectives for various application domains including military, surveillance and monitoring, telecommunications, and delivery of supplies [1]. On top of these traditional applications, as accurate and affordable solutions to certain technical challenges, which can be used to monitor large environments, they are also able to provide Drones in Smart-Cities. http://dx.doi.org/10.1016/B978-0-12-819972-5.00004-5 Copyright © 2020 Elsevier Inc. All rights reserved.

57

58

Krishna Doddapaneni, Enver Ever

solution in critical applications such as disaster relief and healthcare [2–5]. Especially for disaster relief-related scenarios, UAVs can also be employed as aerial base stations to enhance coverage, capacity, reliability, and energy efficiency of the communication systems [6–8]. The traditional research work on UAVs typically focuses on issues of navigation, control, and autonomy. This is mainly because, instead of using UAVs as facilitators of communications, the motivating applications used to be either robotics or military oriented. Instead, recent studies address the communication challenges that appear in various levels for UAVs. In this study, the medium access control and routing protocols, related challenges, and various solution approaches are considered in detail critically.The main contributions can be summarized as follows: • An exhaustive survey on routing protocols in UAVs and their classifications are presented. • Data delivery models in UAVs are discussed. • A survey on the MAC protocols in UAVs is considered critically along with their effects.

2  Routing protocols for UAVs Typical wireless network requirements such as the most efficient routing, scalability, latency reduction especially in mission critical applications, reliability, mobility, and ability to ensure the required QoS are important for UAV-based communication systems as well [9,10]. However, on top of these, routing in UAVs also requires location awareness, energy awareness, and increased robustness to fragile links and dynamically changing network topology. Development of the routing strategies and protocols for UAV networks is still one of the most challenging tasks for UAV-based communication systems [11]. Hence, although there exist a plethora of routing protocols for MANETs and VANETs, they cannot be directly applied for UAVs due to their unique characteristics. Lot of research is being currently conducted on UAV routing [12]. Existing routing protocols for wireless networks have been modified to suit for UAV applications. A variety of categories of protocols have been discussed in this study and present the extent of their suitability for applications involving UAVs. When typical wireless communication networks are considered, it is well known that the system should come with requirements such as an acceptable potential for scalability (depending on the application area), minimum latency (especially for mission critical applications), high reliability, ability to

A survey study on MAC and routing protocols

59

Figure 4.1  Classification of Routing Protocols for UAVs.

adapt various mobility conditions, and providing acceptable levels of other similar QoS measures. In addition to all these, routing in UAVs would also have requirements such as location and energy awareness, and mechanisms to improve robustness to intermittent links as well as to cope with dynamically changing network topology. Routing concept for UAV networks is still one of the most challenging tasks and a hot research topic [11]. Routing is well investigated for MANETs and VANETs introducing a plethora of routing protocols; however, these protocols cannot be directly applied for UAVs due to their unique characteristics. In this section, an overview of the state of the art of routing protocols for UAVs is presented. Routing protocols in UAVs can be classified based on the routing strategy into two major categories as singlehop and multihop routing protocols. Further classification of the protocols can be found in Fig. 4.1.

2.1  Single-hop routing The main purpose of a routing protocol is to transmit/forward the data gathered while increasing delivery ratio and minimizing delays and resource consumption. While the basic operations are handled, the routing protocols should also consider scalability, loop freedom, and efficient use of resources such as energy, memory, and computation time. When UAVs are employed as communication systems, one way to use them is as packet bearers, which transfers the relevant information (packets) when flying from the source to the destination. Fig. 4.2 illustrates a typical scenario for using UAVs as packet bearers. This way, a static routing table can be sufficient since there is no need of updates. In turn, the overhead can be avoided, which means the packets become lightweight [13]. Of course, this approach can mainly be employed for occasions with fixed topology.

60

Krishna Doddapaneni, Enver Ever

Figure 4.2  Single-hop Message Bearer Use Case for UAVs.

When protocols using UAVs as message bearers are considered, load carry and deliver (LCAD) routing can be a good example [14]. In this protocol, one or more UAVs are employed to relay messages from a source to a destination ground station with an aim of maximizing the throughput of UAV-relaying networks. Since the packets are delivered within the UAVs rather than transmissions over RF links, interference and medium access contention can be avoided; however, compared to conventional multihop store and forward routing protocols, the packet delivery latency can become excessive. Therefore, the main trade of would depend on the comparison in how fast the UAVs can travel between stations and how busy the wireless medium is. A well-known static routing algorithm for planning the route of UAVs in single-hop setup is differential evolution with quantumbehaved particle swarm optimization (DEQPSO). It combines two important aspects by providing a hybrid differential evolution (DE) system with quantum-behaved particle swarm optimization (QPSO) with an attempt to further enhance the performance of both algorithms. DEQPSO has been widely used especially in military applications such as fixed-wing UAVs [15] and civilian applications [16].

2.2  Multihop routing Multihop energy-efficient routing is another popular strategy employed to address the challenges of UAV-based communication systems. In multihop routing, packets are forwarded to the destination using other nodes as relays in the communication network. For systems using multihop communications,

A survey study on MAC and routing protocols

61

Figure 4.3  Multihop Routing Scenario for UAV Networks.

selection of the appropriate next hop is a critical factor for the performance. Fig. 4.3 shows a typical multihop routing scenario for UAV networks. Based on the different selection strategies for next-hop node, it is possible to classify multihop routing protocols into two main categories as topology-based and position-based routing protocols. In turn, the topology-based routing protocols can be further divided into four categories as proactive, reactive, hierarchical, and hybrid routing.

2.3  Position-based routing Due to the above-mentioned characteristics of the UAV networks, use of static routing tables or proactive routing protocols may not always be effective. Knowledge of the geographic position of the nodes is assumed in position-based or geographical routing schemes. The nodes send messages to the desired destination without a route discovery mechanism, assuming that the source knows the geographical position of the node. Greedy forwarding is one of the commonly used mechanisms in which each node uses the neighbor node closest to the destination for forwarding the message available.

62

Krishna Doddapaneni, Enver Ever

The decision of the closest node to the destination is made using only the local information available. There are a few approaches considering UAV routing in two dimensions, which have proposed algorithms that uses greedy routing. In these approaches, it becomes challenging to recover from local minima [17]. Furthermore, in UAVs, depending on the application, it is usually ideal to consider the network in three dimensions. When greedy routing protocols are considered, Greedy Distributed Spanning Tree Routing-3D (GDSTR) routing scheme is a popular approach. In order to reduce the likelihood of local minima, and to aggregate 3D node coordinates using two 2D convex hulls, the protocol uses information of two-hop neighbor during greedy forwarding [18]. Instead, another similar approach, Greedy-Hull-Greedy (GHG) protocol, makes use of the routing on the hull in order to escape the local minima. It uses a 3D equivalent of routing for 2D protocols [19].Yet another approach GreedyRandom-Greedy (GRG) protocol utilizes greedy mechanisms to forward packets until a local minimum is reached [20].

2.4  Topology-based routing 2.4.1  Proactive routing In Proactive routing, before sending the data packets, each and every node in the network maintains the current network state information of all the nodes. Whenever the topology of the network changes, the tables need to be updated. In a typical network, a routing table is needed to establish routes based on a predefined criterion including the bandwidth size of the calculated route, shortest path metric, and the forwarding delay between the nodes [21]. As the number of nodes in the network and the mobility factor increase, the overhead of the overall network tends to increase proportionally. As the nodes do not have an idle phase and are always listening to the wireless medium continuously, this impacts the energy level of the nodes [22]. However, in spite of these drawbacks, proactive routing offers fast connections between the nodes since routing information is immediately available in the routing table, minimizing the delay when data are ready to be sent. Because the nodes contain the latest information of the routes, the routing path can be selected to immediately transmit the data, therefore minimizing the overall delay in the network. 2.4.2  Reactive routing In reactive routing, the protocols effectively reduce the control message overhead in the network by finding a routing path on demand.

A survey study on MAC and routing protocols

63

Although the control message overhead is reduced, this type of protocols has the potential to form highly delayed networks unnecessarily in case the topology and link conditions do not change very often. Some common reactive routing protocols include DSR, UVAR, RGR, and AODV [23–26]. 2.4.3  Hierarchical routing In the UAV-based communications used for coverage extension, relaying and data distribution and collection, using UAVs similar to traditional wireless sensor network infrastructures is also becoming quite popular [27,28]. In these setups, the cluster head (CH) itself collects data and it is also responsible from data collection from other UAV cluster members. Furthermore, the data should be transferred toward the sink through inter cluster communications. Frequent change of cluster heads [29] has the potential to impose large overhead on the networks, even more for UAVs. Advanced clustering techniques are employed in order to efficiently utilize the radio channel resources as well as to reduce the routing overhead on the network. By maintaining a multilevel hierarchical topology, the cluster heads at the lower level automatically become the members of the next higher level in the hierarchical link state routing protocol [30]. These new members recursively organize themselves in clusters on the new level. Current research trends show the interest in the routing protocols based on hierarchy. Hierarchical protocols although achieved greater network performance in wired networks, in wireless networks the same cannot be achieved, due to the dynamically changing network hierarchies. 2.4.4  Hybrid routing Proactive and reactive routing protocols together constitute to hybrid routing protocols. The hybrid protocols aim to reduce the overhead, which is usually observed in case of proactive routing protocols and latency, which is a commonly known drawback of reactive routing protocols especially at the stage of initial route discovery. When larger networks are considered, proactive approach is used to perform zonal routing and reactive approach is used to perform inter-zone routing and also to divide the network into smaller zones. Network characteristics are taken in to account for Hybrid routing as it focuses on adjusting the routing based on these factors. A few clustering-based hybrid routing algorithms are proposed in the literature for UAVs, where the popular ones include Mobility Prediction Clustering

64

Krishna Doddapaneni, Enver Ever

Table 4.1  Categorization of data-delivery models with respect to application requirements. Data delivery End-to-end model Interactivity performance Delay tolerance Criticality

Event-driven Query-driven Continuous Hybrid

✓ ✓ ✗ ✓

✗ ✗ ✗ ✗

✗ Query-specific ✓ ✗

✓ ✓ ✓ ✓

Algorithm (MPCA) [31] as well as Hybrid Routing based on Clustering (HRC) [32].

3  Data delivery models in UAVs It is very difficult to analyze each and every application in UAVs from a data delivery perspective.Various classes of applications can be analyzed and classified for data delivery models. In other words, applications with similar requirements can be grouped [33]. Network’s QoS is affected by the choice of method used to transmit the data collected by a UAV to the base station. Considering the requirements of traditional WSN applications [34], the same can be presented for UAV-based applications as presented in Table 4.1. Real-time, energy-efficient, and reliable communication capabilities of UAV networks along with latest advancements in hardware capabilities realize the current trends in UAV applications. For UAVs, the traffic characteristics are very specific to the applications and are, in general, very distinct. For most of the applications, the main traffic flow shows the characteristics of many to one type communication. In other words, the traffic is from many sensor nodes to the UAV and in turn from many UAVs to the base station, which is typically an upstream connection. Upstream traffic delivery can be classified into four different models. The first class is continuous model, the second one is event driven, third type is query driven, and the last one is called as the hybrid-based data delivery model. Each of these data delivery models has different requirements in terms of quality of service and reliability depending on the application areas it is employed in [35]. Protocol design and performance of the network are greatly affected by the choice of the traffic models. The network’s performance especially in terms of energy is very critical while focusing on UAV applications.

A survey study on MAC and routing protocols

65

The four models are presented and their related performance features are discussed below. • When continuous delivery model is considered, first, a delivery rate is specified and each one of the sensors transmits information continuously or periodically, to the base station. It is also possible to use a hybrid model, which uses a combination of continuous, event-driven, and query-driven data delivery models for some applications. Time division multiple access (TMAC) protocols are used in case of continuous data delivery models to achieve significant energy savings. • UAV applications, especially the ones which are event-driven in nature, are mission critical, real time, and are interactive. During an event, the sensor node reports the event and an associated value to the base station. For such applications, the need to send the sensed data quickly and reliably is crucial. In fact, the event-driven and query-driven data delivery models are very similar in nature. The main difference is that in the case of event-driven model, nodes push the data to the base station, whereas for query-driven models, the data are pulled by the base station. • In general, the traffic introduced by a single sensor is significantly lower rate compared to a set of sensors during an event where the traffic is random and bursty in nature. Therefore, the frequency of occurrence of the events highly effect the routing efficiency in event-driven delivery model. As the data generated are sporadic in nature, carrier sense multiple access can be effectively employed in medium access control layer for such applications. • Query-driven UAV-based applications, especially the ones with eventdriven nature, are also mission critical, real time, and interactive similar to event-driven applications. To save energy, which is a key factor for UAV’s life time, queries can be sent on demand. As an explicit request from the user, the sensor node responds by reporting the observed data that are gathered locally. In this delivery model, an automatic path construction is performed at the arrival of a query. The query message is in turn broadcasted by the sink. The sensor node only reports its findings as a response to the request from the user in a query message. • In case the application considered requires sensing of diverse data, use of a hybrid model can be required. In the hybrid data delivery model, a combination of event-driven, continuous, and query-driven data delivery can be used together. As an example, the base station may need to query information from a different sensor, where the data may be reported continuously by some other sensors.

66

Krishna Doddapaneni, Enver Ever

Figure 4.4  Classification of MAC Protocols for UAVs.

4  MAC protocols for UAVs In the MAC layer of the UAVs, self-organization, high-degree of coordination, and management among the sensors are required to support the variety of application areas and make use of UAVs as effectively as possible. A common transmission medium shared as fairly as possible by all the nodes should be defined for this purpose. Therefore, shared access of the channel requires establishment of a MAC protocol among the nodes in the network. In addition, when UAVs are considered, the energy preservation becomes one of the most important factors in the design of a MAC layer protocol that can facilitate these requirements. The choice of MAC protocol highly impacts the performance of the UAVs. Therefore, while facilitating fair and high performance sharing of resources, the MAC-layer protocols also need to reduce the energy wastage due to packet collisions, idle listening, excessive overhead, and overhearing [36]. For UAVs, MAC protocols can be categorized into three main groups: schedule-based, contention-based, and hybrid, as shown in Fig. 4.4. Each of these categories of MAC protocol focuses on certain aspects: schedule-based MAC focuses on extending the networks life time by able to achieve higher levels of energy efficiency. Contention-based MAC protocols, rather than avoiding the collisions completely, tend to minimize them. Hybrid protocols combine the best of both contention-based and schedule-based algorithms, focus on energy saving, and provide better scalability and flexibility. Although contention-based protocols are known to efficiently adapt changing network conditions, they suffer from hidden terminal problems and delays caused due to request to send, clear to send (RTS/CTS) handshakes with increased node densities. Contention-free protocols on the

A survey study on MAC and routing protocols

67

other hand have a better channel utilization at higher loads and tend to eliminate collisions, but face issues with resource synchronization. Hybrid MAC protocols on the other hand tend to save energy while providing better scalability and higher flexibility by combining the best of both schedulebased as well as contention-based protocols. These protocols differ in the mechanism used to minimize the possibility of a collision while minimizing control traffic overhead as well as overhearing, with an overall goal to reduce energy consumption of the network. The nodes consume energy needlessly in idle state as well as when the collisions occur.This is a key limitation of the traditional contention-based channel access protocols defined in literature. In UAVs, collisions cause packet retransmissions and increased latency [37]. In any wireless network, collisions occur when more than one node tries to access the channel to transmit simultaneously or when in a noisy channel, the node assumes that the data are lost due to errors. These lead to retransmission on the packets causing extra energy burden on already constrained system.While collisions lead to reductions in terms of quality of service because of retransmissions and extra latency (network performance degradation), they also cause higher levels of energy consumption [37–39]. Other major effects of frequent collisions and retransmissions in UAVs are dropped in packet success ratio and increased in time for the data packets to reach the base station for processing and of course, the energy consumption in UAV networks. Data packets get corrupted due to data collisions and they have to be discarded. This leads to follow-on retransmissions, which again lead to network latency and depletion of energy sources [37–40]. Apart from considering effects at different layers including topology in which the nodes are deployed, rate of transmission, data aggregation factor, path loss, etc., performance optimization in UAVs must also include the effects of collisions. No MAC protocol has been standardized for UAVs among a variety of protocols available in literature. The choice of MAC protocol to be employed depends on the type of deployment, expected load patterns, the chosen application, underlying hardware’s energy consumption behavior including the costs for idle time, transmitting, receiving, wake-up times, switching between modes, as well as the cost for executing the MAC protocol employed. For UAV networks, time-division multiple access (TDMA) and carrier-sense multiple access (CSMA) are most frequently employed underlying MAC protocols. TDMA employs strict time synchronization between transmitting nodes. While TDMA significantly improves energy efficiency

68

Krishna Doddapaneni, Enver Ever

even in higher traffic scenarios due to its collision-free nature, in low traffic scenarios, due to idle slot times, it has low throughput and higher delay. CSMA on the other hand are contention-based protocols for high traffic scenarios, yield higher delays and lower energy efficiencies. For low traffic scenarios, yield higher throughput and reduced delay. The choice of MAC protocol—TDMA or CSMA—based to be employed needs to be based on the specific application and the required performance requirements. In a UAV network, the energy spent due to idle listening, path loss, control packet overhead, collisions, retransmissions, over-emitting, over hearing, sleep, and wake-up costs are quite crucial factors as they impact the network’s lifetime severely and hence need to be considered during the design of a novel MAC protocol for UAVs. Hence, while designing a MAC protocol, the following attributes need to be considered: • Throughput: The requirements of UAVs in terms of delivery efficiency are very specific to the task and are demanding because of the increasing growth in UAV applications especially involving video and surveillance. • Scalability of the Network: For nodes competing to access the channel, the network must be scalable—either to increase or decrease as needed. • Latency: Most UAV applications are critical where data need to be send in real time and latency needs to be kept to possible minimum. • Energy Efficiency: In order to prolong the network lifetime, the protocol must be as energy efficient as possible.

4.1  Effects of MAC At the cluster head (CH), channel contention causes queuing delays, additional delays and wireless channel errors occur due to packets arriving from nodes in the cluster and data forwarded from other CHs. QoS of the network is determined by the allocation of shared wireless channel and scheduling of the channel usage, which is dependent on the MAC layer. MAC protocols influence the communication mechanisms that effect the utilization of the transceiver. Network’s life time is highly effected as the sensors deplete their energy as transceiver constantly senses the channel. Each MAC protocol is designed with an ideal goal of catering high throughput and QoS. However, higher priority is given to minimize the energy consumption rather than the QoS requirements when MAC protocols designed for UAV networks are considered. Hence, it is important to characterize interarrival distribution correctly. In turn, this would play a significant role in the estimation of the QoS in UAVs. Channel contention results in collisions,

A survey study on MAC and routing protocols

69

energy depletion, retransmissions, and loss of data and hence is a serious problem in UAVs. To resolve contention among the nodes to acquire channel access, MAC protocols employ back-off algorithms. To maintain energy efficiency, the amount of energy wasted on idle listening should be reduced and common contention-based MAC protocols such as CSMA or T-MAC should be employed [41]. Most of the UAV-based MAC protocols are contention-based. MAC protocols designed for UAVs often trade performance characteristics, such as throughput and latency, to minimize the energy consumption and extend their life time. CSMA/CA is one such contention-based technique in which the node senses whether the channel is idle or not before it transmits the packet. Due to the nondeterministic nature of delay, it becomes a key challenge to support real-time data transmissions in such techniques. The node needs to wait for a certain predefined period of time before it can resense the channel if it is not idle at the time of transmission. Hence, the delay time is nondeterministic in a successful transmission. In such cases, if there is a duty cycle employed, as the node back-offs, the node sleeps to save energy (example TMAC). Before the node can wake up and sense the channel again, an extra delay period is expected since all the nodes apart from the sink should periodically turn off their radio to save energy. Every time a node has a packet to send, it sends an RTS beacon similar to CSMA/CA. Carrier sense medium access and random back-off mechanisms are also employed, which again causes delays in the system. Of course, this delay has the potential to cause collisions with higher probability; however, the significant amount of energy that can be saved makes it a preferable mechanism [41]. The transmission time of the packet depends on the data rate set by the radio and also the size of the packet being transmitted, and these can be governed by equations. That said, most of the delay is due to the employed MAC protocol. For example, an instance could be that the MAC layer is waiting for the active period to commence or for the channel to be clear. MAC layer is responsible not only for retransmissions but most of the delay caused by a retransmission packet is also due to the MAC technique employed.Therefore, delay in the transmission of data packets and the overall performance of the network including energy usage is effected due to buffering the packets for retransmission. Such uncertainties due to random back-offs caused due to employing various MAC protocols are not considered in the delay distribution models presented in the literature. Hence, while considering the effects of MAC protocols causing delays, it is fundamental to characterize the inter-arrival distributions at the CH.

70

Krishna Doddapaneni, Enver Ever

The choice of MAC protocol to be employed depends on the type of deployment, expected load patterns, the chosen application, underlying hardware’s energy consumption behavior including the costs for idle time, transmitting, receiving, wake-up times, switching between modes, as well as the cost for executing the MAC protocol employed. Hence, there is no generic MAC protocol. 4.1.1  Schedule-based MAC protocols By preallocating the resources used for transmission in the network, schedule-based protocols eliminate the collisions in the network. There are three conventionally available approaches are available for scheduling wireless channels, which are very well known. These mechanisms are known as time division multiple access (TDMA), frequency division multiple access (FDMA), and code division multiple access (CDMA). These mechanisms rely either on fixed or on-demand allocation of resources and are responsible for preallocating the transmission resources. Each of these MAC mechanisms possesses some challenges. The challenges related to power control in CDMA, and its soft capacity, the time synchronization problem observed for TDMA, and frequency generation/ filtering-related issues of FDMA are also well known [9]. 4.1.2  Prioritized frame selection MAC protocol based on CDMA (PFSC) One of the novel multiple access schemes proposed for wireless sensor networks using UAVs is PFSC. In this scheme, based on some prespecified priorities, the active sensors are divided into a number of subgroups. In turn, CDMA-based transmission technique is used to facilitate the communication of the sensor nodes with UAVs. In other words, Prioritized Frame selection MAC protocol uses a novel PFS scheme as well as a CDMA mechanism in the physical layer. One of the main advantage of employing this technique is that a number of sensor nodes can be allowed to transmit concurrently with Orthogonal CDMA codes resulting in minimum transmission intervals. However, constraints such as multiuser interference, and the failure to deliver the uniformity of data collection [42] can cause some degradations of performance. 4.1.3  Prioritized frame selection MAC protocol based on TDMA (PFST) Another scheme proposed for MAC, which is mainly based on TDMA time slots, is PFST. This technique enables reliable and concurrent transmission

A survey study on MAC and routing protocols

71

between UAVs and relatively large-scale networks (with large number of sensors). In order to change the levels of energy consumption, the UAV can adopt to the environment considering the distances between UAVs and sensors and accordingly changing the beacon transmission power. This way, UAV becomes able to adjust the transmitting power in order to maintain various levels of beacon signal at the sensors. Prioritized Frame Selection divides the active sensors into several subgroups and each is assigned with a proper priority.TDMA scheme has been validated for those multiple data transmissions in each subgroup for effective reasons. In addition, although there are constrains such as the longer data transmission time and partitioned channels, which means reduced bandwidths, using TDMA slots transmission from large numbers of sensors can be facilitated [43]. 4.1.4  Adaptive-opportunistic (AO) aloha CDMA-based MAC The proposed AO Aloha CDMA-based MAC uses adaptive priority-based channel assignment for collision avoidance and a Slot-like beaconing for effective channel allocation. The protocol considers transmission efficiency, energy consumption, and uniformity of data collection. Uniform data collection across the area is ensured by priority-based data collection.The sensor nodes are randomly distributed in the sensing area. Based on the number of nodes in the area, the particular node will be assigned a higher priority if the total nodes are less. The node with highest transmission priority will be allowed to claim the unoccupied channel and then transmits its sensed data packets back to the UAV. When this occurs, in the current slot, all the remaining nodes are in silent mode. All the sensor nodes go to sleep mode when they are not transmitting in order to improve energy efficiency. A beacon signal from the UAV wakes the sensor nodes when it is their turn to transmit data. This process enhances system throughput with the help of a handshake for every transmission’s time interval [44].

4.2  Contention-based MAC protocols In UAVs, collisions and retransmission due to random accessing of channels and path-loss effects degrade the network performance and in turn the lifetime of the network. The most popular is IEEE 802.11 CSMA/CA as it is not very suitable in UAVs due to hidden terminal effect and low power sensors. It is due to these factors, the need for newer contention-based MAC protocols arises in UAVs.

72

Krishna Doddapaneni, Enver Ever

4.2.1  Priority-based contention window adjustment scheme (PCWAS) A conventional CSMA/CA-based protocol is proposed in Priority-Based Contention Window Adjustment Scheme, where an UAV relays data from the sink to collect data from WSN on the ground. Each frame of the UAV’s beacon signal is divided into many frames, and different transmission priorities are assigned for each frame. With this approach, it is possible to introduce a circularly optimized frame selection as well as a priority-based optimized frame selection. The proposed scheme allows higher priority of data transmission to sensors and also minimizes the number of collisions, minimizing the packet loss [45]. 4.2.2  Partnerships-based MAC protocol A flexible and fast approach has been proposed in partnerships-based MAC protocol. In this approach, the mobility of the mobile sink (UAV) and sensor nodes in the network are considered for data collection. By leveraging two factors, a novel frame selection technique that classifies sensor nodes into different frames is adopted—including location of sensor nodes and UAV mobility. This is followed by individually pairing with peers, which permits transmitting data simultaneously. This corporate partnership model reduces the packet loss occurring from the sensors that are far away in the overall coverage area.This pattern normally occurs while the UAV is flying toward the center of the network. A novel data forwarding scheme and partnership model is proposed in order to reduce these packet drops and to guarantee better success rate of packet reception ration and hence overall performance of the network [46]. 4.2.3  Cooperative sensing data collecting framework A novel system architecture and a MAC protocol based on the selection criteria of the target cooperative sensor node is presented. This protocol also focuses on the sensing data that should be retransmitted by using the decided cooperative sensor node. Based on the IEEE 802.11 CSMA/CA MAC protocol, the proposed approach addresses an efficient data collection scheme based on the cooperative communications by using UAVs for WSNs. 4.2.4  Prioritized data gathering MAC protocol The proposed prioritized data gathering MAC protocol is based on the very common 802.11 standard. While the original 802.11 standard is considered, a lot of changes have been made to the original version including modifying the data transmission phase and removing the RTS/CTS mechanism completely.

A survey study on MAC and routing protocols

73

Working in environments with higher node densities can cause high probabilities of collisions that can disrupt the process of data collection in WSNs and especially with UAVs, it can be quite challenging. Use of prioritization mechanism to control the order of sending data to the UAVs has been effectively proposed in this scheme, thus to lower this factor of collisions and handshakes and assuring that every node has the possibility to communicate with the UAV. In this scheme, prioritization becomes the key aspect as it guarantees successful data collection with high densities and limiting the number of retransmissions due to collisions and in high-speed pursuits [47].

4.3  Hybrid MAC protocols Schedule-based MAC protocols and contention-based protocols have their fair share of advantages including elimination of collisions, adapt easily to changing network scenarios, have better channel utilization at higher loads and are better suited for networks with low loads, they have their own limitations. These include proper synchronization of allocated resources and not performing well when the node density increases in the network which cause delays due to collisions due to hidden terminal problems and of course the RTS/CTS handshaking. Therefore, the need for protocols that combine the advantages of both contention-based and schedule-based protocols to provide higher scalability, flexibility and of course to increase the overall lifetime of the network. Hybrid MAC protocols are proposed in an attempt to improve the performance of the network. 4.3.1  Fixed inter Beacon duration and proactive scheduling (FDPS MAC) In this approach, the authors propose a MAC scheme based on hybrid beaconing by combing CSMA/CA with physical parameters-based scheduling. The mobile nodes that receive the beacon will randomly access the channel through CSMA/CA while the UAVs broadcast beacons all across the coverage area. The fixed inter-beacon duration combined with a proactive scheduling MAC to an beacon-based IEEE 802.15.4 are compared. Two models of FDPS, FD-PS MAC I, and FD-PS MAC II are introduced. The advantage of FDPS MAC is that adaptive contention-based and contention-free period enhance performance in FD-PS II. While the limitations include the network experiences delay due to longer contentionbased period [48].

74

Krishna Doddapaneni, Enver Ever

4.3.2  Adaptive inter-Beacon duration and proactive scheduling (ADPS MAC) Beacon-based IEEE 802.15.4 MAC and Data rate/Contact duration time algorithm (DR/CDT) mechanisms are adopted in ADPS MAC. This protocols is introduced not only to minimize the collisions between the joint sensors and the UAV but also to coordinate the control frames between them. In order for scheduling information in the detected nodes, the ADPS MAC combines the beacon-based CSMS/CA along with the DR/CDT, which contains a predefined number of time slots. In ADPS MAC, even though the network topology is dynamic, the Inter-beacon duration is adaptive, which is the most important advantage [49]. 4.3.3  Hybrid medium access control (HP MAC) HP protocol named HP-MAC is a hybrid MAC protocol for data gathering in a UAV-based WSN. In this scheme, the UAV periodically informs its presence to the sensor node by sending a beacon frame. Now, at this point, each sensor node that receives the beacon frame from the UAV sends a registration frame to the UAV.The transmission schedule is now transmitted to the registered node by the UAV in return with the help of a beacon frame. During the registration process of the sensor nodes, HP-MAC uses CSMA/ CA and allocates time slots to the sensor nodes that are registered in the network for data gathering. Based on the priority during the registration process, the transmission schedule of each sensor node is determined based on the time-slot scheme [50–52].

5  Conclusion Since the existing work on MANETs, and VANETs do not address the unique characteristics of the UAV networks—there is a need for newer protocols for applications which use UAVs. Protocols to be developed are required to adapt to high mobility, dynamic topology, intermittent links, power constraints, and changing link quality. Performance evaluation of the developed protocols from various perspectives is therefore very critical.The protocols considered in various levels should be able to support features such as energy efficiency, high connectivity, delay sensitiveness, high reliability, and security. In this study, the existing MAC and routing protocols are considered critically. Their characteristics are analyzed and discussed, which can help the researchers understand which ones would be more suitable for specific application areas.

A survey study on MAC and routing protocols

75

References [1] L. Gupta, R. Jain, G. Vaszkun, Survey of important issues in uav communication networks, IEEE Commun. Surveys Tutorials 18 (2) (2015) 1123–1152. [2] M. Erdelj, E. Natalizio, K.R. Chowdhury, I.F. Akyildiz, Help from the sky: Leveraging uavs for disaster management, IEEE Pervasive Computing 16 (1) (2017), 24–32. [3] S. Ullah, K.-I. Kim, K.H. Kim, M. Imran, P. Khan, E. Tovar, F. Ali, Uav-enabled healthcare architecture: issues and challenges, Future Gen. Comput. Syst. 97 (2019) 425–432. [4] F. Malandrino, C.-F. Chiasserini, C. Casetti, L. Chiaraviglio, A. Senacheribbe, Planning uav activities for efficient user coverage in disaster areas, Ad Hoc Networks 89 (2019) 177–185. [5] Y. Liang, W. Xu, W. Liang, J. Peng, X. Jia,Y. Zhou, L. Duan, Nonredundant information collection in rescue applications via an energy-constrained uav, IEEE Internet Things J. 6 (2) (2018) 2945–2958. [6] S.A.R. Naqvi, S.A. Hassan, H. Pervaiz, Q. Ni, Drone-aided communication as a key enabler for 5g and resilient public safety networks, IEEE Commun. Mag. 56 (1) (2018) 36–42. [7] D. Sikeridis, E.E. Tsiropoulou, M. Devetsikiotis, S. Papavassiliou, Wireless powered public safety iot: a uav-assisted adaptive-learning approach towards energy efficiency. J. Netw. Comput. Appl. 123 (2018) 69–79. [8] R. Shakeri, M.A. Al-Garadi, A. Badawy, A. Mohamed, T. Khattab, A. Al-Ali, K.A. Harras, M. Guizani, Design challenges of multi-uav systems in cyber-physical applications: a comprehensive survey, and future directions, IEEE Commun. Surveys Tutorials (2019). [9] K. Doddapaneni, Energy aware performance evaluation of WSNS. PhD Thesis, 2014. [10] K. Doddapaneni, E. Ever, O. Gemikonakli, I. Malavolta, L. Mostarda, H. Muccini, Path loss effect on energy consumption in a WSN, in: UKSim 14th International Conference on Computer Modelling and Simulation, Cambridge, 2012, pp. 569–574. doi:10.1109/ UKSim.2012.87. [11] J. Jiang, G. Han, Routing protocols for unmanned aerial vehicles, IEEE Commun. Mag. 56 (1) (2018) 58–63. [12] J. Enright, E. Frazzoli, M. Pavone, K. Savla, UAV routing and coordination in stochastic, dynamic environments, in: Handbook of Unmanned Aerial Vehicles, Springer, Dordrecht, 2015, pp. 2079–2109. [13] J.N. Al-Karaki, A.E. Kamal, Routing techniques in wireless sensor networks: a survey, IEEE Wireless Commun. 11 (6) (2004) 6–28. [14] C. Cheng, P. Hsiao, H.T. Kung, D. Vlah, Maximizing throughput of uav-relaying networks with the load-carry-and-deliver paradigm, in: IEEE Wireless Communications and Networking Conference, March 2007, pp. 4417–4424. [15] V. Roberge, M. Tarbouchi, G. Labonté, Fast genetic algorithm path planner for fixedwing military UAV using GPU, IEEE Trans. Aerospace Electron. Syst. 54 (5) (2018) 2105–2117. [16] Y. Fu, M. Ding, C. Zhou, H. Hu, Route planning for unmanned aerial vehicle (UAV) on the sea using hybrid differential evolution and quantum-behaved particle swarm optimization, IEEE Trans. Syst. Man Cybernet.: Syst. 43 (6) (2013) 1451–1465. [17] S.S. Lam, C. Qian, Geographic routing in d-dimensional spaces with guaranteed delivery and low stretch, in: Proceedings of the ACM SIGMETRICS Joint International Conference on Measurement and Modeling of Computer Systems, SIGMETRICS ‘11, New York, NY, USA, ACM, 2011, pp. 257–268. [18] J. Zhou,Y. Chen, B. Leong, P.S. Sundaramoorthy, Practical 3d geographic routing for wireless sensor networks, in: Proceedings of the 8th ACM Conference on Embedded Networked Sensor Systems, SenSys ‘10, , New York, NY, USA, ACM, 2010, pp. 337–350. [19] C. Liu, J. Wu, Efficient geometric routing in three dimensional ad hoc networks, in: IEEE INFOCOM, 2009, pp. 2751–2755.

76

Krishna Doddapaneni, Enver Ever

[20] R. Flury, R. Wattenhofer, Randomized 3d geographic routing, in: IEEE INFOCOM 2008 - The 27th Conference on Computer Communications, April 2008, pp. 834–842. [21] S. Mohseni, R. Hassan, A. Patel, R. Razali, Comparative review study of reactive and proactive routing protocols in manets, in: 4th IEEE International Conference on Digital Ecosystems and Technologies, April 2010, pp. 304–309. [22] K. Doddapaneni, E. Ever, O. Gemikonakli, I. Malavolta, L. Mostarda, H. Muccini, Path loss effect on energy consumption in a WSN, in: 2012 UKSim 14th International Conference on Computer Modelling and Simulation, March 2012, pp. 569–574. [23] A.I. Alshabtat, L. Dong, J. Li, F.Yang, Low latency routing algorithm for unmanned aerial vehicles ad-hoc networks, Int. J. Elect. Comput. Eng. 6 (2011) 48–54. [24] S. Murthy, J.J. Garcia-Luna-Aceves, An efficient routing protocol for wireless networks, Mob. Netw. Appl. 1 (2) (1996) 183–197. [25] J.M.M. Biomo, T. Kunz, M. St-Hilaire, Routing in unmanned aerial ad hoc networks: Introducing a route reliability criterion, in: 2014 7th IFIP Wireless and Mobile Networking Conference (WMNC), May 2014, pp. 1–7. [26] O.S.Oubbati, A. Lakas, F. Zhou, M. Gne, N. Lagraa, M.B. Yagoubi, Intelligent UAVassisted routing protocol for urban vanets, Comput. Commun. 107 (C) (2017), 93–111. [27] J.Wu, L. Zou, L. Zhao, A. Al-Dubai, L. Mackenzie, G. Min, A multi-UAV clustering strategy for reducing insecure communication range, Comput. Netw. 158 (2019) 132–142. [28] G. Wang, B. Lee, J. Ahn, G. Cho, A UAV-assisted CH election framework for secure data collection in wireless sensor networks, Future Gen. Comput. Syst. 102 (2020) 152–162. [29] K. Doddapaneni, F.A. Omondi, E. Ever, P. Shah, O. Gemikonakli, R. Gagliardi, Deployment challenges and developments in wireless sensor networks clustering, in: 2014 28th International Conference on Advanced Information Networking and Applications Workshops, May 2014, pp. 227–232. [30] G. Pei, M. Gerla, X. Hong, C. Chiang, A wireless hierarchical routing protocol with group mobility, in: WCNC. 1999 IEEE Wireless Communications and Networking Conference (Cat. No.99TH8466), 3, Sep. 1999, pp. 1538–1542. [31] C. Zang, S. Zang, Mobility prediction clustering algorithm for uav networking, in: 2011 IEEE GLOBECOM Workshops (GC Wkshps), December 2011, pp. 1158–1161. [32] K. Liu, J. Zhang, T. Zhang, The clustering algorithm of uav networking in near-space, in: 2008 8th International Symposium on Antennas, Propagation and EM Theory, November 2008, pp. 1550–1553. [33] I.M.M.E. Emary, S. Ramakrishnan, Wireless Sensor Networks: From Theory to Applications, Telecommunications Books, Taylor & Francis, 2013. [34] R. Rizk, Quality of services in wireless sensor networks.Wireless sensor networks: from theory to applications, CRC Press, Taylor & Francis Group, USA, 2013. Available from: http://www.crcpress.com/product/isbn/9781466518100. [35] S.Tilak, N.B. Abu-Ghazaleh,W. Heinzelman, A taxonomy of wireless micro-sensor network models, ACM Mobile Comput Commun. Rev. 6 (2002) 28–36. [36] D. Liu, P. Ning, Security for wireless sensor networks. Advances in Information Security, 28. Springer Science + Business Media, LLC, 2007. [37] V. Rajendran, K. Obraczka, J.J. Garcia-Luna-Aceves, Energy-efficient collision-free medium access control for wireless sensor networks, in: Proceedings of the 1st International Conference on Embedded Networked Sensor Systems, SenSys ‘03, New York, NY, USA ACM, 2003, pp. 181–192. [38] C.-K. Lin, V. Zadorozhny, P. Krishnamurthy, Efficient hybrid channel access for data intensive sensor networks, in: Proceedings of the 21st International Conference on Advanced Information Networking and Applications Workshops - Volume 02, AINAW ‘07, Washington, DC, USA, IEEE Computer Society, 2007, pp. 659–664. [39] T. Stathopoulos, R. Kapur, D. Estrin, J. Heidemann, L. Zhang. Application-based collision avoidance in wireless sensor networks, in: 29th Annual IEEE International Conference on Local Computer Networks, November 2004, pp. 506–514.

A survey study on MAC and routing protocols

77

[40] F. Hu, X. Cao,Wireless Sensor Networks Principles and Practice. Auerbach Publications, Boca Raton, FL, 2010. Formerly CIP. [41] T. van Dam, K. Langendoen, An adaptive energy-efficient mac protocol for wireless sensor networks, in: Proceedings of the 1st International Conference on Embedded Networked Sensor Systems, SenSys ‘03, New York, NY, USA, ACM, 2003, pp. 171–180. [42] Tu Dac Ho, Jingyu Park, S. Shimamoto, Novel multiple access scheme for wireless sensor network employing unmanned aerial vehicle, in: 29th Digital Avionics Systems Conference, 5.C.5-1–5.C.5-8, Oct 2010. [43] D. Ho, S. Shimamoto, Highly reliable communication protocol for wsn-uav system employing tdma and pfs scheme, in: 2011 IEEE GLOBECOM Workshops (GC Wkshps), December 2011, pp. 1320–1324. [44] L. Wang, H. Li, Y. Jiang, Adaptive-opportunistic aloha: a media access control protocol for unmanned aerial vehicle-wireless sensor network systems, Int. J. Distrib. Sens. Netw 12 (2016) 12–20. [45] S. Sotheara, N. Aomi, T. Ando, L. Jiang, N. Shiratori, S. Shimamoto, Effective data gathering protocol in wsn-uav employing priority-based contention window adjustment scheme, in: 2014 IEEE Globecom Workshops (GC Wkshps), December 2014, pp. 1475–1480. [46] S. Say, H. Inata, M.E. Ernawan, Z. Pan, J. Liu, S. Shimamoto, Partnership and data forwarding model for data acquisition in UAV-aided sensor networks, in: 2017 14th IEEE Annual Consumer Communications Networking Conference (CCNC), January 2017, pp. 933–938. [47] A.-V.Vladuta, M. Pura, I. Bica, Mac protocol for data gathering in wireless sensor networks with the aid of unmanned aerial vehicles, Adv. Electr. Comput. Eng. 16 (2016) 51–56. [48] X. Ma, R. Kacimi, R. Dhaou, Adaptive hybrid mac protocols for uav-assisted mobile sensor networks, in: 2018 15th IEEE Annual Consumer Communications Networking Conference (CCNC), January 2018, pp. 1–4. [49] X. Ma, Data Collection of Mobile Sensor Networks by Drones, November 2017. [50] M.R. Ramli, J.-M. Lee, D.-S. Kim, Hybrid mac protocol for UAV-assisted data gathering in a wireless sensor network, Internet of Things (2019) 100088. [51] F. Al-Turjman, A novel approach for drones positioning in mission critical applications, T. Emerg., Telecommun. Technol. (2019) doi:10.1002/ett.3603. [52] E. Ever, E. Gemikonakli, H. Nguyen, F. Al-Turjman, A. Yazici, Performance evaluation of hybrid disaster recovery framework with D2D communications, Comp. Commun. J. 152 (15) (2020) 81–92.

CHAPTER FIVE

UAVs: communication aspects and cellular service provisioning Muhammad Asim Jana, Syed Ali Hassana, Haejoon Jungb

School of Electrical Engineering and Computer Science (SEECS), National University of Sciences and Technology (NUST), Islamabad, Pakistan b Department of Information and Telecommunication Engineering, Incheon National University, Incheon, Korea a

Chapter outline 1 Introduction 1.1 UAV communication 2 System model 2.1 Channel model 2.2 User association metric 3 Performance analysis 4 Results 5 Conclusion References

79 80 81 82 85 85 87 90 90

1  Introduction The smart city concept has gained massive popularity over the recent years due to the opportunities that it offers in improving the quality of life of its citizens. This improvement will be brought about through smart and efficient resource optimization in multiple domains ranging from smart homes, smart grids, intelligent transportation system, air quality monitoring, smart parking, smart agriculture, smart health care system, smart energy management and round the clock surveillance [1]. The smart city concept will revolve around deployment of wire-less sensor networks throughout the city and then deploying monitoring mechanism to take decisions based on the sensor information. These services includes, but are not limited to, automated cars, traffic congestion avoidance during peak hours, efficient energy management system to conserve energy, and air pollution monitoring and control [2]. The sensors will be ubiquitous and can be deployed in human body, vehicles, buildings, agriculture fields, etc. Authors in Ref. [3] propose an unmanned aerial vehicle (UAV)-based Internet of Things network based completely in the sky to collect data for provisioning of services in future smart cities. Deployment of sensors for provisioning of smart services along Drones in Smart-Cities. http://dx.doi.org/10.1016/B978-0-12-819972-5.00005-7 Copyright © 2020 Elsevier Inc. All rights reserved.

79

80

Muhammad Asim Jan, Syed Ali Hassan, Haejoon Jung

with the existing cellular traffic is going to put huge constraints on the existing cellular infrastructure. To keep up with the high bandwidth demands of applications inside smart cities, this chapter defines a method for augmenting the existing cellular infrastructure with millimeter-wave (mmWave) UAVsbased communication. The mmWave UAVs will act as small cells in the sky which will effectively increase the capacity of the existing network as proposed in the 5G network architecture [4]. mmWave UAV assisted cellular service provisioning will provide on demand and infrastructure less communication services that will not only reduce user load from the ground base stations, like in case of a crowded hot spot or a traffic jam, but will also meet bandwidth demands of critical applications whenever required. Deployment of UAVs as aerial base stations has grabbed interest of academia and industry in recent times. UAV, due to its capability to manoeuvre in 3D space, offers exciting opportunities and unexplored potential. Its ability to be used only at the required time and place is another exciting area which can be of great interest in futuristic smart cities. UAV can come in handy in a smart city where due to users congestion the ground-based cellular infrastructure is overloaded and UAVs can be sent in at particular place to offload traffic from macro base station (MBS) or ground-based small base station (SBS). Deployment of UAVs to cater for various functions of smart cities has already been proposed in literature including employment of UAV network for provisioning of public safety emergency services [5], intelligent transportation inside smart cities [6] and smart parking solutions [7]. This chapter explains a mmWave UAV assisted cellular network for meeting coverage and data rate requirements of smart services and users to sustain smooth and efficient operations inside a smart city. mmWave spectrum for UAV assisted communication provides us with huge unexploited bandwidth that can easily meet the hyper connectivity requirements inside smart cities [8]. Moreover, in a heterogeneous network (HetNet) existing cellular services in the form of groundbased MBS and SBS are overlaid with mmWave band UAVs. mmWave owing to its inherent high directionality and path loss provides us with adequate protection from interference which can degrade performance of such a HetNet.

1.1  UAV communication UAV-based communication has gained considerable popularity owing to its rapid and on demand deployment capability. Authors in Ref. [9] have provided mechanism for maximizing the coverage based on height of UAV and the environment dependent statistical parameters. Moreover, expression for calculating the geometric line of sight (LoS) from the ground user to the aerial platform has also been calculated. Accurate calculation of LoS

UAVs: communication aspects and cellular service provisioning

81

probability between the mmWave UAV and the ground user is very important as propagation characteristics of mmWave are considerably effected in a nonline of sight (NLoS) environment. mmWave communication is greatly effected by diffraction and the blockages from buildings, therefore LoS path is important for getting optimum performance from a mmWave network. As the LoS probability increases considerably with the altitude of the aerial platform, therefore, UAV-based mmWave network will perform better in mmWave band due to enhanced LoS capability gained due to height of the UAV above the ground. Authors in Ref. [10] have focused on the propagation models in the mmWave spectrum under different environments. To sustain the huge number of smart functions inside the smart cities, we need flexible and diverse communication networks. mmWave UAV communication is an effective mechanism for enhancing the already deployed terrestrial communication network because mmWave UAVs will enable the massive bandwidth exploitation in the untapped mmWave band. Moreover, infrastructure less deployment makes it a cost effective solution at the time of need, be it for the provisioning of emergency communication services in the case of a fire [11] or the handling of crowds in a city center.

2  System model Our case study assumes the scenario, as illustrated in Fig. 5.1, in which a crowded hot spot is generated in a particular region of interest (RoI) of a smart city due to traffic congestion or an exclusive event. The already

Figure 5.1  System model of communication network.

82

Muhammad Asim Jan, Syed Ali Hassan, Haejoon Jung

existing terrestrial cellular infrastructure is insufficient to satisfy the user and smart services demands in the specific RoI.We assume the RoI to be spread over an area of 500 m2.The existing cellular network in the RoI consists of a single MBS operating in the ultra high frequency (UHF) band and multiple SBSs operating in the mmWave band. The capacity of the existing cellular services in the area needs to be enhanced in terms of the data rates. At the same time, we want to offload users from the existing base stations because they are over loaded. A feasible solution to this is to deploy UAVs in the RoI till the crowd disperses and we are back to routine traffic in the RoI. The users in RoI are modelled through a homogeneous poisson point process (PPP) with intensity λu users/m2.The MBS is assumed to be located at the center of the RoI. The SBSs are all operating in the mmWave band and are deployed at a minimum separation distance of 100 meters from the MBS. The distribution of SBSs and UAVs is modelled according to Matern type 1 process with minimum separation, δSBS = δUAV = 40 m , between any two SBSs or UAVs. We deploy the UAVs at a minimum distances of 100 m and 30 m from the MBS and SBSs respectively. The reason for keeping this separation is to effectively maximize the benefit of each deployed UAV in the region. If a UAV is deployed close to the MBS or the SBS, then the user may associate with the MBS or the SBS instead of the UAV, thereby reducing the advantage of UAV deployment in that region. We define the total SBSs in the region as NSBS, and the total UAVs deployed as NUAV. Similarly the total number of users in the RoI is denoted by n.

2.1  Channel model In the considered scenario, the MBS is operating in the sub-6 GHz frequency spectrum while the SBSs and the UAVs are operating in the mmWave bands. However, to minimize the interference between the small cell users and the UAV users, we use frequency bands of 28 GHz and 73 GHz for UAVs and SBSs, respectively. First we model the path loss for the UHF link, which is given, in dBs, as



 4π  LU (d ) = 20 log   + 10α log(d ) + χ u ,  λU 

(5.1)

where d is the distance between user and the connected base station, λU is the wavelength depending upon the carrier frequency, fu, α denotes the path loss exponent for the UHF frequencies and χ u denotes denotes the effects of shadowing for the UHF links.

83

UAVs: communication aspects and cellular service provisioning

Now, we define the path loss model for the SBSs operating in mmWave bands, in dBs, as SBS mm

L 

 ω + 10α SBS log(d SBS ) + χ SBS LoSlink, L mm L  (d ) =  SBS SBS SBS  ω + 10α N log(dmm ) + χ N otherwise,

(5.2)

χ LSBS and χ NSBS denotes where w is defined as ω = 32.4 + 20 log( f cSBS ,mm ) and shadowing effects for LoS and NLoS links, respectively.While α LSBS and α NSBS are the LoS and NLoS path loss exponents depending upon whether the mmWave link established between the user and the SBS is LoS or NLoS, SBS respectively. For simplicity we assume that whenever the link distance dmm is less than or equal to 30 m the established link will be LoS and vice versa. Now that we have clearly defined path loss models for the existing terrestrial network in the RoI, we define the path loss model for our aerial network consisting of UAVs, in dBs, as UAV mm

L 

 γ + 10α UAV log(d UAV ) + χ UAV LoSlink, L mm L  (d ) =  UAV UAV UAV otherwise,  γ + 10α N log(dmm ) + χ N

(5.3)

χ LUAV and χ NUAV denotes where γ is defined as γ = 32.4 + 20log(f cUAV ,mm ) and shadowing effects for LoS and NLoS links, respectively.While α LUAV and α NUAV are the LoS and NLoS path loss exponents depending upon whether mmWave link established between the user and the UAV is LoS or NLoS, respectively. For calculating the LoS probability (PLoS) of the mmWave link established between a user and a UAV, the closed form expression as derived in Ref. [9] is given as: PLoSUAV mm = 

1

UAV 1 + a exp  −b (φmm − a ) 

,

(5.4)

where the values of a and b depend upon the statistical parameters of the UAV environment, and φmm is the elevation angle between the ground user and the UAV, given by



 hUAV UAV φmm = arctan  mm UAV  xmm

 , 

(5.5)

UAV UAV where hmm is the UAV height and xmm is the horizontal distance from the user to the projection of the UAV on ground.

84

Muhammad Asim Jan, Syed Ali Hassan, Haejoon Jung

From Eq. (5.4) we go on to find the probability of nonline of sight (PNLoS) between a ground user and a UAV as: UAV PNLoSUAV mm = 1 − PLoSmm .



(5.6)

Using the probabilities obtained from Eqs. (5.4) and (5.6) we calculate the path loss, for a mmWave link established between a ground user and a UAV as: UAV UAV UAV UAV LUAV mm = PLoSmm × L mm ,L + PNLoSmm × L mm ,N .



(5.7)

The received power at each user from the UHF MBS is given, in dBs, as: Prx ,U = Ptx ,U − LU ,



(5.8)

and for mmWave links the received power is given as: 

υ Prxυ = Ptxυ + G(φmm ) − Lυmm , υ ∈ {SBS,UAV},

(5.9)

where for every mmWave link established, we add the directional gain of transmitting antenna to the received power equation. For UAVs we use the model in Ref. [12], given by UAV G(θmm )=



2 , UAV 1 − cos (θmm / 2)

(5.10)

where



 18 dB if φ UAV ≤ θ UAV , UAV mm mm G(φmm )= 10 dB otherwise, − 

(5.11)

we assume that the maximum mmWave radiation is concentrated in the UAV = 30 degree, subtended from the UAV. Similarly cone with solid angle, θmm directional antenna gain of the SBS is calculated as G(φ 

SBS mm

SBS  θ SBS SBS  18 dB if −θmm ≤ φmm ≤ mm , )= 2 2  −10 dB otherwise, 

(5.12)

SBS is the angle in horizontal plane between the ground user and where φmm the SBS and the mmWave beamwidth of the radiation pattern from the SBS SBS = 30 degree, arc in horizontal plane. is concentrated within an, θmm

UAVs: communication aspects and cellular service provisioning

85

2.2  User association metric Association of user with a particular tier is an important consideration in any heterogeneous network. In the considered scenario, we have a three tier HetNet with the UHF tier consisting of the MBS, while the other two tiers being the SBSs and UAVs are operating in separate mmWave bands. We associate the user with any tier based on the maximum biased average received power given, in dBs, by 

υ Prx ,ζ = Ptx ,ζ − Lζ + βmm + G(φmm ), ζ ∈ {U,SBS,UAV},

(5.13)

where the power received from any tier is denoted by Prx,ζ , the power transmitted is given as Ptx,ζ , the path loss associated with each tier is Lζ , the bias factor for mmWave tier, consisting of both SBSs and UAVs, is denoted by βmm and the directionality gain in mmWave band will be added in case of a user link with a SBS or a UAV. Bias factor is important for mmWave tier because due to the greater propagation losses, power received from the mmWave tier will always be lesser than the UHF tier. To create a balance in the network and offload users from the UHF tier to the mmWave tier, based on the network or the user requirements, we use bias factor for mmWave tier. We can also use separate bias factor values for SBS and UAV tier; however, for our study we assume same bias factor value for mmWave tier consisting of both the ground-based SBSs and the UAVs.

3  Performance analysis For performance analysis of our system, we consider a time division multiple access scheme in our network. The user will connect to any tier based on already discussed user association scheme. For a user connected to UHF tier, there will be no interference as we have assumed only a single MBS in the RoI. Therefore, for every user connected to the MBS, the signal-to-interference and noise ratio (SINR) will be calculated as:



SINRU =

Prx ,U , σ U2

(5.14)

where the noise power, in dBs, is calculated as σ U2 = −174 dBm/Hz +  10 log (BU  ) + NF, and NF is the noise figure of receiver.We have assumed that there is only one MBS in the RoI, therefore interference is assumed to be zero in the UHF tier. Suppose, nu, is the number of users connected to

86

Muhammad Asim Jan, Syed Ali Hassan, Haejoon Jung

the UHF tier or the MBS. For each user connected to the MBS the downlink rate received will be



BU log 2(1 + SINRU ), nu

RU =

(5.15)

where BU is the total bandwidth available at the MBS. The user will connect to mmWave tier only if maximum average biased received power from either SBS or UAV is greater than power received from MBS which is operating in UHF spectrum. In that case, the connected SBS or UAV will be serving base station while the remaining will act as interferers. However, it is to be noted that SBS and UAVs are operating at different mmWave spectrums of 73 GHz and 28 GHz, respectively. Therefore, for user connected to a SBS, only the remaining SBSs will act as interferers. Similarly, for a user connected to a UAV, interference will come only from the remaining UAVs.The SINR for the mmWave tier is given by SINRυ = 

Prx ,υ

∑P

I

+ σ υ2

, υ ∈{SBS,UAV},

(5.16)

I

where PI is the sum of powers received from all interfering base stations which depends upon the user link is established with either a SBS or a UAV. The downlink data rates received by each user in the mmWave tier will be given by



Rυ =

Bυ log 2(1 + SINRυ ). nυ

(5.17)

Now that we have calculated the SINR and the downlink data rates for each user, we evaluate the performance of our HetNet by defining two parameters, that is, SINR coverage and rate coverage. The SINR coverage, at a given threshold τ, is given by 

Pcov (τ ) = P(SINRζ > τ ).

(5.18)

Similarly, for a given rate threshold, Γ, the rate coverage probability is defined as



Γ×nζ   Prate (Γ ) = P(rate > Γ ) = P SINRζ > 2 B − 1 .  

(5.19)

87

UAVs: communication aspects and cellular service provisioning

4  Results For simulation of our system model, we model a system with a single MBS at the center of the RoI and with equal number of SBS and UAVs in the mmWave tier. For analysis purpose we carry out Monte Carlo simulations. Table 5.1 defines simulation parameters for our system model. We analyze performance of the HetNet on already defined metrics of SINR and rate coverage. Different bias factor values, βmm, are used to offload maximum users toward the mmWave tier. Fig. 5.2 shows the SINR coverage probability for the HetNet. The HetNet consists of a UHF MBS and mmWave SBSs and UAVs. The results are obtained at different bias factor values for mmWave tier. The reason is that power received from UHF tier will always be greater due to much better propagation characteristics. On the contrary mmWave encounters much higher losses due to propagation and blockages. As our user association with base station is dependent upon maximum power therefore without using bias factor for mmWave tier, the majority of users will associate with MBS. This will not only overload the MBS, but also reduce the advantage that mmWave SBS and UAVs offer in terms of data rates. Association based on maximum average biased received power means that even if power received from mmWave tier is slightly lower than UHF tier, depending upon the bias factor value, the user will still associate with the mmWave tier. From Fig. 5.2, we observe that as bias factor value is increased, the overall coverage

Table 5.1  Simulation parameters. Parameter Value

Parameter SBS c ,mm

UAV c ,mm

Value

fc,U

2.4 GHz

f

BU

20 MHz

Bmm

400 MHz

Ptx,U

40 dBm

Ptx,mm

30 dBm

30°

θ

30°

Environment

a = 9, b = 0.11

Std ( χ L ,mm )

Users

500

NF

9 dB

α Ht, UAV

θ

UAV mm

, f

SBS mm

Std ( χ N ,mm )

73 GHz, 28 GHz

5.2 7.2 2, 3.3

3

α L ,mm , α N ,mm χu

30 m

N SBS = N UAV

20

4

88

Muhammad Asim Jan, Syed Ali Hassan, Haejoon Jung

Figure 5.2 SINR coverage probability related to bias factors, (MBS = 1, NSBS = 20 , NUAV = 20 ).

of the HetNet reduces. However at βmm = 15dB, and τ = −5dB, the network SINR coverage is still well above 0.9. The real advantage of offloading the users toward the mmWave tier can be observed in Fig. 5.3 when we see the rate coverage of the HetNet. As value for βmm increases, user association with the mmWave tier will start to increase and the network will be able to exploit the huge bandwidth available at the mmWave tier. From Fig. 5.3, we can see that if do not use bias factor value for mmWave tier, then, at rate threshold of Γ = 1 Mbps as low as 5% of users are receiving data rates of 1 Mbps and above. Considering by modern smart city standards this is very low. However, when we use bias factor value, βmm = 15dB, users receiving data rates in excess of 1 Mbps increases to 27%. Which means that we have achieved almost 5 times increase in number of users achieving required data rates. Fig. 5.4 shows the advantage of deploying mmWave UAVs to assist the terrestrial cellular infrastructure. A bias factor value of 15 dB has been used for the mmWave tier so that traffic can be offloaded from MBS to the SBSs and UAVs.When there are no UAVs deployed and only ground-based MBS

UAVs: communication aspects and cellular service provisioning

89

Figure 5.3 Rate coverage probability related to bias factors, ( n = 500 , MBS =1 , NSBS = 20 , NUAV = 20 ).

Figure 5.4 Rate coverage probability related to number of UAVs, (n = 500, MBS =1, NSBS = 20 , β mm =15dB).

90

Muhammad Asim Jan, Syed Ali Hassan, Haejoon Jung

and SBS are providing communication services, the user data rates have dropped considerably. This is due to the fact that the number of users has increased considerably and the existing communication resources are not sufficient to meet the data rates requirements. When we deploy mmWave UAVs in the area, then the average data rates increases in the RoI. As we increase the number of mmWave UAVs, N UAV , from 10 to 40, the number of users receiving data rates of 1 Mbps and above increases from 10% to 42% .

5  Conclusion Future smart cities will require huge bandwidth and communication network resources to sustain its operations. As smart cities emphasize on hyper connectivity and integration, therefore mmWave UAVs offer a flexible and rapid solution for meeting any requirement. Efficient placement of UAVs inside the smart cities will enhance the capacity of the existing network in an effective manner. Moreover, through effective placement of UAVs based on user and network demands we can further enhance communication service provisioning inside smart cities.

References [1] A. Giyenko, Y. Im Cho, Intelligent UAV in smart cities using IoT, in: 2016 16th International Conference on Control, Automation and Systems (ICCAS), IEEE, 2016, pp. 207–210. [2] S.A.R. Naqvi, S.A. Hassan, H. Pervaiz, Q. Ni, Drone-aided communication as a key enabler for 5G and resilient public safety networks, IEEE Commun. Mag. 56 (1), 2018, 36–42. [3] F. Qi, X. Zhu, G. Mang, M. Kadoch, W. Li, UAV network and IoT in the Sky for future smart cities, IEEE Network 33 (2) (2019) 96–101. [4] J.G. Andrews, S. Buzzi, W. Choi, S.V. Hanly, A. Lozano, A.C. Soong, et al. What will 5G be?, IEEE J. Select. Areas Commun. 32 (6) (2014) 1065–1082. [5] S. Wan, J. Lu, P. Fan, K.B. Letaief, To smart city: public safety network design for emergency, IEEE Access 6 (2017) 1451–1460. [6] H. Menouar, I. Guvenc, K. Akkaya, A.S. Uluagac, A. Kadri, A. Tuncer, UAV-enabled intelligent transportation systems for the smart city: applications and challenges, IEEE Commun. Mag. 55 (3) (2017) 22–28. [7] X. Li, M.C. Chuah, S. Bhattacharya, Uav assisted smart parking solution, in: 2017 International Conference on Unmanned Aircraft Systems (ICUAS). 1em plus 0.5em minus 0.4em IEEE, 2017, pp. 1006–1013. [8] T.S. Rappaport, S. Sun, R. Mayzus, H. Zhao, Y. Azar, K. Wang, et al. Millimeter wave mobile communications for 5G cellular: It will work!, IEEE Access 1 (2013) 335–349. [9] A. Al-Hourani, S. Kandeepan, S. Lardner, Optimal LAP altitude for maximum coverage, IEEE Wireless Commun. Lett. 3 (6) (2014) 569–572. [10] T.S. Rappaport, Y. Xing, G.R. MacCartney, A.F. Molisch, E. Mellios, J. Zhang, Overview of millimeter wave communications for fifth-generation (5G) wireless networks

UAVs: communication aspects and cellular service provisioning

91

a with a focus on propagation models, IEEE Trans. Antennas Propagat. 65 (12) (2017) 6213–6230. [11] M. Mezzavilla, M. Polese, A. Zanella, A. Dhananjay, S. Rangan, C. Kessler, et al. Public safety communications above 6 GHz: challenges and opportunities, IEEE Access 6 (2017) 316–329. [12] R. Kovalchukov, D. Moltchanov, A. Samuylov, A. Ometov, S. Andreev, Y. Koucheryavy, et al. Analyzing effects of directionality and random heights in drone-based mmWave communication, IEEE Trans.Vehicul. Technol. 67 (10) (2018) 10064–10069.

CHAPTER SIX

AI simulations and programming environments for drones: an overview Oluwaseun Priscilla Olawalea, Kamil Dimililerb, Fadi Al-Turjmanc

Department of Software Engineering, Research Center for AI and IoT, Near East University, Nicosia, Mersin, Turkey b Department of Electrical and Electronic Engineering, Research Center for AI and IoT, Near East University, Nicosia, Mersin, Turkey c Department of Artificial Intelligence, Engineering, Research Center for AI and IoT, Near East University, Nicosia, Mersin, Turkey a

Chapter outline 1 Introduction 1.1 What is a simulator 2 Why simulators are important in drones 3 Where it is applicable 4 Simulation environments 5 Artificial intelligence types and issues 5.1 AI types in drones’ simulations 5.2 AI Issues in drones’ simulations 5.3 Programming languages 6 Applications 6.1 Performance assessment parameters 7 Other assessment methods 8 Open research issues 9 Conclusion References

93 93 95 96 99 100 100 100 101 101 101 102 103 103 104

1  Introduction 1.1  What is a simulator Simulators are well known as devices used for imitating the actual concept of a system [1] under experimental conditions. In order to simulate any system at all, mathematical models [2] are usually generated and converted to computer models via programming as shown in Fig. 6.1.

Drones in Smart-Cities. http://dx.doi.org/10.1016/B978-0-12-819972-5.00006-9 Copyright © 2020 Elsevier Inc. All rights reserved.

93

94

Oluwaseun Priscilla Olawale, Kamil Dimililer, Fadi Al-Turjman

Figure 6.1  Simulation model.

Simulators are generally used in the field of robotics [3], medical applications to virtually understand a system. They help in economizing cost of the actual implementation of robotic systems [4]. Also, artificial intelligence (AI) when integrated with simulation plays a very important role before the real system is deployed, and even afterwards. For example, the (United States Patent No. US 2019/0176987 A1, 2019) states that in the case of a fire outbreak, AI when applied to the development of drone applications with sensors, are capable of conveying fire extinguishing drones to the fire outbreak location (Fig. 6.2). The system involved achieves this by: 1. receiving data such as locations of possible fire outbreak; 2. analyzing the data received; 3. identifying the main location of the fire outbreak, 4. dispatching the fire extinguishing drone(s), and 5. releasing the content of the fire extinguisher to suppress the fire and continues from (1) above if there exist other locations with a fire outbreak. From the aforementioned examples, we see that drones built and simulated with AI can track fire outbreak by locating the areas affected, and

AI simulations and programming environments for drones: an overview

95

Figure 6.2  Fire extinguishing drone in action. (From Ref. [5].)

suppress the fire. On a good day, controlling this kind of disaster may be extremely dangerous for the fire fighters. But with drones, this operation is 90% faster and possible because sensors and other hardware such as infrared cameras are embedded in them [5]. AI are able to handle specific tasks that humans may not be able to carry out [6]. The main reason for the simulation of systems (such as surgical robots, etc) with AI algorithms is to enable them learn how to perform tasks without the aid of any human intervention.

2  Why simulators are important in drones Drones, otherwise known as quadcopter or unmanned aerial vehicles [7] are robots [8], technically designed for ethereal use. Although they do not require the presence of a pilot, they can be controlled remotely and are majorly used in fields of intelligence, security and military. They are also used for wireless Internet distribution [9], telecommunications [10], commercial delivery [11], disaster control such as wildfires (United States Patent No. US 2019/0176987 A1, 2019), civil, and industry applications [8], surveillance, agriculture, and spying enemy territory [12], education, archaeology, emergency response, weather and storm analysis, photography [3], and movie making [13]. It is very essential for drones to be simulated and evaluated before they are formally set up or put into use [14]. These drones are usually simulated in a virtual air space via controllers. The objective of simulators in drones is to study the behavior of drone-based applications, provide training for

96

Oluwaseun Priscilla Olawale, Kamil Dimililer, Fadi Al-Turjman

navigation and control [3]. Mairaj et al. states that there are generally three important reasons why drones are simulated [13]: • evaluating new technology, • and low-cost training, and • research and development. Simulators are also used to investigate the performance of drone applications [15]. Because drone simulation requires the computation of various variables, the activities involved in its simulation may not be easy. If these variables are not properly selected, the following results may bring about negative effects [15].These simulators are used to create mathematical models that may be complex and extremely similar to the main system, thereby reducing the experimental cost and time.

3  Where it is applicable Drones are built with several sensors and hardware for different purposes and for this particular reason, they are considered as an integral part of IoT.They both comprise of AI, sensing, data/information processing, and communication networks [16] (Fig. 6.3). These sensors send and receive data through AI techniques embedded in their systems, if the systems are actually built with AI algorithms (United

Figure 6.3  Components of drones. (From Ref. [17].)

AI simulations and programming environments for drones: an overview

97

States Patent No. US 2019/0176987 A1, 2019) and perform the required operation afterwards. Usually, we simulate some of the following components of a complete drone (Figs. 6.4 and 6.5):

Figure 6.4  Drone vision control. (From: Ref. [18].)

Figure 6.5  Trajectory planning for drones.

98

Oluwaseun Priscilla Olawale, Kamil Dimililer, Fadi Al-Turjman

• Vision and control Drones perform vision through optical sensing or cameras. This way, they perceive and understand their environment and detect objects for proper navigation [19]. They are also able to carry out control operations through other sensors and actuators. • Trajectory planning Velocity, location, coordinates, and flight path [20] are important parameters to consider when planning trajectory for take-off, landing and safety in drones. Trajectory planning is used to execute air take-off position, target position, goals, and tasks of the flight [21]. This enhances the flight movement of drones especially in environments with many obstacles • Automatic navigation system Simulating drones with AI greatly increases its speed to over 90 miles per hour, which makes the drones more accurate than an onboard pilot [22]. Simulating the automatic navigation system reduces the risk of poor navigation performance [15] and enables the system make use of precise units for the calculation of measurements (Fig. 6.6).

Figure 6.6  Wi-Fi drone communication.

99

AI simulations and programming environments for drones: an overview

• Communication System Drones are the capable of transmitting data such as speed, direction, fixed points [10] with the aid of Wi-Fi for as long, as they are within the same transmission radius [23], which actually reduces the unit of its memory and power. They communicate with other drones, the ground station, and the user applications involved. Parameters used for communication simulation includes bandwidth, latency, throughput, and data rate [10].

4  Simulation environments The Table 6.1 describes some simulation environment that can be used test controllers and algorithms. Table 6.1  Simulation environments.

Reference

Simulation types

Platform

Features

Programming language

http://ardu- Software in Linux, Win- Simulate C++ pilot.org/ the loop dows drones even (SITL) without the simulator hardware https://flytbase.com/

https://developers. google. com/ https://hub. packtpub. com/ https://unity. com/

AI algorithms

Reinforcement learning algorithms FlytBaseIoD — FlytOS C++, Py- Machine platform FlytCloud thon, learning JS Rest, and vision ROS, algoJS Webrithms Socket Google — Image proJavaScript, Machine Earth cessing Python learning engine functions and deep Machine learning learning algorithms AirSim Windows, Deep learning Java, Machine Linux Robotic Python, learning Models C++, algoSensors C# rithms Unity Platform Create other C#, JavaS- Machine indepensimulation cript, learning dent environUnity algoment Script, rithms BOO

100

Oluwaseun Priscilla Olawale, Kamil Dimililer, Fadi Al-Turjman

5  Artificial intelligence types and issues 5.1  AI types in drones’ simulations • Supervised versus unsupervised Although supervised learning is preferable [24] for simulation with AI algorithms, it is concerned with labeled data where input, output and an agent are present.The agent usually maps the output with known inputs. Its algorithms are used for prediction in most cases. Unsupervised learning on the other hand does not deal with labeled data and it does not require the presence of an agent [25]. • Machine learning (ML) versus deep learning (DL) ML and DL are both classified as subgroups of AI and are similar.The difference between them is that DL employs the use of artificial neural networks. While ML requires the presence of humans to be trained, DL generates several layers of network with or without the involvement of any human [26].

5.2  AI Issues in drones’ simulations As much as AI is popularly integrated in the simulation of drones and applied in other facets of life, there still exist some issues with it. Some of these issues include: 1. High cost: Embedding AI in robotic systems, especially drones can be very expensive.This does not leave out the general maintenance, fix, and repair of such systems [27]. 2. Human error: Yes, systems with AI can detect mistakes but are not excluded from making mistakes. They make decisions based on trained data and may not be able to make intelligent decisions that humans would make in some critical conditions. This implies that they can be imperfect, and make mistakes especially when trained with the wrong data. 3. Security and privacy: AI generally pose threats to human safety and privacy. This technology has aided cyber criminals to intrude into the privacy of other individuals [28]. 4. Adaptability: It may be impossible for drones with AI to adapt to changes in environments. For example, drones with AI may find it difficult to steer in environments they are not familiar with [29] since they are trained with data. 5. Required expertise: As AI evolves, there is need for the required skills in programming and other areas of AI which is limited. These experts are scarce because they put a lot of effort in acquiring the skills [30] that are required for developing the systems.

101

AI simulations and programming environments for drones: an overview

Table 6.2  Comparison of C++, java, and python programming languages. ProgramPlatform ming Object IndepenAI Language Main Use Oriented dent Build Code Length Support

C++

Application, system

Yes

No

Java

Application, Yes business, mobile development, server-side, client-side, and web development Application, Yes web development, scripting, scientific computing, and artificial intelligence

Yes

Python

Yes

Compiled Code length Yes is a bit lesser, 1.5 times less that java Compiled Java has Yes quite huge code

Interpreted

Smaller Yes code length, 3–4 times less than java

5.3  Programming languages From Table 6.1, we can deduce that the major languages used in simulation applications are Java, C++, and Python (Table 6.2). Although each of these three programming languages has some unique purpose, python is majorly used because it works best when combined with neural networks and AI [31] (Table 6.3).

6  Applications 6.1  Performance assessment parameters • Coverage radius With antennas, drones have a precise coverage radius that is determined by its altitude. The radius coverage enables the base to receive and send signals [32].This also takes parameters such as throughput and communication into account [32a].

102

Oluwaseun Priscilla Olawale, Kamil Dimililer, Fadi Al-Turjman

Table 6.3  Application of drone simulation with artificial intelligence. Reference Framework Algorithm Application

[12] [8] [15]

[4]

Robot operating AdaBoost classi- To simulate condition in war system (ROS) fier and pinhole zone for spying the enemy Matlab-simulink PD, PID, and To dynamically compute the framework back-stepping quad-copter altitude and control attitude QT C++ frame- General UAV To design a multi-agent applicawork algorithms tion in which a team of UAVs are employed for a certain mission in an environment UTSim (unreal Sense and avoid To simulate UAV air trafengine) (S&A) and colli- fic integration, control, and sion avoidance communication

• Drones’ throughput Drones are expected to forward data to stationary base stations. The rate at which they are able to perform this task makes throughput a parameter for their assessment. Considering data heterogeneity and time sensitivity is a key factor in any good simulator • Scalability Scalability of drones greatly depends on its architecture [33]. For drones to thrive in airspace, their system needs to be robust [34]. • Battery lifetime The batteries of UAVs are restricted in capacity and hence, their lifespan. Its lifespan determines how long it lasts in performing its operations, especially the flight duration. Petrides et al. in their research stated that work is being done by other researchers to discover a way to manage the power consumption of drone batteries.

7  Other assessment methods • UAV testbeds The testbeds serve as a platform for testing and analyzing UAVs with algorithms, models, network protocols, and user applications [35]. • Integrated testbeds and simulation Since there are other subsystems (such as communication, navigation, trajectory and control), we need to evaluate their accuracy of their security and flexibility by integrating them. This produces durable architecture of the testbeds [36].

AI simulations and programming environments for drones: an overview

103

8  Open research issues General open research issues of AI simulations and programming environment may include the following: 1. Behavior and control: Drone systems and applications may seem complicated [37] since they are built with AI algorithms. As a result, it is able to intelligently navigate once it is in the air. The same drone may be remotely controlled too. Controlling it when in motion may be difficult sometimes [19] because the ability of human operators to predict them grows less with time. If it looses control, it may crash and get damaged. 2. Computer vision for drones generally is challenging because of the nature of their structure [38]. 3. Security and viability issues in the GUI applications are on a high rise. This affects the operation of the drone network [13]. 4. Communication: Choosing the network for drone communication may be regarded as a laborious task [39]. The network topology, architectural design, and routing are essential components that have to be considered for achieving efficient communication [40,41]. 5. Vague simulation environments: It is difficult to obtain required results especially when several different simulation environments are used. Simulating the AI algorithms again and again causes delay which may lead to poor documentation. This makes it very difficult to compare the AI algorithms used for simulation and the accuracy of their performance [4].

9  Conclusion We presented an overview of AI simulations and programming environments for drones by reviewing some available literatures from 2010 to November 2019. First, we briefly analyzed the use of simulators with AI, particularly on drones otherwise known as quadcopter or unmanned aerial vehicles. We then compared simulation environments and programming languages used in the development of their system application and finally discussed issues associated with AI and simulation research. AI is “Compulsory” as it plays a very vital role in the simulation of drones. Without it, deploying and maintaining the actual drone system will be extremely expensive. It will also be impossible to program them carry out specific tasks, especially those that endanger human life.

104

Oluwaseun Priscilla Olawale, Kamil Dimililer, Fadi Al-Turjman

References [1] K.P. White, R.G. Ingalls, 2009. Introduction to simulation, in: M.D. Rossetti, R.R. Hill, B. Johansson, A. Dunkin, R.G. Ingalls (Eds.), Proceedings of the 2009 Winter Simulation Conference, Piscataway, New Jersey: Institute of Electrical and Electronics Engineers, pp. 12–23.. [2] C. Kamali, S. Jain, Hardware in the loop simulation for a mini UAV, in: 4th IFAC Conference on Advances in Control and Optimization of Dynamical Systems ACODS 49 (1) (2016) 700–705, doi:10.1016/j.ifacol.2016.03.138. [3] N. Bhushan, UAV: trajectory generation and simulation, 2019. [4] A. Al-Mousa, B.H. Sababha, N. Al-Madi, A. Barghouthi, R. Younisse, UTSim: a framework and simulator for UAV air traffic integration, control, and communication, IJARS (2019) 1–19, doi: 10.1177/1729881419870937. [5] T. Frey. , Using drones to eliminate future forest fires, 2018, Available from: https:// futuristspeaker.com/. [6] Y.K. Ever, B. Sekeroglu, K. Dimililer, Classification analysis of intrusion detection on NSL-KDD using machine learning algorithms, in: In International Conference on Mobile Web and Intelligent Information Systems, Cham, Springer, 2019, pp. 111–122, doi:10.1007/978-3-030-27192-3_9. [7] A. Bermúdez, R. Casado, G. Fernández, M. Guijarro, P. Olivas, Drone challenge: a platform for promoting programming and robotics skills in K-12 education, IJARS 16 (2019) 1–19, doi: 10.1177/1729881418820425. [8] K. Patel, J. Barve, Modeling, simulation and control study for the quad-copter UAV, in: 9th International Conference on Industrial and Information Systems, IEEE, Gwalior, India, 2014, doi:10.1109/ICIINFS.2014.7036590. [9] A. French, M. Mozaffari, A. Eldosouky, W. Saad, Environment-aware deployment of wireless drones base stations with Google Earth simulator, in:Workshop on UNmanned Aerial Vehicle Applications in the Smart City: from Guidance Technology to Enhanced System Interaction, IEEE, 2019, pp. 868–873. [10] F. Al-Turjman, J.P. Lemayian, S. Alturjman, L. Mostarda, Enhanced deployment strategy for the 5G Drone-BS using artificial intelligence, IEEE Access 7 (2019) 75999–76008, doi: 10.1109/ACCESS.2019.2921729. [11] T.R. Gulden, The Energy Implications of Drones for Package Delivery, RAND Corporation, Santa Monica, California, (2017)doi: 10.7249/RR1718.1. [12] M.A. Ma'sum, M.K. Arrofi, G. Jati, F. Arifin, M.N. Kurniawan, P. Mursanto, W. Jatmiko, Simulation of intelligent Unmanned Aerial Vehicle (UAV) for military surveillance, in: International Conference on Advanced Computer Science and Information Systems, 2013, doi:10.1109/ICACSIS.2013.6761569. [13] A. Mairaj, A.I. Baba, A.Y. Javaid, Application specific drone simulators: recent advances and challenges, Simul. Model. Pract. Th 94 (2019) 100–117, doi: 10.1016/j.simpat.2019.004. [14] A.I. Hentati, L. Krichen, M. Fourati, L.C. Fourati, Simulation tools, environments and frameworks for UAV systems performance analysis, in: 14th International Wireless Communications & Mobile Computing Conference, IEEE, 2018, pp. 1495–1500, doi:10.1109/IWCMC.2018.8450505. [15] M.D. Benedetti, F. D’Urso, F. Messina, G. Pappalardo, C. Santoro, 3D simulation of unmanned aerial vehicles, n.d., 7–12. [16] F. Al-Turjman, Y.K. Ever, E. Ever, H.X. Nguyen, D.B. David, Seamless key agreement framework for mobile-sink in IoT based cloud-centric secured public safety sensor networks, IEEE Access 5 (2017) 24617–24631, doi: 10.1109/ACCESS.2017.2766090. [17] Drone Solutions - TE | DigiKey, n.d. Available from: https://www.digikey.be/. [18] O. Liang, Best quadcopter FPV simulators | racing drone trainer, 2015. Available from: https://oscarliang.com.

References

105

[19] S. Jain, How to add person tracking to a drone using deep learning and NanoNets, 2018. [20] L. Tan, J. Wu, X.Yang, S. Song, Research on optimal landing trajectory planning method between an UAV and a moving vessel, Appl. Sci 9 (2019)doi: 10.3390/app9183708. [21] H. Xu, X. Xu, Y. Li, X. Zhu, L. Jia, D. Shi, Trajectory planning of Unmanned Aerial Vehicle based on A* algorithm, in:The 4th Annual IEEE International Conference on Cyber Technology in Automation, Control and Intelligent, Hong Kong, IEEE, China, 2014, doi:10.1109/CYBER. 2014.6917508. [22] J. Swearingen, A.I. is flying drones (very, very slowly). Available from: https://www. nytimes.com/2019/03/26/technology/alphapilot-ai-drone-racing.html, 2019. [23] S. Iranmanesh, R. Raad, A novel data forwarding strategy for a drone delay tolerant network with range extension, Electronics 8 (2019) 659, doi: 10.1109/CYBER.2014.6917508. [24] K. Dimililer, Y.E. Kirsal, S.M. Mustafa, Vehicle detection and tracking using machine learning techniques, in: R.A. Aliev, et al. (Eds.), ICSCCW2019, in: 10th International Conference on Theory and Applications of Soft Computing, Computing with Words and Perceptions, AISC 1095, Springer Nature Switzerland AG2020, 2019, pp. 1–9. [25] S.Y. Choi, D. Cha, Unmanned aerial vehicles using machine learning for autonomous flight; state-of-the-art, Adv. Robot (2019)doi: 10.1080/01691864.2019.1586760. [26] A. Kapoor, Deep learning vs. machine learning: a simple explanation, 2019. Available from: https://hackernoon.com/. [27] DataFlair Team, Pros and cons of artificial intelligence – a threat or a blessing? Available from: https://data-flair.training/. [28] S. Dilek, H. Çakır, M. Aydın, Applications of artificial intelligence techniques to combating cyber crimes: a review, Int. J. Artif. Intel. Appl 6 (1) (2015) 21–39, doi: 10.5121/ ijaia.2015.6102. [29] L. Huang, H. Qu, M. Fu, W. Deng, Reinforcement learning for mobile robot obstacle avoidance under dynamic environments, in: Pacific Rim International Conference on Artificial Intelligence, Cham, Springer, 2018, pp. 441–453, doi:10.1007/978-3-31997304-3_34. [30] I. Tuomi, The impact of artificial intelligence on learning, Teaching and Education (2018), doi:10.2760/12297. [31] L. Whitney, Why Python is considered the top programming language ahead of JavaScript and C++, 2019. Available from: https://www.techrepublic.com/. [32] P.V. Klaine, J.P. Nadas, R.D. Souza, M.A. Imran, Distributed drone base station positioning for emergency cellular networks using reinforcement learning, Cognit. Comput 10 (5) (2018) 790–804, doi: 10.1007/s12559-018-9559-8. [32a] F. Al-Turjman, H. Zahmatkesh, I. Aloqily, R. Dabol, Optimized unmanned aerial vehicles deployment for static and mobile targets monitoring, Comp. Commun. J. 149 (2020) 27–35. [33] S. Hausner, Scalability, safety and security for drones, 2019. [34] C. Downing, , DDS and drones: scalability, safety and security in the sky, 2019. Available from: https://www.rti.com/. [35] M. Imran, A.M. Said, H. Hasbullah, A survey of simulators, emulators and testbeds for wireless sensor networks, in: 2010 International Symposium on Information Technology, IEEE, Kuala Lumpur, Malaysia, 2010, doi:10.1109/ITSIM.2010.5561571. [36] H. Neema, B. Potteiger, X. Kousoukos, G. Karsai, P. Volyesi, S. Janos, Integrated simulation testbed for security and resilience of CPS, in: The 33rd ACM Symposium on Applied Computing - Cyber-Physical Systems Track, Pau, France, 2018. doi:10.1145/3167132.3167173. [37] A. Ilachinski, 2017, AI, robots, and swarms: issues, questions, and recommended studies, CNA Corporation. Available from: https://www.cna.org/cna_files/pdf/DRM-2017U-014796-Final.pdf.

106

Oluwaseun Priscilla Olawale, Kamil Dimililer, Fadi Al-Turjman

[38] M. Sanfourche, J. Delaune, G.L. Besnerais, H.D. Plinval, J. Israel, P. Cornic, et al. , Perception for UAV: vision-based navigation and environment modeling. AerospaceLab; 2012, p. 1–19. [39] V. Sharma, Advances in drone communications, state-of-the-art and architectures, Drones 3 (21) (2019) doi:10.3390/drones3010021. [40] L. Gupta, R. Jain, G.Vaszkun, , Survey of important issues in UAV communication networks, IEEE Commun. Surv. Tutor. (99) (2015). [41] F. Al-Turjman, M. Abujubbeh, A. Malekoo, L. Mostarda, UAVs assessment in softwaredefined IoT networks: an overview, Comp. Commun. J. 150 (15) (2020) 519–536. [42] J.E. Beecham, United States Patent No. US 2019/ 0176987 A1, 2019.

CHAPTER SEVEN

Smart agriculture framework using UAVs in the Internet of Things era Fadi Al-Turjmana, Hamit Altiparmakb

Department of Artificial Intelligence Engineering, Research Center for AI and IoT, Near East University, Nicosia, Mersin, Turkey b Department of Computer Engineering, Near East University, Nicosia, Mersin, Turkey a

Chapter outline 1 Introduction 1.1 What is smart agriculture? 1.2 Why we need it? 1.3 Where can we apply it? 2 System model 2.1 Multispectral imaging device 2.2 Unmanned aerial vehicle 3 The smart framework 3.1 Factors affecting grain production 3.2 Plant nutritional elements 4 Discussions 5 Concluding remark References

107 108 109 109 110 111 112 112 116 117 118 120 120

1  Introduction This chapter aims to provide a framework for the smart agriculture based [1] on scientific data, to implement it in collaboration with farmers and agricultural experts and to perform comprehensive performance measurements in agricultural fields. Multispectral imaging (MSG), unmanned aerial vehicles (UAVs), and IoT will be used to realize the vision of this chapter. When this framework is realized, it can contribute to both the agricultural industry and its natural resources. Farmers can be provided with a means to optimally control and respond to factors affecting crop quality (e.g., irrigation and fertilization, plant diseases, weeds, animal pests and extreme climate changes) [2]. One of the factors that disrupt the standard of agricultural production is that the soil structure varies from region to region. Apart from the variability Drones in Smart-Cities. http://dx.doi.org/10.1016/B978-0-12-819972-5.00007-0 Copyright © 2020 Elsevier Inc. All rights reserved.

107

108

Fadi Al-Turjman, Hamit Altiparmak

Table 7.1  The used abbreviations in this chapter. Abbreviation Explanation

MSG NIR NDVI IoT IAS TRNC HLB IAS UAVs MMHOS/CM O2 CO2 NM

Multispectral imaging Near InfraRed Normalized differential vegetation index Internet of Things Intelligent agriculture sensor Turkish Republic of Northern Cyprus Huanglongbing Intelligent agricultural sensors Unmanned aerial vehicles Millimhos per centimeter Oxygen Carbon dioxide Nanometers

from region to region, no land has a homogeneous structure in itself. When a production area is examined carefully, it can be easily observed that the plants develop very well from time to time, weak places and even drying and deaths occur in some places. So, what makes these differences within the same terrain? Of course, it is the physical, chemical, and biological structure of the soil that can change every decare or even every square meter. In this section, we overview the targeted smart agriculture scope, including the importance and the main motivations behind the usage of UAVs in it [3,4]. For more readability, we summarize all used abbreviations in this chapter as shown in Table 7.1.

1.1  What is smart agriculture? This variability lies in the emergence of many elements of intelligent agriculture [5]. Smart agriculture is a technique that enables land and crop management to increase agricultural productivity, minimizes environmental damage through more economical use of resources. In this context, it is aimed to abandon the classical production and to implement an application form that handles the land with an inhomogeneous variable approach. The main objective is to use the inputs applied in agricultural production in the amount, where and when needed. Smart agriculture is a modern agricultural production technology based on the intervention to be done in terms of spatial and temporal requirements in the field of crop cultivation in an agricultural enterprise considering these location and time criteria. Smart agriculture aims to prevent waste of resources, increase the gross return of the product and minimize the environmental pollution caused by produc-

Smart agriculture framework using UAVs in the Internet of Things era

109

tion with the use of advanced information and control systems [6].With the development of UAVs day by day and increasing technical specifications, it is a fact that it facilitates our works in many areas.

1.2  Why we need it? Plants use the energy stored in chemical bonds within carbohydrate molecules to survive. Plants go through a chemical process called photosynthesis to produce carbohydrates. Light, carbon dioxide, and water are required for photosynthesis. In order for plants to be healthy and highly efficient, photosynthesis processes should be experienced without any problems. For this, the correct irrigation and fertilization of the plant; in addition, plant diseases, weeds, animal damage and extreme climate changes should be effectively tackled. In this chapter, it is aimed to maximize the photosynthesis of plants, that is health and crop yields, by using IoT and UAVs technology [7]. With the help of the IoT and UAVs technology, it is aimed to solve all possible problems in advance by processing all the values of the soil to the cloud database instantly by using various sensors. The objectives of smart agriculture are to reduce chemical costs such as fertilizers and pesticides; protection of the environment by reducing these uses; providing high quantity and quality product; to provide more effective information flow for business and aquaculture decisions, and to establish a registration order in agriculture [5]. Some of them have already been implemented with advanced technology intelligent agriculture systems in the future;We will have the opportunity to monitor the entire farm with cloud-connected UAVs and IoT, control natural elements such as humidity, temperature with digital sensors, prevent unnecessary use of resources such as water and electricity, and reduce water pollution. In addition, farmers will be able to evaluate real-time production performance with these systems and analyze all of their products and resources in detail. With the spread of IoT in agriculture, productivity will increase significantly. Analysis of unwanted substances such as heavy metals in the soil with advanced technology, remote operation and control ability, protection of natural resources and green energy, the ability to harvest products without decay are the main objectives of companies in the agricultural sector and renewing the machines with the industrial revolution, quality and efficient agriculture is being made with each passing day [8].

1.3  Where can we apply it? Citrus, Cereal, and Olive cultivation, which are three important species in Northern Cyprus agriculture, can be selected for the implementation and

110

Fadi Al-Turjman, Hamit Altiparmak

performance measurements of the chapter.The Ministry of Agriculture and Natural Resources of TRNC and volunteer agricultural experts will be consulted and the most suitable species and areas will be determined in order to determine the type of agriculture and the areas where the framework will be implemented. When this framework is implemented, it will contribute to both, the agricultural industry and natural resources of Northern Cyprus. Farmers will be provided with means to best control and intervene the factors affecting the quality of the crop. To increase the quality and quantity of the crop, we recommended monitoring the following NDVI values over time: • extreme climate changes, • weeds, • animal damage, • plant diseases, • over sowing, • incorrect irrigation, • false fertilization, and • poor sewerage network Determining factors that threaten crop production will prevent crop losses. Furthermore, more efficient agriculture will preserve natural resources. Water is a natural resource that is limited for our country and should be used very efficiently. Another advantage of efficient agriculture is the use of less chemicals (pesticides). This is of utmost importance for both the environment and human health [9].

2  System model In order to achieve the aim of the chapter, the photosynthesis levels of the plants should be measured first. This level reports the overall health status of the plant. The level of photosynthesis can be measured by nearinfrared (NIR) electromagnetic waves at frequencies that are not visible to the eye. In plants, photosynthesis occurs in chlorophyll-containing cells called chloroplasts. These cells are mostly located in the leaves of the plant. As shown in Fig. 7.1, the health status of the plant can be determined by looking at the NIR reflection level; NIR reflection level is high in healthy leaves, medium in stressed leaves and low in dead leaves [10]. The relationship between the photosynthesis process and the wavelengths between 700 nanometers (nm) and 1000 nm in the NIR band is shown in more detail in Fig. 7.2. The importance of monitoring these

Smart agriculture framework using UAVs in the Internet of Things era

111

Figure 7.1  Identifying healthy, stressed, and dead leaves by looking at the NIR reflection level [18].

Figure 7.2  Recognizing healthy and unhealthy plants by looking at the NIR reflection index [19].

wavelengths is that the greatest changes occur in this spectrum when elements that threaten the health of plants occur [10].

2.1  Multispectral imaging device The first device required for accurate and high precision NIR measurements in our framework is MSG. Fig. 7.3 shows a high quality and affordable MSG device.

112

Fadi Al-Turjman, Hamit Altiparmak

Figure 7.3  Multispectral imaging device (Parrot Sequoia) [20].

2.2  Unmanned aerial vehicle The MSG device must be mounted on an UAV, also called the drone, to take an image of the work area. Fig. 7.4 shows a high quality, affordable drone.

3  The smart framework After the installation of the MSG device on the drone, the most widely used “normalized differential vegetation index” (NDVI) information can be obtained to measure the health status of the vegetation of the study area. To obtain this information, multispectral photographs of the workspace are processed through special software. The NDVI scale ranges from +1 to the presence of healthy vegetation and −1 to indicate that no vegetation is present [11]. Using NDVI images, vegetation density can be displayed on a dot basis.

Figure 7.4  Unmanned aerial vehicle (DJI Phantom) [21].

Smart agriculture framework using UAVs in the Internet of Things era

113

Figure 7.5  Sample images. (A) Multispectral imaging. (B) NDVI image.

Fig. 7.5 A and B illustrate MSG and an example of the resulting NVDI image [22]. Fig. 7.5A shows an image encoded by the false-color technique. All the plants actively involved in photosynthesis are shown in this reddish-purple color. Fig. 7.5B shows the NDVI version of the same image. High efficiency photosynthesis is performed at red appearing points in NDVI image [22]. Another example showing the use of NDVI images in agriculture is given in Fig. 7.6. While the plants in the area seen with red color perform photosynthesis at high efficiency, the plants in the area seen with green

Figure 7.6  An example NDVI weather map for the agricultural area [23].

114

Fadi Al-Turjman, Hamit Altiparmak

Figure 7.7  Agricultural field fertilization application using NDVI air map [24].

color perform photosynthesis at medium activity. Agricultural experts/ farmers can do the work needed to increase the overall health and yield of plants, that is, crop protection, plant diseases, weeds, animal pests, and get higher crops than agriculture [12]. In another example, Fig. 7.7, using NDVI air map of North Dakota State University, measured Nitrogen in the agricultural area and studied the optimal fertilization of the area. MSG can also be used for the diagnosis and rapid intervention of plant diseases. Fig. 7.8A and B illustrate the diagnosis of huanglongbing (HLB), a very serious disease in citrus farming, by MSG [13]. Using this identification technique, trees in an agricultural area can be distinguished as healthy (marked with the letter H in the image) and HLB-infected (marked with the letter D in the image) as shown in Fig. 7.9 and intervened without spreading the disease. Using the MSG technique, NDVI image as well as many informational parameters of vegetation can be obtained. A few examples of these parameters are the plant's water state, chlorophyll state, and nitrogen state. After processing multispectral images taken from the air, “intelligent agricultural sensors” (IAS) which works continuously and collects information on land will be used in order to understand which factors are causing the output (increase / decrease of crop quality, onset of plant diseases, animal damages). These sensors will measure air temperature, humidity and

Smart agriculture framework using UAVs in the Internet of Things era

115

Figure 7.8  Diagnosis of HLB disease by multispectral imaging technique [25]. (A) Healthy plant. (B) HLB infected plant.

116

Fadi Al-Turjman, Hamit Altiparmak

Figure 7.9  Marking healthy (marked H in the image) and HLB-infected (marked D in the image) trees using multispectral imaging in the field of agriculture [26].

pressure, soil moisture, leaf wetness, and luminance. It will also work with solar panel. Fig. 7.10 shows the high quality and affordable IAS. In addition to IAS measurements, soil analysis will also measure 16 different nutrients in the soil with one or more specific functions in the growth and development of plants. Factors and nutrients to be measured by soil analysis are as follows.

3.1  Factors affecting grain production • Soil humidity (∼25% and above) This factor represent the soil moisture or the amount of water in the soil, and therefore the air capacity of the soil is vital for plants. In order for plants to develop normally, there must be sufficient water and air in the soil.The deterioration of this equilibrium, that is, the dehydration of the soil or the decrease in air capacity due to the presence of too much water, causes diseases in plants [14].

Smart agriculture framework using UAVs in the Internet of Things era

117

Figure 7.10  Intelligent agriculture sensor [27].

• Soil pH (between 6 and 7) At high pH values, microorganism activities in the soil are reduced and as a result, the usability of the plant nutrients in the soil is greatly reduced [14,15]. • Soil salinity (4 MMHOS/CM and under) One of the most important problems of agriculture is salinity. More than one hundred countries in the world are affected by the salinity problem. The problem of salinity in irrigation of plants arises when the salt in the irrigation water is collected at the root of the plant and affects the yield [15]. • Soil temperature (2°C and over) Soil temperature is one of the most important factors that direct the physical, chemical, and biological events taking place in the soil, the physiological events taking place in the plant and the germination of the seeds and the growth and development of the plants.The main source of soil temperature is the sun's rays [16].

3.2  Plant nutritional elements Plants take a large number of elements from the environment in which they develop with their underground and above-ground organs. Scientists have determined that plants take 74 elements from the environment in which they develop. However, a very small proportion of these elements are

118

Fadi Al-Turjman, Hamit Altiparmak

essential for plants. Such elements are called absolute essential nutrients for plant growth. The absolute essential nutrients are listed below [17]. • NITROGEN (N) NO3-NH4 • PHOSPHORUS (P) P2O5 • POTASSIUM (K) K2O • CALCIUM (Ca) Ca+2 • SULFUR (S) SO4-2 • MAGNESIUM (Mg) Mg+2 • CARBON (C) HYDROGEN (H) OXYGEN (O) CO2 (from air) • IRON (Fe) Fe+2, Fe+3 • ZINC (Zn) Zn+2 • BOR (B) B4O7-2, HBO3-2, HBO3-2 • COPPER (Cu) Cu+2 • MOLYBDEN (Mo) MoO4-2 • SODIUM (Na) Na+ • CHLOR (Cl) Cl-

4  Discussions Agricultural products have been produced by farmers for centuries using traditional methods. Since each country, region and city is located under different climatic conditions, the regions have their own production, irrigation, and harvesting techniques. These techniques are suitable for producing the appropriate product for that region, with the most suitable climate, ventilation, irrigation time and frequency, intensity and duration of illumination, and nutrient type, amount, and concentration in water. It is necessary to determine the values and ensure that the plant has appropriate conditions. This situation necessitated that agriculture should be done by using modern and smart techniques with scientific infrastructure instead of traditional and local methods. Smart farming: It has three basic components—measuring all ambient conditions in the production environment, calculating nutrient, water, lighting values, and ensuring that the plant has suitable environmental conditions. Measurement of ambient conditions: Temperature, humidity, light level, CO2 ratio, O2 ratio of the plantation area, and pH and electrical conductivity values of water/nutrient mixture used for irrigation of the plant are measured. Although these values are measured in many soilless agricultural greenhouses, they are generally only used to provide instant information to the farmer and are not sent to any server or automation system.

Smart agriculture framework using UAVs in the Internet of Things era

119

Calculation of appropriate environmental values on the basis of plant species: The optimum environmental conditions for the cultivation of different plant species with the highest yields are determined by scientific studies conducted in the faculties of agriculture and biology institutes. In standard soilless agricultural systems, suitable values for plants are determined by the farmer and greenhouses are provided to reach these values by manual methods. In intelligent farming systems, the environmental conditions that the plant must possess in order to reach these optimum conditions are calculated with the help of the data collected in the sensors and submitted to the approval of the farmer. Reaching the plant to suitable environmental conditions: In intelligent agricultural systems, server-controlled automation systems are used to achieve the appropriate environmental conditions calculated in Point 2. Processes such as operating the lighting modules at appropriate times, periods and with sufficient light intensity, watering the plants in the appropriate time and time, ensuring that the water and nutrient mixture have the necessary and sufficient nutrient ratio are realized by the automation system. Intelligent agricultural systems generally use Internet of Things (IoT) and UAVs-based technologies. All sowing areas are equipped with sensors. Each sensor is connected to the central server via a computer network. The fact that no sensor can transmit data instantly does not adversely affect the overall operation of the system, and obtaining a statistically consistent and accurate efficiency from a sensor is sufficient for the consistent operation of the system. Therefore, the sensor calibrations must be accurate and repeated at regular intervals. Smart agriculture applications include sensors, remote sensing drone (UAV) and satellite technologies, artificial intelligence, robotics, and image processing technologies. Measuring the temperature of the soil and air with the sensors allows for more conscious and decision-making applications such as irrigation and spraying. Local forecasts can be reached with local weather stations. In this way, farmers can make decisions about the weather conditions in their region. Smart irrigation systems are also one of the technologies that make life easier and save water.There are now systems that can be controlled even from a mobile phone. Farmers have the chance to irrigate their fields from where they are, without the hassle and expense of going to the field. With satellite image processing systems, it is possible to detect the productivity of the terrain and the factors affecting this efficiency without ever going to the terrain.With the help of image processing technologies, it can detect diseases and pests very easily. In addition, with

120

Fadi Al-Turjman, Hamit Altiparmak

smart agriculture applications, types and amounts of fertilizers, irrigation time, soil condition, and the estimated harvesting time can be monitored in detail and in real time. The intelligent irrigation system is an IoT-based device that automates irrigation.The user can switch the engine on or off at the touch of a button. The soil moisture sensor checks whether it reaches the threshold, if it exceeds the threshold, the pump motor will shut down, otherwise the process will start automatically. This saves you water and electricity by assisting you in processing according to weather conditions such as rain etc. When the power supply is cut off and the engine is turned off, it will start automatically when there is an available power supply.The user will not worry about manually restarting the pump motor. In addition, various sensor-like data such as humidity sensor, temperature sensor can be graphically displayed in the cloud system.

5  Concluding remark The use status of connecting agricultural equipment data to manufacturers is predictive maintenance. Sensor placement and connection allow manufacturers to monitor the use of the product. They can detect performance loss and perform preventive maintenance. In addition, the collected data can help them better understand the manufacturer's needs and use them to improve the product range. The use of data collected to directly improve products and vehicles is very promising as significant productivity increases in fertilizer and fuel use can lead to optimizations. With early estimates, it can reduce fertilizer and fuel consumption and increase revenue. Finally, another important problem in the adoption of IoT in agriculture is the development of communication infrastructure in rural areas. Current wireless networks are distributed with a focus on urban areas. Therefore, communication networks need to be developed in rural areas. For rural development in Europe, wireless coverage is around 40%, largely despite the use of the agricultural fund.

References [1] F. Al-Turjman, The road towards plant phenotyping via WSNs: An overview, Comput. Elect. Agr. 161 (2019) 4–13. [2] R. Noland, S. Wells, C. Sheaffer, U.S. Patent Application No. 15/663,112, 2018. [3] F. Al-Turjman, S. Alturjman, Confidential smart-sensing framework in the IoT era, J. Supercomput. 74 (10) (2018) 5187–5198.

Smart agriculture framework using UAVs in the Internet of Things era

121

[4] F. Al-Turjman, S. Alturjman, 5G/IoT-enabled UAVs for multimedia delivery in industryoriented applications, Multimed, Tools Appl. (2018) doi: 10.1007/s11042-018-6288-7. [5] G. Singh, F. Al-Turjman, A data delivery framework for cognitive information-centric sensor networks in smart outdoor monitoring, Elsevier Comput. Commun.J. 74 (1) (2016) 38–51. [6] F. Al-Turjman, Intelligence in IoT-enabled Smart Cities, CRC Press, New York, USA, (2018). [7] Y.A.K. Utama,Y. Widianto,Y. Hari, M. Habiburrahman, Design of weather monitoring sensors and soil humidity in agriculture using Internet of Things (IoT), Trans. Mach. Learn. Artif. Intel. 7 (1) (2019) 10–110. [8] S. Liu, L. Guo, H.Webb, X.Ya, X. Chang, Internet of Things monitoring system of modern eco-agriculture based on cloud computing, IEEE Access 7 (2019) 37050–37058. [9] R. Filgueiras, E.C. Mantovani, D. Althoff, E.I. Fernandes Filho, F.F.D. Cunha, Crop NDVI monitoring based on sentinel 1, Remote Sensing 11 (12) (2019) 1441. [10] F. García-Sánchez, L. Galvez-Sola, J.J. Martínez-Nicolás, R. Muelas-Domingo, M. Nieves, Using near-infrared spectroscopy in agricultural systems, Developments in Near-Infrared Spectroscopy, IntechOpen, London, UK, (2017) p. 97. [11] R. Houborg, M. McCabe, High-resolution NDVI from Planet's constellation of earth observing nano-satellites: a new data source for precision agriculture, Remote Sensing 8 (9) (2016) 768. [12] H. Altiparmak, M. Al Shahadat, E. Kiani, K. Dimililer, Fuzzy classification for strawberry diseases-infection using machine vision and soft-computing techniques, in:Tenth International Conference on Machine Vision, vol. 10696, p. 106961N, International Society for Optics and Photonics, 2018. [13] N. Uechi, H. Katoh, T. Iwanami, T. Fujikawa, S.I. Miyata, K.Yasuda, et al. Management of huanglongbing (HLB) by an intensive vector and disease control in the surroundings of the orchard, in addition to planting HLB-free trees in Okinawa, Japan, JARQ 53 (2) (2019) 103–108. [14] K.W.T. Goulding, Soil acidification and the importance of liming agricultural soils with particular reference to the United Kingdom, Soil Use Manag. 32 (3) (2016) 390–399. [15] W. Halecki, D. Młyński, M. Ryczek, E. Kruk, A. Radecki-Pawlik, Applying an artificial neural network (ANN) to assess soil salinity and temperature variability in agricultural areas of a mountain catchment, Polish J. Environ. Stud. 26 (6.) (2017) 2545–2554. [16] R. Liu, H.L. Hayden, H. Suter, H. Hu, S.K. Lam, J. He, et al. The effect of temperature and moisture on the source of N 2 O and contributions from ammonia oxidizers in an agricultural soil, Biol. Fert. Soils 53 (1) (2017) 141–152. [17] M. Biglarbegian, F. Al-Turjman, Path planning for data collectors in precision agriculture WSNs, Advances in biodegradation and bioremediation of industrial waste, Nicosia, Cyprus, (2014), pp. 483–487. [18] What is CIR Imagery and what is it used for, 2016. Available from: https://www.altavian.com/blog/2016/8/cir-imagery. [19] Awesome Farm Drones. Available from: https://www.instructables.com/id/AwesomeFarm-Drones/. [20] Parrots Sequoia. Available from: https://www.parrot.com/business-solutions-us/parrot-professional/parrot-sequoia. [21] Phantom 3 adv. Available from: https://www.dji.com/phantom-3-adv. [22] A. Navarro, J. Catalao, J. Calvao, Assessing the use of sentinel-2 time series data for monitoring cork oak decline in Portugal, Remote Sensing 11 (21) (2019) 2515. [23] Plot 30-1. (2013). Available from https://optoknowledge.com/∼ag2020/2003/30-1. html. [24] M. Carter, Saving money with precision agriculture. (2018). Available from: https:// ag.purdue.edu/extension/WIA/Documents/5C_Saving_Money_with_Precision_Agriculture[1].pdf.

122

Fadi Al-Turjman, Hamit Altiparmak

[25] S. Sankaran, J. Maja, S. Buchanon, R. Ehsani, Huanglongbing (citrus greening) detection using visible, near infrared and thermal imaging techniques, Sensors 13 (2) (2013) 2117–2130. [26] F. Garcia-Ruiz, S. Sankaran, J.M. Maja,W.S. Lee, J. Rasmussen, R. Ehsani, Comparison of two aerial imaging platforms for identification of Huanglongbing-infected citrus trees, Comput. Electr. Agri. 91 (2013) 106–115. [27] M.S. Farooq, S. Riaz, A. Abid, K. Abid, M.A. Naeem, A survey on the role of IoT in agriculture for the implementation of smart farming, IEEE Access 7 (2019) 156237– 156271.

CHAPTER EIGHT

Coastline change determination using UAV technology: a case study along the Konyaaltı coast, Antalya, Turkey Rifat Tura, Levent Uzunsakala, Ali Danandeh Mehrb Civil Engineering Department, Akdeniz University, Antalya, Turkey Civil Engineering Department, Antalya Bilim University, Antalya, Turkey

a

b

Chapter outline 1 Introduction 2 Study area 3 Methodology 3.1 Flight planning with UAV 3.2 Cameras and systems used in UAVs 3.3 Distortion and calibration 3.4 Distortion error correction 3.5 Calibration of the camera used in the present study 4 Results and discussion 4.1 West of the study area 4.2 East of the study area 5 Summary and conclusions References

123 125 127 129 131 132 133 135 136 137 138 139 140

1  Introduction The coastal areas are subject to serious changes under the influence of natural events and man-made activities. The causes and origins of the events/activities affect the extent and nature of the changes. In particular, any morph in the shoreline as a result of man-made activities often has negative effects in and around the coastal area. It is of great importance to measure the change occurred under the natural or artificial activities quickly and makes necessary decisions and onset action plans. Undoubtedly, the more accurate the measurement, the better the solution. In coastal engineering, terrestrial acquisition methods using traditional or modern geodesic measurement tools such as geographical positioning system (GPS), satellite image analysis, laser scanning methods, and experimental model studies are typically used to study coastal morphology and measure Drones in Smart-Cities. http://dx.doi.org/10.1016/B978-0-12-819972-5.00008-2 Copyright © 2020 Elsevier Inc. All rights reserved.

123

124

Rifat Tur, Levent Uzunsakal, Ali Danandeh Mehr

the potential or existing shoreline changes. Among these methods, the terrestrial ones are more time-consuming and costly. GPS provides a much faster measurements than the traditional tools, yet the terrestrial acquisition methods remain very slow compared to the nonterrestrial methods. When the satellite images are used, both the desired resolution and the difficulty of reaching the specific time periods are the challenging tasks. Besides, they are insufficient to evaluate the changes in short-time periods. For examples of the use of satellite images in coastal engineering the interested reader is referred to Refs. [1–6]. On the other hand, although the use of laser scanning method for point cloud formation or light detection and ranging (LiDAR) helps to overcome the earlier-mentioned shortcomings, they are preferred to be applied in a much wider area due to being very expensive in today’s technology. All these issues have led to the need to try unmanned aerial vehicles (UAVs) in recent coastal engineering studies. The authors review revealed that the UAV technology provides desired precision amounts/ resolution in the most of the river and coastal engineering works (e.g., [7–9], [10–12]). While UAVs were very expensive in previous years, they are easily available and frequently applied for detecting coastline changes in recent studies. For example, the UAV technology was used by Niculescu et al. [13] in order to investigate the weakness of natural habitats and changes caused by the human and natural factors in the touristic coastal zones of southern Romania. The study also showed promising application of 3D point clouds for database creation. Puijenbroek et al. applied UAV to examine the changes in the size of the dunes on the beach on the Coast of Texel Island in the Netherlands [14]. Similarly, the UAV Papakonstantinou et al. to 3D coastal mapping sand dune and sea grass classifications on the shores of the Lesbos Island after the heavy storm that occurred in 2016 [15]. Mahabot et al. compared the coastal changes in the La Reunion Island considering three methods of differential GPS (DGPS), terrestrial laser scanning and UAV technology and reported that UAVs producing structure-from-motion data with a sensitivity of 10 cm gives more accurate results [16]. Returning to earlier studies in the literature, Guariglia et al. reported that orthophoto map gives more accurate results compared to satellite images [17]. Şeker et al. demonstrated that the profile changes on the shoreline can be detected by use of point clouds obtained with the laser scanner [18]. Avdan et al. and Yılmaz et al. compared the accuracy of the data produced by UAV and GPS [19,20].

Coastline change determination using UAV technology

125

This chapter investigated the change in the Mediterranean shoreline due to sediment movement along the coastline of the Konyaaltı coast by using UAV technology. In addition, seasonal changes in coastal profile were determined by using both UAV and DGPS and the results were compared with each other. The study predicts man-made morphs in the coastal area that yields in the erosion/sedimentation problems. Thus the results would be beneficial for the wise planning and operation of the Konyaaltı coast, which has of great importance in truism industry of Antalya, Turkey.

2  Study area The Konyaaltı Beach (Fig. 8.1), one of the most attractive beaches in Antalya, Turkey, has a length of approximately 7.5 km, which is limited to the Varyant in the east and Antalya harbor in the west. The majority of

Figure 8.1  Location of the cost of Konyaaltı and the Bog˘ açay River inlet in Antalya, Turkey.

126

Rifat Tur, Levent Uzunsakal, Ali Danandeh Mehr

the coastline material consists of sand and gravel. The Bog˘ açay River (see Fig. 8.1), a perennial river with the approximate length of 25 km, is the most important river in the study area that transports significant amount of sediment to this beach every year. The beach has taken its current shape with alluvial soil and pebbles conveyed by the Bog˘ açay River. Approximately 4.5 km of this pebbly coastline with 70 meters wide is a blue flag public beach. With its eye-catching environmental features, the Konyaaltı beach is one of the most attractive destinations in the worldwide tourism industry and so, very important and valuable for the city, dwellers and its decision makers. The average width of the beach is generally about 30–40 m; however, it decreases down to 10 m in some places. The Bog˘ açay mouth, where the erosion causes the coastline to fall to 10–20 m is investigated in this study. The Bog˘ açay River basin with the irrigation area of 822.23 km2 located in the west of Antalya (Fig. 8.2). It is surrounded by mountains in the north and west, rocky terrain in the east, and opens to Mediterranean Sea in the south. As illustrated in Fig. 8.2, the main tributaries includes Karaman, Doyran, and Çandır streams. The floodplain is mostly covered by vegetation; however, a large part of it exposed to rapid urbanization and therefore, increasing runoff in the catchment is expected. The main channel starts 10 km away from northwest of Antalya and crosses the city in the Konyaaltı district.Table 8.1 presents the flood flow rates (Q) at different return periods (T) reported by Tür and Og˘ uz [21]. A flood event with Q = 1900 m3/s was measured on December 23, 2003. Huge amount of tangible and intangible

Figure 8.2  The Bog˘açay River catchment.

127

Coastline change determination using UAV technology

Table 8.1  Expected flood discharge in the Bog˘açay River.

T (year) Q (m3/s)

2 660

5 970

10 1162

25 1399

50 1571

100 1736

500 2133

Figure 8.3  The Bog˘ açay River mouth and partly destroyed bridge at the flood event occurred in 2003.

damage was occurred along the Bog˘ açay River floodplain and the concrete bridge nearby Mediterranean Sea was partly destroyed (Fig. 8.3). To determine the maximum coastal erosion, the Bog˘ açay inlet where the river discharges to Mediterranean Sea is considered as the center of our study area and approximately 1 km to the east and 1 km to the west, a total of 2 km coastline, have been selected as the whole study area. This is the region that exposed to the most hydrodynamic effects and shoreline changes in comparison to the other parts of the Konyaaltı coast. For this reason, determining the extent of erosion and sedimentation in the chosen area is of great importance for the planning and operation of the beach as well as sustainable tourism industry and economy of Antalya.

3  Methodology Temporary and fixed benchmarks at the study area marked on the ground after the navigation through Google Earth. The geodesic and photogrammetric measurements on the shore were carried out on the

128

Rifat Tur, Levent Uzunsakal, Ali Danandeh Mehr

north-south line up to the Konyaaltı coast road. In addition, accuracy analysis between the methods was performed. Geodesic data obtained from the ground surveys were analyzed with respect to the benchmarks. Flight planning was conducted and UAV technology was used to obtain the data required for the fieldwork. By means of field works, averages of the data obtained from drone within the 50 cm diameter of the same profile points obtained by GPS and the shoreline profile data were found and compared with each other. The difference between the two methods was calculated to be 2%–3% for X and Y coordinates and 7%–8% for Z coordinates. After it was determined that the coordinate values were close to each other and the point cloud data was verified, the UAV measurements were continued. Using UAV, point clouds of different times were collected. Threedimensional terrain model and orthophoto map were obtained by using these point clouds. The obtained map together with the Konyaaltı orthophoto image, which is officially available since 2006 (see Fig. 8.4), were compared by means of different software including Agisoft Photoscan and DJI Ground Station. Consequently, we were able to recognize the change in the shoreline. In the production of orthophoto images, the image coordinate system and the ground coordinate system must be correlated and the error caused by the differences in height on ground must be corrected (rectification of the image). Before any aerial photogrammetry, one must select some benchmarks on the ground that are known as ground control points (GCPs). The GCPs

Figure 8.4  Orthophoto image of the Konyaaltı coast.

Coastline change determination using UAV technology

129

Figure 8.5  GCP and Geodetic GCP measurement (CORS GPS with altitude parameter western mediterranean geoid correction).

are built on the ground so that they would be recognizable in the taken aerial photogs.The associated coordinates can be achieved through the filed measurement methods. To this end, we initially divided our study area into several blocks and then, GCPs were built with predefined distances (see Fig. 8.5). The surveying results is integrated with an appropriate scale map of the project area so that the distances between GCPs are calculated based on the photographic scale. For this purpose, the GCP’s of the images are used. The wise selection of the points, their frequency and distribution, and the accuracy in field measurements are important as they can affect the results of rectification. The use of temporary control point may also be useful to increase accuracy of measurements. In this study, 10 fixed (permanently placed on the ground) plus some temporary control points were adopted as exhibited in Fig. 8.6. The relevant coordinates for the fixed ones were tabulated in Table 8.2.

3.1  Flight planning with UAV In the first stage of flight planning with UAV, the DJI Ground Station PC program was installed to be used in the devices produced by the same company. The Ground Station is a software package for 3D modeling of photogrammetry data applicable in both personal computers and tablets.To use the program, DJI’s Phantom 2, Phantom 3, Phantom 4, Inspire 1, Inspire 2, and

130

Rifat Tur, Levent Uzunsakal, Ali Danandeh Mehr

Figure 8.6  Location of the fixed GCP and temporary control points in the study area.

Mavic Pro devices would be enough. Since the Phantom 2 and Phantom 4 devices were used during the present study, these programs were considered to make flight planning. As the study area is separated to the east and west of the Bog˘ açay River inlet, two separate plans to complete the flights on the same day were designed. Fig. 8.7 shows Ground Station flight planning. Regarding the implemented coordinate data, the longitudes are used in 3- or 6-degree coordinate system. In the routine similar surveying studies, 3-degree system is commonly used. To convert the geographic coordinate system from WGS84 into ITRF96, the number of the current zone covering the study area must be known. The Y coordinates are repeated in each zone. If the Y coordinate other than 400,000-600,000 appears when the measurement is taken, the selected zone number is incorrect. It is not difficult to find the correct zone number (the middle Meridian of the zone). It is necessary to look at the degree part of the longitude value at the current location. In Turkey, 3 degree slice middle meridians are 27-30-33-36 Table 8.2  Coordinate information for fixed GPC points. GCP

Y

X

Z

Latitude (degree)

Longitude (degree)

S1 S2 S3 S4 S5 S6 S7 S8 S9 S10

4,079,756.850 4,079,898.780 4,079,993.330 4,080,178.060 4,080,281.170 4,080,403.520 4,080,606.410 4,080,764.210 4,080,973.770 4,081,163.470

555,232.610 555,329.210 555,442.730 555,622.560 555,733.610 555,862.910 556,078.520 556,236.570 556,453.520 556,680.880

5.258 5.551 5.311 5.418 5.877 5.765 4.071 3.900 4.092 3.988

36.847417 36.848690 30.620364 36.851189 36.852112 36.853206 36.855022 36.856434 36.858310 36.860005

30.619271 30.620364 30.621644 30.623674 30.624926 30.626385 30.628818 30.630602 30.633050 30.635613

Coastline change determination using UAV technology

131

Figure 8.7  The Konyaaltı coast Ground Station pro tablet flight planning for flight with DJI Phantom 4 at east (left) and west (right) of the Bog˘ açay inlet.

and 39. The zone number, which is closer to the zone number, is selected as the study area middle meridian. Longitude in the Bog˘ açay River inlet is approximately 30.625928 for the middle point. In this case, the zone number will be selected as 30, which is why the TM30 is taken. EPSG represents the Geodetic Parameter Data Set. The EPSG code for Antalya is 5254.

3.2  Cameras and systems used in UAVs Multicopters are able to transmit images through the systems connected to the motherboard and carry out the commands sent from the ground to the device. Image transmitter, camera, GPS, datalink, and some other equipment are the systems used by the device.When the cameras are examined, it is seen that there are camera systems integrated with the device in today’s technology. When looking at the older generation devices, it should be mentioned that there are camera systems that operate separately from the device. In the present study, a DJI Phantom 2 V2 model multicopter was used. The implemented camera was GoPro Hero 4+ which noun as fisheye with very wide-angle sports lens (Fig. 8.8).

Figure 8.8  The UAV multicopter (DJI Phantom 2 V2) and DJI Phantom 4 Pro used in the present study.

132

Rifat Tur, Levent Uzunsakal, Ali Danandeh Mehr

Figure 8.9  DJI Phantom 2 V2 produced with fisheye distortion error global model (Agisoft PhotoScan).

The camera is intentionally manufactured with distortion. Looking at sensor features of the camera, 12 MP 4000×3000 (4:3), it has a shooting angle of 170 degree. The first point cloud was created by calibrating the distortion errors in order to avoid the fisheye effect while creating the orthophoto map and model. Otherwise, the model will contain many errors (Fig. 8.9). The program also utilizes Brown’s checkers model to correct distortion errors.

3.3  Distortion and calibration In geometric optics, distortion is a deviation from rectilinear projection; a projection in which straight lines in a scene remain straight in an image. There are two main kinds of optical distortion: barrel distortion (see Fig. 8.10 left) and pincushion distortion (see Fig. 8.10 right). In fact, it is the refraction of the ray inclined on the optical axis to surfaces with different refractive indexes. The ray coming from a point P will make the angle τ with the camera axis, and the camera will emerge by making the angle τ′ in the photographic space due to defects in the lens system. As a result as illustrated in Fig. 8.11, the projection of point P would be at the point (P′) rather than

Figure 8.10  Barrel distortion (left) and pincushion distortion (right).

Coastline change determination using UAV technology

133

Figure 8.11  Distortion error.

the point P′. The error, equals to ∆r, is the distortion of the camera that affects the geometry of photos [22]. Distortion causes the change of the position of the subject on a photo. Therefore, the presence of any geometric distortion is very important in photogrammetry. If any metric measurement on a photo is needed, potential distortion should be taken into account and eliminated by geometric calibration of the camera. Each camera has its own calibration values determined by measurements in laboratory conditions. These values must be considered in order to obtain the accurate results from the relevant software. Some software uses the camera information encoded in the pictures taken, recognizes the cameras used and uses the photos that are automatically calibrated in the final stage as they contain the distortion values of these cameras in their data. But it should be noted that camera distortion is a fact and even if a camera has never been used, it may cause distortion in its first application.Therefore, it is necessary that the distortion values of the camera to be addressed before each use. In order to minimize distortion in cameras, all precautions are taken during the manufacturing process. However, this error cannot be avoided completely. Camera distortion is detected via high sensitive measurements in laboratories, and reported to the users. Distortion error may be 20–25 microns in older aerial cameras, but it remains below 10 microns in recent ones. Nowadays, this error is below 5 microns.

3.4  Distortion error correction Even if very carefully designed during the production, tangential lens distortion cannot be avoided. However, their effects can be reduced to very

134

Rifat Tur, Levent Uzunsakal, Ali Danandeh Mehr

low amounts. The diametric distortion value is the diametric distance between the distorted position and the distortion-free position of the point. Diameter of lens distortion is determined from its calibration reports. For this, the polynomial curve is drawn according to the amount of change corresponding to the diametric distances. The polynomial diametric lens distortion is given by the following equation. (8.1) ∆r = k1r1 + k2r3 + k3r5 where ∆r is radial distortion amount, r is radial distance to prime and k1, k2, and k3 are polynomial coefficients. These coefficients are determined according to the least squares method using the distortion values in the calibration report.As illustrated in Fig. 8.12, the distance r from the image point to the prime point is calculated to correct the x and y position of the image point and used to calculate ∆r in the equation [23]. (8.2) r = ( x − x 0 ) + ( y − y0 ) , 2

2

∆r ∆x ∆y . = = r x − x 0 y − y0 (8.3)

Figure 8.12  Correction of distortion error.

Coastline change determination using UAV technology

135

And thus, ∆r (8.4) ∆x = ( x − x 0 ) , r ∆r (8.5) ∆y = ( y − y 0 ) . r Corrected coordinates can be extracted using Eqs. (8.6) and (8.7). ∆r (8.6) x = ( x − x 0 ) 1 −  , r   ∆r (8.7) y = ( y − y0 ) 1 −  . r  

3.5  Calibration of the camera used in the present study As previously mentioned Agisoft Photoscan (Demo Version) is the package program used for modeling, orthophoto map and point clouds during the study.When the program is installed on the computer another package with the name of Agisoft Lens distortion errors is installed that can to produce the necessary parameters to eliminate distortion errors. This program provides the opportunity to calculate and correct the lens errors in the camera. For this, the program uses Brown’s checkers model (Fig. 8.13). Table 8.3 presets the calibration report for the GoPro Hero4+ camera via Agisoft PhotoScan. We carried out the first stage with the Phantom 2 V2, as the GoPro Hero 4 Black Edition+ camera is suitable for it. Despite working with such lenses, distortion values are needed to be known or obtained to achieve accurate results. To increase the accuracy in our study, the available distortion

Figure 8.13  Brown’s checker board model (calibration pattern).

136

Rifat Tur, Levent Uzunsakal, Ali Danandeh Mehr

Table 8.3  Calibration report for the GoPro Hero4+ camera. Parameter Value Standard Error

Image width Image height Focal length Principal point (x) Principal point (y) Affinity B1 Skew B2 Radial K1 Radial K2 Radial K3 Radial K4 Tangential P1 Tangential P2

3000 2250 1750.78 15.2055 −32.0247 13.5354 −0.00510936 −0.268438 0.119624 −0.0428787 0.00736563 2.34437E-05 9.95205E-05

1.06617 0.451601 0.406621 0.0854521 0.0806343 0.00100331 0.00212492 0.00179012 2.61187E-07 0.000025129 3.21872E-05

values of GoPro Hero 4 were not used. Instead, a calibration test was performed and the obtained values were employed. These parameter values were entered into Agisoft Photoscan program during the model and orthophoto map creation phase. The GoPro lens, which is otherwise close to the fisheye lens, is spherically formed with a 170 degree aperture. The lenses in the cameras used for calibration create new distortions depending on the time and the type of usage (Fig. 8.14). Hence, at certain time intervals, each camera must be recalibrated and the newly formed parameters entered into the programs used for photogrammetry.

4  Results and discussion As previously mentioned, the available 2006 orthophoto map of Antalya region and that of created by the authors in the same coordinate system were used to determine and compare the coastline change between

Figure 8.14  Calibration report of radial (left) and tangential (right) distortion [24].

Coastline change determination using UAV technology

137

the period 2006 and 2017. Accordingly, the regions with coastal erosion and deposition have been identified. The results are of great importance to determine the possible impacts of shoreline change on the harbor and newly constructed recreational area along the coastline. The findings obtained from the study area are presented in two following subsections describing deposition and erosion zones in the west and eastern part of the Bog˘ açay inlet.

4.1  West of the study area Between 2006 and 2017, it was recognized that there were some changes in the coastline in the western part of the Bog˘ açay River. On the orthophoto map of 2006, the existing coastal pavement, banquets and enterprises are shown with a green line and compared with the orthophoto map created in 2017. As shown in Fig. 8.15, in the orthophoto map of 2017, there is an erosion in the coastline at the 305 m section of the coastal area west of the Bog˘ açay during the 12-year time period since 2006. When the entire line was considered, it was found that erosion was generally observed between Antalya Port and the Bog˘ açay inlet. The amount of erosion determined at some predefined locations (i.e., points D to F) in the western region of the Bog˘ açay is given in Table 8.4. Similarly, erosion amounts in selected coordinates are shown in Fig. 8.15.

Figure 8.15  Loss of coast at point D (equal to 11.32 m), E (equal to 10.66 m) and F (equal to 11.43 m).

138

Rifat Tur, Levent Uzunsakal, Ali Danandeh Mehr

Table 8.4  Erosion of certain points at the west of the Bog˘açay. Point Name Y X Latitude Longitude

D E F

555,676.218 4,080,168.060 555,609.140 4,080,111.982 555,517.012 4,080,023.603

Erosion (m)

36.851096 30.624275 11.32 36.850594 30.623519 10.66 36.849804 30.622479 11.43

4.2  East of the study area Comparing to the orthophoto map of 2006, the orthophoto map of 2017 indicates sediment accumulation (deposition) on the entire study area at eastern side of the Bog˘ açay inlet. The amount of deposition determined at some predefined locations (i.e., points A to C) in the region is tabulated in Table 8.5. Similarly, erosion amounts in the given points are marked on the orthophoto maps (Fig. 8.16).

Table 8.5  Deposition in the certain points at east of the Bog˘açay River. Point Name Y X Latitude Longitude

A B C

555,962.531 4,080,430.074 36.853440 555,979.114 4,080,450.702 36.853625 556,080.679 4,080,568.486 36.854680

Deposit (m)

30.627504 9.44 30.627692 9.04 30.628839 7.91

Figure 8.16  Creation of coast (deposition) at point A (equal to 9.44 m), B (equal to 9.04 m) and C (equal to 7.91 m).

Coastline change determination using UAV technology

139

5  Summary and conclusions In recent years, coastal erosion has been experienced frequently in coastal cities as a result of both human activities and the natural impact of climate change. The long-term and periodic identification of potential changes in the coastline, not only creates knowledge about the dynamics of sediment along costal area, but also provides the evidences to make wisdom decisions in case of any problems for the coast environment and available infrastructures. Identifying such changes by classical geodesic methods has some disadvantages such as being expensive and time-consuming. For this purpose, the use of faster and more economical photogrammetric methods has made great progress especially in the coastal engineering literature. In this study, first, the differences between the measurements of classical geodesic and photogrammetric methods have been determined. Then, the calibration was done, and it was shown that UAV could determine the coastline change with 2–3 cm accuracy. Especially important in terms of tourism, the dynamics of the Konyaaltı beach was investigated using UAV technology. To this end, Orthophoto maps produced for the Antalya region in 2006 were compared with that of produced in this study in 2017. The orthophoto map, which was formed along a line of approximately 2 km, provided a suitable tool to determine the erosion and deposition zones appeared within the period of 2006–17. The results showed that the sediment accumulation occurred in the east coast of the Bog˘ açay inlet and the continues erosion was in the west of the Bog˘ açay. In 1995, the Local Environmental Board issued a decision not to extend the licenses following the expiry of the licensed gravel mining activities in the Bog˘ açay, and since then, new requests and license extension requests have not been taken into consideration. According to Tür et al., the gravel mining activities of all enterprises were terminated in early 2000 [21]. Our results prove that with the termination of this activities, the coastal loss was decreased, the continuity of the sediment forming the beach was ensured, and the shore was fed again. In the eastern part of the study area, coastal erosion was observed. Antalya Harbor, which is located at the beginning of the coastline, prevents the sediment transport by coastal currents to the east of the harbor due to the disruption of the coastline flows. This is the reason behind the rapid coastal erosion that was observed in the coastal line between the Bog˘ açay and Antalya Port. All in all, our results suggest long-term and continuous monitoring of the changes in coastal areas. It is of great importance for the further

140

Rifat Tur, Levent Uzunsakal, Ali Danandeh Mehr

development of the Konyaaltı beach which is the focal point of tourism in Antalya.

References [1] X. Zhang, Y. Zhang, Y. Ji, Y. Zhang, Z. Yang, Spatial and temporal shoreline changes of the southern Yellow River (Huanghe) Delta in 1976–2016, Marine Geology 395 (2018) 188–197. [2] M. Jayanthi, S.Thirumurthy, M. Samynathan, M. Duraisamy, M. Muralidhar, J. Ashokkumar, et al., Shoreline change and potential sea level rise impacts in a climate hazardous location in southeast coast of India, Environmental, Monitoring and Assessment 190 (1) (2018) 51. [3] Q.X. Liu, J. Trinder, I.L. Turner, Automatic super-resolution shoreline change monitoring using Landsat archival data: a case study at Narrabeen-Collaroy Beach, Journal of Applied Remote Sensing 11 (1) (2017) 016036. [4] N.Q. Tuan, H.C. Tin, L.Q. Doc, T.A. Tuan, Historical monitoring of shoreline changes in the Cua Dai Estuary, Central Vietnam using multi-temporal remote sensing data, Geosciences 7 (3) (2017) 72. [5] F.Temiz, S. Durduran, Monitoring coastline change using remote sensing and GIS technology: a case study of Acigol Lake Turkey, in: IOP Conference Series: Earth and Environmental Science, vol. 44, 2016. [6] M. Aydın, M. Uysal, Risk assessment of coastal erosion of Karasu coast in Black, Sea J. Coast. Conservat. 18 (2014) 673–682. [7] S. Harwin, A. Lucieer, Assessing the accuracy of georeferenced point clouds produced via multi-view stereopsis from unmanned aerial vehicle (UAV), Imagery Remote Sensing 4 (2012) 1573–1599. [8] P. Bellezza, F. Grimaccia, Masini, Airborne unmanned monitoring system for coastal erosion, Assess. Eng. Geol. Soc. Territ. 4 (2014) 115–120. [9] E. Casellaab, A. Roverede, A. Pedroncinig, L. Mucerinoa, M. Casellaf, L.A. Cusatig, et al. Study of wave runup using numerical models and low-altitude aerial photogrammetry, Estuar. Coast. Shelf Sci. 149 (2014) 160–167. [10] D. Niculescu, E. Vlasceanu, I. Omer, R. Mateescun, The monitoring of the romanian littoral cliffs using automatic flight systems, in: International Multidisciplinary Scientific GeoConference - SGEM, 2015, pp. 1187–1194. [11] C. Kyoungah, L. Impyeong, Accuracy analysis of coastal area modeling through UAV photogrammetry Korean, J. Remote Sens. 32 (2016) 657–672. [12] C. Yoo, T.S. Oh, Beach volume change using UAV photogrammetry Songjung Beach, Korea, Int. Arch. Photogram. Remote Sens. Spat. Info. Sci. 41 (2016) 1201–1205. [13] D. Niculescu, E.Vlasceanu, Unnamed aeral vehicle (UAV) technology in monitoring of coastal cliffs, J. Environ. Prot. Ecol. (2017) 1202–1212. [14] M. Puijenbroek, C. Nolet, A. Groot, J. Suomalainen, M. Riksen, F. Berendse, et al. Exploring the contributions of vegetation and dune size to early dune development using unmanned aerial vehicle (UAV) imaging, Biogeosciences 14 (2017) 5533–5549. [15] A. Papakonstantinou, K. Topouzelis, G. Pavlogeorgatos, Coastline zones identification and 3D coastal mapping using UAV spatial data, ISPRS Int. J. Geo-Inf. 5 (2016) 75. [16] M.-M. Mahabot, M. Jaud, G. Pennober, The basics for a permanent observatory of shoreline evolution in tropical environments; lessons from back-reef beaches in La Reunion Island Comptes Rendus, Compt. Rendus Geosci. 349 (2017) 330–340. [17] V.D. Guariglia, A multisource approach for coastline mapping and identification of shoreline changes, Annals of Geophysics 49 (1) (2006) 295–304.

Coastline change determination using UAV technology

141

[18] D.Z. Şeker, A. Körog˘lu, M.S. Kabdas¸lı, H. Karabörk, A. Göktepe, E. Varol, Yersel Lazer Tarayıcıların Kıyılardaki Profil Deg˘ is¸imlerinin İzlenmesinde Kullanımı, 7. Kıyı Sempozyumu, Trabzon. 2011, pp. 537–543 [19] U. Avdan, E. Şenkal, R. Çömert, S. Tuncer, İnsansız Hava Aracı ile Olus¸turulan Verilerin Dog˘ ruluk Analizi, V. Uzaktan Algılama ve Cog˘rafi Bilgi Sistemleri Sempozyumu UZAL-CBS 2014, İstanbul, 2014. [20] V. Yılmaz, A. Akar, Ö. Akar, O. Güngör, F. Karslı, E. Gökalp, İnsansız Hava Aracı ile Üretilen Ortofoto Haritalarda Dog˘ ruluk Analizi, Türkiye Ulusal Fotogrametri ve Uzaktan Algılama Birlig˘ i VII. Teknik Sempozyumu (TUFUAB’2013), 2013. [21] Tür R., Og˘ uz C., Bog˘ açayı Havzasında İllegal Malzeme Alımı Ve Çevreye Olumsuz Etkileri, Antalya Yöresinin İns¸aat Mühendislig˘ i Sorunları Kongresi, Antalya,Türkiye, 2005. [22] J.Weng, P. Cohen, M. Herniou, Camera calibration with distortion models and accuracy evaluation, IEEE Trans. Patt. Anal. Mach. Intel. 14 (10) (1992). [23] A. Yas¸ayan, M. Uysal, A. Varlık, U. ve Avdan, Fotogrametri, Anadolu Üniversitesi Açıkög˘ retim Yayını, No: 1292, Eskis¸ehir, 2011. [24] L. Uzunsaka, Kıyı Profilinin İnsansız Hava Araçları Yardımı ile Belirlenmesi, Yüksek Lisans Tezi (MSc Thesis), Akdeniz Üniversitesi Fen Bilimleri Enstitüsü İns¸aat Mühendislig˘ i Ana Bilim Dalı,Haziran, 2018.

CHAPTER NINE

Explorative analysis of AUV-aided cluster-based routing protocols for Internet of intelligent underwater sensors Abdulazeez Femi Salamia, Emmanuel Adewale Adedokunb, Fadi Al-Turjmanc, Habeeb Bello-Salaub, Bashir Olaniyi Sadiqb, Eustace M. Dogod

Department of Computer Engineering, University of Ilorin, Ilorin, Nigeria Department of Computer Engineering, Ahmadu Bello University, Zaria, Nigeria Department of Artificial Intelligence Engineering, Research Centre for AI and IoT, Near East University, Nicosia, Mersin, Turkey d Institute for Intelligent Systems, University of Johannesburg, Johannesburg, South Africa a

b c

Chapter outline 1 Introduction 2 Internet of intelligent underwater sensor networks (IoIUSN) application: AUV-aided WQM 2.1 Conceptual definition, framework and features of IoIUSN 2.2 Conceptual framework and features 2.3 Characteristics and functional requirements of AUV-aided WQM 2.4 Implementation challenges 3 CBR protocols for UWSN 3.1 Technical definitions, features, and architecture 3.2 Expository analysis 3.3 Network performance and statistical analysis 4 Results and discussion 5 Conclusion References

143 145 145 147 147 148 148 148 154 154 175 182 183

1  Introduction Underwater sensor networks (UWSNs) is a booming and versatile technology with a plethora of practical and interesting applications [1–3]. In UWSN, a number of submerged sensors are systematically deployed and distributed in an aquatic field of interest [4–5]. In some strategic applications, these smart underwater sensors are specially attached to and operate Drones in Smart-Cities. http://dx.doi.org/10.1016/B978-0-12-819972-5.00009-4 Copyright © 2020 Elsevier Inc. All rights reserved.

143

144

Abdulazeez Femi Salami, Emmanuel Adewale Adedokun, Fadi Al-Turjman, Habeeb Bello-Salau, Bashir Olaniyi Sadiq, Eustace M. Dogo

with different types of unmanned underwater vehicles (UUVs) such as autonomous underwater vehicles (AUVs), remotely operated underwater vehicles (ROUVs), and autonomous underwater gliders (AUGs) [6–7]. The aquatic field of interest can correspond to a single depth layer, region (range of depth layers with similar aquatic features), or multiple regions (different ranges of depth layers with unique underwater properties) [8–9]. Therefore, the purpose of these smart underwater sensors is to collaboratively gather vital data from the aquatic field of interest and relay the aggregated data to one or more surface sinks from where the aggregated data are transferred to a remote coastal base station for intelligent data processing and analytics in order to extract useful (predictive or decision-making) information [10–11]. In addition to this, it must be mentioned that the designed UWSN should be self-organizing, scalable, adaptable, reliable, resilient, and resource-efficient [12–13]. This will ensure swift and seamless exchange of location, configuration and mobility information within the network without interfering with other on-going shipping or monitoring activities [14–15]. Some of the unique challenges facing the design and performance of UWSN are the high cost and nondisposability of underwater instruments coupled with the high packaging and deployment costs [1,9]. Furthermore, UWSN deployments are comparatively less dense, planned for shorter periods, and aimed at longer range in harsher environments than terrestrial wireless sensor networks (TWSNs) [3,9]. Since UWSN deployments are designed for longer range, acoustic mode of communication is usually preferred because optical, radio, and electrostatic communication signals fade or weaken rapidly within a short range of approximately 1–10 m [5,9,11]. However, acoustic communication in aquatic environment is severely affected by node failures from biofouling and corrosion, Doppler spread, multipath and fading effects, surface noise, long propagation delays, salinity, narrow bandwidth, shadow zones, high temperature gradients, unstable topology, and erratic node mobility as a result of water currents [1,8,9]. These challenges are cogent factors contributing to high-energy consumption and network performance deterioration [3,5,8,9]. These challenges have also instigated UWSN researchers to proffer solutions in the form of unique routing techniques, which are custom-made for the underwater environment as existing routing schemes for TWSN cannot be directly employed in UWSN due to wide environmental and operational differences [4,6,14].

Explorative analysis of AUV-aided cluster-based routing protocols

145

Cluster-based routing (CBR) is one of these proposed solutions where the network adopts a dynamic hierarchical process of logically grouping the nodes into cluster heads (CHs) and cluster members (CMs) with respect to well-defined performance indicators [14,16,17]. Researchers have also established that CBR protocols are relatively more versatile and capable of yielding better performance in terms of fault tolerance, resource awareness, and route efficiency for large-scale UWSNs [5,6,18,19]. This chapter, therefore, investigates the peculiarities and analyzes the performance of several proposed CBR protocols for UWSN. Section 9.2 of this work discusses the architecture, network model, and technical features of AUV-aided water quality monitoring (WQM) as a target application for the Internet of intelligent underwater sensors. This research furthermore conducts an explorative analysis of state-of-the-art CBR protocols for UWSNs in Section 9.3. This work conducts simulation-based network and statistical analysis to provide useful technical insights on the performance analysis of selected CBR protocols as technically demonstrated in Section 9.3. Section 9.4 concludes this research work.

2  Internet of intelligent underwater sensor networks (IoIUSN) application: AUV-aided WQM 2.1  Conceptual definition, framework and features of IoIUSN Internet of Intelligent Underwater Sensor Networks (IoIUSNs) is an evolving branch of wireless sensor network (WSN) and Internet of things (IoT) technologies that finds numerous practical applications in underwater monitoring domains as earlier discussed in Section 9.1 and further shown in Fig. 9.1. Applications of IoIUSN include environmental monitoring, underwater resource explorations, civilian and military surveillance, disaster prevention, and many other interesting applications. IoIUSN can be defined, as small- or large-scale deployment of interconnected smart underwater uniquely identifiable objects or ‘things, operating without human intervention, with data capturing, communication and actuation capabilities, using intelligent interfaces and with security considerations [20–22]. One of the challenging tasks in UWSN research for WQM is an effective water parameter data-gathering system. This is due to the unique characteristics associated with water body environment [22]. Hence, a method that is gaining popularity is using AUV-aided UWSN over acoustic signals for an efficient data capturing of large underwater region. This

146

Abdulazeez Femi Salami, Emmanuel Adewale Adedokun, Fadi Al-Turjman, Habeeb Bello-Salau, Bashir Olaniyi Sadiq, Eustace M. Dogo

Figure 9.1  Application Areas of IoUT. (Reproduced with permission from Ref. [20].)

will ensure an enhanced communication sensing range that should cater for low signal attenuation, absorption rate, and an optimal transfer rate [22]. It must be mentioned, however, that acoustic signals experience interoperability difficulties with existing RF-based routing protocols. Fig. 9.2 depicts

Figure 9.2  AUV-aided and Controlled Model for UWSN. (Reproduced with permission from Ref. [23].)

Explorative analysis of AUV-aided cluster-based routing protocols

147

Figure 9.3  Conceptual Framework of IoIUSN for WQM. (From Ref. [24].)

a conceptualized AUV-aided IoIUSN model. This deployment in Fig. 9.2 is based on three broad assumptions, namely, central power resting with the sink, nodes are assumed to be always sending data, and nodes have intelligent self-awareness capabilities (location/position awareness and adjustment of energy levels). However, it is worth noting that these assumptions are dependent on the desired nature of deployment and modeling scenario under investigation.

2.2  Conceptual framework and features Fig. 9.3 depicts the conceptual framework for UWSN-based WQM system. It comprises both static and movable underwater sensors for gathering vital water quality information. From Fig. 9.3, the network section is employed for transmitting data to the sink or base station, which is usually the monitoring center.The data transmitted to the sink are further analyzed and visualized for enhanced monitoring experience and water quality condition monitoring.

2.3  Characteristics and functional requirements of AUV-aided WQM The following are generic functional requirements for underwater sensor nodes deployed for AUV-aided WQM [25]:

Abdulazeez Femi Salami, Emmanuel Adewale Adedokun, Fadi Al-Turjman, Habeeb Bello-Salau, Bashir Olaniyi Sadiq, Eustace M. Dogo

148

Table 9.1  Challenges and operational differences between TWSN and UWSN [20]. S/N Challenges TWSN UWSN

1

Transmission media

2 3

Propagation speed Transmission range

4

Transmission rate

5

Accessibility and recharging difficulty Sensor node mobility Relatively static sensor High (due to dynamic positions network topology changes) Link reliability Relatively stable link Low (Unstable link reliability)

6 7

• • • • •

Radio signal waves

Acoustic or sound signal waves 300,000,000 m/s 1500 m/s Shorter range 10–100 m Longer range and prune to interference Relatively wider band- Narrow bandwidth, width, ∼250 kpbs ∼10 kbps Low High

Information gathering by deployed sensors, Network communication, Sensors environmental adaptive function, Data processing, Information browsing and condition monitoring.

2.4  Implementation challenges The challenges associated with IoIUSN research can be viewed in terms of the difference between TWSN and UWSN as discussed in [20]. The summary of these challenges is outlined in Table 9.1.

3  CBR protocols for UWSN 3.1  Technical definitions, features, and architecture Routing protocols are systematic rules guiding the format, path, and reliable transmission of data from source nodes to the desired destination [6,14]. Routing is an indispensable component of UWSN design where it is essential to adaptively discover and maintain energy-efficient paths with respect to the dynamic underwater environment [4,5,7]. Most routing protocols take these key issues into account, namely, rapid energy consumption, which reduces network lifetime, and erratic node mobility, which degrades network reliability [8,12,13]. Conventional routing protocols proposed for TWSN are not suitable for UWSN routing. The

Explorative analysis of AUV-aided cluster-based routing protocols

149

Table 9.2  Routing protocol considerations for TWSN and UWSN [6]. Factors TWSN UWSN

Deployment scenario Deployment region

Dense Two-dimensional space

Sparse Three-dimensional volume Deployment mode Deterministic and uniRandom and form non-uniform Battery specifications Larger size, replaceable, Smaller size, difficult to rechargeable replace or recharge Energy consumption Low High Network architecture Static sensor nodes Dynamic sensor nodes Nature of network Homogeneous Heterogeneous Network reliability and Nodes less prone to errors Nodes have more problifetime and failures, longer ability of errors/failoperational time ures, shorter operational time Noise effects Low High Received signal strength Strong; lesser interference Weak; more interference Sink location Fixed or preplanned Changing with random movement water currents Mobility pattern Tractable Complex Link stability High Low Communication medium Radio frequency (RF), Acoustic, optical (Li-Fi) Optical, electrostatic Propagation speed High Low Time synchronization and Realizable with high ac- Very difficult due to localization curacy inaccurate synchronization Incorporation of global Fully supported and Partially practicable positioning techniques feasible Number of routing hops Function of network size Function of monitoring (space) region (depth) Data rate High Low Robustness Easily ensured with error Difficult due to narrow detection and recovery bandwidth and high schemes latency

technical factors differentiating routing protocol considerations for both networks are outlined in Table 9.2. The following paragraphs discuss the broad categories of UWSN routing such as cross-layer (CLR), layered (LR), path cost calculation-based (PCCBR), reinforcement learning-based (RLBR), flooding-based (FBR), geo-based (GBR), multipath-based (MPBR), location-based (LBR),

150

Abdulazeez Femi Salami, Emmanuel Adewale Adedokun, Fadi Al-Turjman, Habeeb Bello-Salau, Bashir Olaniyi Sadiq, Eustace M. Dogo

opportunistic (OR), adaptive (AR), source (SR), hop-by-hop (HBHR), spatio-temporal multicast (STMR), and CBR protocols. CLR focuses on sharing interlayer information within the protocol stack in order for network performance to be globally optimized [13,14]. Power control schemes, layer-specific functions, and information and frame collision control are strategies employed by CLR to ensure smooth interlayer interactions, energy consumption minimization, and network performance maximization [4,12]. On the other hand, LR is concerned with minimizing the number of retransmissions from packet collisions by forming a virtual hierarchical structure around the surface sink [6,8]. This structure is based on logically dividing the monitoring field of interest into layers where relay nodes are elected from each layer to monitor and manage network conditions (errors, route failures, packet loss) and preserve residual energy [5,14]. LR relies on CDMA (incorporating orthogonal spreading sequence) for packet collision minimization [4,12]. However, LR still suffers from highenergy consumption from recurrent retransmissions in the event of multiple route failures in the UWSN [7,13]. In PCCBR, the cost of selected path selection metrics is computed to possibly find the optimal route based on two phases, namely, cost establishment and data forwarding [3,4]. The cost establishment stage entails broadcasting of HELLO packets from the sink to other nodes in order to periodically estimate and update relative distance (cost) to the sink [6,9]. The data forwarding utilizes residual energy-based ordering, holding time and least-cost path to prioritize forwarding of data packets from the sending node to the desired destination [7,14]. RLBR is concerned with learning the network conditions in order to compute the optimal path with the aid of the Q-learning algorithm [4,12]. The purpose of this is to intelligently adapt the network to run-time topological variations without the need for offline training. In order to extend network lifetime, the reinforcement function depends on residual energy for each node and group energy distribution to update and enhance the routing process [3,6]. FBR is one of the simplest routing schemes, which demands little knowledge about the underlying network because a sending node simply broadcasts its packet to all nodes within the reachable transmission range [9,13]. Delivery is thereby guaranteed but high-energy consumption due to packet duplication is one of the major issues in adopting FBR [12,14]. LBR was proposed to address the multiple broadcasting challenges of FBR by introducing a virtual routing pipe with predefined radius, which serves as the transmission threshold [7,8]. The basic assumption in LBR is that

Explorative analysis of AUV-aided cluster-based routing protocols

151

nodes know their own location and operate with an INFO packet containing location information of the original sender, neighboring forwarder and desired receiver [4,14]. Upon reception of the INFO packet, nodes with distance costs lesser than the transmission threshold are selected into the forwarder, position field for data forwarding [12,13]. LBR performs poorly for sparse UWSN and the computational overheads for determining and updating the transmission threshold are high [6,8]. In GBR, positions of the submerged sensors are used to determine the best path for forwarding data packets based on a well-defined networking policy [6,14]. It is assumed in GBR that the underwater sensors are aware of (1) the monitoring field of interest, (2) their positions, and (3) positions of all adjacent nodes [4,5]. Global positioning techniques and received signal strength-based coordinate estimation are presently the widely used methods for localization (determining position information) [7,14]. However, as highlighted in Table 9.3, global positioning techniques are partially practicable with poor performance in underwater environment while receiving signal strength-based coordinate estimation is severely attenuated due to high interference [6,12]. In MPBR, the formation of more than one route from a given source to a desired sink is employed to ensure robustness and reliability [13,14]. Establishment of multiple paths addresses critical networking issues such as route failures, transmission errors, and packet drop rates [4,12]. OR technique exploits packet transmission opportunities as nodes move and broadcast ADV packets [6,8]. This helps in quickly computing next hop based on specified routing metrics and determining suitable adjacent nodes that will serve as potential forwarding nodes [7,13]. The potential forwarding nodes are thereafter prioritized, and selected nodes with high priorities are employed for packet transmission [4,12]. OR technique takes advantage of channel transmission features to enhance network reliability and throughput [5,14]. In AR, issues of link and node failures are addressed by switching data packet transmission to another suitable route in order to adjust to prevailing network conditions [3,4]. AR relies on a number of resilient algorithms such as interlaced MPBR model with disjoint main and backup routes, and fault-tolerance (path repair, traffic switching) schemes [4,8]. SR gives a certain degree of permission to the originating sender to specify the desired path for the packet, which is embedded, with all relay node information for the specified path [6,13]. Route construction is done by flooding RREQ packets by the source while the destination replies with RREP packet, which contains route information tracing the path back to

Interlayer information sharing

LR

Logical formation of Static and single sink layered structure scenarios around the sink

PCCBR

Cost of route selection metrics evaluated to find the best path Next hop determined by evaluating reinforcement function Reliance on packet broadcasts to the entire network for path discovery Reliance on GPS

RLBR

FBR

GBR MPBR

Static and single sink scenarios

Static and single sink scenarios

Energy consumpProtocol complexities, computational tion minimization, overhead network performance maximization Connectivity loss reduc- High processing overhead due to tion, improved netnumber of layers, high energy work management consumption due to recurrent retransmissions Network stability, ability Best path routing is not guaranteed, to cope with heavy high computational overhead traffic

Single sink scenarios

Network lifetime maxi- Expensive and high-performance mization nodes required

Static and multiple sinks scenarios

Packet delivery is guaranteed

Limited and partially prac- The proposed network ticable model is simple Formation of alterna- Single and multiple sink Fault tolerance, rotive routes from scenarios bustness, enhanced source to sink bandwidth

High-energy consumption due to packet duplication; increased data redundancy High-frequency bands are impractical and not supported for UWSN High communication and computational overhead due to backup route discovery and maintenance

Abdulazeez Femi Salami, Emmanuel Adewale Adedokun, Fadi Al-Turjman, Habeeb Bello-Salau, Bashir Olaniyi Sadiq, Eustace M. Dogo

CLR

Weakness

152

Table 9.3  Comparative analysis of UWSN routing protocol categories [26]. Routing protocol Features Relevant scenario Strength category

OR

AR

SR HBHR STMR

CBR

Relies on virtual routing pipe with pre-defined radius Opportunistic selection of adjacent nodes as potential forwarding nodes Adjust to network conditions by switching packet transmission to suitable routes Route specified by the originating source Next hop specified by a relay node Relies on AUV to gather data from underwater sensors for real-time delivery Nodes are logically grouped into clusters

Static sink scenarios Multiple sinks mobile scenarios

Improved throughput Poor network performance for sparse performance, network UWSN, high latency and compuscalability tational overhead Enhanced network reli- Expensive and high-performance ability and throughput nodes required

Single and multiple sink scenarios

Resilient and fault-toler- High end-to-end delay, high commuant, adaptive nication and processing overhead

Static and single sink scenarios

Route maintenance cost Increased overall routing cost and reduction packet overhead

Both single and multiple sink scenarios AUV or mobile sink scenarios

Network flexibility and Suboptimal final route scalability Location-awareness, Expensive for real-time monitoring resource-awareness applications, high communication and processing overhead

Versatile

Adaptive, resource awareness, route efficiency, reduced redundancy

Explorative analysis of AUV-aided cluster-based routing protocols

LBR

Cluster management overhead

153

154

Abdulazeez Femi Salami, Emmanuel Adewale Adedokun, Fadi Al-Turjman, Habeeb Bello-Salau, Bashir Olaniyi Sadiq, Eustace M. Dogo

the source [4,14]. SR reduces route maintenance cost but increases overall routing cost and packet overhead [13,14]. This leads to rapid network performance degradation for mobile UWSN [4,12]. On the other hand, HBHR permits each relay node to choose its next hop with respect to the node's local information of the network [6,8]. The benefit of this approach is increased flexibility and scalability but the drawback is that the final established route is suboptimal [5,14]. STMR was proposed for UWSN applications requiring node mobility where AUV is used to gather sensed data from submerged sensors [4,12].This approach utilizes energy-efficient mechanisms such as effective mobi-cast packet delivery, maintenance of SLEEP-WAKE node schedule, three-dimensional zone referencing and forwarding, adaptive zone size adjustment based on network density, and AUV speed adjustment based on ocean current [6,8]. In CBR, the network adopts a dynamic hierarchical process of logically grouping the nodes into CHs and CMs with respect to well-defined performance indicators (residual energy, relative position) [6,18]. CMs sense the desired data and forward data packets to their respective CHs [14,17]. CHs are responsible for gathering and processing packets from their CMs, computing transmission schedule for their CMs, and establishing communication with other clusters [8,19]. CBR leads to significant energy consumption minimization and reduced redundancy by leveraging on data aggregation, internode connectivity, and network adaptability [14,17]. It is difficult to easily conclude that a particular category of routing protocol can effectively support all ranges of UWSN applications because each of the previously mentioned routing categories is application-specific, tailored to particular situations, and has its own respective strengths and weaknesses as shown in Table 9.3. However, researchers have established that CBR protocols are relatively more versatile and capable of yielding better performance in terms of fault tolerance, resource awareness, and route efficiency for large-scale UWSNs [6,18,19].

3.2  Expository analysis In Table 9.4, a comprehensive explorative analysis of state-of-the-art CBR protocols for UWSN is given.

3.3  Network performance and statistical analysis Based on the explorative analysis conducted in the preceding subsection, four CBR techniques (ACUN, CBE2R, EERBLC, EULC) were consequently chosen as baseline protocols due to their special emphasis on depth-based

Method

Assumption

Strength

Weakness

SPB-WD- Residual energy, Prevent FADpacket delivery sudden DBR & ratio, end-to-end network BFSPBdelay failure for WDFADnetwork DBR [27] lifespan enhancement SOSNET [28]

FBC [29]

Adaptive transThree-dimensional High packet High computamission range underwater envidelivery tional costs of adjustment to ronment, multiple ratio, relithe clustering address void hole sinks, sinks have ability algorithms lead issues, Dijkstra higher energy and to more energy and breadth-first self-knowledge of consumption search algorithms the location of all and end-to-end for reliable shortnodes delay est route selection Grid size, transmis- Efficient Dynamic clusterNodes have fixed po- Reduced Computational sion range, node routing ing, moth flame sitions, node mobilpacket routcomplexity, density for smart optimizer for deity is very slow ing cost, communication oceans termining cluster node-level overhead number energy conservation Energy consump- Cluster-based Fuzzy logic-based Random deployment, Network Sum of squared tion, dead nodes, data agclustering, data homogeneous and lifetime error issues number of clusgregation aggregation using static nodes, unlimmaximizafor increased ters, alive nodes similarity funcited energy for the tion, energy CH selection, tion with Euclidsink, 3-d underconsumpredundant data ean distance water scenario, tion minitransmission full-duplex commization munication

Explorative analysis of AUV-aided cluster-based routing protocols

Table 9.4  Explorative analysis of CBR techniques. CBR technique Metrics Focus

(Continued) 155

Assumption

Strength

Weakness

SD-UASN [30]

Abdulazeez Femi Salami, Emmanuel Adewale Adedokun, Fadi Al-Turjman, Habeeb Bello-Salau, Bashir Olaniyi Sadiq, Eustace M. Dogo

Computational Network Software-defined Fixed underwater Cluster stabil- Processing overcomplexlifetime clustering mecha- region, random deity, cluster head, computaity, number of maximizanism ployment (normal number tional complexsurviving nodes, tion or Poisson distribuminimizaity survival ratio of tion) tion CMS, cluster counts, cluster coverage ratio, changing CH counts E2R2P [31] Propagation delay, Power con- Dynamic reconfigu- Nodes always have a Improved fault High computareachability, trust sumption ration, trust value closest neighbor, recoverabiltional cost to value, error rate minimizaevaluation-based knowledge of closity, reduced find new trust tion data transmission est neighbor exists retransmisvalues with sion increased node failures severely drains battery power FBCPSO Number of clusters, Topology Fuzzy logic for Random deployment, Death rate High redundancy [32] dead nodes, control for cluster formation, fixed underwater of nodes introduced, sum total messages enhanced particle swarm sensors, unlimited reduction, of squared error received, alive network optimization for energy for the sink, network problems with nodes lifetime electing CHS data fusion is indelifetime enthe increase in pendent of number hancement cluster numbers of cluster nodes

156

Table 9.4  Explorative analysis of CBR techniques. (Cont.) CBR technique Metrics Focus Method

Stability, energy exhaustion rate

Network stability improvement

(Continued)

157

Particle swarm Water waves move Improved net- High latency isoptimization and slowly with a simple work stabilsues, high comc-fuzzy technique pattern, random ity, reduced putational cost for cluster mandeployment death rate of agement process nodes EH-ARStability period, Prolonging Piezoelectric Availability of sufReduced Varying and high CUN [34] end-to-end delay, operational energy harvestficient energy harpacket aggregate energy packet delivery duration of ing mechanism, vesting relay nodes, forwardconsumption by ratio UWSN amplify and forno direct communi- ing load on relay nodes, lack ward cooperative cation link between the source, of local relay technique source and sink improved selection process packet delivery ratio FSO [35] Energy exhaustion TransmisFirefly swarm Deployment in a Reduced High processing rate, stability sion time optimizationtwo-dimensional network overhead, netreduction based clustering underwater plane connectivity work becomes process failures unstable with increased network operation I-HENPC Network capacity Energy efJellyfish breathing Poisson node distriEnhanced Single-layer network [36] rate, number of ficiency for process-based CH bution, all sensors equalization issues, expensive nodes, network prolonging election,Voronoi have the same for residual for scenario using lifetime network diagram/autoradius, computing energy, mobile nodes, lifespan matic adjustment and battery resources, improved surface collision algorithm for the three-dimensional network issues, operational clustering process underwater space capacity and maintenance complexity

Explorative analysis of AUV-aided cluster-based routing protocols

OL-FCT [33]

Assumption

Strength

Weakness

Abdulazeez Femi Salami, Emmanuel Adewale Adedokun, Fadi Al-Turjman, Habeeb Bello-Salau, Bashir Olaniyi Sadiq, Eustace M. Dogo

CMDG [37] Energy tax, tour Strike a Meandering curStatic underwater Good scalabil- Not applicable for length, endbalance rent mobility to sensors, depth of ity perscenarios with to-end delay, between model movement underwater sensors formance, multiple sinks or relay hop count, energy of underwater are fixed, CMS are reduced AUVS, perforisolated sensors, conservasensors, mobile always connected to tour length, mance drops number of CHS, tion and data aggregation their CHS network with dense network lifetime, data agstrategy, distributed lifetime UWSN packet delivery gregation clustering process maximizaratio latency tion QL-EDR Optimization efResource Energy-Delay Three-dimensional Reduced Not applicable for [38] ficiency, network managerouting based underwater space, transmisdense or largelifetime, energy ment in on q-learning, random deployment sion delay, scale UWSN, consumption, hierarchienergy-delay with hexagonal network cannot work end-to-end delay, cal UWSN adjustment with characteristics lifespan enwith a flexible throughput regulatory factor hancement network with and metrics mobile nodes DPDCValue of informa- Data collec- Three-dimensional Uniform deployment, Load balanc- Network conflict EDC [39] tion, energy eftion and spatial monitorstatic and pre-clusing, reduced issues with ficiency, accomcontrol of ing region divitered nodes, nodes end-to-end multiple AUVS, plishment time, unexpectsion, Q-learningcan self-organize delay effect of underthroughput ed emerbased access path and determine their water movement gencies in planning for AUV locations on AUV is not UWSN in sub-regions accounted for

158

Table 9.4  Explorative analysis of CBR techniques. (Cont.) CBR technique Metrics Focus Method

SH-FEER [41]

DQELR [42]

AUV rounds, currency factor, residual energy

Overall network lifespan enhancement

AUV traversal algo- Homogeneous under- Reduced Lack of interrithms, sectorwater sensors, AUV processing cluster coming mechanism, knows the location time, battery munication, lack time slot-based of all sensors, all resources of flexibility as intra-cluster sensors are always conservation multi-hop comcommunication listening and waitmunication is mechanism ing for AUV not relied upon Energy exhaustion Improve Particle swarm Water waves move Energy con- High latency, rate, stability network optimizationslowly with a simple sumption computational stability for based CH (sinusoidal) pattern, is balanced and processing minimizelection, fuzzy-c random deployment via unequal overhead ing death means clustering clustering rate of strategy nodes End-to-end latency, Prolong Deep Q-network Underwater sensors Reduced High computanetwork lifetime, network with on/off can self-recharge latency, tional costs, packet delivery lifetime policies for deand have selfenergy condelays and ratio, energy eftermining global knowledge of sumption is complexities ficiency optimal routes, location, topology minimized from obtaining hybrid of unicast does not change for and updating and broadcast a short operational q-values communication period techniques for reducing overhead

Explorative analysis of AUV-aided cluster-based routing protocols

AEC [40]

(Continued) 159

Assumption

Strength

Weakness

Abdulazeez Femi Salami, Emmanuel Adewale Adedokun, Fadi Al-Turjman, Habeeb Bello-Salau, Bashir Olaniyi Sadiq, Eustace M. Dogo

SP-CBE2R Throughput, ener- Enhance Static courier nodes The sink is on water Reduced rate Switching of & IMgy consumption, network to avoid void surface, source nodes of void hole topology for CBE2R packet delivery throughput hole problem, are at deepest underoccurrence, transmission [43] ratio, end-to-end any-cast geowater layer, underwaimproved takes time and delay opportunistic apter network divided throughput leads to high laproach and Dijks- into seven equidistant tency, scalability tra's algorithm for layers with seven issues for larger route selection courier nodes network size MERP [44] Congestion, netEnsuring Multi-objective ge- Homogeneous under- Avoidance of Effect of mobilwork lifetime, qualitynetic algorithmwater sensors, sink route loops, ity of nodes not packet delivery aware data based cluster and underwater senreduced considered, high ratio, residual en- gathering management sors have self-knowlredundant delay introduced ergy, packet error process edge of their location transmission in sink-to-base rate, throughput, and depth informafrom multistation link delay tion, movement of hopping nodes due to water currents negligible PTSDAR Center selection pe- Efficient Aggregation based Static and self-config- Network High computa[29] riod, aggregation packet on palm tree urable underwater lifetime is tional cost and ratio and energy, routing structure, agents sensors, periodic maximized high latency energy consumpand data path traversal event monitoring, for switching tion, number aggregaalgorithm between master of aggregation tion and local centers, centers, network the effect of molifespan bility of nodes is not considered

160

Table 9.4  Explorative analysis of CBR techniques. (Cont.) CBR technique Metrics Focus Method

(Contiuned)

161

Number of CHS, Enhance Depth-based unThree-dimensional Energy con- Network topology energy conenergy efequal layering cubic volume sumption is not optimal, sumption, numficiency of and formation deployment, nodes is balanced, no support ber of surviving underwater of variable size have self-knowledge enhanced for differentinodes, period acoustic clusters to address of locations, the network ated services for first and 30% sensors hot spot issue sink is at center of lifetime to deliver QoS, of nodes death, UWSN, need for security of data residual energy, periodic data coltransmission is packets received lection not ensured by the sink CBE2R [46] Throughput, Node mobil- Weighted cluster- Cluster formation is Quality of Packet delivery ranetwork lifeity control based routing limited to pre-decommunitio severely suftime, energy for efprocess, division fined monitoring re- cation link fers when water consumption, ficient batof the field of ingion, the sink is on is improved, pressure breaks end-to-end delay, tery power terest into seven water surface, source effective multiple routes, packet delivery utilization layers, adoption nodes are at seabed, managesevere fading ratio of courier nodes underwater network ment of bat- effects lead to for packet fordivided into seven tery power high end-to-end warding layers with seven delay courier nodes EERBLC Network lifetime, Reduce Formation of lay- Sink has unlimited en- Enhanced data High energy [47] stability period, energy ers and unequal ergy, homogeneous transmission consumption throughput, deconsumpclusters, next forunderwater sensors and network during cluster livery ratio tion, error warder electionlifespan formation at the rate and based routing, network initiapropagaperiodic cluster tion phase tion delay update and maintenance

Explorative analysis of AUV-aided cluster-based routing protocols

EULC [45]

CDBR [48] Network lifetime, Overall end-to-end delay, energy packet drop, enconservaergy consumption tion

ESM [50]

Strength

Weakness

Adaptive depthbased routing

Nodes near the water Reduced Problems of node surface have short energy confailures and lifespan and they are sumption, damages with major contributors balanced large-scale to overall network node distriUWSN due to lifetime deteriorabution high network tion load and high processing demands Multiple noise Ensure Weighted CH elec- Random deployment, Enhanced Clock mismatches level, scalability, robustness tion process, CH nodes are smart and overall between nodes, data rate variaand relichain formation pre-programmed network timing inaccution ability for routing with monitoring reliability racy effects target, nodes have full-duplex communication links and adjustable power levels Mobile sink utility Packet drop Usage of 2 mobile Random deployImproved High packet ratio, network ratio minisinks, effective ment, all nodes are throughput drop rates and lifetime, residual mization data collection homogeneous with perforlatency issues for energy, packets by harness- and aggregation a fixed and same mance, effidense UWSN, throughput, ing mobile process energy cient energy network stability period, sinks utilization implementapacket drop ratio tion limited to a two-dimensional plane

Abdulazeez Femi Salami, Emmanuel Adewale Adedokun, Fadi Al-Turjman, Habeeb Bello-Salau, Bashir Olaniyi Sadiq, Eustace M. Dogo

JCRP [49]

Assumption

162

Table 9.4  Explorative analysis of CBR techniques. (Cont.) CBR technique Metrics Focus Method

(Contiuned)

163

Throughput, ener- Effective node On-demand packet Random deployReduced data Scalability isgy consumption, mainterouting, signal to ment, homogeneous transmission, sues, clustering cluster overheads, nance and noise ratio-based underwater sensors network processing is network lifetime energy clustering process with same sensor lifetime ennot flexible and conservacharacteristics hancement dynamic tion CBEER [52] Energy consump- Network Event-driven Homogeneous under- High packet High processing tion, alive nodes, lifetime evolutionary water sensors with delivery and computapacket delivery maximizaalgorithm-based same sensor characratio pertional overhead ratio tion clustering process teristics, nodes have formance, lead to latency simplex commureduced issues nication links with energy confixed transmission sumption range, nodes have self-knowledge of position ACUN [53] Energy consump- Minimize Multi-Level Communication Improved High deploytion, survival rate, energy loss network hierarlink is full-duplex cluster man- ment costs and network lifetime in packet chy formation and symmetrical, agement, scalability issues transmisprocess, energy physical and chemieffective as the number sion consumption balcal properties of balancing of of nodes to ancing based on seawater for the energy conbe deployed is routing rules and monitoring region sumption uncertain and hop adjustment is uniformly distribdepends on schemes uted monitoring area requirements, high communication overhead

Explorative analysis of AUV-aided cluster-based routing protocols

ODTMC [51]

Assumption

Strength

Weakness

MICUN [54]

Abdulazeez Femi Salami, Emmanuel Adewale Adedokun, Fadi Al-Turjman, Habeeb Bello-Salau, Bashir Olaniyi Sadiq, Eustace M. Dogo

Network coverage, Effective Murmuration Mobile underwater Reduced Redundant data energy connetwork (animal swarming sensors, two-dimen- clustering transmissions, sumption reconfigupattern)-based sional underwater overheads, high computaration clustering process, environment, comnetwork tional complexadaptive network munication link is coverage ity reconfiguration symmetric, all nodes enhanceshare same sensor ment characteristics EHO [55] Execution time, Prolong Elephant herd op- Nodes share common Overall Accuracy issues efficiency, cost network timization-based identifiers, CHS network and optimizafactor, overhead lifespan dynamic clusterare responsible for lifetime imtion errors for and complexity, ing process identifying destinaprovement dense UWSN, reliability tion and route manhigh computaagement tional cost CADC [56] Tour length, CH Balance the Improved branch- Static and homogeScalable, reHigh energy connumber, end-toenergy and-bound techneous underwater duced tour sumption during end delay, energy conservanique for optimal sensors, employlength data gathering, tax, relay hop tion and route selection, ment of AUV with high complexity count, packet latency randomized unlimited memory of branch-anddelivery ratio of data adaptive greedy and energy rebound techgathering search for routing sources nique makes it trade-off impractical for dense UWSN

164

Table 9.4  Explorative analysis of CBR techniques. (Cont.) CBR technique Metrics Focus Method

165

(Contiuned)

Explorative analysis of AUV-aided cluster-based routing protocols

A-DBR, Packet delivery Prolonged Adaptive transmission Three-dimensional Reduced void High-energy C-DBR, ratio, energy tax, network range adjustment, multiple sink scehole probconsumption B-DBR & end-to-end delay, lifetime by backup recovery nario, sinks comlem, end-to- for clustering CA-DBR propagation optimizing mechanism for municate with one end delay is operations in [57] distance available alternative route another using radio minimized sparse UWSN network selection, packet links, sink have selfregions, high resources collision miniknowledge of all propagation mization scheme, nodes information delays adoption of smallsized clusters for data aggregation FDRT [58] Delay, packet deliv- Intra-cluster Fuzzy logic-based Static underwater Rapid failure Severe drop in ery ratio, energy fault detec- backup clustering sensors, random detection packet delivconsumption, tion and process, TDMA deployment and recovery ratio with packet drop recovery scheduling for ery, reduced higher cases of data transmission energy confaults and aggregation sumption SEEC, Packet delivery Energy hole Sparsity and density Random deployment, High cluster Low throughput CSEEC & ratio, network avoidance search algorithms, one static sink lostabilissues, processCDSEEC lifetime, residual by reduclarge-scale cluscated at the center ity, network ing costs and [59] energy, throughing energy tering, reliance of UWSN and two lifetime enalgorithmic put, energy consumpon sink mobility, mobile sinks hancement complexities for consumption tion in hop minimizasearch algosparse retion using depth rithms lead to gions and threshold metric latency issues decreasing data load in dense regions

ABA [60]

Strength

Weakness

AUV traversal Underwater creatures Improved Lack of fault-toleralgorithm, path and obstacles do cluster man- ance and secudiscovery using not affect deployagement rity mechanisms mobile agents ment, homogeneous process, high for the mobile and clones, dyunderwater sensors packet deagents lead to namic clustering with similar sensor livery ratio, severe privacy mechanism characteristics, preless data and reliability programmed depth aggregation issues and location adjusttime ment algorithms Residual energy, Ensuring Optimal cluster Nodes fall into logical Network Communication dead nodes, endenergyformation using cubic groups, nodes lifetime is and computato-end delay efficiency small cubic pathave information improved, tional overhead, in UWSN terns, effective correlation abilreduced multiple failures communidata transmission ity for comparison energy concan lead to cation schemes with other nodes sumption severe network performance deterioration

Abdulazeez Femi Salami, Emmanuel Adewale Adedokun, Fadi Al-Turjman, Habeeb Bello-Salau, Bashir Olaniyi Sadiq, Eustace M. Dogo

EEGCMR [61]

Cluster formation Devising period, energy agentconsumption, ag- based gregation period, routing packet delivery strategy ratio, end-to-end delay

Assumption

166

Table 9.4  Explorative analysis of CBR techniques. (Cont.) CBR technique Metrics Focus Method

Energy consump- Low-cost Switching between Underwater sensors Reduced Heavy processing tion, data variaenergylocation-aware are semi-mobile, communioverhead for tions, stability efficient and location-free node position afcation cost, CH selection, factor, packet size routing for transmission, fected by water network data conflict isimpact UWSN nodal confidence currents, data traffic lifetime is sues as gathered applicalevel computais stable for 70% enhanced sensed data is tions tion, dynamic of the monitoring not differentiload adjustment period ated and priorifor cluster cotized for varying ordinators, relay ocean depths nodes and CHS QERP [63] Delay, packet deliv- Improve data Reliance on routing Underwater sensors Reduced Lack of considery ratio, energy transfer retable and dyshare same senlatency, eration for full consumption liability for namic power adsor characteristics, improved node mobility real-time justment scheme, symmetric compacket descenarios, high UWSN QoS-aware short- munication link, livery ratio packet collisions applicaest route selection two-dimensional for large-scale tions process, evolumobility allowed, and time-critical tionary clustering nodes can adjust applications technique power dynamically, the sink can simultaneously receive multiple packets

Explorative analysis of AUV-aided cluster-based routing protocols

E-CBCCP [62]

(Contiuned)

167

EBECRP [64]

Strength

Weakness

Stability period, Load reduc- Mobile sinks-based Sinks have selfStabilLow throughput network lifetime, tion on load adjustment, knowledge of dense ity period is issues, prone to throughput, nodes near data compression and sparse regions, maximized, network conflict packet drop, the sink scheme the sink does not network and packet collipacket actraverse a path prelifetime sion issues from ceptance ratio, viously visited by maximizaoverlapping residual energy another sink tion mobile sink interactions End-to-end delay, EnergyFuzzy logic cluster Cluster members sense Improved Lack of considerpacket delivery efficient formation prosimilar event and packet deation for node ratio, energy clustering cess, minimum they are close to livery ratio, mobility, packet consumption scheme for average routing each other, node reduced collision and agUWSN path selection mobility is not energy congregation delay applicafor intra-cluster allowed, coordisumption issues tions transmission, nates of nodes are hierarchiobtained through cal multi-path position calculation routing-based leach is used for inter-cluster transmission

Abdulazeez Femi Salami, Emmanuel Adewale Adedokun, Fadi Al-Turjman, Habeeb Bello-Salau, Bashir Olaniyi Sadiq, Eustace M. Dogo

IICC [65]

Assumption

168

Table 9.4  Explorative analysis of CBR techniques. (Cont.) CBR technique Metrics Focus Method

(Contiuned)

169

Number of clusters, Address Fuzzy logic-based Cluster management Reduced Not applicable to clustering precisparseness multi-cluster forprocess need to be network solve clustering sion, the time issues in mation process dynamically mainoverhead, issues in dense complexity clustering tained and updated enhanced UWSN, mobile and data overall scenarios and aggreganetwork other application for performance tions UWSN EEHC [67] Path loss, endQoS-aware Reliance on Random deployment, Improved High processto-end delay, and enerenergy-saving localization-free throughing cost and network lifetime, gy-efficient underwater senrouting process, put peralgorithmic throughput, routing for sors, balancing of nodes are equipped formance, complexity lead transmission loss UWSN energy consumpwith depth sensors network to heavy packet tion by switchto find their own lifetime is burden during ing transmission depth information maximized network initiamode tion phase OVAR [68] Packet delivery Improve Clustering coupled Random deployment, Low-cost High commuratio, energy throughwith adjacency nodes are equipped solution to nication and consumption, put and graph construcwith depth sensors void hole computational end-to-end delay reliability tion, forwarding to find their own problems, overhead, local in sparse set selection using depth information, improved routing inand lossy distributed beavertical node mothroughput formation are UWSN coning scheme, bility is negligible, performance invalid for dense balancing of underwater sensors UWSN applicaenergy consumpare homogeneous tions tion and reliabiland share same senity trade-off sor characteristics, lossy channel

Explorative analysis of AUV-aided cluster-based routing protocols

FULCAN [66]

URSA [69]

Network coverage, Distributed reliability, energy self-deconsumption ployment scheme

Optimize network energy utilization with reduced clustering load

Strength

Weakness

Uneven clustering Node sensing based Network con- Algorithm lacks process, hybrid on Boolean percepnectivity adaptability radius route setion model, nodes guaranteed, and practicallection technique are isomorphic enhanced ity as obstacles and can self-adjust reliability are not considtransmission power, ered, clustering only one sink with process suffers unlimited memory from position and power resources adjustment and mobile node deployment issues Layered dynamic All nodes have unique Prolong net- Lack of flexibility dual-cluster route ids, data are received work lifesleads to scalabilformation using by sink periodically pan, energy ity issues with krill herd optimiwithout delay, sink consumed different cluster zation, Lagrange lacks mobility and for clusters sizes in the clusmodel for guidcommunication is reduced tering process ing nodes to high features likelihood regions

Abdulazeez Femi Salami, Emmanuel Adewale Adedokun, Fadi Al-Turjman, Habeeb Bello-Salau, Bashir Olaniyi Sadiq, Eustace M. Dogo

DC-KH [70] Energy consumption, surviving nodes, network lifetime

Assumption

170

Table 9.4  Explorative analysis of CBR techniques. (Cont.) CBR technique Metrics Focus Method

Network coverage, network connectivity, reconstruction rate, network lifetime

CBKU [72] Network lifetime, energy consumption, data transfer volume

KEER & EKEER [73]

Energy consumption, network lifetime

Improving network coverage

(Contiuned)

171

Clustering with Random distribution, Node disHigh energy conheterogeneous three-dimensional placement sumption issues, communication cubic volume avoidance, lack of accuracy range, usage of deployment, nodes enhanced and exactness as aggregate contrishare similar sensor network in real UWSN bution degree characteristics, nodes coverage scenarios, node have self-knowledge drift is unavoidof location and can able due to move in all directions water currents Avoid unbal- An improved kHomogeneous under- Energy con- Processing and anced means technique water sensors, CHS sumption computational clustering for cluster manmaintain direct is balanced, overhead lead to issues agement communication enhanced rapid death of with sink, threedata transnodes and high dimensional static mission energy conunderwater envisumption issues ronment, symmetric communication link, sink operate with acoustic and radio Effective K-means clustering Homogeneous under- Prevention of Shortest path is routing for used in a cluswater sensors, nodes overconnot adopted for energytering process, periodically sense at sumption routing, effects constrained reliance on data a fixed rate of battery of node mobilUWSN aggregation and power reity and undermulti-hop transsources water conditions mission not considered

Explorative analysis of AUV-aided cluster-based routing protocols

NNDBC [71]

Assumption

Strength

Weakness

EGRC [74] Energy efficiency, reliability, endto-end delay, alive nodes

Abdulazeez Femi Salami, Emmanuel Adewale Adedokun, Fadi Al-Turjman, Habeeb Bello-Salau, Bashir Olaniyi Sadiq, Eustace M. Dogo

Impact of Three-dimensional Three-dimensional Energy con- Lack of fault-tolUWSN geospatial divicubic volume deservation, erant and error deploysion, information ployment, random network recovery mechament on storage mechadeployment, nodes lifetime is nisms can lead energy nism, duty-cycled pre-programmed extended to severe netconsumpscheduling for with positioning work deterioration node operation algorithm, homogetion for UWSN neous underwater deployed to sensors with unique monitor critiids, fixed sink, all cal underwater nodes move with infrastructures the same speed TCBR [75] Node mobility, en- Delay-toler- Reliance on Node operation Cost of netHigh communicaergy consumpance and strategic data needs a customized work mantion overhead tion, reliability energy-efforwarding nodes, mechanical module, agement is due to indirect ficiency for CH-to-sink mobile and homoreduced transmissions, UWSN communication geneous underwater high design and uses multi-hop sensors maintenance transmission cost for special mechanical module

172

Table 9.4  Explorative analysis of CBR techniques. (Cont.) CBR technique Metrics Focus Method

Throughput, delay, Effective col- Reliance on relay Uniform deployment, Improved Network conflict density of nodes, lision and nodes for data static sink located at packet deand channel probability of flooding packet forwardthe center of monilivery ratio, content issues packet forwardavoidance ing, implementatoring region enhanced due to the ining for routing tion of clusternetwork creasing number in UWSN based urgent lifetime of relay nodes route formation for immediate packet forwarding DUCS [77] Routing overhead, Effects of Adoption of data Underwater sensors Low comSevere network packet delivery node aggregation techalways in sending munication deterioration ratio, alive nodes mobility niques for minimode, periodic data overhead, due to single and delaymizing redundant collection, nodes energy conpoint of failure tolerance data transmission have self-knowledge sumption is issues, high on UWSN of adjacent cluster minimized processing and perforlocation compression mance costs MCCP [78] Cluster distribution, Energy op- Exploiting nonRandom deployNetwork High energy connetwork lifetime timization overlapping clusment, nodes have adaptability, sumption issues (temporal & for UWSN ters for low-cost self-knowledge of reduced data as the algorithm capacity) applicapacket forwardadjacent and postredundancy do not suptions ing, reliance on adjacent clusters port multi-hop cluster-based location transmission, direct single-hop high control & communication communication overhead

Explorative analysis of AUV-aided cluster-based routing protocols

E-PULRP [76]

173

174

Abdulazeez Femi Salami, Emmanuel Adewale Adedokun, Fadi Al-Turjman, Habeeb Bello-Salau, Bashir Olaniyi Sadiq, Eustace M. Dogo

multilayer hierarchical cluster formation strategies that are in line with the aim and future works of this research. The performance metrics considered in this research are residual energy ( ∑ ), cluster centrality (µ ), received packets (θ), cluster proximity (ψ), end-to-end delay (∆), clustering coefficient (δ), routing overhead ( Ω  ), cluster dissimilarity (λ), computational complexity ( Π ), cluster connectivity (τ  ), cluster coverage (Γ ), and CH overhead ( Φ ). The ∑-metric measures the aggregate remaining energy (in Joules) for all the underwater sensor nodes after a specified round of network operation.The µ-metric determines the significance or centrality of a CH in the UWSN by evaluating how near CMs are to their respective CHs for any given cluster. This metric displays the effect of CH-based routing operation on the data traffic flow.The normalized form of ∑-metric is defined in Eq. (9.1) as [79,80]: NC − 1 µ= Σ CM ∈C dist ( CH,CM ) (9.1) From Eq. (9.1), NC is the number of underwater sensors in a cluster (C) and dist() is the function computing distance between the CMs and their CH. The θ-metric is the sum total of received data packets at the sink from all the underwater sensor nodes after a specified round of network operation. The ψ-metric measures CH proximity to its CMS by using shortest path distance. This metric is critical for ensuring energy conservation during the data transmission process. The normalized form of ψ-metric is expressed as [79,80]:

(

)

prox CM i ,CM j |CH 2 ψ− × Σ CM ≠ CM ≠ CH ∈C ( N C − 1)( N C − 2 ) ΣC prox CM i ,CM j (9.2) i

j

(

)

From Eq. (9.2), prox() is a function computing shortest distance between CMs and their CH.The ∆-metric measures the aggregate delay (in seconds) taken to forward packets from source nodes to the sink for a specified round of network operation.The δ-metric quantifies the cluster formation rate for CMs around their CH. This metric explains the relationship between cluster density, node distribution and intercluster data transmission. The normalized form of δ-metric is given as [79,80]:

(

)

dist CM i ,CM j 1 δ= × Σ CM ≠ CM ≠ CH ∈C (9.3) NC N C ( N C − 1) i

j

Explorative analysis of AUV-aided cluster-based routing protocols

175

The Ω-metric gives the ratio (in percentage) of routing processing time to the network operational period. The λ-metric evaluates the degree of intersecting CMs shared between adjacent cluster sets. This measure is essential for ensuring fair utility of CMs and load balancing. The normalized form of λ-metric is expressed as [79,80]:

Ε CMi ∈Ci CM i − Σ CMj ∈C CM j λ = N Ci + N Cj − 1 (9.4) j

The ∏-metric measures the total execution time (in seconds) for the CBR algorithms for a given round of network operation.The τ-metric determines the minimal number of hops for intracluster packet transmission. This metric is useful for understanding the effects of multihop data transmission on energy consumption pattern. The normalized form of τ-metric is given as [79,80]: prox ( CM i ,CH ) 2 τ= × Σ CMi ≠ CH ∈C N C ( N C + 1) ΣC prox ( CM i ) (9.5) The Γ-metric gives the ratio (in percentage) of successfully clustered nodes to the node density (including clustered and orphaned nodes). The Φ-metric is the ratio (in percentage) of cluster formation and processing time to the network operational period. The simulation parameters employed for the entire experiment are as indicated in Table 9.5. From Table 9.5, N, NoS, PKS, A, DEP, ASR, CTR, ED, LoS, MS, RTT, FREQ, LD, CLD, SAL, TEMP, PRS, DST, MAC, and ENV represent number of nodes, number of sinks, packet size, UWSN area, deployment nature or distribution, average of statistical results, critical transmission range, energy distribution, location of sinks, mobility support features, number of random topologies tested, acoustic signal carrier frequency, layer depth, cluster diameter, water salinity, water temperature, water pressure, water density, underwater medium access control protocol, and UWSN environment type, respectively. It must be mentioned that the original weight coefficients for route development, competition radius, next hop selection cost, and CH selection cost functions employed in the four baseline CBR protocols were kept intact to ensure fair comparison.

4  Results and discussion The quantitative results obtained using the statistical metrics ( µ, ψ, δ, |λ|, τ ) are presented in Table 9.6.

Abdulazeez Femi Salami, Emmanuel Adewale Adedokun, Fadi Al-Turjman, Habeeb Bello-Salau, Bashir Olaniyi Sadiq, Eustace M. Dogo

176

Table 9.5  Simulation parameters. Parameters Values

N NoS PKS A DEP ASR CTR ED LoS MS RTT FREQ LD CLD SAL TEMP PRS DST MAC ENV

400 B 2 200 500 m × 500 m × 500 m Uniform 75 100 m 5 J (constant) Surface center Horizontal, random mobility and variable current (0.2–3 m/s) 50 10 kHz 60 m 30 m Variable (32–38 ppt) Variable (3–30°C) Variable (2–50 bars) 1025 kg/m3 802.11-DYNAV 3-D

Table 9.6  Statistical performance metrics. Metric ACUN EULC

CBE2R

EERBLC

µ ψ δ |λ| τ

0.2390 0.3136 0.2932 0.0840 0.6641

0.3025 0.4459 0.3784 0.1609 0.3532

0.1950 0.2829 0.1707 0.0147 0.0209

0.1348 0.2216 0.2032 0.0450 0.0672

From Table 9.6, it is observed that CBE2R and EERBLC exhibit better performance than ACUN and EULC with respect to µ, ψ, δ and τ. The implication of this observation is that CH and CMs are more quickly configured and utilized for energy-efficient data transmission in CBE2R and EERBLC than in ACUN and EULC. The reason for this is the periodic cluster update and maintenance, weighted CBR process, flexible cluster size and dynamic layer formation, and other algorithmic enhancements introduced into CBE2R and EERBLC. However, it is observed that ACUN and EULC exhibit better performance than CBE2R and EERBLC with respect to |λ|. This means that there are relatively fewer cases of redundant

Explorative analysis of AUV-aided cluster-based routing protocols

177

packet transmission from intersecting cluster sets, which ensures reduced end-to-end delay.The reason for this is the hop adjustment schemes, inbuilt routing rules, multilevel hierarchy formation, and other simple protocol enhancements adopted in ACUN and EULC. Comparative results obtained using the network performance metrics (Φ, Γ, Π, Ω, ∆, ∑, θ  ) are given in Figs. 9.4–9.10.

Figure 9.4  Residual Energy Analysis.

Figure 9.5  Received Packets Analysis.

178

Abdulazeez Femi Salami, Emmanuel Adewale Adedokun, Fadi Al-Turjman, Habeeb Bello-Salau, Bashir Olaniyi Sadiq, Eustace M. Dogo

Figure 9.6  End-to-End Delay Analysis.

From Fig. 9.4, it is observed that ACUN approximately consumes a similar amount of energy to EULC, CBE2R, and EERBLC at lower network rounds (