MAR 2021 IEEE Spectrum

Table of contents :
Spectrum0321_cover1
Spectrum0321_cover2
Spectrum0321_01
Spectrum0321_02
Spectrum0321_03
Spectrum0321_04
Spectrum0321_05
Spectrum0321_06
Spectrum0321_07
Spectrum0321_08
Spectrum0321_09
Spectrum0321_10
Spectrum0321_11
Spectrum0321_12
Spectrum0321_13
Spectrum0321_14
Spectrum0321_15
Spectrum0321_16
Spectrum0321_17
Spectrum0321_18
Spectrum0321_19
Spectrum0321_20
Spectrum0321_21
Spectrum0321_22
Spectrum0321_23
Spectrum0321_24
Spectrum0321_25
Spectrum0321_26
Spectrum0321_27
Spectrum0321_28
Spectrum0321_29
Spectrum0321_30
Spectrum0321_31
Spectrum0321_32
Spectrum0321_33
Spectrum0321_34
Spectrum0321_35
Spectrum0321_36
Spectrum0321_37
Spectrum0321_38
Spectrum0321_39
Spectrum0321_40
Spectrum0321_41
Spectrum0321_42
Spectrum0321_43
Spectrum0321_44
Spectrum0321_45
Spectrum0321_46
Spectrum0321_47
Spectrum0321_48
Spectrum0321_49
Spectrum0321_50
Spectrum0321_51
Spectrum0321_52
The Institute0321_TI01
The Institute0321_TI02
The Institute0321_TI03
The Institute0321_TI04
The Institute0321_TI05
The Institute0321_TI06
The Institute0321_TI07
The Institute0321_TI08
The Institute0321_TI09
The Institute0321_TI10
The Institute0321_TI11
The Institute0321_TI12
The Institute0321_TI13
The Institute0321_TI14
The Institute0321_TI15
The Institute0321_TI16
Spectrum0321_cover3
Spectrum0321_cover4

Citation preview

WHEN CHAOS IS A GOOD THING Taking stress testing to an extreme P. 04

FOR THE TECHNOLOGY INSIDER | 03.21 FOR THE TECHNOLOGY INSIDER | 03.21

FINDING ANOTHER EARTH It’s out there. Here’s how we’ll find it P. 22

RADIO’S NEXT REVOLUTION Truly duplex radios could energize 5G P. 30

HOW INSTACART WON US OVER It wasn’t just the pandemic P. 36

POWERING CARGO SHIPS WITH AMMONIA A bold plan for zeroemission shipping P. 44

Sensor Control and Characterization. All in one.

Lock-in Amplifiers and more

experience

Numeric

Scope

Spectrum

Sweeper

PID

PLL

LabOne

Applications

Your Benefits

∏Inertial devices ∏Capacitive and resistive sensors ∏Light detectors ∏Hall sensors

∏Simplify your setup with a single instrument for characterization and control ∏Automate your workflow with full computer control ∏Find feedback settings quickly with the PID Advisor ∏Track and control multiple resonances simultaneously

Zurich Instruments

→ Challenge us with your requirements

www.zhinst.com

CONTENTS_03.21

THE AMMONIA SOLUTION

Cargo ships with ammonia engines and fuel cells could slash carbon emissions. by M a r i a Ga llucci

Page 44

EIDESVIK

On the cover Illustration for IEEE Spectrum by Peter Sanitra/ NoEmotion

22 HOW WE’LL FIND ANOTHER EARTH Better tools may locate a world similar to our own. By Jason Wright & Cullen Blake

30 THE RADIO THAT CAN HEAR OVER ITSELF Wireless networks gain efficiency with self-interference cancellation. By Joel Brand

36 THE ALGORITHMS THAT MAKE INSTACART ROLL Massive databases, machine learning, and apps power the shopping system. By Sharath Rao & Lily Zhang

04 NEWS 14 HANDS ON 18 CROSSTALK 51 PAST FORWARD

The Institute

TI-12 THE AUTOMATED DISHWASHER Water pressure was the key to a socialite’s invention.

IEEE SPECTRUM (ISSN 0018-9235) is published monthly by The Institute of Electrical and Electronics Engineers, Inc. All rights reserved. © 2021 by The Institute of Electrical and Electronics Engineers, Inc., 3 Park Avenue, New York, NY 10016-5997, U.S.A. Volume No. 58, Issue No. 3. The editorial content of IEEE Spectrum magazine does not represent official positions of the IEEE or its organizational units. Canadian Post International Publications Mail (Canadian Distribution) Sales Agreement No. 40013087. Return undeliverable Canadian addresses to: Circulation Department, IEEE Spectrum, Box 1051, Fort Erie, ON L2A 6C7. Cable address: ITRIPLEE. Fax: +1 212 419 7570. INTERNET: [email protected]. ANNUAL SUBSCRIPTIONS: IEEE Members: $49.95 included in dues. Libraries/institutions: $399. POSTMASTER: Please send address changes to IEEE Spectrum, c/o Coding Department, IEEE Service Center, 445 Hoes Lane, Box 1331, Piscataway, NJ 08855. Periodicals postage paid at New York, NY, and additional mailing offices. Canadian GST #125634188. Printed at 120 Donnelley Dr., Glasgow, KY 42141-1060, U.S.A. IEEE Spectrum circulation is audited by BPA Worldwide. IEEE Spectrum is a member of the Association of Business Information & Media Companies, the Association of Magazine Media, and Association Media & Publishing. IEEE prohibits discrimination, harassment, and bullying. For more information, visit https://www.ieee.org/web/aboutus/whatis/policies/p9-26.html.

SPECTRUM.IEEE.ORG | MAR 2021 |  01

BACK STORY

TAP. CONNECT. NETWORK. SHARE. Connect to IEEE– no matter where you are–with the IEEE App.

THE SHIPPING NEWS

Schedule, manage, or join meetups virtually Get geo and interest-based recommendations Read and download your IEEE magazines Create a personalized experience

03.21

Locate IEEE members by location, interests, and affiliations

02  | MAR 2021 | SPECTRUM.IEEE.ORG

eing a tech jour na list can take you to unusual places.

Just ask IEEE Spectrum contributing editor Maria Gallucci, shown here in late 2017 aboard an experimental cargo ship in the middle of the North Sea. The trip was the capstone of her shipping-focused itinerary, which began at the United Nations’ annual climate conference in Bonn, Germany. Ahead of the conference, Gallucci met representatives of Norsepower, a Helsinki-based company that makes rotor sails. The novel wind-propulsion system allows boats to throttle back their main engines. (The tall white structures in the photo above are the rotor sails.) They invited Gallucci to take a three-day round-trip voyage aboard the M/V Estraden from Rotterdam, in the Netherlands, to northern England. Yes, she’d be the only passenger and the sole woman. But as a reporter who had recently become interested in cleaner types of propulsion, she was intrigued. “For most people, the shipping industry is out of sight, out of mind,” Gallucci notes. “But it’s a significant contributor to greenhouse-gas emissions. I wanted to look at what kinds of tech will be needed to decarbonize the industry.” She’s spent the last three years doing exactly that. She’s written about ships that run on hydrogen fuel cells and batteries and maritime biofuels. And in this issue, she reports on one of the most promising efforts: engines and fuel cells that run on ammonia, which has low environmental impact when produced using renewable energy. “I first heard about ammonia at the conference in Bonn,” Gallucci recalls. “It almost seemed like an aside at the time, so I was surprised to see the industry starting to coalesce around ammonia.” Among other big developments, the International Maritime Organization has set a goal of halving shipping’s greenhouse-gas emissions by midcentury. “That’s the reason ammonia is getting all this attention,” Gallucci says. “Before it was just a few companies trying different initiatives. Now there’s a real sense of urgency.” ■

MARIA GALLUCCI

B

Stay up-to-date with the latest news

simulation case study

Looking beyond our solar system with ray tracing simulation... Astronomers detected an Earth-like planet 11 light-years away from our solar system. How? Through data from an échelle spectrograph called HARPS, which finds exoplanets by detecting tiny wobbles in the motion of stars. Engineers looking to further the search for Earth-mass exoplanets can use ray tracing simulation to improve the sensitivity of échelle spectrographs. learn more comsol.blog/echelle-spectrographs

The COMSOL Multiphysics® software is used for simulating designs, devices, and processes in all fields of engineering, manufacturing, and scientific research.

News CHAOS ENGINEERING SAVED YOUR NETFLIX Extreme stress testing of online platforms has become its own science

To hear Greg Orzell tell it, t he i r ­c oron av i r u s- c omprom i s e d the original Chaos Monkey circumstances. tool was simple: It randomly Chaos engineering is a kind of highpicked a virtual machine hosted some- octane active analysis, stress testing where on Netflix’s cloud and sent it a taken to extremes. It is an emerging “Terminate” command. Unplugged it. approach to evaluating distributed netThen the Netf lix team would have to works, running experiments against a figure out what to do. system while it’s in active use. CompaThat was a decade ago now, when nies do this to build confidence in their ­Netflix moved its systems to the cloud operation’s ability to withstand turbuand subsequently nav igated itself lent conditions. around a major U.S. East Coast serOrzell and his Netflix colleagues built vice outage caused by its new partner, Chaos Monkey as a Java-based tool from ­A mazon Web Services (AWS). the AWS software development kit. The Orzell is currently a principal software tool acted almost like a number generaengineer at GitHub and lives in Mainz, tor. But when Chaos Monkey told a virtual Germany. As he recently recalled the machine to terminate, it was no simulaearly days of Chaos Monkey, Germany got tion. The team wanted systems that could ready for another long round of COVID- tolerate host servers and pieces of applirelated pandemic lockdowns and deathly cation services going down. “It was a lot fear. Chaos itself raced outside. easier to make that real by saying, ‘No, But while the coronavirus wrenched no, no, it’s going to happen,’ ” Orzell says. daily life upside-down and inside out, “We promise you it will happen twice in a prac t ice c a l led c haos eng i neer- the next month, because we are going to ing, applied in computer networks, make it happen.’ ” might have helped many par ts of In controlled and small but still signifithe networked world limp through cant ways, chaos engineers see if systems

04  | MAR 2021 | SPECTRUM.IEEE.ORG

work by breaking them—on purpose, on a regular basis. Then they try to learn from it. The results show if a system works as expected, but they also build awareness that even in an engineering organization, things fail. All the time. As practiced today, chaos engineering is more refined and ambitious still. Subsequent tools could intentionally slow things down to a crawl, send network traffic into black holes, and turn off network ports. (One related app called Chaos Kong could scale back company servers inside an entire geographic region. The system would then need to be resilient enough to compensate.) Concurrently, engineers also developed guardrails and safety practices to contain the blast radius. And the discipline took root. At Netf lix, chaos engineering has evolved into a platform called the Chaos Automated Platform, or ChAP, which is used to run specialized experiments [see infographic, “Spawning Chaos”]. Nora Jones, a software engineer, founder and chief executive of a startup called Jeli, says teams need to understand when and where to experiment. She helped implement ChAP while still at Netflix. “Creating chaos in a random part of the system is not going to be that useful for you,” she says. “There needs to be some sort of reasoning behind it.” Of course, the novel coronavirus has added entirely new kinds of chaos to network traffic. Traffic fluctuations during the pandemic did not all go in one direction either, says AWS principal solutions architect Constantin Gonzalez. Travel

investment institution in Singapore with US $437 billion in assets—they helped. DBS is three years into a network resiliency program, s­ ite-reliability engineer Harpreet Singh says, and even Netflix users 5 Netflix engineer as the program got off the ground in Netflix early 2018 the team was experimenting “front door” Internet with chaos-engineering tools. 1 And chaos seems to be catching. Jones’s Real-time ChAP monitoring Jeli startup delivers a strategic view on 6 External system cloud-service what she calls the catalyst events (events provider that might be simulated or sparked by 4 Netflix chaos engineering), which show the difcontrol ference between how an organization plane thinks it works and how it actually works. Gremlin, a four-year-old San Jose venture, 2 3 Continuousdelivery offers c­ haos-engineering tools as sersystem vice products. In January, the company also issued its first “State of Chaos EngiA Netflix engineer uses the company’s Chaos Automation Platform (ChAP) [1] to connect neering” report for 2021. In a blog post with its continuous-delivery system [2]. The system reaches Netflix’s external cloudannouncing the publication, ­Gremlin service provider [3], which, per ChAP’s orders, will slightly modify operations on a number vice president of marketing Aileen of test subjects to determine where the stress points in the system’s “control plane” [4] are. When subscribers cross the service’s virtual front door [5] and select a video to ­Horgan described ­c haos-engineering watch, a few are quietly delivered the altered service as part of their viewing experience. conferences these days as topping (One example Netflix describes is intentionally altering the bookmarking service for its 3,500-plus registrants. Gremlin’s user test users: If they exit from Netflix in the midst of watching a video, the service might lose its ability to remember the users’ stopping point.) ChAP then instructs its real-time base alone, she noted, has conducted monitoring system [6] to observe the test users’ experience to ensure that ChAP’s small nearly 500,000 chaos-engineering syserrors are compensated for and don’t result in cascading failures or crashes. tem attacks to date. Gonzalez says AWS has been using chaos-engineering practices for a long time. This year—as the net worked services like the German charter giant ­moonshot-era rocket science: If some- world, hopefully, recovers from stress Touristik Union International (TUI), for thing can go wrong, it will go wrong. that tested it like never before—AWS is instance, drastically pulled in its sails as It’s tough to say that the practice launching a fault-injection service for traffic ground to a halt. But the point in kept the groaning networks up and its cloud customers to use to run their building resilient networks is to make running during the pandemic. There own experiments. them elastic, he says. are a million variables. But for those Who knows how they’ll be needed. Chaos engineering is geared for this. using chaos-engineering techniques— —Mich a el Dumi a k As an engineering mind-set, it alludes even for as far-flung and traditional a POST YOUR COMMENTS AT to Murphy’s Law, developed during business as DBS Bank, a consumer and spectrum.ieee.org/chaos-mar2021

SOURCE: A. BASIRI ET AL., “AUTOMATING CHAOS EXPERIMENTS IN PRODUCTION,” PROCEEDINGS OF THE 41ST INTERNATIONAL CONFERENCE ON SOFTWARE ENGINEERING, 2019

SPAWNING CHAOS NETFLIX ORIGINATED “CHAOS ENGINEERING.” HERE’S HOW THEY DO IT

NEWS SPECTRUM.IEEE.ORG | MAR 2021 |  05

BETTING BIG ON HYDROGEN: In 2020, the world’s largest hydrogen plant and greenenergy center opened in Japan—the Fukushima Hydrogen Energy Research Field.

ual’s radiation dose to 1 millisievert a year, a generally accepted international standard. Nevertheless, some 337 square kilometers within seven Fukushima municipalities continue to be designated “difficult-to-return zones,” while a critical Greenpeace radiation survey report published in 2019 warned that forests in the region, which have never been decontaminated, “will continue to be long-term sources of recontamination.” To help both revitalize the stricken area and advance the country’s decarbonization efforts, the government in 2014 established the Fukushima Renewable Energy Institute, AIST (FREA) in Koriyama, Fukushima prefecture, says Masaru Nakaiwa, FREA’s ­director-general. (“AIST” stands for the National Institute of Advanced Industrial Science and Technology.) FREA’s mandate is to work with industry and academia to improve photovoltaic and wind-turbine performance, optimize ground-source heat pumps and geothermal resources, and develop technologies for hydrogen-energy carriers and hydrogen-energy systems. Ten years after the Fukushima disaster, the nation “Fukushima prefec tural governdoubles down on decarbonizing—with sustainable ment has set a target of producing all energy in the driver’s seat of ­Fukushima’s energy demands from renewable sources by 2040,” says Nakaiwa. To do this, the government is working with FREA, industry, and When the tsunami generated vide 20 percent of the nation’s power by universities to help commercialize by the Great East Japan Earth- 2030), the prospect of a z­ ero-carbon research in renewable technologies quake struck the Fukushima future in Japan still leaves the lion’s and increase the use of solar, biomass, Daiichi Nuclear Power Plant on 11 March share to renewables. and wind generation in the prefecture. 2011, it not only knocked out the plant The magnitude 9.0 earthquake also Hydrogen is also viewed as an important but eventually led to the shutdown of killed nearly 20,000 people, with 2,500 new energy resource. The prefecture is all the country’s 54 nuclear reactors as still missing. As of last December, some now home to the Fukushima Hydrogen a safety precaution. Ten years on, just 42,000 of the total 470,000 evacuees Energy Research Field, the world’s largnine reactors have come back on line. remained evacuated, even as the disas- est green-hydrogen production facility, And while nuclear energy in Japan today ter’s 10th anniversary loomed. The capable of supplying 1,200 cubic meters is anything but dead (the central gov- government has directed its decontam- of hydrogen an hour. This new focus is ernment now hopes nuclear could pro- ination efforts to reducing an individ- in keeping with past and recent cen-

06  | MAR 2021 | SPECTRUM.IEEE.ORG

KYODO NEWS/GETTY IMAGES

JAPAN’S RENEWABLES RENAISSANCE

tral government announcements technologies powered by renewon hydrogen and the goal to make able energies, says Manabu Ihara, Japan carbon neutral by 2050. director of the Tokyo Tech AcadAc h i ev i n g t h e 2 05 0 t a r ge t emy of Energy and Informatics at won’t be easy. Whereas nuclear the Tokyo Institute of Technology. accounted for 30 percent of the FREA has already demonstrated country’s energy use before the a green-hydrogen supply chain accident, today it provides just and a hydrogen cofiring genera6 percent. Making up the short- tor system, as well as the successful fall, Japan now relies more on coal synthesis of ammonia (NH3) from (25 percent), natural gas (23 per- green hydrogen, and its use to fuel cent), and oil (39 percent), with a modified micro gas-turbine genrenewables and hydro accounting erator. (Hydrogen could also be for the rest, as of April 2018. used in ammonia-powered cargo To encourage industry to work ships; see p. 44 in this issue.) Curtoward carbon neutrality, the gov- rently FREA is working with IHI ernment will provide capital invest- Corp. and Tohoku University to ment, tax relief, and deregulation in develop larger generator systems areas such as wind power; carbon using liquid ammonia spray injeccapture, utilization, and storage; tion, says Hirano. and the mass production of storOther countries are also developage batteries. ing green-hydrogen projects. China At the end of 2018, some 55 giga- has a major project underway in watts of solar power equipment had Inner Mongolia slated to produce been installed around Japan, put- 454,000 metric tons annually; the ting the country on track to surpass European Union estimates spendthe government’s target of 64 GW ing €430 billion (about US $520 bilby 2030. Regarding wind power, lion) over the next 10 years on however, Japan had only 3.6 GW of hydrogen technologies, while South equipment installed in 2018, hence Korea is aiming to become a leader Japan’s Ministry of Economy, Trade in developing clean hydrogen. and Industry noted it as technology Meanwhile, Japan is creating to invest in. international supply chains for More notable is the country’s shipping green hydrogen and embrace of hydrogen as a versa- “blue” hydrogen (using carbon tile energy-storage medium. Hydro- capture and storage) to the coungen can be produced from various try, and has established pilot projkinds of natural resources, in par- ects in Brunei and Australia to test ticular the water used for electroly- the feasibility of the scheme. These sis, which removes carbon dioxide, overseas and domestic sources of says Satoshi Hirano, FREA’s deputy clean hydrogen fueling large-scale director-general. And hydrogen can modified gas turbines will eventube compressed, stored, transported, ally take on the role of supplying and converted into electricity or base load power to the electric grid heat when needed, without emit- that can replace nuclear power, ting CO2. says Ihara, of the Tokyo Institute Hydrogen’s major downside is of Technology. “And we should see the high cost of production. Hence this partly realized before 2030.” FREA and other national research –John Boyd institutes are developing efficient, POST YOUR COMMENTS AT low-cost hydrogen-production spectrum.ieee.org/renewables-mar2021

IEEE Spectrum’s Fukushima Coverage How our contributors tackled an unprecedented disaster

When a magnitude 9.0 earthquake struck off the Pacific coast of Tōhoku, Japan, in March 2011, the tsunami it triggered devastated the Fukushima Daiichi Nuclear Power Plant—and sent two Spectrum contributors in particular into a 24-hour cycle of near-constant coverage. As Senior Editor Eliza Strickland recalls, she would work on stories throughout the day based in our New York office, while Japan-based contributor John Boyd (see main story) would work dawn to dusk on his portion of our extensive Fukushima reporting. Strickland and Boyd—along with at least a half dozen other editors and contributors— kept readers abreast of the region’s many various mitigation, cleanup, and emergencyresponse efforts. During March 2011 alone, Spectrum’s blogs featured 23 articles on the constantly changing circumstances postdisaster. Our special report on Fukushima helped us win a National Magazine Award for General Excellence in 2012. And these days, as Boyd reports, Japan is pushing hard toward renewable-energy generation and infrastructure, including at a renewableenergy research institute established 67 kilometers from the disaster site. “It’s heartening to see Fukushima prefecture recovering,” Boyd now says, “though the accident’s impact is still very much evident.” –MARK ANDERSON

NEWS SPECTRUM.IEEE.ORG | MAR 2021 |  07

JOURNAL WATCH

Superconducting CPUs

–MICHELLE HAMPSON

NEWS 08  | MAR 2021 | SPECTRUM.IEEE.ORG

BOSTON DYNAMICS & HYUNDAI: LET’S DANCE  A newly acquired robotics giant finds its footing and cuts a rug The robotics company Bosto n D y n a m i c s re g u l a rly attracts millions of YouTube views with videos depicting its robots performing remarkable feats of agility. These extremely popular shorts typically depict Spot (a quadrupedal robot the size of a large dog) or Atlas (a humanoid robot) performing menial work tasks—including moving boxes and other items around a warehouselike environment—as well as acrobatic maneuvers that would challenge even a nimble human. Founded in 1992 as a spinoff of MIT, Boston Dynamics spent decades working on bespoke projects for government agencies like the Defense Advanced Research Projects Agency before Google acquired it in 2013. In 2017, Google sold Boston Dynamics to Japan’s SoftBank Group. This past December, Hyundai Motor Group acquired Boston Dynamics in a deal that values the robotics company at US $1.1 billion. 

Around the same time the South Korean automaker’s purchase was announced to the public, Boston Dynamics released one of its most popular videos yet. In it, Spot, Atlas, and a hybrid robot with legs and wheels called Handle dance to the Contours’ 1962 hit song “Do You Love Me”—clocking an impressive 27 million views in just its first month online. This latest chronicle of robotic prowess demonstrates a charming side to the company’s clear engineering brilliance. And while the robots make it look effortless, it took a talented and dedicated team of humans to get them there. Boston Dynamics engineers collaborated with a group of professional dancers, led by Boston-based choreographer Monica Thomas, to create an initial concept for the dance by composing individual dance moves and assembling them into a routine. But transferring that routine onto Boston Dynamics’ robots took months of work, and even some BOSTON DYNAMICS

According to one recent estimate, data centers consume 2 percent of the world’s electricity, a figure that’s expected to climb to 8 percent by 2030. To turn back the age of the grid-greedy CPU, a group of researchers in Japan has developed a superconducting microprocessor— one with zero electrical resistance. The new device, the first of its kind, is described in a study published in December in the IEEE Journal of SolidState Circuits. The research group sought to create a superconducting microprocessor that’s adiabatic—meaning that, in principle, energy is neither gained nor lost from the system during the computing process. The device is composed of superconducting niobium and relies on hardware components called adiabatic quantum-flux-parametrons (AQFPs). “The data-processing part of the microprocessor can operate up to a clock frequency of 2.5 gigahertz, making this on par with today’s computing technologies,” says Christopher L. Ayala, an associate professor at the Institute of Advanced Sciences at Yokohama National University, in Japan, who helped develop the new microprocessor. “Even when taking [the] cooling overhead into account,” Ayala says, “the AQFP is still about 80 times more energy efficient when compared to the state-of-the-art semiconductor electronic device, [such as the] 7-nanometer FinFET [fin field-effect transistor], available today.”

TERPSICHORE WITH ACTUATORS: Newly acquired by Hyundai, Boston Dynamics released a highly popular video featuring creatively choreographed dancing robots.

hardware upgrades, as Aaron ­Saunders, Boston Dynamics’ vice president of engineering and Atlas program team lead, explained when we interviewed him shortly after the video’s release. IEEE Spectrum: What was the most challenging part of teaching your robots to dance? Aaron Saunders: One of the challenges, and probably the core challenge for Atlas in particular, was adjusting human dance moves so that they could be performed on the robot. To do that, we used simulation to rapidly iterate through movement concepts while soliciting feedback from the choreographer to reach behaviors that Atlas had the strength and speed to execute. It was very iterative—[the troupe] would literally dance out what

they wanted us to do, and the engineers would look at the screen and go, “That would be easy” or “That would be hard” or “That scares me.” And then we’d have a discussion, try different things in simulation, and make adjustments to find a compatible set of moves that we could execute on Atlas. Throughout the project, the time frame for creating new dance moves got shorter and shorter as we built new tools. Eventually we were able to use that tool chain to create one of Atlas’s ballet moves in just a single day, the day before we filmed. And it worked! Spectrum: Were there some moves that were particularly difficult to translate from human dancers to Atlas? Saunders: Some of the spinning turns in the ballet parts took more iterations to get to work. They were the furthest from leaping and running and some of the other things that we have more experience with, so they challenged both the machine and

the software in new ways. We definitely learned not to underestimate how flexible and strong dancers are—when you take elite athletes and you try to do what they do but with a robot, it’s a hard problem. It’s humbling. Fundamentally, I don’t think that Atlas has the range of motion or power that these athletes do, although we continue developing our robots toward that. We believe that in order to broadly deploy these kinds of robots commercially, and eventually in a home, they need to have this level of performance. Spectrum: How were you able to create dance moves for robots that have very different form factors from humans? Saunders: The dancers and choreographer who we worked with had a lot of talent for thinking about motion, and thinking about how to express themselves through motion. And our robots do motion really well. They’re dynamic, they’re exciting, they balance. What we found was that the dancers connected

ROBOTIC END - EFFECTORS

Measure all six components of force and torque in a compact, rugged sensor. Interface Structure

high-strength alloy provides IP60, IP65, and IP68 environmental protection as needed

Sensing Beams and Flexures

designed for high stiffness and overload protection without compromising resolution

High-Speed Electronics

interfaces for Ethernet, PROFINET, EtherNet/IP, Analog, USB, CAN EtherCAT, Wireless, and more

Silicon Strain Gages

provide high noise immunity, accuracy, and high factor-of-safety, standard on all F/T models

Engineered for high-performance and maximum stiffness, with the highest resolution and accuracy available, it’s the ultimate force/torque sensor. Only from ATI. www.ati-ia.com 919.772.0115

SPECTRUM.IEEE.ORG | MAR 2021 |  09

with the way the robots moved and then shaped that into a story, and it didn’t matter whether there were two legs or four legs or wheels. When you don’t necessarily have a template of animal motion or human behavior, you just have to think a little harder about how to go about doing something, and that’s true for more pragmatic commercial behaviors as well. Spectrum: Are the videos that Boston Dynamics creates accurate representa-

tions of the general capabilities of your robots? Saunders: We work hard to make something, and once it works, it works. For Atlas, most of the robot control existed from our previous work, like the work that we’ve done on ­[ obstacle-course-inspired] parkour, which sent us down a path of using model predictive controllers that account for dynamics and balance. We used those to run on the robot a set of dance steps that we’d designed offline

IEEE MEMBER DISCOUNTS Exclusive offers are another big plus to IEEE Membership. In addition to helping you stay abreast of the latest developments, network with your peers, and gain access to high-quality resources, an IEEE Membership will also give you exclusive offers and discounts on products and services from many trusted brands. Visit www.ieee.org/discounts for Member savings and country-specific insurance. See the Member Discounts Marketplace for new, time-limited deals.

+ +

+ ++ +++ + ++ + + + ++ 10  | MAR 2021 | SPECTRUM.IEEE.ORG

19-FA-032 01/20

+

+

with the dancers and choreographer. Dancing required a lot of strength and speed, so we even upgraded some of Atlas’s hardware to give it more power. Dance might be the highest-power thing we’ve done to date; even though you might think parkour looks way more explosive, the amount of motion and speed that you have in dance is incredible. That also took a lot of time over the course of months—creating the capability in the machine to go along with the capability in the algorithms. There were definitely some failures in the hardware that required maintenance, and our robots stumbled and fell down sometimes. These behaviors are not meant to be productized and to be 100 percent reliable, but they’re definitely repeatable. We try to be honest with showing things that we can do, not a snippet of something that we did once. I think there’s an honesty required in saying that you’ve achieved something, and that’s definitely important for us. Spectrum: Has the experience of teaching robots to dance informed your approach to robotics for commercial applications? Saunders: We think that the skills inherent in dance and parkour, like agility, balance, and perception, are fundamental to a wide variety of robot applications. When you push limits by asking your robots to do these dynamic motions over a period of several days, you learn a lot about the robustness of your hardware. Spot, through its productization, has become incredibly robust and required almost no maintenance. It could just dance all day long once you taught it to. And the reason it’s so robust today is because of all those lessons we learned from previous things that may have just seemed weird and fun. You’ve got to go into uncharted territory to even know what you don’t know. Maybe more importantly, finding that intersection between building a new robot capability and having fun has been Boston Dynamics’ recipe for robotics—it’s a great way to advance. —Evan Ackerman POST YOUR COMMENTS AT spectrum.ieee.org/robotsdance-mar2021

CONTRIBUTORS_ CONTRIBUTORS EDITOR IN CHIEF

Susan Hassler, [email protected] Glenn Zorpette, [email protected]

ADVERTISING PRODUCTION

+1 732 562 6334

EXECUTIVE EDITOR

ADVERTISING PRODUCTION MANAGER

EDITORIAL DIRECTOR, DIGITAL

Felicia Spagnoli, [email protected]

Harry Goldstein, [email protected] MANAGING EDITOR Elizabeth A. Bretz, [email protected]

Nicole Evans Gyimah, [email protected]

SENIOR ADVERTISING PRODUCTION COORDINATOR

SENIOR ART DIRECTOR

EDITORIAL ADVISORY BOARD, IEEE SPECTRUM

Mark Montgomery, [email protected] PRODUCT MANAGER, DIGITAL Erico Guizzo, [email protected]

Susan Hassler, Chair; David C. Brock, Robert N. Charette, Ronald F. DeMara, Shahin Farshchi, Lawrence O. Hall, Jason K. Hui, Leah Jamieson, Mary Lou Jepsen, Deepa Kundur, Peter Luh, Michel Maharbiz, Somdeb Majumdar, Allison Marsh, Carmen Menoni, Sofia Olhede, Wen Tong, Maurizio Vecchione

SENIOR EDITORS

Evan Ackerman (Digital), [email protected] Stephen Cass (Special Projects), [email protected] Jean Kumagai, [email protected] Samuel K. Moore, [email protected] Tekla S. Perry, [email protected] Philip E. Ross, [email protected] David Schneider, [email protected] Eliza Strickland, [email protected] DEPUTY ART DIRECTOR Brandon Palacio, [email protected] PHOTOGRAPHY DIRECTOR Randi Klett, [email protected] ONLINE ART DIRECTOR Erik Vrielink, [email protected] NEWS MANAGER Mark Anderson, [email protected]

ADMINISTRATIVE ASSISTANT

Ramona L. Foster, [email protected] CONTRIBUTING EDITORS Robert N. Charette, Steven Cherry, Charles Q. Choi, Peter Fairley, Maria Gallucci, W. Wayt Gibbs, Mark Harris, Jeremy Hsu, Allison Marsh, Prachi Patel, Megan Scudellari, Lawrence Ulrich, Emily Waltz EDITOR IN CHIEF, THE INSTITUTE

Kathy Pretz, [email protected] ASSISTANT EDITOR, THE INSTITUTE

Joanna Goodrich, [email protected] Peter Tuohy Michael Spector ASSOCIATE ART DIRECTOR, PUBLICATIONS Gail A. Schnitzer DIRECTOR, PERIODICALS PRODUCTION SERVICES MULTIMEDIA PRODUCTION SPECIALIST

IEEE  BOARD OF DIRECTORS

Susan K. “Kathy” Land, [email protected]

+1 732 562 3928 FAX: +1 732 981 9515 PRESIDENT-ELECT TREASURER SECRETARY

K.J. Ray Liu

Mary Ellen Randall Kathleen A. Kramer

PAST PRESIDENT

Toshio Fukuda

VICE PRESIDENTS

Stephen M. Phillips, Educational Activities; Lawrence O. Hall, Publication Services & Products; Maike Luiken, Member & Geographic Activities; Roger U. Fujii, Technical Activities; James E. Matthews, President, Standards Association; Katherine J. Duncan, President, IEEE-USA DIVISION DIRECTORS

Alfred E. “Al” Dunlop (I); Ruth A. Dyer (II); Sergio Benedetto (III); Manfred “Fred” J. Schindler (IV); Thomas M. Conte (V); Paul M. Cunningham (VI); Miriam P. Sanders (VII); Christina M. Schober (VIII); Rabab Kreidieh Ward (IX); Dalma Novak (X) REGION DIRECTORS

Eduardo F. Palacio (1); Barry C. Tilton (2); Jill I. Gostin (3); Johnson A. Asumadu (4); James R. Look (5); Timothy T. Lee (6); Jason Jianjun Gu (7); Antonio Luque (8); Alberto Sanchez (9); Deepak Mathur (10) DIRECTOR EMERITUS

MANAGING DIRECTOR, PUBLICATIONS

Steven Heffner

EDITORIAL CORRESPONDENCE

IEEE Spectrum, 3 Park Ave., 17th Floor, New York, NY 10016-5997 TEL: +1 212 419 7555 FAX: +1 212 419 7570

ASSOCIATE EDITORS

Willie D. Jones (Digital), [email protected] Michael Koziol, [email protected] SENIOR COPY EDITOR Joseph N. Levine, [email protected] COPY EDITOR Michele Kogon, [email protected] EDITORIAL RESEARCHER Alan Gardner, [email protected]

PRESIDENT & CEO

EDITORIAL ADVISORY BOARD, THE INSTITUTE

Kathy Pretz, Chair; Qusi Alqarqaz, Philip Chen, Shashank Gaur, Lawrence O. Hall, Susan Hassler, Peter Luh, Cecilia Metra, San Murugesan, Mirela Sechi Annoni Notare, Joel Trussell, Hon K. Tsang, Chenyang Xu

Theodore W. Hissey

IEEE STAFF EXECUTIVE DIRECTOR & COO Stephen Welby +1 732 562 5400, [email protected] CHIEF INFORMATION OFFICER Cherif Amirat +1 732 562 6017, [email protected]

BUREAU

Palo Alto, Calif.; Tekla S. Perry +1 650 752 6661

DIRECTOR, BUSINESS DEVELOPMENT, MEDIA & ADVERTISING

Mark David, [email protected]

Naylor Association Solutions, Erik Henson +1 352 333 3443, [email protected] ADVERTISING INQUIRIES

+1 212 221 9595, ext. 319 Articles may be photocopied for private use of patrons. A per-copy fee must be paid to the Copyright Clearance Center, 29 Congress St., Salem, MA 01970. For other copying or republication, contact Managing Editor, IEEE Spectrum. REPRINT SALES

REPRINT PERMISSION / LIBRARIES

IEEE Spectrum is a registered trademark owned by The Institute of Electrical and Electronics Engineers Inc. Responsibility for the substance of articles rests upon the authors, not IEEE, its organizational units, or its members. Articles do not represent official positions of IEEE. Readers may post comments online; comments may be excerpted for publication. IEEE reserves the right to reject any advertising. COPYRIGHTS AND TRADEMARKS

PUBLICATIONS Steven Heffner +1 212 705 8958, [email protected] CHIEF MARKETING OFFICER Karen L. Hawkins +1 732 562 3964, [email protected] CORPORATE ACTIVITIES Donna Hourican +1 732 562 6330, [email protected] MEMBER & GEOGRAPHIC ACTIVITIES Cecelia Jankowski +1 732 562 5504, [email protected] STANDARDS ACTIVITIES Konstantinos Karachalios +1 732 562 3820, [email protected] EDUCATIONAL ACTIVITIES Jamie Moesch +1 732 562 5514, [email protected] GENERAL COUNSEL & CHIEF COMPLIANCE OFFICER

Sophia A. Muirhead +1 212 705 8950, [email protected] CHIEF FINANCIAL OFFICER Thomas R. Siegert +1 732 562 6843, [email protected] TECHNICAL ACTIVITIES Mary Ward-Callan +1 732 562 3850, [email protected] MANAGING DIRECTOR, IEEE-USA Chris Brantley +1 202 530 8349, [email protected] IEEE  PUBLICATION SERVICES & PRODUCTS BOARD

Lawrence O. Hall, Chair; Sergio Benedetto, Edhem Custovic, Stefano Galli, James Irvine, Clem Karl, Hulya Kirkici, Fabrizio Lombardi, Aleksandar Mastilovic, Sorel Reisman, Gaurav Sharma, Isabel Trancoso, Maria Elena Valcher, Peter Winzer, Bin Zhao IEEE OPERATIONS CENTER

445 Hoes Lane, Box 1331 Piscataway, NJ 08854-1331 U.S.A. Tel: +1 732 981 0060 Fax: +1 732 981 1721

John Boyd Within 48 hours of the earthquake that led to the Fukushima Daiichi nuclear disaster, Boyd, a longtime Japan-based contributor, was filing the first of a series of stories for IEEE Spectrum documenting efforts to contain the radioactivity. Ten years on, Boyd reports on the region’s renewable energy renaissance [p. 6]. “It’s heartening to see Fukushima prefecture recovering,” he says, “though the impact of the accident is still very much evident.”

Joel Brand Brand is vice president of product management at Kumu Networks, in Sunnyvale, Calif. In “The Radio That Can Hear Over Itself” [p. 30], he explains how Kumu and other startups are developing radios capable of self-interference cancellation, which allows the radios to transmit and receive signals on the same frequency at the same time. Brand says that improvements in spectral efficiency have slowed recently, but self-interference cancellation “is something that provides a massive boost.”

Sharath Rao Rao is director of machine learning at Instacart. He applied machine learning to online advertising and search at Yahoo and developed strategies for pricing mortgage derivatives at Morgan Stanley. With Lily Zhang, Instacart’s director of software engineering, Rao explains how Instacart uses machine learning and apps to get groceries from shelves to doorsteps [p. 36]. Rao dreams of someday building an app that counts calories just from a photo of a meal.

Peter Sanitra Sanitra and Marek Denko run the NoEmotion studio in Prague, where they use computer graphics to create images for movies, video games, advertisements, and magazines. Sanitra, who has an interest in space exploration, helped illustrate “How We’ll Find Another Earth” [p. 22] as well as the cover image. He started by fabricating many different virtual worlds, including “water and continents, dry planets covered with dust clouds, and ice planets with very sparse clouds.”

Jason T. Wright Wright, an astronomer at Pennsylvania State University, and Cullen Blake of the University of Pennsylvania, hunt exoplanets, work they describe in “How We’ll Find Another Earth” [p. 22]. Perhaps the strangest exoplanet observations that Wright has been involved with concerned “Tabby’s Star,” which some experts thought might be surrounded by alien technology. “It became a sensation,” says Wright, who notes that news of the star was even reported on “Saturday Night Live.”

SPECTRUM.IEEE.ORG | MAR 2021 |  11

SOCIALLY DISTANCED CELEBRATION LONDON’S NEW YEAR’S EVE FIREWORKS display, initiated as a special event to ring in the year 2000, normally attracts about 3 million people to the edge of the River Thames. As the calendar turned to 2021, a virtual audience watched from home because of COVID-19. The made-for-TV event let viewers witness festivities at locations around the city, including the O2 entertainment complex, Tower Bridge, and Wembley Stadium.

NEWS 12 | MAR 2021 | SPECTRUM.IEEE.ORG

THE BIG PICTURE PHOTOGRAPH BY

Victoria Jones/PA Images/Getty Images

SPECTRUM.IEEE.ORG | MAR 2021 | 13

Hands On

HANDS ON BY STEPHEN CASS 14  | MAR 2021 | SPECTRUM.IEEE.ORG

ILLUSTRATIONS BY

James Provost

HACK THIS $50 WATCH GET BACK INTO MOBILE LIFE WITH THIS ARDUINOCOMPATIBLE SMARTWATCH Like many of you, for the past year I’ve led a pretty static existence living and working in my apartment. But with the rollout of COVID-19 vaccinations, I can actually think of returning to a more ambulatory way of life. Which also means that I’m thinking about mobile devices for the first time in a long time, making it just the right moment to spot the recent release of the Watchy, an open-source, programmable, wireless-enabled wristwatch with a lowpower display. Could I use it to create some personalized electronics for when I go back out and about? With its gray-tinted screen, ­Squarofumi’s Watchy inevitably conjures echoes of the Pebble smartwatch, which made a huge splash in 2012 when it raised over US $10 million on Kickstarter. Pebble ­ultimately had its lunch eaten by Apple and others, but Watchy is different in a few key respects: It is not trying to be a mass-­market

TIME PIECES: The Watchy is quickly assembled, and held together with some squares of adhesive tape. Slots on the central PCB are provided to thread a fabric watch strap through, but you may wish to 3D-print a case that allows standard watch straps to be used; if so, I recommend not taping the battery down to make it easier to position in the case.

device. It is unashamedly for those willing to tangle with code. It’s also inexpensive— just $50 versus the Pebble’s $150, let alone the Apple Watch’s $400 price tag. Watchy is based around an ESP32 microcontroller, a popular alternative to AVRor Arm-based microcontrollers because of its built-in Wi-Fi and Bluetooth capabilities that can be programmed via the Arduino IDE.

Surrounding hardware includes a 1.5-inch e-paper display, a real-time clock module, a vibration motor, a three-axis accelerometer, and four control buttons. Assembling the Watchy takes little time. It comes in just four components: a fully populated printed circuit board, a 200 ­milliampere-hour lithium polymer battery, the display, and a fabric wristband.

DEPARTMENTS SPECTRUM.IEEE.ORG | MAR 2021 |  15

HANDS ON BY STEPHEN CASS

FACE FORWARD: Here’s the custom ogham-based face I created for the Watchy [upper left], along with some of the sample watch faces included with the Watchy software. Time can be displayed using either complete fonts or bitmaps for the numerals. The ogham reads 12:33.

­ dhesive tape keeps the screen and battery A in place. A microUSB socket charges the battery and provides the link for uploading programs for new watch faces. Once I had everything put together, I followed Squarofumi’s instructions to install compiler support for the ESP32 along with the Watchy library and example face code. I soon ran into my first problem—none of the sample code would compile. A little poking around online revealed that the macOS version of the Arduino IDE currently has a compatibility problem with the ESP board. However, I was able to grab the latest ­release candidate for the next version of the ESP ­library, and all was well. At least until I tried to upload a face to the Watchy. Despite much fiddling, I could not get a response when I tried storing code in the Watchy’s flash memory. Wondering if this was another macOS problem, I fired up ­Windows 10 on my iMac, but no joy. Querying the official support forum on GitHub got a suggestion from Squarofumi that I test 16  | MAR 2021 | SPECTRUM.IEEE.ORG

the connection using the “esptools” Python ­library by erasing the flash directly, but this also produced a negative result. As I sagged, frustrated in my chair, my eye fell on one of the Raspberry Pi’s floating around my overcrowded desk, and on a hunch, I plugged my Watchy into it. I installed esptools on my Pi—and it recognized and wiped the Watchy’s memory! With that done, I plugged the Pi into my iMac and—miracles— uploaded a program without problem from the Arduino IDE. So now, every time I want to upload new code I have to run it through my Pi first, in a manner reminiscent of an oldschool EPROM ultraviolet eraser. As of this writing, Squarofumi is still working out why I have to perform this ritual of obliteration, but fingers crossed, there should be a better solution soon. From there, I put my hand to creating my own watch face. Unfortunately, there are quite a few “coming soon” pages under the Docs section of the Watchy website, but between what is available and the cleanly writ-

ten code samples, it didn’t take me too long to get the hang of how to code a new face. Arduino programs are automatically structured to call first a setup function, and then a main loop function, and typically the bulk of the work is done in the latter. Watchy programs are unusual in that the loop function is left empty, and the setup function is where the action happens. This structure allows the Watchy to conserve power by sleeping most of the time, typically waking up only to ­refresh once per minute to trigger functions and update the display, although faster updates are possible. The Watchy can display gray-scale fonts and bitmaps up to 200 by 200 pixels. I decided to display the time in a way that no conventional numerical display could—­using ogham, a centuries-old alphabet used to inscribe the Irish language vertically on the edges of stone monuments. Rambling in the southwest of Ireland in younger days, I would often come across these monoliths with their enigmatic markings, so I associate them with the feel of a good long walk. As ogham doesn’t have numerals, I used an online ogham transliterator, to make images of the words “zero” through “nine,” and converted them to ­Arduino arrays using another online tool recommended in the Watchy docs. Each array represents a narrow bitmap 50 pixels wide, so I could tile them four across to represent hours and minutes. The result was surprisingly effective (and has had the side effect that I can now sight-read a lot of ogham letters for the first time!). Watchy faces can be expanded beyond simply telling the time: The wireless connection and a JSON parser library allow you to extract information from online services about things such as the weather or subway arrival times, and the accelerometer allows for gesture controls. So once I do ­finally get vaccinated, I will have my personalized mobile tech to remind me it’s nice to go ­outdoors. —STEPHEN CASS POST YOUR COMMENTS AT spectrum.ieee.org/watchy-mar2021

PROOF 3 CLOSED

5/15/20 @ 11:27 am

EV INT

Please return to:

Careers

JNL MK by

JEANNETTE PLANTE

SAVING NASA MISSIONS BEFORE THEY GO WRONG

NASA

NASA’s spacecraft have to function with exquisite precision in extraordinarily harsh envi­ronments. An essential part of making that happen is rigorous quality assurance on the systems that go into these spacecraft. ­Jeanette Plante, a quality-engineering technical fellow in NASA’s Office of Safety and ­Mission Assurance, is one of the people charged with ensuring that rigor, which goes far beyond the traditional concept of qualityassurance teams performing inspections as items are assembled. The work at NASA “is less about inspections than you might think,” says Plante. “A lot of our work goes into deciding what we want in the first place and what measures we will use to determine conformance. Getting the requirements right first and building in quality is critical.” Another wrinkle is that many of NASA’s missions are tied to timelines that are as unalterable as the motions of the planets themselves. “Quality assurance [for] missions has to not only include tests that address failure modes but also fit within often-inflexible schedules for testing-facility availability and launch windows,” says Plante, who also notes, “A lot of the hardware we build or buy can’t be repaired while in use [while] terrestrially based systems might be maintained, replaced, or repaired throughout their service life.” Plante came to NASA through “a simple twist of fate” rather than an interest in space

or rocketry. “When I graduated in 1987 with my bachelor’s in electrical engineering, I had no idea where that would lead me,” recalls Plante. “I looked through newspaper and magazine job ads. And I also went to my college career center.” That’s where Plante found out about a contract job at NASA, doing electrical, e ­ lectronic, electromechanical, and electro-optic part assurance at NASA Goddard Space Flight Center in Greenbelt, Md. In the nearly three and a half decades since, Plante has gone from contractor to employee, and risen to workmanship standards program manager, to division chief for quality and reliability at Goddard, to her present position as technical fellow, working at NASA headquarters in Washington, D.C. “I’ve learned about manufacturing, materials engineering, how contracts and procurements work, relationships in the supply chain, and the physics of failure,” says Plante. “I expanded my knowledge from electronic parts to assemblies,

and then to supply chains, and found my way to having a view of all hardware manufacturing.” The goal, explains Plante, is to examine “all the aspects of how the hardware is designed and built, including what kinds of process controls will go into building a finished product.” NASA projects that Plante did QA for include the X-Ray Spectrometer (XRS), the Mercury Laser Altimeter (MLA), the Tracking and Data Relay Satellite (TDRS), and the James Webb Space Telescope (JWST). Currently, says Plante, “I develop policy for quality assurance for all of our missions, particularly for mission hardware—writing the p ­ olicy as well as creating a toolbox of methods and techniques that can be used by projects for QA.” For any engineers who might be interested in QA, Plante has some advice. ­“Industrial engineering can be a foot in the door [but also] take some formal courses in statistics, in quality, and in manufacturing control. And get fluent with new technologies and methods. For example, NASA is doing more with additive manufacturing, augmented ­reality, and digital twins [where a precise digital replica of an object or process is created]... and all of these either need or advance quality assurance.” To work at NASA specifically, “getting experience in a manufacturing environment would be an asset,” she says “You also need resilience and patience,” adds Plante. “If the problem wasn’t hard, they wouldn’t be calling on engineers to solve it.” One final job-finding tip: “Don’t just look within NASA—also look at the NASA supply chain,” Plante suggests. “Most of our hardware is made by private industry.” —DANIEL P. DERN POST YOUR COMMENTS AT spectrum.ieee.org/ plante-mar2021

DEPARTMENTS SPECTRUM.IEEE.ORG | MAR 2021 |  17

CrossTalk THE TOMATO’S ENERGY FOOTPRINT THE WORLD’S FOOD SUPPLY is now unimaginable without the tomato, which although often called a vegetable is actually a fruit of the Solanum lyco­persicum species, native to Central and South America. It gradually diffused through the world after 1500, finally becoming the world’s largest “vegetable” crop, with recent harvests around 180 million

metric tons a year, nearly twice that of dry onions, five times that of peppers, and more than six times that of cauliflower and broccoli. The tomato’s versatility explains its rise: eaten unadorned when fully ripe, as part of countless salads, made into purees ­(Italian ­passata), soups and sauces, sliced, chopped, baked, fried, and dried.

Given about 90 days of warm weather, growing tomatoes in your garden isn’t too hard. When cultivated in modern intensive ways, it requires fuels and electricity to make synthetic fertilizers and pesticides and to produce and power the machinery and supplementary irrigation. Field tomatoes need as little as 0.8 megajoule (about 190 ­kilocalories) per harvested kilogram. A typical tomato, bought at retail, weighs about 125 grams, of which 95 percent is water and most of the rest is carbohydrate and fiber. Its nutritional value lies in its vitamin C and A content. A tomato of that size contains only about 22 kcal (or 22 calories, in the

NUMBERS DON’T LIE BY VACLAV SMIL 18  | MAR 2021 | SPECTRUM.IEEE.ORG

PHOTOGRAPH BY

Dan Saelinger

colloquial but incorrect usage that is common in the United States). That is just slightly less than the commercial energy, in fuel and electricity, that is needed to produce it. An energy economist might call them 1:1 tomatoes. But, increasingly, store-bought tomatoes are grown not in fields but in greenhouses, usually hydroponically—without soil—either under glass or in long, plasticcovered tunnels. Here the energy cost of production is substantially higher. Direct energy inputs include electricity, gasoline, and diesel fuel; indirect energy costs involve the production of fertilizer (particularly nitrogen), pesticides, fungicides, and plastics and metals (not only for covers but also for cultivation troughs and copious pipes and heaters). In the United States, tomato greenhouses are concentrated in California, Minnesota, and New York. The world’s largest concentration of plastic green-

houses lies in the southernmost part of Spain’s Almería province; you can see it on satellite images, and you can even move between some of them using ­dystopic Google street views. A 125-gram tomato of Almería that is grown in an unheated plastic tunnel requires about 150 kcal; one grown in heated structures, about 560 kcal. It provides about 22 kcal of food energy, making the two kinds approximately 1:7 and 1:25 tomatoes. And with economies of scale favoring large-scale centralized production and long-distance shipping, the required storage, packing, and trucking to a regional distribution center raises the total energy cost to 460–875 kcal/125 g, raising the ratio to between 1:21 and 1:40. And because these tomatoes are grown in the relatively warm Mediterranean climate, they are far from being the world’s most energy intensive: Tomatoes produced

in heated greenhouses in many European countries require 40 to 150 times as much commercial energy input as they yield in edible food energy. Perhaps the best way to convey the energy cost of a 1:60 tomato is to express it in equivalent terms, as tablespoons (each equal to 14.8 milliliters) of diesel fuel to be poured over a sliced tomato instead of the classic oil-and-vinegar dressing. For a 125-gram tomato it would come to 10 tablespoons. For a family-size salad, requiring a kilogram of those regularly sized greenhouse tomatoes (cluster of eight, still attached to a strong green vine), you’d incur an energy production cost equivalent to 80 tablespoons, or 5 cups of diesel fuel. This is a perfect illustration of how our food production and distribution depend heavily on substantial fuel and electricity inputs! n POST YOUR COMMENTS AT spectrum.ieee.org/tomato-mar2021

U.S. SAYS TOMAY-TO, WORLD SAYS TOMAH-TO This luscious fruit originated in the Americas, then transformed the culinary practices of the world. It requires warmth, light, and fertilizer, often from artificial sources, which is why each tomato on your plate represents a great investment of energy.

U.S. TOMATO HARVEST

WORLDWIDE TOMATO HARVEST

Source: Food and Agriculture Organization of the United Nations Statistics Division (FAOSTAT)

CROSSTALK SPECTRUM.IEEE.ORG | MAR 2021 |  19

INTERNET OF EVERYTHING BY STACEY HIGGINBOTHAM

SMARTER SMART CITIES SMART CITIES, LIKE many things, took a beati n g i n 2 02 0. Alphabet, Google’s parent company, pulled its Sidewalk Labs subsidiary out of a smart-city project in Toronto. Cisco killed its plans to sell smart-city technology. And in many places, city budgets will be affected for years to come by the pandemic’s economic shock, making it more difficult to justify smart-city expenses. That said, the pandemic also provided municipalities around the world with reason to invest new technologies for public transportation, contact tracing, and enforcing social distancing. In a way, the present moment is an ideal time for a new understanding of smartcity infrastructure and a new way of paying for it. Cities need to think of their smart-city efforts as infrastructure, like roads and sewers, and as such, they need to think 20  | MAR 2021 | SPECTRUM.IEEE.ORG

about investing in it, owning it, maintaining it, and controlling how it’s used in the same ways as they do for other infrastructure. Smart-city deployments affect the citizenry, and citizens will have a lot to say about any implementation. The process of including that feedback and respecting citizens’ rights means that cities should own the procurement process and control the implementation. In some cases, citizen backlash can kill a project, such as the backlash against Sidewalk’s Toronto project over who exactly had access to the data collected by the smart-city infrastructure. Even when cities do get permission from citizens for deployments, the end results are often neighborhood-size “innovation zones” that are little more than glorified test beds. A truly smart city needs a master plan, citizen accountability, and a means of funding that grants the city ownership.

CROSSTALK

One way to do this would be for cities to create public authorities, just like they do when investing in public transportation or even health care. These authorities should have publicly appointed commissioners who manage and operate the sensors and services included in smartcity projects. They would also have the ability to raise funds using bond issues backed by the revenue created by smartcity implementation. For example, consider a public safety project that requires sensors at intersections to reduce collisions. The project might use the gathered data to meet its own safety goals, but the insights derived from analyzing traffic patterns could also be sold to taxi companies or logistics providers. These sales will underpin the repayment on bonds issued to pay for the technology’s deployment and management. While some municipal bonds mature in 10- to 30-year time frames, there are also bonds with 3- to 5-year terms that would be better suited to the shorter life spans of technologies like traffic-light sensors. Even if bonds and public authorities aren’t the right way to proceed, owning and controlling the infrastructure has other advantages. Smart-city contracts could employ local contractors and act not just as a source of revenue for cities but also as an economic development tool that can create jobs and a halo effect to draw in new companies and residents. For decades, cities have invested in their infrastructure using public debt. If cities invest in smart-city technologies the same way, they could give their citizens a bigger stake in the process, create new streams of revenue for the city, and improve quality of life. After all, people deserve to live in smarter cities, rather than innovation zones. n POST YOUR COMMENTS AT spectrum.ieee.org/smartcities-mar2021

ILLUSTRATION BY

Greg Mably

MACRO & MICRO BY MARK PESCE

CROSSTALK

SPOTTY CONNECTIONS QR (QUICK RESPONSE) codes have undergone something of a renaissance during the pandemic: Where I live in Australia, we use them to check in at a restaurant or cinema or classroom. They anchor the digital record of our comings and goings to better assist contact tracers. It’s annoying to have to check into such venues this way—and even more annoying when it doesn’t work. Not long ago, I went to a local café, only to have the check-in process fail. I thought the problem might be my smartphone, so my companion gave it a go, using a different phone, OS, and carrier. No luck. We later learned the entire check-in infrastructure for our state had gone down. Millions of people were similarly vexed—I would argue completely needlessly. Nearly all the apps on our smartphones— and on our desktop computers—rest on networked foundations. Over the last ILLUSTRATION BY

Dan Page

generation, network access has become ubiquitous, useful, cheap, and stable. While such connectivity has obvious benefits, it seems to have simultaneously encouraged a form of shortsightedness among app developers, who find it hard to imagine that our phones might ever be disconnected. Before the arrival of the smartphone, we encountered connected devices only rarely. Now they number in the tens of billions—each designed around a persistently available network. Cut off from the network, these devices usually fail completely, as the contact-tracing app did for a time for so many Australians. People who develop firmware or apps for connected devices tend to write code for two eventualities: One assumes perfect connectivity; the other assumes no connectivity at all. Yet our devices nearly always reside in a gray zone. Sometimes things work perfectly,

sometimes they don’t work at all, and sometimes they only work intermittently. It’s that third situation we need to keep front of mind. I could have checked into my café, only to have the data documenting my visit uploaded later (after some panicked IT administrator had located and rebooted the failed server). But the developers of this system weren’t thinking in those terms. They should have taken better care to design an app that could fail gracefully, retaining as much utility as possible, while patiently waiting for a network connection to be reestablished. Later that same day, I tried to summon my smartphone-app-based loyalty card when checking out at the supermarket. The app wouldn’t launch. “Yeah,” the teenage cashier said, “The mobile signal is horrible in here.” “But the app should work, even with a crappy signal,” I insisted. “Everything’s connected,” he counseled, leaving it there. Like fish in water, we now swim in a sea of our own connectivity. We can’t see it, except in those moments when the water suddenly recedes, leaving us flapping around on the bottom, gasping for breath. Typically, such episodes are not much more than a temporary annoyance, but there are times when they become much more serious—such as when a smartwatch detects atrial fibrillation or a sudden fall and needs to signal that the wearer is in distress. In a life-ordeath situation, an app needs to provide defense in depth to get its message out by every conceivable path: Bluetooth, mesh network—even an audio alarm. As engineers, we need to remember that wireless networking is seldom perfect and design for the shadowy world of the sometimes connected. Our aim should be to build devices that do as well as they possibly can, connected or not. n POST YOUR COMMENTS AT spectrum.ieee.org/spottynetwork-mar2021

SPECTRUM.IEEE.ORG | MAR 2021 |  21

Atomically precise sensors could detect a planet much like our own

EARTH 2.0: This artist’s rendition shows how an Earthlike exoplanet might appear.

By Jason Wright & Cullen Blake

GUTTER CREDIT GOES HERE

22  | MAR 2021

ILLUSTRATION BY

Peter Sanitra/NoEmotion

GAZING INTO THE DARK and seemingly endless night sky above, people have long wondered: Is there another world like ours out there, somewhere? Thanks to new sensors that we and other astronomers are developing, our generation may be the first to get an affirmative answer—and the earliest hints of another Earth could come as soon as this year. Astronomers have discovered thousands of exoplanets so far, and almost two dozen of them are roughly the size of our own planet and orbiting within the so-called habitable zone of their stars, where water might exist on the surface in liquid form. But none of those planets has been confirmed to be rocky, like Earth, and to circle a star like the sun. Still, there is every reason to expect that astronomers will yet detect such a planet in a nearby portion of the galaxy. So why haven’t they found it yet? We can sum up FUNNEL FOR LIGHT: The 3.5-meter the difficulty in three words: resolution, contrast, mirror of the WIYN telescope channels starlight into a glass fiber, which carries and mass. Imagine trying to spot Earth from hunit to the NEID spectrograph, located in dreds of light-years away. You would need a giant the basement of the observatory. telescope to resolve such a tiny blue dot sitting a mere 150 million kilometers (0.000016 light-year) from the sun. And Earth, at less than a billionth the brightness of the sun, would be hopelessly lost in the glare. Geneva and the European Southern Observatory For the same reasons, current observatories—and even the next is commissioning an instrument called ESPRESSO generation of space telescopes now being built—have no chance of ­(Echelle SPectrograph for Rocky Exoplanet and Stasnapping a photo of our exoplanetary twin. Even when such imagble Spectroscopic Observations), in Chile, to hunt ing does become possible, it will allow us to measure an exoplanet’s in the southern skies. size and orbital period, but not its mass, without which we cannot All three groups are integrating Nobel Prize–­ determine whether it is rocky like Earth or largely gaseous like Jupiter. winning technologies in novel ways to creFor all these reasons, astronomers will detect another Earth first by ate ­c utting- edge digit al spec t ro graphs of exploiting a tool that’s been used to detect exoplanets since the mid­femtometer-scale precision. Their shared goal is 1990s: spectroscopy. The tug of an orbiting exoplanet makes its star to achieve the most precise measurements of stel“wobble” ever so slightly, inducing a correspondingly slight and slow lar motions ever attempted. The race to discover shift in the spectrum of the star’s light. For a doppelgänger of Earth, that the elusive “Earth 2.0” is afoot. fluctuation is so subtle that we have to look for a displacement measuring just a few atoms across in the patterns of the rainbow of starlight fallUSING SPECTROSCOPY ing on a silicon detector in the heart of a superstable vacuum chamber. to pick up an exoplanet-induced Our group at Pennsylvania State University, part of a team funded wobble is an old idea, and the basic by NASA and the U.S. National Science Foundation (NSF), has conprinciple is easy to understand. As structed an instrument called NEID (short for NN-explore Exoplanet Earth circles the sun in a huge orbit, Investigations with Doppler spectroscopy), for deployment in Arizona it pulls the sun, which is more than 300,000 times to search the skies of the Northern Hemisphere. (The name NEID— as massive, into a far smaller “counter orbit,” in pronounced “NOO-id”—derives from an American Indian word meanwhich the sun moves at a mere 10 centimeters per ing “to see.”) A Yale University team has built a second instrument, second, about the speed of a box turtle. called EXPRES (for EXtreme PREcision Spectrometer), which will also Even such slow motions cause the wavelengths of operate in Arizona. Meanwhile, a third team led by the University of sunlight to shift—by less than one part per billion—

U

24  | MAR 2021 | SPECTRUM.IEEE.ORG

NATIONAL OPTICAL-INFRARED ASTRONOMY RESEARCH LABORATORY/KPNO/AURA/NSF

G

toward shorter, bluer wavelengths in the direction of its motion and toward longer, redder wavelengths in the other direction. Exoplanets induce similar Doppler shifts in the light from their host stars. The heavier the planet, the bigger the Doppler shift, making the planet easier to detect. In 1992, radio astronomers observing odd objects called pulsars used the timing of the radio signals they emit to infer that they hosted exoplanets. In 1995, Michel Mayor and Didier Queloz of the University of Geneva were using the Doppler wobbles in the light from ordinary stars to search for orbiting companions and found a giant planet orbiting the star 51 Pegasi every four days. This unexpected discovery, for which Mayor and Queloz won the 2019 Nobel Prize in Physics, opened the floodgates. The “Doppler wobble” method has since been used to detect more than 800 exoplanets, virtually all of them heavier than Earth. So far, almost all of the known exoplanets that are the size and mass of Earth or smaller were discovered using a different technique. As these planets orbit, they fortuitously pass between their star and our telescopes, causing the apparent brightness of

the star to dim very slightly for a time. Such transits, as astronomers call them, give away the planet’s diameter—and together with its wobble they will reveal the planet’s mass and density. Since the first transiting exoplanet was discovered in 1999, the Kepler space telescope and Transiting Exoplanet Survey Satellite (TESS) have found thousands more, and that number continues to grow. Most of these bodies are unlike any in our solar system. Some are giants, like Jupiter, orbiting so close to their host stars that they make the day side of Mercury look chilly. Others are gassy planets like Neptune or Uranus but much smaller, or rocky planets like Earth but much larger. Through statistical analysis of the bounteous results, astronomers infer that our own Milky Way galaxy must be home to billions of rocky planets like Earth that orbit within the habitable zones of their host stars. A study published recently by scientists who worked on the Kepler mission estimated that at least 7 percent—and probably more like half—of all sunlike stars in the Milky Way host potentially habitable worlds. That is why we are so confident that, with sufficiently sensitive instruments, it should be possible to detect other Earths. But to do so, astronomers will have to measure the mass, size, and orbital period of the planet, which means measuring both its Doppler wobble and its transit. The trove of data from Kepler and TESS—much of it yet to be analyzed—has the transit angle covered. Now it’s up to us and other astronomers to deliver higher-precision measurements of Doppler wobbles. N E I D , O U R C O N T E S T A N T in the race to find Earth 2.0, improves the precision of Doppler velocimetry by making advances on two fronts simultaneously: greater stability and better calibration. Our design builds on work done in the 2000s by European astronomers who built the High Accuracy Radial Velocity Planet Searcher (HARPS), which controlled temperature, vibration, and pressure so well that it faithfully tracked the subtle Doppler shifts of a star’s light over years without sacrificing precision. In 2010, we and other experts in this field gathered at Penn State to compare notes and discuss our aspirations for the future. HARPS had been producing blockbuster discoveries since 2002; U.S. efforts seemed to be lagging behind. The most recent decadal review by the U.S. National Academy of Sciences had recommended an “aggressive program” to regain the lead. Astronomers at the workshop drafted a letter to NASA and the NSF that led to them commissioning a new spectrograph, to be installed at the 3.5-meter WIYN telescope at Kitt Peak National Observatory, in Arizona. (“WIYN” is an acronym derived from the names of the four

N

SPECTRUM.IEEE.ORG | MAR 2021 |  25

founding institutions.) NEID would be built at Penn State by an international consortium. Like HARPS before it, NEID combines new technologies with novel methods to achieve unprecedented stability during long-term observations. Any tiny change in the instrument—whether in temperature or pressure, in the way starlight illuminates its optics, or even in the nearly imperceptible expansion of the instrument’s aluminum structure as it ages—can cause the dispersed starlight to creep across the detector. That could look indistinguishable from a Doppler shift and corrupt our observations. Because the wobbles we aim to measure show up as shifts of mere femtometers on the sensor—comparable to the size of the atoms that make up the detector—we have to hold everything incredibly steady. That starts with exquisite thermal control of the instrument, which is as big as a car. It has required us to employ a vacuum chamber with 26  | MAR 2021 | SPECTRUM.IEEE.ORG

precise thermal-feedback systems and also to design special electronics that prevent the variable heating that sensors of the type we use normally experience. Yet we know we cannot build a perfectly stable instrument. So we rely on calibration to take us the rest of the way. Previous instruments used specialized lamps containing thorium or uranium, which give off light at hundreds of distinct, well-known wavelengths—yardsticks for calibrating the spectral detectors. But these lamps change with age. NEID instead relies on a laser frequency comb: a Nobel Prize–winning technology that generates laser light at hundreds of evenly spaced frequency peaks. The precision of the comb, limited only by our best atomic clocks, is better than a few parts

CLOCKWISE FROM TOP: NATIONAL OPTICAL-INFRARED ASTRONOMY RESEARCH LABORATORY/ KPNO/AURA/NSF; GUÐMUNDUR KÁRI STEFÁNSSON; JOE NINAN

SKY-HIGH ADVANCE: The optical components of the NEID spectrograph [bottom left] are housed in a large vacuum chamber [bottom right]. NEID is a key addition to the Kitt Peak National Observatory, in Arizona [top].

930 nm

Frequency-comb calibration mark

S

Absorption feature

0.024 nm

SOMEWHERE OUT THERE, another Earth is orbiting its star, which wobbles slowly in sympathy around their shared center of gravity. Over the course of a year or so, the star’s spectrum shifts ever so slightly toward the red, then toward the blue. As that starlight reaches Kitt Peak, the telescope brings it to a sharp focus at the tip of a glass optical fiber, which guides the light down into the bowels of the observatory through a conduit into a custom-designed room housing the NEID instrument. The focus spot must strike the fiber in exactly the same way with every observation; any variations can translate into slightly different illumination patterns on the detector, mimicking stellar Doppler shifts. The quarter-century-old WIYN telescope was never designed for such work, so to ensure consistency our colleagues at the University of Wisconsin–Madison built an interface that monitors the focus and position of the stellar image hundreds of times a second and rapidly adjusts to hold the focal point steady. Down in the insulated ground-floor room, a powerful heating-and-cooling system keeps a vacuum chamber bathed in air at a constant temperature of 20 °C. NEID’s spectrograph components, held at less than one-millionth of the standard atmospheric pressure inside the vacuum chamber, are protected from even tiny changes in pressure by a combination of pumps and “getters”—including 2 liters of cryogenic charcoal—to which stray gas molecules stick. A network of more than 75 high-precision thermometers actively controls 30 heaters placed around the instrument’s outer surface to compensate for any unexpected thermal fluctuations. With no moving parts or motors that might generate heat and disrupt this delicate thermal balance, the system is able to hold NEID’s core optical components to exactly 26.850 °C, plus or minus 0.001 °C. As the starlight emerges from the optical fiber, it strikes a parabolic collimating mirror and then a reflective diffraction grating that is 800 millimeters long. The grating splits the light into more than 100 thin colored lines. A large glass prism and a system of four lenses then spread the lines across a siliconbased, 80-megapixel charge-coupled device. Within

Light from star

550 nanometers

per quadrillion. By using the comb to regularly calibrate the instrument, we can account for any residual instabilities.

380 nm

RAINBOWS UP CLOSE: The NEID spectrograph analyzes wavelengths ranging from 380 to 930 nanometers. The instrument uses calibration marks from a device called a laser frequency comb to detect subtle shifts in absorption features caused by gas in stellar atmospheres. The spectrograph is capable of sensing shifts that are just one forty-thousandth of the spacing between such marks.

that CCD sensor, photons from the distant star are converted into electrons that are counted, one at a time, by supersensitive amplifiers. Heat from the amplifiers makes the detector itself expand and contract in ways that are nearly impossible to calibrate. We had to devise some way to keep that operating heat as constant and manageable as possible. Electrons within the CCD accumulate in pixels, which are actually micrometer-size wells of electric potential, created by voltages applied to minuscule electrodes on one side of the silicon substrate. The usual approach is to hold these voltages constant while the shutter is open and the instrument collects stellar photons. At the end of the observation, manipulation of the voltages shuffles collected electrons over to the readout amplifiers. But such a technique generates enough heat— a few hundredths of a watt—to cripple a system like NEID. Our team devised an alternative operating method that prevents this problem. We manipulate the CCD voltages during the collection of stellar photons, jiggling the pixels slightly without sacrificing their integrity. The result is a constant heat load from the detector rather than transient pulses. Minor imperfections in the CCD detector present us with a separate engineering challenge. State-of-the-art commercial CCDs have high pixel counts, low noise, and impeccable light sensitivity. These sensors nevertheless contain tiny variations in the sizes and locations of the millions of individual pixels and in how efficiently electrons move across the detector array. Those subtle flaws induce smearing effects that can be substantially larger than the femtometer-size Doppler shifts we aim to detect. The atomic ruler provided by the laser frequency comb allows us to address this issue by measuring the imperfections of our sensor—and appropriately calibrating its output—at an unprecedented level of precision. We also have to correct for the motion of the telescope itself as it whirls around Earth’s axis and around the sun at tens of kilometers per second—motion that creates apparent Doppler shifts hundreds SPECTRUM.IEEE.ORG | MAR 2021 |  27

RADIAL-VELOCITY (“DOPPLER WOBBLE”) METHOD No Doppler shift Partial Doppler shift

Star

Center of mass

Planet

Planet Brightness

I N J A N U A R Y 2 0 2 0 , NEID achieved its “first light” with observations of 51 Pegasi. Since then, we have refined the calibration and tuning of the instrument, which in recent months has approached our design target of measuring Doppler wobbles as slow as 33 cm/s—a big step toward the ultimate goal of 10-cm/s sensitivity. With commissioning underway, the instrument is now regularly measuring some of the nearest, brightest sunlike stars for planets. Meanwhile, in Chile’s Atacama Desert, the most sophisticated instrument of this kind yet built, ESPRESSO, has been analyzing Doppler shifts in starlight gathered and combined from the four 8.2-meter telescopes of the Very Large Telescope. ESPRESSO is a technological marvel that achieves unprecedentedly precise observations of very faint and distant stars by using two arms that house separate optics and CCD detectors optimized for the red and blue halves of the visible spectrum. ESPRESSO has already demonstrated measurements that are precise to about 25 cm/s, with still better sensitivity expected as the multiyear commissioning process continues. Astronomers have used the instrument to confirm the existence of a supersize rocky planet in a tight orbit around Proxima Centauri, the star nearest the sun, and to observe the extremely hot planet WASP-76b, where molten iron presumably rains from the skies. At the 4.3-meter Lowell Discovery Telescope near Flagstaff, Ariz., EXPRES has recently demonstrated its own ability to measure stellar motions to much better than 50 cm/s. Several other Doppler velocimeters, to be installed on other telescopes, are either in the final stages of construction or are coming on line to join the effort. The variety of designs will help astronomers observe stars of differing kinds and brightnesses. Since that 2010 meeting, our community of instrument builders has quickly learned which design solutions work well and which have turned out to be more trouble than they are worth. Perhaps when we convene at the next such meeting in California, one group will have already found Earth 2.0. There’ll be no telling, though, whether it harbors alien life—and if so, whether the aliens have also spied us. n

I

Dopplershifted light

TRANSIT METHOD

of thousands of times as great as the ~10 cm/s wobbles caused by an Earth-like exoplanet. Fortunately, NASA’s Jet Propulsion Laboratory has for decades been measuring Earth’s velocity through space to much better precision than we can measure it with our spectrograph. Our software combines that data with sensitive measurements of Earth’s rotation to correct for the motion of the telescope.

Star

Light curve

Time

HOW TO WEIGH A PLANET T H E F I R S T S T E P in estimating the mass of an exoplanet is to estimate the mass of the star it orbits. That can be done based on the spectral type of that star, for example. The mass of a planet can then be gauged by measuring how it influences the motion of the star. Although you might think that a planet orbits a star, in truth the planet and star both orbit their shared center of mass. Because the star is much more massive than the planet, their center of mass is located close to the center of the star; it may even be located within the star. In any case, the star moves around this point as the planet travels in its orbit. Astronomers can measure this stellar “wobble” by virtue of subtle Doppler shifts in the light the star emits [top diagram]. When the star moves in the direction of Earth, characteristic absorption features in the spectrum of light it emits will be shifted toward shorter wavelengths. When the star wobbles in the opposite direction, these absorption features will be shifted toward longer wavelengths. By measuring these shifts over time, astronomers can work out the orbital period of the planet. Knowing the mass of the star, they can also determine the size of the planet’s orbit and thus how fast it moves. If you know how fast the star is moving when it wobbles, you’ll have all the information needed to determine the mass of the planet. The problem is that the Doppler shifts reveal only the velocity of motion toward or away from Earth. If Earth is in the plane of the planet’s orbit, astronomers will observe a Doppler shift of maximal value [blue line]. If Earth is located at 90 degrees to that plane, no Doppler will be seen [orange]. If Earth is at some intermediate angle, the Doppler shift will be of some intermediate value [green]. Lacking additional information, this angle is unknown, and all that can be determined is a minimum estimate of the planet’s mass. Fortunately, sometimes more is known. If Earth is in the orbital plane of the planet or very close to that, the planet may periodically pass in front of the star, blocking a portion of its light [bottom diagram]. If astronomers detect the planet transiting in front of its star, they will know that Earth must be in the orbital plane of the planet and that their estimate of mass is not just a minimum value. They will thus have a good measure of the planet’s mass. The amount of light blocked also allows them to estimate the size, and thus the density, of the planet. 28  | MAR 2021 | SPECTRUM.IEEE.ORG

POST YOUR COMMENTS AT spectrum.ieee.org/anotherearth-mar2021

The World’s Best

OP2 Courtesy of ROBOTIS

ROBOTS GUIDE Is Here!

ROBOTS.IEEE.ORG

IEEE Spectrum’s new ROBOTS site features more than 200 robots from around the world. • Spin, swipe and tap to make

robots move. • Read up-to-date robotics news.

• Rate robots and check

their ranking. • View photography, videos and technical specs.

• Play Faceoff, an interactive

question game.

Check out

Robots.ieee.org on your desktop, tablet, or phone now!

The Radio That Can Hear Over Itself By Joel Brand

Self-interference cancellation allows radios to transmit and receive on the same frequency

30  | MAR 2021 | SPECTRUM.IEEE.ORG

ILLUSTRATION BY

Harry Campbell

S E L F - I N T E R F E R E N C E C A N C E L L AT I O N

W

IRELESS ISN’T ACTUALLY UBIQUITOUS. We’ve all seen the effects: Calls are dropped and Web pages sometimes take forever to load. One of the most fundamental reasons why such holes in our coverage occur is that wireless networks today are overwhelmingly configured as star networks. This means there’s a centrally located piece of infrastructure, such as a cell tower or a router, that communicates with all of the mobile devices around it in a starburst pattern.

Ubiquitous wireless coverage will happen only when a different type of network, the mesh network, enhances these star networks. Unlike a star network, a mesh network consists of nodes that communicate with one another as well as end-user devices. With such a system, coverage holes in a wireless network can be filled by simply adding a node to carry the signal around the obstruction. A Wi-Fi signal, for example, can be strengthened in a portion of a building with poor reception by installing a node that communicates with the main router. However, current wireless mesh-­ network designs have limitations. By far the biggest is that a node in a mesh network will interfere with itself as it relays data if it uses the same frequency to transmit and receive signals. So current designs send and receive on different frequency bands. But spectrum is a scarce resource, especially for the heavily trafficked frequencies used by cellular networks and Wi-Fi. It can be hard to justify devoting so much spectrum to filling in coverage holes when cell towers and Wi-Fi routers do a pretty good job of keeping people connected most of the time. And yet, a breakthrough here could bring mesh networks into even the most demanding and ­s pectrum-intensive networks, for example ones connecting ­assembly-floor robots, self-­driving 32  | MAR 2021 | SPECTRUM.IEEE.ORG

cars, or drone swarms. And indeed, such a breakthrough technology is now emerging: self-interference cancellation (SIC). As the name implies, SIC makes it possible for a mesh-network node to cancel out the interference it creates by transmitting and receiving on the same frequency. The technology literally doubles a node’s spectral efficiency, by eliminating the need for separate transmit and receive frequencies. There are now tens of billions of wireless devices in the world. At least 5 billion of them are mobile phones, according to the GSM Association. The Wi-Fi Alliance reports more than 13 billion

Frequency-division duplexing

Time-division duplexing

Full duplex on the same frequency

C E L L TOW E R

CELLPHONE

Radios, like the one in your cellphone, typically communicate using different frequencies, or the same frequency at different times, to send and receive signals. These techniques are half as efficient at using spectrum as using the same f­requency at the same time.

Wi-Fi equipped devices in use, and the ­Bluetooth Special Interest Group predicts that more than 7.5 billion Bluetooth devices will be shipped between 2020 and 2024. Now is the time to bring wireless mesh networks to the mainstream, as wireless capability is built in to more products—bathroom scales, tennis shoes, pressure cookers, and too many others to count. Consumers will expect them to work everywhere, and SIC will make that possible, by enabling robust mesh networks without coverage holes. Best of all, perhaps, they’ll do so by using only a modest amount of spectrum.

C

ellphones, Wi-Fi routers, and other two-way radios are considered full-duplex radios. This means they are capable of both sending and receiving signals, oftentimes by using separate transmitters and receivers. Typically, radios will transmit and receive signals using either frequency-division duplexing, meaning transmit and receive signals use two different frequencies, or time-division duplexing, meaning transmit and receive signals use the same frequency but at different times. The downside of both duplexing techniques is that each frequency band is theoretically being used to only half of its potential at any given time—in other words, to either send or receive, not both. It’s been a long-standing goal among radio engineers to develop full duplex on the same frequency, which would be able to make maximal use of spectrum by transmitting and receiving on the same band at the same time. You could think of other full-duplex measures as being like a two-lane highway, with traffic traveling in different directions on different lanes. Fullduplexing on the same frequency would be like building just a single lane with cars driving in both directions at once. That’s nonsensical for traffic, perhaps, but entirely possible for radio engineering.

TRANSMITTING ANTENNA

R EC E I V I N G A N T E N N A

AT T E N UATO R

Radiofrequency layer

P H AS E S H I F T E R

T I M E D E L AY

6

3

4 Intermediate frequency layer

7

2

10 1

8 Digital canceler

Digital layer 5

9 Digital tuner

An SIC component in a radio samples the transmit signal at the digital [1], IF [2], and RF [3] layers. At the IF and RF layers, the sampled signal is adjusted by several components [4] to create inverses of the samples. At the digital layer, algorithms cancel out alterations in the signal caused by environmental reflections [5]. When the signal is received, the SIC component cancels it at the RF [6], IF [7], and digital [8] layers. The transmit signal is also sampled [9] by a digital tuner that will adjust SIC components [10] to better cancel it next time.

To be clear, full duplex on the same frequency remains a goal that radio engineers are still working toward. Self-­ interference cancellation is bringing radios closer to that goal, by enabling a radio to cancel out its own transmissions and hear other signals on the same frequency at the same time, but it is not a perfected technology. SIC is just now starting to emerge into mainstream use. In the United States, there are at least three ­startups bringing SIC to real-world applications: ­G enXComm, L ­ extrum, and Kumu ­Networks (where I am vice president of TECHNICAL ILLUSTRATIONS BY

Erik Vrielink

product management). There are also a handful of substantial programs developing self-cancellation techniques at universities, namely Columbia, Stanford (where Kumu Networks got its start), and the University of Texas at Austin. Upon first consideration, SIC might seem simple. After all, the transmitting radio knows exactly what its transmit signal will be, before the signal is sent. Then, all the transmitting radio has to do is cancel out its own transmit signal from the mixture of signals its antenna is picking up in order to hear signals from other radios, right?

In reality, SIC is more complicated, because a radio signal must go through several steps before transmission that can affect the transmitted signal. A modern radio, such as the one in your smartphone, starts with a digital version of the signal to be transmitted in its software. However, in the process of turning the digital representation into a radiofrequency signal for transmission, the radio’s analog circuitry generates noise that distorts the RF signal, making it impossible to use the signal as-is for selfcancellation. This noise cannot be easily predicted because it partly results from ambient temperatures and subtle manufacturing imperfections. The difference in magnitude of an interfering transmitted signal’s power in comparison with that of a desired received signal also confounds cancellation. The power transmitted by the radio’s amplifier is many orders of magnitude stronger than the power of received signals. It’s like trying to hear someone several feet away whispering to you while you’re simultaneously shouting at them. Furthermore, the signal that arrives at the receiving antenna is not quite the same as it was when the radio sent it. By the time it returns, the signal also includes reflections from nearby trees, walls, buildings, or anything else in the radio’s vicinity. The reflections become even more complicated when the signal bounces off of moving objects, such as people, vehicles, or even heavy rain. This means that if the radio simply canceled out the transmit signal as it was when the radio sent it, the radio would fail to cancel out these reflections. So to be done well, self-interference cancellation techniques depend on a mixture of algorithms and analog tricks to account for signal variations created by both the radio’s components and its local environment. Recall that the goal is to create a signal that is the inverse of the transmit signal. This inverse signal, when combined with the original receive signal, should ideally cancel out the original transmit signal entirely—even with SPECTRUM.IEEE.ORG | MAR 2021 |  33

S E L F - I N T E R F E R E N C E C A N C E L L AT I O N

the added noise, distortions, and reflections—leaving only the received signal. In practice, though, the success of any cancellation technique is still measured by how much cancellation it provides. Kumu’s SIC technique attempts to cancel out the transmit signal at three different times while the radio receives a signal. With this three-tier approach, Kumu’s technique reaches roughly 110 decibels of cancellation, compared with the 20 to 25 dB of cancellation achievable by a typical mesh Wi-Fi access point. The first step, which is carried out in the analog realm, is at the radio-­ frequency level. Here, a specialized SIC component in the radio samples the transmit signal, just before it reaches the antenna. By this point, the radio has already modulated and amplified the signal. What this means is that any irregularities caused by the radio’s own signal mixer, power amplifier, and other components are already present in the sample and can therefore be canceled out by simply inverting the sample taken and feeding it into the radio’s receiver. The next step, also carried out in the

analog realm, cancels out more of the transmitting signal at the intermediatefrequency (IF) level. Intermediate frequencies, as the name suggests, are a middle step between a radio’s creation of a digital signal and the actual transmitted signal. Intermediate frequencies are commonly used to reduce the cost and complexity of a radio. By using an intermediate frequency, a radio can reuse components like filters, rather than including separate filters for every frequency band and channel on which the radio may operate. Both Wi-Fi routers and cellphones, for example, first convert digital signals to intermediate frequencies in order to reuse components, and convert the signals to their final transmit frequency only later in the process. Kumu’s SIC technique tackles IF cancellation in the same way it carries out the RF cancellation. The SIC component samples the IF signal in the transmitter before it’s converted to the transmit frequency, modulated, and amplified. The IF signal is then inverted and applied to the receive signal after the receive signal has been converted to the interme-

CELLPHONE

3 CELLPHONE

R E L AY N O D E

1 C E L L TOW E R

R E L AY N O D E

2

CELLPHONE

34  | MAR 2021 | SPECTRUM.IEEE.ORG

When a cellphone is close enough or has line of sight to a cell tower, it is able to communicate easily with established duplexing techniques [1]. Relay nodes can extend a cell tower’s signal range without hogging spectrum by using self-interference cancellation (SIC). The best results will be when the cellphone is positioned directly opposite the relay node from the cell tower [2]. At an angle, the cancellation required to keep communications clear gets trickier as the signals begin to interfere with one another [3].

diate frequency. An interesting aspect of Kumu’s SIC technique to note here is that the sampling steps and cancellation steps happen as an inverse of one another. In other words, while the SIC component samples the IF signal before the RF signal in the transmitter, the component applies cancellation to the RF signal before the IF signal. The third and final step in Kumu’s cancellation process applies an algorithm to the received signal after it has been converted to a digital format. The algorithm compares the remaining received signal with the original transmit signal from before the IF and RF steps. The algorithm essentially combs through the received signal for any lingering effects that might possibly have been caused by the transmitter’s components or the transmitted signal reflecting through the nearby environment, and cancels them. None of these steps is 100 percent effective. But taken together, they can reach a level of cancellation sufficient to remove enough of the transmit signal to enable the reception of other reasonably strong signals on the same frequency. This cancellation is good enough for many key applications of interest, such as the Wi-Fi repeater described earlier.

A

s I mentioned before, engin e e r s s t i l l h ave n’t f u l ly realized full-duplex-on-thesame-frequency radios. For now, SIC is being deployed in applications where the transmitter and receiver are close to one another, or even in the same physical chassis, but not sharing the same antenna. Let’s take a look at a few important examples. Kumu’s technology is already commercially deployed in 4G Long Term ­E volution (LTE) networks, where a device called a relay node can plug coverage holes, thanks to SIC. A relay node is essentially a pair of two-way radios connected back-to-back. The first radio of the pair, oriented toward a 4G cell tower, receives signals from the network. The second radio, oriented toward the cov-

erage hole, passes on the signals, on the same frequency, to users in the coverage hole. The node also receives signals from users in the coverage hole and relays them—again, on the same frequency— to the cell tower. Relay nodes perform similarly to traditional repeaters and extenders that expand a coverage area by repeating a broadcasted signal farther from its source. The difference is, relay nodes do so without amplifying noise because they decode and regenerate the original signal, rather than just boosting it. Because a relay node fully retransmits a signal, in order for the node to work properly, the transmitter facing the 4G base station must not interfere with the receiver facing the coverage hole. Remember that a big problem in reusing spectrum is that transmit signals are orders of magnitude louder than receive signals. You don’t want the node to drown out the signals it’s trying to relay from users by its own attempts to retransmit them. Likewise, you don’t want the transmitter facing the coverage hole to overwhelm signals coming in from the cell tower. SIC prevents each radio from drowning out the signals the other is listening for, by canceling out its own transmissions. Ongoing 5G network deployments offer an even bigger opportunity for SIC. 5G differs from previous cellular generations with the inclusion of small cells, essentially miniature cell towers placed 100 to 200 meters apart. 5G networks require small cells because the cellular generation utilizes higher-frequency, millimeterwave signals, which do not travel as far as other cellular frequencies. The Small Cell Forum predicts that by 2025, more than 13 million 5G small cells will be installed worldwide. Every single one of those small cells will require a dedicated link, called a backhaul link, that connects it to the rest of the network. The vast majority of those backhaul links will be wireless, because the alternative—fiber optic cable—is more expensive. Indeed, the 5G industry is developing a set of standards called Integrated Access and Backhaul

(IAB) to develop more robust and efficient wireless backhaul links. IAB, as the name suggests, has two components. The first is access, meaning the ability of local devices such as smartphones to communicate with the nearest small cell. The second is backhaul, meaning the ability of the small cell to communicate with the rest of the network. The first proposed schemes for 5G IAB were to either allow access and backhaul communications to take turns on the same high-speed channel, or to use separate channels for the two sets of communications. Both come with major drawbacks. The problem with sharing the same channel is that you’ve introduced time delays for latency-­sensitive applications like virtual reality and multiplayer gaming. On the other hand, using separate channels also incurs a substantial cost: You’ve doubled the amount of often-expensive wireless spectrum you need to license for the network. In both cases, you’re not making the most efficient use of the wireless capacity. As in the LTE relay-node example, SIC can cancel the transmit signal from an access radio on a small cell at the backhaul radio’s receiver, and similarly, cancel the transmit signal from a backhaul radio on the same small cell at the access radio’s receiver. The end result is that the cell’s backhaul radio can receive signals from the wider network even as the cell’s access radio is talking to nearby devices. Kumu’s technology is not yet commercially deployed in 5G networks using IAB, because IAB is still relatively new. The 3rd Generation Partnership Project, which develops protocols for mobile telecommunications, froze the first round of IAB standards in June 2020, and since then, Kumu has been refining its technology through industry trials. One last technology worth mentioning is Wi-Fi, which is starting to make greater use of mesh networks. A home Wi-Fi network, for example, now needs to reach

PCs, TVs, webcams, smartphones, and any smart-home devices, regardless of their location. A single router may be enough to cover a small house, but bigger houses, or a small office building, may require a mesh network with two or three nodes to provide complete coverage. Current popular Wi-Fi mesh techniques allocate some of the available wireless bands for dedicated internal communication between the mesh nodes. By doing so, they give up on some of the capacity that otherwise could have been offered to users. Once again, SIC can improve performance by making it possible for internal communications and signals from devices to use the same frequencies simultaneously. Unfortunately, this application is still a way off compared with 4G and 5G applications. As it stands, it’s not currently cost effective to develop SIC technology for Wi-Fi mesh networks, because these networks typically handle much lower volumes of traffic than 4G and 5G base stations. Mesh networks are being increasingly deployed in both cellular and Wi-Fi networks. The two technologies are becoming more and more similar in how they function and how they’re used, and mesh networks can address the coverage and backhaul issues experienced by both. Mesh networks are also easy to deploy and “self-healing,” meaning that data can be automatically routed around a failed node. Truly robust 4G LTE mesh networks are already being greatly improved by full duplex on the same frequency. I expect the same to happen in the near future with 5G and Wi-Fi networks. And it will arrive just in time. The tech trend in wireless is to squeeze more and more performance out of the same amount of spectrum. SIC is quite literally doubling the amount of available spectrum, and by doing so it is helping to usher in entirely new categories of wireless applications. n

POST YOUR COMMENTS AT spectrum.ieee.org/ selfinterference-mar2021

SPECTRUM.IEEE.ORG | MAR 2021 |  35

THE S M H T I R O G AL THAT MAKE T R A C A T S IN ROLL

HOW MACHIN LEARNING AN E D OTHER TECH TOOLS GUID YOUR GROCE E RIE FROM STORE S TO DOORSTE P By Sharath Rao & Lily Zhang

36

MAR 202 1

SPECTR UM. IEEE .ORG

IN THESE COVID TIMES: Instacart orders surged during the pandemic, with the company adding some 300,000 shoppers and an option for customers to avoid human contact: “Leave at My Door Delivery.”

CLOCKWISE FROM TOP LEFT: LAURA MCDERMOTT/THE NEW YORK TIMES/REDUX; MICHAEL LOCCISANO/GETTY IMAGES (2); JOANNE RATHE/THE BOSTON GLOBE/GETTY IMAGES; CHENEY ORR/REUTERS; EVELYN HOCKSTEIN/ THE WASHINGTON POST/GETTY IMAGES; TED HSU/ALAMY

IT’S SUNDAY MORNING, and, after your socially

distanced morning hike, you look at your schedule for the next few days. You need to restock your refrigerator, but the weekend crowds at the supermarket don’t excite you. Monday and Tuesday are jam-packed with Zoom meetings, and you’ll also be supervising your children’s remote learning. In short, you aren’t going to make it to the grocery store anytime soon. So you pull out your phone, fire up the Instacart app, and select your favorite grocery store. You click through your list of previously purchased items, browse specials, search for a new key-lime sparkling water a friend recommended, then select a delivery window. About 2 hours later, you watch a shopper, wearing a face mask, place bags on your porch.

The transaction seems simple. But this apparent simplicity depends on a complex web of carefully choreographed technologies working behind the scenes, powered by a host of apps, data science, machine-learning algorithms, and human shoppers. Grocery delivery isn’t a new concept, of course. In our great-grandparents’ day, people could select items at a neighborhood store and then walk home emptyhanded, the groceries to follow later, likely transported by a teenager on a bicycle. Customers often had basics like milk

and eggs delivered weekly. But with the advent of the fully stocked supermarket, with its broad selection of staples, produce, and specialty foods, customers shifted to selecting goods from store shelves and toting them home themselves, though in some cities local stores still offered delivery services. Then in 1989, Peapod—followed in the mid-1990s by companies like Webvan and HomeGrocer—tried to revive grocery delivery for the Internet age. They invested heavily in sophisticated warehouses with automated inventory sys-

tems and fleets of delivery vehicles. While these services were adored by some for their high-quality products and short delivery windows, they never became profitable. Industry analysts concluded that the cost of building up delivery networks across dozens of large metro areas rapidly ate into the already thin margins of the grocery industry. Timing, of course, is everything. Cloud computing and inexpensive smartphones emerged in the decade after the launch of the first-generation online grocery companies. By 2012, when Instacart began, these technologies had created an environment in which online grocery ordering could finally come into its own. Today, retailers like Target and Whole Foods (via Amazon) offer delivery and pickup services, using their existing brick-and-mortar facilities. Some of these retailers run their delivery businesses from warehouses, some pull from the stocked shelves of retail stores, and some fulfill from a mix of both. Small, online-only companies like Good Eggs, Imperfect Foods, and Thrive Market offer curated selections of groceries sourced from local farms and suppliers. Meanwhile, food and grocery delivery services emerged to bring brickand-mortar restaurants and stores into the online economy. These businesses— which include DoorDash, Shipt, and Uber Eats in the United States, and Buymie, Deliveroo, and Grofers, based elsewhere—

APP TO APP: A customer selects groceries and chooses a delivery window using the Instacart app. Then the Shopper app takes over, guiding Instacart’s contractors to and through stores as they fulfill grouped orders, suggesting replacements, and, finally, directing them to customers’ doors. 38  | MAR 2021 | SPECTRUM.IEEE.ORG

INSTACART (8)

have built technology platforms and fulfillment networks that existing stores and restaurants can use to reach customers online. In this model, the retailer’s or restaurant’s physical location nearest the customer is the “warehouse,” and a community of independent contractors handles fulfillment and delivery. Our employer, Instacart, is the North American leader in this type of online grocery service, with more than 500 grocers, including Aldi, Costco, Food Lion, Loblaws, Publix, Safeway, Sam’s Club, Sprouts Farmers Market, and Wegmans, encompassing nearly 40,000 physical store locations in the United States and Canada. At the onset of the COVID-19 pandemic, as consumers heeded stayat-home orders, we saw our order volume surge by as much as 500 percent, compared with the volume during those same weeks in 2019. The increase prompted us to more than double the number of shoppers who can access the platform from 200,000 in early March to 500,000 by year-end. Here’s how Instacart works.   F R O M T H E C U S TO M E R ’ S perspective, the ordering process is simple. Customers start by opening a mobile app or logging on to a website. They enter their delivery zip code to see available retailers. After choosing a store or retail chain, they can browse virtual aisles of produce, deli, or snacks and search for specific products, clicking to add items to an online

shopping cart and specifying weights or quantities as appropriate. When finished, they see a list of available 2-hour delivery windows, from later the same day to a week or more in the future. Customers can adjust their orders up until the shoppers start picking their items off store shelves, usually an hour or two before the delivery window. They can enter preferred substitutions beforehand or chat with their shoppers in real time about what’s available. Once the groceries are out of the store and on the move, customers get alerts when their orders are about to be delivered. That’s Instacart from a customer’s perspective. Behind the scenes, we face huge technical challenges to make this process work. We have to keep track of the products in nearly 40,000 grocery stores—billions of different data points. We have to predict how many of our 500,000-plus shoppers will be online at a given time in a given area and available to work. We have to group multiple orders from different customers together into batches, so that the designated shopper can efficiently pick, pack, and deliver them. When products run out, we suggest the best replacements. Finally, we dispatch shoppers to different delivery addresses along the most efficient route. We’re crunching enormous volumes of data every day to keep the customer-facing Instacart app, our Shopper app, our business management tools, and other software all humming along.  

LET’S START WITH how we keep track of products on the shelf. The average large supermarket has about 40,000 unique items. Our database includes the names of these products, plus images, descriptions, nutritional information, pricing, and close-to-real-time availability at every store. We process petabytes daily in order to keep these billions of data points current. Back in 2012, when Instacart started making its first deliveries in the San Francisco Bay Area, we relied on manual methods to get this product data into our system. To stock our first set of virtual shelves, our founders and a handful of employees went to a store and purchased one of every item, filling up cart after cart. They took everything back to the office and entered the product data into the system by hand, taking photos with their phones. It worked, but it obviously wasn’t going to scale. Today, Instacart aggregates product data from a variety of sources, relying on automated rule-based systems to sort it all out. Many stores send us inventory data once a day, including pricing and item availability, while other retailers send updates every few minutes. Large consumer products companies, like General Mills and Procter & Gamble, send us detailed product data, including images and descriptions. We also purchase specialized data from thirdparty companies, including nutrition and allergy information. One listing in our database could have information from dozens of sources that SPECTRUM.IEEE.ORG | MAR 2021 |  39

must be sorted out. Let’s say a popular apple juice just underwent a rebranding, complete with new packaging. Our system has to decide whether to use the image provided by a third-party data provider last week, the image sent in by the local store last night, or the image submitted by the manufacturer earlier that morning. Our rules address this problem. Usually images and other data provided by the manufacturer on the morning of a rebrand will be more up-to-date than data provided by individual stores the night before. But what if a store and the manufacturer upload data at about the same time? In this case, our rules tell the system to trust the image provided by the manufacturer and trust the price and availability data provided by the store. Our catalog updates automatically and continuously around the clock to account for all sorts of incremental changes—more than a billion data points every 24 hours on average.

B E C A U S E I N S TA C A R T D O E S N ’ T own and operate its own stores or warehouses, we don’t have a perfect picture of what is on the shelves of a particular store at any moment, much less what will be there later that day or several days in the future. Instead, we need to make well-informed predictions as we stock our virtual shelves. There’s a lot to consider here. Stores in certain regions may get produce shipments on, say, Monday mornings and meat shipments on Thursday evenings. Some stores restock their shelves periodically throughout the day, while others just restock at night. We’ve built two machine-learning models to help us understand what’s on each store’s shelves and manage our customers’ expectations about what they will actually receive in their grocery bags. Our Item Availability Model predicts the likelihood that popular items are in stock at any location at any given time. We trained this model using our own data set, which includes millions of anony40  | MAR 2021 | SPECTRUM.IEEE.ORG

mized orders from across North America. Some items—like a particular brand of organic eggs, chips, or seasoned salt, or niche items like fresh-made stroopwafels— are considered “active,” meaning they’re regularly ordered year-round from a particular store. “Non-active” items include discontinued products as well as seasonal items like eggnog, Advent calendars, and Peeps marshmallows. The model looks at the history of how often our shoppers are able to purchase the items consumers order most. For each, it calculates an availability score ranging from 0.0 to 1.0; a score of 0.8 means the item has an 80 percent chance of being found in the store by a shopper. We can update this score in real time because our shoppers scan each item they pick up or else mark it as “not found.” Having this score enables us to reduce the chances our customers will order items that won’t be on store shelves when our shoppers look for them, whether that’s a few hours away or days ahead. We do this in several ways. For example, if a customer’s favorite type of peanut butter has a very low availability score, we will automatically bump that listing down in the search results and, in turn, bump up similar products that have a higher availability score. In these times of supply-chain shortages, we’ll also add “out of stock” labels to affected items and prevent customers from adding them to their carts. The COVID-19 pandemic pushed our Item Availability Model in a number of other ways and challenged our assumptions about customer behavior. In March 2020, at the start of the U.S. shelter-inplace orders, we saw massive spikes in demand for common household products like toilet paper and disinfecting wipes. Such items were flying off the shelves faster than retailers could stock them. Consumers behaved in new ways— instead of buying their preferred brand of toilet paper, they grabbed any kind of toilet paper they could find. In response, we broadened the number of products our availability model scores to include

INSTACART’S APPS AND ALGORITHMS IN ACTION An Instacart customer uses the company’s app, either on a mobile device or using a website, to order groceries. Behind the scenes, databases, machinelearning models, and a variety of algorithms interact with that app to help a customer finetune the order. They then hand the order over to a shopper, who is guided through the store and to the customer’s doorstep with the help of more models and algorithms. Finally, the system projects an arrival time, alerts the customer that the shopper’s arrival is imminent, and prompts the customer to leave a tip.

lesser-known products. We also tweaked the model to give less weight to historical data from several weeks earlier in favor of more recent data that might be more representative of the times. If a customer adds an item with a low availability score to her cart, a second machine-learning model—our Item Replacement Recommendation Model— gets to work, prompting the customer to select a replacement from a menu of automatically generated alternatives in case the first-choice item isn’t in stock. Giving customers great replacements is a critical part of making them happy. If you’re shopping in-store for yourself, our research suggests that you’ll have to find replacements for about 7 percent of the items on your list. When Instacart shoppers are shopping for you, they can’t just leave out an item—some items may be critical for you, and if you have to make your own trip to the store after unpacking an Instacart order, you might be less

Inventory and pricing data from stores Independent nutrition information

Product database

Drive Time Model

Product details from manufacturers and others

Item Availability Model

Item Replacement Recommendation Model

Instacart app

Capacity Model

Instacart app

Selects items in app Chooses delivery window

Parking Model

Matching Algorithm

Shopper app

Selects a batch of orders

Checks out

likely to use the service again. But our shoppers aren’t mind readers. If the store is out of your preferred brand of creamy peanut butter, should a shopper replace it with crunchy peanut butter from the same brand? What about a creamy peanut butter from a different brand? We trained our Item Replacement Recommendation Model on a range of data inputs, including item name, product description, and five years of customer ratings of the success of our chosen replacements. When we present a menu of replacement choices, we rank them according to scores assigned by this model. If you select one of the replacements, we’ll remember it for your future orders; if you don’t, our shopper picks from products our model recommends.   T H AT ’ S H O W M A C H I N E learning helps us set expectations with our customers as they fill their shopping carts. ILLUSTRATION BY

Chris Philpot

Itempicking order

Shopper app

Takes items from shelves to cart

Once an order is placed, another piece of technology enters the picture: the Shopper app. The vast majority of Instacart’s shoppers are independent contractors who have signed up to shop for us, meeting requirements and passing a background check. They drive to the stores, select items off the shelves, check out, and deliver the orders. They can choose to work at any time by logging onto the Shopper app. In certain high-volume stores, we also directly employ part-time shoppers who pick and pack orders and then hand them off to contractors for delivery. The Shopper app includes a range of tools meant to make it easy to access new orders, address issues that shoppers encounter, and guide checkout and delivery. When shoppers are ready to work, they open up the app and select batches of orders. As they go through the store and fill orders, they can communicate with the customers via in-app chat. The Shopper app suggests an item-

Routing Algorithm

Instacart app

Sees alert when arrival is imminent

Selects tip

picking order to help the shopper navigate the store efficiently. Generally, this picking order puts refrigerated and frozen items, along with hot or fresh deli preparations, near the end of a shopping trip. Meanwhile, customers can watch their shopper’s progress via the Instacart app, tracking each item as it’s scanned into the shopper’s cart, approving replacement items, and viewing yetto-be-shopped items. When shoppers check out, they can charge the order to a physical card that Instacart mails to them or use a mobile payments system in the Shopper app. If they encounter a problem, they can communicate with our help team through the app. And when they complete a delivery, they can use the app to transfer their earnings to their bank accounts.   WE HAVE MANY ORDERS coming in at once to the same stores, slated to be SPECTRUM.IEEE.ORG | MAR 2021 |  41

delivered in the same general vicinity. In a major metropolitan area, we may get more than 50 orders a minute. So we typically group orders into batches to be picked off store shelves at the same time. Here, our Matching Algorithm comes into play. This technology applies rules of thumb and machine-learning models to try to balance the number of shoppers with customer demand in real time. The algorithm benefits from scale—the more orders we have in a given area, the more options we can give the algorithm and the better decisions it can make. It considers things like a shopper’s age: If shoppers are not yet 21, they may not be eligible to deliver orders containing alcohol. We rerun the Matching Algorithm as often as every few minutes as we get new information about orders and delivery locations. The algorithm works hand in hand with our Capacity Model. This model calculates how much delivery capacity we have throughout the day as conditions on the ground change. We used machine learning to build this system; it takes demand predictions based on historical data and historical shopping speeds at individual stores and couples them with real-time data, including the number of shoppers completing orders and the number of orders waiting in a queue for each store. We rerun this model every 2 minutes to ensure that we’re getting a close-toreal-time understanding of our capacity. That’s why a customer may log on at 1:00 pm and see only one l­ ate-evening delivery slot remaining, but when they look again at 1:30 pm, they see a host of afternoon delivery slots pop up.   WHILE THESE MODELS are critical to Instacart’s operation, other tools are crucial for getting the groceries from the store to the customer smoothly and predictably. Our Drive Time Model uses historical transit times and real-time traffic data to estimate when a shopper will arrive at the store. Our Parking Model calculates how long it can take the shopper to get in 42  | MAR 2021 | SPECTRUM.IEEE.ORG

and out of a particular store’s parking lot. If a shopper is likely to spend 10 minutes cruising for a spot in a small, crowded parking lot, that needs to be built into delivery-time estimates for that store. Once the shopper is ready to make deliveries, our Routing Algorithm comes into play. This model is our take on the classic “traveling salesman” problem. Given three customers at three different addresses in the same city, what’s the most efficient route from the store to the first location and from there to the next two? That’s tricky enough, but Instacart has to work with added complexity. For example, in highly dense areas like New York City, some shoppers may walk to their destinations. And we need to ensure that all three deliveries are made within their designated delivery windows— if a customer isn’t home, too early can be just as bad as too late. So our algorithm considers the projected arrival time, using real-time traffic conditions, to create a delivery route. Our system also sends the projected arrival time to the customer and an alert when the shopper is just a few minutes away.   ALL OF OUR DATABASES, machinelearning models, and geolocation technologies work in concert to build an efficient system. But no model is perfect. And the COVID-19 pandemic proved to be an unexpected stress test for our systems. As stay-at-home orders rippled across North America, with more data flowing into the platform than ever before, we had to repeatedly reconfigure our databases and tools to keep up with the new demand. At the peak, we found ourselves making upgrades multiple times a week. We also had to speed up the rollout of a new feature we had just started testing: Leave at My Door Delivery, which allows shoppers and customers to remain socially distant. Shoppers can drop groceries on the porch of a house or the reception or lobby area of an apartment building and send customers a photo

of their completed orders at the site. We are continually looking at ways to optimize our technology and operations. Right now, we are exploring how to improve the suggested picking orders in the Shopper app. Today we rely on a set of rule-based formulas guided by human intuition—for example, that it’s best to pick up fresh vegetables and fruit together, since they’re usually in the same section of the store. But not all stores have the same layout, aisles in a given store can be rearranged, and items may get moved around the store seasonally. We’re hoping we can use machine learning to develop an algorithm that determines such “rhythms” in the way a location should be shopped, based on historical item-picking data along with seasonal additions to store shelves and regular changes in store layouts. As we add retailers and brands and serve more customers, our algorithms and technologies continue to evolve. We retrain all of our models over and over again to better reflect new activity on our platform. So the next time you click on the Instacart app and order groceries to get you through a busy week, know that anonymized data from your order and from your shopper will get fed into this feedback loop, informing the models we train and the technologies we build. We are proud that our system has been able to keep groceries flowing to people across North America who have been sheltering at home during the pandemic, especially those who are particularly vulnerable to the novel coronavirus. These are extraordinary times, and we’ve taken our responsibility to serve our customers, shoppers, partners, and corporate employees very seriously, as well as to keep them safe. As the world continues to shop from home, we hope that our investments in machine learning will continue to make it easier for everyone to get access to the food they love and more time to enjoy it together. n POST YOUR COMMENTS AT spectrum.ieee.org/ instacart-mar2021

Harness the publishing power of IEEE Access. ®

IEEE Access is a multidisciplinary open access journal offering high-quality peer review, with an expedited, binary review process of 4 to 6 weeks. As a journal published by IEEE, IEEE Access offers a trusted solution for authors like you to gain maximum exposure for your important research.

Explore the many benefits of IEEE Access: • Receive high-quality, rigorous peer review in only 4 to 6 weeks • Reach millions of global users through the IEEE Xplore® digital library by publishing open access • Submit multidisciplinary articles that may not fit in narrowly focused journals • Obtain detailed feedback on your research from highly experienced editors

Learn more at ieeeaccess.ieee.org

• Establish yourself as an industry pioneer by contributing to trending, interdisciplinary topics in one of the many topical sections IEEE Access hosts • Present your research to the world quickly since technological advancement is ever-changing • Take advantage of features such as multimedia integration, usage and citation tracking, and more • Publish without a page limit for $1,750 per article

THE AMMONIA SOLUTION

Ammonia engines and fuel cells in cargo ships could slash their carbon emissions BY MARIA GALLUCCI

ILLUSTRATIONS BY

MCKIBILLO

SPECTRUM.IEEE.ORG | MAR 2021 |  45

There’s a lot to like about ammonia.

This colorless fuel emits no carbon dioxide when burned. It’s abundant and common, and it can be made using renewable electricity, water, and air. Both fuel cells and internal combustion engines can use it. Unlike hydrogen, it doesn’t need to be stored in high-pressure tanks or cryogenic dewars. And it has 10 times the energy density of a lithium-ion battery. For all these reasons, ammonia (NH3) is gaining favor in the global shipping industry, a multitrillion-dollar machine in need of cleaner fuels to power the freighters and tankers that haul manufactured goods and bulk materials across the ocean. Shipping companies seek climate-friendlier alternatives to petroleum that can propel their behemoth vessels for days or weeks at sea and still leave room on board for cargo.

46  | MAR 2021 | SPECTRUM.IEEE.ORG

Maritime shipping contributes nearly 3 percent of annual carbon-dioxide emissions, according to the International Maritime Organization (IMO), the United Nations body that regulates the industry. In 2018, delegates agreed to reduce emissions by 50 percent from 2008 levels by 2050. Meeting that target will require swift and widespread development of diesel-fuel alternatives and new designs for freighters, tankers, and container ships. Shipowners and industry analysts say they expect ammonia to play a pivotal role in decarbonizing cargo ships. But there’s a crucial caveat: No vessels of any size today are equipped to use the fuel. Even if they were, the supply of renewable, or “green,” ammonia produced using carbon-neutral methods is virtually nonexistent. Most ammonia is the product of a highly carbon-intensive process and is primarily used to make fertilizers and chemicals. Recently, though, a handful of projects aim to change that. Finland’s Wärtsilä plans to begin testing ammonia in a marine combustion engine in Stord, Norway, by late March. Germany’s MAN Energy Solutions and Korean shipbuilder Samsung Heavy Industries are part of an initiative to develop the first ammoniafueled oil tanker by 2024. Also by 2024, the Viking Energy is poised to become the first vessel propelled by ammonia fuel cells. The Norwegian energy company Equinor (formerly Statoil) charters this offshore supply vessel, which currently runs on liquefied natural gas. Chemical giant Yara will provide the green ammonia, which it plans to produce at a plant in southern Norway. The initiative “will open up a completely new option for zero-emission shipping,” says Henriette Undrum, Equinor’s vice president of renewable and low-carbon technology. “We are not just solving one small problem for one ship. It’s part of the bigger picture. It will be a starting point to build up the market for zero-carbon fuels.”

O2

+

–

H2

NH3

Electrolysis

O2

Air separation

SOURCE: MAN ENERGY SOLUTIONS

HOW TO PRODUCE “GREEN” AMMONIA

Still, industry experts say that revamping the global shipping fleet will be extraordinarily expensive. Researchers estimate that up to US $1.4 trillion will be needed to achieve the IMO’s emissions-reduction target. And fully eliminating emissions will require an additional $500 billion, according to a January 2020 study by a panel of maritime experts. A number of climate-friendly technologies are being considered to reach that goal, including fuel cells, hydrogen-storage systems, and large battery packs. Spinning metal cylinders, towing kites, and other propulsion methods are already helping to curb diesel-fuel consumption by harnessing the wind. But ammonia will likely dominate among ocean-crossing vessels, which sail for

N2

The traditional Haber-Bosch process is used to produce virtually all of the world’s ammonia, but it is energy and carbon intensive. To decarbonize the ammonia-making process, electricity from renewable sources, such as wind and solar power, is used to electrolyze water, yielding hydrogen (as well as oxygen). The electricity is also used to separate air, yielding nitrogen (as well as oxygen and some argon and carbon dioxide). The hydrogen is then reacted with the nitrogen to produce ammonia, NH3. Cargo ships equipped with ammonia-burning internal combustion engines or ammonia fuel cells are expected to help the shipping industry halve its carbon-dioxide emissions by midcentury.

days or weeks between refueling and rely For ammonia-fueled shipping to on common infrastructure worldwide. become a reality, though, several things For such ships, “ammonia is the lowest- need to go right. Manufacturers and engicost zero-emission fuel that we could neers must overcome technical hurdles find,” says Tristan Smith, a researcher and safety issues in the design of ammoat University College London’s Energy nia engines and fuel cells. Port operaInstitute, which evaluated more than tors and fuel suppliers must build vast 30 different shipping fuels. “bunkering” infrastructure so ships can Smith predicts that green ammonia fill ammonia tanks wherever they dock. will be produced in large volumes and And energy companies and governments will start to be used on ships during will need to invest heavily in solar, wind, the coming decade. Other researchers and other renewable-energy capacity to make similar predictions. According to produce enough green ammonia for thoua September 2019 report from the inter­ sands of ships. Globally, ships consume an national consultancy DNV, ammonia estimated 300 million tons of marine fuels could make up 25 percent of the mari- every year. Given that ammonia’s energy time fuel mix by midcentury, with nearly density is half that of diesel, ammonia all newly built ships running on ammo- producers would need to provide twice nia from 2044 onward. as much liquid ammonia, and ships will SPECTRUM.IEEE.ORG | MAR 2021 |  47

Diesel shortages prompted the first realworld use of ammonia as a fuel. In 1942, German-occupied Belgium was struggling to find enough diesel to run its public buses, just as ridership was increasing. Engineers considered using compressed coal gas, but that fuel’s low energy density and awkward storage requirements made it impractical. In April 1943, Ammonia Casale (now part of the Swiss fertilizer maker Casale) introduced an internal combustion engine that could run on a blend of ammonia and coal gas. Some 100 buses in Belgium adopted the system, dubbed Gazamo. But the bus operator returned to using diesel once supplies reappeared. SHIP SHAPE: The supply vessel Viking

Energy is being retrofitted with a 2-megawatt ammonia fuel-cell system. Ammonia has only half the energy density of traditional fuel, so storing it on board requires more space.

48  | MAR 2021 | SPECTRUM.IEEE.ORG

Over the ensuing decades, research on ammonia engines has come in fits and starts, even as ammonia supplies have soared. In the 1930s, worldwide production of ammonia was about 300,000 metric tons per year. Today, it’s about 150 million metric tons. While ammonia is valuable as a chemical feedstock, the transportation sector has had little incentive to use it as a fuel. Petroleum has a higher energy density and is easier and cheaper to produce. “Now, with a focus on having carbonneutral fuels, it’s obviously a different discussion. The economics around [ammonia] are very different,” says Peter Kirkeby of MAN Energy Solutions. “Everybody wants to know, ‘When can we have the ammonia engine?’ ” Kirkeby spoke from Copenhagen, where the company has a large waterfront facility on the city’s south harbor. MAN, a subsidiary of Volkswagen, develops multimegawatt diesel engines for ships and power generators. The Danish outpost is looking beyond diesel, designing marine engines that run on methanol, liquefied natural gas, liquid petroleum gas, and other alternative fuels. Kirkeby says the industry’s recent push for ammonia comes as renewableenergy producers are seeking new mar-

kets, and as shipping companies look for emission-cutting solutions. Ammonia is a simple molecule, composed of three hydrogen atoms bonded to a single nitrogen atom. Today, most industrial hydrogen is produced using an energy-intensive method called steam methane reforming, which causes the methane in natural gas to react with steam and releases hydrogen, carbon monoxide, and a small amount of carbon dioxide. Nitrogen is mainly produced by cooling air to separate it into its constituent gases: nitrogen, oxygen, argon, and carbon dioxide. To make ammonia, hydrogen and nitrogen are reacted with a catalyst at high temperature (about 500 °C) and high pressure (20 to 40 megapascals) via an industrial process developed by the German chemists Fritz Haber and Carl Bosch more than a century ago. To be stored in large quantities, ammonia can be liquefied by being pressurized (to about 1 MPa at 25 °C) or refrigerated (to –33 °C). All told, the Haber-Bosch process accounts for 1.8 percent, or half a billion metric tons, of human-caused global CO2 emissions each year. If ammonia is to play a part in reducing maritime emissions, the fuel must be made in a cleaner way. For example, the hydrogen can be made through electrolysis, splitting water into hydrogen and oxygen using electricity from a renewable source such as wind or solar power. Renewable energy can also be used to separate nitrogen from air. Boosting fuel supplies and building fueldistribution infrastructure are the biggest challenges to ammonia-powered shipping, experts say. Only tiny amounts of green ammonia are now being produced. A trial plant at the Fukushima Renewable Energy Institute in Japan uses solar power and water electrolysis to produce 20 to 50 kilograms of green ammonia per day. A demonstration system at the Rutherford Appleton Laboratory, in Oxfordshire, England, is powered by an on-site wind turbine and makes up to 30 kg of green

EIDESVIK

need to accommodate larger storage tanks, potentially eating into cargo space. But if these efforts succeed, it will mark a dramatic revival for a transportation fuel that’s largely sat on the sidelines since World War II.

100% 90% 80% 70% 60% 50% 40% 30% 20% 10%

CHART: MICHAEL SOLITA; DATA SOURCE: AMERICAN BUREAU OF SHIPPING

Biofuels

Liquefied natural gas (LNG)

2049

2050

2047

2048

2046

2044

2045

2043

2041

Liquefied petroleum gas (LPG)

2042

2039

2040

2037

Methanol

2038

2036

2034

2035

2033

2031

2032

2029

Ammonia/ hydrogen

2030

2027

2028

2025

2026

0%

Oil based

PROJECTED MARINE FUEL USE TO 2050

As the shipping industry moves to reduce greenhouse-gas emissions, as mandated by the International Maritime Organization, ammonia is projected to be a leading alternative to traditional oil-based fuels by 2050. For that to happen in an environmentally sound way, renewable energy sources to supply green ammonia will need to be built, along with infrastructure for distributing ammonia to far-flung ports.

ammonia daily. [For a look at how a farmer in Iowa is using solar power to produce green ammonia, see “The Carbon-Free Farm,” IEEE Spectrum, November 2019.] Larger initiatives are underway in Australia, Chile, and New Zealand. In Queensland, for example, the Australian Renewable Energy Agency recently backed a A$3.9 million (US $3.0 million) feasibility project for a plant that could produce 20,000 metric tons of ammonia annually, using 208 gigawatt-hours

of electricity from solar and wind. The global shipping industry used the equivalent of 3.05 million GWh in 2015. Substituting just 10 percent of that total with green ammonia will require some 550,000 GWh of renewable electricity, according to the Korean Register of Shipping.

As green ammonia slowly scales up, the shipping industry will have to solve

some other problems. The top concern is ammonia’s toxicity. In concentrated form, the pungent, colorless gas can be deadly. In January 2020, a spill of nearly 3,000 liters of liquefied ammonia fertilizer in Illinois sent more than 80 people to the hospital with chest pain, eye irritation, cough, and severe headache. Ammonia manufacturers and distributors must follow strict handling and safety guidelines to minimize the potential for disaster. To use ammonia fuel, ships will need additional safety equipment, such as emergency ventilation and gas-absorption systems. Fortunately, operators of chemical tankers—large vessels designed to transport hazardous products—already have experience handling ammonia. About 10 percent of annual production is transported by sea. These ammonia tankers may be among the first vessels to use the chemical for fuel, in the same way that today’s liquefied natural gas carriers burn some of their own cargo while sailing. Still, using ammonia in the engine room poses new risks. MAN’s engine will likely include double-walled fuel pipes to prevent gas from escaping should the inner pipe leak or rupture. A mechanical ventilation system will intercept any leaking gas and alert the ship’s crew. Ammonia is also corrosive to some alloys containing copper, nickel, and certain plastics. The fuel is difficult to ignite and doesn’t sustain combustion well. Engineers could solve the ignition problem by combining ammonia with a liquid pilot fuel, such as diesel, though that would boost the ship’s carbon footprint. Or they could potentially combine it with better-burning liquid hydrogen; that would require adding hydrogen tanks or equipment to separate hydrogen from the ammonia as needed. Air pollution from burning ammonia presents another puzzle for engineers to solve. When burned at high temperatures, ammonia produces nitrogen dioxide, which contributes to smog SPECTRUM.IEEE.ORG | MAR 2021 |  49

and acid rain and can harm people’s de l’Atlantique and the Swiss line MSC the IMO will ultimately enforce the respiratory systems. Combustion also Cruises are spearheading the initiative. rules. Regulators will need to comyields small amounts of nitrous oxide— Although the fuel cell will initially run pel, not just encourage, companies a greenhouse gas that’s significantly more on liquefied natural gas, it will also be to eliminate greenhouse-gas emispotent than carbon dioxide and methane. compatible with ammonia, methanol, sions, Raucci says. “There is a need for If necessary, shipbuilders could install and other gaseous fuels, the partners say. ­policy-driven objectives to decarbonize special equipment, such as for selecIn the near term, fuel cells are expected the shipping industry.” tive catalytic reduction, to avoid such to play only a complementary role on A May 2020 survey by the American outcomes. Japan Engine Corp. and the ships, supplying electricity for auxiliary Bureau of Shipping captures the uncerNational Maritime Research Institute, systems and navigational equipment. If tainty sowed by today’s vague poliin Tokyo, evaluated such devices on a developers can scale up the technology cies. Nearly two-thirds of shipowners 7.7-kilowatt, single-cylinder engine using to propel large ships and bring down and operators said they have no decara diesel-ammonia mixture. manufacturing costs, fuel cells could bonization strategy in place. Even so, Another option for eliminating harm- eventually provide the least expensive nearly 60 percent of respondents said ful emissions is to use fuel cells rather way to operate ammonia vessels, says they view hydrogen and ammonia as the than an internal combustion engine. In Carlo Raucci, who was a principal con- most attractive fuel choices in the long simple terms, a fuel cell converts chem- sultant of University Maritime Advisory term—even if they don’t have plans to ical energy into electrical energy with- Services, in London, at the time of our use them yet. out burning the fuel, thus avoiding the interview. A big container ship would “We think that the major reason behind release of harmful gases or particles into need more than 60 MW of fuel-cell capac- this [disparity] was the lack of regulathe air. Although existing fuel cells don’t ity, while a small bulk carrier might need tory framework so far,” says Sotirios have an adequate power capacity for only 2 MW, he says. Mamalis, who manages the American ships, experts believe the devices will Other experimental systems aim Bureau of Shipping’s sustainability, fuels, eventually be able to provide higher effi- to prove the viability of ammonia and technology program from Houston. ciency and lower emissions than internal at sea. MAN Energy Solutions plans “A lot of the owners, management comcombustion engines can. to st ar t full- sc ale tests on a t wo - panies, and operators are not necessarily About two dozen projects have suc- stroke ammonia-burning engine in aware of what they need to do in order cessfully demonstrated that fuel cells can Copenhagen this year, Kirkeby says. to develop a decarbonization strategy.” power and propel smaller vessels. Many In 2019, the company partnered with One policy tool would be to set a of these involve the electrochemical reac- Japan’s Kyushu University to assess the global price on CO2 emissions, Raucci tion of hydrogen and oxygen in what’s combustion and heat-release character- says. This would make it more expenknown as a proton-exchange membrane istics of ammonia on a smaller combus- sive to use fossil-fuel products, allowing fuel cell, which operates at low tempera- tion rig. Separately, MAN is developing alternative fuels like ammonia to comture and pressure. But ammonia is not a an ammonia engine for a medium- pete. International regulators could also suitable fuel for these devices. size container vessel in a project with establish standards limiting a fuel’s carNH3 is also more difficult to oxidize the Shanghai Merchant Ship Design & bon content by mass, similar to existthan hydrogen is, and so it requires Research Institute. ing restrictions on the sulfur content higher temperatures to speed up the “On the technology side, we see some of fuels. reaction. Researchers say a better fit work ahead for ammonia,” Kirkeby says. The new initiatives by MAN Energy for ammonia may be the solid-oxide fuel “But it’s doable.” Solutions, Samsung, Equinor, and other cell, which uses a solid ceramic material companies will be critical for determinsuch as zirconia as the electrolyte. These ing ammonia’s potential within the shipdevices can operate at high temperaping industry. Given that vessels can tures of about 1,000 °C. A 2-megawatt All of the forecasting and specula- operate for decades, companies “need system is being installed on the Viking tion around ammonia, fuel cells, and to make sure that they’re investing in a Energy supply ship in Norway and will the like assume that the shipping fuel that has a good chance of being used be tested beginning in 2024. industry will embrace such climate- long-term,” Raucci says. “The maritime In France, meanwhile, a new cruise friendly approaches. Critics say the industry at this moment has a very comvessel will demonstrate a 50-kW solid- International Maritime Organization’s plex choice to make.” n oxide fuel cell system when deliv- emission-reduction goals aren’t ambiPOST YOUR COMMENTS AT spectrum.ieee.org/ ered in 2022. Shipbuilder Chantiers tious enough, and it’s unclear how ammonia-mar2021 50  | MAR 2021 | SPECTRUM.IEEE.ORG

PAST FORWARD BY ALLISON MARSH

CATFISH, ROBOT, SWIMMER, SPY

↗ For more on the history of underwater robots, see spectrum.ieee.org/ pastforward-mar2021

51  | MAR 2021 | SPECTRUM.IEEE.ORG

CIA MUSEUM

Charlie is not your average catfish. A master of disguise, this fish is in fact an unmanned underwater vehicle developed by the CIA’s Office of Advanced Technologies and Programs in the 1990s. Its mission: to collect water samples without being detected. Charlie’s human handler, perhaps disguised as a humble fisherman, controlled the robotic fish via a line-of-sight radio handset. At 61 centimeters long, the catfish wouldn’t set any biggest-fish records. Whether Charlie reeled in any useful intel is unknown, as details of its missions are still classified. n

New York, NY (MSG Entertainment Group, LLC) – Lead Full Stack Engineer Conceptualize and architect enterprise scale applications using Microsoft and open source technologies. Manage timely completion of systems and development tasks for business applications. Min. Req: Master’s degree or equivalent in Computer Science, or a related technical field and 2 years of experience in object-oriented design and development or in the job offered or related occupation.  Must have 2 years of experience in each of the following:   working with AWS Elastic Beanstalk; Angularjs; Bootstrap; SOA; .NET 4.x; architect enterprise; managing timely completion of systems and development task for business applications; and generating testing and documenting program code. Qualified applicants send resumes to: Emily Pantofel, Job Code: IS100, MSG Entertainment Group, LLC, 2 Penn Plaza, New York, NY 10121.

New York, NY (MSG Entertainment Group, LLC) – Manager, Technical Developer - Reporting to VP of Corporate Applications, develop, create and modify computer applications software and analyze customers’ needs to develop software solutions. Oversee the design, implementation and deployment of technical integration solutions for HR, Payroll, Financials, Purchasing, EPM and Venue operational systems. Min. Req: Bachelor’s degree or foreign equivalent in Electronic Engineering, Computer Science, or a related technical field and 5 years of technical integration experience with inbound and outbound financial, human capital management and workforce management systems. Must have 5 years of experience in each of the following: designing and testing RICE objects associated with Oracle R12 E-Business Suite; developing and debugging PL/SQL stored procedures, packages, functions, triggers and indexes; and modifying Oracle 11i interface programs for implementation in R12. Qualified applicants send resumes to: Emily Pantofel, Job Code: IS126, MSG Entertainment Group, LLC, 2 Penn Plaza, New York, NY 10121.

What + If = IEEE

420,000+ members in 160 countries. Embrace the largest, global, technical community. People Driving Technological Innovation.

ieee.org/membership #IEEEmember

One of the most influential reference resources for engineers around the world. For over 100 years, Proceedings of the IEEE has been the leading journal for engineers looking for in-depth tutorial, survey, and review coverage of the technical developments that shape our world. Offering practical, fully referenced articles, Proceedings of the IEEE serves as a bridge to help readers understand important technologies in the areas of electrical engineering and computer science.

To learn more and start your subscription today, visit

ieee.org/proceedings-subscribe 52 | MAR 2021 | SPECTRUM.IEEE.ORG

The Institute VOLUME 4 5 I S S U E 1

News of the IEEE

Making Quantum Tech Work

Verizon’s Jean McManus is leading development of a new encryption method to protect subscriber data P. TI-4 THE INSTITUTE  | 

MAR 2021  |  TI- 1

president’s column

IEEE Membership: Make It Your Own and volunteers and how together we could make positive changes to our profession, I was hooked! This is what I want to share with every individual working in technology. This is an essential point. It is important to understand that IEEE brings together and welcomes to its membership not only engineers but also technologists from a variety of fields, including computer science, information technology, physical science, biological and medical science, mathematics, technical communications, education, management, law, and policy. IEEE is home to some of the highest caliber individuals with whom I have ever been associated. It also has always been a place where my ideas are welcome and participation encouraged. My desire is to see IEEE continue to be a place where future members­—particularly those from underprivileged or underrepresented

NEW OPPORTUNITIES FOR ENGAGEMENT

I

EEE is built on the strength of its members and their volunteer efforts throughout the organization. Volunteering not only benefits IEEE as an organization but is also one of the key benefits available to members. Two new opportunities hope to simplify the process and offer flexibility to volunteering within IEEE. The IEEE Volunteering platform enables members to search for opportunities across the organization, be it short- or long-term, local or remote. Those who are

TI-2  | MAR 2021 | THE INSTITUTE

groups including women, students, and young professionals as well as those in less advanced economies—seeking professional growth can participate and contribute. I would like to help these colleagues in their quest. As never before, they need guidance and support navigating expanding markets and gaining the necessary knowledge to become more competitive across the broad expanse of technical careers. We live in a diverse world full of complex problems. Solving them requires an

looking for helpers can post the positions they need to fill. The portal was developed by the IEEE Young Professionals group but is open to all IEEE members. By using the platform, IEEE leaders, volunteers, and members can connect with each other according to their schedule, talents, and interest. The objective of the new IEEE Volunteer STEM Portal is to leverage IEEE’s global community to engage and impact as many students as possible and to increase volunteer engagement in preuniversity science, technology, engineering, and math activities. It serves as a centralized volunteer resource for all such activities across IEEE’s operating units.

PREVIOUS PAGE: VERIZON. THIS PAGE: IEEE.TV

I

EEE membership means something different to each individual. For some, being a member means coordinating a conference. To others it is chairing a standards working group or editing a publication. Some individuals become members because they want to show the world the “boots on the ground” vision of IEEE’s mission: advancing technology for the benefit of humanity. When people join, they are looking for IEEE to fulfill a need. Over the years, we have conducted a number of membership surveys. I have looked carefully at the data from those conducted from 2015 to 2020. What I found is consistent. Our members want IEEE to help them remain technically current, engage and network with others, and enhance their careers. One of the best choices I made more than 20 years ago was to become involved in IEEE. It helped define who I am both personally and technically. Once you become involved with the organization, you see how collaborative and how effective each individual can be. When I saw how I could empower other members

editor’s note Members Blaze a Path to the Future  BY KATHY PRETZ Editor in chief, The Institute

RANDI KLETT

environment of collaboration with active and open engagement. IEEE can uniquely provide this environment, where every individual’s passion and commitment is represented and respected, where they feel engaged and empowered, and where we can all work together to support IEEE’s mission. I believe that it is vital that IEEE live up to the ideals expressed in our Code of Ethics: that IEEE be a place where people feel respected and included. We should welcome new and challenging ideas from everyone. It is my hope that we work together to ensure that IEEE continues to support robust diversity of thought and build a community where differences of opinion can be discussed and resolved collegially and where we can find consensus in expanding and furthering IEEE’s mission. I hope that we can work together to create safe and supportive environments, free from bullying, that attract and nurture individuals who strive for professional and technical growth. Let’s work together to accelerate and nurture innovation, helping to make IEEE the technical professional’s lifelong network of choice and the first place people go for the highest quality technical information. I encourage all our members to get involved, use your membership to its fullest, and be part of the drive toward fulfilling our mission of advancing technology for the benefit of humanity. It is remarkable what we can accomplish working together!

Many IEEE members oversee groups at large organizations that are working on developing cuttingedge technologies. In this issue, we highlight two. Jean McManus is executive director of emerging technologies for telecom giant Verizon’s technology and product development group. In addition to other technologies, her team is developing quantum key distribution [page TI-4]. The emerging cryptographic method could use the quantum nature of photons to help protect Verizon’s subscriber data. The IEEE member and her team recently conducted one of the first commercial trials of the technology in the United States. Sethuraman “Panch” Panchanathan, director of the U.S. National Science Foundation, is responsible for guiding the agency’s mission [page TI-6]. The NSF plays a critical role in U.S. science and engineering because it supports basic research. In a Q&A with The Institute, the IEEE Fellow talks about his vision for the foundation, the impact of the COVID-19 pandemic on the scientific community, and what the NSF is doing to encourage more students to study STEM subjects. Also in this issue, we profile Diana Dabby, who put her passion for music to work by establishing the Olin Conductorless Orchestra. No one person leads the orchestra; instead, engineering students work together to perfect their performances [page TI-9]. In addition to serving as a creative outlet, the orchestra teaches students leadership, teamwork, and communication skills. On page TI-12, learn how Josephine Cochran, a socialite with no technical background, invented the automatic dishwasher. On pages TI-14 and TI-15, meet the candidates running for IEEE president-elect and learn who received the IEEE Medal of Honor. Also learn who the winners of the IEEE 2020 election are on page TI-16. n

—SUSAN K. “KATHY” LAND, IEEE president and CEO Share your thoughts with me at [email protected].

FOR UPDATES ABOUT IEEE AND ITS MEMBERS, VISIT US AT SPECTRUM.IEEE.ORG/THE-INSTITUTE THE INSTITUTE  |  MAR 2021  |  TI-3

Later she joined Bell Atlantic, which merged with GTE to form Verizon. Today the IEEE member is executive director of emerging technologies with Verizon’s technology and product development group in Waltham, Mass. McManus also mentors the company’s young engineers, especially women. In addition, she works with professionaldevelopment programs run by Verizon and IEEE to connect with young professionals and help them reach career goals. She was profiled in the 2016 book The Internet of Women: Accelerating Culture Change, which highlights standouts in science, technology, engineering, and math who are making historic contributions to their field.

Jean McManus

The Woman Behind Verizon’s Big Bet on Quantum Cryptography

J

ean McManus seemed to be destined to work for Verizon. After all, every telephone company she worked for in her 24year career was eventually acquired by or merged with Verizon. Her first job out of college was with Contel, which was acquired by GTE in 1990.

TI-4  | MAR 2021 | THE INSTITUTE

VERIZON

DESTINED FOR VERIZON

McManus was inspired to pursue engineering by her father, who was an electrical engineer. She received a bachelor’s degree in electrical engineering in 1987 from Duke University, in Durham, N.C. She then joined Contel as a systems engineer and later transitioned to security engineer. After the company was acquired by GTE, McManus decided to return to school, and she earned a master’s degree in electrical engineering in 1993 and a Ph.D. in systems engineering in 1996 from the University of Pennsylvania, in Philadelphia. For her doctoral dissertation, she focused on the delivery of video over networks. She then joined Bell Atlantic in Arlington, Va., and worked as an engineer on the company’s Video Delivery Through Networks project. The company was conducting a trial in Toms River, N.J., that was related to her dissertation, and “it was a perfect fit,” she says. When Bell Atlantic refocused its efforts on broadband and DSL, McManus became the lead architect in building the company’s DSL architecture. “It was really exciting,” she says, “because at that time, most of my colleagues were experts on narrowband technologies. Meanwhile, I was more familiar with broadband technologies.” McManus was able to define the architecture of broadband technologies such as customer premises equipment (Wi-Fi routers, cable TV boxes, and telephone sets), edge routers and switches, and associated protocols, she says. Four years after she joined Bell Atlantic, it merged with GTE to form Verizon. After the merger, McManus was named a Verizon Fellow— which, she says, provided her with a great opportunity to work with the company’s leaders. She began working in the network architecture department and, in 2014, was promoted to executive director of emerging technologies. She leads a proof-of-concept lab and is responsible for product-focused technology innovation.

profile “Becoming a manager wasn’t in my original career plan,” she says. “I wanted to be an individual contributor. But after 10 years I realized that being a manager would give me more opportunities and would challenge me in new ways.” In her current position, she says, “not only do I have to keep up with technology but I also need to motivate my team of engineers and developers to move forward with their ideas.” McManus and her team—which consists of network architects, engineers, and software developers—are working on quantum key distribution and advancing GPS technology using the satellite navigation technique real-time kinematics (RTK). Quantum key distribution is a “new encryption method that uses photon properties to protect subscriber data,” according to a Verizon news release.

encryption keys to the two sites and were able to detect if someone was eavesdropping on that connection.” Her team is developing software to enhance GPS location data using RTK. The software “makes GPS more accurate and precise, reducing the error to the 2-centimeter range,” McManus says. “My team has figured out a way to scale it such that we can support the large number of Internet of Things devices.” Among those devices are drones. The software McManus and her team developed can make them more vertically accurate, giving the pilots a better measure of how high off the ground the drones are and helping to avoid crashes into telephone wires and power lines.

Employer: Verizon

MENTORING THE NEXT GENERATION

McManus takes pride in being a mentor to McManus’s expertise lies in telecommunicaTitle: Executive women. She says she was inspired to become director of tions, specifically on protocols, architecture, a mentor when she “saw a lot of women were emerging and security. She also has worked on technolstruggling with how to navigate Verizon.” She technologies ogies such as carrier Ethernet (the use of highwanted to provide others the support she didn’t Member grade: bandwidth Ethernet technology), subscriber receive early in her career. Member data management, and network virtualization. “There wasn’t such an emphasis on mentorJoined IEEE: 1993 When she was offered the opportunity to be ing at that time,” she says. Alma mater: involved in product development, she took it. She works with Verizon’s Women of the World, University of “I’m still doing technical work, but it’s now a seven-month-long career-development proPennsylvania more product-focused,” McManus says. “This gram that aims to help employees develop [position] gives me the opportunity to think effective communication skills, personal brand differently about technology and how we can support our development, and self-leadership. The participants are put customers.” into groups led by managers. McManus has two responsibilities: staying up to date with “Jean shared her experiences working as an executive directechnology and exploring areas in telecommunication that tor and encouraged the women in our group to openly discan be improved in support of Verizon’s products and services. cuss our goals, take action, and share our career successes “One part of my day is spent looking to see what’s happenwith each other,” says Sharon Muli, who participated in the ing, whether it be research labs, academia, what other comprogram. “I also met individually with Jean, and she offered panies are doing, or just trying to understand how people me guidance on pursuing various career paths and training are applying technology,” she says. “But, more importantly, opportunities. She utilized her connections to introduce what technologies are going to be coming within the next me to individuals on other teams and strengthen my netseveral years. work at Verizon.” “The other part of my day is spent interacting with my team. McManus, who joined IEEE as a student member, has spoI try to engage with them as much as possible from an innoken at IEEE conferences such as the 2019 IEEE Women in vation perspective.” Engineering Forum, where she was on a panel discussing 5G. “It’s great to see the support [IEEE] is giving to women in DEVELOPING NEW TECH engineering,” she says. “Some of the things it’s doing to help Last year McManus led the development of several innovadevelop women engineers are not just at the start of their tive technologies. careers but also making sure that they’re staying in those She and her team conducted one of the first commercial areas. There’s just a lot of opportunity to really embrace the trials of quantum key distribution in the United States. industry as a whole and develop yourself.” “We connected three Verizon sites in the Washington, —JOANNA GOODRICH D.C., area and sent a video between two of the sites,” McMThis article originally appeared online as “This Executive Director anus says. “Using quantum key distribution, we received Is Leading Verizon Into the Future Through Quantum Computing.” FROM ENGINEER TO LEADER

THE INSTITUTE  |  MAR 2021  |  TI-5

Sethuraman Panchanathan 

S

ethuraman “Panch” Panchanathan left his academic career at Arizona State University in June to start a six-year appointment as director of the U.S. National Science Foundation. There the IEEE Fellow oversees the foundation’s 2,100 employees and its day-to-day operations. Panchanathan is also responsible for directing the agency’s mission, including supporting all fields of fundamental science and engineering in such areas as artificial intelligence and quantum computing. He also has a large budget to manage: US $8.3 billion. That is about 25 percent of the total amount the U.S. government spends to support basic research. The money goes to nearly 2,000 colleges, universities, and institutions across the country. TI-6  | MAR 2021 | THE INSTITUTE

Panchanathan is no stranger to the NSF. He was appointed in 2014 to serve on its National Science Board, a 25-member group that establishes the foundation’s overall policies.  He spent 23 years at ASU in Phoenix, where he developed people-centric technologies and fostered innovative research. He helped found the university’s School of Computing, Informatics, and Decision Systems Engineering and its Center for Cognitive Ubiquitous Computing. He also led its Knowledge Enterprise, which supports entrepreneurs with research, strategic partnerships, international development, and other activities. Panchanathan holds a bachelor’s degree in physics from Vivekananda College—now the University of Madras—in India, and a bachelor’s degree in electronics

STEPHEN VOSS/NATIONAL SCIENCE FOUNDATION

Director Outlines His Vision for U.S. National Science Foundation

profile TI: What has been the impact of the COVID-19 pandemic on scientific research, labs, conferences, and research directions?  Panchanathan: The research community is displaying resilience under tremendous pressure. It makes me proud to be a scientist and an engineer. The role of NSF and other science agencies is to enhance our support to this community. And that’s what we’re working to do. We are all facing new and unique challenges as we deal with COVID-19, and NSF is prioritizing the health and safety of our community. NSF recognizes the many concerns related to the effects this will have on NSF-funded research and facilities, and is committed to providing the greatest flexibility to support researchers’ health and safety. NSF is consistently updating its guidance and resources to keep the scientific community informed. Additionally, NSF reacted right away to the pandemic through its Rapid Response Employer: U.S. Research funding mechanism for nonmedNational Science ical research to understand the spread of Foundation COVID-19, provide education about the sciTitle: Director ence of virus transmission, and encourage Member grade: the development of actions to address this Fellow global challenge. To date, we have funded Joined IEEE: 1987 more than 1,000 coronavirus research projects totaling more than $198 million. Alma mater:

and communication engineering from the Indian Institute of Science in Bangalore. He also holds a master’s degree in electrical engineering from the Indian Institute of Technology, also in Madras. He began his teaching career at the University of Ottawa, after earning his Ph.D. in electrical and computer engineering there in 1989. He left in 1997 to join ASU as an associate professor in the Department of Computer Science and Engineering. Because of his busy schedule, The Institute conducted this interview via email in October. We asked him about his vision for the foundation, how he plans to increase partnerships between industry and academia, and how his membership in IEEE has advanced his career. His answers have been edited for clarity.

The Institute: What inspired you to become an engineer? Panchanathan: At a young age, I was curious about basic science and how things work. My father was my inspiration to become an engineer. He was a scientist, and his work was on upper-atmospheric physics. His quest for scientific exploration, for discovery, for academic achievement, for solving real problems, for University of Ottawa understanding the universe and how it works to TI: What are your thoughts on the need for how people work—all of that has always inspired more students to study STEM subjects, and me and motivated me to want to pursue scihow is the NSF addressing that? ence and engineering.  Panchanathan:  Ensuring inclusivity and broadening My mom ensured that we valued education. So the combiparticipation is an important priority of mine. Diversity nation of my mom and dad’s implicit role modeling was the enriches innovation to solve problems. We must inspire talideal incubator for me to pursue science and engineering. ent in every corner of our nation and empower role models at every level of leadership. I want students to feel empowTI: Where would you like to see the NSF in five years?  ered and excited to pursue science. Panchanathan: The foundation plays a critical role in U.S. Of course, NSF is not the only entity that can do that. A science and engineering because it supports basic research in number of entities are coming together through partnerall these fields. We enable researchers to explore fundamenships, including other federal agencies, industry, nonprofits, tal scientific questions about everything from the forces that foundations, states, and academia. I am deeply committed govern the universe to the biological, chemical, and social to partnerships in all forms. systems that make us who we are. So the question then becomes: How do you partner effecI have identified three pillars for my vision: advancing tively across all entities to build better futures for our nation? research into the future, ensuring inclusivity, and continuIt is going to take commitment and participation with all ing global leadership in science and engineering. players in the STEM community, including K–12 education This is a defining moment. The intensity of global comand informal learning environments. For example, the NSF petition, the urgent need for domestic talent at scale, and Includes program was created to identify best practices and the broad support for science as the path for solving global provide resources to people across the country working to grand challenges all motivate us to strengthen discovery and broaden participation in STEM. translation. Partnerships and innovative mindsets ensure One acknowledgement built into Includes is that broadenwe rapidly seize opportunities and accelerate progress at ing participation is too complex a problem for one-size-fits-all speed and scale. solutions. Something that works in one region or for one THE INSTITUTE  |  MAR 2021  |  TI-7

profile population might not work elsewhere. That is why Includes is helping create education experiences that are tailored to the communities they serve. This is going to require an intense collaboration and intentional, strategic actions. It will not happen unless it is a priority. That is the kind of coalition I envision NSF helping to build. Success takes a village, right?

TI: One way to increase the number of STEM students and STEM workers is to recruit them from other countries. Many U.S. universities and companies have criticized the increased U.S. restrictions on immigration and visas—which have made recruiting difficult. What, if anything, will you do as NSF head to address the situation? Panchanathan: International collaboration enhances U.S. global leadership and ensures that the U.S. research community participates in the best science and has access to the best resources around the world. NSF is committed to sustaining the country’s position as a global innovation leader as well as contributing to its economic strength and national security through basic research. Openness, transparency, and collaboration are essential for basic research. NSF and our fellow federal agencies are continuing to embrace and promote international collaboration. For NSF, this collaboration entails establishing joint projects between researchers at U.S. institutions and those at organizations in other countries. These collaborations will continue because they enable the best science. I would encourage anyone thinking about working or pursuing a career in the United States to do so, as we provide great opportunities for students to express their talents in unimaginable positive ways. TI: How will you foster more partnerships between universities and industry?  Panchanathan:  Partnerships between academia and industry are critical to the rapid advancement of science and engineering, ensuring national prosperity. I am deeply committed to not only strengthening existing frameworks of academia-industry partnerships but also, more importantly, evolving new frameworks for robust collaboration.  The frameworks get researchers from both university and industry to share different perspectives that not only enrich research outcomes but also inspire unparalleled talent, leading to an innovative workforce of the future. They also help evolve new models of partnerships and frameworks. For example, we need to design and build Bell Labs–like entities across the nation through public-private partnerships where curiosity-driven research and translational research are working synergistically to enrich each other, unleashing transformative outcomes for the future.  TI-8  | MAR 2021 | THE INSTITUTE

TI: In your recent interview with Science, you talked about your support for “use-inspired research.” How will the NSF balance funding for use-inspired research and basic research? Panchanathan: What we are talking about at NSF is useinspired basic research, which in some cases may lead to applied research outcomes and commercialization. Our focus should also be to identify the gaps in our knowledge that are holding us back from advancing in some of the most competitive fields of science and engineering. When you look at it from that perspective, you will find that NSF and other supporters of basic research have already been funding use-inspired research for several decades.  NSF has the unique ability to be strategic in how we inspire researchers to cultivate both curiosity-driven and use-inspired mindsets. One example of how NSF will undertake this is our support for convergent research. Scientific knowledge leads to actionable progress, which in turn enriches the scientific process. In other words, science and technology are intertwined. NSF advances technological progress because it is already intrinsic to everything we do. NSF is making this translation happen through several programs. For example, NSF began funding the Laser Interferometer Gravitational-Wave Observatory project decades ago. Some doubted it would ever be possible for LIGO to detect the minute distortions of gravitational waves. LIGO was not a theoretical problem, they feared, but a technological limitation. Science drove the development of technological capabilities necessary to detect gravitational waves. Now that technology will open up new ways to do science, and we continue to see new discoveries from that technology. TI: How has IEEE helped your career?  Panchanathan:  Being a member and Fellow of IEEE has been an important part of my career as an educator, researcher, and leader. In my early career, I had the opportunity to publish several scientific papers in IEEE conference proceedings and journals. Attending the various conferences helped me to gain valuable insights and feedback from leaders in the research community that shaped my research trajectory. I also had the opportunity of serving as a conference organizer, panelist, and editorial board member, and as editor-in-chief of IEEE MultiMedia magazine. These experiences provided me with opportunities to further enrich my knowledge and to contribute to the engineering and scientific community. —KATHY PRETZ This article originally appeared online as “Q&A: U.S. Science Foundation Director on His Vision for the Agency.”

ALEXANDER BUDNITZ

Diana Dabby, creator of the Olin Conductorless Orchestra, interacting with the players.

Conductorless Orchestra Helps EE Students Fine-Tune Their Career Skills

D

iana Dabby  grew up surrounded b y music—both her parents were pianists. The IEEE member followed in their footsteps and earned a bachelor’s degree in music from Vassar College, in Poughkeepsie, N.Y. After graduating, she moved to New York City and worked as a pianist, performing at venues including Merkin Hall and Weill Recital Hall. Although Dabby was passionate about music, she had an unsettling feeling that something was missing. That something turned out to be engineering— which she discovered after she read journal articles about engineering’s relationship to music. She decided to pursue a graduate degree in the field.

THE INSTITUTE  |  MAR 2021  |  TI-9

profile After earning a doctorate in electrical loved that process,” she says. “The idea engineering from MIT, Dabby became of reaching one’s full potential was very an engineering and music professor. She powerful to me.” taught at Tufts University, MIT, and The She says she enjoyed taking risks in Juilliard School. She also continued to play order to achieve her goal of bettering concerts, performing at Jordan Hall, Tan- her skills as a musician. glewood, and other ven“I built up a very strong ues in Massachusetts. track record with taking In 2000 Dabby joined risks,” she says, “whether Employer: Olin the Olin College of EngiCollege of during a performance or Engineering neering, in Needham, in my professional life.” Mass., where she was Title: EE professor And taking a risk is and music program one of 12 founding facexactly what Dabby did director ulty members. In 2002 after she came across Member grade: she established the Olin an engineering journal Member Conductorless Orchesat the New York Public Joined IEEE: 2006 tra (OCO), which comLibrary for the Performpleted its 19th season Alma mater: MIT ing Arts. The journal last year. contained articles by No one person leads engineers whose avothe orchestra; instead, the students work cation was music, and they inspired together to perfect their performances. Dabby to ask: “What if a professional The program is designed to give talented musician, one of my colleagues, or I engineering students an expressive outlet acquired the tools of an engineer? Would while also helping them develop profes- we invent something new for music in sional skills such as leadership, teamwork, our own time?” and communication. That idea pushed her to pursue a gradIn 2019 Dabby won a Best Paper Award uate degree in engineering while workfrom the American Society for Engineer- ing as a performer and freelancer. ing Education. Her winning paper—“The In order to apply to graduate proEngineers’ Orchestra: A Conductorless grams, she had to supplement her music Orchestra for Developing 21st-Century bachelor’s degree with postbaccalaureProfessional Skills”—describes the pro- ate classes. gram’s benefits. “I had to [earn] around 127 credits because I had no math or science backTAKING A RISK ground,” she says. She did so at the City Dabby says music has always been an College of New York. extension of herself, and she enjoyed the “I retaught myself algebra and discovfocus and expressivity that came with ered that I loved it,” she says. “Engineerpreparing for her concerts. ing became this wonderful respite from Performing “just kept accentuating performing. The engineering felt fresh. and improving my musicianship, and I The music felt fresh.”

I built up a very strong track record with taking risks, whether during a performance or in my professional life. TI-10  | MAR 2021 | THE INSTITUTE

After Dabby completed the credits she needed, she was accepted to MIT. For her doctoral thesis, she merged engineering and music. She devised a chaotic mapping tool—a representation of chaotic behavior that is typically used in mathematics—that could be used to make musical variations. The variations, which could be either changes in pitch or in the rhythmic sequence of a piece, could be close to the original work or mutate almost beyond recognition. Dabby has been granted four U.S. patents for her work. She says she wanted to “come up with something for music in the 21st century that wouldn’t necessarily occur to those who were not performers or professional musicians.” CONDUCTORLESS ORCHESTRA

In fall 2000, when the Olin College of Engineering assembled a leadership team and faculty to begin from scratch, it paid attention to a list of skills the U.S. National Academy of Engineering wanted in engineering students. The list included leadership skills, effective communication, and the ability to work as part a team. The Olin faculty members brainstormed how they could help their students develop the skills, and that’s when the OCO was born. The idea “just popped into my head in our first meeting,” Dabby says. “I thought, Oh my gosh, this could mean a conductorless orchestra. Everyone leads, and everyone follows.” The students learn how to collaborate with one another and how to communicate effectively. The musicians learn to watch one another to ensure everyone starts and ends together, as well as adjust balance, dynamic levels, and tempo by listening intently and cueing one another, Dabby says. “It requires the musicians to actively listen to their parts within the context of a larger whole and adjust accordingly,” she wrote in her chapter of the

Engineering became this wonderful respite from performing. The engineering felt fresh. The music felt fresh. book Creative Ways of Knowing Engineer- International Conference, Dabby says. ing. The chapter describes the OCO. “There’s always an upcoming perforOlin had only 75 students in its first mance, and it’s another chance for stuyear, and the first conductorless orches- dents to raise the bar,” she says. “For tra was composed of five engineering stu- students, it’s a challenge and a neat way dents, with Dabby at the piano. These to become better while doing something days there are between 12 and 22 stu- they love.” dents, all selected by audition, in the OCO. The students select a piece to play, and PLAYING THROUGH THE PANDEMIC Dabby creates an arrangement, adjusting Like many other schools, Olin closed its the piece according to the instruments campus due to COVID-19. But that didn’t the students play. stop Dabby from continuing to help stuEach year the musicians elect two to dents develop their performance and profour navigators who work with Dabby fessional skills. to ensure rehearsals run smoothly and This year’s OCO navigators, Caitlin communication lines remain open within Coffey and Jack Mao, continue to select the group. Together, along with two pieces for the orchestra, and the group rehearsal leaders, they come up with meets virtually.  the agenda for that week’s rehearsal. Because the students can’t play together, During rehearsals, orchestra members each player records his or her pieces and can share their thoughts regarding the sends them to Dabby, who mixes together different interpretations of the piece the the audio to create the performance as if group chose to play. The members play the students had played as a whole. each interpretation, and the orchestra “We learn so much by hearing ourselves votes on which version it wants to perform. on playback,” Dabby says. “We hear what All involved in the OCO learn how to lis- can be improved, and that inspires practen, when to speak, and when to refrain tice in order to reach a higher level of from sharing their thoughts. performance.” “Employers see the Olin Conductorless Although the orchestra has often Orchestra on résumés and they’re curi- recorded itself as a whole, its players ous,” Dabby says. “It’s actually helped have rarely had the opportunity to record students get jobs.” themselves playing their own parts.  The program also has helped students “The pandemic has unwittingly given us during their time at the college. a chance to help each musician increase “It’s a stress-reliever,” Dabby says. The his or her musicianship and technical OCO “gives [students] balance in their skills through recording,” Dabby says. “Or lives.” as the old saying goes: When life gives you The orchestra performs at school func- lemons, make lemonade!” tions and travels once a year to play at —JOANNA GOODRICH other venues. In 2019 it received a standing This article originally appeared online as ovation after performing at the American “Conductorless Orchestra Helps EE Students Society for Engineering Education Zone 1 Fine-Tune Their Professional Skills.” THE INSTITUTE  |  MAR 2021  |  TI-11

Josephine Cochran developed the first dishwasher that used water pressure instead of scrubbers to clean dishes.

This Socialite Hated Washing Dishes, So She Automated the Dishwasher

T

he dishwasher, a popular appliance in kitchens around the world, has gone through a number of iterations throughout its 170-year history. The first dishwasher to be granted a patent was invented in 1850 by Joel Houghton. It was a wooden box that used a hand-turned wheel to splash water on dirty dishes, and it had scrubbers. Ten years later, inventor L.A. Alexander improved on Houghton’s machine by adding a “geared mechanism that allowed the user to spin racked dishes through a tub of water,” according to an entry on reference website ThoughtCo. But the person we have to thank for the modern-day dishwasher is Josephine Cochran (sometimes spelled Cochrane). Her machine was the first to use water pressure instead of scrubbers to clean dishes—which made it more efficient than Houghton’s or Alexander’s versions. For Cochran’s invention, she was inducted into the U.S. National Inventors Hall of Fame in 2006. Her technical achievement is worthy of being named an IEEE Milestone, according to the IEEE History Center, but no one has proposed it yet. The Milestone program honors significant achievements in the history of electrical and electronics engineering.

PHOTO-ILLUSTRATION: GLUEKIT

FED UP WITH DIRTY DISHES

Cochran’s dishwashing woes began after she married wealthy merchant William Cochran in 1858. As a socialite, she was expected to hold frequent dinner parties. She served the meals on her expensive, heirloom china. When the household staff hand-washed the dishes, the delicate china often got chipped. She opted to wash the dishes herself, but after she damaged many a plate, she decided to design and build a machine that could handle the task—faster and more carefully.

According to a profile of Cochran on the U.S. Patent and Trademark Office website, she vowed: “If nobody else is going to invent a [mechanical] dishwashing machine, I’ll do it myself.” Although she had no technical background, she came from a family of engineers and inventors. Her father, John Garis, was a civil engineer who supervised a number of mills near the Ohio River in Illinois. Her great-grandfather John Fitch invented the first steamboat to be granted a U.S. patent. She designed her first model in the shed behind her house in Shelbyville, Ill. Her lack of formal engineering education, however, became an obstacle, so she sought out someone who could help. Mechanic George Butters agreed to assist her in building the prototype. To make the machine wash dishes efficiently, Cochran measured the width, height, and length of plates, cups, and saucers and constructed wire compartments for the china to sit in. The compartments separated each piece of dishware. At the bottom of the machine was a container that held soap. The compartments were placed inside a wheel that laid flat within a copper boiler, according to the Lemelson-MIT program’s profile of Cochran. A motor powered the wheel, which turned as soapy water was squirted on the dishes to clean them. Cochran was granted a U.S. patent in 1886 for her machine, which she named the Cochran dishwasher. She advertised her invention in local newspapers and built the machines for friends and family.

her connect with not only restaurants and hotels interested in buying her dishwasher but also with investors. Many potential investors asked Cochran to resign, however, so the company could be sold to a man, according to the Patent and Trademark Office article. She refused and continued to fund the business herself. To increase sales, Cochran displayed her machine at the 1893 Chicago World’s Fair, where she won an award for the machine’s design and durability. Thanks to that visibility, orders came pouring in and she was able to open a manufacturing facility near Chicago. Her dishwashers became popular with the hospitality industry, but it wasn’t until the 1950s that dishwashers caught on with the public. “Some homemakers admitted that they enjoyed washing dishes by hand, and the machines reportedly left a soapy residue on the dishes,” the LemelsonMIT article says. Many homes built before the 1950s used a furnace to heat water, and not all furnaces at the time could produce enough hot water to run a dishwasher. Thanks to changing attitudes about technology and housework, though, the dishwasher’s popularity grew over time. Cochran never saw her machines become sought-after household appliances. She died in 1913. In 1926 her company was acquired by KitchenAid, now a part of Whirlpool. Any IEEE member can submit a milestone proposal to the IEEE History Center. The center is funded by donations to the IEEE Foundation.

COCHRAN MEANT BUSINESS

—J.G.

To expand the market for her machine, she founded Garis-Cochran Manufacturing in the early 1890s in Shelbyville. The business was renamed Cochran’s Crescent Washing Machine Co. in 1897. It helped

This article originally appeared online as “This Socialite Hated Washing Dishes So Much That She Invented the Automated Dishwasher.”

THE INSTITUTE  |  MAR 2021  |  TI-13

briefings S.K. Ramesh

Francis Grosz

T

he IEEE Board of Directors has nominated Fellow S.K. Ramesh and Life Senior Member Francis Grosz as candidates for IEEE presidentelect. The winner of this year’s election will serve as IEEE president in 2023. Ramesh is a professor of electrical and computer engineering at California State University Northridge’s college of engineering and computer science, where he served as dean from 2006 to 2017. While dean, he established centers on renewable energy, entrepreneurship, and advanced manufacturing. He created an interdisciplinary master’s degree program in assistive technology engineering to meet emerging workforce needs. Ramesh is the founding director of the university’s nationally recognized TI-14  | MAR 2021 | THE INSTITUTE

Attract, Inspire, Mentor, and Support Students program, which advances the graduation of underrepresented minorities in engineering and computer science. He has been an IEEE volunteer for almost 40 years and has served on the IEEE Board of Directors, Awards Board, Educational Activities Board, Publication Services and Products Board, and Fellows Committee. As the 2016–2017 vice president of IEEE Educational Activities, he championed several successful programs including the IEEE Learning Network and the IEEE TryEngineering Summer Institute. He expanded chapters of IEEE’s honor society, Eta Kappa Nu (IEEEHKN), globally to serve all 10 regions,

and he increased industry support as the society’s 2016 president. Ramesh was elevated to IEEE Fellow in 2015 for “contributions to entrepreneurship in engineering education.” He serves on the board of ABET, the global accrediting organization for academic programs in applied science, computing, engineering, and technology, and is an experienced program evaluator. Ramesh has served IEEE Region 6 at the section, chapter, and area levels. He currently serves on the IEEE Buenaventura (California) Section member development team, which received a 2020 Gold Award for its work. His many recognitions include the 2004 IEEE Region 6 Community Service Award and the 2012 John

LEFT: S.K. RAMESH; RIGHT: FRANCIS GROSZ

2022 President-Elect Candidates Announced

J. Guarrera Engineering Educator of the Year Award from the Engineers’ Council. Grosz spent the majority of his career in industry before retiring in 2012. He also served as an assistant professor of engineering at the University of New Orleans for six years and an adjunct professor for two years, as well as an adjunct engineering professor at Tulane University, also in New Orleans, for two years. He designed systems for defense contractors Litton Data Systems, Omni Technologies, and the U.S. Naval Research Laboratory. He was granted two U.S. patents—one for a method of transmitting data through a ship’s bulkhead and the second for a NASA fiber-optic communication system for rocket engine testing. Grosz has been an IEEE volunteer for more than 35 years, serving at the section, region, and institute levels. He has held almost all offices at the section level, including chair, secretary, and vice chair of the IEEE New Orleans Section, and he has been a member of the IEEE Region 5 executive committee for 18 years. He served on the IEEE Board of Directors as the 2016–2017 Region 5 director and the 2019 vice president for IEEE Member and Geographic Activities (MGA). He was 2017 chair of the audit committee and cochair of the 2019 ad hoc committee on member engagement, which included three subcommittees examining member value and leading MGA efforts in realigning IEEE’s regions. Grosz, a member of IEEE-HKN, has received several recognitions including an IEEE Third Millennium Medal, the 2008 IEEE Region 5 Outstanding Member Award, and a 1999 NASA Space Act Award, which recognizes a technical innovation of significant value to the agency’s activities. An amateur radio operator, his call sign is K5FBG. JACOB ZIV

—J.G. This article originally appeared online as “S.K. Ramesh and Francis Grosz Run for 2022 President-Elect.”

IEEE Medal of Honor Goes to Data Compression Pioneer

I

EEE Life Fellow Jacob Ziv will receive this year’s IEEE Medal of Honor “for fundamental contributions to information theory and data compression technology, and for distinguished research leadership.” Ziv and Abraham Lempel developed two lossless data compression algorithms: Lempel-Ziv 77 in 1977 and LZ78 the following year. The two procedures enable perfect data reconstruction from compressed data and are more efficient and projects in science, technology, than previous algorithms. They engineering, and math. allowed for the development of Ziv has been a professor of elecGIF, PNG, and ZIP files. trical engineering since 1970 at the “The LZ algorithms were the first Technion Israel Institute of Technolmajor successful universal compres- ogy, in Haifa. He served as dean of the sion algorithms,” says one engineer EE faculty from 1974 to 1976 and vice who endorsed Ziv for the award. president of the school’s academic “These algorithms, and Jacob’s affairs department from 1978 to 1982. analysis of them, [have] formed the Born in Israel, he began his engibasis for most subsequent work on neering career in 1955 as senior universal algorithms.” research engineer in the scientific Ziv pioneered universal source cod- department of the Israel Ministry of ing, a series of algorithms that com- Defense, where he conducted R&D press data without needing to know in communication systems. anything about the inherent inforHe moved to the United States to mation. Such algorithms reduce the pursue a Ph.D. in electrical engineerrequired data rate needed to recon- ing from MIT. After he received his struct images from undistorted as doctorate in 1962, he moved back well as distorted data. to Israel to rejoin the Ministry of Ziv also contributed to the theory Defense and head its communicaof low computational complexity tions division. decoding of error-correcting codes. He returned to the United States in He has received several recogni- 1968 to join AT&T Bell Laboratories, tions including the 1995 Marconi in Murray Hill, N.J., as a member of Prize, a 2008 BBVA Foundation Fron- the technical staff. He left there in tiers of Knowledge Award, and a 2017 1970 to join the Technion. EMET Prize—known as Israel’s Nobel The IEEE Foundation sponsors the Prize—in the exact sciences category. IEEE Medal of Honor. In 1997 he established the Israeli —J.G. National Infrastructure Forum for This article originally appeared online Research and Development, which as “IEEE Medal of Honor Goes to Data strives to promote R&D programs Compression Pioneer Jacob Ziv.” THE INSTITUTE  |  MAR 2021  |  TI-15

of note

Countdown to the IEEE Annual Election

O

n 1 May, the IEEE Board of Directors is scheduled to announce the candidates to be placed on this year’s ballot for the annual election of officers—which begins on 16 August. The ballot includes IEEE presidentelect candidates as well as nominees for IEEE Standards Association presidentelect and IEEE board of governors members-at-large, IEEE Technical Activities vice president-elect, and IEEE-USA president-elect. New to the annual

2020 Election Results Here is the Tellers Committee tally of votes counted in the 2020 annual election and approved in November by the IEEE Board of Directors.

IEEE president-elect, 2021 K.J. Ray Liu: 21,120 Saifur Rahman: 15,781 S.K. Ramesh: 12,852 IEEE division delegateelect/director-elect, 2021 Division I Franco Maloberti: 1,838 Rakesh Kumar: 1,198 Samar K. Saha: 767 TI-16  | MAR 2021 | THE INSTITUTE

election ballot is the position of chairelect for the IEEE Women in Engineering committee. Those elected take office next year. IEEE members who want to run for an office but who have not been nominated need to submit their petition intention to the IEEE Board of Directors by 15 April. Petitions should be sent to the IEEE Corporate Governance staff in Piscataway, N.J. To ensure voting eligibility, members are encouraged to review and update

Division III Khaled B. Letaief: 2,177 Robert S. Fish: 1,694 Andrzej Jajszczyk: 917 Division V Cecilia Metra: 3,331 Elizabeth “Liz” L. Burd: 2,592 Division VII Claudio Cañizares: 3,432 Lalit K. Goel: 2,325 Division IX Ali H. Sayed: 2,161 John R. Treichler: 2,147 Fabrice Labeau: 1,498 IEEE region delegateelect/director-elect, 2021-2022 Region 2 Andrew D. Lowery: 1,644 Philip M. Gonski: 1,549

their contact information and election communication preferences on the election preferences web page (www.ieee. org/election-preferences) by 30 June. Given ever-changing global conditions, members may wish to consider voting electronically. For more information about the offices up for election, the process of getting on the ballot, and deadlines, visit the IEEE annual election web page (www.ieee.org/elections) or write to [email protected].

Region 4 Vickie A. Ozburn: 1,435 Tarek Lahdhiri: 826 Region 6 Kathy Hayashi: 3,091 Chris R. Gunning: 1,476 Scott M. Tamashiro: 1,155

IEEE Standards Association board of governors memberat-large, 2021-2022 Subhas Chandra Mondal: 918 Haiying Lu: 524

Region 8 Vincenzo Piuri: 6,643 Igor Kuzle: 2,536

IEEE Technical Activities vice president-elect, 2021 Bruno Meyer: 15,837 F.D. “Don” Tan: 14,484

Region 10 Chun Che “Lance” Fung: 5,686 Celia Shahnaz: 3,305 Supavadee Aramvith: 3,213 Norliza M. Noor: 2,235

IEEE-USA presidentelect, 2021 Deborah M. Cooper: 14,825 Keith A. Moore: 6,221

IEEE Standards Association board of governors member-atlarge, 2021-2022 Stephen D. Dukes: 978 Mehmet Ulema: 479

Because someone looks up to you.

Providing a secure financial future is in your hands.

IEEE Member Group 10-Year Level Term Life Insurance. To learn more*, visit IEEEinsurance.com/10Year *For information on features, costs, eligibility, renewability, limitations and exclusions. The Group 10-Year Level Term Life Insurance Plan is available only for residents of the U.S. (except territories), Puerto Rico and Canada (except Quebec). This plan is underwritten by New York Life Insurance Company, 51 Madison Ave., New York, NY 10010 on Policy Form GMR. This plan is administered by Mercer Consumer, a service of Mercer Health & Benefits Administration LLC. This coverage is available to residents of Canada (except Quebec). Mercer (Canada) Limited, represented by its employees Nicole Swift, Pauline Tremblay and Suzanne Dominico, acts as a broker with respect to residents of Canada.

Program Administered by Mercer Health & Benefits Administration LLC In CA d/b/a Mercer Health & Benefits Insurance Services LLC AR Insurance License #100102691 CA Insurance License #0G39709 95755 (3/21) Copyright 2021 Mercer LLC. All rights reserved.

MATLAB SPEAKS

MACHINE LEARNING With MATLAB® you can use clustering, regression, classification, and deep learning to build predictive models and put them into production.

@2021 The MathWorks, Inc.

mathworks.com/machinelearning