TOYOTA AUTOMATED DRIVING - Whitepaper | August 2020
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
TABLE OF CONTENTS Message from Toyota Research Institute CEO Gill Pratt 1 Introduction 2 The Promise of Automated Driving Technology 3 Core Principles 5 The Future of the Driver - Vehicle Relationship 7 The Automated Driving Landscape 9 Toyota’s Approach to Vehicle Automation 11 Automated Driving Deployment: Personally-Owned Vehicles 14 Automated Driving Deployment: Mobility as a Service 15 Elements of Automated Driving 17 Challenges Facing Automated Driving Development 20 Toyota’s Automated Driving Programs, Partnerships, and Investments 22 Technology on Toyota Vehicles 30
MESSAGE FROM TOYOTA RESEARCH INSTITUTE CEO GILL PRATT As Toyota President Akio Toyoda has said, “the threatening illness. Traffic crashes are sufficiently automotive industry is facing sweeping, once-in- rare to conceal the potential safety benefits of a-century changes.” They are driven by innovations automated driving, and news of technology- in four technology categories – Connected, caused crashes tend to be greatly amplified in Automated, Shared and Electrified – each with its human perception. Societal acceptance will likely own challenges and opportunities. take time and will not be easy. For automated driving, the challenges include As a result, some may ask why any company complex technical problems in perception, would take on the challenge of developing prediction and planning, and a range of social, a system for automated driving. For Toyota, ethical and policy issues for citizens, regulators two of the most compelling reasons are: 1) and lawmakers to consider. automated driving technology offers the chance to dramatically reduce the more than a While simpler to focus on technical issues, the million human-driven traffic fatalities that occur social, ethical, and policy challenges facing around the world every year, and: 2) automated automated driving are the most difficult. Since driving technology could provide mobility for no technology can perfectly predict human a rapidly growing segment of society – the behavior, some crashes will still occur wherever elderly – for whom loss of mobility leads to loss vehicles with automated technology travel of independence and lower quality of life. As we the same roads with human-driven vehicles have seen since Toyota first began researching and pedestrians. Even if automated driving automated driving in the 1990s, we continue to technology can be made, on average, much see the technology’s potential to help everyone safer than human driving, it may be difficult for get where they want and need to be... safely. society to accept the remaining inevitable injuries and fatalities when a vehicle with automated We hope the public shares our belief mobility can technology is involved in a crash. be a source of inspiration, and automated driving offers the chance to share joy in the experience Of course, society has faced similar questions of driving and a way to improve quality of life before. In medicine, for example, patients for everyone. Most importantly, because risk must sometimes risk adverse outcomes in can never be completely eliminated, we hope the hope of being cured. In general, we have the public will work with us as partners in the come to accept such risks, so there is reason to development of automated driving technology to hope a similar understanding could evolve for improve the safety, availability of mobility and, in vehicles with automated technology. But going turn, quality of life for as many people as possible. for a drive is not the same as fighting a life- Dr. Gill Pratt Chief Scientist and Executive Fellow for Research, Toyota Motor Corporation Chief Executive Officer, Toyota Research Institute 1
INTRODUCTION More than 80 years ago, Kiichiro Toyoda That is why we are working to help consumers launched a department inside Toyoda Automatic both enjoy the benefits of this research in the Loom Works to investigate what was then a near term, while we continue to work on the relatively new technology: automobile engines. most advanced technologies for the future. We This move, which laid the foundation for the have made advanced driver assistance systems Toyota Motor Corporation, reflected a critical (ADAS) available across nearly all new Toyota insight: Automobiles would revolutionize society and Lexus vehicles through Toyota Safety Sense – not because the automobile was a machine and Lexus Safety System+. These packages, that could move, but because it was a machine which include underlying automated driving that could dramatically amplify human mobility. components like Automatic Emergency Braking Today, technology is once again poised to and Lane Keep Assist, are available as standard expand what is possible for mobility. Connected or optional equipment on new Toyota and Lexus vehicles – the power of big data – sharing, vehicles in Japan and Europe, while nearly and vehicle electrification change the concept every model and trim level sold in the United of the car. At the same time, steep declines States provides it as standard equipment. in the cost of sensors, exponential growth in computing power and the rapid improvement of Artificial Intelligence (AI) systems have opened The value offered by today’s the door to advanced systems for automated groundbreaking technology is not driving. At Toyota, we see the extraordinary potential for this technology. Just as we did in the machines themselves, but in so many decades ago, we are remaking our what they offer to society. company to meet this new challenge. Importantly, Kiichiro Toyoda’s insight in the Looking ahead, we follow those same priorities 1930s remains true today. The value offered in our research. We are working aggressively by today’s groundbreaking technology is not to develop vehicles that will use the “Toyota in the machines themselves, but in what they Chauffeur” automated driving approach but offer to society. As a result, our ultimate goal is have also prioritized research and development not to create automation for cars, but rather to of nearer term “Toyota Guardian™” active expand autonomy for people. We are working safety technology. We believe Toyota Guardian to create the tools that will help people get – which uses Toyota’s underlying automated where they want and need to be, and to create driving technology to support and protect a a society where mobility is safe, convenient, driver by amplifying the driver’s capabilities – is enjoyable, affordable and available to everyone. an approach that can save more lives sooner, As we pursue this vision, we are guided by a prior to bringing Chauffeur to market. commitment to safety – both in how we research, develop, and validate the performance of vehicle As with any technology revolution, the impact technology and in how we aim to benefit society. of automated driving will go well beyond Despite their tremendous positive benefits, cars and trucks to include new business cars and trucks are involved in crashes that models and product categories, ranging from annually result in more than a million fatalities mobility service platforms to personal robotics worldwide. Our research into automated driving solutions. Major challenges remain, but we is an extension of Toyota’s long-standing are inspired to help lead the way toward the focus on improving automobile safety, and it future of mobility as we continue to focus on advances our ultimate goal of helping realize enriching lives around the world with safe a future without traffic injuries or fatalities. and responsible ways of moving people. 2
THE PROMISE OF AUTOMATED DRIVING TECHNOLOGY Toyota is guided by its Global vision, IMPROVING SAFETY which calls on the company to “lead the way to the future of mobility, enriching First and foremost, Toyota is committed to lives around the world with the safest and automated driving because the technology most responsible ways of moving people.” offers the promise of a world with far fewer This mission influences everything we casualties from crashes. According to the do, including making high-quality cars World Bank, there are about 1.25 million and trucks, constantly innovating and worldwide traffic fatalities annually and working to safeguard the environment. far more traffic injuries. Automated driving technology could drastically reduce this Automated driving is a natural next number by helping to prevent crashes step in Toyota’s growth as a mobility caused by human error. Our work on this company. It expands upon traditional technology follows from our commitment automotive capabilities to help people to safe driving with a goal of developing get to where they want and need to be, while offering the potential to benefit a car that would never be responsible all of society by helping to eliminate for causing a crash and that can avoid or traffic fatalities and injuries, reshape mitigate many crashes caused by other cities, reduce emissions, and achieve vehicles or external factors on the road. our ultimate goal: mobility for all. This focus on safety carries through to all of Toyota’s research and development of systems for automated driving. This means testing and validation to help ensure the proper performance of new technologies before their market introduction and a focus on expanding the adoption of potentially life-saving features to non-self-driving vehicles. 3
MORE EFFICIENT TRANSPORTATION A REVOLUTION BEYOND CARS Beyond their core safety benefits, systems Toyota believes the technology behind for automated driving may help make traffic automation will bring sweeping benefits and smoother and more efficient. This could provide mobility solutions that extend far beyond cars meaningful improvements to air quality through and trucks. AI offers the potential to revolutionize reduced emissions and to the quality of life of and improve the daily lives of millions by creating drivers through shortened and more predictable new categories of technologies and services. travel times. The technology would also expand Toyota is leveraging its research in automated the number of people able to experience driving to create new mobility solutions, such as the joy of driving, while also significantly robots with enhanced perception, reasoning and improving the quality of time spent in a vehicle manipulation that can offer expanded freedom of during routine commuting or long drives. movement for all, including people with limited Vehicle automation also holds the promise of mobility associated with age, illness or disability. increasing access to vehicle travel to a larger For example, assistive robots could someday portion of the world’s population through empower and enable people who might affordable, on-demand mobility models. otherwise be restricted in their ability to move Combining automated technology with around their homes or in their communities. Mobility as a Service (MaaS) will help increase These systems show promise in helping personal mobility, especially for people with seniors “age in place” with dignity rather physical limitations. This type of on-demand than moving into assisted living facilities – transportation could potentially transform cities an important potential benefit given current and support more economically vibrant and demographic trends in many countries. efficient communities. For example, parking areas in urban centers could be repurposed for people rather than vehicles, with MaaS vehicles using automation helping cities evolve into more environmentally-friendly spaces with greatly reduced emissions, traffic and noise. 4
CORE PRINCIPLES SAFETY APPROACH We accomplish this by working in each category to improve safety along three Since the 1990s, Toyota has worked to distinct pillars: people, vehicles and develop automated driving technologies the traffic environment. This approach through the framework of our Integrated allows us to expand our focus, looking Safety Management Concept. This beyond the potential of new vehicle safety approach focuses on technologies technologies to include expanded safety that could mitigate the risk of collision education programs and new partnerships at each stage of driving, including with governments and other stakeholders Parking, Active Safety, Pre-Collision that can improve the construction of Safety, Passive Safety and Rescue. roads and other traffic infrastructure. Collision Pre-Collision Emergency Parking Active Safety Passive Safety Response Panoramic View Dynamic Radar Cruise Blind Spot Monitor Pre-Collision Monitor Control (DRCC) (BSM) System (PCS) GOA Automatic Collision Vehicle Dynamic Notification (ACN) Control Pedestrain VDIM Injury-reducing Body Lane Keeping Assist Lane Departure Alert VSC TRC Back Guide (LKA) (LDA) Monitor Basic Function BA ABS Alert Pre-Collision Seatbelts Intelligent Adaptive VDIM Brake Assist Airbag Front-lighting System Adaptive High-Beam (AFS) System (AHS) Pre-Collision Braking Intelligent Parking Assist (IPA) Automatic High VSC Beam (ABS) Type Traction Control Regular Type: (TRC) PCS to help prevent Pop-up Hood Night View rear-end collisions Intelligent Clearance Sonar (ICS) Advanced Type: Brake Assist (BA) PCS to help prevent Navigation Cooperative ITS collisions with Coordination System pedestrians Antilock Braking System (ABS) Drive-Start Control ITS: Intelligent Transport Systems VDIM: Vehicle Dynamic Integrated Management PCS: Pre-Collision System GOA: Global Outstanding Assessment VSC: Vehicle Stability Control BA: Break Assist ABS: Antilock Braking System 5
In addition to looking forward, we also THE MOBILITY TEAMMATE CONCEPT strive to learn from the past. This includes Toyota’s development philosophy for automated studying actual collisions in order to meet driving is the Mobility Teammate Concept and push beyond the industry’s ever-higher (MTC), an approach built on the belief people safety standards. Toyota analyzes the causes and vehicles can work together in the service of vehicle-related collisions and occupant of safe, convenient and efficient mobility. injuries by using a range of investigation data. Using technologies like Toyota’s The MTC combines all of Toyota’s research into Total Human Model for Safety (THUMS), automated driving and merges it into a vision in we simulate crashes to help develop safety which people and vehicle “team up” to monitor technologies. In addition, we conduct and assist each other whenever necessary. In experiments on actual vehicles before we the near term, this approach capitalizes on the launch new technologies. Post-launch, we different skills humans and machines bring to evaluate the effectiveness of the technologies the challenge of safe driving. Indeed, thanks by assessing any crashes that might occur. to the power of connected systems and cloud- based technology, this sharing of responsibilities Collectively, this strategy has resulted in means intelligent vehicles might one day improve meaningful safety improvements, as each continually, with every car and truck potentially new advancement in technology or design benefiting from the experience of many drivers. expands the range of potential collisions the vehicle can prevent or mitigate. Importantly, MTC is a philosophy built on the belief people should have choices, and technology should amplify rather than replace human capability. Instead of removing humans from any engagement with their own mobility, it allows people to enjoy the freedom and joy of driving if they choose, while also benefiting from the capabilities of automated driving when they need or wish. 6
THE FUTURE OF THE DRIVER - VEHICLE RELATIONSHIP The emergence of automated driving lives we lead. Most importantly, the has raised new questions about car amplifies us – allowing us to travel the future of the personally-owned further, faster and with less effort in the vehicle and about the long-term company of and with the goal of joining relationship between car and driver. loved ones. What’s more, we control the car. We press the accelerator, and the car For some, this is an easy question with little moves us the way our legs do, but with significance. To them, a car is just a means more power. We turn the steering wheel, of transportation used to move from one and it is as if we ourselves are changing place to another, not much different than our body’s direction of exploration. a train or a plane. At Toyota, we see a car as something more – a feeling we know Each customization we make to our car many people share around the world that also helps to build a deep and meaningful is captured by the Japanese word “Aisha,” relationship between it and us. Our or “My car, I love it.” And we do not think car becomes the only one like it in the this relationship will cease any time soon. world: a unique reflection of our own life. We pamper, abuse and use our cars What is it about the connection extensively, and they carry the marks of between people and cars that this relationship: a sticker here, a coffee generates so much love? stain there, and that time where we made We believe it’s a relationship that is a tear in the back seat and never bothered fundamentally different from those to get it fixed. At a more social level, cars we have with many other machines. belong to families and are sometimes A car is a safe, personal space with a passed on to the next generation, carrying comfortable and relaxing environment. with them the histories of our lives and Almost everything about it, inside and relationships. Over time, a car stops out, can be customized to express our being just a mode of transportation unique identity. Over time, the car itself and becomes something we love. changes to reflect who we are and the 7
For Toyota, the relationship between cars Of course, there are also people who want and people is fundamental to everything a specific driving experience and expect the we do, including our research into vehicle to perform the way they want, when automated technology. We use the word they want. That may mean a sporty driving “teammate” to describe our concept for mode in some situations and a smooth automated driving because of our belief ride in others. We believe advanced vehicle car and driver can help each other to make technologies should respect these unique driving safe, comfortable and fun. and changing needs and respond with the capability a driver desires while maintaining What does this relationship mean in practice? the appropriate level of safety support. When it comes to safety, driving skills don’t just differ between people. Experience, age, medical In short, Toyota believes cars should learn from conditions or simple fatigue mean individual and be responsive to their owners. Even in a driving ability can change from moment to future where driving is automated, we believe moment. That’s why Toyota designs automated this connection means cars will continue to be driving technologies with the aim of meeting loved, with the relationship deepening as they these shifting needs, helping to support safe meet our needs and grow to reflect our unique driving regardless of the driver’s condition. tastes over the course of our lives together. 8
THE AUTOMATED DRIVING LANDSCAPE AUTOMATED OR AUTONOMOUS? By contrast, much of what is currently described in public discourse as an Over decades, vehicles have become more “autonomous vehicle” is often not truly sophisticated. Computers can increasingly “autonomous” from human oversight handle direct involvement with key or driving responsibility. Care in using functions, such as acceleration and these terms is important, as their braking. More recently, new technologies application to vehicles in the market have emerged to perform additional tasks, may impact consumers’ expectations such as helping to keep a vehicle in its lane and understanding about how those or intervening if a collision is imminent. vehicles perform. As we implement With the rise of systems that can perform these technologies in passenger some or all driving tasks, a host of terms vehicles, we believe it is important have entered the market to describe to describe accurately, or to use them. These include “automated,” terminology that suggests, the actual “highly automated,” “semi-” and function the vehicle can perform. We “partially autonomous,” “self-driving,” believe helping consumers understand and “driverless,” to name a few. automated driving technology is another way of promoting safety in its use. For general descriptive purposes, Toyota uses the word “automated” to describe a Regardless of whether one prefers wide range of technology that augments “automated” or “autonomous,” the or replaces direct human control during range of capabilities of these emerging some period of time. We use the term technologies cannot be described in “autonomous” to describe only those just one word. For that, international vehicles where an automated system standards organizations have developed can perform the complete dynamic documents to address technical driving task, essentially replacing the aspects of automation for vehicles. job of the driver, during some period of time (not necessarily indefinitely). 9
STANDARDS ORGANIZATION ACTIVITIES FOR DRIVING AUTOMATION One example of activity is SAE International’s issuance of recommended practice J3016. That document categorizes automation based on whether an automation system: • performs acceleration/deceleration or steering, • performs acceleration/deceleration and steering, • monitors and responds to the driving environment, • performs fallback of the dynamic driving task (functions required to operate a vehicle), • is limited by an operational design domain (conditions under which a vehicle can operate). SAE has defined levels of automation that range from no automation (labeled as “Level 0”) to full automation (labeled as “Level 5”) as displayed in the following chart. SAE J3016TM LEVELS OF DRIVING AUTOMATION LEVEL 0 LEVEL 1 LEVEL 2 LEVEL 3 LEVEL 4 LEVEL 5 You are driving whenever these driver support features You are not driving whenever these driver support are engaged — even if your feet are off the pedals and features are engaged — even if your feet are off the What does the you are not steering pedals and you are not steering human in the driver’s seat You must constantly supervise these support features, When the feature These automated driving features have to do? you must steer, brake or accelerate as needed to requests, will not require you to take maintain safety over driving you must drive These are driver support features These are automated driving features These features These features These features These features can drive the vehicle This feature are limited to provide provide under limited conditions and will can drive the providing steering OR steering not operated unless all required vehicle under What do these warnings and brake/ conditions are met all conditions AND brake/ features do? momentary acceleration acceleration assistance support to support to the driver the driver automatic lane centering lane centering traffic jam local driverless same as emergency chauffeur taxi level 4, but OR AND braking feature Example adaptive cruise adaptive cruise pedals/ can drive blind spot Features warning control control at the steering wheel everywhere same time may or may not in all lane departure be installed conditions warning Copyright © 2014 SAE International. The summary table may be freely copied and distributed provided SAE International and J3016 are acknowledged as the source and must be reproduced AS-IS. These levels provide valuable and useful guidance, and they serve an important role in facilitating clarity in global discussions and potential regulations around automation 10
TOYOTA’S APPROACH TO VEHICLE AUTOMATION Toyota’s research into automation is driven TOYOTA GUARDIAN™ by two related but distinct objectives: To more quickly realize the safety benefits 1) improving driving safety and of automated driving technology and to expand the ability of current drivers to 2) improving access to and the experience the joy of driving, Toyota is convenience of mobility. developing an approach to active safety We are pursuing two distinct concepts to called Guardian. Guardian aims to safely achieve those goals with the development blend vehicle control between the driver and an AI system, sharing roles to take of each built on similar perception, best advantage of their individual skills. prediction and planning technology. We call this blended envelope control. As in the flight control systems modern CHAUFFEUR fighter airplanes use, the human driver retains control, but the AI translates his or For those who cannot or choose not her commands to keep the vehicle within to drive because of age, infirmity or a defined safety envelope. Guardian will any other reason, we are working to constantly monitor the environment and develop automated driving technology step in to amplify the driver’s capabilities that will allow a vehicle to drive on and may intervene when a collision or its own without human oversight or other safety-critical incident is imminent. fallback responsibility. We call this This is not done by braking alone, as eventual capability Toyota Chauffeur. in current pre-collision systems, but by employing the full range of driving actions, We are designing Chauffeur to use an including accelerating, changing lanes and AI system to completely perform the preventing the driver from over-reacting to driving function. With Chauffeur, the a situation. Fundamentally, the system is human will not have a role for the task being designed to help prevent the vehicle of driving while the system operates from being hit, to avoid hitting anything (similar to SAE Levels 4 and 5). else and to stop it from going off the road. 11
Toyota has made significant progress applying • Navigating the traffic environment can at blended envelope control between the driver times be exceptionally complicated in ways and the system to automobiles. Rather than that go beyond developing technology that switching between the driver and the system can handle varied environmental conditions in the event of a safety issue, there is a near- and obey the rules law defines. A system fully seamless blend of both working as teammates capable of staying within appropriate lanes, to extract the best input from each. stopping and starting at traffic lights, and Guardian exists on a spectrum of capability navigating a route from one point or another that defines how much it can assist to protect would also encounter a traffic context defined vehicle occupants. This includes mistakes or by people and can change at any time. other errors made by the driver and from external At present, many deployments of Level factors on the road, such as other vehicles, 4-equivalent systems address these issues obstructions or traffic conflicts. The higher the Guardian capability, the greater the number and by operating in highly restricted Operational types of crashes it can help protect against. Design Domains, which limit the vehicles to operate in conditions, such as low speeds, fair For example, at a modest level of Guardian weather, simple geography, light traffic and few capability, systems like Lane Departure Alert other variables on the road, like pedestrians (LDA) and Automatic Emergency Braking (AEB) or cyclists. Wider deployment will require a can help prevent some crashes. At the highest significant technical and sociological leap so level, Guardian capability is being developed vehicles can recognize the road environment’s with the goal to ensure a human-driven vehicle physical elements and the social context and would never be responsible for causing a interactions between the people who use it. crash and also avoid many crashes that other vehicles or external factors would cause. We are designing Guardian to avoid many of these constraints. The underlying automated technology only intervenes when necessary. This allows human drivers to leverage their social awareness to navigate through the complicated and rapidly shifting traffic environment. In so doing, Toyota believes the Guardian approach may help us realize the safety benefits of automated technology sooner by enabling us to introduce higher levels of driver assistance into production vehicles in the near term as BENEFITS OF THE GUARDIAN MODEL we continue our progress toward Chauffeur. To ensure these benefits are available to as Guardian offers an important path to vehicle many drivers as possible, Toyota plans to automation because it avoids several challenges offer Guardian to others in the industry. Chauffeur faces. Chauffeur is an important technology that will make a meaningful impact In addition, just as Guardian is designed on the lives of people who cannot or do to amplify the human driver’s ability rather not wish to drive. But its development faces than replacing it, it might also add an extra significant technical and social challenges measure of support to another automated that will take time to address. In particular, the driving system, whether provided by Toyota or Chauffeur automation model faces several key any other company. For example, Toyota has challenges when applied to automated driving: announced its intention to include Guardian • Humans suffer from a basic psychological as standard equipment on all Toyota e-Palette challenge known as “vigilance decrement,” platforms we build for the MaaS market, which makes it hard to keep one’s attention offering intervening support regardless of which and awareness engaged when monitoring automated system a fleet buyer chooses to use. for rare events over long time periods. 12
What’s more, Toyota also believes Guardian envelope control, the vehicle might one day will, over time, build much-needed trust and leverage Guardian to help prevent a crash acceptance for highly automated driving when driving difficulty is higher than the technology. Automated systems are often driver’s capability to navigate safely. At times viewed with suspicion or fear because they of low driving difficulty, Guardian technology remove the human from vehicle control. will maintain vigilance to help make minor Guardian may foster a different reaction corrections (i.e., lane keep assist) to help because it is designed to amplify human keep a vehicle in its lane, for example. performance rather than replace it and because it may highlight those situations At present, Guardian’s technical abilities where the automated technology might be higher than the driver’s skill only in provided a safety benefit. some limited circumstances. Over time, as technology improves and the system learns, Guardian capability will grow steadily toward HUMAN DRIVER CAPABILITY AND an ideal goal of creating a vehicle that would ENVIRONMENTAL DIFFICULTY never be responsible for a crash, regardless of human driver errors. At the same time, Toyota’s research on automated driving Chauffeur capability will advance toward a goal technology goes beyond seeking to expand of being able to drive safely in all conditions. vehicle capabilities. It also considers the capability of human drivers and the Importantly, driving environments can be difficulty of the driving environment. extremely complex and difficult, and no automated system – regardless of how Importantly, driver capability and driving capable it may be – is likely to prevent crashes difficulty are not static. They rise and fall over entirely. A fundamental question yet to be time. Driver capability depends on factors, such addressed is “how safe is safe enough?” The as skill, fatigue and distraction level. Driving answer will depend on government regulation, difficulty shifts based on issues ranging from liability risks, societal acceptance and what weather to traffic or construction. Most of the is technically possible. In general, we believe time, driver capability is sufficient to prevent a systems providing Chauffeur capability might crash. It is the times when driving difficulty rises need to be significantly safer than average above the driver’s skill that a crash is likely. human drivers to be accepted by society. Automated technology offers the potential By contrast, Guardian capability might be to contribute at periods of both high and low judged against the standard of, on average driving difficulty. With the power of blended and as often as possible, “do no harm.” Far Future Driver Capability Near Autonomy Future Capability Possbile Crash #2 Present Possbile Crash #1 Time 13
AUTOMATED DRIVING DEPLOYMENT: PERSONALLY-OWNED VEHICLES In 2003, Toyota introduced its first millimeter sign recognition system covering more wave radar-based Pre-Collision System regions and countries than before. (PCS). Soon after, Toyota rolled out the This approach of beginning with system to more-affordable vehicles, personally-owned vehicles is a proven such as Prius. Developing advanced and valuable method for technology technologies first, then finding reasonable development, as it speeds the introduction ways to bring them to a more popular and of advanced systems that can help affordable range of vehicles, continues improve safety, reduce accidents and to be Toyota’s strategy. This two-axis ease traffic. Today, thanks to rapid approach allows us to rapidly spread the component and information technology availability of safety technology, and it also developments, we can often reach applies to automated driving technologies. mass-vehicle deployment of new safety The results can be seen most clearly in systems much faster than before. Toyota Safety Sense and Lexus Safety Toyota is committed in the near term System+. As research and development to bringing vehicles with automated progresses, future generations of both driving capabilities to market. Teammate, systems will continue to expand to widen targeted for commercial availability the range of safety systems and automated in 2021, is expected to initially enable technologies available to the general driver-supervised automated driving public. For example, in 2018, Toyota began (SAE Level 2) on highways. The system to roll out the second generation of Toyota will evaluate traffic conditions, make Safety Sense and Lexus Safety System+ on decisions and take action during highway vehicles benefiting from a major model driving. Potential capabilities include change. In this second generation, we changing lanes (with driver approval), added new capabilities to enhance safety, maintaining a safe distance from including the ability to detect pedestrians preceding vehicles and exiting highways. in low-light conditions, the ability to detect bicycles, and an expanded road 14
AUTOMATED DRIVING DEPLOYMENT: MOBILITY AS A SERVICE USING MAAS TO ACCELERATE In addition, MaaS provides a unique AUTOMATED DRIVING DEVELOPMENT opportunity to deploy vehicles equipped with SAE Level 4 automated systems Through its own programs and in sooner. Fleets do not need the same partnership with various companies flexibility as personally-owned vehicles in the mobility services space, Toyota and can be limited to simplified operating plans to leverage a range of MaaS domains, such daylight hours, good platforms to support and accelerate weather and/or known fixed routes. the development and deployment of automated driving technology. Toyota believes the deployment of a high level of automation to support MaaS can MaaS addresses one of the key help lower costs per passenger mile to challenges in developing automated create new waves of consumer demand driving systems – the need for significant and a virtuous cycle of affordable mobility, driving data to improve core technologies. safety and convenience. It can also help Initial component costs mean personally- improve automated driving technology owned vehicles with systems for automated and support greater societal acceptance driving are likely to be expensive and and consumer adoption. Taken together, sell in small numbers. This, combined the system will bring forward key benefits with the low usage rates of private of automated driving much faster than vehicles, limits how much data they through private ownership alone. will collectively generate in order to help improve the system. In a MaaS application, costs can be amortized across a fleet and higher utilization rates will increase data gathered. As a result, the system’s capabilities can be extended more quickly, benefiting both MaaS and personal vehicle applications. 15
AUTONOMOUS MOBILITY AS A • Toyota also plans to deploy automated SERVICE (“AUTONO-MAAS”) driving technology on the mobility networks operated by third parties. In August 2018, To fully realize the potential synergies we announced a new collaboration with of automated technology and MaaS Uber Technologies designed to advance and networks, Toyota is pursuing a range of bring to market autonomous ride-sharing technology and vehicle platforms. as a mobility service at scale. As part of the • e-Palette, first announced at the 2018 agreement, Toyota is developing an Autono- Consumer Electronics Show, is a purpose- MaaS fleet based on the Sienna Minivan built mobility commerce platform designed to platform, which will be equipped with Uber’s support Autono-MaaS business applications. Autonomous Driving System and use Toyota’s The open, flexible platform will be easily Guardian as a safety support system. Pilot- adapted to support a range of uses, including scale deployments are intended to begin ride-sharing, delivery and retail. Currently on the Uber ride-sharing network in 2021. under development in coordination with partners, including Amazon, DiDi, Mazda, Pizza Hut and Uber, e-Palette will be controlled by SAE Level 4 Automated technology or, if desired, by a user’s proprietary system for automated driving. In either case, e-Palette will include Toyota’s Guardian technology, which will act as a safety net. Toyota plans to deploy mobility solutions like the e-Palette at the Tokyo 2020 Olympic and Paralympic Games, which now take place in 2021. 16
ELEMENTS OF AUTOMATED DRIVING FUNDAMENTALS OF AUTOMATED (INU), among other inputs – to collect TECHNOLOGY and interpret information about the vehicle’s current situation and its In general, driving automation operates relationship to its environment. This via interactions between fundamental includes the location and movement of systems. Various combinations of these the full range of obstacles, both static systems enable vehicles to understand and dynamic, including infrastructure, the driving environment, make intelligent vehicles, pedestrians, bicycles and more. decisions and, in the case of higher The amount and complexity of data automation levels, navigate safely for analysis makes this one of the most to a destination. These fundamental challenging steps in automated driving. systems are described as follows: • Prediction helps the vehicle estimate • Localization and Mapping determines where other vehicles, pedestrians, where the vehicle is within its bicycles, etc. are likely to be in the environment. This requires building future. Often there are multiple possible a specialized map of the surrounding predictions (known as hypotheses). environment, either from scratch or Currently, humans do a much better by drawing from a baseline of prior job of prediction than any machine. knowledge that is well-understood and This area will continue to challenge the trusted to be mostly correct, and then development of automated driving. localizing the vehicle within that map. This system helps a vehicle correctly • Planning determines one or more interpret the data its sensors gather. safe courses of travel for the vehicle, including decisions such as which lane • Perception combines information to travel, where to position the vehicle from the Localization and Mapping relative to other dynamic objects and system with data from vehicle sensors how much space to afford obstacles. – including cameras, LiDAR, RADAR, Importantly, the Planning system makes Global Navigation Satellite Systems decisions about how to safely guide (GNSS) and inertial navigation units the vehicle under uncertain conditions, 17
such as when other vehicles on the road •A rtificial Intelligence (AI) is a broad term for may be blocked from view or if they behave technology that processes information and in unexpected ways. Multiple hypotheses makes decisions to achieve a certain goal. may lead to multiple possible plans, with the This may be accomplished via a rule-based ultimate choice depending on the actions system, such as if a vehicle perceives a stop of other vehicles, pedestrians and more. sign and follows a programmed command to stop, or via machine learning, in which a system • Control executes the planned driving might process large volumes of information trajectories set by the planning system, which are updated constantly based on • Computer Vision is the process of gathering new information. This is accomplished using information from sensors and using it to actuators that direct vehicle driving functions. perceive the surrounding environment. This process leverages AI to draw •Coordination communicates with other vehicles, knowledge from the data, identifying and the road infrastructure and cloud databases. differentiating individual elements, such •External Human Machine Interaction as cars, pedestrians, trees and roads. (e-HMI) manages the communication of • Predictive Algorithms are used to anticipate information between the vehicle and humans the likely behavior of other objects in the in the traffic environment. Importantly, while road environment, such as the expected communication between driver and vehicle is future position of another vehicle on obviously important, particularly in managing the road based on its current trajectory the transfer of vehicle operation, so too is and proximity to other vehicles. communication between the vehicle and humans outside the vehicle, such as drivers of • Decision Algorithms choose the vehicle’s other vehicles, first responders and pedestrians. proper path based on the predicted behavior of others on the road. Importantly, KEY TOOLS FOR AUTOMATED DRIVING decision algorithms must operate despite uncertainty, which varies based on conditions, An overlapping set of core technologies and including visibility and traffic congestion. tools help make the fundamentals of automated driving possible. The core technologies • Maps are baseline representations of the and tools are described as follows: core elements of the physical world the vehicle occupies. These can be generated • Classical controls, including proportional- ahead of time, such as high-definition maps, integral-derivative controllers and non- to be used by a vehicle when it enters an linear gain scheduling, are used for low- environment, or generated in real time (on-the- level controller tasks like set-point tracking fly) using algorithms, such as Simultaneous and regulation within actuators, such as Localization and Mapping (SLAM). steering actuator or throttle actuator. • Sensors gather data from the driving • Modern controls include state-space environment or from the vehicle itself. These representations, robust control, nonlinear include systems that gather data about the control and adaptive control. These mostly world, such as cameras, sonar, LiDAR and explicit mathematical equations solve RADAR; those that track location, such as GNSS; apriori (offline) for the control value. and those that monitor the movement of the vehicle itself, such as inertial measurement • Optimizations, such as model predictive control units or wheel speed and angle monitors. (MPC) or receding horizon control (RHC), are used to generate path trajectories, which lower-level controllers subsequently track. Optimizations rely on numerical methods to solve/converge in real-time for the control value. 18
• Actuators are used to control the physical • V2X Communication consists of a direct operation of the vehicle, opening or closing the information exchange between vehicles with throttle, turning the wheels or engaging the roadside traffic management systems and brakes. Importantly, while computers rapidly with pedestrians via dedicated short-range perform much of automated driving, actuators communication (DSRC) and/or via cellular are limited by physical constraints, including networks. These vehicle-to-vehicle (V2V), vehicle dynamics and the speed of the vehicle-to-infrastructure (V2I) and vehicle- actuator itself. Thus, automated drive systems to-pedestrian (V2P) communications share must account for the lag between issuing information about road signals, signs, road commands and the vehicle’s physical response. conditions and other vehicles or pedestrians that may be difficult to see. They can also alert • Simulation is used to test the performance drivers of approaching vehicles, pedestrians, of automated driving software in a virtual red lights, and slow or stationary vehicles. environment. Data gathered from real-world Vehicle-to-network (V2N) supports map data testing is used to recreate a variety of traffic generation and map data updates, as well scenarios digitally, which are then used as various kinds of information delivery and to test and measure system response and remote control. Together, V2X communications to promote proper operation. Automated provide an additional means for automation to driving software can be tested in both the gain knowledge about surrounding traffic. The car and simulator at the same time, while obtained information is combined with data engineers shift back and forth seamlessly from on-board sensors to enable the vehicle to allow for continuous integration testing. to make better decisions for vehicle control, Driving systems can also be operated in traffic safety, efficiency and driver interaction. multiple simulation instances at the same time, allowing for the relatively-rapid creation of billions of miles of training and test data. 19
CHALLENGES FACING AUTOMATED DRIVING DEVELOPMENT The industry faces various challenges For example, in Japan, SIP (Cross to realize and popularize automated ministerial Strategic Innovation Promotion driving. These include legal/regulatory Program) under the Japan Cabinet Office, developments, social-system reform identifies areas of collaborative work and and the time needed for societal promotes research and development acceptance, with the detailed situation among relevant stakeholders. SIP varying between countries and regions. collaboration scope includes dynamic mapping and cybersecurity. Automated driving technology also depends on other industries that provide For some technologies, regional or key technologies that together make up global harmonization and collaboration the automated driving ecosystem. Key across borders would be required subjects, listed below, typically reflect to establish a common shared geographical or cultural differences, foundation. With such a foundation, globally and regionally. Therefore, we can accelerate our technology one key and unique success factor development toward improving our for automated driving is the level of customers’ safety and toward enhancing collaboration and cooperation with freedom and mobility efficiency. various stakeholders who are not always actors in the traditional auto industry. Realizing and popularizing automated driving will call for appreciating their expectations. Industry collaboration in non-competitive areas that act as a foundation of systems for automated driving and vehicles, such as infrastructure or social systems, is an effective way forward. 20
LEGAL / REGULATORY SOCIAL CHALLENGES FRAMEWORK CHALLENGES • Regional Differences (Custom & Behavior, Tacit • Infrastructure (Traffic Design and Driving Manner and Rules, Ethical Perceptions) Management, Road Construction) • Social Acceptance (Safety Concerns) • Cybersecurity • Sustainable solutions • Data Privacy (Smart Cities, Urban Planning) • Safety Assurance (Design, Construction, Performance, Validation) • Social System (Vehicle Registration, Licensing, Driving Education and Training, Traffic Rules, Insurance, Law Enforcement, Crash Investigation, Safety and Emission Inspections, etc.) 21
TOYOTA’S AUTOMATED DRIVING PROGRAMS, PARTNERSHIPS, AND INVESTMENTS TOYOTA COMPANIES WORKING •T oyota Central Research & Development ON AUTOMATED TECHNOLOGY Laboratories: TCRDL contributes to the present and future businesses of the JAPAN Toyota Group by conducting research in a • Toyota Motor Corporation: TMC’s variety of fields. At the same time, it surveys Advanced R&D and Engineering Company global technology trends and explores leads the company’s efforts to develop new fields of science to propose a vision of automated driving technologies at the future that will lead to new businesses the global level, pulling together the and contribute to the advancement of resources and work of all the other entities science, technology and industry. listed here, in addition to organizing the research, development and testing of these technologies by all relevant functions at Toyota Motor Corporation. • Toyota Research Institute – Advanced Development (TRI-AD): Launched in March 2018, TRI-AD is a company Toyota, Aisin and Denso organized that works in collaboration with the U.S.-based Toyota Research Institute (TRI). The company’s mission is to leverage the cutting-edge research in AI for automated driving created by TRI into real products that can have a substantial positive impact on mobility for the world. 22
NORTH AMERICA It pioneers the development of Toyota’s Mobility Services Platform (MSPF), a global, cloud- • Toyota Research Institute (TRI): Toyota Research based digital ecosystem that provides the tools Institute is a wholly-owned subsidiary of Toyota necessary to bring to market mobility services. Motor North America under the direction of Dr. Gill Pratt. The company, established in • Toyota AI Ventures: Toyota AI Ventures 2015, aims to strengthen Toyota’s research is a Silicon Valley-based venture capital structure and has four initial mandates: subsidiary of Toyota Research Institute that 1) enhance the safety of automobiles, 2) invests in promising startups from around increase access to cars to those who cannot the world. The fund focuses on companies otherwise drive, 3) translate Toyota’s expertise developing solutions in AI, robotics, in creating products for outdoor mobility autonomous mobility, data and cloud into products for indoor mobility and 4) technology that share its mission of improving accelerate scientific discovery by applying the quality of human life through AI. techniques from AI and machine learning. • Toyota Motor North America Research and Development (TMNA R&D): Founded in 1977 in Ann Arbor, Michigan, and formerly a division of Toyota Motor Engineering & Manufacturing North America, Inc. (TEMA), TMNA R&D provides everything from engineering design and prototype building to testing and evaluation of vehicles, parts and materials for Toyota’s North American vehicles. It includes research divisions studying Next Generation Mobility and Society Research, Cloud Infrastructure Architecture, Intelligent Computing, Network, EUROPE and System & Software. It is also the home of Toyota’s newly established North American • Toyota Research on Automated Cars in Europe purchasing and supplier center. With facilities (TRACE): Toyota Motor Europe’s Advanced in Michigan, California and Arizona, TMNA Research team in Brussels collaborates with R&D also handles emissions certification, experts across Europe in the field of computer technical research, regulatory affairs and more. vision for automation. It is organized loosely around a lab structure named TRACE (Toyota • Toyota Collaborative Safety Research Research on Automated Cars in Europe). Center (CSRC): CSRC partners with leading universities, hospitals, research institutions Current partners include KU Leuven, University and federal agencies, with a focus on safety of Cambridge, CTU Prague, Max Planck Institute research projects aimed at developing and Saarbrücken and ETH in Zürich. Each partner bringing to market new and advanced safety contributes with unique research algorithms, and technologies. Research areas include active/ all elements are integrated into the experimental passive integration, human experience, vehicles under the responsibility of KU Leuven. driver state detection and big data/safety analytics, with results shared publicly Current activities include state-of-the-art with other companies and academia. deep learning algorithms for object detection, robust and precise tracking, and full scene • Toyota Connected: TOYOTA Connected leads segmentation and classification. Monocular Toyota’s global development of mobility and stereo-camera algorithms provide services for consumers, businesses and long-range depth measurements. The government, powered by vehicle data science, objective is real time-free space estimation machine learning and contextual data services. for path planning and vehicle control. 23
• University of Michigan Transportation Research Institute: An investigation into kinematics of minimally aware adult occupants exposed to Automatic Emergency Braking (AEB) and evasive maneuvers on a test track. • University of Iowa: Research attempting to measure the response characteristics and estimated benefit with respect to reduction in injury/fatalities of adaptive headlamp system that highlights detected pedestrians and bicyclists using both driver and pedestrian/bicycle simulator study. TOYOTA COLLABORATIONS FOR AUTOMATED TECHNOLOGY DEVELOPMENT • TASI – Indiana University, Purdue University at Indianapolis: A project attempting to TOYOTA RESEARCH INSTITUTE develop test scenarios and methods for the evaluation of vehicle road departure warning • Massachusetts Institute of Technology: and assist and control systems on a test track. The CSAIL-Toyota Joint Research Center projects range from autonomy to self-awareness. • Massachusetts Institute of Technology Research is aimed at furthering the development Age Lab: A project attempting to develop a of automated driving technologies. deep-learning-based, full-scene recognition of vehicle environment from a vision sensor. • Stanford University: Stanford’s SAIL laboratory Examples are vehicles, pedestrians, bicyclists, is engaged in research projects that include traffic signs, buildings, curbs, etc. human-computer and human-robot interactions. • University of Wisconsin: A project attempting The focus is on developing innovative and to provide a theoretical and mathematical impactful approaches, algorithms, and data. framework of how drivers communicate The approach includes research in perception, with each other in an intersection. learning, reasoning and interaction. • University of California at San Diego: A project • University of Michigan: TRI’s Ann Arbor research attempting to provide a computational facility with University of Michigan focuses on prediction model for a transfer of control research into enhanced driving safety, partner between the automation and the human robotics and indoor mobility, automated driver. The model has factors originated driving, and student learning and diversity. from human motor and perceptual behaviors, scenarios and environments. COLLABORATIVE SAFETY RESEARCH CENTER (CSRC) • University of Iowa – National Advanced • Children’s Hospital of Philadelphia, University Driving Simulator: A project attempting to of Virginia, Ohio State University: A project provide a meaningful and useful dataset attempting to quantify key occupant responses of driver behaviors when encountering (kinematics, kinetics and muscle activity) to situations where transfer of control between evasive swerving and emergency braking using automation and the human is required. both adult and child subjects on a test track. • Virginia Tech: A research study attempting to estimate the Residual Safety Problem after Integrated Safety Systems (ISS) is deployed in 2025. ISS consists of all active (auto braking for vehicle, pedestrian, bicyclist, lane keeping, etc.) and passive safety systems (advanced airbag, curtain shield airbag, roof strength, pedestrian protection active hood, etc.). 24 24
You can also read