A MATTER OF2.0FORD’S APPROACH TO DEVELOPING SELF-DRIVING VEHICLESUpdated June 2021OUR PURPOSEOUR PURPOSE IS TO HELP BUILD A BETTER WORLD, WHERE EVERY PERSON IS FREE TO MOVE AND PURSUE THEIR DREAMS.2CONTINUES TO BE OUR GUIDING PRINCIPLEAT FORD, we are building a self-driving service for ride hailing and goods delivery. Our goal? To ultimately make people’s lives better by providing a safe, trusted and affordable mobility solution.We know that the services enabled by self-driving technology will be nothing without establishing trust among the cities and people we serve. And, building trust while treating customers like family has been at the heart of what we do since 1903. Over the last century, Ford has designed vehicles with the customer at the center and manufactured safe, quality vehicles at high volumes to meet the needs of people all over the world. We’ve taken the same approach with building our self-driving business — working closely with cities, neighborhoods and the people who live in them to establish trust in an entirely new way to move.Since our safety report was first released in 2018, we have been bringing together all of the complex pieces needed to launch a self-driving service—the vehicle is just one part of the broader picture. We have also been developing a robust Automated Driving System (ADS) with our technology partner Argo AI; building an exceptional customer experience; the transportation-as-a-service (TaaS) software; and our fleet operations expertise—which will be key to a trusted, reliable service; as well as collaborating with government partners at a federal, state and local level to apply this technology to help solve the mobility challenges cities face.In 2018, Ford announced Miami, Florida, would serve as our launch city for self-driving services. Since then, we have established operations to launch our services in two additional cities: Austin, Texas; and Washington, DC. Not only have each of these cities become proving grounds for building our business and testing our technology, they have become home for us. In addition to our three launch cities, we are also testing the ADS with Argo AI in Detroit, Michigan; Palo Alto, California; and Pittsburgh, Pennsylvania. With testing being conducted in six markets, Ford and Argo AI have what may be the largest, most diverse active urban-testing footprint of any self-driving vehicle developer. Testing in multiple cities is important to building a self-driving service because it exposes the ADS to a variety of complex scenarios, behaviors and imagery, enabling the system to be able to scale quickly.Our goal is to continue to be transparent about the services we are building, and the impact they will have in our cities. Through this Voluntary Safety Self-Assessment (VSSA) report, you will learn about the progress we are making and our continued commitment to building these services in a safe and trusted manner.Self-Driving Defined: Throughout the VSSA we use the term “self-driving vehicle” to refer to a vehicle equipped with a Level 4 or Level 5 Automated Driving System (ADS). The terms Automated Driving System (ADS), Self-Driving System (SDS) and Virtual Driver System (VDS) have been used interchangeably in the past. Ford has decided to align with the Society for Automotive Engineering (SAE) J3016 standard’s usage of Automated Driving System, or ADS. Argo AI continues to refer to its system as Argo AI’s SDS.Dive Deeper + Stay Connected: You can dive deeper and continue to stay connected, as we launch our services by checking out Ford’s Self-Driven blog: https://medium.com/self-driven and Argo AI’s Ground Truth blog: https://groundtruthautonomy.com/.3TABLE OF CONTENTS5 Our Technology Partner: Argo AI 6 Ford’s History With Automated Drivingand Driver-Assist Technologies 7 Frequently Asked Questions: Ford’s Self-Driving Service 10 Frequently Asked Questions: Ford and Argo AI Testing 11 Mobility Services for People and Cities 12 What Our Self-Driving Vehicles Will Enable 13 How Self-Driving Vehicles Work 15 Safety, Reliability, Valuable Experience 23 Our Collaboration With Cities 24 Researching a Self-Driving Language 25 Safety Elements4Our Technology Partner:ARGO AIArgo AI is an artificial intelligence and robotics company, led by some of the industry’s most experienced self-driving vehicle experts. Ford is working with Argo AI to develop the “brains and senses,” or what Argo refers to as the Self-Driving System (SDS), for our self-driving vehicles.Argo AI is applying the latest advancements in computer vision and machine learning to help build safe and effective self-driving vehicles. The team also has extensive experience in commercializing robotics and sensing technologies with a history of designing and deploying integrated systems.With their global headquarters in Pittsburgh, PA, and engineering centers in Palo Alto, CA; Detroit, MI; Cranbury, NJ; and Munich, Germany, the 1,300 people on staff come from diverse educational backgrounds with a high concentration of Master’s and PhD degrees. To foster continued learning and enable the roboticists of tomorrow, Argo AI has formed research collaborations with Carnegie Mellon University and Georgia Institute of Technology.Argo AI’s ability to build a scalable software architecture with productionquality code from the start plus Ford’s expertise with vehicle integration and manufacturing give us the confidence to build high-quality self-driving vehicles. This unique partnership unites the benefits of a technology start-up with the experience and discipline of an automaker.You can read more about Argo AI and their safety report here, as well as their Ground Truth blog: https://groundtruthautonomy.com/.Ford refers to Argo’s SDS as the Automated Driving System (ADS)5FORD’S HISTORY WITH AUTOMATED DRIVING AND DRIVER-ASSIST TECHNOLOGIESFord has a long history of research, development and commercialization of driving automation systems, although we’ve clearly split them into dual pathways depending on how the technology is applied for customers. While this report is focused on describing our approach to SAE J3016 Level 4 self-driving or ADSs, the timeline highlights when some of our active safety systems and driver-assist1 technologies were introduced to the Ford and Lincoln portfolios.We use the term driver-assist technology to describe features that assist drivers with convenience or safety benefits, but they are not autonomous because they require the human driver to always remain engaged and available to take control of the vehicle. These features are defined by SAE J3016 as Levels 0-2 of automation, which leverage technology to augment a human driver but not replace them. For instance, the blind spot information system uses radar technology to detect if there is a vehicle in a driver’s blind spot, thus augmenting their ability to look and see for themselves.Active Safety and Driver-Assist Technology MilestonesLane departure warningBlind Spot Information System (BLIS)Adaptive Cruise ControlActive park assistForward Collision Warningwith Brake SupportCurve ControlPre-Collision Assist with Pedestrian Detection360-degree view cameraFord introduces Co-Pilot360, the most comprehensive standard active safety and driver-assist technology package available with Post-Collision Braking, Evasive Steering Assist and Lane CenteringPro Trailer Backup AssistAdaptive cruise control with stop-and-goFord Active Drive Assist2004200520062007200820092010201120122015201620172018201920202021Self-Driving Research + Development MilestonesFord employees volunteer at DARPA Grand Challenge to research technologypotential to improve safetyFord enters DARPA Grand Challenge with a self-driving Ford F-250Super Duty and qualifies for the finalsFord reaches the finals at the DARPA Urban ChallengeLaunch of fleet of ten Fusion HybridAutonomous Research Vehicles for testing anddevelopment1SAE now refers to Level 1 and Level 2 driving automation systems as “driver support features” as of the SAE J3016 2021-04 release. https://www.sae.org/news/press-room/2021/05/sae-international-and-iso-collaborate-to-update-and-refine-industry-recognized-sae-levels-of-driving-automationFord begins testing self-driving vehicleson public roads in MichiganAnnounces plan to launchcommercial service in 2021Ford secures a license to test self-drivingvehicles in California and opens itsPalo Alto Research and Innovation CenterBeginning of testing at McityInvestment and partnership with Argo AI for ADSdevelopmentPartners with Domino’s Pizza and Lyft to begin testing of self-driving customer experienceLaunch ofLaunch ofRevealesInstalls firstself-drivingself-driving fourth-generation Ford-designedbusinessbusinesstest vehicle andsmart nodeoperationsoperations launch vehicle platform: in Miamiin Miamiin AustinEscape HybridIntegratingExpansion ofLaunches moving fourth-generationself-drivinggoods pilot in Escape Hybrid testbusiness operationsMiami withvehicle into fleetsin Washington, DCfirst integration ofArgo AI’s self-drivingPartners with Postmatescapabilities with Ford’sand Walmart to researchcustomer-facing partnerships.self-driving deliveryexperienceInstalls first Ford-designedsmart node in Saline, Michigan,to begin researching complexintersectionsGiven the impact of COVID-19on consumer behavior, Fordshifts commercial servicelaunch to 20226FREQUENTLY ASKED QUESTIONS: FORD’S SELF-DRIVING SERVICE❚ WHAT DOES FORD MEAN BY SELF-DRIVING VEHICLE?Today, customers can benefit from experiencing automation systems that enhance safety or convenience in cars they can buy now. We call these features driver-assistance or partial automation, because they are designed to help improve the driving experience for a human driver and are set by the SAE International (SAE J3016) as Level 0-2. You may have heard of some of these features such as blind spot monitoring, lane keeping assist and adaptive cruise control. To be clear, these are not “self-driving” or “autonomous”—and we will never refer to them as such— because they require the human driver to supervise and maintain control of the vehicle.We use the term “self-driving vehicle” to refer to a vehicle equipped with an Automated Driving System (ADS) that meets the definition set by the SAE J3016 for Level 4 or Level 5 automation. The self-driving feature of our vehicle will be Level 4, meaning the ADS is capable of performing the dynamic driving task (DDT) without the need for a driver under specific conditions (known as an Operational Design Domain, or ODD – see page 28 for details). When the ADS is engaged, it has full responsibility for the task of driving.❚ WHAT DO YOU MEAN BY ADS?To make SAE J3016 defined Level 4-capable autonomous vehicles, which do not need a driver to take control, the car must be able to perform what a human can perform behind the wheel. Our ADS is designed to do just that, with the help of:• Sensors — lidar, cameras and radar to “see” all around the vehicle• Algorithms running on powerful computers to determine the traffic situation all around the vehicle and plan where to go next• Highly detailed 3D maps• Precise inertial measurement systems0No AutomationThe driver performs all driving tasks1Driver AssistanceThe driving automation system controlsacceleration/braking or steering while driverperforms remainder of the DDT2Partial AutomationThe driving automation system controlsacceleration/braking and steering while driverremains active and engaged3Conditional AutomationThe ADS performs the driving task, the driver must be ready to take control when requested and is responsible forDDT fallback4High AutomationThe ADS performs the entire DDT under specific conditions5Full AutomationThe ADS performs the entire DDT under all conditionsDriver Support feature performs part of the object and event detection and response (OEDR), but human is responsible for OEDR and monitoring the feature's performanceADS feature is responsible for the complete object and event detection and responseSOURCE: The SAE J3016 standard - Taxonomy and Definitions for Terms Related to On-Road Motor Vehicle Automated Driving Systems7FREQUENTLY ASKED QUESTIONS: FORD’S SELF-DRIVING SERVICE❚ WILL FORD’S SELF-DRIVING SERVICE BE TRULY DRIVERLESS?Yes, our self-driving service will eventually operate without the need for a driver or test specialist within a defined geo-fence service operations area. For our testing, we use test specialists who are responsible for monitoring the road ahead and monitoring the performance of the ADS.❚ WHEN WILL FORD REMOVE THE SAFETY OPERATOR?We are not in a race to remove the safety operator, which Ford and Argo refer to as test specialists – we are focused on doing this correctly and safely. We will work with Argo AI to assess the need for the test specialist and make a decision based on applicable regulations, safety performance milestones of the ADS, an appropriate level of community acceptance, and other factors to enable truly driverless operation within our defined service operation area.❚ WHEN AM I GOING TO BE ABLE TO EXPERIENCE FORD’S SERVICE?Ford and our technology partner Argo AI are currently testing with our self-driving vehicles on public roads with test specialists. Therefore, you might read about or see our selfdriving vehicle operating on city streets as we are testing. Ford is also building our business to move both people and goods in parallel. To do this, Ford will conduct business pilots for ride-hailing and goods delivery services; build our customer experience and fleet operations to ensure our vehicles are utilized and sanitized; and continue to collaborate with the local communities.ADS: Automated Driving SystemWe plan to launch our initial self-driving service in Austin, Miami and Washington, DC, in defined geo-fenced areas with plans to expand to additional cities. If you are within these areas you could potentially use Ford’s service.Over the next few years, Ford will continue to build its business and work with Argo AI to develop the ADS across multiple dynamic cities across the U.S. like Austin, Miami and Washington, DC. You can stay up to date on our business plans and latest news on our Self-Driven blog at medium. com/self-driven.❚ CAN I BUY A FORD SELF-DRIVING VEHICLE?No, our self-driving vehicles won’t initially be sold to customers in the way that cars are today. We believe we can offer the best value to our customers by providing the technology through a fleet service, similar to the way Ford currently offers specially engineered vehicles for taxi and police fleets. You’ll be able to experience these vehicles through a ride-hailing or goods delivery service.❚ WHAT VEHICLE WILL FORD USE TO LAUNCH ITS SELF-DRIVING SERVICE WITH?The vehicle platform and architecture customers can expect to see when we bring our service online is the Escape Hybrid. The team will modify the exterior and interior of the vehicle to build a safe and customer-centered experience – whether someone is having a package delivered or hailing a ride.❚ WILL THE INTERIOR LOOK LIKE THE ESCAPE DOES TODAY?No. We will be modifying the interior to build the best experience to move both people and goods.❚ WHY IS FORD LAUNCHING A SERVICE WITH A HYBRID VS. AN ALL-ELECTRIC VEHICLE?Ford envisions a future where our self-driving vehicles are all-electric, but we also need to find the right balance that will help develop a profitable, viable business model. The current state of battery technology doesn’t support our business or utilization model. While we will launch with a hybrid platform initially, we will progress to an all-electric platform as soon as it is viable for the business.8FREQUENTLY ASKED QUESTIONS: FORD’S SELF-DRIVING SERVICE❚ HOW IS FORD WORKING WITH THE BROADER INDUSTRY TO COLLABORATE TO SAFELY LAUNCH ITS SERVICE?Ford is a member of several business associations and consortiums whose members are also testing self-driving vehicle technologies. Some of these groups work with lawmakers, regulators, and the public to realize the safety and societal benefits of fully self-driving vehicles, while others seek to educate the public about autonomous technologies. Additionally, Ford participates in various consortiums addressing the safety challenges self-driving vehicles face through the development of industry best practices and standards (see page 15 for more details).Ford is also actively seeking a federal framework that will promote safety and innovation in the self-driving vehicle space. This is critically important to the future of mobility by building public trust in these technologies and ensuring our country’s competitiveness in this new automotive era. Among our ongoing autonomous vehicle (AV) regulatory and policy efforts are the Alliance for Automotive Innovation and the Self-Driving Coalition for Safer Streets, where we are working with industry to advocate for the safe and expeditious deployment of AVs while promoting the benefits of the technology.In 2021, Ford joined the National Highway Traffic Safety Administration’s (NHTSA) AV Test Initiative. Through this voluntary initiative, NHTSA aims to provide an interactive tool that will keep the public up to date on the states and companies using ADS technology on public roads, and we will support by sharing details about our self-driving vehicles, testing and operations.ADS: Automated Driving System SOURCE: https://www.selfdrivingcoalition.org/about❚ I LOVE TO DRIVE. WHY WOULD I WANT TO USE A SELF-DRIVING SERVICE?Building cars and trucks that are fun to drive is at the heart of what we do at Ford, and we’re not going to stop any time soon. But, we believe there are many benefits to using a self-driving service for ride-hailing including increased flexibility, control and freedom to make changes before or during your trip; control over having a private environment, providing increased productivity; and most importantly increased safety. Our technology will focus initially on urban areas where vehicle ownership is challenging and expensive, so our service will complement, not conflict. It can ultimately provide value and a consistent, enjoyable experience.❚ WILL SELF-DRIVING VEHICLES IMPROVE SAFETY?The U.S. Department of Transportation has projected that self-driving vehicles could significantly reduce the severity and frequency of crashes and fatalities. We believe that both the safety technology that we offer on our vehicles today and the self-driving vehicles of the future, have the potential to reduce accidents and injuries and save lives. But while we see tremendous potential in these technologies, they are not our only solutions. We continue to improve upon the design of our vehicles to ensure protection of all our passengers. We offer technologies to reduce distractions, like voice-enabled SYNC, and enable emergency services, like 911 Assist, which can help connect you to emergency services when you need it. And we work to improve road safety through driver education and training, like our Driving Skills for Life program.❚ WILL SELF-DRIVING CARS “SOLVE” CONGESTION IN OUR CITIES?There’s no one silver bullet for the problem of urban congestion, so we’re working on comprehensive solutions rather than thinking self-driving vehicles alone will solve traffic congestion. We believe self-driving vehicles will be part of that larger solution. For instance, we believe selfdriving vehicles have the potential to decrease the number of vehicles required on the road to move people and goods. In addition, their ability to access road information in real-time could help them take more efficient routes.❚ WILL PEDESTRIANS BE SAFE LIVING IN A CITY WITH SELF-DRIVING CARS?Self-driving vehicles should enable safer streets for everyone, including cyclists and pedestrians, not just those utilizing a vehicle. The ADS is constantly surveying its surroundings taking into account pedestrians, cyclists and other road users to predict how to best react. In fact, our perception system can detect, track and classify things on the road – including pedestrians, strollers, cyclists, mopeds and more.9FREQUENTLY ASKED QUESTIONS: FORD AND ARGO AI TESTINGADS: Automated Driving System❚ HOW DO FORD AND ARGO AI TEST AUTONOMOUSLY?Before operating autonomously on public roads, Argo AI maps the roads where it intends to test to build the ADS’s knowledge of the area. The map includes information such as traffic signals, speed limits and bike lanes. The team then conducts a rigorous testing method – including simulation testing and closed-course driving on a test track, and then the test specialists take the vehicle on public roads to safely test.❚ HOW ARE FORD AND ARGO AI SAFELY TESTING?Ford and Argo AI currently have test specialists in each vehicle – a driver and a co-driver. The driver in the leftseat is responsible for safely operating the vehicle by focusing on the road to monitor for pedestrians and other immediate objects external to the test vehicle. The co-driver in the right-seat maintains focus on a laptop that displays what the ADS’s sensors “see” to communicate and coordinate with the driver. As the ADS matures, it will meet safety performance milestones that will eventually enable reduction to a single driver, and eventually, driverless operation within its operational design domain.❚ HOW ARE FORD AND ARGO AI CONFIDENT THAT TEST DRIVERS WILL INTERVENE WHILE OPERATING AUTONOMOUSLY ON PUBLIC ROADS?The test specialists are highly trained and go through a rigorous training program, developed in partnership with Argo AI. The program improves manual-driving skills in the classroom and then verifies those skills driving on closed courses. Vehicle operator training also provideseducation on self-driving technology. The training program is designed to be difficult and requires candidates to pass tests to advance through each phase.❚ DO FORD’S SELF-DRIVING VEHICLES OPERATE IN SCHOOL ZONES?Yes. Working closely with city officials and schools to clearly communicate our operations, Ford and Argo AI’s self-driving vehicles can operate in school zones.❚ WHERE ARE FORD AND ARGO AI TESTING?Ford and Argo AI are testing across six U.S. cities, including: Austin, Detroit, Miami, Palo Alto, Pittsburgh and Washington, DC. ❚ WHY IS IT IMPORTANT TO TEST IN MULTIPLE CITIES? WHY NOT JUST TEST IN ONE TO TWO CITIES?Argo AI has what may be the largest, diverse active urbantesting footprint of any self-driving vehicle developer. This is important because it exposes the ADS to a variety of complex scenarios, behaviors and imagery, making it safer and enabling the system to be able to scale or grow more quickly.Additional information about how Argo AI tests can be found at https://www.argo.ai/how-we-test/.10MOBILITY SERVICES FOR PEOPLE AND CITIESWe recognize that just inserting new mobility technologies and services into a city or neighborhood won’t solve its existing challenges and may even make them worse. That is why we are committed to collaborating with city leaders where we operate. We are learning how each city works, what its needs are, and how our technology can help adapt and support each city’s unique transportation system. If applied correctly, new technologies can enable solutions to help improve the quality of life for everyone by providing additional access to transportation. We’re developing a portfolio of solutions that can help a city improve its transportation system and the ever-growing mobility options emerging every day.Self-driving services is one part of the solution to help improve mobility. For example, Fordowned Spin partners with cities, campuses and community groups to provide dockless e-scooter micromobility services in the U.S. and in Europe. Also, Ford’s Safety Insights platform tool, available through StreetLight Data’s platform, uses machine learning to combine crash data, proprietary vehicle data, and traffic volume data to help cities identify streets and intersections that could be improved to increase safety. Finally, Ford’s TransLoc provides technology solutions and orchestration software to transit providers in municipalities, universities, airports, hospitality and private business campuses.As we develop our mobility solutions, we aren’t ignoring how the COVID-19 pandemic has changed the world, including how people workand shop. Prior to the pandemic, many urban centers were dealing with record levels of traffic and pollution, and we began to see a shift from individual vehicle ownership to increased individual and shared mobility services. Due to the pandemic, we have seen a shift in consumer behavior, including a dramatic shift in people’s willingness to embrace goods delivery. In a June 2020 report, consulting firm Accenture found an increase of consumers using home delivery services with 45% saying they expect to continue to do so. In terms of shared mobility, now more than ever before, consumers are looking for a safe and sanitary environment.Coming out of a pandemic, our self-driving service fleet operations are going to be more important as people will likely continue to be wary of gathering in large numbers or in small spaces for some time. Understanding how people view mobility services will be critical to incorporating the right elements into our user experience – and that is one of the many things we are looking at as we build a safe service. No matter what, we will want to put people’s minds at ease by making sure they know they, or their packages, are in a safe and protected environment when they get in our vehicles.At Ford, we tend to think these changes will not simply go away as the virus subsides. Whether permanent or temporary, a change in customer behavior is something we cannot ignore as we build mobility services designed around people’s needs.11WHAT OUR SELF-DRIVING VEHICLES WILL ENABLETo understand the potential ways to use self-driving vehicles, we must first think about them in a different way from how most of us use cars and trucks today. Initially, self-driving vehicles will work best to move people and goods using a different business model – one where vehicles are accessed and shared as a service versus owned and driven. We think self-driving vehicles can enhance mobility services by being more accessible, affordable and by complementing a city’s existing transportation ecosystem. For example, these services could provide a solution for people who may not be able to drive themselves or provide a last-mile solution for a commuter. In the realm of goods delivery, we see an opportunity to help with the increasing demand generated by the convenience of ordering from retailers or supporting a business to provide additional ways to move goods. As they reach scale, self-driving vehicles have the potential to transform cities in countless ways – including improving safety, reducing congestion and making people’s lives better. Furthermore, people will have an opportunity to hail selfdriving rides to avoid parking costs and reduce traffic frustrations, while shoppers will seek efficient, effortless delivery via self-driving services. We believe that people and packages can move more safely, efficiently and affordably in the self-driving future.SOURCE: https://www.selfdrivingcoalition.org/12HOW OUR SELF-DRIVING VEHICLES WORKFor an ADS to achieve capabilities similar to a human, our test vehicles are fit with the latest in cutting-edge sensing and computing technology. Yet, the components that make up the vehicle itself—such as the suspension, brakes and electrical system—are also critical because a self-driving vehicle must be designed and developed as an integrated system that works in unison. Otherwise, we’re not able to achieve the needed performance levels capably and reliably. Today, we’re testing with our third-generation Ford Fusion Hybrid and our fourth generation Ford Escape Hybrid self-driving test vehicles. Each vehicle is equipped with hardware that make up the ADS. This system is responsible for the perception, motion planning and controls, which together enable the vehicle to drive itself, including planning a route, detecting and tracking objects, and anticipating how other road users will behave. • T he perception system is made up of sensors including cameras,radar and lidar and compute, that enable the ADS to understand the traffic scene around the vehicle. • T he motion planning system is a sophisticated set of software and hardware that combines the perception information and with route objectives to determine potential vehicle trajectories through a traffic scene. • T he control system then follows the desired trajectory by directing the powertrain, braking and steering systems to guide the vehicle.13HOW OUR SELF-DRIVING VEHICLES WORK (cont.)MAPSOne fundamental element of our self-driving vehicles is the detailed high-definition map of the roads on which they will be operating. The ADS uses maps to aid localization, perception, prediction, and motion planning decisions.The maps contain detailed information beyond just roads and turns — they also contain detailed 3D information of the entire environment built for Ford’s self-driving vehicles.In fact, the map imagery is created by the self-driving vehicle’s sensors as the test specialists manually drive our test vehicles throughout a city. During these initial drives, sensors built-in to the vehicle scan the roads, sidewalks and buildings, allowing for the creation of a 3D map. The base map is annotated with all the relevant traffic regulations and guidance, including road edges, directions of travel, speed limits, traffic signals, stop signs, crosswalks and yielding relationships. This map forms the basis of a complex system of perception and tracking, so the onboard computing system can use it to understand the static world before it even starts driving autonomously.PERCEPTIONJust like human drivers, the ADS needs information to operate. The ADS takes sensor data from the cameras, lidar and radar to infer what is happening in the world around the vehicle. The system observes other road actors, such as vehicles, pedestrians and bicyclists, noticing how they move and what their likely intended actions are. The ADS also detects road debris that the vehicle must avoid and, of course, detects the state of traffic lights relevant to the vehicle motion. Achieving human-level perception is a tremendous challenge and requires bringing state-of-theart machine learning techniques to bear on the problem. The ADS combines information from all sensor sources in a carefully designed and thoroughly tested architecture that provides redundant sensing 360° around the vehicle. Redundancy is essential because different sensors have different failure modes.MOTION PLANNING AND CONTROLAt the core of self-driving is decision making. In order to drive, the ADS needs to continually receive information from the perception system, evaluate and select from a range of driving actions and execute those precisely to enable it to drive efficiently and safely. The vehicle takes actions that allow it to drive in a contextually safe manner that is predictable to other road users.To do this, the vehicle continually evaluates different maneuver possibilities. Would it be better to slow down and follow a slow-moving vehicle or change lanes? The best decision depends upon a range of factors including traffic laws, the predicted motion of other road users and the vehicle’s own driving route. Changing into a faster moving left lane only makes sense, for instance, if the vehicle is not planning to make a right turn a short distance ahead.When planning any maneuver, the ADS plans to keep a safety envelope around the vehicle at all times. This envelopedescribes a safe following distance behind other road users and a lateral margin that gives other road users appropriate space as the vehicle drives by. In the final stage of the motion planning and control system, the ADS determines appropriate steering, brake and acceleration commands so that the vehicle executes the desired maneuver as designed and enforces the safety envelope.REMOTE ASSISTANCEArgo AI has established Remote Guidance capability to provide human support to the ADS in the event there is a unique and challenging situation which requires additional guidance – for example, an unexpected road closure. Using cellular connectivity, the operator will assess the event and deliver information to the self-driving vehicle. As this is not the operator controlling the vehicle remotely, the ADS still executes the driving task based on guidance from the human operator when it is safe to do so.14SVAARFLEEULTIAAYBBLIELIETXYPERIENCE2PEGASUS: Project for the Establishment of Generally Accepted quality criteria, tools and methods as well as Scenarios and Situations for the release of highly-automated driving functions 3ISO: International Organization for Standardization 4PAS: Publicly Available Specification 5SAE: SAE International, previously the Society of Automotive EngineersAs we develop our self-driving service, we are focused on three ideals that we believe are crucial to earning trust: safety, reliability and valuable experiences. Ford is committed to the promotion of the safe implementation of selfdriving vehicles not only in the U.S., but around the world. We are steadfast in cooperating with others who share our belief in ensuring the safe operation of self-driving vehicles. To this end, we have joined various consortiums, including the following:AVSC – Ford is a founding member of the Autonomous Vehicle Safety Consortium (AVSC), convened by the SAE International and SAE Industry Technologies Consortia (SAE ITC). This consortium is working together to advance safer testing, development, and deployment of AVs through the establishment of industry best practices. PEGASUS2 – Along with other AV developers, Ford has joined PEGASUS, in the effort to identify areas for international collaboration and harmonization on safety assurance for ADS, and initiate steps to make progress towards collaboration and harmonization. AMP – Ford has joined with other AV developers in the U.S., to form the Automated Mobility Partnership (AMP), which is geared towards the use of real-world driving data to inform the development of ADS systems. Ford continues to play a leadership role in the development of various industry automotive standards, including: ISO3 26262 – Road vehicles – Functional safety ISO/PAS4 21448 – Road vehicles – Safety of the intended functionality ISO/SAE5 21434 – Road vehicles – Cybersecurity Engineering Safety is a top priority in developing the processes we use for designing, testing and manufacturing our vehicles. We believe that the safety processes we have developed and demonstrated throughout Ford’s history are key to earning trust. We’re applying those time-tested processes to the development of self-driving vehicles, and we’ve created additional, new safety processes for self-driving technology. We are also designing for reliability through durability testing, using our extensive experience in building dependable vehicles that operate day after day, in hot or cold weather, and under a variety of road conditions.Our end goal, as our self-driving service comes together, is to create an experience that people enjoy, find valuable and trust.15SAFETYAPPLYING SAFETY PROCESSESWhile the field of self-driving vehicle development may only be a decade old, Ford brings more than a century of experience in vehicle safety processes to the table. In partnership with Argo AI, Ford is integrating decades of vehicle safety experience with the latest advancements in industry best practices around the integration of vehicle hardware and software that enable autonomy into a safe, efficient and enjoyable experience for our customers.TEST SPECIALIST TRAININGThroughout the development period, safety operators, or test specialists, are utilized in our test vehicles until the ADS meets the necessary performance and safety milestones to operate without them. We currently have two-person teams – a driver in the left-seat and a co-driver in the right-seat – in all our test vehicles. Also, before anyone is put in the driver or passenger seat of our test vehicles, they go through rigorous training and certification. The goal of the training, defined with our partner Argo AI, is to develop a standard professional framework for test specialists through a program that improves manual-driving skills in the classroom and then verifies those skills driving on closed courses. Test specialist training also provides education on self-driving technology. The training program is designed to be difficult and requires candidates to pass tests to advance through each phase.Our test vehicles make it easy for test specialists to override the ADS in any situation. If the test specialist controls the vehicle in any way, by steering, throttle or braking, the vehicle returns full control with notification to the driver. The test specialist can also disconnect the ADS or disable the powertrain completely by pressing a fail-safe button in the center console. Once the test specialists complete training, they continue to work in teams of two with the test specialist in the left-hand seat focused on the road ahead and the test specialists in the right-hand seat monitoring the performance of the ADS via a laptop computer. Furthermore, test specialists cannot use a cell phone unless the vehicle is in “Park.”We also subscribe to a philosophy of continuous education and learning. There are mandatory pre- and post-shift daily briefings to share information amongst the teams, and there are regular classroom sessions to educate about new software releases and capabilities. Test specialist training is a critical element of deploying our development vehicles, and this is a view also shared by the AVSC. We have helped establish the AVSC Best Practice for safety operator selection, training, and oversight procedures for automated vehicles under test.16DESIGNING FOR RELIABILITYRELIABILITYFor self-driving vehicles to be embraced, the public needs to trust the technology. People tend to gain trust in something when they can predict what it will do. Part of earning the public’s trust is for the self-driving vehicle to drive in ways that other motorists, cyclists and pedestrians expect; we call this “naturalistic driving.” Each city has its own culture and expectations for how vehicles should drive. That’s why we drive extensively in the actual cities where we will eventually deploy our self-driving vehicle fleet before we allow the vehicles to drive autonomously with the test specialists. In this way, we learn the roads, the challenging intersections and the local behaviors, so we can develop the system to operate as people expect. As a company, we’re known for delivering dependable vehicles that meet and exceed customer expectations, in both private vehicle sales and fleet sales, like police, taxi and commercial trucking services. We are bringing that same focus on reliability to our self-driving vehicles.To make our vehicles capable of self-driving, we redesigned and modified them to work with the ADS. We have added several new sensors and computing systems to understand the traffic and environment around the vehicle and make decisions for what to do next. We call this group of components and software the ADS. We also modified the traditional components to allow the ADS to control the vehicle with its computer-generated commands and capably handle system faults. We call this adapted set of traditional components the Autonomous Vehicle Platform, and it’s designed to be applied to a broad range of different vehicles types.AUTONOMOUS VEHICLE PLATFORMContrary to what some might think, building an autonomous car isn’t as easy as putting sensors and a computer into an existing car. The vehicle’s traditional components and associated software must be developed and engineered to work together with the self-driving hardware and software to deliver a great experience. The end result needs to be a high-quality, energy-efficient vehicle people trust to serve their needs — just as they trust Ford vehicles today. Here’s one of the challenges we face: in today’s cars, if the rare event of a power failure, the driver is still able to mechanically push the brake pedal and steer to bring the vehicle to a controlled stop. Many of our self-driving vehicle systems have redundancies and are designed to be fail-functional. For instance, we have multiple power systems to ensure the ADS and platform receive power. If there is a loss of operation in one system, the ADS can still bring the vehicle to a controlled stop.LONG RANGE LIDAR FAR-FIELD CAMERAMICROPHONES RADARRADARSHORT RANGE LIDAR NEAR FIELD CAMERARADARRADAR SHORT RANGE LIDAR NEAR FIELD CAMERA RADAR MICROPHONESCAVS (COMPUTE MODULE)AVS (COMPUTE MODULE)SOURCE: ARGO AI17RELIABILITYOUR NEXT-FGorEd N TEST VEHICLE and Argo AI’s fourth-generLataioyninseglf-dtrhiveingFtoeustnvedhaictleion for Ford and Argo AI’s fourth-geOneurartiSonesleflf--dDrivrinivg itensgt veBhiuclesinessGen 4 FFAEouosfI-csrrtdaiehopvenaee-nHaaHdlyreytbAdbrsrrigdieitdonsspseAnillnaefI-gtwhdfoaerairvsvnmetidn.sTnibbgnTceeehethotlweneeefem-eessnrdmlytsepastordvstuittpveeeeotedihesminfsntirtttgcasgothvGTnilnetweeteesheitsetsnuwheeencet.spgi’oecsrrc4Ivtmypesnmhleaestovsin2thpsntnreeoeo0tiocadmhcfsolnc2rtotliseoretche0vpgewmenl“ef,oeyslheetcFa.mr,tsi’oaarcouiTbemlnwettnrehiuuidrgpsnncecreihocaglaiebwaotnr-nsieeErilenniiopldtzn“sstitonhelaottceaArsvataaunotoierwttonpthguieunw”cneoeer-hrg.bi-lieilnnltiteoevnoetuw”r illHybrid is also the architecture and bpelantefeodremd tFoosurdpphoartsccomhomseercnialization. to bring its autonomous vehicle service online.The systems we’re incorpoGraenti1nginto our newest test vehiclesGen 1are “launch-intent” in termsof the components we believewill be needed to supportcommercialization. What thismeans is that with a well-defined architecture andplatform in the Escape Hybrid,our team can continuouslytest and refine performance tobetter prepare us for laGuennc2h.Gen 2Everything we learn while usingthem can be channeled directlyinto our self-driving serviceas soon as it starts servingcustomers.We are integrating thesefourth-generation vehicles intoour multi-city testing effortsalongside our Fusion Hybrids inAustin, Detroit, Miami, Palo Alto,Gen 3Pittsburgh and WashingtoGne,nD3C.We have said it before, but the vehicle is just one part of bringing together the future of self-driving services. With our fourth-generation test vehicle, we have everything we need from a vehicle to stand up our self-driving service.Upgraded Tiara Sensing SuiteUpgraded Tiara Sensing Suite• Ahseilgln-hsneien•rwgAhsreeiltgllnsoo-hsonneihelnrguewgr-tlertpilosaooopnnnhlrguge1o-t2elrpvi8oaLipndn-iDbrge1o2eeAva8aLiRd-miDbeweAaaiRtmhwith 360-de36g0re-edevgierewe viewUpgraded Sensing Suite • The addition of new near-field cameras andUpgradesdhoSrte-nrasnignegLSiDuAiRteto help improve detection of • The addfiixteiodnaonfdnmeowvinnegaorb-jfeiectlds cclaomseer troasthaenvdehicle short-r•aNnegaer-LfiieDldAcRatmoehraeslpanimd sphroorvte-radnegteecLtiDioAnRof fixed anldoomkionvgianhgeoabdjaencdtstocltohseesridtoe othf tehveevheihcilcele.Rear-facing sensing suite that keeps track of• Near-fiewldhacta’smgeorinags oannbdeshhinodrtt-hraenvegheicLlieDAR looking ahead and to the side of the vehicle. Rear-facing sensing suite that keeps track of what’s going on behind the vehicleAttention to Detail Sensor Cleaning At•tHeindtdioennt, ofoDrceetadi-laSirecnlseoarnCinlegacnhinagmbers • Hi–dd3e6n0, fdorecgerde-eaicroclveearnainggechambers –• 3In6c0redaegseredescporvaeyrangoezzles and coverage • Incarreeaassedfosrpirmaypnroozvzeledscalnedancoinvgerageareas for improved cleaning• Increased air pressure to aid • Inccrleeaasneidngaisr ppreeesdsure to aidcleaning speedIncreased Electrification Capabilities • Next-gen efficient hybrid powertrain withIanncurnedaesrefldooErlleiqcutirdi-fciocoalteidobnaCttaepryadbeisliigtines • •MNoedxiftie-gdehnigehffviocliteangtehbyabttreidrypwoiwtheartdrdaiitniownaitl h banttuernydcerllfsl,owohr ilciqhuhiedl-pcsosoulpepdobrtatthteery design •tsoeMtlfao-lddpriifoviweindegrhsryiegsqhtueivmroemlwtaehngileteshbreealqptutineirgreydtowbryiettdhhueacedditional gbasaotltienerycocenlslusm, wphtiiocnh helps support thetotal power requirements required by the self-driving system while helping to reduce gasoline consumption18RELIABILITYSENSING THE ROAD AHEAD… AND ALL AROUNDVehicle DirectionFord’s self-driving vehicles incorporate multimodal sensing that provides a 360-degree view of monitoring the environment in which they operate. It incorporates lidar, camera and radar providing overlapping perception capability, ensuring that at least two sensor types are monitoring at a minimum of 200m range – with some sensors exceeding 400m, which is more than three football fields in length. This means more than one type of sensor is observing the same area, and each detecting it a different way. The radar helps identify objects and the rate of speed of other vehicles. The cameras generate an image that can help the ADS recognize, classify and track objects including pedestrians and bicyclists.CAMERA COVERAGE - GROUND PLANE Near Field Cameras Far Field Cameras VehicleLIDAR COVERAGE - GROUND PLANE Short Range Mid Range & Long Range VehicleRADAR Mid Range Long Range VehicleSOURCE: ARGO AI19RELIABILITYKEEPING OUR SENSING SYSTEMS CLEANKeeping our sensors clean from rain, dirt, debris and even insects is very important to ensure our vehicles can better “see” the world around them in a variety of challenging conditions.We are constantly researching and improving upon our systems to ensure they are safe, capable and precise. Think about all of the dust, dirt, grime and bird droppings that accumulates on your windshield while driving. The same thing happens on our self-driving vehicles on the “tiara,” the structure that sits on top of all our self-driving vehicles and holds the collection of cameras, lidar and radar that helps the car “see” where it’s going. It is important that the tiara is always clean so that it provides the best picture of the world while driving. As a result, we wanted to create a cleaning system that could be fully integrated into the self-driving vehicle’s tiara.The initial cleaning system we developed for our third-generation self-driving vehicle test fleet feature next generation nozzles next to each camera lens that can spray washer fluid as needed to clean the sensors. The cleaning system uses advanced software algorithms that help our self-driving vehicles determine when a sensor is dirty and it can specifically hone in on one or multiple dirty sensor lenses to efficiently clean each one. After a sensor has been sprayed down, our tiara has an innovative way of drying the sensor lenses. It releases air through an air nozzle which quickly “dries” the face of the lens.We initially tested the cleaning system using the bug launcher, a tool our team of engineers at Ford created to test and understand how we can best clean insects off. The bug launcher uses compressed air to propel the bugs onto the sensors.The cleaning system was equipped on our third-generation self-driving test vehicles in Detroit, Austin, Pittsburgh, Miami-Dade County and Washington DC to test in a variety of environments. As a result of the real-world testing, we have refined the cleaning system for our fourthgeneration self-driving test vehicle. Some of the key new enhancements include: • Hidden, forced-air cleaning chambers that surround the camera lenses and LiDAR sensors toensure their field of view is clear while providing 360-degree cleaning coverage • Increased the number of spray nozzles and coverage areas for improved liquid cleaning; and • Increased pressure to aid cleaning speed.Additionally, we’ve extended these new cleaning designs to our added near-field cameras and lidar sensors as well. With these enhancements and improved hydrophobic coatings, our latest test vehicles are much better equipped.20VALUABLE EXPERIENCESOUR SERVICE: RIDE HAILING AND DELIVERYFORD believes self-driving vehicles have the potential to make a positive impact in our communities by providing greater access to transportation and delivery services.Collaboration is at the center of our approach as we build our business in each of our launch markets. We are committed to working closely with local officials and the community to understand how our service offerings can integrate within the existing transportation system and how they can provide value. To do this, we are continuing to conduct pilots that will help us stand up our self-driving business to move people and goods. For example, in Miami we wanted to build upon our past pilots and also find a way to help the community as it deals with the challenges of the global pandemic. As part of this, we teamed up with a local non-profit to deliver fresh produce and school supplies to local Miamians. This was the first time we integrated the self-driving capabilities from Argo AI with our customer-facing partnerships. As a result, this provided us meaningful real-world insights into what is required to run an efficient business.Ford recently teamed up with The Education Fund, a local non-profit in Miami-Dade County, building on previous business pilots to now conduct deliveries with our self-driving vehicles in autonomous mode.21VALUABLE EXPERIENCESCUSTOMER EXPERIENCE: THE CORE OF OUR SERVICESelf-driving technology has tremendous potential to change people’s lives and the communities in which they will operate. In addition to developing the technology in a safe and reliable manner, we are making sure it is being applied in ways that will deliver experiences that are valued. As we design the ride-hailing and goods delivery services that self-driving technology will enable, we strive to put people at the center of our thinking. We want to ensure that self-driving technology works to enhance people’s lives. Our customer experience will be our differentiator and is at the core of our service. We are building a great experience that offers our customers the peace of mind of knowing they, or their packages, are safe and in a protected environment. The foundation of this experience will be a Transportation-as-a-Service (TaaS) software platform, powered by the cloud, that can integrate seamlessly with other businesses, dispatch vehicles and orchestrate the fleet. We must provide our riders and goods-delivery customers transparency, with access to key info including ETA, route flexibility and control. To do this, we’re beginning to design and test customer experiences in parallel with our efforts to develop the technology.22OUR COLLABORATION WITH CITIESEvery city has unique transportation needs and FORD understands you can’t just copy and paste a service from one city to another.By working closely and collaboratively with city officials, community leaders and local businesses, we will build and deploy a self-driving service that is part of the city’s transportation ecosystem.In addition, Ford will invest in the local community through real estate development, job creation and provide a mobility service that has the potential to reduce congestion and accidents. Before a single one of our test vehicles starts driving autonomously in a city, Ford, Argo AI and the city work together to identify where self-driving cars could make the biggest difference, considering a number of factors, including where the demand is, what the market wants, and how we can meaningfully contribute to the transportation network already in the city.For example, we are testing smart infrastructure technology using Ford-designed smart nodes in Miami to research complex intersections and scenarios our selfdriving vehicles may encounter. The node is equipped with sensors like radar, lidar and cameras and situated above the intersection, so it can offer a bird’s-eye view of the area to support a self-driving vehicle as it analyzes its surroundings at a street level. While it is not required for a Level 4 self-driving vehicle to operate safely, this sensor node can quickly relay even more information to a self-driving vehicle, providing an additional layer of information about the situation it’s about to encounter long before it approaches the intersection. We worked closely with Miami-Dade County, the City of Miami Beach and the Florida Department of Transportation to make this possible, and they have been key collaborators.Ford’s smart node installed just below the street sign on the right at a complex intersection in Miami’s South Beach.Austin Assistant City Manager Gina Fiandaca speaks at press conference announcing plans for Ford’s self-driving service launch in Austin, Texas.23RESEARCHING A SELF-DRIVING LANGUAGEToday, pedestrians and bicyclists can look into a car and make eye contact with the driver to assure themselves that they are seen. A driver usually nods their head or waves a hand to indicate it’s okay to cross the road. But, what about in a driverless future?We believe a standard way for selfdriving vehicles to easily communicate their intention to people on the streets where they will operate could help communities trust and interact with the technology.As a result, we set out to develop a common language that self-driving vehicles can share with other road users. To communicate with people outside the vehicle, we designed a light bar to be placed in an area where a pedestrian or cyclist may look for cues from a human driver. We continue to investigate what light bar signals are most effective, such as the full light bar being in a steady on state when the vehicle is accelerating or driving at speed.Communicating effectively with limited assumptions on common language, or symbology, is very challenging. There are thousands of languages throughout the world, and many countries have more than one official language, which makes it nearly impossible to ensure text-based signals can effectively communicate with the greatest number of people.Additionally, human-driven vehicles typically communicate what they are doing – not what others, such as pedestrians, should do. Red brake lights indicate a vehicle is slowing or stopped. A turn signal indicates it is turning right or left. So, it’s important to build off the current applications of light communications on vehicles today. Also, many regulators around the world restrict what colors can be used for lighting on the front of a vehicle, so white or amber is the best option.To test out the signal patterns in the real world, we worked with the Virginia Tech Transportation Institute to create a “simulated” autonomous vehicle. Through our testing, which encompasses more than 180 hours and 2,300 miles in a dense urban area, we have results that suggest there may be a societal benefit to creating a standardized communications method.Our light signals still need to undergo additional research, but we believe development and adoption of a global standard is critical to support the deployment of self-driving vehicles. Ford is leading an International Organization for Standardization (ISO) initiative to develop the ergonomic design guidance for external visual communication from self-driving vehicles to other road users.
下载完整pdf文档
关注微信
