Download PDF Summer Bridge: A Vision for the Future of America’s Infrastructure June 15, 2018 Volume 48 Issue 2 The articles in this issue, by academic and industry experts, focus on what’s needed to prepare US infrastructure systems for the coming decades. The Role of Infrastructure in an Automated Vehicle Future Friday, June 15, 2018 Author: Ryan J. Harrington, Carmine Senatore, John M. Scanlon, and Ryan M. Yee Automated vehicles (AVs) have the potential to revolutionize road transportation. In the United States, approximately 94 percent of all crashes can be attributed to human error (Singh 2015) and the cost of crashes is more than $250 billion annually (Bamonte 2013). Automated vehicles are anticipated to reduce the number of crashes and also improve mobility for underserved segments of the population, reduce commute burdens, and increase road use. As AV development and deployment continue to advance, a parallel and synergistic opportunity to improve roadway infrastructure exists. This article considers opportunities and challenges associated with the improvement of roadway-related infrastructure to support automated vehicles. Introduction In 1997 the government of Sweden announced Vision Zero, a national target of zero traffic deaths (Swedish Ministry of Transport and Communications 1997). Since then, a number of other countries have adopted this goal. In the United States the National Highway Traffic Safety Administration (NHTSA), Federal Highway Administration (FHWA), Federal Motor Carrier Safety Administration, and National Safety Council are collaborating to achieve the goal (NHTSA 2016). The FHWA, which has oversight of the construction and maintenance of the nation’s highways, bridges, and tunnels, plays an important role in any improvements to this infrastructure, which can help maximize the safety benefits of automated vehicles. Accidents can be contextualized in terms of contributions from the human, vehicle, and environment. Progress toward removing the human from the operation of a fully automated vehicle will increasingly emphasize the roles of the vehicle and the environment, including the roadway and other infrastructure. Infrastructure improvements might help address challenging operational design domains (ODDs) such as driving in the snow at night with little visibility or correctly interpreting differing traffic controls and signage across the 50 states. While the technology is still maturing, the aging US infrastructure, which also faces funding uncertainty, is being pressured by continuously and quickly evolving AV technology. Fortunately, although automation is posing challenges, it is also revitalizing the conversation around infrastructure and its role in the transportation ecosystem. This article briefly describes the state of AV technology, the levels of automation for AVs, and the sensing suites used to perceive the environment, including the surrounding infrastructure. It explains how the current infrastructure can be modified to improve AV performance, and then reviews challenges to continued progress. The State of AV Technology The last few decades have seen the emergence and deployment of active safety systems throughout the automotive industry. Unlike passive safety technologies, such as airbags and seatbelts, which aim to protect occupants in the event of a crash, active safety systems aim to proactively mitigate or eliminate crashes altogether. AV technologies comprise the most advanced active safety systems. Rather than relying solely on the driver, they are designed to assist or take full control of safely operating the vehicle. A vehicle’s level of automation is defined by the -level of human monitoring and supervision required and the ODD in which the vehicle is capable of operating. The available technologies to achieve a certain level of automation vary in complexity, and different combinations of sensors and actuators can be used to achieve the same level of automation. In other words, a given level of automation is not defined by the suite of sensors and actuators installed on a vehicle. The technologies involved are highly proprietary and at times cost sensitive, so systems differ across manufacturers and vehicle models. Figure 1 The current standard adopted by NHTSA, SAE J3016 (SAE International 2016), defines six levels of automation (figure 1), ranging from none (level 0) to a fully automated vehicle (level 5). The simplest active safety systems (SAE level 0) rely solely on the driver to operate the vehicle. Common examples of level 0 automation are lane departure warning (LDW) and forward collision warning (FCW) systems that deliver an alert to the driver in the event of an imminent lane departure or frontal impact, respectively, but rely on the driver to take evasive action (General Motors 2018; Mercedes-Benz USA 2018; Tesla 2018; Volvo Car Corporation 2015). The most advanced technologies on the market are SAE level 2, partial automation; SAE level 3 systems, for conditional automation, are slated to enter the market later in 2018 (Audi Media Center 2017). An example of level 2 automation is adaptive cruise control in conjunction with a lane keeping assist (LKA) system (Audi Media Center 2017; General Motors 2018; Mercedes-Benz USA 2018; Tesla 2018; Volvo Car Corporation 2015); together, they maintain the vehicle’s position on the roadway and relative to other vehicles. However, to ensure the safe operation of the vehicle, the driver must continue to monitor the driving environment. Although several companies are testing highly automated vehicles, consumers cannot currently purchase vehicles that are capable of operation without any -driver supervision (SAE levels 4 and 5). How AV Systems Work The function, hardware, and objective of AV systems vary widely across the industry, but the framework for how they help navigate vehicles is largely consistent: The environment is monitored using a combination of sensors (e.g., cameras, radar, ultrasound, and lidar). The vehicle’s onboard computer processes the information relayed from the sensors and combines it with global positioning system (GPS) data, the known vehicle state (e.g., speeds, orientation, steering, brake application), and 3D mapping data to estimate the vehicle’s absolute position. These steps create a virtual representation of the world, which includes the subject vehicle as well as all other road users (including bicyclists, pedestrians, and other vulnerable road users), objects, and their intended path. The vehicle determines an appropriate course of action (e.g., avoiding a collision) while obeying -traffic laws. In the most basic systems (SAE level 0) the vehicle monitors only a narrow set of variables (e.g., distance from the closest-in-path vehicle) and simply delivers a warning to the driver when certain conditions are met. The most advanced technologies actively plan the path of the vehicle and can modify the lateral and longitudinal dynamics of the vehicle within that space. AV technologies rely on information sent from vehicle-based sensors to provide the driver with extra “eyes” that continuously scan the area around the vehicle. Cameras, for example, are widely used in AV systems. With machine vision techniques, incoming video data can be processed in real time to identify objects (including their location and trajectory) or determine the position of the vehicle in the roadway (Dabral et al. 2014; Lee 2002; Papageorgiou and Poggio 1999). LKA and LDW both use cameras to detect lane markings for determining the lateral position of the vehicle in the roadway. But when the camera is unable to detect lane markings—during poor weather conditions (e.g., snow), inadequate lighting conditions, or on roads without lane markings—LKA technologies may be ineffective. This is a particularly limiting factor given that a third of drift-out-of-lane road departure events occur on roads without lane markings (Scanlon et al. 2016a). Radar, lidar, and ultrasonic sensors are also involved in AV technologies to determine the location of roadway obstacles and other roadway users (Hatipoglu et al. 2003; Pakett 1994; Schubert et al. 2010; Tesla 2018). They are capable of varying detection ranges, field of view, and resolution, but all require line of sight and are therefore of limited effectiveness in detecting oncoming vehicles in certain driving scenarios. As an example, a vehicle making a left turn at a signalized intersection may encounter sightline restrictions due to vehicles in the opposite lanes (Scanlon et al. 2017). And when vehicles approach from lateral directions at intersections, line of sight may be impaired by roadside objects such as signs or foliage, or by roadway geometry such as curves or hill crests (Scanlon et al. 2016b). GPS data are used to determine vehicle position in the roadway, but accuracy limitations and degraded signal near buildings or other obstructions compromise the reliability of these data (DOD 2008). Automated vehicles will use data fusion techniques to incorporate GPS data with other sensing equipment, such as -cameras or inertial measurement units, to improve reliability (Caron et al. 2006; Chang et al. 2010; El Faouzi et al. 2011; Milanés et al. 2008). Automated vehicles are expected to eventually close the gap with humans in terms of adaptability and resilience to unstructured environments and to be capable of operating, initially, in selected “geofenced” environments and under prescribed ODDs. The Role of Infrastructure Current infrastructure is designed and built to accommodate human abilities and information needs. Road signs, for example, are sized and positioned based on human perception capabilities in relation to speed limits and local traffic patterns. To align with advances in AV technologies, the infrastructure will likely need to evolve in three ways: (1) account for AV sensing capabilities, (2) provide complementary sensing capabilities, and (3) adapt to the requirements of transportation modes enabled by AVs. AV technologies are currently being designed to operate with little or no support from the infrastructure, but the burden of perception and path planning will be increasingly shared and integrated with the infrastructure. Some argue that AVs should be capable of navigating using the same infrastructure that human drivers use today. But technologies can equip AVs with sensing range and accuracy beyond human drivers’ capabilities. For instance, humans drive with limited exchanges of information with other human drivers, but vehicle-to-vehicle communication can facilitate AV navigation and planning by sharing information, even in the absence of line of sight. Deeper integration of vehicles and infrastructure will increase AV sensitivity to infrastructure conditions and inconsistencies, while at the same time granting additional layers of robustness, making AVs arguably safer. Technological Enhancements for Infrastructure Certain physical infrastructure elements such as lane markings, signage, and signals can be designed to facilitate AV perception and interpretation. Infrastructure can also act as a distributed sensor network, supporting data sharing and providing information to vehicles. And technologies such as variable speed limits, traffic detection at signalized intersections, and traffic signal coordination are already moving the infrastructure in this direction. It is expected that this digital infrastructure will become the cyberphysical backbone for AVs: using an Internet of Things approach, it will be capable of sensing the environment and sharing useful information with vehicles (figure 2). For instance, precipitation sensors may alert AVs to potentially hazardous driving conditions, and smart traffic cones may be capable of repositioning themselves safely on the road while communicating to nearby vehicles about their placement and the reason for their presence. Figure 2 A constant exchange of information between vehicles and the infrastructure will facilitate the updating of digital maps in real time. Many AVs now rely heavily on such maps to ascertain precise location and safely navigate the environment. With the environment continuously changing—because of road work, local road closures, weather, and other factors—access to updated maps in real time has direct repercussions on AV performance. The constant exchange of information between infrastructure and AVs can facilitate the identification of nonconformities and road hazards, establishing a virtuous cycle of data sharing that benefits the safety and mobility of both drivers and the public at large. Finally, with the introduction of AVs, the infrastructure will have to accommodate new driving behaviors and traffic patterns. A prime example is parking. In London, an estimated 8,000 hectares of land are occupied by parked cars. However, in a driving landscape dominated by AVs it may not be necessary to find a parking spot close to the drop-off location since vehicles will be able to drive away to park (if necessary) where space allows, thus operating similarly to a taxicab. As an additional benefit, AVs will enable better use of land allocated for parking by parking closer to each other. Challenges As AV technology is continuously progressing, infrastructure changes will have to accommodate new and unforeseen technologies. The increased interaction between technologically sophisticated vehicles and infrastructure will require closer collaboration between the automotive, technology development, and infrastructure communities as well as road owners and operators, transportation planners, and federal, state, and local agencies. Although updating the infrastructure can be daunting and expensive, its benefits will likely extend beyond AVs to human drivers as well. The difference in the deployment time horizons for sensor and vehicle technologies, often measured in years, and for infrastructure, measured in decades, will create planning, design, and funding challenges. Current infrastructure decisions will impact and define AV operation for decades to come, so communication and coordination among the automotive, technology development, and infrastructure communities will be essential: The infrastructure community will benefit from a better understanding of current and future AV technology needs, which will allow the implementation of infrastructure enhancements that can support the safe and efficient operation of AVs into the future. The automotive and technology development communities should consider and design within the context of infrastructure planning, funding, and maintenance. Technology developers should plan for the availability and deployment of future infrastructure. The infrastructure community needs to stay abreast of vehicle and sensor technology development to understand how infrastructure may impede or accelerate the adoption of sensor technologies and AVs. As data sharing between vehicles and infrastructure expands, securing and leveraging these data communications will require coordination among the three communities. In the short term, the most relevant infrastructure features for AV safety, efficiency, and performance should be identified and evaluated in the context of the level of automation. For instance, well-maintained lane markings are critical for LKA technologies. Harmonization of lane markings, signage, and traffic signals across all states is equally important. The type and periodicity of maintenance and repairs (e.g., road markings and pavement quality) need also to be considered for the effective implementation of AV technology. As an example, careful attention should be paid to the conditions of road signs to ensure maximum visibility in all seasons and weather. As technology evolves, some constraints may be relaxed, but even if a new technology is more robust to infrastructure inconsistencies, full market penetration will be gradual and could take decades. Therefore the needs and limitations of the current AV fleet must be considered well into the future. The infrastructure community also needs to assess the impact of AVs on road capacity and land use. Will AVs increase or reduce vehicle miles traveled (VMT) and thus road use? How will AVs affect traffic flow and volume? How will land use change as the need for surface and garage parking evolves? These are only a handful of questions that need to be addressed in anticipation of the release and widespread adoption of AVs. Finally, infrastructure for automotive transportation is under pressure from increasing vehicle electrification. AVs are not necessarily electric, but electrification in the automotive field is gaining traction and the infrastructure must account for this regardless of AV penetration. Electrification can offer synergies with certain aspects of AVs—for instance, by streamlining data sharing—and will play an increasingly important role in infrastructure. Conclusion In a fast-paced technological landscape, it is challenging to identify the needs of the next 50 years. And at a time of uncertainty in infrastructure funding, it may be even more difficult to plan and implement infrastructure for AV technologies that are still in development. In remote areas with low traffic volume, for example, it may be cost prohibitive to install adequate infrastructure to fulfill current AV needs. On the other hand, future AVs are expected to be more robust and resilient to infrastructure deficiencies and may be capable of navigating those remote areas even without particular infrastructure support. Given the current trajectory of AV technology, infrastructure modifications could enhance and expedite the development and deployment of these systems to support the vision of zero road traffic fatalities. Achieving this vision will require collaboration between the automotive, technology development, and infrastructure communities as well as federal, state, and local agencies. It is vital to plan for and implement infrastructure solutions that are agnostic to specific technologies, benefit both AVs and human drivers, and prioritize short- and medium-term needs while keeping a long-term view. Areas for immediate action include traffic control harmonization, continuous engagement between -parties, and pilot demonstration projects. Uniform signage and road marking across jurisdictions can be achieved through the updating and implementation of the -Manual of Uniform Traffic Control Devices. Infrastructure planners and engineers should maintain constant communication with AV developers to make sure they have their finger on the pulse of the industry and understand AV needs. This process will necessarily evolve through pilot demonstration projects that can inform the interaction between infrastructure and AVs while offering opportunities to engage and educate the public. The limitations of today may be overcome by the breakthroughs of tomorrow. Acknowledgment The authors would like to acknowledge Cameron Fletcher for her in-depth review and skillful editing of this article. References Audi Media Center. 2017. On autopilot into the future: The Audi vision of autonomous driving. Ingolstadt, Germany. Bamonte TJ. 2013. Autonomous vehicles: Drivers of change. Transportation Management & Engineering, July 24. Online at https://www.trafficandtransit.com/autonomous-vehicles- drivers-change. Caron F, Duflos E, Pomorski D, Vanheeghe P. 2006. GPS/IMU data fusion using multisensor Kalman filtering: Introduction of contextual aspects. Information Fusion 7(2):221–230. Chang BR, Tsai HF, Young C-P. 2010. Intelligent data fusion system for predicting vehicle collision warning using vision/GPS sensing. Expert Systems with Applications 37(3):2439–2450. Dabral S, Kamath S, Appia V, Mody M, Zhang B, Batur U. 2014. Trends in camera based automotive driver assistance systems (ADAS). Proceedings of the IEEE 57th International Midwest Symposium on Circuits and Systems (MWSCAS), pp. 1110–1115. DOD [US Department of Defense]. 2008. Global Positioning System Standard Positioning Service Performance Standard, 4th ed. Washington. El Faouzi N-E, Leung H, Kurian A. 2011. Data fusion in intelligent transportation systems: Progress and challenges—A survey. Information Fusion 12(1):4–10. FHWA [Federal Highway Administration]. 2012. Manual of Uniform Traffic Control Devices. Washington. General Motors. 2018. Super Cruise. Online at www.cadillac.com/world-of-cadillac/innovation/super-cruise. Hatipoglu C, Özgüner Ü, Redmill K. 2003. Automated lane change controller design. IEEE Transactions on Intelligent Transportation Systems 4(1):13–22. Lee JW. 2002. A machine vision system for lane-departure detection. Computer Vision and Image Understanding 86(1):52–78. Mercedes-Benz USA. 2018. S-Class Operators Manual, P-6515-2164-13.pdf. Stuttgart. Milanés V, Naranjo JE, González C, Alonso J, de Pedro T. 2008. Autonomous vehicle based in cooperative GPS and inertial systems. Robotica 26(5):627–633. NHTSA [National Highway Traffic Safety Administration]. 2016. US DOT, National Safety Council launch “Road to Zero” coalition to end roadway fatalities. Press release, October 3. Washington. NHTSA. 2017. Automated Driving Systems 2.0: A Vision for Safety. Washington: US Department of Transportation. Pakett AG. 1994. Smart blind spot sensor. Google Patents. Papageorgiou C, Poggio T. 1999. Trainable pedestrian detection. Proceedings of the 1999 International Conference on Image Processing (Cat. 99CH36348), October 24–28, Kobe, Japan. SAE International. 2016. Taxonomy and definitions for terms related to driving automation systems for on-road motor vehicles, J3016_201609. Warrendale PA. Online at https://www.sae.org/standards/content/j3016_201609/. Scanlon JM, Kusano KD, Gabler HC. 2016a. Lane departure warning and prevention systems in the US vehicle fleet: Influence of roadway characteristics on potential safety benefits. Transportation Research Record 2559:17–23. Scanlon JM, Page K, Sherony R, Gabler HC. 2016b. Using Event Data Recorders from Real-World Crashes to Investigate the Earliest Detection Opportunity for an Intersection Advanced Driver Assistance System. SAE Technical Paper 2016-01-1457. Warrendale PA: SAE International. Scanlon JM, Sherony R, Gabler HC. 2017. Earliest sensor detection opportunity for left turn across path opposite direction crashes. IEEE Transactions on Intelligent Vehicles 2(1):62–70. Schubert R, Schulze K, Wanielik G. 2010. Situation assessment for automatic lane-change maneuvers. IEEE Trans-actions on Intelligent Transportation Systems 11(3):607–616. Singh S. 2015. Critical reasons for crashes investigated in the national motor vehicle crash causation survey. -Traffic -Safety Facts Crash Stats Report No. DOT HS 812 115. -Washington: National Highway Traffic Safety -Administration. Swedish Ministry of Transport and Communications. 1997. En route to a society with safe road traffic. Memorandum DS 1997:13. Norrköping. Tesla. 2018. Model S Owner’s Manual. Palo Alto. Online at https://www.tesla.com/sites/default/files/model_s_-owners_ manual_north_america_en_us.pdf. Volvo Car Corporation. 2015. 2016 Volvo XC90 Owners -Manual. Gothenburg, Sweden. Online at http://esd.-volvocars.com/local/us/volvo-2016-xc90-owners- -manual-v3.pdf.  The 2019 Audi A8 “Traffic Jam Pilot,” for instance, allows -drivers to travel hands-free up to 35 mph on a limited-access divided highway.  The Manual of Uniform Traffic Control Devices (FHWA 2012) defines standards and recommendations for state and local authorities. About the Author:Ryan Harrington is a principal and Carmine Senatore a senior associate in Exponent’s Vehicle Engineering practice in Natick, MA. John Scanlon is an associate in Exponent’s Vehicle Engineering practice in Philadelphia. Ryan Yee is a senior associate in Exponent’s Vehicle Engineering practice in Menlo Park, CA.