Advanced Ground Vehicle Technologies for Airside Operations (2020)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

2 AGVT include automated vehicle (AV) and connected vehicle (CV) technologies, as well as other technologies, software, and systems that increase safety and efficiency. Interest in AGVT in the roadway sector is significant for a number of reasons. These technologies may free drivers from the driving task, and as a result, increase efficiency since drivers can use the time they previously spent driving for other tasks. AGVT may increase safety, reduce injuries, save lives, and reduce the cost of crashes since human error is a contributing cause in many crashes. AGVT may lower costs by using infrastructure more efficiently: allowing closer following distances, ensuring optimal routes to reduce fuel consumption, and facilitating shared vehicles to reduce the cost associated with private vehicle ownership. AGVT may also provide benefits for sectors that have an inadequate labor supply. Although less often acknowledged in the popular press, scholarly journals, or aviation trade magazines, the airside implementation of AGVT may introduce similar benefits: increase safety, increase efficiency, and reduce costs, injuries and ground damage. AGVT for airside operations may also present potential challenges, including the need to work within a complex environ- ment, not only with respect to operations but also with respect to regulations and security. Each airport may present unique operating challenges due to the different airfield layouts and the variety of aircraft and airside activities. FAA, the International Civil Aviation Organization (ICAO), and Transportation Security Administration (TSA) represent regulatory stakeholders that must be considered, and AGVT must be integrated into the existing framework for air traffic control (ATC) and ramp control. These considerations highlight that AGVT will be incor- porated into a different framework at each airport since each airport is unique in terms of its activities, stakeholders, geography, and facilities. This background chapter includes information on enabling technologies, levels of auto- mation as defined by the Society of Automotive Engineers (SAE), and examples of vehicle technology applications for different levels of automation. Enabling Technologies A wide range of enabling technologies, technology components, and systems have been developed for AGVT by long time participants in the auto industry, as well as established software and technology companies, and numerous start-up companies. A detailed description of enabling technologies is provided in Appendix A, and a more general overview is provided in this section. Technology companies and applications may be categorized into the following sectors, which illustrate the breadth of potential technologies used, and the wide variety of participants involved (Stewart, 2017): • Onboard sensors—cameras, global position system (GPS), LiDAR, and radar; • Intelligent manufacturing—advanced materials, 3D printing, and automated assembly lines; C H A P T E R 2 Background

Background 3 • Infrastructure and connected cars—fleet and traffic management, data platform, and software development for connected cars; • Mapping, simulation, and image recognition; • In-car intelligence and assistance—vehicle diagnostics, passenger focused sensors, personal assistance, infotainment, and displays; and • Safety and security and services—parking, route planning, and carpooling. Widely used technologies include sensors, machine learning, and maps. Sensors allow vehicles to assess the environment and often include radar, cameras, and LiDAR (less commonly, ultra- sound or infrared sensors may also be used for specialized applications). Machine learning allows vehicles to recognize objects in their environment and respond appropriately. Automated systems typically rely on a baseline map, which provides a frame of reference for operation. Additional details are provided below (Davies, 2018a). • Radar (radio detection and ranging) sends out radio waves and has been used for decades for aviation (e.g., ATC), defense, and weather applications. Radar may be limited by line-of- sight, noise, interference, and may not detect smaller objects (e.g., a pedestrian’s arm making a turn signal) but is reliable, has good range, and works well in fog, snow, and rain. Radar is also less expensive than LiDAR (Brandt, 2017). • Cameras, combined with machine vision (image-based inspection), allow lane lines, road signs, and traffic lights to be recognized and interpreted. • LiDAR (light detection and ranging) sends out infrared pulses and measures how fast the pulses return. The data is used to create a 3D image or map of the surroundings. Although LiDAR is a less mature technology (and more expensive), it is easier for computers to inter- pret and convert to a 3D rendering than images from cameras. Lidar can build a 360-degree 3D image by obtaining data from a device that rotates on the roof of a vehicle, and can be used for positioning within a lane since it can detect the intensity and brightness of the lane marking. Newer systems are experimenting with different systems in an effort to increase capabilities and bring down costs. Alternatives include tiny mirrors to direct the laser rather than spinning (Davies, 2018b), 3D flash, optical phase array, and solid state (Gain, 2017). • Machine learning, a subset of artificial intelligence (AI), allows computers to discern vehicles, signs, pedestrians, and cyclists in the roadway environment. Typically, machine learning is accomplished by training the computer via millions of examples. The challenge is that the roadway environment can be complex and in some cases, hard for a machine to interpret, since there are such a wide variety of possible scenarios. Machine learning allows cars to learn from experience to build their own database from which to draw. • Maps provide a reference for AGVT and are typically developed based on LiDAR and radar. The reference map provides a baseline for sensor data. Some cities and states have mapped corridors and networks. • Communications are important and necessary for CV applications, including vehicle-to- vehicle (V2V), vehicle-to-infrastructure (V2I), and vehicle-to-everything (V2X). Communica- tion options include dedicated short-range communication (DSRC), which provides one- or two-way radio communication, and 5G, which provides cellular-based commu nication. 5G may have better range, performance, and reliability, although perspectives on this topic differ. Some experts think the potential impact of V2X is much greater than the impact of advanced vehicle automation technologies, which are more vulnerable to sensor and software problems, poor weather, and unexpected conditions (Knight, 2018). Although an initial Notice of Proposed Rulemaking (NPRM) published in January 2017 required that all light vehicles use a DSRC device to broadcast vehicle characteristics (e.g., speed, heading, and brake status), as of November 27, 2018, the U.S. DOT had not made a final decision regarding a mandate for V2V communication. DSRC provides 360-degree awareness up to 1,000 feet, and 75 MHz of spectrum in the 5.9 GHz band has been allocated for advanced vehicle technology applications (U.S. DOT, n.d.).

4 Advanced Ground Vehicle Technologies for Airside Operations Terminology and Levels of Automation In 2016, the National Highway Traffic Safety Administration (NHTSA) adopted the conven- tion for levels of automation as defined by SAE (shown in Table 1) and defined levels of auto- nomy from Level 0 (L0), which reflects no automation, to level 5 (L5), which reflects full automation. This framework provides a common vocabulary for government agencies and industry. As automation increases, the responsibility for vehicle operation transitions from the driver to the advanced vehicle technologies. The driver has the responsibility for environ- mental monitoring during conventional operation (L0, L1, and L2) with advanced driver assis- tance systems (ADAS) providing support in L1 and L2. Automated driving systems (ADS) are responsible for environmental monitoring in L3, L4, and L5. Many companies may skip L3 due to challenges related to driver vigilance and the need for drivers to respond appropriately and quickly when needed (Naughton, 2017). It is expected that L4 will be realized before L5; an example of L4 is automated operation on an interstate during good weather. L5 includes fully automated operation on any public street in any weather condition, including complex geometries such as one-way streets, roundabouts, and roads with different rules for different times (e.g., school zones and peak-hour turn restrictions). L5 also includes fully automated operation in urban areas with a mix of passenger vehicles, motorcycles, and delivery trucks, as well as pedestrians and cyclists, and suburban streets where pedestrians and cyclists are less common but may be present in very low volumes. Airside applications of AGVT typically represent L4, due to the limited geographical area and well-defined operational domain; however, in some cases (e.g., on the ramp near the terminal) the operational domain in the airside environment is very complex, much more than opera- tion on a limited access freeway. In some cases, applications may not fit clearly in this framework. For example, remote operation, platoon operation, and/or conditional automation (L3) with oversight from a command center are not explicitly addressed by this framework. When discussing terminology, it is important to acknowledge that there is a variety of termi- nology and different terminology is used by different organizations. Moreover, there is ongoing critique regarding the definitions put forth (for example, A Critique of the SAE Conditional Driving Definition, and Analysis of Options for Improvement, published in 2017). SAE suggests that terms such as automated, autonomous, and robotic should not be used in Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles (originally issued in 2014 and revised in 2016 and 2018). However, the U.S. DOT uses the term auto- mated vehicle in Preparing for the Future of Transportation: Automated Vehicle 3.0 (2018) and continues to use the term automated vehicle in publications in 2019. The International Air Responsible for monitoring environment Level Description Human operator 0 No automation: human driver does everything 1 Driver assistance: ADAS provides limited assistance with some functions (e.g., backup warning beep, automatic headlights) 2 Partial automation: ADAS can conduct some parts of driving task, human monitors and controls other tasks (e.g., cruise control) Automation 3 Conditional automation: vehicle ADS can conduct some parts of task and monitor driving environment but human ready to take back control 4 High automation: vehicle ADS can conduct driving task and monitor driving environment but only in certain environments and under certain conditions 5 Full automation: vehicle ADS performs all tasks under all conditions Source: NHTSA, 2016. Table 1. SAE International definitions for levels of autonomy.

Background 5 Transport Association (IATA) also uses the terms automated and autonomous in their white paper Simplifying the Business (StB) (2017), as do numerous states and NHTSA. IATA uses the terms autonomous vehicle and automated vehicle to refer to vehicles that achieve at least L3 and, under most circumstances, would operate under the control of the ADS. This is fairly consistent with terminology used in the vernacular of the popular press. A final example of the differences between SAE and other entities is that although SAE considers the term robotic a deprecated term (one that should not be used), the mowing industry uses the term robotic (for example, Particular Requirements for Robotic Battery Powered Electrical Lawnmower, the draft ANSI/OPEI safety standard published in 2019). SAE acknowledges the use of the terms automated and autonomous in “colloquial discourse” (SAE, 2018) and the concern that these terms have technical specificity. In this document, the terms automated and autonomous are both used, which is consistent with the terminology used by IATA and the U.S. DOT. Any associated loss of technical specificity is not considered a problem given the conceptual level of evaluation provided in this report and the fact that the audience for this publication is airport professionals and not automotive engineers or technical professionals specializing in automated driving technology. Example of Advanced Driver Assistance Systems There are a variety of ADAS that provide assistance with some functions during L1 or L2 operation. Many of these systems rely on information from vehicle cameras, radar, and/or LiDAR systems. Examples of common ADAS include the following (Consumer Reports, 2016): • Brake assist applies maximum break force for the shortest possible stopping distance and is activated when the driver initiates a panic stop rather than a typical gradual stop used in normal driving conditions. • Forward collision warning warns the driver with audible or visual signals if they are approaching a vehicle and a crash is imminent. • Automatic emergency braking (AEB) initiates braking to avoid a potential collision. • Pedestrian detection provides a warning or AEB if a pedestrian is detected in the vehicle path and a collision is likely. • Adaptive cruise control automatically maintains a constant and safe following distance behind the forward vehicle. Some systems automatically increase speed when traffic (and the forward vehicle) speeds up again. • Blind-spot warning warns the driver of nearby vehicles by illuminating a light or icon on the outside mirror and/or with an audible warning during lane changes. Some systems may break or steer back to the center of the lane to avoid a potential collision. Activated by infor- mation from cameras or radar when the turn signal is turned on or if the vehicle initiates a lane change. • Rear cross-traffic alert warns the driver of traffic behind the vehicle when the vehicle is in reverse and some systems automatically brake to avoid a collision. • Lane-departure warning warns the driver with audible, visual, or haptic signals (vibration of the steering wheel or seat) if the vehicle leaves the lane without the activation of turn signals. • Lane-keeping assist senses when the vehicle is leaving the lane and provides correction with a minimum steering input. • Parking assist systems have sensors in the front and/or rear bumper, which alert the driver at parking speeds when obstacles are getting close. Many of these systems are maturing, and some of these systems have been available for more than a decade. For example, Toyota’s Japanese Prius hybrid vehicle was one of the first vehicles to offer parallel parking assistance in 2003, followed by Lexus (LS sedan), Ford (Active

6 Advanced Ground Vehicle Technologies for Airside Operations Park Assist in 2009), and BMW (2010) (Dormehl and Edelstein, 2018). Although technological hurdles may be surmounted, the regulatory framework for self-driving cars is still evolving, as discussed below. Federal, State, and Local Legislative Framework Federal, state, and local legislation for autonomous vehicle systems may serve as a frame- work or guidance for rules associated with automated systems in airside operations. There has been significant legislation at many levels to provide a safe process for autonomous vehicles to be tested on public roadways. In 2016, NHTSA published the Federal Automated Vehicles Policy, which provided guidance rather than regulation as well as a standard framework for levels of autonomy (shown in Table 1) and vehicle performance. This was replaced by NHTSA’s Automated Driving Systems 2.0: A Vision for Safety in September 2017. Automated Driving Systems 2.0 provided a streamlined approach intended to facilitate and support the safe deploy- ment of ADS, clarified the federal role and best practices for states, and laid the framework for forthcoming legislation. In October 2018, the U.S. DOT released updated guidance and invited comment on Preparing for the Future of Transportation: Automated Vehicles 3.0 (referred to as AV 3.0). This builds on the previous framework, and broadens the scope with a multimodal framework for AV deployment. The document lays out plans for the U.S. DOT to support pilot programs with the intent of supporting the United States as a leader in AV technology in the future. AV 3.0 acknowledges that AVs should learn from automation in aviation, specifically the use of automation in the cockpit, and the need to manage unusual situations, as well as issues such as workload and distractions. As a result of automation, aviation safety has increased. Aviation automation has successfully addressed human-machine interface (HMI) or pilot interface and the need for training to understand new systems and avoid skill degradation; moreover, automation in the cockpit has not replaced pilots nor reduced their compensation (U.S. DOT, 2018). All vehicles sold in the United States must meet Federal Motor Vehicle Safety Standards (FMVSS), which implicitly makes vehicle safety standards the responsibility of the federal government. Manufacturers must obtain an exemption to FMVSS for any vehicle that does not have a steering wheel, accelerator, or brake pedal, or if it does not meet the FMVSS for any other reason. Currently, NHTSA grants exemptions for up to 2,500 vehicles per year. The voluntary guidance and associated safety elements in Automated Driving Systems 3.0 focus on L3 to L5, when the system takes over full control of the vehicle including monitoring of the environment (NHTSA, n.d.). Federal legislation proposed in 2017 included the Safely Ensuring Lives Future Deployment and Research in Vehicle Evolution Act (SELF DRIVE) passed by the U.S. House of Represen- tatives (United States HR 3388) and the American Vision for Safer Transportation through Advancement of Revolutionary Technologies Act (AV START, 2017) passed by the U.S. Senate Commerce, Science, and Transportation Committee (United States S 1885). Both SELF DRIVE and AV START proposed to maintain common law liability under state law, limit state regula- tions for ADS, exclude large commercial vehicles from the legislation, expand the FMVSS to include ADS operation, and increase the number of exemptions from FMVSS to allow more vehicle sales (e.g., from the current limit of 2,500 per year up to 100,000 per year per manu- facturer in SELF DRIVE). The expansion of the FMVSS to include ADS operation contradicts the traditional framework in which states license vehicle drivers. AV START also required the submission of a safety evaluation from manufacturers to the U.S. DOT, and focuses more attention on data use, data access, consumer privacy, consumer education, and cybersecurity.

Background 7 The proposed federal legislation would address ongoing concerns regarding the current framework in which there is different AV legislation in different states. Thirty states and the District of Columbia have enacted legislation and seven governors have issued executive orders, as of September 19, 2019, based on data from the National Conference of State Legislatures (shown in Figure 1). State legislation may reflect traditional state responsibilities such as driver and vehicle licen- sure and registration, operator requirements, insurance and liability, and vehicle inspection, as well as vehicle testing and operation, commercial vehicle operation (including vehicle platoons), and privacy (Hubbard, 2017). In addition to federal and state legislation, vehicles must also operate within constraints of local laws (aka ordinances), including speed limits and other motor vehicle laws. Some states allow local jurisdictions to promulgate AV laws, and other states, such as Texas, Colorado, and North Carolina, explicitly restrict local juris- dictions from regulating AV operation. Nevada recognizes the sovereignty of airports and allows them to charge a fee for a business license for AV if it is consistent with standard charges for permits. Just as Nevada recognizes the sovereignty of airports to create their own rules, airside operation of AGVT may be exempt from state and local laws in some cases. An examination of the laws related to AV provides Figure 1. States with autonomous vehicle legislation (as of September 19, 2019). Based on data from the National Conference of State Legislatures, 2019. Image created with MapChart.net.

8 Advanced Ground Vehicle Technologies for Airside Operations an opportunity to examine topics and evaluate how they should be addressed in the airside environment, including airport rules and regulations. Topics for consideration may include defining a safe framework for testing and operation, liability, and coordination with all stakeholders that may be affected. AV Testing Activities Roadway AV testing has occurred in many states, and the lack of state legislation does not necessarily preclude testing. Testing has been publicized in Arizona, California, Michigan, Ohio, Pennsylvania, Texas, and Virginia. The U.S. DOT identified 10 AV testing sites in Pittsburg, Pennsylvania; Arlington, Texas (Proving Grounds Partnership); Aberdeen, Maryland; Willow Run, Michigan; Contra Costa and San Diego, California; Iowa City, Iowa; Madison, Wisconsin; Central Florida (with a variety of partners, including Orlando, Central Florida Expressway, and Kennedy Space Center); and North Carolina (Turnpike) (U.S. DOT, 2017). Other states also have ongoing testing activities. For example, Ohio has a Smart Mobility Corridor, the Smart Mobility Advanced Research and Test (SMART) Center, and the state capital (Columbus) was designated a SMART City, leveraging $10M from industry and $40M from the U.S. DOT as a catalyst for $500M in investments (Maddox, 2017). The San Diego and Texas testing activities that are part of the U.S. DOT program are notable in that they both involve partnerships with airports, namely the San Diego Airport and the Austin Bergstrom International Airport. In Austin, the airport may serve as a controlled environ- ment for testing AVs; the simplified street network and low speeds at the airport provide benefits for testing (Texas AV Proving Ground Partnership, n.d.).