Challenges to CV and AV Applications in Truck Freight Operations (2017)

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.

CRP Project HR 20-102(03) 6 C H A P T E R 3 Research Findings Technology Overview A range of technologies comprise today’s CV and AV concepts and solutions, many dating back across decades of research and development by both industry and governments. This section introduces five key enabling technology-focused areas: (1) truck systems, specifically major components of truck systems needed to enable automated solutions; (2) sensor-based technologies that are added to trucks to enable many of the connected, autonomous, automated, and connected automated applications; (3) wireless communications technologies that provide connectivity between vehicles and the infrastructure; (4) cybersecurity needs including privacy topics; and (5) technical standards needs. Figure 1-1 shows a conceptual overview of the envisioned deployed environment. This is a general diagram that applies broadly to the family of CV and AV applications rather than to any one or more specific applications. It depicts a three-level hierarchy, beginning at the bottom with the truck. In addition to changes made to existing truck systems (not depicted), trucks may have a range of new sensor technologies added, quite similar to those being added to passenger vehicles that the public is becoming increasingly aware of. As the truck enters the road environment, as depicted in the middle level, sensor- based data inputs might be augmented by a range of wireless communications technologies permitting the truck to communicate with other, similarly-equipped vehicles (and not just limited to other trucks), and to parts of the fixed road infrastructure. Finally at the top level, CV and AV applications further extend trucks’ and trucking fleets’ integration into connected, real-time transportation systems. These are discussed in more detail later in this chapter, but considered as an example trucks being actively managed from a centralized control center to form and dissolve platoons with other trucks. Truck Technologies Most manufacturers have the automation components needed for full autonomy already in development or installed on the vehicle for advanced safety emergency interventions like vehicle stability control and advanced cruise control applications. The most common automation hardware is certainly the braking systems due to the prevalence of anti-lock braking systems. Also for longitudinal control many trucks have some form of cruise control. For full autonomy longitudinal velocity control requires some type of automatic transmission. Often a torque converter automatic transmission is not suitable for the desired fuel efficiencies of the trucking industry. The manufacturers have instead implemented the automatic manual transmission to achieve the advantages of the manual transmission but remove the need for a driver to actively shift the gears. On the lateral end of safety, the Anti-lock Braking System (ABS) hardware can provide basic traction control but can also be used to stabilize some forms of excessive yaw rate induced maneuvers. Additional advancements in the cruise control area have led to lane keeping capabilities wherein steer actuators are part of the cruise control packages from the manufacturer. The steer hardware however is not prevalent in the market just yet but has been heavily researched in the lane assist area. The move toward autonomy with respect to automation hardware is almost fully complete even though currently the autonomous hardware is used primarily for cruise control and safety. It’s likely that a few

CRP Project HR 20-102(03) 7 hardware modifications are needed to step to upper levels of autonomy. Certainly, higher level control software and system integration concerns will need to be taken into account, but the automation hardware is there currently or will be mostly standard in the next several years. However, there are sensor packages is there currently or will be mostly standard in the next several years. However, there are sensor packages needed for the fully autonomous truck to be a robust system. Figure 1-1. High-Level Overview of Components of CV and AV Trucking Applications.

CRP Project HR 20-102(03) 8 hardware modifications are needed to step up to upper levels of autonomy. Certainly, higher level control software and system integration concerns will need to be taken into account, but the automation hardware is currently specified on new model tractors or will be mostly standard in the next several years. However, there are sensor packages needed for the fully autonomous truck to be a robust system. Sensor Technologies Trucks with varying degrees of autonomy are deployed with sensors including position and velocity measurements from GPS, as well as range to adjacent vehicles via information from radar and dedicated short range communication (DSRC) units. Command coordinated vehicle accelerations must also be communicated between vehicles. The vehicle-to-vehicle (V2V) communication requirement is handled by a DSRC system. The sensors and system processing allow for significantly quicker intervention and corrective action in comparison with the response time of a human. Status information from the lead vehicle including but not limited to brake and throttle information aids the controller performance and adds to the safety of operation of the follow truck while in platoon. Thus, the much shorter following distances are safely facilitated by these systems compared to following distances safely achieved with manual driving. The various classes of sensors needed for autonomy and coordination of vehicles are described below. Communication systems are then discussed below in the Wireless Communications Technology Section. Charge-Coupled Device (CCD) Cameras Traditional visible-spectrum cameras are often used as a ranging and obstacle detection management system. In general, CCD cameras are fairly reliable in obstacle detection, however during suboptimal conditions, traditional visible light cameras suffer disproportionately to other sensors. In shadows, extremely bright light, rain and other inclement weather, or against low-reflectivity surfaces, a traditional CCD camera suffers. Due to this large restriction on environmental conditions in which traditional cameras operate successfully, they are typically used as a redundant measurement rather than the primary, although some groups have successfully deployed stereoscopic camera systems as an effective dual- purpose ranging and detection measurement device. Typically, CCD cameras are used in pre-existing lane detection systems already implemented by companies to bolster the sensors intended for platooning applications. For a stereoscopic system, two cameras are placed at a calibrated distance laterally away from each other. Knowing certain parameters about the cameras, such as their focal length and the distance between them, known as the baseline, an object’s spatial coordinates can be determined. Combining this with knowledge of the stereoscopic system to the vehicle, it can then evaluate the safety of the road and potential obstacles in its field of view. While fairly robust as a ranging system, stereoscopic CCD cameras adhere to the same principles as a single CCD camera when concerned with their limitations. Inclement weather, as well as poor lighting situations, create scenarios which require another sensor to accurately determine both obstacles and range to an object. Stereoscopic solutions are further limited by the focal length of the cameras involved, with high accuracy results extending sometimes as short as only 50-75 ft. The combination of cameras exacerbates the problem of focal length, since the camera’s shared field of view is even further limited. For specific applications, such as low-speed, highly controlled environments, CCD cameras are suitable for an extremely low cost ranging and obstacle detection solution. An example of such an environment is mining and construction sites with set paths and low speeds, such as the prototype being developed in KTH’s iQMatic project discussed later in this chapter.

CRP Project HR 20-102(03) 9 Differential Real-Time Kinematic (D-RTK) GPS New algorithms developed for GPS systems involve combining the messages from two GPS receivers in order to achieve extremely high precision position solutions. D-RTK GPS can be used to achieve centimeter-level precision in the relative position between vehicles. In terms of the overall cost of infrastructure required to implement the platooning system, D-RTK GPS proves much cheaper than standard RTK, and proves extremely beneficial when used as a filtering agent for a primary ranging mechanism. This can be seen in platooning systems such as Peloton’s prototype. The high resolution between vehicles is extremely useful in mitigating the largest problem in other sensors. The inherent issues in using GPS systems, long time-to-solution, urban canyons and poor reception are frequently not commonalities with other ranging solutions, making GPS a strong complement to other ranging techniques, particularly when filtering the relatively high noise radar and lidar measurements. Previously, the poor precision with single systems, along with transmission issues in transporting the large data required to calculate differenced positions proved prohibitive. Technological advancements in transmission as well as the new D-RTK algorithms have pushed GPS into viability for platooning systems. In general, D-RTK proves more useful than RTK in the near-term, due to the high initial cost of the required infrastructure to support RTK’s extremely high global location precision. Radar Reliable in a large variety of situations, radar proves to be a highly robust primary ranging mechanism. Its largest downside lies in its relatively noisy signal, which thus requires some type of filtering to achieve meaningful results for the platooning solution. Despite this drawback, radar’s low cost and high reliability makes it the current ideal solution for primary ranging, particularly when coupled with GPS. Radar suffers from some of the same problems as CCD cameras, where poor visibility and inclement weather cause the system to register more false readings, increasing its already noisy signal. Fortunately, GPS solutions are typically degraded much less than other sensors in these conditions, which proves a valuable asset in constructing the final range determination, for example in platooning systems. Lidar The current gold standard in terms of reliability and robustness, lidar is the newest primary ranging instrument. Lidar’s massive field of view, coupled with its high reliability even in inclement weather, poor visibility, and other hard-to-handle situations makes it extremely valuable. Recently, several companies have begun marketing lower cost lidars, which mitigates the one of the disadvantages of using lidar as the primary ranging instrument. Additionally, while lidars are extremely valuable in their large field of view, and high accuracy, this comes at a great cost to the computational load demanded by the sensor. High resolution, plus large field of view creates massive datasets that must be manipulated to yield accurate obstacle detection. Advancements in machine learning and computational speed have enabled companies such as Google to utilize lidar as their primary ranging and obstacle detection device. Companies have begun to utilize lower cost single-pass lidar units in Adaptive Cruise Control (ACC) systems. By using a single-pass lidar, the data management issue can be skirted, by vastly reducing the field of view of the sensor. This trades off some of the robustness of the sensor, but gains large decreases in both the cost and computational load. Wireless Communications Technologies Communication technologies offer a great potential for improving the safety and reliability of automated vehicles by connecting the vehicles to one another and to a host of off-board resources. The continuous progress of these technologies has been essential in CV/AV growth and has moved connected

CRP Project HR 20-102(03) 10 fleet from being a dot-on-a-map to a dynamic fleet network. These technologies include dedicated short range communications (DSRC), cellular Long-Term Evolution (LTE), Wi-Fi, satellite, peer-to-peer technologies, visible light communication, device-to-device (D2D) communication, and machine-to- machine (M2M) communication, among others. Moreover, with more interaction and integration of vehicles, smartphones, and cloud connectivity, flow of information among drivers and the infrastructure and, therefore, improved mobility have become more commonplace. More specifically for the connected/automated fleet, there has been a shift from satellite GPS to cellular and cloud connectivity. Below is a brief discussion of the more promising technologies that are being used or tested for use in the near future. DSRC DSRC, which operates in 5.9 GHz frequency band is currently the leading wireless medium tested for V2V communication. DSRC is supported by standards such as SAE J2735 and the IEEE 1609 suite that establish message types, message applications, and their structure and has been tested rigorously for V2V cooperative safety applications. DSRC remains the most reliable, safe, low latency technology to provide non-line-of-sight short range communication. Although DSRC has been mostly tested for light vehicles, the same concepts are applicable for connectivity among trucks as well. For example, in an attempt to improve their work zone safety, Texas DOT has been using their attenuator trucks and lead vehicles equipped with DSRC technology for better communication, especially in the areas with no cellular coverage (Dellenback 2015). RFID Radio-frequency identification (RFID) is the wireless use of electromagnetic fields to transfer data, for the purposes of automatically identifying and tracking tags attached to objects. The feasibility of using the RFID for building intra-vehicle sensor networks has been studied and researched for a while. RFID could be used for connecting sensors to the central unit (instead of using wires); this will be both cost effective and fuel efficient. For trucks, RFID tags can be very efficient for toll/weight stations as well. LTE 4G/5G There is currently active research and developments in using 4G/5G LTE for vehicle connectivity and automation. Of note are Qualcomm’s LTE modems that work within major RF bands and can provide up to 450 Mbps coverage in support of vehicle automation and advanced telematics. The fifth generation (5G) mobile network technology represents the next major phase of mobile telecommunications standards. Machine-to-machine (M2M), and device-to-device (D2D) communications are becoming more possible as 5G technology matures. 5G promises to have speeds of up to 10 Gbps, five times lower latencies than 4G, and vastly improved network coverage. 5G will also offer a more intelligent use of spectrum and greater reliability. Ericsson recently reported achieving 5 Gbps speed in a live test of pre-standardized 5G, using a new radio interface concept combined with advanced multiple-input multiple-output (MIMO) technology with wider bandwidths, higher frequencies and shorter transmission time intervals. Relevant applications such as automated trucks continuously updating a map for autonomous driving should be enabled through this technology. Although 5G has significant potential, issues (such as standardization) remain before it can move into widespread deployment. 5G is expected to be a most promising wireless technology in the next five to ten years and will most likely play a role in vehicle connectivity and automation. However, the future of CV and AV fleets will ultimately depend on a host of wireless technologies. For instance, the safety critical communications will

CRP Project HR 20-102(03) 11 likely utilize the DSRC network while communications for other applications could use LTE Direct or 4G/5G, and the internal vehicle connection can be over peer-to-peer networks. M2M and Internet of Things (IoT) M2M refers to technologies that enable communication with other similar devices. M2M does not refer to specific wireless or wired technology, and is an integral part of the IoT. The key components of an M2M system are: wireless devices with embedded sensors or RFIDs with complementary wired or wireless access such as cellular communication, Wi-Fi, ZigBee, WiMAX. M2M examples include a device such as a sensor or meter to capture an event such as temperature that is communicated through a wired/wireless network to a software application. M2M communication has expanded beyond a one-to- one connection and has changed into a system of networks that transmits data to personal appliances. M2M has already been utilized for fleet management collecting data from fleet systems and sensors, engine data, cargo-related data, driver-centric data, and on-board safety systems data. M2M and IoT are currently being used to share real-time traffic and road condition data to increase road safety and improve traffic flow by enabling better communications. IoT and connectivity to the cloud will play a significant role in integrating and analyzing the information collected from the vehicles and disseminating the necessary information back to the vehicles and infrastructure for better traffic flow and mobility. IoT/M2M is especially useful for connected and automated trucks for dynamic routing and scheduling, fleet monitoring on the map, collecting data from rolling stock and remote equipment, in-cab navigation, and extracting information from a component of a truck (e.g. load, truck speed, engine status, etc.). Some notable, current industry examples include:  Drivewyze offers a cloud-based weigh station bypass system.  Daimler Trucks North America’s Detroit subsidiary offers telematics and remote diagnostic services on Freightliner and Western Star trucks.  Navistar’s OnCommand Connection system uses the connection from a fleet’s third party telematics provider to generate engine performance and diagnostic information in near real-time.  Volvo’s remote diagnostic platform monitors vehicle performance and sends alerts to its 24-hour call center. Volvo personnel contact the appropriate personnel at the carrier and advise them on whether or not the detected problem is serious enough to warrant shutting down the truck immediately or whether the truck can continue driving. D2D D2D is considered one of the key technology components of the evolving 5G architecture, enabling devices in the network to function as re-transmission nodes to enable a large and ad hoc web of communication. D2D is also referred to as cooperative communications that represent a new class of wireless communication techniques in which network nodes help each other in relaying messages. Researchers who studied over 100 D2D communications in cellular networks, determined that using D2D and cellular together, applying special techniques, could lead to a more efficient use of spectrum and energy (Asadi et al. 2014). However, interference between D2D and cellular networks could be a key challenge. Visible Light Communication In addition to RF communication technologies, VLC is a promising non-RF technology currently under research and development. VLC is a scalable solution in high vehicle density and fast topology changing scenarios in vehicular communication. The design includes VLC transmitters connected to the headlamps and taillights, and photodetectors to enable short range optical communications. An example of VLC

CRP Project HR 20-102(03) 12 research has been in platooning applications, using LED lighting infrastructures as communications equipment to exchange information about the relative directional position of each member of the platoon. VLC holds the potential to be a solution for truck platooning, especially when accessing the RF spectrum is difficult. Cybersecurity Cybersecurity has become a rapidly increasing concern within the automotive world in the past few years. As vehicles become more connected through V2V communications, infotainment systems, on- board diagnostics systems, and other technology, vulnerabilities to cybersecurity attacks naturally increase. Most people within the automotive field are familiar with recent demonstrations of what an attacker with the necessary resources can accomplish with select vehicles, from turning on windshield wipers to taking control of the breaking and steering systems. At the same time, the government and industry are taking steps to mitigate these vulnerabilities and threats by building cybersecurity into technology development programs and vehicle design. Examples include the Security Credentials Management System (SCMS) to facilitate secure, authenticatable communications among connected vehicles, infrastructure, and wearable devices, as well as the development of new standards such as SAE J3061 Cybersecurity Guidebook for Cyber-Physical Vehicle Systems, SAE, January 14, 2016 (an update to this standard is already in progress); SAE J3101 (work in progress) Requirements for Hardware- Protected Security for Ground Vehicle Applications, Society of Automotive Engineers; and the draft Framework for Cyber-Physical Systems, Release 1.0, Cyber-Physical Systems Public Working Group, National Institute of Standards and Technology, May 2016. With all of the vehicle cybersecurity concerns, the question of whether government should regulate vehicle cybersecurity has been asked many times by the public and members of Congress. Given the number of current efforts led by government research programs, standards bodies, and industry organizations, along with the constantly evolving nature of cybersecurity, the government may be hesitant to take a traditional regulatory or rulemaking approach. Rather, we are seeing an interest on the part of government agencies at the federal and state levels, as well as among inter and intra-industry trade groups and organizations to develop guidelines, practices, and policies in coordination with multiple stakeholder groups that are invested and affected by the cybersecurity issues at hand. Trucking-Specific Issues and Concerns In most cases, trucks and heavy vehicles are much larger, more expensive, more complex in terms of their internal systems, and pose a greater physical risk than do light or passenger vehicles. Heavy vehicles may also carry high value and/or hazardous cargo that create a higher impact in the event of physical accidents or cybersecurity incidents. These factors converge to present targets that are attractive in terms of impact and potential return on investment for attackers. The motivation for attackers will likely be financial (e.g., bricking the vehicle until a ransom is paid) and companies would likely pay a higher price to regain control of vehicles hauling high value and/or hazardous cargo. Heavy vehicles have a number of unique characteristics that may require different and/or more stringent security requirements than light vehicles. Heavy vehicles of all classes are being equipped with aftermarket features and functions (e.g., wireless interfaces and automation) that result in new attack surfaces. These surfaces have already been demonstrated to serve as entry points and launching pads for attacks on passenger cars. Many of these newer features facilitate information exchange and are connected to electronic and control systems, including interfaces to the V2V communications systems and processors. In addition, fleets that outfit a large number of heavy vehicles with vulnerable aftermarket devices create inviting targets for attackers who can develop and use a single attack across a large number of targets. A number of heavy vehicles operating on the highway in cooperative platoons might also represent an inviting target for an attacker who simply wants to disrupt the platoon's efficiency, change

CRP Project HR 20-102(03) 13 the platoon’s course, apply unintended braking, or use the platoon as a weapon. Heavy vehicle transport that can be slowed or stalled could have a significant impact on the economy and the delivery of goods and raw materials. Heavy commercial vehicles may also have a longer life than passenger cars; thus, the risks over the extended life of a heavy vehicle could be more significant. Older trucks often have legacy systems and components, such as the engine, cab, and chassis, from various companies, and these may be technologically easier to attack. Technical Standards Standards of various sorts are required to fully and successfully deploy these application areas. Safety- oriented vehicular standards, such as the Federal Motor Vehicle Safety Standards (FMVSS) for the U.S. market, and other ‘standards’ or guidelines that are regulatory in nature, are discussed below in the Legal, Regulatory, and Policy Topics Section. A variety of other technical standards are also needed. These are generally not regulated, though there are exceptions such as the standards that have been developed to support NHTSA’s V2V rulemaking. Broadly, the need for standards is multi-faceted, including to ensure safety, promote interoperability, and enable a competitive marketplace where vendors can create equipment that meets requirements. Standards are important because it signals to stakeholders (technology providers, local implementers, etc.) that they can move forward with increased assuredness of the vitality of a particular solution. Standards may be local, national, regional, or international in scope depending on factors such as the extent to which the equipment market is local versus national or international; cost, and willingness of stakeholders to harmonize standards inter-regionally; and the criticality against target deployment timelines. Except in instances where a vendor’s solution becomes a de facto standard, true standards are developed deliberately by consensus processes that tend to be long in duration compared with the more rapid pace of technology developments. Some of the technical standards needed for truck environments are covered in the passenger vehicle space and so not discussed in this report. But there are unique trucking applications, where technical standards are needed and these have data element and messaging needs. The technical standards needed to deliver the applications discussed in this report encompass communications protocols, data elements, data directories, applications and application messages, test devices and procedures, and management and governance protocols. Leading standards development organizations for development of the technical standards that will be used in various systems includes the International Organization for Standardization (ISO) and SAE International. Technical committees within ISO focused in particular on the vehicle (Technical Committee 22) and on intelligent transport systems (Technical Committee 204) have as part of their charter to develop standards needed for truck-based applications (many of which are developed equally for light and heavy-duty vehicles). Many legacy standards are mature and are or will be used in development of final systems and solutions. Examples of work items that are within these standards development organizations’ work programs, that apply in part or in whole to trucking include:  Expansion of the Basic Safety Message from one focused on rigid-bodied vehicles to one that robustly accounts for articulated vehicles such as a tractor-trailer.  Messaging and performance requirements around Cooperative Adaptive Cruise Control (CACC) and platooning (this is an example where there is shared leadership within the standards organizations, as the European Telecommunications Standards Institute has a significant role in this – as well as many other – relevant standards).  Other messaging and performance requirements are under development within SAE J2735 and J2945 and include consideration for trucking applications. For example, an initiative planned to begin soon on-road weather information includes a use case around freight carriers.

CRP Project HR 20-102(03) 14  A suite of technical standards providing the basis for regulated applications (referred to as Telematics Applications for Regulated Commercial Freight Vehicles) such as weigh-in-motion and electronic toll collection.  Other vehicle/roadway warning and control systems applications such as Cooperative Forward Vehicle Emergency Brake Warning, Road Boundary Departure Prevention System, and Partially Automated Lane Change Systems.  Test device requirements for vehicles, vulnerable road users, and other objects.  ISO 26262 is a voluntary industry standard that defines comprehensive automotive safety requirements around electronic and software intensive features in road vehicles. While a mature standard, its current efforts include focus on trucks (its original focus was on light-duty vehicles), and extensions into areas pertinent to AV operations such as parameters around fail-operational scenarios. Other standards are needed related to infrastructure, and while technical in nature, they may also have a public policy component to them. An example of an infrastructure standard with increased importance in the CV and AV environment is pavement markings. Delphi reports that during one of its tests of a highly automated vehicle, “…despite nominally uniform standards across states, the pavement markings were actually all different. Even within one state, the standards can vary” (Vock 2016). The needs that CV and AV have in particular for such infrastructure-based standards have a public policy component to them, as governments decide how to prioritize standards’ needs.

CRP Project HR 20-102(03) 15 Application Scenarios Trucking CV and AV applications span a range of application areas and uses. In this section, we group the applications in four major categories and outline their development progress, deployment prospects, challenges, and opportunities: Over-the-Road CV applications are SAE level 0 applications, providing a range of applications that are conjoined with or in addition to similar applications under development for the passenger vehicle market. Many applications here are stepping stones to higher levels of automation. The second category discussed looks specifically at truck platooning, which includes options for higher levels of automation and has an emerging distinct market potentially forming. The third category extends the automation applications further into trucks that have SAE levels 3-5 capability. Finally, applications that are specific to ports, including freight terminals and border crossings, are discussed. Over-the-Road Connected Vehicle Applications Research and development in CV applications is highly centered on the passenger vehicle market, many of which have extensions into the trucking sector. This section discusses the main CV applications that are applicable to trucking. We first define and identify the concepts within the context of passenger vehicles when this is applicable, then provide specific discussion of trucking applications including current status, relevant technology issues, and potential benefits. In general, a significant benefit of CV applied to heavy trucks is for safety improvements, which has both public sector interest as well as helping make the investment case for these applications. Most deaths in large truck crashes are passenger vehicle occupants. Large trucks often weigh 20-30 times as much as passenger vehicles. They are taller and have greater ground clearance than cars, which means that lower- riding vehicles can slide beneath truck trailers, with deadly consequences. Furthermore, truck braking capability is a factor in many crashes. Compared with passenger vehicles, stopping distances for trucks are much longer, made worse on wet and slippery roads or if the brake systems are poorly maintained. Large trucks are also prone to rolling over. A requirement for electronic stability control takes effect for most new truck tractors in 2017 and is expected to reduce crashes. CV applications for trucks also provide mobility, efficiency, and environmental benefits. Inspections and weighing processes often consume large amounts of time resulting in lower efficiencies. It is possible to improve the times spent on these necessary procedures by conducting electronic screening or weigh-in-motion. Furthermore, CV applications have the potential to reduce the number of stops, the number of unproductive moves, and the length of segments traveled by trucks by providing real-time data that optimize freight routes. This mobility and efficiency improvement also translates into environmental benefits reflected by a reduction in emissions and fuel consumption. Promise of Connected Vehicles is Large "Our trucks fully connect with their environment, becoming part of the internet and continuously sending and receiving information. All those involved in the logistical process can use these real-time data for their needs. In the future it will be possible to reduce waiting times while loading and unloading, reduce paperwork and avoid traffic jams. An enormous opportunity to intelligently cope with the growing volume of goods traffic. We intend to use it." Dr. Wolfgang Bernhard, Member of the Board of Management of Daimler AG. Daimler Trucks & Buses (Daimler 2016)

CRP Project HR 20-102(03) 16 Lane Departure Warning (LDW) or Mitigation Systems LDW systems alert drivers when they unintentionally drift too close to the edges of the lane. The warning is provided using audible warnings, haptic warnings like making the driver’s steering wheel or seat to vibrate, or creating a feeling like driving over a rumble strip. LDW systems are different from Lane Departure Mitigation systems (sometimes referred to as lane keeping assistance). Warning systems provide a warning, but leave any corrective actions up to the driver, while prevention systems maneuver the truck (or car) by steering it to re-center it in the lane. Relevant Technology Issues. A simulation study based on Japanese crash data found potential benefits to be dependent upon the timing of warnings (Tanaka et al. 2012). In contrast with estimates from Field Operation Test (FOT) and other simulation modeling studies, an analysis of insurance claim data suggests that LDW alone (without active lane keeping assistance) may actually increase crashes (Lund 2013). Only Volvo LDW systems were associated with decreases in claim frequency; this is likely due to the pairing of LDW with forward collision warning and forward collision mitigation technologies (autonomous braking systems). Many systems rely upon having good, visible road markings and may not be able to detect an unmarked road edge. Lane departure warning systems rely on the ability of the sensors to register lane markings, which may be problematic on roads that are not well marked or are covered with snow. Many current LDW systems have reduced lane detection rates in low light or inclement weather compared to normal light and weather conditions (Sayer et al. 2010). Many crash avoidance technologies rely on the driver to take action (Insurance Institute for Highway Safety (IIHS) 2012). The effectiveness of these systems depends on whether drivers accept the technologies, understand the information from the system, and respond appropriately. Costs and Benefits. Mobileye offers a safety solution for collision prevention and mitigation that includes LDW. The aftermarket cost for the Mobileye 560 is $850 with $150 in installation labor costs. The LDW application is primarily designed to reduce high speed accidents on highways and freeways; it has the potential to reduce single vehicle crashes, head-on crashes, and sideswipes. The IIHS analysis reports that a reduction of 179,000 crashes annually, 38,000 non-fatal crashes and 7,529 fatal crashes may be achieved with LDW. This amounts to 3%, 5.4% and 23% reduction of crashes in each of those categories. Based on the published 25–30% effectiveness estimates for rumble strips, a more realistic estimate of crashes that may be prevented by LDW systems would be 45,000–54,000 per year (Jermakian 2011). An FOT suggests a possible reduction of crashes between 1% and 8%. It was conducted in 2004 and 2005, and it estimated that LDW systems could reduce road departure crashes by between 9,400 and 74,800 annually if all passenger vehicles were equipped with these systems and they worked as intended (Wilson et al. 2007). Furthermore, a statistical estimation study using crash data from the United Kingdom reported a benefit potential in the range of 7% to 29% for fatality reduction and 13% to 34% for serious injuries (Robinson et al. 2011). State of the Practice. LDW is an SAE level 0 vehicle automation system (no-automation). An area which may be worthy of more research is the issue of driver annoyance with false alarms, which may be particularly an issue with LDW systems. A study presents consideration of an adaptive LDW design which was intended to reduce frustration by reducing false alarms. However, a trade-off is involved between the number of times the system should warn the driver versus the increase in chances of failures to alarm (Tijerina et al. 2010).

CRP Project HR 20-102(03) 17 Applicability to Freight. A project using statistical modeling and simulation of a Volvo pre-production LDW system produced an estimated benefit of a 47% reduction in lane departure crashes, assuming ideal conditions (Gordon et al. 2010). A likely more realistic evaluation considering variable lane markings, non-ideal weather conditions, etc., produced an estimated target crash reduction of 33%. Taking further factors into consideration, a final estimated crash rate reduction was placed in the range of 13% to 31%. In a survey of Volvo owners, 77% reported that the LDW system had never failed to warn them when they believed they were at risk of drifting out of their lane. However, 17% reported that it had. That is, the system failed to warn them when it should have (a “false negative”). The most frequently reported situations in which this happened included missing or unclear lane markings (60%), inclement weather (1%), driving at slow speeds (7%), and driving in the dark (7%) (Eichelberger and McCartt 2012). Forward Collision Warning or Mitigation Systems (FCWS) Forward Collision Warning (FCW) systems alert the driver when the vehicle is about to collide with another vehicle some distance ahead. This is a very common type of collision that results in about 1.4 million crashes per year, or about a quarter of all collisions. FCW systems may be most helpful in alerting the driver of dangerous situations, helping him or her to respond more quickly as the need arises. The type of warning that the systems use vary between vehicles; some use a flashing light, while others use an alarm sound or vibration. The system monitors the relative speed and following distance from the forward vehicle, or the distance to an unmoving object if it is estimated to be in the forward path of the vehicle. When the combination of speed and distances (i.e., the time headway or the time to collision) becomes critical, a signal (audible, haptic, visual, or some combination) is presented to alert the driver. The area in front of the vehicle is monitored by a sensor (e.g., radar, lidar, and/or camera). Forward Collision Warning systems should not be confused with Forward Collision Mitigation (FCM) systems. Warning systems simply warn the driver when a collision is likely, but do not automatically apply the brakes. It is also important to keep in mind that different vehicles have the ability to detect different kinds of crashes. Some vehicles will only sound the alarm if it is about to collide with another moving vehicle, for example. Some systems use a flashing light to indicate a possible crash, while others play an alarming sound. Some systems only detect possible collisions with moving vehicles, whereas others work with both moving and stationary vehicles. Some FCW systems can also “prepare” the brake to make braking more effective. The Forward Collision Mitigation is also referred to as collision imminent braking, autobrake, and autonomous braking. FCM systems detect the distances and closing speeds of objects in the path of the vehicle and automatically decelerate or stop the vehicle if the driver does not respond to the alarm provided by the system. Crashes that are potentially preventable may still occur due to late braking and/or braking without sufficient force. Many drivers are not used to dealing with safety critical braking situations and do not apply enough braking force to avoid a crash. FCM is designed to reduce the number and severity of these types of collisions. Implementations of this class of technology may include FCW and Brake Assist (BA) technology that pre-primes the brake system. FCM is an SAE level 1 vehicle automation system (Function-Specific Automation) (NHTSA 2013). Relevant Technology Issues. FCW systems are camera-based or radar-based. Camera-based systems are less effective in low light conditions than radar-based systems and can be “blinded” by direct sun light (e.g., early sunrise and late sunset). The effectiveness of both radar and camera-based systems can be compromised by snow/ice build-up in front of the sensors. While a forward collision warning system may provide assistance in directing a driver’s attention to a potential collision event, drivers need to be aware that FCW technology still requires an appropriate response by the driver to avoid or mitigate the severity of a potential crash and that the technology is not

CRP Project HR 20-102(03) 18 designed to or able to warn for all potential crash situations. Drivers need to maintain an appropriate level of attention to the driving environment at all times even when driving with an FCW system active. Costs and Benefits. As previously noted, the aftermarket cost for Mobileye 560 is $850 with $150 in installation labor costs. As for benefits, studies have shown that FCW systems can reduce rear-end collisions by about 10%. Insurance studies have also shown that owners of cars equipped with FCW have lower claim rates than owners who do not have FCW. Overall benefit estimates for potential fatality reduction of 17% (5,633 cases out of 33,035), non-fatal injury reduction of 21% (146,000 cases out of 698,000), and crash rate reduction of 25% (1,453,000 cases out of 5,825,000) are given. (Case counts are based on combining annual relevant front-rear crashes and relevant single vehicle crashes.) Furthermore, IIHS has reported that insurance data on property damage loss claims show a reduction of 5% to 7% for vehicles with FCW and suggest that this translates into a 10% to 15% reduction in rear crashes (Lund 2013; Lund 2014). Besides, modeling using the German In-Depth Accident Study (GIDAS) database and effectiveness estimates based on reactions to a Bosch system by researches affiliated with the manufacturer estimated that FCW features would translate into a safety benefit of a 38% reduction in rear-end crashes (Georgi et al. 2009). In a simulation study, it was estimated that FCW could reduce rear-end crashes at a relative velocity of 20 km/hr by 30% (Yasuda 2011). Vehicle models equipped with autonomous braking are more effective than similar vehicles equipped only with forward collision warning. State of the Practice. FCW is most commonly available in luxury and higher-end vehicles, but that is expected to change in the next few years. Since all drivers need some help monitoring their surroundings, all types of drivers are expected to benefit equally. Early simulator based research suggested that FCW can redirect the driver’s attention to the road and improve reaction time (Lee et al. 2002). In a field study evaluating the impact of an integrated crash warning system, it was found that drivers were slightly more likely to maintain shorter headways; more time was spent at time headways of one second or less with the integrated system in the treatment condition (24%) than in the baseline condition (21%) (Sayer et al. 2011). Applicability to Freight. The FCW’s brake support can only help reduce the speed at which a collision occurs if the driver applies the vehicle’s brakes. The brake pedal must still be pressed, as in a typical braking situation. FCW is thus a warning only, and is to be distinguished from Forward Collision Mitigation, where the system may actually apply the brakes. For freight, the FCW along with FCM systems may be beneficial in preventing crashes. The detection systems however, have to be superior in order to provide warnings well ahead of time to allow the trucks to come to a safe stop, owing to their longer stopping distances. Blind Spot Detection Systems (BSDS) The BSDS is a vehicle-based sensor device that detects other vehicles located to the driver’s side and rear. Warnings are provided visually, audibly, or through vibrations. Blind spot monitoring systems utilize radar-based technology that is either mounted below side mirrors or in the rear bumper of the vehicle. The radar sensors emit and receive electromagnetic waves that are tuned to a specific frequency and distance by the system suppliers. When an approaching vehicle is within the range of the electromagnetic wave, it is reflected off of the approaching vehicle and sent back to the primary vehicle. This information is then processed by the Blind Spot Monitoring controller to determine whether or not an alert condition exists. If an alert condition exists, the warning lights in the A-pillar or side mirror will be illuminated (Consumer Reports 2013).

CRP Project HR 20-102(03) 19 Relevant Technology Issues. BSDS uses sensors mounted on each side of the rear bumper, corresponding warning lights for the interior, and an audible alarm. The system warns of other vehicles but it can also provide false warnings for guardrails and other objects. It is helpful when the BSDS is adjusted to its lowest setting of sensitivity to get warnings only when a turn signal is activated. Painting of the sensors should be done with caution. A single thin coat of base color and clear coat are recommended, only if required. Certain paints and/or multiple coats will affect performance and possibly delay detections and/or cause false detections. During certain slowing and stopping maneuvers, the blind spot system can produce false alerts. Blind spot systems should never be used at slowing or stopped speeds. A study used motorcycles and cars as the objects to be detected in the blind spot. When comparing the baseline pass-by tests with the target vehicle versus the target motorcycle, the motorcycle was detected on average 26% later than the target full-size sedan. As a result, the average separation distance of the vehicle test versus the motorcycle also resulted in a reduced distance of 14% on average (AAA Automotive Engineering 2014). The road’s curvature affects the performance of the BSDS. More research is required to mitigate lack of detection due to the curvature of the roads (Lee and Yoo 2013). Cost. The aftermarket cost for BSDS is estimated to range between $250 and $850. State of the Practice. Most radar-based blind spot monitoring systems utilize a short range pulse that has a potential range of approximately 100’. While the sensors have the ability to measure up to the maximum distance, their detection range is tuned by the manufacturer to a distance that is determined to provide adequate response yet not too far to create an annoying alert when vehicles are driven in typical traffic conditions. Most manufacturers do not use an audible alert for blind spot warning systems due to the annoyance potential of an audible alarm in traffic conditions. If a driver attempts to make a lane change toward a vehicle in their blind spot, the system may issue an audible alarm or flash the Blind Spot Monitoring warning light to indicate a potential collision. Some aftermarket sensors can detect only up to 10’ (standard mode). Some systems use a combination of cameras and sonar technologies which offer reliable detection in less than 120 ms. Mercedes and Infiniti use selective rear-wheel braking to steer the vehicle away from a potential accident. The Blind Spot Intervention system is only activated when a potential collision scenario is detected. Seventy-five percent of the 2014 model year vehicles offer blind spot detection as an option. The AAA evaluation of these systems reported some delayed warnings by the blind spot monitoring technologies and blind spot monitoring systems had difficulty detecting fast-moving vehicles, for example when merging onto a busy highway. Alerts were often provided too late for evasive action. Motorcycles were detected by blind spot monitoring systems significantly later (26%) than passenger vehicles, particularly when motorcycles were traveling at higher speeds; in some instances, systems missed motorcycles altogether (Mohn 2014). Heavy Truck Applications. Truck BSDS applications include those which sense the presence of objects located to the rear of the vehicle (referred to as Rear Object Detection Systems, or RODS), and those which sense the presence of objects on the right side of the vehicle (referred to as Side Object Detection Systems, or SODS). The rearward sensing systems are intended to aid drivers when backing their vehicles, typically at very low speeds, so they do not damage parked cars or other fixed objects, strike pedestrians, or impact loading docks at too high of a speed. Six rear systems are typically available per vehicle.

CRP Project HR 20-102(03) 20 The right side looking systems are intended primarily as supplements to outside rear-view mirror systems and as an aid for detecting adjacent vehicles when making lane change or merging maneuvers. Lane changes, especially those to the right, often present difficult challenges for drivers of heavy trucks, particularly in dense traffic situations. Four of these systems, two commercially available systems and two prototypes, were evaluated. Both rear mounted video cameras and ultrasonic-based RODS may have the potential to improve safety and reduce accidents while backing. A rear mounted video camera (with microphone) greatly improves a driver’s ability to see and hear what is behind the vehicle. The field of view of the camera evaluated in this program is adequate for most backing situations. The information acquired by the camera is transmitted to the driver in an easily understood form. Since this technology primarily increases the area a driver can see, it is very easy for the driver to determine whether or not the video camera is working. This increases driver confidence in the obstacle detection technology. In terms of driver confidence, there is no easy way for drivers to be sure that the ultrasonic RODS are actually working. When combined with the reliability problems that have been noted, drivers are reluctant to trust these systems enough to really gain benefits from them. In general, the results of these tests and evaluations tend to indicate that near object detection technology is still in the early stages of its development. Commercial truck drivers appreciate the value of RODS and SODS but improvements in the technology are needed before drivers can realize their full potential for preventing crashes. Improvements should focus on improving system reliability and the human factors aspects of the control and display interface. Emergency Electronic Brake Light (EEBL) EEBL is a V2V CV application that enables a vehicle to broadcast a self-generated emergency brake event to surrounding vehicles. Upon receiving the event information, the receiving vehicle determines the relevance of the event and if appropriate provides a warning to the driver in order to avoid a crash. This application is particularly useful when the driver's line of sight is obstructed by other vehicles or bad weather conditions (e.g., fog, heavy rain). It is another application that allows vehicles to “talk” to one another and with the infrastructure, auspiciously to provide more information to drivers about their surroundings. As with many CV applications, its success is in part a function of the depth of penetration of the V2V CV technology. It is intended for both light and heavy-duty vehicles. State of the Practice. Ford’s experimental “Electronic Brake Light” transmits a wireless signal to following cars, warning their drivers in a harder-to-ignore manner of what the car in front is doing. Instead of scanning the road ahead, drivers need only look at their dashboard to see if the car in front is slamming on its brakes. Cars equipped with the system can signal each other during emergency braking situations, with following cars illuminating dashboard lights to alert their drivers. Ford tested this tech on 20 S-Max vehicles in Frankfurt, Germany. The tests were part of Safe Intelligent Mobility: Test Field Germany, an industry-wide study of potential smart safety systems. Mercedes-Benz has similar, though broader technology it is looking to introduce, with its “Car-to-X” system. It will allow cars to observe road hazards and transmit vital information to following vehicles, giving drivers advance warnings of crashes, bad weather, and other obstacles (Edelstein 2013). Intersection Movement Assist (IMA) IMA is a V2V CV application that warns a vehicle’s driver when it is not safe to enter an intersection due to high collision probability with other vehicles at stop sign controlled and uncontrolled intersections. This application can provide collision warning information to the vehicle operational systems which may perform actions to reduce the likelihood of crashes at the intersections. While primarily developed for the

CRP Project HR 20-102(03) 21 light-duty market, research has been underway to address the unique needs of the heavy trucking environment (such as articulated tractor-trailers). Costs and Benefits. IMA and associated Left Turn Assist applcations could save as many as 1,083 lives and prevent up to 592,000 crashes annually. NHTSA estimates that the V2V equipment and supporting functions would cost about $341 to $350 per vehicle in 2020. That cost might dip to approximately $209 to $227 by 2058, after manufacturers gain experience producing the equipment, according to a NHTSA report (Harding et al. 2014). IMA has the potential to bring significant safety savings. As designed, IMA should address five types of junction-crossing crashes. These crashes, which collectively represent 26% of all crashes occurring in the crash population and 23% of comprehensive costs, can be categorized as follows: straight crossing paths at non-signal, left turn into path at non-signal (LTIP), right turn into path at signal (RTIP), running red light, and running stop sign. State of the Practice. Initial results from the U.S. DOT’s Safety Pilot Model Deployment indicate that this application may issue false warnings in a real world environment. In fact, various roadway geometries (e.g., cloverleaf, on-ramp, exit ramp) that do not represent a crash-imminent situation, can be incorrectly classified as conflict situations by the system. Improvements to the IMA algorithm for the second stage of driver evaluations indicate that these false warnings can be improved as the algorithms mature through additional testing. It may be necessary to develop new performance and test metrics that are designed to mitigate false warnings on different roadways such as curved roads and at non-perpendicular intersections. Curve Speed Warning (CSW) The CSW application allows a connected vehicle that is approaching a curve to receive information along with the recommended speed for the curve. This capability allows the vehicle to provide a warning to the driver regarding the curve and its recommended speed. In addition, the vehicle can perform additional warning actions if the actual speed through the curve exceeds the recommended speed (Stephens et.al. 2012). Relevant Technology Issues. The overall performance of digital map-based CSW can be significantly improved with the addition of vision-based sensors such as Lane Detection and Tracking (LDT) and vision-based rain sensing. The lane detection module can provide additional information about the shape and distance to the approaching bend, refining in this way the estimate of the maximum recommended speed. The vision-based rain sensor is able to detect rain-drops on the windshield and estimate the reduced road friction, an important parameter for the CSW module (euroFOT 2016). A CSW system was tested via a 4G network on a selected highway route with seven curves where vehicle speed accuracy and curve detection accuracy were evaluated. While a couple of curves were displayed in the wrong sequence due to their close proximity, curve detection results show that 85% of curves were successfully detected by the system (Qin et. al. 2015). Costs and Benefits. An advanced curve warning system was installed on five curves along I-5 in a mountainous portion of rural northern California. A before-and-after evaluation at two sites showed a significant reduction in truck speeds on downgrades greater than 5%. In a survey completed ten months after installation of a CSW application in northern California, 70% of commercial vehicle drivers indicated the signs were useful. Sixty-nine percent of the drivers indicated they reduced their speed through the curves in response to the signs (Tribbett et al. 2000). A dynamic curve speed warning sign using radar detection is estimated to cost between $9,000 and $14,000.

CRP Project HR 20-102(03) 22 Applicability to Freight. CSW holds the highest benefit for freight. A Roll Advisor and Control (RA&C) system provides added value to the CSW system. The RA&C system is designed to assist commercial vehicle drivers, especially drivers of tanker trucks, in avoiding rollover crashes. Based on driving data collected during a Freightliner FOT, in-vehicle rollover advisory control warning messages were expected to prevent 20% of rollover crashes caused by excessive speed in curves. For the national fleet of approximately 110,000 tanker trucks, the warning messages were seen to have the potential to prevent 34 crashes, 21 injuries, and two to three fatalities per year (Battelle 2003). Bridge Height Inform (BHI) The BHI application intends to provide the driver with information about bridge heights in the general area. Retrofit safety devices can be used to provide bridge height information to trucks. These are available as software retrofit devices that can be used in heavy trucks. The Oversize Vehicle Warning (OVW) application is similar to the BHI application. It uses external measurements taken by the roadside infrastructure, and transmitted to the vehicle, to support in-vehicle determination of whether an alert/warning is necessary. Specifically, the infrastructure data equipment detects and measures the approaching vehicle's height and width. The infrastructure component of the application transmits the vehicle measurements, along with bridge, overpass, or tunnel geometry, to the OVW vehicle application. The vehicle application utilizes these data to determine whether the vehicle can clear the bridge or tunnel. If deemed necessary, the driver is alerted to the impending low height and/or narrow horizontal clearance bridge or tunnel prior to a decision point, enabling the vehicle to reroute and avoid a collision. If the vehicle driver ignores the alert and continues along the route, the vehicle will generate a warning indicating an impending collision at a point near the bridge or tunnel approach. To support unequipped vehicles the infrastructure will display warning or reroute information when the measurements indicate that a vehicle does not have adequate height or width clearance. This application could be expanded to consider overweight as well as height and width. Smart Roadside Initiative Electronic Screening. E-Screening is a key component of the information collection systems and communications networks that support commercial vehicle operation, referred to as the Commercial Vehicle Information Systems and Networks (CVISN). E-Screening involves automatic identification and safety assessment of a commercial vehicle in motion. With E-Screening, safe and legal vehicles are allowed to continue on their route. Enforcement resources can be used to target noncompliant, unsafe vehicles and carriers. Currently, E-Screening occurs at fixed stations and on-demand verification sites. A study details the development of a CVISN electronic screening (E-screening) inspection site in Schodack, New York. The Schodack inspection site is fully operational with the following e-screening tools: permanent weigh-in-motion technologies, automatic vehicle identification sensors, and license plate recognition camera system. In addition, a local wireless network configuration is included in the e- screening system supporting on-site communications between equipment and providing enforcement personnel with 24-hour remote access to the system. Furthermore, the Schodack site is capable of collecting and monitoring bidirectional data and providing a data feed to the regional transportation management center (Clough Harbour 2012). In an effort to improve the detection of noncompliant commercial drivers and vehicles, researchers updated the E-screening system at the Schodack inspection site in New York. The inspection site was already equipped with 5.9 GHz DSRC to transmit automatic vehicle identification (AVI) information, automated license plate readers (ALPR), automated U.S. Department of Transportation (AUT) readers, and an overview camera. The following additional e-screening systems were tested across 240 vehicles: trailer/rear ALPR, vehicle over height detection sensors, and hazardous materials (HazMat) placard

CRP Project HR 20-102(03) 23 readers. The researchers validated the e-screening systems by extracting vehicle data recorded by the rear ALPR and the HazMat placard readers for the last ten minutes of every hour during a 24-hour period. The researchers analyzed both the rear ALPR and HazMat placard data for “read accuracy” (correct, incorrect, did not read). The researchers suggest that the rear ALPR and HazMat placard reader further increase the automation of roadside inspections, resulting in benefits such as reduced congestion, improved out-of- service rates, increased inspector productivity and efficiency, increased fuel savings, and reduced greenhouse emissions (Hamadeh 2013). Drivers in Georgia, Kentucky, Tennessee, and North Carolina participated in a year-long program testing the effectiveness of an infrared brake screening system. The infrared screening system identified 84% of vehicles for further inspection, compared to traditional screening systems which only identified 34% of vehicles. Truck drivers indicated increased time efficiency, while vehicle inspectors indicated increased accuracy as benefits of electronic screening systems (Maccubbin et.al. 2008). Virtual Weigh Stations. Virtual Weigh Stations/Electronic Permitting was the focus of an Enforcement Technologies Study conducted in 2008 and 2009. The focus of the study was to develop the foundation for roadside technologies that can be used to improve truck size and weight enforcement. The Virtual Weigh Station concept will further increase the number of electronic screenings and, depending upon the virtual weigh station configuration, will provide a more enhanced safety and credentials assessment. Wireless Roadside Inspection (WRI). WRI Program research is being conducted to increase the number and frequency of safety inspections at the roadside and obtain data about the commercial vehicle and its driver. It is capable of monitoring and flagging compliance violations in hours of service, licensing, and regulation. Studies are examining technologies that can transmit safety data directly from the vehicle to the roadside and from a carrier system to a government system. The safety data being considered for transmission include basic identification data (for the driver, vehicle, and carrier), the driver’s hours of service record, and sensor data that provide information on weight, tire, and brake status. Enforcement systems and staff will use this data set to support E-Screening and inspections at locations such as staffed roadside sites, virtual weigh stations, and on-demand verification sites. Smart Truck Parking. Truck parking research and ITS-based project deployments provide commercial vehicle parking information so that commercial drivers can make advanced route planning decisions based on hour-of-service constraints, location and supply of parking, travel conditions, and loading/unloading. To address the truck parking issue in Florida, researchers deployed a “smart truck parking” pilot study. Each truck parking site in the study received a rating of low, medium, or high truck capacity. Truck capacity was determined through the use of wireless ground sensors. Based on the truck data gathered, the researchers incorporated geographic information systems to develop real-time parking information maps, as well as a report generation module that provided historical parking capacity information among the various sites. The research suggests that drivers could access this information through devices such as smart phones (Fender 2014). Truck Platooning Platooning is commonly used term representing use cases that pair together two or more vehicles with electronic communication between them, and sufficient on-board controls to maintain safe following distances under most circumstances. Human operators (drivers) in the following vehicle(s) cede control of the spacing of the vehicle, and possibly also its lateral positioning (steering), to the system. The lead vehicle is driven more or less as usual, and technology replaces drivers in following vehicle(s) to maintain a more optimal and consistent spacing and with the ability to react more quickly and safely as needed.

CRP Project HR 20-102(03) 24 Platooning is certainly not limited to trucks, and indeed mixed truck and passenger vehicle platoons are possible. In the trucking context, platooning sometimes is referred to as “road train” and “electronic towbar”, though today platoon is the most commonly used term. The U.S. trucking industry has used the term “Driver Assistive Truck Platooning” in its tests, to more explicitly recognize that the testing performed to date involves a significant and active role for drivers. Platooning concepts have been studied for many years. Prior to the viability of V2V and V2I solutions, concepts were constrained to heavy infrastructure investment to provide the necessary guiding mechanism and therefore were of more limited practical utility. A more contemporary application of platooning has emerged that uses V2V communications to help maintain situational awareness and coordinate actions between vehicles. This technology, coupled with other improvements such as bandwidth availability, software and algorithms available on the truck, and quality of the sensors, lead many to believe that solutions are getting close to commercial viability. A definitional overview of platooning might best begin with Adaptive Cruise Control (ACC), a feature that is well known and has been deployed on Class 8 trucks in the U.S. since 2008 (ATA TMC 2015). ACC automates the throttle and braking such that a consistent time gap is maintained to the vehicle directly in front of the subject vehicle, introducing radar technology to the vehicle to detect the vehicle directly in front. Cooperative Adaptive Cruise Control (CACC) represents various vehicle-following control concepts, mostly basically by adding V2V CV capabilities to ACC. CACC automates the vehicle speed through exercising throttle or brakes, but the driver remains in full control of all other driving operations and thus remains alert to driving conditions at all times. By definition, CACC represents SAE level 1 automation. It might be loosely referred to as a simple implementation of platooning, though research led by the California PATH Program (Nowakowski et al. 2015) offers three differences: (1) CACC research usually uses a constant-time-gap approach to how far the vehicle is kept behind the vehicle in front, mimicking how people typically drive, whereas platooning relies more on constant-distance-gap strategies given the importance for the distance between vehicles in accruing the draft-related benefits; (2) CACC does not include automated lateral control where platooning implementations might be expected to; and (3) commercial implementations of platooning require some additional degree of a business infrastructure around them, for example to manage the formation and dissolution of platoons in manners viable to the platoon’s participant(s). Testing of V2V-based platooning has used DSRC as the wireless communications channel, although depending on characteristics of an operational deployment, such as minimum headways, it could at least in theory be possible to deploy with a communications technology other than DSRC. In addition to this basic V2V communications mechanism, CACC includes forward sensing (radar), positioning and actuation. A Human-Machine Interface (HMI) must also be provided for the truck driver to engage with and disengage the platoon. That HMI will likely be different for the lead driver as for following drivers. Effecting lateral automation through steering control is itself a meaningful step in vehicular automation, for functions such as automated lane keeping that further reduce the responsibilities of the driver. Many of the Original Equipment Manufacturers (OEMs) worldwide have hardware (whether developed themselves or available through vendors) for lateral control and/or have discussed adding features requiring lateral control to the platooning packages in the future. Adding lateral control comes with the trade-off of a lightweight system. Adding lateral control significantly increases the complexity of the control systems implemented. In the near-term, this drives the initial investment cost of systems that utilize lateral control higher, while also increasing their potential benefits. Environmental Benefits Platooning offers the potential of meaningful and measurable environmental benefits, and tests in the U.S. and elsewhere have quantified those benefits. Trucks traveling in a platoon operate with less headway than in non-platooned, traditional operations, and demonstrations have shown that this

CRP Project HR 20-102(03) 25 difference in headways causes platooned trucks to operate with reduced aerodynamic drag, which leads to fuel savings and a reduction in emissions. This occurs because the back truck(s) operate within the slipstream of the front truck, thus reducing the pressure gradient on the truck front. Further, because the back truck(s) disrupt the formation of low-pressure vortex structures behind the lead truck, the lead will experience some fuel savings as well. Because this is the major source of benefit, and because its effect is minimized at slower speeds, platooning is most applicable in highway environments rather than, for example, at ports and other terminals. The extent of drag reduction, and the associated improvements in fuel efficiency and emissions reductions, has been shown by these tests to be a function of several operating factors including:  Number of trucks and position of the truck within the platoon  Headway between trucks  Truck geometry, noting that the majority of Class 8 trucks in the US are nosed  Lateral offset of the trucks, which can be significant if the platoon is not operating with any automated lateral control  Speed of the platoon  The truck’s weight As aerodynamic drag is responsible for over 60% of fuel use at highway speeds (National Academy of Sciences (NAS) 2010), the drag improvement has a direct translation to fuel savings. Several tests of Class 8 trucks have been conducted in the U.S. and elsewhere to empirically measure the fuel consumption reduction due to platooning. Among these tests are those performed and reported by the National Renewable Energy Laboratory (NREL) (Lammert et al. 2014), California Partners for Advanced Transit and Highways (PATH) (Browand 2004), the Federal Highway Administration (FHWA) (Humphreys et al. 2016; Auburn University 2016), the North American Council for Freight Efficiency (NACFE) (Roeth 2013), the Japan Ministry of Economy, Trade, and Industry (METI) (Tsugawa 2012), and Safe Road Trains for the Environment (SARTRE) (Davila 2013). Sample results of these tests are given in Table 3-1. Though direct comparisons are impossible due to differences in the tests’ operating characteristics, there does appear to be a level of consistency in the general magnitude of savings due to platooning. Table 3-1. Comparisons of Fuel Consumption Savings Measured in Selected Platooning Tests. Test, Year, and Country Example of Reported Savings for Lead Truck (%) Example of Reported Savings for Following Truck(s) (%) Notes FHWA/Auburn test, 2015, U.S. 2.0 10.2 Peloton system run on test tracks; driver steering; results shown for 65k lbs Gross Vehicle Weight, 50’ following distance and platoon speed of 65 mph NREL test, 2014, U.S. 3.1 9.2 Peloton system run on test tracks; driver steering; results shown for 65k lbs Gross Vehicle Weight, 50’ following distance and platoon speed of 65 mph

CRP Project HR 20-102(03) 26 Test, Year, and Country Example of Reported Savings for Lead Truck (%) Example of Reported Savings for Following Truck(s) (%) Notes NACFE test, 2013, U.S. 4.5 10.0 Over-the-Road (OTR) tests on Utah I-80 of 2 fully loaded trucks, using Peloton system and with driver steering; results shown for 36’ following distance and platoon speed of 64 mph SARTRE test, 2012, E.U. ~ 4.8 ~ 9.5 2 Volvo trucks run on test track; results shown for 12 m gap and platoon speed of 84 km/h “Energy ITS” Project, 2010, Japan 7.5 17.0 3 truck test on flat Expressway, empty- loaded; automated steering; results shown for 10 m gap and platoon speed of 80 km/h PATH, 2003 6 10 2 empty-loaded trucks; 10 m gap and platoon speed of 55 mph Note: Source citations for these tests are provided in the accompanying text. Within the differences in parameters used by each of these tests, there has been rough, general agreement in the results (each of the tests shown in the table made many runs, not shown here, to vary one or more particular parameters). The relationship between fuel savings and gap/headway between vehicles has documented that savings decrease as the gap increases toward a natural headway that today’s non- automated trucks run, with the important exception of trucks running at short gaps and laterally offset from one another. The FHWA/Auburn test controlled for this using simulations, showing that lateral offsets between platooning trucks significantly reduce the savings, particularly at closer gaps. As little as a two foot offset between two platooning trucks reduced by one-quarter the percentage drag reduction (which directly correlates to fuel reduction) on the back truck at gaps of 50 feet (Humphreys et al. 2016). Thus the environmental benefits of a platooning system that includes automated lateral control should be expected to be greater than one without this feature. Fuel use, specifically the combustion of fossil fuels, is the primary greenhouse gas emitted from human activities. The U.S. Environmental Protection Agency (EPA) has estimated that the transportation sector contributed 1,742 million metric tons of equivalent carbon dioxide from fossil fuel combustion in the U.S. during 2014, or about one-third of all CO2 emitted from all sources (EPA 2016). They estimated that 23.1% of the sector’s CO2 emissions were attributable to medium- and heavy-duty trucks. There is a direct correlation of reduction in fuel use through platooning, and reduction in greenhouse gas emissions. Reduced fuel use encourages the business value proposition for trucking organizations and ultimately their customers. The environmental benefits accrue in a less tangible way to these concerns, and in a direct way to the missions of public sector organizations such as the U.S. DOT. A test in Japan (part of the Energy ITS project) used a simulation to estimate the CO2 emission reduction due to platooning vehicles. Considering a Tokyo area expressway which was populated 31% by heavy vehicles, and 40% of those were platooning, the simulated emissions reductions due to reduced aerodynamic drag was 2.0% for platoon gaps of 10m and vehicles traveling 80 km/hr, and 3.5% when the gap was decreased to 4m (Tsugawa 2012). Operational considerations affect fuel savings and therefore emissions reductions. For example, the means by which platoons form may involve extra miles being driven to meet with a partner truck, or it could involve a trailing truck increasing its speed to meet the partner truck.

CRP Project HR 20-102(03) 27 Operational and Deployment Considerations Platooning systems are in active development. While solutions that do not include automated steering control could be deployed, in a broader market sense it is probable that solutions including this lateral control (SAE level 2) would be more popular to capture driver benefits as well as the aforementioned full aerodynamic benefits. Many questions currently exist regarding how platoons will operate. Indeed, how these questions are resolved will drive the eventual success and depth of deployment. Platoon Establishment and Dissolution. A driver intending to join a platoon may need to request the desired maneuver and wait for approval and further instructions from the lead vehicle before initiating any maneuvers. Early adopters are likely to be operators with large fleets and a degree of predictability to their routes. Such adopters will have ultimate control over the protocols used to form and dissolve platoons. There is an added layer of complexity if or when operators decide to permit platooning with other fleets. Platoon Management. Some nature of centralized control function is seen as needed to manage platoons. (A longer term future may include more fully automated trucks that could form platoons in a decentralized manner. Such a scenario is briefly discussed below in the Highly Automated Trucks Section and is not included here.) Management largely includes successfully forming and dissolving platoons. Platoons would form when the business decision is favorable for a formation and any external variables such as weather conditions are favorable. A vital consideration in formation is the ordering of the trucks. Because trailing trucks enjoy greater benefits (from fuel savings, as discussed below), this becomes a business issue that requires deliberate processing. More importantly, the ordering could be a safety issue if a trailing truck has a longer braking distance that a lead truck. Factors affecting braking distances such as brake maintenance status and trailer loadings are a challenge to reliably measure in an operational environment. Different types of organizations might vie for this new business opportunity of providing the management function. Two disparate examples of private companies that have publicly stated their interest in operating such a function include a technology provider (Peloton 2016) and a truck OEM (Scania 2016). Highway Infrastructure. Truck platoons do not require dedicated lanes to operate. However, some jurisdictions could require this, or just have a bias toward dedicated lane solutions for platoons, as a way to mitigate concerns about mixed traffic environments and the difficulties of individual non-platooned vehicles engaging in the lanes occupied by a platoon. At this time, empirical evidence to justify, allay, or qualify such concerns is lacking. Jurisdictions that do consider dedicated truck lanes for platooned or otherwise automated trucks would likely realize the following benefits above and beyond those benefits from platoons deployed in mixed traffic lanes:  Safely shorten the headways between trucks.  Address concerns surfaced by road users of how to interject non-platooned vehicles within a platoon, for example to move to rightmost lanes to access highway exit ramps.  Address concerns of how a tightly coupled platoon would safely respond to exceptional situations occurring in general traffic, such as swerving drivers or quick application of brakes by drivers. Managed lanes might be created more generally as a solution to automated vehicular traffic. Platoons may be a part of this, and mixes of platoons that are not truck only could be possible. Whether such a scenario would provide adequate benefit to jurisdictions that pursued it is an open question. Answers are likely to be forthcoming as small-scale deployments begin in the U.S. and elsewhere.

CRP Project HR 20-102(03) 28 Safety as a Benefit in a Business Case “Safety technologies must be seen as preventing future costs caused by accidents. The cost of accidents can be significant for a carrier, but they occur intermittently. Smaller carriers tend to view these potential costs as risks they must take because of the steep up-front cost of implementing new technologies.” ITS America (ITS America 2015) Another potential concern regards the increased degradation of infrastructure elements such as bridges would incur due to the closer spacing of trucks. Research is needed to estimate this impact. Platoon Lengths. Most testing, certainly within the U.S., has been on two- and three-truck platoons (see the Predicted Timing Section below for examples). The operational use of longer platoons is both possible and anticipated at least in an exploratory phase, and could accrue proportionally more benefits as a higher proportion of platooned vehicles enjoy the slipstream benefits of being behind a truck to the front of them. There are no limiting technical constraints with the technology solutions to running platoons of longer than three trucks. With the likelihood of using DSRC-based communications, a technical limitation on platoon length comes as the communications range of the DSRC technology begins to be reached (around 300m). As an example, eight tractors linked to 53’ trailers, platooned with 50’ headways, might safely operate within this framework. Operational and policy constraints would likely intercede on maximum platoon lengths before such a technology-driven constraint became active. Specifically, finding and coordinating many trucks traveling together along the same corridor at the same time will be challenging. The public’s acceptance of very long platoons is likely to be a constraint, particularly with early deployments. Unease over the concept of platoons, including how non-platooned vehicles would navigate through such a trail of platooned vehicles, may limit this acceptance. Over time, if it is demonstrated through deployments that this concern is not real, then public concern may dissipate. Business Case for Platooning Industry needs to understand the business case that these new technology-driven opportunities present, and ensure that new solutions meet thresholds for adoption. Governmental agencies need to know that these business cases exist, as one input to creating a proper regulatory environment that respects the vitality of the private enterprises. Some research results are emerging, notably from a 2015 study sponsored by the FHWA (Auburn University 2015). There is also an important body of work emerging from E.U. based research including assessment of platooning business cases (Janssen et al. 2015). Results from one region may not map directly into other regions’ contexts, but can certainly help inform them. Substantial unknowns remain however. At the present time, the existence and parameters around a viable business case are a key factor for industry as the technology nears market readiness. Additional research, testing, and vendor solution development is needed. User business cases are particularly critical in CV and AV trucking applications, since the users are public and private companies versus the general public. Carriers must understand their return on investment to justify deploying these technologies. A 2015 study sponsored by the Intelligent Transportation Society (ITS) of America (ITS America 2015) of advanced safety technologies revealed that 13% of carriers expect a return on investment within 12 months, 40% within two years, 39% within three years, and only 8% longer than three years. Thresholds for adoption vary greatly within the trucking industry. Class 8 trucks controlled by owner-operators number around 235,000 in the U.S. (Heine 2013), and in a 2015 survey conducted by ATRI, those owner-operators who indicated a willingness to pay for the system (who were only 30% of all responding

CRP Project HR 20-102(03) 29 owner-operators) had a median response of six months as their acceptable break-even threshold for willingness to adopt a platooning system (Auburn University 2015). For-hire and private carrier fleet managers responded to the same question with a break-even threshold of 18 months. Safety Benefits. Truck platooning holds a promise of improving safety through the reduction in frontal collisions, the most common highway accident type for heavy trucks in the U.S. Additional testing is needed to both validate this assertion and look more holistically at how highway safety might be affected by platooning deployments. Conceptually, the automated control of a truck’s movements, whether lateral control is included or not, should reduce accidents as the driver reaction time is eliminated as a concern. Even if the platooning function is not engaged in a platoon-ready truck, the radar and CV technologies can at least shorten the driver awareness of a situation. Improved safety affects a business case via higher system productivity and fewer injuries and damage costs. Labor Impact. The effect of labor costs on a business case depends substantially on the platooning concept under consideration. Phased implementations are likely, beginning with early systems that might provide only longitudinal control, to future states including additional levels of automation including lateral control and possibly even in which the technical, operational, and legal environment permits platoons with a driver in the lead truck only. Labor costs would obviously be greatly reduced in this driverless scenario; however, it is distant enough in the future that it does not affect a current business case analysis. Also, the penetration of an operationalized concept, that is, the extent to which OTR routes can be platooned, also greatly affects the labor savings possible. Asset Utilization Savings. Platooned trucks should be operated more efficiently, though this is highly a function of the specific platooning concept deployed. The extent to which trucks in a fleet are physically able to create a platoon is also a major factor, that is, traveling along the same corridor at roughly the same time, and in sufficiently compatible equipment. Companies operating large private fleets, with homogeneity in their tractor OEM, and predictable routes may gain the best savings along their high density routes. Truckload (TL) and Less-Than-Truckload (LTL) trucking operations would also be early candidates to gain the best utilization savings. Fuel Savings. As discussed above, a substantial body of research exists to help trucking organizations understand the savings due to a truck operating in a particular position within a platoon. Savings of roughly 4-6% for the lead vehicle, and 8-10% for following vehicles, are likely, though with the many caveats that were discussed above. Societal Value. The public sector should be interested in public level benefits of platooning. Road capacity utilization improvements should accrue, though whether at a level that is meaningful is yet to be determined. Traffic flow should be improved, though this is currently unproven. On a lightly used roadway there might be little effect in traffic flow. Likewise, in stop-and-go traffic there may be little effect. But in conditions where a road is approaching capacity, there may be traffic flow improvements due to the predictability and reduction in the inter-truck gaps. Highly Automated Trucks Highly automated freight trucks, particularly those functioning at SAE levels 4 and 5, hold the prospect of significantly reducing labor costs by eliminating driving functions for meaningful portions of the goods movement process. Such trucks are in use today in very limited and controlled environments, such as in mining operations and farming (that is, not on shared roadways). It is therefore an area that is

CRP Project HR 20-102(03) 30 commanding attention and level of research investment. Many of the technology-oriented barriers are successfully being addressed. “Softer” barriers such as managing public perceptions have had less attention to date. Trucks that might be described as “highly automated” still have drivers on-board. The notion that a heavy-duty commercial freight hauling trucks might operate over its entire route from origin to destination with no driver is one that is unlikely to be realized in any near- or mid-term timeframe. Among the barriers to deploying trucks without any drivers are:  Difficulty of operating in the complexities of urban environments  Resistance from the general public  Indeterminate fail-safe solutions for when a highly automated truck faces a situation that it is unable to successfully negotiate  Inability of today’s automation solutions to safely handle all situations, including weather, construction zones, blocked lanes, etc.  Other security and safety concerns with the truck, trailer, and cargo Nonetheless, there are elements of highly automated trucking and particular use cases that are more likely to emerge in a near-term future and therefore that warrant the attention of the stakeholder community. In OTR operations, the earlier discussion in this chapter of truck platooning discussed elements of automation such as brake, throttle, and steering control. With additional functionality such as automated lane changing and fail-operational or fail-safe implementation, a truck could be technically capable of driving meaningful portions of an assignment (1) without a human driver and (2) without the need for a driver to be readily available to take over the driving function. This earlier discussion made the assumption of trucks with drivers that become a part of a platoon through some central management and control function. In a context of higher level automated trucks, one can certainly extend the platooning concept such that self-driven trucks are able to find other suitable trucks to create ad hoc platoons, thus gaining not only the labor advantage but also the fuel and environmental benefits of the platoon. At the SAE level 3 automation level, there are applications such as Traffic Jam Assist that OEMs are likely to deploy as part of broader technology packages. Traffic Jam Assist essentially combines Adaptive Cruise Control and a Lane Keeping Assist function for operation at slow speeds including “stop-and-go”, intended for heavy traffic situations. These applications hold a promise of reducing fuel usage and reducing driver fatigue. It is likely that all the major OEMs selling in the U.S. market are invested in some aspects of highly automated technologies, though only certain lines such as Freightliner and Peterbilt are actively publicizing their developments. Particularly in the U.S. market, some of the capabilities will be delivered by Tier 1 suppliers. Acceptance of highly automated trucks is a critical issue. The public conditioning that is beginning in the highly automated passenger vehicle space is applicable to trucking, in that the degree and speed of the public’s acceptance of highly automated cars will be a stepping stone to its acceptance of the much larger highly automated heavy trucks. Likewise, barriers that might arise in the passenger market could affect the nature or speed of acceptance of inherently truck-based solutions. A promising deployment scenario of such a system is in long-haul routes currently driven by a team. Reducing a team of two drivers to a single driver would be a direct and meaningful cost savings to the operator, while maintaining human presence on the truck. This would largely address one of the major concerns with driverless trucks, that the physical security of the truck including trailer and its contents could be jeopardized without the presence of a professionally trained driver. With the remaining driver able to take a rest while the automated system was deployed, this seems to conform to the driver’s Hours of Service Rule on-duty time, while keeping the truck productively moving.

CRP Project HR 20-102(03) 31 Ottomotto Creating Autonomous Retrofit Kit “From a technology perspective we felt we can solve the problem sooner than later because the vast majority of truck miles are on-highway miles, and highway by nature is a much more constrained environment to introduce self-driving technology …It’s just simpler…There are no pedestrians and there are a handful of trucking corridors, so it’s easier to map.” Lior Ron, Ottomotto Founder As with passenger vehicles, one critical criterion for the success of the concept is the ability to truly disengage drivers from the driving experience and for long periods of time without a need for the driver to re-engage on a moment’s notice. In the trucking sector, the technology would need to be approved by NHTSA and FMCSA. If a solution pairs the automated system with a single driver, for example replacing a driving team with a single driver, it is less clear whether federal action is required since Hours of Service Regulations count the driver as “off the clock” if resting in a cab’s sleeper berth. It is also unclear at this time whether a highly competitive market will emerge for automated technology in trucking the way that has recently been observed in the passenger vehicle market. In 2016 both Tesla and Ottomotto (subsequently acquired by Uber) publicly announced intentions to bring to market such a system (and others are believed to be in a research and development phase but have not made public announcements). In 2016 Ottomotto, the state of Colorado, and Anheuser-Busch announced a successful test of a product delivery using an Ottomotto-retrofitted truck in which a 120-mile interstate portion of the trip was accomplished with a driver on-board but not operating the truck. Inter-city heavy-duty trucks spend a preponderance of time on limited access highways, which presents a significantly more feasible environment. Indeed, Ottomotto has publicly stated that their aim is “to focus on less- complicated highway driving so the technology can quickly pass muster” (Ramsey 2016). Whatever level of competition emerges, there is a “first-to-market” interest that entices such companies to want to deploy quickly, thereby introducing pressure on public regulators to act appropriately while not stifling innovation. With the promotion of quick deployment, characteristics of a solution are likely to include:  Operating on roads that are known to be compatible with the technology  Curtailing automated operations in certain defined situations, such as inclement weather  Operating on limited access highways only  Operating in states with a favorable climate to permitting such vehicles These solutions have requirements related to road infrastructure, such as lane markings with a quality that vehicle-based sensors can consistently read them. Safety-focused stakeholders including operators of highly automated trucks would want to know the state of the road infrastructure before traversing it. A conceivable public sector solution might be a certification scheme for roads, indicating that a road meets certain well-defined standards. Or the operator or technology provider may effectively fill this role. Another application area is in automating yard functions such as those at distribution centers including dock locating, auto coupling, backing to a gate and refueling. Such applications would generally occur in controlled, privately owned and operated space and therefore make some aspects of deployment easier. The questions that need to be answered for such applications are similar to those in the OTR scenarios, but with different dynamics – safety considerations, evaluation of infrastructure changes needed (for example, whether concrete sensors may be needed), and concepts of operations that can drive articulation of a business case (for example, if a driver is not needed to back a truck to a dock, can labor savings really accrue given the other responsibilities that the driver has around the docking function?). Most of the above discussion is of applications outside of the urban environment (Traffic Jam Assist being the exception). The deployment of highly automated trucks in urban environments will likely lag that of the passenger vehicle segment, which itself has uncertain timing. Trucking has many of the same challenges as passenger vehicles have, plus additional operational, safety, and security concerns, for

CRP Project HR 20-102(03) 32 example maneuvering trucks in complex geographies and pedestrian environments. OEMs may build and deploy information and communications technologies such as location-based services that apply mainly to urban-oriented “last mile delivery” use cases, but these may be technically outside the specific CV and AV domains. There is substantial research and development activity occurring in general in the urban freight movement area in the U.S., and even more in other regions, particularly the E.U. While outside the scope of this research project, it is noted here that the public sector at all levels should monitor developments around the coupling of automated trucks with electric truck technologies, unmanned aerial vehicles (e.g., drones), and other cutting edge technologies that could further disrupt this market. Driver Issues Issues surrounding the driver are central to when and how the technologies discussed throughout this report will be successfully deployed, increasingly so in the higher level automated solutions. The trucking industry will need to (and indeed has begun) to achieve a level of understanding of these issues. The context is larger though, as consideration must be made from stakeholders including labor, technology providers, local government planners, and regulators. The pervasiveness of driver issues is evidenced by the many places in this report raising driver issues, including discussion of the business case, regulatory issues, and local workforce planning. Driver acceptance of CV and particularly AV solutions will be determined as deployments begin. Driver perspectives should be actively brought into the developmental phases. The HMI is an example of this need. While this report details many examples of how elements of the passenger vehicle market extend into trucking, that is less so the case in this area with the professional, trained workforce operating the vehicles, and the physical characteristics of heavy-duty vehicles obviously very different than passenger vehicles (for example, whether and how haptic responses are provided to a driver is going to be very different). How drivers respond to new features will be critical. Every nature of technology, safety, and legal consideration for a feature such as automated lane changing may be solid, but if drivers do not feel the feature is safe, or not appropriately applied (for example, it is too timid), then acceptance will be difficult and this will likely translate to lack of market acceptance. If systems are brought to market in a manner that drivers find attractive, then in addition to the business success, driver approval and acceptance could help the driver profession rather than be seen as a threat. There is anecdotal evidence within the industry that younger driving recruits view future AV solutions with a certain level of ‘cool’ factor. CV and AV technologies have the potential of reducing driver stress, and cutting monotony on long trips. It could also have a positive effect on driver health. Daimler Trucks North America conducted a study showing that driver sleepiness was reduced by as much as 25% when operating their SAE level 3 AV test truck and when the drivers reported they felt more rested (Lockridge 2015b). A primary motivator for interest in AV for trucking is attributable to the prospect, as yet unproven, of labor savings by permitting drivers to conduct other tasks while in an automated truck, or ultimately eliminating the driver and the associated labor cost. The driving function in general, and for heavy trucks particularly, requires quick decision making in a great variety of situations (and not always highly patterned after past events), and precision judgments often in a very short amount of time. Additionally, the responsibilities of a driver extend beyond simply operating the truck. Commercial freight hauling environments require drivers to perform other functions often including security, safety, managing aspects of a route, complying with various rules, and performing functions related to loading and unloading at terminals, among others. The integration of AV technologies into this environment is a complicated proposition, and one that requires active input and consideration from the industry, labor, and regulators.

CRP Project HR 20-102(03) 33 Port CV and AV Applications In the near future, CV applications and AVs are expected to open new opportunities to enhance port operations, from ship side to road and rail. CV systems are increasingly coming online and will be common by 2021. While widely deployed AVs may be further out, AVs are already operating in a number of ports, moving containers between storage areas; these capabilities will be considered in each operational phase. CV applications leverage data flowing between vehicles, infrastructure units, and wireless access points. One suite of applications developed under the U.S. DOT’s Dynamic Mobility Applications (DMA) program is the Freight Advanced Traveler Information System (FRATIS). FRATIS applications provide optimization that can be applied to port drayage, and provide pre-trip and real-time en-route information for truck fleet operations. FRATIS also offers route guidance aimed at reducing delays. In addition, other DMA applications like Freight Signal Priority and Intelligent Signal Systems on arterial and local roads, CACC and Speed Harmonization on freeways offers potential mobility benefits to trucks using the facilities in and around ports. For AV concepts, possible solutions include platooning in drayage operations, autonomous sea to shore vehicles used in ports, and driverless container transfer at intermodal terminals. Automation has become common in many European ports, starting in 1990 when the first automated stacking cranes were installed in Rotterdam. Today, extensive automation is in the initial stages of rollout; as an example, Shanghai, the world’s busiest container port, is almost entirely manual, while Rotterdam’s Maasvlakte II terminal, which opened in April 2015, has no personnel inside its cargo-handling section. Keying on ports using automation, scans examine select operations during each phase focusing on automation that gains best efficiencies. One area of particular focus is the backup of road vehicles (trucks) at the gate phase, as was seen in Los Angeles/Long Beach in 2015, a result of lengthening queues from labor shortages, and which caused road traffic to further back up in the surrounding areas. Automated systems that carry out ship-to-shore movements result in time, fuel, and wage savings. Beyond automated equipment for unloading and stacking, autonomous vehicles are expected to offer significant improvement to logistics in general. Such technologies include Daimler driverless trucks, Google’s driverless cars, and even the Rolls Royce prototype drone ship unveiled in 2014. Implementing automated trucks and delivery vehicles at terminals and within logistics networks will optimize supply chains. When coupled with information and communications technologies focused on end-to-end logistics scheduling, automation can improve throughput and reduce traffic congestion in and near ports, which reduces emissions and mitigates environmental impacts.

CRP Project HR 20-102(03) 34 Legal, Regulatory, and Policy Topics Laws and regulations relating to autonomous vehicles are still in their infancy. Since the technology has been evolving rapidly, federal and state legislators, regulators and policymakers have been hesitant to enact laws before they understand how the technology will develop and what the problems that need to be addressed are. Current laws and vehicle regulations are often centered on the driver and the driver’s decision making. Many changes need to be incorporated into the legal code to incorporate CV and particularly AV technologies and autonomous vehicle operation. A handful of states have passed laws to address some of the issues around autonomous and connected vehicles. Most of these laws have expressed a general desire to promote CV and AV technologies, and specific trucking applications like platooning. There is growing concern among industry and public policy makers that a patchwork state- by-state approach to some of these issues could serve as a barrier to implementing CV and AV technologies. In addition, uncertainty for industry about how these new technologies will be treated could serve as a barrier to implementation. The following section highlights the current legislative, regulatory and policy environment for autonomous vehicles and offers some considerations for improving it. Current Legislation Legality of Autonomous Vehicles The federal government and states are still working out the right balance between providing a legal framework to promote CV and AV technologies, and taking a hands-off approach to let the technology develop. In its 2013 Preliminary Statement of Policy Concerning Automated Vehicles, NHTSA notes that states are well suited to address issues related to licensing, testing and operation of self-driving vehicles (NHTSA 2013). NHTSA did not recommend that states permit operation of self-driving vehicles for purposes other than testing. NHTSA released the Federal Automated Vehicles Policy in September 2016 which provides a model state policy and describes a framework for the regulation of vehicle testing. In the absence of specific laws that regulate automated vehicles, current law in most states would appear to allow them, as long as a driver is present. A comprehensive review of the laws governing motor vehicles conducted in 2014 concluded that “computer direction of a motor vehicle’s steering, braking, and accelerating without real-time human input is probably legal” under existing law (Smith 2014). The analysis began with the principle that “everything is permitted unless prohibited” and examines three legal regimes: the 1949 Geneva Convention on Road Traffic, regulations enacted by NHTSA, and the vehicle codes of all fifty U.S. states. The Geneva Convention does not prohibit automated driving. The treaty does require uniform rules that include requiring “every vehicle or combination thereof to have a driver who is at all times able to control it.” Vehicles that have a human who is able to intervene in an automated vehicle’s operation would appear to satisfy this requirement. The FMVSS, promulgated by NHTSA, requires new vehicles to be certified, but “do not generally prohibit or uniquely burden automated vehicles” (Smith 2014). State vehicle codes typically assume the presence of a licensed human driver who is able to exercise human judgment. These laws would be consistent with autonomous systems where a human driver is available to take control and exercise this judgment when needed. There are some state laws that may limit the utility of autonomous driving systems. For instance, New York law requires that a driver keep one hand on the wheel at all times. Many states have rules mandating “reasonable, prudent, practicable, and safe” driving. There is some uncertainty as to how these rules apply to automated vehicles and their passengers. Many states have following distance requirements that would limit the lawful operation of truck platoons.

CRP Project HR 20-102(03) 35 States with Autonomous Vehicle Legislation Seven states and Washington DC have passed significant legislation related to autonomous vehicles. In 2011, Nevada was the first state to explicitly authorize their operation. In the years since, six other states have enacted AV-related legislation, including California, Florida, Michigan, North Dakota, Tennessee and Utah. State laws have explicitly authorized testing of autonomous vehicles on roadways in California, Michigan, Nevada, and DC with a driver in the vehicle. The Governor of Arizona issued an executive order to promote the testing of these vehicles as well (State of Arizona 2015). The order provided for pilot programs at selected universities and established a Self-Driving Vehicle Oversight Committee within the governor’s office. In December 2016, Michigan expanded its autonomous vehicle legislation, to allow the operation of autonomous vehicles on state roads by the public, to permit automated vehicle platoons, and to authorize on-demand autonomous vehicle networks. In addition, the legislation outlined specific parameters for entities that wish to offer on-demand autonomous vehicle networks to the public (Elfin 2016). Some key features of these bills are highlighted in the table below. While it is not a comprehensive list of every provision in every law, it provides an overview of the wide variety of provisions addressed in state legislation. Note that for any one of these provisions, the legislative language may vary significantly from state to state. California, Nevada, Florida, and Michigan have passed the most comprehensive bills with respect to testing and operation of autonomous vehicles. While a separate legal framework for trucks is not typically spelled out, Florida law has mandated a study of truck platooning safety and provides for a pilot platooning program on completion of the study. California law requires the study what is an appropriate (and safe) following distance for trucks in a platoon and how many vehicles could be safely allowed. A number of smaller targeted state legislative bills have also addressed research on CV and AV issues more generally. North Dakota and Utah have adopted bills that mandate the study of autonomous vehicles. Table 3-2. State Autonomous Vehicle Legislation Elements. Bill provisions CA DC FL MI ND NV TN UT Defines autonomous vehicle      Authorizes autonomous vehicles to be tested on public roadways     Authorizes use of autonomous vehicles by the public    Allows for the testing of “driverless vehicles” under certain conditions   Requires an instrument of insurance, surety bond or self- insurance prior to the testing of a vehicle     Requires state agency to recommend or adopt safety standards and performance requirements to ensure the safe operation or testing of autonomous vehicles    Requires a special license for the driver of a test vehicle 

CRP Project HR 20-102(03) 36 Bill provisions CA DC FL MI ND NV TN UT Requires completion of a manufacturer AV course    Mandated study of AVs     Allows operation of vehicles without drivers  Requires means to engage and disengage the autonomous technology which is easily accessible to the human operator of the autonomous vehicle   Authorizes truck platooning   Addresses liability of the OEM of a vehicle on which a third party has installed an automated system      Permits the use of cell phones and other wireless vehicles in an autonomous vehicle  Prohibits local governments from banning autonomous vehicles  Allows a motor vehicle to be operated, or to be equipped with, an integrated electronic display visible to the operator while the motor vehicle's autonomous technology is engaged  Establishes certification program for manufacturers of autonomous vehicles before such vehicles may be tested, operated, or sold  Requires a human driver conducting testing to be "prepared to take control of the autonomous vehicle at any moment”   Restricts conversion to recent vehicles  The autonomous vehicle has a visual indicator inside the cabin to indicate when the autonomous technology is engaged  The autonomous vehicle has a separate mechanism, in addition to, and separate from, any other mechanism required by law, to capture and store the autonomous technology sensor data for at least 30 seconds before a collision 

CRP Project HR 20-102(03) 37 Bill provisions CA DC FL MI ND NV TN UT …while the vehicle is operating in autonomous mode The manufacturer of the autonomous technology installed on a vehicle shall provide a written disclosure to the purchaser of an autonomous vehicle that describes what information is collected by the autonomous technology equipped on the vehicle  The comprehensiveness of state legislation varies significantly. Although there are many commonalities between states, each state has chosen to address a somewhat different set of issues in each of their respective AV laws. While some states have chosen to explicitly authorize the testing or use of autonomous vehicles, this does not mean it is necessarily prohibited in other states. Manufacturers and firms involved in testing autonomous vehicles initially pushed to enact autonomous vehicle laws to provide some legal certainty for their activities. Nonetheless, most of these laws have provided additional restrictions and regulation of the testing and use of AVs. There are numerous differences between state laws that could serve as a barrier to widespread testing and adoption. The definition of an autonomous vehicle differs between states, as does many other legislative provisions such as the insurance required for vehicle testing, whether autonomous vehicles can be used by the public or whether specific vehicle features are required. California requires a number of specific features on the vehicle, including internal indicators of when the vehicle is operating in autonomous mode and a “black box” to collect data when the vehicle is operating in autonomous mode. Tennessee state law requires a similar indicator of autonomous operation for the driver and sets up a certification program for manufacturers of autonomous vehicles. Florida law allowed unoccupied operation of AVs starting July 1, 2016. The California Department of Motor Vehicles (DMV) published a draft rule requiring licensed drivers behind the wheel of AVs. States vary in how they treat following distance for trucks. Many state laws require that this distance be reasonable and prudent. Other laws require that trucks follow at specific distances, or at a certain duration of time behind another vehicle. A less common standard used in state law is the “Sufficient space to enter and occupy without danger” standard. In 2014 the ATA documented how state laws differ with respect to following too closely (Scott 2014). In July 2016 the Competitive Enterprise Institute (CEI) released the report, “Authorizing Automated Vehicle Platooning: A Guide for State Legislators” (Scribner 2016) that provides suggested language for each state to adjust their laws to allow for truck platooning. The report provides both strong language that would allow a legislature to provide for truck platooning, and weak language that would provide for a regulatory approach to platooning. Another issue with truck platooning is that some platooning concepts provide a live video feed to the following truck of what the lead truck driver sees. State laws prohibit video displays that are visible to the driver though, so changes would be needed in state laws to make this legal. Tennessee has acted to make the use of a video display legal for the driver when the vehicle is in autonomous mode. Considerations for Improving Laws Related to Autonomous and Connected Vehicles Bryant Walker Smith’s review of state laws and their application to autonomous vehicles includes draft language for U.S. states that wish to clarify the laws that apply to autonomous vehicles (Smith 2014). In March 2016, NHTSA released an update to “Preliminary Statement of Policy Concerning Automated

CRP Project HR 20-102(03) 38 Vehicles.” NHSTA noted that it is “working with states to craft and propose model policy guidance that helps policymakers address issues in both the testing and the wider operational deployment of vehicles at advanced stages of automation and offers a nationally consistent approach to autonomous vehicles” (NHTSA 2016a). In September 2016, NHTSA released its Federal Automated Vehicles Policy. This policy provides model state policy guidance and is applicable to the heavy-duty vehicle segment (NHTSA 2016b). If not initially, the U.S. DOT has solicited and encouraged trucking industry input as this policy is finalized. A number of recommendations to increase legal certainty without prematurely regulating a technology that is rapidly evolving have been identified in the literature (Smith 2014). We review some of these suggestions, providing updated examples of ongoing initiatives and identify areas for further progress. Develop Common Legal and Regulatory Definitions Standardizing the legal terminology used for autonomous vehicles will likely be important to clarifying and harmonizing the law between states. Currently there are two similar, but different classification schemes used to define vehicle autonomy. The SAE system has, as noted in Chapter 1, five levels of autonomy plus one “no-automation” level. NHTSA has a separate classification system with four levels of autonomy plus the “no-automation” level, though it has indicated willingness to standardize on the SAE system. Currently state laws do not use either of these classification systems, but rather each state defines what an autonomous vehicle is in their own way. The University of Washington’s Technology Law and Policy Clinic (UWTLPC) prepared a report for the Uniform Law Commission (ULC). The report examines the legal terms and definitions in existing state autonomous vehicle laws (UWTLPC 2014a). They make numerous recommendations on how existing state laws can improve their legal definitions, including being clear about whether and how SAE level 2 vehicles are included in the definition of autonomous. In addition, they note it is important to clarify how the duration of time vehicles can operate autonomously will affect their autonomous classification. Since SAE level 3 autonomous vehicles can only be operated under limited conditions, there may also be a need to standardize the geographic and environmental condition types. For instance, Nevada law defines the following geographic road types for vehicle testing:  Interstate highways  State highways  Urban environments  Complex urban environments  Residential roads  Unpaved or unmarked roads Nevada uses these geographic categories to limit licenses for on-road testing to certain areas. Nevada law also uses the following categories to define the environmental conditions that may be included in vehicle test permits to limit their operation:  Night driving  Rain  Fog  Snow/Ice  High crosswinds (gusts above 30 mph) If the general use of autonomous trucks by businesses is restricted to certain geographic areas or environmental conditions in the future, it will be important to clarify and harmonize these categories to facilitate interstate commerce. Some designation of roadways with respect to the quality of lane markings and signage could also be important for future efforts to license autonomous vehicles for broader public use. With SAE levels 4 and

CRP Project HR 20-102(03) 39 5 autonomous vehicles, it will likely be important to consider whether there are detailed digital maps available for roadways. For trucks specifically, it will also be important to identify roadway geometric limitations and turning radius restrictions for trucks in these maps. Will information be readily available to autonomous trucks on bridge weight and height restrictions or designated truck routes, for example? Lastly some autonomous truck technologies are limited to specific speeds. For instance, technology providers are working to develop autonomous queuing systems that would keep trucks in line at ports or other facilities, but could only be used at lower speeds. The consideration of these and other factors and how they may be defined legally and incorporated into the laws that regulate autonomous trucks will be important in the future. Autonomous Vehicle Legislation and the International Context United States legislators and regulators need to monitor efforts to amend or interpret the 1969 Vienna Convention, which contains language similar to the Geneva Convention, but does not bind the United States. In March 2016, the Vienna Convention was updated to specifically allow “automated driving technologies transferring driving tasks to the vehicle” as long as these technologies conform to other United Nations vehicle regulations and can be “overridden or switched off by the driver” (United Nations Economic Commission for Europe (UNECE) 2016). Currently this convention does not define any truck specific requirements. Some state laws have adopted the requirement that the driver must be able to switch off the technology. Scope and Schedule of Regulatory Action When key federal and state agencies indicate the likely scope and schedule of potential regulatory action, legislators and policymakers will be able to coordinate their activities. As noted, NHTSA released a Preliminary Statement of Policy Concerning Automated Vehicles and provided an update to this in 2016, as well as releasing model state policy guidance that should help states harmonize state laws and regulations. The U.S. DOT has been working with Google, BMW, General Motors and other companies developing driverless and partially autonomous cars to adapt existing safety rules to the new technologies. They also have created a federal advisory committee to plan how to “approach autonomous vehicles and artificial intelligence more generally” (Mitchell 2016). FMCSA is still in the preliminary research stages on its regulatory actions. The National Conference of State Legislatures (NCSL) tracks autonomous vehicle legislation, some of which will require DMVs and other state agencies to issue regulations (NCSL 2016). There will be a need to coordinate how state laws may affect truck platooning. Specifically, different states currently specify different following distances. Some have speculated that truck platooning should be restricted to rural highway segments where there are few exits, interchanges or merging traffic that could conflict with truck platoons. Additional coordination and communication between state and federal policymakers on this issue could be helpful. Vehicle Automation and Vehicle Codes U.S. states need to analyze how their vehicle codes would apply to automated vehicles, including those that have an identifiable human operator and those that do not. A number of initiatives have been completed or are ongoing in this area. Iowa DOT funded a report “Review of Automated Vehicle Technology: Policy and Implementation Implications” (McGehee et al. 2016) that reviews some state laws and policies. Project NCHRP 20-102(07), Implications of Automation for Motor Vehicle Codes, is in its early stages and will help answer some of these questions. As noted above, the UWTLPC prepared a report that identified best practices for developing an autonomous vehicle law. Utah is currently examining NHTSA and American Association of Motor Vehicle Administrators (AAMVA) standards and

CRP Project HR 20-102(03) 40 best practices and evaluating appropriate safety features and regulatory strategies and developing recommendations for their state. Although many states incorporate federal rules by reference for trucking operations (e.g., FMCSA, FMVSS), so that changes will be automatic when federal agencies revise rules, others adopt them in full into state statutes. Some exceptions that were “grandfathered in” (e.g., California’s rules) may need to be reviewed. FMCSA regulates the operation of interstate motor carriers, including the insurance. Legislators will need to assess whether legislative action in their state is needed to incorporate any FMCSA rule changes related to autonomous vehicles. Another complexity is that state statutes are also organized in different ways, placing authorities and laws that are relevant to AVs in varying agencies and varying chapters. State regulatory codes track this diversity. States do not necessarily use the same terms or define common terms in the same way even within a single state’s statutes. Although states have liability responsibility provisions that apply to motor vehicles, states’ general civil liability regimes apply to any tort, warranty or contractual action for damages. These civil liability regimes will be applied to AVs. All states have statutes prohibiting various forms of computer fraud and abuse, and computers will either assist in operating or operate AVs. Application of these statutes to AVs is nearly certain to vary among the states. It is not clear whether state and local roadway infrastructure will need to be adapted to AVs. At present, state laws confine the use of low-speed-vehicle forms of AVs to specific types of roadways. In the future there could be roadways set aside for advanced AVs, which can travel faster with less headway than other vehicles. NCHRP Legal Research Digest 69 provides a broad review of many legal issues associated with driverless vehicles (Glancy et al. 2016). One important policy issue for vehicle platooning is existing legal restrictions on signage and lights on platooned vehicles. Currently the only flashing lights that can be placed on vehicles are for emergency vehicles and they require standardized indicators. There are exceptions to this, which include tow trucks and school buses. Platooning vehicles need an indicator light of some kind, and this issue is regulated at the state level. For wide scale adoption it would be desirable to have standardization of requirements for indicators, perhaps at least on the National Highway Freight Network. Harmonizing Laws Between States The Iowa DOT report mentioned above also compiled recommendations on regulatory and legal issues that could be targets to increase harmony between the states. One area of great uncertainty relates to the impact of closely spaced truck platoons on roadway pavements and bridges. There is a concern that closely spaced vehicle platoons could cause significantly greater wear and tear on roads and bridges. Additional research is needed in this area to more fully understand the implications. States differ with respect to the spacing that is currently legal between trucks in platoons. Some states also have grandfathered exemptions to state size and weight rules that allow unique oversize and overweight truck configurations as well. States also issue numerous special permits to allow oversize and overweight vehicles to operate for non-divisible loads. There has been little legal consideration of the implications of truck platoons on state size and weight laws, but this is an area that may be difficult to harmonize due to current differences between states on this issue. There may also be a need to address the legality of autonomous vehicles or truck platoons moving oversize and overweight loads specifically. Evaluate the Likelihood of Federal Preemption. In general states that seek to revise or harmonize their laws for AVs and CVs will need to consider which components of law may be preempted by the federal government. A recent report, “The Risks of Federal Preemption of State Autonomous Vehicle Regulations” (UWTLPC 2014b) examines what is likely to happen in this regard. They argue that NHTSA is unlikely to preempt state laws and regulations in the areas of testing, permitting, licensing, test-driver training, and conditions for the operation of specific types of AVs. NHTSA has expressed the intention to create broad safety regulations and standards for new AVs and automating equipment, which

CRP Project HR 20-102(03) 41 will likely preempt states in this area. This includes working with states on inspection regimes to monitor aftermarket modifications that turn traditional vehicles into AVs. This could become an important issue with companies promising to soon sell equipment and software that will allow existing heavy-duty trucks to be retrofitted with autonomous technology (note the discussion above in the Highly Automated Trucks Section regarding the startup Ottomotto). NHTSA only has the power to preempt state tort law where it conflicts with a significant regulatory objective. Recent federal policy has not preempted state tort laws even where such conflicts exist. FMCSA does have the authority to require interstate carriers to maintain specific levels of insurance to guarantee financial responsibility. Regulatory Topics Federal regulations will play an important role in the deployment of autonomous vehicles. Many in the industry have argued that full deployment must start with the regulatory process (Trop 2016b). NHTSA has a mandate to ensure that vehicles manufactured for use on public roadways are safe. FMCSA has authority over the operation of heavy-duty commercial trucks operating in interstate commerce. We discuss these below in the context of automated truck technologies (noting that much of it is related to passenger vehicles as well). Federal Motor Vehicle Safety Standards NHTSA promulgates regulations and enforces the FMVSS. NHTSA has recognized the great promise of automated and connected vehicle technology. As the quote from the administrator in the text box to the right shows, highly automated vehicles provide a substantial, albeit largely unproven opportunity to improve safety. NHTSA will need to engage in the practical work of adjusting existing regulations and promulgating new rules to facilitate the widespread adoption of the family of technologies. Current Regulatory Barriers Manufacturers of new motor vehicles and motor vehicle equipment are required to self-certify that their vehicles and equipment are in compliance with the minimum safety performance requirements outlined in the FMVSS. The most recently released Federal Automated Vehicles Policy also suggests that manufacturers submit a voluntary safety assessment (NHTSA 2016b). Many of the regulations in the FMVSS assume a human driver and require a specific design of the vehicle with respect to that driver. For instance, they require that the brake pedals and steering wheel be operable by a driver in the front seat. They require that the gauges and side mirrors are visible by the driver. After a comprehensive review of the FMVSS, the Volpe National Transportation Systems Center concluded that few of the existing FMVSS standards present barriers to the introduction of vehicles that “have automated capabilities but retain the overall design, seating arrangement, and human-machine interfaces of a conventional” vehicle. They note that as manufacturers build increasing levels of automation into their NHTSA Recognizes the Promise of Automated Technologies “Whereas new drivers must learn on the road and make the same mistakes as thousands before them, automated vehicles will be able to benefit from the data and learning of all others on the road….new highly automated vehicles provide an enormous opportunity for learning that has rarely existed before…When something goes wrong, or a highly automated vehicle encounters an edge case—something it hasn’t been programmed to deal with—that data can be taken, analyzed, and then the lessons can be shared with more than the rest of that vehicle fleet. It could be shared with all automated vehicles.” Mark Rosekind NHTSA Administrator, at the Automated Vehicles Symposium July 20, 2016

CRP Project HR 20-102(03) 42 Software Validation in Automated Vehicles is Critical “It is important to emphasize that refinement and validation of the software will take much longer than putting in place the cameras, radar, lidar and computing hardware … Even once the software is highly refined and far better than the average human driver, there will still be a significant time gap, varying widely by jurisdiction, before true self-driving is approved by regulators.” Elon Musk vehicles, and use novel designs, such as omitting steering wheels and control interfaces the existing regulatory framework would serve as a barrier to such vehicle designs (Kim et al. 2016). NHTSA has committed itself to encouraging the deployment of autonomous vehicle technologies that can be documented to improve safety. NHTSA has also encouraged manufacturers to seek the use of NHTSA’s existing exemption authority to conduct field tests that can demonstrate the safety benefits of fully autonomous vehicles (NHTSA 2016a). NHTSA has further indicated that it may need additional authorities to address autonomous vehicles. The Federal Automated Vehicles Policy defines the federal role in regulating autonomous vehicles versus the state role, and suggests laws and regulations states might want to adopt. There is a concern that without NHTSA guidance, a patchwork of state and local laws could emerge that could hinder adoption of the technology. The policy does not change the overall structure of the FMVSS. Industry groups, such as the Alliance for Transportation Innovation, have noted that the current structure of the FMVSS may limit rapid innovation in the long run. Autonomous vehicles must still demonstrate compliance with FMVSS. As noted above, there are numerous references to human drivers and these standards mandate traditional design features for human drivers. “Any fully self-driving vehicle would violate more than one-third of those standards, not to mention half of the series 100 Crash Avoidance standards built into the compliance structure” (Brubaker 2016). While there are temporary exemptions available, it can take a significant amount of time to obtain these exemptions, and this may present a barrier to technology companies whose business model may depend on them rapidly deploying new technology. There is also concern that the voluntary safety assessments described in the Federal Automated Vehicles Policy indicate that NHTSA will expand its role and become more involved in the development of technologies on the front end rather its traditional role of setting prescriptive standards. There is concern that a greater role for NHTSA during technology development could slow down the development process. Testing and Validation Needs In a recent letter to Google, NHTSA indicated that they could consider the self-driving system (SDS) in the Google car as the “driver” for purposes of regulation (Hemmersbaugh 2016). Therefore, vehicle controls could simply be operable by the SDS and vehicle indicators could be available to the SDS. The same logic would presumably be applicable to heavy-duty vehicles, although these were not specifically mentioned in this communication. NHTSA noted that they still could not certify a fully autonomous vehicle with a novel design because they had “no defined way at this time of verifying Google’s compliance with this interpretation.… NHTSA would be unable to conduct confirmatory testing to satisfy ourselves that the Google vehicle is compliant. Therefore, unless and until NHTSA has a standard and testing procedures to confirm compliance… it cannot conclude that Google’s SDS is compliant with those requirements. In order to determine what ‘operable by’ and ‘visible’ to the SDS mean, and to establish procedures for testing compliance with those requirements using its existing regulatory tools, NHTSA would be required to conduct a rulemaking” (Hemmersbaugh 2016). Industry leaders expect that significant testing will be required to validate any highly autonomous vehicle. Tesla recently announced that they plan to build autonomous heavy-duty trucks. CEO Elon Musk has estimated that it will require data from six billion miles of vehicle operation to test self-driving

CRP Project HR 20-102(03) 43 vehicle software and obtain broad worldwide approval regulatory approval. As indicated by the quote in the text box from Musk, validating and testing software is expected to be much more complex and difficult than testing the underlying hardware (Lockridge 2016). Tesla is currently accumulating approximately three million miles per day of experience with their autonomous passenger vehicles. At this rate, it would take over five years to accumulate enough miles for Musk’s estimate of full regulatory approval. While Tesla is expected to expand sales of autonomous passenger vehicles in the near future, it is likely that heavy-duty vehicles will accumulate experience with autonomous technology at a much slower rate. Unless autonomous testing experience is considered transferable between the heavy and light-duty sectors, it is likely that full regulatory approval will occur at a slower pace. Mandates for SAE Level 1 Autonomous Technology NHTSA’s focus on safety has led it to mandate or suggest mandates for certain SAE level 1 automation technologies for trucks. NHTSA announced that all new Class 7 and Class 8 heavy-duty trucks will be required to be equipped with stability control systems in August 2017 (Jones 2015). An agreement between NHTSA, the insurance industry and automakers will also put automatic braking systems into most light vehicles by 2022. Developing automated emergency braking systems for large commercial trucks presents challenges that don’t exist in passenger vehicles. Because of the weight and size of heavy-duty trucks, it is critical that automated emergency braking systems maintain the stability of a truck to ensure that it doesn’t jackknife or tip over. Implementation of automatic emergency braking for trucks must ensure that rapid braking will not cause the driver to lose control of the vehicle. Heavy-duty trucks haul a wide range of trailers and transport cargo loads with a wide range of sizes and weights. Creating software that can safely and reliably apply the brakes in emergency situations is thus much more complex in heavy trucks than it is in passenger vehicles. Another complexity is that the diverse braking technology in large trucks makes universal implementation more challenging for vehicles that weigh between 10,000 and 26,000 pounds, where many different technologies are used (Stoltzfus 2016). Product Defect Investigation Another function of NHTSA is to investigate product defects that cause crashes. NHTSA's Office of Defects handles product defect investigations. With heavy-duty trucks, this can be a more complex process. Heavy-duty vehicles are more often customized or retrofitted by third parties. They are also often operated as combination vehicles that can use multiple configurations and carry loads of various weights and dimensions. For example, in connected vehicle pilot programs, implementing connected vehicle crash avoidance technologies is more challenging in combination vehicles since this requires that the location of the vehicle be identified very exactly in space. Combination vehicles have more varied vehicle dimensions than passenger cars, making it more difficult for these systems to be implemented. Determining the party responsible for the equipment failure could also be complex. Autonomous truck technology could further complicate this situation. For instance, the agency could be called upon to assess the safety performance of multiple vehicles operating in a truck platoon. Assessing whether the leading or following vehicle was the cause of a crash could be difficult. One could imagine a case where the unexpected performance of both vehicles operating in combination could cause a crash. Future NHTSA rules for autonomous trucks are likely to include a requirement that vehicles record whether autonomous features are enabled in the event of an accident. While federal regulations do not currently require trucks to have systems to record vehicle data during crashes (event data recorders), almost all truck engines today have an electronic control module that has the capability to record this information. The electronic control module (ECM) is standard equipment on all diesel fuel injection systems and allows companies to monitor and analyze a wide range of data including trip times, speeds, total idle time and the existence and number of hard stops. Collecting event data from the ECM is typically implemented with software. Collecting information from “event data recorders” on the accidents

CRP Project HR 20-102(03) 44 that do occur will likely prove critical to assist product defect investigations and helping manufacturers improve their technology. These data can also be useful for helping manufacturers assess their liability in accidents, and defend themselves legally when the accident was not caused by a product defect. Federal Motor Carrier Safety Regulations The FMCSA regulates interstate trucking, promulgating the FMCSRs and providing enforcement to ensure that the operation of interstate carriers is consistent with these regulations. The regulatory and enforcement environment for commercial trucking is somewhat complex. States that accept MCSAP funds are required to adopt the FMCSRs for all commercial motor vehicles (CMVs) that require a commercial driver’s license (CDL) and apply those rules to intrastate operations. There are a few grandfathered exceptions. Intrastate safety regulations can be more stringent than federal rules. In addition, states are free to exempt intrastate CMVs that do not require a CDL to drive (i.e., less than 26,001 pounds) and almost every state has adopted such exemptions. The lead agencies enforcing motor carrier safety regulations and the specific methods used may also differ between states. States receive federal MCSAP funds, but also spend their own money on these programs, and have some autonomy in operating them, and crafting regulations and enforcement strategies that are consistent with federal standards. Hours of Service Regulations The FMCSA regulates the hours of service and schedule of breaks/rest for truck drivers. This includes daily and weekly limits on driving and on-duty time, and mandates periods of rest for truck drivers. The Hours of Service Rules were put into place to reduce crashes caused by driver fatigue. Because these rules limit the number of hours that can be driven, they are an important determinant of industry productivity. In addition, they affect the price and level of service trucking carriers can provide their customers, by limiting how fast a truck with a single driver can move freight over long hauls. Some industry and technology insiders have been optimistic that the adoption of new CV and AV technologies will allow changes to the Hours of Service Rules. The CEO of Daimler Truck North America (DTNA) noted that “if DTNA can bring scientific evidence to regulators that drivers operating under autonomous mode experience decreased fatigue (and it's already found that's the case in tests on tracks), it could possibly get regulations changed for more flexible and even longer hours of service, which in turn could improve productivity" (Lockridge 2015c). Ottomotto (which as noted above is building an autonomous retrofit kit for existing trucks) claims that trucks fitted with their software can drive more than double their normal daily mileage and will provide “a very strong return on investment" (Sage 2016). They are already conducting on-road tests of the technology and have argued that the technology for an SAE level 3 highway autopilot system is much easier to implement for trucking than in other vehicle markets (Ohnsman 2016). Estimates of productivity increases assume that Hours of Service Regulations would be changed to allow drivers to operate the vehicle for longer. CV and AV technologies may make possible a number of different types of changes to the Hours of Service Regulations. These possibilities could include at least the following:  If it can be shown that vehicle automation reduces fatigue, then drivers of these vehicles could be allowed to be on-duty or drive for longer.  If the technology was good enough, the driver might even be able to sleep while the vehicle was operating in autonomous mode. There is also the possibility that trucks could be driven as drones. The driver could sleep in the back while it was being operated remotely.  The Hours of Service Rules could account for the time that the driver spends in the cab when he is using the autopilot mode differently than driving – perhaps allowing extended operation in this mode, or perhaps providing flexibility for the driver to get home or make it to the next trucking parking or rest area.

CRP Project HR 20-102(03) 45  For trucks platooning with SAE level 2 or higher technologies, an alternative duty status for the driver in the following vehicle(s) could be created.  SAE level 3 or higher platooning technologies could enable the driver in the following vehicle(s) to rest. This would improve the ability to operate trucks continuously with a smaller team of drivers.  Autonomous technology could be used in stop-and-go traffic or waiting in a queues to reduce on-duty time (ATA TMC 2015). Changing the Hours of Service Rules will likely be a highly contentious issue that will engage the attention of safety advocates, unions, enforcement officials, industry and the broader public. Previous changes in the rule have been tied up by litigation in the courts for years and been subject to changes in policy between administrations. A very clear documentation of the safety benefits of autonomous technology will likely be needed to make any changes to existing rules. A significant volume of data collected from the real world operations of trucking companies would be the most credible source. It is unclear how such data would be collected under existing rules, but one solution might be to set up a legal or regulatory framework that allows states to establish pilot programs to test new Hours of Service Rules with autonomous vehicle technologies. A mechanism allowing states and individual businesses to voluntarily opt in if they agree to collect the requisite data and make it available to policymakers could provide an incentive to participate. Financial Responsibility FMCSA requires truck carriers in interstate commerce to maintain minimum levels of financial responsibility – essentially requiring specific levels of insurance or surety bonds for each truck. FMCSA regulations require a minimum of $750,000 for carriers of property, $1 million for oil and $5 million for hazardous materials (FMCSA 1985). Assuming that autonomous vehicles were significantly safer or less likely to get involved in serious accidents, these levels could potentially be reduced. The existing levels would not serve as barriers to the adoption of the technology, although reducing these levels for autonomous vehicles could help to incentivize companies to invest in autonomous trucks. Enforcement There are more than 3.5 million roadside inspections conducted annually to enforce FMCSA regulations on the operation of commercial vehicles. These inspections are typically unannounced and vehicles that have serious violations are immediately placed out-of-service and must fix any deficiencies found before driving further on public roadways. The vast majority of roadside safety inspections are conducted by the states, which receive annual grants from FMCSA through the MCSAP. States have some latitude in how they organize to enforce federal motor carrier regulations, including which lead agency will oversee these activities and some discretion in the how these inspections are undertaken. Federal regulations identify a number of different levels of roadside inspection. The most common types include the following:  Level I North American Standard Inspection – This is the most comprehensive of the inspections. It includes Roadside Inspection of Automated Vehicles May Need Adaptations “Inspectors currently examine brakes with a flashlight and pocket ruler. Stability control is mandated now, but how do you inspect it? There is no physical device added to the vehicle. It is a computer. Are we going to eventually have to start plugging into the vehicle and reading fault codes versus physically walking around the vehicle? The inspection of the vehicle may be an over the air data dump in the future. There are other automated technologies like auto breaking and adaptive cruise control. Are we going to start inspecting radar and lidar cameras to decide if the automated vehicle is in working order?” FMCSA Official

CRP Project HR 20-102(03) 46 examination of compliance with the critical elements of both driver and vehicle regulations.  Level II Walk-around Driver-Vehicle Inspection – This is similar to a Level I inspection, but the inspector does not check items that require the inspector to physically get under the vehicle.  Level III Driver/Credential Inspections – This is an examination of documents relating to the driver and hazardous materials (if applicable). These include a commercial driver's license, medical certificate, logbook and hours of service, and documentation of the annual vehicle inspection. The presence of hazardous materials is also checked.  Level V Vehicle-Only Inspections – This includes all the vehicle components in the Level I inspection and may take place without a driver present. It is often conducted at a carrier's place of business during a compliance review. Current methods of vehicle and driver inspection rely on a physical inspection of the vehicle and an in- person inspection of driver records. Increasingly advances in autonomous truck technologies may require changes in how inspections are conducted. Complex computer hardware and software systems to control vehicle functions will require new inspection methods and technologies. Connected vehicle technologies also provide a large opportunity to streamline inspection procedures and increase enforcement productivity and this could have a large and positive impact on regulatory compliance and safety. Wireless Roadside Inspection Automated and wireless roadside inspections of commercial vehicles could provide significant savings for industry and allow for more comprehensive inspections of heavy-duty vehicles. With the exception of load-securement, assessing drug and alcohol use and noting whether a driver is wearing the seatbelt, most of the key vehicle and operator condition criteria lend themselves to on-board electronic monitoring and diagnostic assessment (Hultin et al. 2008). There are two major pilot programs that are demonstrating the viability of wireless roadside safety inspections. Wireless Roadside Inspection Pilot. FMCSA is funding the Wireless Roadside Inspection pilot (sometimes called the Trusted Truck® pilot). This program is currently piloting wireless inspection technology to test the feasibility of “electronically assessing truck and motor coach driver and vehicle safety at least 25 times more often than is possible using only roadside physical inspections.” FMCSA’s Trusted Truck® research pilot has accomplished the following:  Phase I Transmitted real-time truck brake condition data to the roadside inspection officer.  Phase II research added functionality including initial implementation of the Trusted Truck® Management Center (TTMC), upgrading roadside communications, and increasing the number of safety-related items included in the wireless roadside inspection to include tractor and trailer weight, trailer tire pressure and temperature, trailer ID, and shipment data. They are also demonstrating a “trusted” vehicle bypassing a roadside inspection using the TTMC as the method of delivering the inspection results.  Phase III of FMCSA’s Wireless Roadside Inspection research is currently testing trucks equipped with electronic logs and telematics devices that transmit operator hours of service and credentials to the roadside for inspection without the necessity of stopping. There are currently at least 20 Phase III test sites located along roadways in Kentucky, Tennessee, Mississippi, North Carolina and Georgia. New vehicle safety technologies provide additional capabilities that can be used to automate inspections. For example, on new vehicles the electronic stability control system knows how much the vehicle weighs. Information on the trailer weight is transmitted to the tractor and thus the CAN Bus in the tractor knows the vehicle weight. This information can be queried electronically by enforcement to automate truck weight enforcement.

CRP Project HR 20-102(03) 47 Drivewyze e-Inspection. Drivewyze is currently testing a technology to transmit most of the information in a Level 3 roadside inspection in a four-state field test in Virginia, Maryland, Delaware and Pennsylvania. This is being done in cooperation with enforcement departments in those states. Trucks are equipped with electronic logging devices capable of transmitting information to the roadside and the Drivewyze PreClear bypass system. The technology can pre-populate almost all of the data elements on the Level 3 inspection form. The inspection officer then can walk up to the truck with an inspection form that is nearly completely filled out, and assess drug and alcohol use and whether the driver is wearing a seat belt. The inspector then presses a button to complete the inspection, greatly increasing the efficiency of the process. Drivewyze includes the capacity to automatically screen for compliance. Under current rules it is not possible to completely file a Level 3 roadside inspection without officer interaction with the driver. The Commercial Vehicle Safety Alliance (CVSA) guidelines for such inspections require that officers walk up to the vehicle, greet the driver, and assess drug and alcohol use and whether a driver is wearing a seatbelt. Given advances in technology, there may be a need for a different CVSA inspection procedure for electronic inspections. Currently CVSA has a committee considering what should be included in the e- inspection and how such inspections might be incorporated into FMCSA’s Compliance, Safety, Accountability safety-scoring program (Dills 2016a). Some in the industry have argued that wireless/electronic inspection programs should evolve into voluntary-participation programs. These Voluntary programs could extend compliance-scoring credit to carriers in exchange for real-time roadside access to carrier, driver and vehicle information (Dills 2016b). These programs could run in parallel to official inspection programs that are similar to what currently exists. The technology of wireless inspection raises multiple questions. Many are concerned that their ease of use will lead to their wide scale application and a loss of privacy. If connected and autonomous vehicles that are electronically inspected can bypass inspection stations or easily build a positive inspection record, it could create a regulatory enforcement equity issue with those operating older trucks, which tend to include more owner-operators and small trucking companies. In general, as inspection technology becomes more advanced, FMCSA will need to figure out how to blend data from high tech and low tech inspections to assess safety performance. Another question relates to the role of the public sector. FMCSA is currently developing and pilot testing technologies that could compete with private sector solutions. In the FAST Act, Congress is requiring “an FMCSA due diligence on the program’s potential conflicts with existing, privately available screening technologies.” The FAST Act requires the U.S. DOT to submit a report that certifies that the program doesn’t do any of the following three things:  Conflict with existing non-federal electronic screening systems;  Create capabilities already available; and  Require additional statutory authority to incorporate generated data into safety determinations. In addition, the FAST Act also requires that the U.S. DOT Secretary certify that privacy concerns of truckers and other participants in any wireless roadside program are addressed. This will likely require some type of comprehensive outreach to the industry and its workers. One barrier to use of automated inspections is government rules concerning the use of information technology. For instance with EOBRs, some enforcement agencies can’t plug thumb drives into their systems. Some local police agencies don’t allow blue tooth linking. There is little standardization of rules for information technology and security. For instance, police and fire departments in the same town can have different protocols.

CRP Project HR 20-102(03) 48 Vehicle Retrofits FMCSA has authority over some vehicle retrofits, including the installation of cameras to replace side mirrors. Cameras can have better visibility can be better than mirrors. Image processing can reduce glare and provide enhanced night vision. Cameras can also eliminate blind spots. Two groups have asked for waivers to install this equipment, but FMCSA has denied both pilot programs. They provide a driver assist technology with built-in algorithms to process the image. The cameras can provide 1.5% improvement in fuel economy. The E.U. is already allowing mirrors to be removed for the 2018 model year. Liability Issues With conventional trucks, liability for crashes is typically attributed to the operator of the vehicle, and the company they work for, if the crash is caused by human error. Most crashes are caused by human error and the vast majority can be attributed to decisions made by the driver in one of the vehicles in the crash. Liability for crashes that involve mechanical defects in the design or manufacture of the vehicle fall on the manufacturer, or the upfitter in the case the defect can be attributed to third party modification. SAE levels 4 and 5 autonomous trucks have the potential to fundamentally alter how liability is allocated in cases of crashes. If the vehicle is operating in a fully autonomous mode, the liability for truck crashes would likely fall entirely on the manufacturer of the vehicle. If autonomous vehicles are significantly safer than conventional vehicles, then the costs of truck crashes should decrease. It is quite possible that crash costs could decrease, but the overall liability for companies could increase since juries often award large punitive damage costs to plaintiffs who sue large companies that manufacture vehicles. This could alter the benefit cost calculation for bringing law suits, and encourage broader scale litigation and higher awards, even in the face of increased safety. Some vehicle manufacturers have argued that they will not produce SAE level 4 autonomous vehicles because of the liability which they would incur. Some have argued that manufacturers need some kind of law pre-empting state liability, others have made the case that the existing system of torts can address this issue. A number of states have passed laws to limit the liability of the original manufacturer of a vehicle on which a third party has installed an automated system. For instance Michigan recently enacted such a law. There are also some unresolved issues with the liability of companies in truck crashes where some of the vehicles are operating in a truck platoon. Are there circumstances when the liability in the crash of a following vehicle should be attributed to the leading vehicle? Since the technology is so new, many of these issues have not been worked out. The U.S. DOT has generally encouraged technological innovation. In a recent report, the RAND Corporation argued that “overall, the guiding principle for policymakers should be that AV technology ought to be permitted if and when it is superior to average human drivers. For example, safety regulations and liability rules should be designed with this overarching guiding principle in mind. Similarly, this principle can provide some guidance to judges struggling with whether a particular design decision was reasonable in the context of a products liability lawsuit” (Anderson et al. 2016). While this is a very rational approach to this issue, there are no guarantees that individual courts will use this framework to assess responsibility and damages. Some have suggested that NHTSA should investigate all autonomous vehicle accidents in the near term in the same manner that the NTSB investigates all airplane accidents. This might provide some assurance to manufacturers that there would be a rigorous process to assign responsibility for these accidents in the early years, mitigating some of the legal risks that they might face.

CRP Project HR 20-102(03) 49 Open Standards Make Heavy-Duty Trucks More Vulnerable to Hacks “If you wanted to hijack someone’s car, you’d have to know the make and model and tailor the (hacking) attack…with trucks, it’s all open, so you can just craft one attack.” Leif Millar, University of Michigan researcher who recently demonstrated the hack of a heavy-duty truck Cybersecurity and Privacy Cybersecurity is a significant concern for policymakers. Highly automated features could make trucks more vulnerable to being hijacked. Hackers who gain access to trucks through connected vehicle features could obtain access to data on the vehicle, firm or driver, compromising confidential and private information. While previously criminals or terrorists have required physical access to a vehicle, new digital features on heavy-duty vehicles provide avenues for of attack according to experts in this field (Eisenstein 2016). There is also the possibility that terrorists could hack into trucks and use them as a weapon like the attack that was launched in Nice, France in June 2016. If terrorists were to hack into autonomous vehicles they could cause vehicle crashes, detonate explosives or use them to drive trucks into crowds (Eisenstein 2016). More sophisticated governments could perhaps exploit security loop holes to engage in cyber warfare. A group of University of Michigan researchers recently demonstrated the hack of a heavy-duty truck in a paper presented at the USENIX Workshop on Offensive Technologies in August 2016 (Burakova et al. 2016). The researchers connected a laptop to the vehicles using the on- board diagnostic port. The fact that heavy-duty vehicles use the J1939 open standard allowed the researchers to easily look up the commands needed to communicate with the vehicles internal network, instead of reverse engineering them. They were able to send signals on the internal network of the truck to change the readout of the truck’s instrument panel, trigger unintended acceleration, or disable one form of the trucks brakes. The use of the J1939 standard in heavy trucks makes them easier to hack than passenger cars. While the hack demonstrated required physical access to the OBD port, it is commonly believed that hackers will find weaknesses in wireless systems that connect trucks. Most large fleets are connected via telematics systems to company databases and OEMs to help manage fleet operations and maintenance. In 2016 a security researcher identified a security weakness in thousands of trucks that left them open to over-the-Internet attacks via an insecure telematics dongle that tracked gas mileage and location (Greenberg 2016). Vehicle vulnerabilities to hackers fall into three categories: direct access, which includes the on-board diagnostics port; mid-range access, including Bluetooth and Wi-Fi; and long-range access, which involves breaching systems through cellular or radio functions such as truck telematics systems. Knowledgeable stakeholders have identified mid-range and long-range wireless hacks as a real and significant risk. These types of attacks have the potential to affect more trucks and cause greater harm (GAO 2016). It will be important to understand how V2V technologies affect existing – or create additional – paths for cyber-attack that could affect heavy-duty truck security. This is particularly true for technologies that need to access information from the in-vehicle network. A variety of specific technologies may pose an enhanced risk. Automated vehicle technologies that can intervene to prevent an accident may pose more cybersecurity risk than those that merely warn drivers of a potential hazard. It may also be harder to protect V2I technologies such as electronic truck inspection technologies that access vehicle fault codes or collect other data from the vehicle. Obtaining a better understanding of the risks involved and educating potential customers on these could be important to promoting the acceptance of these technologies (NHTSA 2014). This may be particularly important for state DOTs that are considering or pilot testing electronic enforcement and inspection technologies. GAO recently reviewed vehicle cybersecurity efforts to identify key practices to identify and mitigate vehicle cybersecurity risk. These practices included:

CRP Project HR 20-102(03) 50  Conducting risk assessments;  Incorporating security-by-design principles;  Creating domain separation for in-vehicle networks;  Implementing a layered approach to security;  Conducting penetration testing;  Conducting code reviews; and  Developing over-the-air update capabilities to allow security holes to be patched quickly (GAO 2016). While the report did not focus on commercial vehicle cybersecurity vulnerabilities or those that may emerge as newer types of connected vehicle technologies emerge in the future, many of the practices would likely be applicable to these new technologies as well. To date, there has been more emphasis on passenger vehicle cybersecurity than on those for commercial trucks. For instance, MAP-21 directed NHTSA to complete an examination of the need for safety standards with regard to electronic systems in passenger motor vehicles, and to consider various topics, such as the security needs for those electronic components to prevent unauthorized access (Public Law 112-141 2012). The FAST Act requires that the U.S. DOT submit a report to Congress on the activities of The Council of Vehicle Electronics, Vehicle Software and Emerging Technologies which was established by MAP-21 as a forum for research, regulation and enforcement. While there is significant overlap between cybersecurity for passenger and commercial vehicles, commercial trucks use a variety of different communications systems that could pose a risk – including mileage and fuel tracking systems, fleet and driver management software, maintenance telematics links, wireless systems to streamline size and weight enforcement among others. In the future the driver hours of service and other roadside safety inspections data may be communicated wirelessly to streamline enforcement. These communications systems may be installed by different vendors and managed by third parties. The combination of so many different systems and parties using these systems likely creates greater cybersecurity risk. Given that commercial motor vehicles could pose larger security risks and tend to more extensively use existing telematics technology, targeted research focused on the security of heavy-duty trucks may be warranted. NHTSA is significantly engaged in rulemakings and other activities to improve cybersecurity, so summarizing the risks for commercial vehicles and communicating the ongoing activities to reduce these risks could be an important part of this project. DOTs infrastructure is also vulnerable to hacks as was demonstrated by recent incidents where variable message signs were hacked (Kovacs 2014). A primary cause is believed to be vendor equipment shipping with the same default password that DOTs are expected to change but sometimes do not. As equipment and systems become more complex, more subtle and less predictable security risks are likely to develop. The combination of complex equipment and software from multiple vendors or the complexity of individual systems themselves can make it difficult to manage risks. The cybersecurity of infrastructure such as DSRC enabled traffic signals, vehicle routing systems, traveler advisory systems, automated enforcement systems will likely become a more significant issue in the future. While the U.S. Department of Homeland Security’s “Transportation Systems Sector Cybersecurity Framework Implementation Guide” provides high level guidance on evaluating the cyber risk of infrastructure, more practical guidance for practitioners is needed. There are also longer term concerns about how advances in technology may reduce the security of current wireless systems that depend on encryption technology. The development of large scale quantum computers, forecast by some to occur in the next ten years, could render current encryption technology obsolete. Replacements for conventional encryption technologies are not currently available.

CRP Project HR 20-102(03) 51 Privacy Ensuring that connected truck technologies safeguard the privacy of users is important to obtain the widespread acceptance and adoption of this technology. Industry is concerned about access of data such as tracking to outside entities including competitors and government regulators. To achieve a future state of connected trucks, documenting how V2V systems will safeguard privacy will be critically important. The general framework of standards and technologies that V2V communications are expected to employ, including DSRC and the Basic Safety Message elements, are designed to maintain the privacy of users. The U.S. DOT notes the following:  V2V systems will not collect or store any personally identifiable information about individuals or vehicles;  The safety messages exchanged by vehicles cannot be used by law enforcement or private entities to identify a speeding or erratic driver;  The V2V system will not permit tracking through space or time of specific owners, drivers, or passengers; and  The V2V system will not provide a “pipe” into the vehicle for extracting data (FHWA JPO 2016). There are some uncertainties and concerns about how V2V technologies and data might be combined with other data and exploited. The E-Government Act requires the federal government to assess and be transparent about the impacts of its activities affecting individual privacy. As required by law, NHTSA has conducted a Privacy Impact Assessment (PIA) and identified a number of risks to privacy posed by V2V technologies (NHTSA 2015). These include:  Basic Safety Message elements, alone or when combined with certificates, may act as “quasi- identifiers” for tracking;  Vehicle tracking history may be combined with other information (address information, public records, cameras) to personally identify the vehicle driver or owner; and  Radio fingerprinting could also identify the origins of messages and facilitate vehicle tracking. The government is conducting additional work to determine the magnitude of these risks. Congress also recently expressed concern that the commercial uses of DSRC could result in security breaches that could compromise confidential data and user privacy. They argued that the Federal Communications Commission (FCC) “should consider reserving the DSRC spectrum for V2V safety systems and not commercial applications that may make vehicles more vulnerable to safety, cyber, and privacy threats" (Eggerton 2016). Specifically, there is concern that commercial applications of DSRC such as paying for tolls, parking or gas could compromise the privacy and security of locational data. These systems could also be entry points for hackers to spread malware into vehicle systems. An advocacy group Public Knowledge has proposed a different approach that would have the “FCC mandate that anyone licensed to use DSRC, including commercial entities, submit privacy and cybersecurity plans to the FCC, require them to periodically update such plans, and require breach notifications” (Eggerton 2016). There are also industry concerns about the privacy impacts of some electronic inspection technologies. The extent to which electronic logging devices used for hours of service enforcement could violate the privacy rights of truckers is currently being litigated. The outcome of this case could have implications for the use of other electronic inspection technologies that depend on V2V or V2I technologies. The Owner Operator Independent Driver Association (OOIDA) has argued that the FMCSA electronic logging device (ELD) mandate violates the Fourth Amendment rights of truckers against unreasonable searches and seizures. They have made the case that for many truckers their truck is also an office and a home away from home. Monitoring the movement and activities of trucks for law enforcement under these conditions could be construed as an intrusion into the privacy of truckers. The Supreme Court has previously found that prolonged use of a warrantless GPS tracking device on a vehicle is a ‘search’ within

CRP Project HR 20-102(03) 52 the meaning of the Fourth Amendment. The Supreme Court is expected to rule on OOIDA’s case in December 2016. The outcome of this case could conceivably affect other types of connected vehicle applications, particularly those related to electronic inspections that rely on federal mandates of V2V or V2I technologies. The privacy implications of electronic inspections have been a continuous subject of Congressional scrutiny in recent years. MAP-21 directed FMCSA to “institute appropriate measures to preserve the confidentiality of any personal data contained in an electronic logging device” and to ensure that any data collected be used “by enforcement personnel only for the purpose of determining compliance with hours of service requirements.” The FAST Act followed up with a broader mandate that requires the U.S. DOT Secretary to certify that privacy concerns of truckers and other participants in any wireless roadside program are addressed.  

CRP Project HR 20-102(03) 53 Freight Planning Topics There are multiple issues from the planning perspective that need to be addressed to harness the maximum benefits of CV and AV technologies for freight. This section deals with the critical planning issues affecting freight operation in such environments. Relationship of CV and AV Adoption to Freight Planning The relationship between CV and AV adoption in freight and freight planning needs to be seen from multiple angles. First, the application of CV and AV will expand the capacity of existing roads due to platooning. At the same time, it will require dedicated facilities and infrastructure improvements. There is consensus among researchers that automated trucks will require less space headway than is required by manually driven trucks. This means that the capacity of existing infrastructure will increase after the introduction of automated trucks. However, it is uncertain how much capacity increment will be achieved given the uncertainties regarding headway requirements for automated trucks at different speeds and varying loads. In addition, automated trucks may share the road with non-automated trucks as well as passenger cars which may also be automated and non-automated; making the scenario even more complex. This warrants advanced and detailed research to understand the impact of the introduction of CV and AV technologies on infrastructure capacity and congestion. In addition, the application of CV and AV technologies will provide greater flexibility in the operation and routing of trucks than exists today. This flexibility can facilitate truck movement on selected routes within specified times of day, achieve network level objectives such as optimal travel times, or energy savings and emission reduction. There are uncertainties about the rate of market penetration of this disruptive technology in the trucking industry as different studies suggest different timelines (Litman 2014; Mashayekh et al. 2014). Therefore, the advancement in truck technology needs to be closely monitored. These benefits will only be realized when automated trucks achieve significant market penetration. Planning and operation of automated trucks will require new workforce requirements and this may occur at the expense of losing other jobs. Creation of a suitable workforce will require proper training that needs to be planned well in advance. These and other issues are discussed in more detail in the following Sections. Planning and Environmental Aspects CV and AV will have a significant impact on infrastructure capacity. Although passenger vehicles occupy a major role in the determination of road capacity the consideration of trucks is also crucial. CV and AV will enable shorter headways, permitting higher volumes at high speeds (Childress et al. 2015). In addition, approximately 60% of all congestion is attributed to non-recurring sources such as crashes and incidents (FHWA 2004). Freight transportation is responsible for a large fraction of all crashes (Kunze et al. 2010) and CV and AV can alleviate many of them. Therefore the application of CV and AV for freight can lead to increases in capacity particularly on rural and interstate highways where truck platooning is more likely. But, it is challenging to anticipate increases in roadway capacity with certainty. Studies from the literature suggest that application of CV and AV can lead to increases in capacity that can vary from 40% to 400% (Tientrakool et al. 2011; Shladover et al. 2013; Fernandes and Nunes 2012). Capacity will also depend on the safety level adopted which will dictate headway requirements (Mammar et al. 2004), especially important for trucks operating under varying loads. There is a strong relationship between freight demand and truck operating cost. The operating cost of trucks is likely to go down especially when SAE level 3 and higher technologies are operational due to better fuel economy and decreases in driver and safety-related costs. The decrease in operational cost can lead to increases in truck freight demand: (1) higher demand for transported goods than local goods, and

CRP Project HR 20-102(03) 54 (2) modal shift from other modes to truck. As a result, truck volume may go up. Although induced demand will be more pronounced in the passenger vehicle segment (Kim et al. 2015), significant induced demand is also likely in the freight segment. The increase in volume will also affect planning practice and needs. The application of CV and AV technologies will allow trucks to operate at decreased headways. This can lead to higher capacity in the existing highway network. At this juncture it is difficult to estimate how the dynamics between these two factors will play in the future but it is intuitive that the focus should shift to some extent from capacity enhancement (road widening) to ride quality/surface quality improvement, enhanced road marking, and improvement in geometric design. Decreased Parking Requirements As a consequence of CV and AV applications in freight, the parking requirement and rest area demand for trucks could go down, though further study is needed to understand and quantify the effect. The parking requirement may go down as the applications lead to less active driving, thereby relaxing driver rest time requirements. This may release large parcels of land that could potentially be used for other purposes. In addition, the introduction of CV/AV applications will also impact the consideration of firms when they select their sites for warehousing. The change in warehousing locations can change the truck volumes on roads. But this consideration is external to planning and may be difficult to estimate. This will necessitate the involvement of private party stakeholders of freight such as trucking companies, third party logistics providers and large shippers in the planning process. Environmental Considerations The better control of operations can enable CV and AV freight traffic to achieve greater optimization leading to savings in system level operations cost and fuel/energy (Besselink et al. 2015), but also to decreases in the associated environmental impacts. This can change the benefit cost analysis of potential projects drastically and can be especially useful in regions of non-conformity as per the definitions of US- EPA. Two contrary effects need to be analyzed with respect to environmental considerations: (1) increases in truck volume due to automation, and (2) decreases in fuel consumption and advancement in engine efficiency leading to decreased environmental impacts. High volumes of trucks may hold health hazards for some areas due to high emissions particularly if diesel engines are employed. Presently electric trucks are not a feasible option for operators due to range constraints. Evolving dynamic wireless charging (DWC) technology can enable trucks to operate across longer distances with small battery packs. Lack of Data and Planning Under Uncertainties The major issue in this domain is the lack of data and uncertainties related to vehicular characteristics and network performance when trucks are operating in a CV or AV environments. The macroscopic planning models require link performance functions for estimating the link travel times based on the volume of vehicles on the links. But there is no proven link performance function known for data under the AV traffic or mixed traffic condition. It is not known whether the automated trucks will use fossil fuels or other energy sources. Will there be requirements for charging stations, gas stations or will roads be enabled with dynamic wireless charging, or a mix fleet of electric and internal combustion engine vehicles requiring all of these three types of facilities? Such questions are very critical for facility location planning but presently there is a lack of data and information about these issues. It is also unknown how quickly the trucking companies will shift to automated vehicles and how the market penetration will grow in the future. These gaps related to truck data make it challenging for optimal short and long-range planning.

CRP Project HR 20-102(03) 55 We may be transitioning from an environment of data scarcity to one of data abundance. Traditionally freight planning has been constrained by a scarcity of data. The scarcity of data not only constrains the model calibration and modeling process but also makes it challenging to validate model output. The current practice of household travel surveys (among other data sources) does not support truck model development; however, CV and AV data would potentially be a better source of data (Wallace et al. 2015). Once automated freight becomes a reality large amounts of data will be available. This can become very useful for planning purposes, not only for better understanding individual truck activity patterns, but also, and perhaps, more importantly for generating much improved estimation of aggregate O-D flows (currently the weakest link in freight planning datasets). Although, the disaggregated nature of CV and AV data makes it a potentially rich data source for activity-based/tour-based models, this new data source may prove to be useful for four-step models after proper aggregation. Therefore there is need to prepare and plan for managing and processing huge data sets that will be coming from CV and AV enabled trucks in order to increase their usefulness for planning needs as we transition from the age of data scarcity to data abundance. Employment and Workforce Issues The trucking industry currently faces a labor shortage estimated at anywhere between 20,000 and 35,000 drivers. According to the American Trucking Associations, 240,000 additional drivers will be needed by 2020. Though heavy truck drivers are on the list of occupations with job growth through 2022, the projected shortfall of drivers is 11.3% according to the Bureau of Labor Statistics. As current drivers retire, not enough drivers from the next generation are replacing them. Young people are encouraged to go to college instead of seeking blue-collar work. Over-regulation, pay, and employee engagement are factors that may influence the desirability of a trucking career, though ultimately the long-haul lifestyle may just not be appealing to millennials (Lockridge 2015a). Applications of CV and particularly AV have the potential to mitigate these issues in multiple ways. AV has the potential to alleviate the problem of driver fatigue, although in the closer future lower-level (SAE level 3) automated trucks will still require a driver to act as a decision maker rather than as an active driver and it is unclear to what extent fatigue might be alleviated. Rather than remove drivers from the truck, highly automated technologies may attract new drivers by improving the quality of life of drivers and health and decreasing stress and job burdens. Younger recruits may be enticed by the opportunity to interact with such technology. By taking control of up to 85% of active steering, CV and AV can reduce driver sleepiness by as much as 25% (Lockridge 2015b). Increased safety, ability to engage in other productive activities while on the road, and ability to connect virtually to the outside world instead of staring at a highway might attract new drivers. At the same time related industries could face job loss due to CV and AV deployments including: insurance companies, small automotive repair shops, etc. The ability of AV to assume some portion of the driving role may play in to industry and government ruminations regarding allowing individuals younger than 21 to hold a CDL, and this could be another way in which the technology helps address the labor shortage. Planning and implementation of automated trucks will require new tools as well as new skills for both public and private sector workers. For example, travel demand modeling, in particular traffic assignment may require new algorithms that need to be incorporated into planning software. The route choice (RC) models will need to be updated to reflect the route choice of automated trucks; in particular when there are mixed traffic conditions. An important aspect related to route choice will be to decide whether the route selection is left on the on-board system of vehicles or whether they are provided through route guidance by a central server at the time of departure to achieve system wide optimal performance. This issue is important as the control of automated vehicles will determine the extent of system wide benefits. Operation and control of automated vehicles may require knowledge and expertise of emerging technologies which may be critical especially under conditions of emergency such as heavy snow, storms, floods etc. Both planning and the operation of freight under CV and AV environments will require training of personnel to achieve optimal operation. There is also need for new fields of expertise, for

CRP Project HR 20-102(03) 56 example hardware, software and communication experts related to operating and maintaining these CV/AV environments. In addition, some effort will be required to educate and increase the awareness of stakeholders regarding CV and AV. Some education is also required for the general public to spread the awareness about the technology and its benefits, and to provide response to their concerns related to truck platooning and safety. Therefore, there is need to consider the following synergistic efforts: (1) State freight planning to integrate CV and AV technologies, (2) MPO planning will need to develop planning and modeling tools and methods to consider CV and AV) operations in their planning and operations, and build awareness of CV and AV technology developments in their continuing planning functions. In summary, operational and technology needs will require new workforce training programs through professional institutions, schools and universities, and government agencies so that traffic engineers, transportation planners, automotive technicians, driver education teachers, and driver license examiners are equipped for the integration of CV and AV. The training and education of the workforce needs to be planned in advance for smooth transitioning from manual to automated driving. Truck Platooning, Human Factors and Safety Automation of heavy vehicles vs. passenger cars has different implications for driver training and truck platooning (Nowakowski et al. 2015). The primary motivation for automation of passenger cars is to free people from the task of driving, heavy vehicles will likely be automated as part of a commercial fleet for economic reasons. For heavy vehicles that are part of a commercial fleet, companies will need to establish driver training programs that are focused on: preventing driving skill degradation as level of automation increases, preventing handling system failures, and coupling and decoupling procedures for platooning. Truck platooning will require considerations such as: maximum platoon size, acceptable following distance, effect on surrounding traffic and drivers, lane usage, and how trucks should join a platoon. Adoption of automation technologies will create mixed traffic situations where drivers in non- automated vehicles are next to platoons of heavy vehicles with short time headways. Although truck automation promises to improve safety issues at full market penetration, it may adversely affect the safety of non-automated vehicles under mixed traffic conditions consisting of both human driven vehicles and automated vehicles especially trucks. In addition, driver behavior will be impacted while operating in a mixed CV and AV environment. For example, drivers in the vicinity of platoons may demonstrate behavioral adaptation by reducing their own time headways (Gouy et al. 2014). Therefore, transitional periods of technology adoption may decrease safety for conventional vehicles. The expectation for zero fatalities is unrealistic for any type of self-driving vehicle (Sivak and Schoettle 2015) due to reasons such as mechanical failure or software failure. Such incidents, if they occur, may result in more fatalities involving heavy vehicles. The focus on truck platooning has been largely concerned with energy savings (Shladover 2012) but safety impacts of truck automation require attention. Historically many projects have been approved on the basis of improving safety. Although, at this point it is difficult to quantify safety benefits due to CV and AV freight applications, it is likely to have significant impacts on project evaluation and selection. Dedicated Freight Related Infrastructure Improvements Infrastructure needs must be informed by trucking needs. The following section focuses on the need for infrastructure improvements especially for optimal operation of automated trucks. Infrastructure needs will be influenced by whether the approach toward CV and AV is considered as a partially automated automotive product operating in mixed traffic or as a fully automated public transportation technology using a dedicated right of way (Shladover 2012). Although, the basic infrastructure requirements for the operation of automated vehicles in the form of better road infrastructure such as better ride quality and visible road markings will be dictated by both passenger and freight vehicle segments, the efficient operation of automated trucks will require additional

CRP Project HR 20-102(03) 57 The Need for Dedicated Truck Lanes “To get the long term benefit of truck platooning we have to look at potentially separating some of the flow of trucks. I will share my experience that auto and other drivers are scared of trucks. You can only imagine the fear when people can see four or five truck platoons that are spaced may be three feet apart. I personally have concern about public acceptance of truck platooning even though we know that effective strategies can be deployed in the long run.” John Orr, Atlanta Regional Commission dedicated freight facilities. This is particularly important for interstate highways where truck platooning is likely to be most effective. As the trucks will be entering and exiting in the form of platoons from interstate highways, even though with small headways, they will necessitate longer on and off ramps than what exists today. The uninterrupted entry and exit of truck platoons will also require better geometric design such as smoother curves for on ramps and off ramps, and gentle vertical curves. Proper planning will also be required for the placement of V2I infrastructure. The location of road side units (RSUs) and their transmission power levels will depend on the traffic composition and the volume of trucks in addition to other factors. This may also involve trade-offs between effect on human health vs communication reliability (Gozálvez et al. 2012). There will also be the need for transnationally harmonized road markings and signage characteristics (Nitsche et al. 2014) especially for optimal operation of automated trucks that may cross multiple borders. Further, to harmonize freight traffic operation and to improve safety issues some dedicated freight corridors need to be constructed or some existing roads may be designated for trucks only use. This will necessitate digitization of available freight routes for CV and AV equipped trucks, preparation of digital maps and interfacing protocols with truck operation. The operating characteristics of heavy vehicles are different than light vehicles. Mixed traffic conditions are potentially dangerous in high speed conditions. It may be challenging to resolve safety issues for mixed traffic where CV and AV technologies are operating for both heavy and light vehicles especially at small headways. The potential of CV and AV to shift trucks more onto local roads makes this issue even more challenging. This may lead to the need for dedicated lanes/roads at two levels; (1) autonomous trucks and non-autonomous trucks (2) heavy vehicles (trucks) and light vehicles (passenger cars). Dedicated lanes may also be required given some environmental considerations. As mentioned earlier, high volumes of trucks in large platoons may pose health hazards especially for urban areas due to high emissions. Such health risks may be reduced or eliminated by moving from fossil fuel based trucks to electrified freight systems using electric trucks (ETs) powered by DWC. However, optimal efficiency of such systems may entail dedicated lanes for ETs. The need for the dedicated lanes for automated trucks may also be desirable given safety concerns. The potential of adoption of dedicated lanes with dynamic wireless charging for electric automated trucks needs to be evaluated from three aspects: (1) What should be the price mechanism (VMT or VMT Weighted by load or monthly/annual subscription or any other mechanism), (2) Benefit cost analysis of such systems, and (3) Willingness to pay for such systems. Some studies have focused on dedicated lanes (Toffetti et al. 2009; Mashayekh et al. 2014; Williams 2013) but more analysis is needed to fully understand their effectiveness. It may be beneficial to separate the automated trucks from the rest of traffic to facilitate the formation of truck platoons. However, demarcating lanes or links for automated trucks when their penetration is small may worsen the system level performance in terms of fuel use, congestion, and travel time. However, at full market penetration when all trucks are automated this step may naturally be obviated although separation of truck and passenger cars may result in significant benefits.

CRP Project HR 20-102(03) 58 Investments in Infrastructure Traditionally most of the infrastructure investment for roads has been used for road widening or adding lanes to existing transportation links and for the construction of new road links. But as introduction of CV and AV technologies may increase capacity, the existing infrastructure may be sufficient even after discounting the increase in truck volume resulting from induced demand. For example, research into heavy truck platooning by PATH found that operating tractor-trailer trucks in close-formation automated platoons of three trucks could enable a capacity of about 1,500 trucks per lane per hour, which is twice the capacity achievable with trucks driven individually (Bergenheim et al. 2012). A survey of 25 of the largest MPOs in the U.S. (Guerra 2016) indicates that investment decisions have not been affected to date by AV, though several interviewees voiced concern that currently planned investments might be unnecessary if AV brings increased roadway capacity. The operation of automated vehicles requires visible lane markings and may require intelligent on-road and road side infrastructure such that they can talk to on-board devices to inform the vehicles regarding traffic signal cycles, capacity of signal preemption to allow a platoon of traffic to cross an intersection etc. In addition, for a truck platoon to move seamlessly from freeway to arterial or to rural roads or to navigate a corridor with intersections without speed reduction will require improved geometric design. For example, gentle curves for exit ramps, longer exit ramps, and curved right turns at intersections as compared to perpendicular right turns may be some of the desired characteristics. In addition, in order to accommodate the high truck volumes with small headways will necessitate better surface quality of roads and thicker pavements. In essence, new improved infrastructure might be needed. Acquiring additional land for building improved infrastructure may be costly and difficult. This is especially so in urban areas, though the impact of highly automated trucks in urban environments may be at least several decades away. The new infrastructure can be built sequentially in the same place where the existing infrastructure exists. The major consideration is that the focus of infrastructure investments will need to shift from largely road widening to better geometric design, road markings and improved thicker pavement. Further, as infrastructure interventions to improve the surface quality and geometric design cannot be taken simultaneously for the entire transportation network. Therefore, staged or prioritized road infrastructure improvements needs to be implemented.

CRP Project HR 20-102(03) 59 Stakeholder Engagement The project team supported its research with participation at industry events and targeted interviews of leaders in stakeholder communities related to CV or AV in commercial trucking. Sentiments are represented in much of the content within this chapter. Indeed, while there is a lack of consensus on what is likely to deploy when, there does seem to be general consensus regarding needs of the federal government, state and local governments, and industry to work cooperatively as much in this domain as in any, to promote what most believe to be an exciting new world of possibility to improve many facets of freight trucking. Federal regulations are necessary, but so are guidelines that appropriately guide states into a consistency of environments established without being heavy-handed. There’s some evolution in the federal role that seeks to address gaps in private sector activity. As the private sector works to create new business segments and business cases for technology implementations at existing concerns, government can play a role particularly in pre-competitive research and development activity. In return this helps inform the regulatory arms of government authorities, which must be more closely aligned with their research counterparts. Likewise in planning areas such as workforce development, the opportunity for collaboration and coordination between, for instance, technology providers, fleet operators, and state agencies focused on economic development is important to be exploited using existing mechanisms and forums or new ones. The following individuals were interviewed as part of the research team’s effort:  Kevin Dopart, Vehicle Safety & Automation Program Manager, U.S. Department of Transportation  Paul Enos, Chief Executive Officer, Nevada Trucking Association  Elizabeth Fretheim, Senior Director, Logistics Sustainability, Walmart Stores  Vince Garcia, GIS/ITS Manager, Wyoming Department of Transportation  Nicole Katskides, Deputy Director, Office of Planning and Preliminary Engineering, Maryland DOT State Highway Administration  Tom Kearney, Freight Program Specialist, U.S. DOT FHWA  Anita Kim, Technology Policy Analyst, U.S. Department of Transportation  Robert Kreeb, Chief, Intelligent Vehicle Technologies Research Division, U.S. DOT NHTSA  Randy Mullett, Principal, Mullett Strategies  John Orr, Manager of Transportation Access and Mobility Division, Atlanta Regional Commission  Mo Poorsartep, Research Scientist, Texas A&M Transportation Institute  Brian Routhier, Transportation Specialist Technology Division, U.S. DOT FMCSA  Ted Scott, Director, Engineering, American Trucking Associations  Katrin Sjöberg, Connected Vehicle Technology Specialist, Volvo Group Trucks Technology

CRP Project HR 20-102(03) 60 Deployment Our exploration of how deployments might be expected to occur begins with a definition of a ‘baseline,’ that is, a summary of the many initiatives that are already under way. Some of this information has been presented above in the discussion of various applications scenarios. This section aims to collect a summary of significant initiatives that the research team is aware of. Connected and Automated Vehicle Pilot Research Projects Many states are conducting research and pilot projects to prepare them for a future with ubiquitous connected and automated vehicles. We identify some of the most important CV and AV projects for trucks by state in the table below. More detailed summaries of each program are presented in Appendix A. Table 3-3. Connected and Automated Vehicle Pilot Research Projects. State Program Location Year Activities AL Driver-Assistive Truck Platooning Auburn 2013-2016 Measuring fuel savings from two- truck platoons CA PATH Research on Platooning Berkeley 2003 & 2010-2011 Two- and three-truck platoons, V2V cooperative adaptive cruise control CA California Connected Vehicle Test Bed Palo Alto 2005- present Many CV applications tested. The Multi-Modal Intelligent Traffic Signal System was used to test freight signal priority. FL Tampa Hillsborough Expressway Authority (THEA) Connected Vehicle Pilot Program Tampa 2015- present V2P, V2V and V2I for the exchange of safety and traffic information between vehicles, pedestrians and crosswalks, traffic signals and other elements of the infrastructure. FL Orlando Connected‐Vehicle Test Bed Orlando 2010- present V2V & V2I FL Miami Dade Pilot Project, AV/CV/ITS Freight Applications Miami Ongoing Connect freight vehicles to traffic signals to provide priority in off peak hours MI Safety Pilot Model Deployment Ann Arbor 2012-2013 Test safety applications using V2V technology in 2,800 cars, trucks and transit vehicles. MI U.S. Army Tank Automotive Research I-69 near Flint 2016 V2I communication, driverless platooning MI MCity Test Facility 2015- present Testing in simulated urban driving environment MO Virtual Weigh Station Route 67 & Route 55. ongoing Magnetometer loop detector (MLD) used to “fingerprint” the commercial motor vehicle on the mainline. Used in conjunction with prepass system. NV Nevada Public Road Testing State-wide 2015-2016 Freightliner granted license to test autonomous truck on public roads NJ The Center for Automated Road Transportation Safety Fort Monmouth Ongoing. Focus on research, development, certification, and commercialization of autonomous collision-avoidance technology for trucks and buses

CRP Project HR 20-102(03) 61 State Program Location Year Activities NJ Development of algorithm to reduce truck idling at adaptive signals U.S. 1 in Mercer and Middlesex Counties Ongoing Adaptive signal communicates with truck to allow engine shut-off when waiting at light and provides two second warning for green signal to allow engine re-start NY New York City Department of Transportation Connected Vehicle Pilot Program Midtown Manhattan Ongoing V2V, V2I and signal upgrades. 10,000 city-owned vehicles. Does not target freight trucks. NY 5.9 GHz Dedicated Short Range Communication Vehicle-Based Road and Weather Condition Application New York, on the Long Island Expressway 2014 Test the acquisition of road and weather condition information from DSRC-equipped public agency vehicles OH Columbus Ohio Smart City Program Columbus 2016 Smart phone application giving truck and freight drivers access to real-time traffic information and routing PA CMU Cranberry Township and Pittsburgh Test Bed: Pittsburg & Cranberry Township Ongoing Traffic signals equipped with DSRC radios PA PennDOT Ross Township Test Bed Ross & McCand- less Townships Adaptive traffic control signals and Dedicated Short Range Communication (DSRC) TN Trusted Truck ® Knoxville 2004-2010 Wireless safety inspections TX I-35 Connected Work Zone Texas I-35 corridor 2014-2018 Optimize commercial freight movements through work zones using CV technologies TX TTI Truck Platooning Study Texas corridor TBD 2015-2016, plus future phases (unfunded) Feasibility study of a small-scale SAE level 2 truck platooning deployment VA Virginia Connected Vehicle Test Beds Blacksburg & Fairfax County 2012- present Instrumented vehicles include semi-truck UT Truck Automation Pilot Route 80 through Salt Lake City ongoing Testing weather probes and truck platooning WY ICF/Wyoming Connected Vehicle Pilot Program I-80 corridor 2015- present V2I & V2V technology Provide weather advisories, roadside alerts, parking notifications and dynamic travel guidance for approximately 500 vehicles. Multi- state Wireless Roadside Inspection (WRI) Research Project Test sites in KY, TN, MS, NC, GA 2007-2017 600 commercial vehicles outfitted with telematics that allow for the wireless transmission of operator hours and credentials to the roadside for inspection. Multi- state Drivewyze E-inspections Field Test VA, MD, DE, PA 2016 Level 3 electronic inspections Multi- state Multi-Modal Intelligent Traffic Signal System AZ, VA 2014-2015 Tested and simulated freight signal priority impacts Multi- state Commercial Vehicle Infrastructure Integration Greensboro, NC and the Long Island Ongoing Testing CV VII compliant On-Board Equipment (OBE) system including HMI for communication and 5.9

CRP Project HR 20-102(03) 62 State Program Location Year Activities Expressway & Spring Valley corridor in NY GHz DSRC. Experiences Outside the U.S. There is substantial interest around the world in CV and AV for trucking, specifically in those countries and regions that are taking a leadership position in the development and implementation of connected and automated vehicle solutions generally. Beyond the U.S., much of the exploration of truck platooning systems in particular has been conducted by research groups in the E.U. (particularly Germany) and Japan Key reasons for this are the higher traffic density in these countries, as well as the ratification of the Kyoto Accords by most European countries and a partial commitment from Japan. Kyoto Accords are committed to a reduction in fuel consumption as well as greenhouse gas production, specifically a mandate on the limitation on the creation of CO2. This focus on emissions reduction has led to vehicle automation projects primarily focused on increasing the fuel savings on vehicles. Three such projects outside the U.S. are SARTRE, iQMatic, and Japan’s Energy ITS project. European Union E.U. developments in truck platooning range from OEM developments, road tests, and political developments aimed at creating an environment where deployment can be expected to proceed in the relative near-term future. The E.U. has been a leader in advancing early automation in trucks, such as the requirement for automatic emergency braking on its trucks. The European Commission is an active funding source for many of the initiatives; individual E.U. member countries also provide funding. Selected projects that are active or recently completed are described below. Other older projects such as the German project KONVOI and PROMOTE-CHAUFFEUR investigated topics such as technical feasibility, environmental impacts, and driver acceptance of truck platoons. European Truck Platooning Challenge. The Netherlands launched this Challenge during its 2016 term as president of the Council of the European Union. Six OEMs participated, running two- and three-truck platoons, some but not all with automated steering control, on public roads from several European cities to the Netherlands during March and April 2016. The Challenge culminated in a “Declaration of Amsterdam”, signed by all 28 E.U. member states, that creates a common policy framework under which truck platooning and other connected automated systems can be deployed. The Challenge was perhaps the most comprehensive to date in terms of recruiting the various stakeholder communities needed in an eventual deployment, including the aforementioned OEMs, fleet operators, alliances of fleet operators and shippers, and road operators. It was also meaningful in that it was run on open highways, unlike many other tests to date that have been run on test tracks or in otherwise controlled environments. The results were sufficiently positive that a relatively aggressive agenda for further development and deployment is being advanced, with particular but not exclusive support from the Dutch government. Visionary goals for this development include multi-brand truck platoons legally operating within and across E.U. borders by 2020, with three or more trucks platooned with headways as small as 0.3 sec.; and SAE level 4 automated platoons operating with following drivers legally able to disengage from the driving task and, for example, have a rest period, by 2025. iQMatic. The industry is investing in developing autonomous heavy-duty vehicles to try to solve problems such as traffic accidents and traffic congestions. iQMatic is a project in collaboration between

CRP Project HR 20-102(03) 63 the KTH Royal Institute of Technology, Linköping University, Autoliv, Saab, Combitech and Scania CV AB with the aim of developing a fully autonomous heavy vehicle for goods transport and other industrial applications. It focused on introducing autonomous trucks in controlled environments. There, the truck must be able to perform autonomously difficult tasks such as navigation, obstacle avoidance and decision making (KTH Royal Institute of Technology 2016). U.K. Driverless Lorry Trials. The U.K. government has recently committed to conduct road trials of platooning lorries. This development builds upon prior work supported by the government that demonstrated truck platooning’s technical feasibility. The test will be particularly interesting given the stated intentions of running in mixed traffic and on a busy roadway with a density of entrance and exit ramps. E.U. Cooperative ITS Corridor. A coalition called the Amsterdam Group has been working to deploy an intelligent highway corridor traversing through the Netherlands, Germany, and Austria. Such a corridor will nominally be a location where a variety of applications can test and deploy CV/AV technologies in a multi-country environment. The initial focus has been on V2I, and the specific application of work zone warnings involves trucks. SARTRE. This was a three-year research project completed in 2012 with primary focus on the development of a highway platooning system that is compatible with multiple vehicle types (e.g. freight trucks, passenger buses, and personal vehicles). Two main objectives were reducing environmental impact through reduced fuel consumption, and reduced congestion through the more efficient use of the road network. The system relied on a trained professional driver in the lead vehicle but provided SAE level 2 automation of following vehicles (a following truck followed by three passenger vehicles). This allowed drivers in these following vehicles to take their attention off the road. SARTRE used 5.9 GHz DSRC radios for V2V communication as well as many other sensors currently in production vehicles such as radar GPS and stereo cameras. V2V information was sent to the other vehicles at a rate of 40 Hz instead of the 10 Hz typically used. This allowed for higher bandwidth in the controller. The primary focus of this research was on the environmental impact of such systems, thus extensive fuel tests were conducted as well as the creation of Computational Fluid Dynamics (CFD) drag models to provide analytical solutions along with empirical data. The SARTRE study conducted safety tests including cut- ins on private tracks as well as public highways. The primary vulnerability of this system is on its heavy reliance on the decisions of the lead vehicle driver. Because each preceding vehicle is essentially performing waypoint following of the lead vehicle’s path any driving errors in the lead vehicle can result in much larger problems for vehicles downstream. COMPANION. This three-year E.C. funded research project, ending in fall 2016, looked specifically at issues around operationalizing platooning, including HMI interfaces, the legal and regulatory environment, and the management schemes needed for the creation and dissolution of platoons. Japan Japanese industry and government have been active in developing CV and, more recently, AV applications in trucking (in addition, of course, to passenger vehicles). Energy Intelligent Transportation Systems (ITS) Project. Following along the same environmental emphasis as SARTRE, Japan’s METI partnered with the Japan Automobile Research Institute and others to develop CACC and platooning systems for heavy trucks. This project ran from 2008-2012 and was designed to reduce the following gap between vehicles thus allowing for decreased aerodynamic drag on all vehicles yielding a reduction in fuel consumption. The project was designed to provide longitudinal

CRP Project HR 20-102(03) 64 and lateral control to all vehicles in the platoon, and used sensor packages available on many vehicles currently in production such as millimeter wave radar, lane detecting cameras, GPS and IMU. In their field testing, a fleet of trucks traveled safely at 80 km per hour with a spacing of just 10 meters. Documented benefits included a reduction in wind drag and elimination of uneven braking and acceleration. Results of the project have helped inform and lead to additional developments. Specifically, Japan’s Ministry of Land, Infrastructure, Transport and Tourism (MLIT) established an autopilot system study group to further examine issues and policies needed to enable automated driving (of passenger vehicles as well as trucks). Smart Logistics Pilot. This MLIT-sponsored CV test uses Japan’s Electronic Toll Collection (ETC) 2.0 infrastructure, and includes an element looking at vehicle fleet management. The intention is to collect truck positional data, and vehicle operational data such as sudden braking or swerving, with the objectives of improving truck operational efficiency (through better routings) and road safety. Also included in the concept is determination of overloaded vehicles and ability to issue warnings. Singapore AV FOTs. The Singapore Land Transit Authority is engaged in a series of FOTs around automated vehicles, including the consideration of various freight concepts such as platooning. Autonomous Inter-Port Shuttles. Singapore’s Ministry of Transport has announced an agreement with the Port of Singapore Authority (PSA) to develop autonomous truck platooning technology for transporting cargo between port terminals, and with Sentosa Development Corporation and ST Engineering to trial self-driving shuttle services across Sentosa. Canada ecoTECHNOLOGY for Vehicles Program. This umbrella program sponsored by Transport Canada tests the performance of advanced heavy-duty vehicle technologies including CV and AV, as well as that of passenger vehicles, and with a focus on safety and environmental measures. As a part of this program, the government has sponsored initial work to understand the feasibility of truck platooning systems. Australia Australia is particularly active developing heavy truck CV applications, including Truck-activated Speed Advisory Signs, Intelligent Speed Adaptation, and Lane and Road Departure Warning. Korea Korea’s Ministry of Land, Infrastructure and Transport is active in a C-ITS Pilot Project, which includes a specific application in Road Work Zone Warning. China Among the Cooperative Intelligent Transportation System work being pursued in China is a project investigating and conducting a demonstration test of commercial vehicle V2I technology and policy (Song 2016).

CRP Project HR 20-102(03) 65 Future Deployments The lists above, as well as discussion throughout this report, demonstrate the substantial level of activity underway in development of CV and AV applications for trucking. The current focusses are on research, development, testing, and business and business case formation. Development work continues in truck platooning, but is moving closer to early deployment. Commercial applications on select roads might have an expected deployment around 2020. Smaller test deployments, such as that noted in this report under consideration in Texas, might occur earlier. It will likely take a decade or more before a broad commercial deployment on a robust network of corridors is possible. During this time, uncertainties over the business case may be resolved, and carriers’ interests confirmed. It is imperative too that the federal and state DOTs actively monitor results and continue to provide funding for pre-competitive research and for results that help inform regulatory action. While it has been pointed out in this report that the business case is still largely an unknown, it is nonetheless probable that early deployments will be pushed by the private sector (and indeed the evidence is already supporting this). The legal environment is sufficiently in place in some states, and for the sake of small-scale deployments, road infrastructure can be hand-picked. Those operations for which platooning would be most economically beneficial will have opportunities to engage in tests and limited deployment. So platooning is expected to migrate from testing to commercial application in a highly gradual mode. With expected sharing of experiences, with advocates for information sharing coming from vendors, shippers, and the government, this will inform the characteristics of future deployments. The present unknowns in the business case are more likely to affect whether platooning ends up as a small niche or something substantially widespread. Terminal specific applications such as automated trailer backing require additional development of the concept including articulation of the business case for those who operate yards. This might be expected within the next 5 years. Other applications such as Traffic Jam Assist, as well as SAE Level 0 CV applications, can likewise be expected during this timeframe. Highly automated trucks will be deployed for the foreseeable future with human drivers on-board. A possible first application might be in replacing a driver team with a single driver. Issues of regulation, public acceptance, fail-safe behavior, etc. will need to be worked out and this creates the uncertainty of specific deployment timeframe. There is a potential for meaningful economic return for adopters who are able to reduce the labor needed to haul freight, and this will help keep pressure to actively pursue these applications. Over the next several years, the experiences that the community will realize in these deployments will of course greatly inform longer term developments. And history has shown that unexpected external actions such as the marketing of new technologies will likely arise and further shape a longer term future. At this time though, it seems that the economic attractiveness of operating trucks with no human driver will entice (continued) active exploration of suitable solutions. It also seems likely that any movement to such a future state will be evolutionary in nature, for example being constrained to highway lanes, possibly dedicated, or to following trucks in platoons led by a truck with a driver on-board. This sort of transition, if it occurs, might most likely occur in the 2030s.