The Impacts of Vehicle Automation on the Public Transportation Workforce (2022)

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.

21   C H A P T E R 3 There are many different ways in which automated transit vehicles could be used to provide automated transit services. The impact of vehicle automation on the transit industry will depend on the degree of automation and the rate at which HAVs are introduced into the market. Instead of attempting to determine the total impact of automation on the transit workforce, the research team selected five transit automation use cases using HAVs and estimated the workforce effects of those five use cases, applied to specific existing conventionally driven transit services. All but one of these five use cases (i.e., automated local bus service) are specifically a part of the FTA’s STAR Plan (Machek et al. 2018): 1. Bus Automation for Maintenance and Yard Operations. 2. Low-Speed Automated Shuttles. 3. Automated BRT. 4. Automated MOD. 5. Automated Local Bus Service. Although the following sections of this chapter describe each use case individually, there are some important factors that the research team considered generally applicable across all use cases unless otherwise stated: • The objectives for any transit agency in an initial HAV deployment will be demonstrat- ing capabilities and customer value, as well as learning how to safely deploy, operate, and maintain that technology (Horadam 2018). There are several challenges that will need to be addressed prior to full deployment, including public acceptance and trust, regulatory stan- dards, liability and insurance issues, ethical questions regarding the vehicles’ algorithms, and how the vehicles will operate in mixed traffic (i.e., alongside non-autonomous vehicles, pedes- trians, and other vulnerable road users). Transit agencies do not serve private investors and therefore will probably not push technological boundaries for competitive advantage. Since transit agencies are risk-averse, the first deployments probably will not remove labor from vehicle operations entirely. However, the uses cases and their workforce effects are described from the perspective of full technological readiness. That is, the workforce effects of use cases are calculated as if the technology for the use case is 100 percent safe, reliable, and accepted by passengers. • Elderly passengers or those who have disabilities may still need assistance boarding and exiting HAVs. The research team recognized this fact, but ultimately the goal of this research was to estimate the impact of transit automation on the transit workforce. To do so, the research team had to assume that either HAVs would be designed in such a way that passenger assistance could be handled remotely or transit agencies would figure out operational models that would allow passenger assistance to be provided on-call (e.g., from a staff member in the field). Transit Vehicle Automation Use Cases

22 The Impacts of Vehicle Automation on the Public Transportation Workforce • Many passengers may not feel comfortable riding without a human presence on board. However, to estimate the potential effect of the use cases on the transit workforce, the research team had to assume that the use cases, as designed and implemented, were accepted by passengers and embraced by transit agencies. • If the technology makes it possible to safely remove the driver, transit agencies will have to decide whether to convert the driver job to a customer service job (with some potential requirements for driving) or eliminate the position altogether. There are some potential ben- efits to keeping an onboard attendant job. For example, automated service attendants would not have to drive, eliminating the stress and physical exhaustion of driving for long hours, potentially increasing operational safety and driver retention—with reduced operating cost advantages. No longer driving, bus operators could devote most or all of their time to cus- tomer service to greatly improve the rider experience. The research team built two operational models—the in-person operations model and the remote operations model—to help handle the uncertainty about whether humans will continue to be onboard automated transit vehicles. In the in-person operations model, an employee is onboard every HAV; in the remote opera- tions model, employees would monitor and potentially control HAVs remotely. (Section 6.2.3 provides more details about these operational models.) • Each automated transit use case will likely have different timelines, which depend heavily on both regulatory and legal developments. HAVs must be able to operate legally and to have reasonable liability in the event of an incident. The research team had to assume that the regulatory and legal environment would be adequately established to support the adop- tion of each use case in order to estimate the potential effects of the use cases on the transit workforce. • For all except Use Case #1 (bus automation for maintenance and yard operations), the use cases could potentially be directly operated by transit agencies, through purchased trans- portation agreements with traditional private operators (e.g., First Transit, TransDev, etc.), or by partnerships with vehicle manufacturers or technology companies (e.g., Waymo, EasyMile, etc.). However, because this research focused on the workforce effects of replac- ing current conventionally driven services with automated transit services (see discussion in Section 2.6.2), the research team focused the analysis on current bus transit employees— that is, employees who work for public transit agencies or their private operators— and did not estimate impacts on other parts of the U.S. workforce (e.g., employees who work for EasyMile). The following use case descriptions are not meant to provide an exhaustive review of what is known about each use case and are based on both academic and non-academic sources, combined with the research team’s professional judgments and opinions. In fact, it is likely that between the time of the writing of this report and its publication, more developments and changes have occurred. The goal of the use case descriptions was to describe key characteristics of each use case in order to provide the background information necessary to understand poten- tial workforce effects. Information for each use case includes: • A general description of the use case. • Supporting technology: Briefly summarizes what HAV technologies are necessary to support the use case. • Current deployments: Lists a few of the known current demonstration, pilot, or full deploy- ments of the use case. • Potential operational impacts and workforce effects: Describes the types of services that the use case might impact—for example, rural or urban areas, fixed-route or demand- response, etc. Also describes the day-to-day operations and activities that the use case might change within a transit agency and the resulting effects those changes may have on the transit workforce.

Transit Vehicle Automation Use Cases 23   • Implementation horizon and considerations: Discusses any known information regard- ing when the use case may be implemented in the United States beyond demonstrations and pilots. Describes some of the factors that would likely influence which transit agencies might implement the use case and to what degree and presents the research team’s adoption assumptions when workforce effects were later estimated in the partial adoption scenario. (See Section 2.6.3 for a brief discussion of the adoption scenarios.) • Planning and policy decisions: Lists what relevant planning and policy decisions will sig- nificantly impact the service, operational, and workforce effects resulting from the use case. These planning and policy decisions were taken to transit industry representatives for their opinions (see Chapter 5). 3.1 Use Case #1: Bus Automation for Maintenance and Yard Operations This use case includes the automation of bus movements in transit agencies’ operations and maintenance (O&M) facilities, such as movements to and from maintenance bays and vehicle storage, parking, washing, fueling, and farebox vaulting areas. Because these are con- trolled environments, this use case represents L4 technology. In the FTA’s STAR Plan, this automation use case includes (a) precision movement for fueling, service bays, and bus wash and (b) automated parking and recall (Machek et al. 2018). Since the systems of this technol- ogy are not currently available, the following subsections describe assumptions based on a review of available literature. 3.1.1 Supporting Technology This use case will need the baseline technology architecture identified in Section 2.2. The tran- sit agency’s maintenance facilities (e.g., refueling, cleaning, and inspection and service bays) and parking areas may need to be equipped with supportive devices for interactions with vehicles (i.e., V2I communication) (Brewer et al. 2019). Also, to take full advantage of this use case, sta- tionary equipment and devices used for farebox vaulting, fueling and servicing, interior clean- ing, and washing might also require modification or some form of automation. The use case may require intensive mapping of the facilities or, if needed, reconfiguration of the infrastructure at the facility (Peirce et al. 2019). Other forms of localization technologies [e.g., embedded mag- nets or radio-frequency identification (RFID)] could also be used to support movements inside buildings where GPS may be less reliable. 3.1.2 Current Deployments In Europe, several projects related to automation for bus maintenance were conducted. In 2017, a German research university, Karlsruhe Institute of Technology (2017), investigated an autonomous bus depot system. This system included automation for refueling, preliminary cleaning, and vehicle washing before the bus began its service. In Paris, France, a fully autono- mous bus garage was initiated in 2018 by the RATP Group (the state-owned public transport operator), the CEA (the French Alternative Energies and Atomic Energy Commission), and Iveco Bus (RATP 2018). This research project, called Intelligent Garage and Predictive Mainte- nance, included two use cases: (1) on arrival of a bus driver, an autonomous bus (L4) leaves its parking space and goes to the exit of the garage (the driver takes control at the exit); and (2) at the end of the service, the driver leaves the bus at the entrance to the garage, and then the bus parks in a preassigned parking space in the garage. Both projects appeared to be successful tests of the feasibility of the use cases; however, neither project examined or collected data on work- force effects related to the deployment.

24 The Impacts of Vehicle Automation on the Public Transportation Workforce 3.1.3 Potential Operational Impacts and Workforce Effects Although the main objective of automation for yard operations is to increase efficiency within transit agencies’ O&M facilities, the use case has the potential to also improve safety within the yard by reducing conflicts between vehicle and maintenance staff and to enhance the efficiency of the overall transit system and on-time service by simplifying yard movement procedures needed for daily O&M (Machek et al. 2018). This use case may reduce labor expenses if fewer yard staff are needed; however, it is uncer- tain whether this will be the case (Machek et al. 2018). Reductions in pre- and post-trip driving tasks performed by bus operators could also result in labor hour reductions; however, reduc- tions may not translate into reduced pay hours and instead be reinvested in additional revenue hours. In addition, with more precise docking and maneuvering technology, the system may increase safety and at the same time optimize the use of available space in the transit agency’s maintenance or storage areas if vehicles can be parked closer together (RATP 2018; Karlsruhe Institute of Technology 2017). For example, at one bus facility, Peirce et al. (2019) estimated that automated buses could be parked in one-third of the current space needed for convention- ally driven buses. However, results would vary significantly based on bus size and O&M facil- ity configuration. Despite the vehicle automation within the facility, a human operator and/or maintenance staff would still be required for some daily O&M tasks, such as putting in the hose to refuel diesel buses (Machek et al. 2018), unless those systems are also automated. This use case does not include autonomous operations outside the O&M facility and therefore does not have any direct service impacts. 3.1.4 Implementation Horizon and Considerations In the United States, according to the FTA’s STAR Plan, the demonstration of automation for maintenance and yard operations were to be completed by federal fiscal year 2021 (Machek et al. 2018). One of the Automated Bus Consortium’s pilot projects includes automated yard movements at MetroLINK’s Operations and Maintenance Center Yard (Automated Bus Consortium 2019). The applicable fleet for this use case included typical 40-foot buses, cutaway buses, or articulated buses (Machek et al. 2018), so the use case could be deployed at any type of transit agency with bus service. How- ever, urban transit agencies with moderate to large bus fleets may be more likely to implement this use case because savings in yard capacity or reductions in staff time dedicated to yard movements will likely be too minuscule at smaller transit agencies (e.g., rural transit agencies) to justify the investment in both the bus and the yard infrastructure to sup- port the use case (e.g., see Machek et al. 2018; Appendix D). However, the cost-benefit analysis of implementing this use case will depend heavily on local conditions and the configurations of individual bus yards. 3.1.5 Planning and Policy Decisions Several key planning and policy decisions will need to be made that will have implications on how bus automation for maintenance and yard operations will affect the transit workforce, including: • Will transit agencies opt to deploy this use case at all or only select bus yards? Within an automated bus yard, will transit agencies opt to equip all buses with the technology or only some buses? For workforce effect estimates under the partial adoption scenario, the research team applied the bus automation for maintenance and yard operations use case to small and large urban transit agencies that had at least one mode of bus service.

Transit Vehicle Automation Use Cases 25   • Will the reduced time associated with pre- and post-trip activities (e.g., bus retrieval, vaulting, and parking) be reallocated to revenue service or other productive activities, or will operators’ schedules have fewer paid hours? • Will the reduced time associated with yard movements during daily cleaning and fueling activities (often occurring overnight) be reallocated to new tasks, allowing for more time to perform current manual tasks, or will the number of hours needed for cleaning and fueling activities be reduced (potentially reducing staffing requirements)? • Will the reduced time associated with yard movements during maintenance activities (e.g., retrieving buses from the yard to perform preventive maintenance) be reallocated to new tasks, allowing for more time to perform current maintenance activities, or will the num- ber of hours needed for maintenance activities be reduced (potentially reducing staffing requirements)? • What additional maintenance activities and skills will be required to support any infrastruc- ture needed to enable precision movements of buses in yards and possibly garages? 3.2 Use Case #2: Low-Speed Automated Shuttles In 2018, the USDOT described automated shuttles as a new category of vehicles and associated services that are best defined by a set of common features (Cregger et al. 2018). These features are: • Fully automated driving: Vehicles are intended for use without a driver or operator onboard. • Restricted ODD (L4): – Operation is intended for protected and less complicated environments. – Service is generally limited to 25 miles per hour (or lower), with cruising speeds around 10–15 miles per hour. • Shared service: Vehicles are designed to carry 4–15 passengers, including unrestrained passengers and standees. • Shared right of way: Vehicles share the right of way with other road users, either at designated crossing locations or along the right of way itself. As L4 vehicles, these shuttles do not require a human operator as long as the shuttles stay within their ODD. However, early demonstrations included an onboard human attendant to observe passengers, record data, answer questions, and serve as a safety operator if needed (Machek et al. 2018). 3.2.1 Supporting Technology This use case will need the baseline technology architecture identified in Section 2.2. However, to serve as fully autonomous transit vehicles for automated transit services, shuttles will also need at least some of the specific automation-supporting technologies for automated transit services discussed in Section 2.3 (e.g., interior sensing and automated fare collection). The vehicle types are typically small shuttle buses available from suppliers such as EasyMile, Local Motors, SB Drive, AURO, and Navya. All suppliers are working on a number of different concepts; all suppliers but SB Drive are building around electric powertrains. While less typical, automated shuttles also operate as smaller pod vehicles serving as few as two passengers, as well as much larger shuttles built to accommodate up to 24 passengers. Many of these shuttles that are manufactured to be driverless may not come equipped with a steering wheel, brake pedal, accelerator, or mirrors. As such, many shuttles do not meet the National Highway Traffic Safety Administration’s Federal Motor Vehicle Safety Stan- dards (FMVSS) and are not allowed to operate on public roadways without a special waiver from NHTSA.

26 The Impacts of Vehicle Automation on the Public Transportation Workforce 3.2.2 Current Deployments Many people may encounter their first HAV as an automated shuttle. In fact, recent research found that significantly more survey respondents would likely use low-speed auto- mated shuttles rather than automated ride-hailing services or privately owned vehicles (Zmud and Sener 2019). Europe was an early leader in piloting low-speed automated shuttles, with its CityMobil and CityMobil2 projects, as far back as a decade ago. As of the beginning of August 2018, there were more than 260 demonstrations and pilots (with some planned, some ongoing, and some completed) in North America, Europe, Asia, Oceania, and Africa (Cregger et al. 2018). Pilots are longer than demonstrations and often last anywhere from a few months to an entire year or more. Demonstrations last anywhere from a few hours to several days. Both domestically and abroad, deployments are split roughly equally between demonstrations and pilots. Typically, deployers involved in pilots are interested in learning about longer-term aspects of operating a shuttle, including service capabilities, costs, and user acceptance. In many cases, the pilot is used to determine if there is a business case for operating a shuttle in revenue service. Many auto- mated shuttle pilots have run into unanticipated costs and delays, notably through unexpected increases in staff time (Hughes-Cromwick and Dickens 2019). Most demonstrations and pilots continue to keep a safety attendant on board every shuttle, and third-party operators (e.g., TransDev, First Transit) often handle much of the O&M of the shuttles (Haque and Brakewood 2020). However, the low-speed automated shuttle practice area is changing rapidly. More details about low-speed automated shuttles, including current deployments, operational models, management and funding, and accessibility, can be found in TCRP Research Report 220: Low-Speed Automated Vehicles (LSAVs) in Public Transpor- tation (Coyner et al. 2021). Although there have been many low-speed automated shuttle demonstrations and pilots, the workforce effects of low-speed automated shuttles have not yet been well researched. 3.2.3 Potential Operational Impacts and Workforce Effects Low-speed automated shuttles have been assumed to offer an opportunity for lower-cost public transportation rather than traditional public transit because of (a) reduced labor costs associated with not employing a driver or other onboard attendant; (b) reduced capital and operational costs associated with smaller, lower-capacity vehicles; and (c) lower fuel and main- tenance costs compared to conventional vehicles if deployed as electric vehicles (EVs) (James 2018; Kamiya and Teter 2019). These assumptions depend on the evolution of the technol- ogy to a level of capability that would (a) enable operation without an onboard attendant to achieve labor cost savings; and (b) enable economies of scale, improve reliability, and justify the investment relative to other options to achieve vehicle cost savings (Kamiya and Teter 2019). Expanded deployment of automated low-speed shuttles without onboard attendants may be expected to reduce employment for transit operators and thus may face opposition from transit employees and labor unions, as well as other stakeholders, including the general public (Cregger et al. 2018). Even deployments with onboard safety or customer service attendants may be opposed because of the perception that such projects would represent a step in the direction of operator and maintenance job losses or wage reductions. While EVs tend to be more expensive to purchase, they also tend to have lower fuel and main- tenance costs than conventional vehicles. Low-speed EV shuttles operating in geographically constrained environments may be able to overcome some of the operational challenges faced by more free-floating on-demand systems. These challenges are limited driving range, access to charging infrastructure, and charging time management (Kamiya and Teter 2019).

Transit Vehicle Automation Use Cases 27   Due to their low speeds, most automated shuttle deployments are being designed to operate on predefined routes or geofenced locations, such as corporate or university campuses, or to provide last-mile travel between transport hubs and final destinations. These unique and limited environments present a less challenging operating environment than that of open-road driving, reducing the number of extreme edge cases that suppliers would have to account for. The typical service design for automated shuttles is along a set route, stopping at all stations along the route on passenger request. The shuttles might operate in a loop, going in one direction, or traverse back and forth along a single route without turning around (Cregger et al. 2018). However, the shuttles could also operate on demand, being hailed by a passenger using a smartphone app (Cregger et al. 2018). Potential transit service applications include (a) circulator bus service— fixed-route or flexible service between two or more points, and (b) feeder bus service—connections to fixed-route transit stations. At the time of the writing of this report, automated shuttles and supporting technologies were mostly managed and maintained by either original equipment manufacturers (OEMs) that provided automated shuttles or third-party operators (Haque and Brakewood 2020). Transit agency employees were usually not troubleshooting, maintaining, or upgrading shuttles and their equipment. As such, most of the workforce effects of low-speed automated shuttles may fall on these private entities. However, if automated shuttles eventually become an additional vehicle type in a transit agency’s fleet, the quantity and skills of agency-employed vehicle mechanics and technicians (as well as many other technology-supporting roles) may need to increase. Automated shuttles would also significantly change other roles and functions for people who work directly with shuttle services, including training, planning and scheduling, customer service, and street supervision and dispatching. 3.2.4 Implementation Horizon and Considerations With all the past and current activity on low-speed automated shuttles, this use case will likely have the shortest implementation timeline compared to other use cases. The FTA’s STAR Plan indicates that the automated shuttle demonstrations in the plan would be completed by the end of federal fiscal year 2020 (Machek et al. 2018). Of course, much has changed regarding low-speed automated shuttles since the STAR Plan was published. Whether these demonstrations and pilots have solidified the business case for further large-scale deploy- ment of the technology remains to be seen; however, interest in low-speed automated shuttles both globally and in the United States continues to expand (Coyner et al. 2021). The research team believes that low-speed automated shuttle technology will continue to develop, and deployments will likely become longer-term and more robust given the continued interest in these vehicles. Low-speed automated shuttles represent a particular market and service niche that has attracted substantial research and development investment from new manufacturers, as well as interest from local, regional, and state governments and transpor- tation agencies (Cregger et al. 2018). So far, low-speed automated shuttle pilots and demonstrations have often occurred in urban areas, college campuses, business parks, and other well-developed areas and on mostly fixed and short routes (Haque and Brakewood 2020; see also Coyner et al. 2021). Although rural areas could benefit from shuttle deployment due to aging and low-income populations, the research team found it unlikely that rural areas will have a significant deployment of low-speed automated shuttles due to For workforce effect estimates under the partial adoption scenario, the research team applied the low-speed automated shuttles use case only to fixed-route feeder or circulator services at small and large urban transit agencies.

28 The Impacts of Vehicle Automation on the Public Transportation Workforce the generally longer distances that those shuttles would have to travel (at low speeds) to accom- modate passenger travel needs. For the purposes of this study, the research team limited the application of low-speed automated shuttles to current fixed-route feeder or circulator-type service at small urban and large urban transit agencies. At least currently, most deployments focus on this type of service, and although overall a very flexible vehicle (once allowed on public roads), many low-speed automated shuttles cannot operate on public roadways without an NHTSA waiver due to their lack of compliance with the FMVSS. 3.2.5 Planning and Policy Decisions Several planning and policy decisions will need to be made that will have implications on how low-speed automated shuttles will affect the transit workforce, including: • Will transit agencies mainly use automated shuttles to replace current services, to create new services, or both? • Will automated shuttles likely be provided as turnkey operations in which private firms provide the shuttles, technology, and O&M staff? Or will transit agencies eventually take over parts or all the low-speed shuttle operations? • Will dedicated positions for human operators/attendants be needed, and if so, at what position-to-vehicle ratio? • Will shuttle maintenance, if performed in-house, be performed by a new classification of mechanics or by current mechanics after receiving training? 3.3 Use Case #3: Automated Bus Rapid Transit BRT provides qualities of a light rail system, such as speed and efficiency, because it often oper- ates in dedicated lanes that have permanent stations with elevated platforms (FTA 2015). Further efficiencies are created if offboard fare collection and traffic signal priority are also implemented. These features eliminate causes of delay that typically slow regular bus services (e.g., being stuck in other road traffic and collecting fares on board). Standard 40-foot coaches are used to provide BRT services due to their lower operating costs and quicker acceleration; however, 60-foot articulated units are sometimes deployed in high-demand corridors (VTA Transit 2007). HAV technology may be easier to deploy for BRT versus other transit use cases, primarily because BRT may operate in designated rights of way (Horadam 2018). Interactions with pedes- trians, bicyclists, and other vehicles can be controlled, though not completely eliminated. Also, automated BRT is an L4 use case, for which many of the required automated technologies are available now, including (Mudge and Lutin 2020): • Precision docking. • Automated lane keeping. • Automated bus platooning. • Automated collision avoidance and emergency braking. • Smooth acceleration and deceleration. For this reason, BRT is generally considered to be a potential early application of full-size automated transit vehicles. 3.3.1 Supporting Technology This use case will need the baseline technology architecture identified in Section 2.2, and for automated BRT to operate without a human on board, vehicles will need the additional automation-supporting technologies discussed in Section 2.3.

Transit Vehicle Automation Use Cases 29   In addition, other supporting technologies or technological approaches may be available to automated BRT in cases where the service operates on dedicated rights of way. The technol- ogy architecture could also include V2V and V2I connectivity for applications like platooning and to reduce vehicles’ reliance on cameras, GPS, and other sensors for exterior sensing and localization. 3.3.2 Current Deployments The research team identified two current pilots of automated BRT. A pilot of the Mercedes- Benz Future Bus was taking place in the Netherlands (Daimler 2019). The pilot was based on the autonomously driving Mercedes-Benz Actros truck with the Highway Pilot technology intro- duced in 2014. The technology has been customized for use in a city bus, with numerous added functions. The vehicle can recognize traffic lights, communicate with them, and safely negotiate junctions controlled by them. The vehicle can also recognize obstacles, especially pedestrians on the road, and brake autonomously. In Japan, several companies are involved in a self-driving bus test project on East Japan Railway Company’s (JR East) BRT lines (Intelligent Transport 2019). The project is evaluating self-driving technologies for bus transit applications, including automated lane-maintenance control, speed control, parking assist, and alternating passage tests on JR East’s BRT lines. In 2021, Robotic Research LLC and New Flyer Industries will be deploying full-size autono- mous buses on the new CTfastrak Hartford–New Britain BRT line in Connecticut with funding from FTA and the Connecticut Department of Transportation (Mudge and Lutin 2020), which will be the first automated BRT deployment in the United States. 3.3.3 Potential Operational Impacts and Workforce Effects Automated BRT may help improve the overall operations of current BRT lines. Automated buses could precisely align with elevated station platforms, improving accessibility for people with disabilities and the elderly and eliminating the need to deploy boarding ramps or lifts. V2I communications would enable signal prioritization to improve the consistency of headways. Automated BRT could scale up to as many vehicles as required by demand through vehicle platooning. An automated BRT system could support functions such as lane centering in nar- row lanes and precision docking at boarding platforms (Brewer et al. 2019; Gregg and Pessaro 2016). This use case could involve bus platooning to dynamically couple two or more buses dur- ing periods of demand surges. Operating with advanced safety technology and without a driver, automated platooned BRT could offer increased safety, reduced liability, reduced environmental impacts, increased service availability, and operational efficiency (Machek et al. 2018). With improvements in service availability and operational efficiency, there could be increased cus- tomer satisfaction. Automated BRT could offer an opportunity to lower the operating cost of BRT service by potentially reducing the labor costs associated with employing an operator (either through posi- tion reductions or through reduced wages). Transit agencies might be able to use the saved operating costs in expanding service capacity or increasing customer-oriented services. Automated BRT could also be seen as a lower-cost alternative to light rail projects, assuming that automated BRT combined with platooning can provide similar passenger-to-operator ratios to light rail, but with lower investment in right-of-way infrastructure (Mudge and Lutin 2020). Depending on the design of BRT route termini, buses may need to execute end-of-route maneuvers in mixed traffic or at multi-route transit stations. These maneuvers may be more

30 The Impacts of Vehicle Automation on the Public Transportation Workforce complex than operating in a dedicated guideway; however, if needed, onboard or remote opera- tors could take over driving functions. Automated BRT would impact not only bus operators but also transit jobs that work directly with BRT routes, including training, planning and scheduling, customer service, and street supervision and dispatching. Also, automated BRT will likely require increased maintenance and technician qualifications and may create a need for additional maintenance staff to keep all automation-supporting systems in full working order and to troubleshoot vehicles in service. 3.3.4 Implementation Horizon and Considerations The FTA’s STAR Plan (Machek et al. 2018) indicates an intent to demonstrate automated BRT during federal fiscal years 2021 and 2022. Given the current state of technology, the increased likelihood that BRT routes have a well-defined ODD (e.g., a dedicated right of way and controlled interactions with other road users), and the industry’s general interest in BRT, the research team believes that automated BRT will likely be one of the earliest use cases deployed to a substantial portion of existing BRT routes or used to create new high-capacity services. In general, BRT is considered a high-capacity transit service and, as such, is implemented in areas where passenger demand is high—where the investments associated with dedicated rights of way, offboard fare payment, and other BRT features could be amortized over higher volumes of passengers. It is hypothetically possible that automated BRT could change the cost-benefit analysis of implementing BRT, making automated BRT attractive for corridors that normally would not support conventional BRT. However, this research focused on the workforce effects associated only with replacing current conventional service with auto- mated transit service (see discussion in Section 2.6.2) and did not con- sider potential BRT service expansions. For the purposes of this research, the adoption of automated BRT was limited to the 11 urban transit agencies that had existing conventional BRT services in operation in 2018 NTD data. Although there was one rural transit agency with BRT service (i.e., the Roaring Fork Trans- portation Authority), the research team decided not to estimate the workforce effects of adopting the automated BRT use case at this transit agency because doing so would single out that agency in the results. 3.3.5 Planning and Policy Decisions Several planning and policy decisions will need to be made that will have implications on how automated BRT will impact the transit workforce, including: • Will dedicated positions for human operators/attendants be needed, and if so, at what position-to-vehicle ratio? • Will transit agencies explore the use of platooning in automated BRT services? • Will transit agencies mainly use automated BRT to replace current services, create new services, or both? 3.4 Use Case #4: Automated Mobility on Demand According to the USDOT (2016), MOD is a demand-responsive approach to mobility that leverages emerging types of services, such as ride-hailing, integrated transit networks and opera- tions, real-time data, connected travelers, and cooperative intelligent transportation systems, For workforce effect estimates under the partial adoption scenario, the research team applied the automated BRT use case to the 11 large urban transit agencies that currently operate BRT service (as of 2018).

Transit Vehicle Automation Use Cases 31   to allow for a more traveler-centric transportation system. Vehicles used are cars, vans, and shuttles. Traditional MOD is best known through the services provided by Uber and Lyft. Auto- mated MOD service requires either L4 or L5 automation, depending on whether the HAV has a limited ODD (e.g., a specific service area) or the HAV is capable of performing all driving functions under any and all roadway and environmental conditions (Brewer et al. 2019; Machek et al. 2018). One of the oft-touted benefits of automated MOD service is the opportunity to lower operat- ing costs for the fleet of vehicles due to reduced labor costs (Zmud and Reed 2019). This desired benefit would be the same for private and public operators but more important for the valua- tions of private-sector players. For example, Uber estimated it lost almost $1 billion in the first quarter of 2019 due to increased competitive pressures from Lyft in certain markets. Being able to lower costs by circumventing the need to recruit, hire, and pay drivers has been labeled an economic holy grail for the firm. Evidence that traditional MOD has had a negative impact on public transit usage is accumu- lating (Clewlow and Mishra 2017; Rayle et al. 2016; Schaller Consulting 2018; Sener et al. 2018). Thus, the imminent launch of privately operated automated MOD services is of interest to the transit industry because of the potential risk to transit ridership. Transit agencies that want to maintain a competitive edge over private mobility companies (or want to take advantage of other benefits of automation) may want to pursue the adoption of their own automated MOD service. However, because of the opportunity costs of testing and deploying automated MOD, few agencies, if any, have begun to test or deploy automated MOD on their own (Hughes- Cromwick and Dickens 2019). Automated MOD already has significant traction as a potential service to be provided by the private sector (see the current deployments in Section 3.4.2). Currently, private developers/ operators are split between traditional manufacturers of vehicle systems (such as automobile manufacturers and their tier 1 suppliers) and new entrants and start-ups from the technology industry (such as Waymo, Uber, and Aptiv). All private-sector players are seeking new revenue streams in light of the changing mobility market. Given the uncertain financial footing of privately operated automated MOD services, these services would most likely appear in densely populated urban areas, potentially impacting the transit ridership of large urban transit agencies but having little to no impact on small urban and rural agencies. If ridership losses materialize, funding streams related to ridership (e.g., fare revenue and some aspects of federal formula funds) may decrease, potentially leading to service reductions and job losses for bus operators. Although automated MOD could be operated by either private technology or auto manufac- turing firms [such as Uber and General Motors (GM)], this research focused on the impacts of automated MOD services operated by transit agencies themselves as part of their portfolio of transit services. Automated MOD services could be operated directly using traditional private contractors (e.g., First Transit and TransDev) or using the private firms already in the auto- mated MOD market (e.g., Waymo) but would be branded as the transit agency’s service in all cases. Key applications for transit-agency-operated automated MOD service include (a) auto- mated alternatives to ADA paratransit, (b) automated first/last mile, and (c) on-demand shared ride (Machek et al. 2018). 3.4.1 Supporting Technology This use case will need the baseline technology architecture identified in Section 2.2. More- over, for automated MOD to operate without a human on board, vehicles will need the addi- tional automation-supporting technologies discussed in Section 2.3.

32 The Impacts of Vehicle Automation on the Public Transportation Workforce Also, automated MOD may need to establish trip management platforms and/or call centers to help customers request trips and to provide customers with real-time information to effec- tively manage the on-demand fleets (these types of platforms are already in development from companies like Via and others). 3.4.2 Current Deployments Private-sector players are investing aggressively to be able to operate automated MOD ser- vices. In 2016, Uber began a pilot project on public roads in Pittsburgh, Pennsylvania. In 2017, GM, Waymo, and Lyft initiated pilots in San Francisco, California; Phoenix, Arizona; and Boston, Massachusetts, respectively. Waymo increased its driverless taxi operations in Phoenix in August and September of 2019, using Chrysler Pacifica minivans equipped with Waymo’s fourth-generation self-driving system (Bigelow 2019). In addition, in 2017, the San Francisco County Transportation Authority announced its plan to test self-driving shuttles on Treasure Island, California, in cooperation with EasyMile, by 2020 to provide first- and last-mile trans- port (Siu 2017). Most recently, Waymo launched another test of its robo-taxi technology in San Francisco in February 2021; the rides are limited to Waymo employees (Wiggers 2021). According to the FTA’s Transit Bus Automation Quarterly Update (FTA 2021), there are at least two public-agency-automated MOD projects completed or planned in the United States, including those at Valley Metro in Arizona and the city of Arlington in Texas. 3.4.3 Potential Operational Impacts and Workforce Effects A highly anticipated benefit of automated MOD would be improved service to people with disabilities or other mobility limitations and older adults. Paratransit service for these individuals is often inconvenient, unreliable, and expensive (Lutin 2018). When automated MOD service is adopted as an alternative for ADA paratransit, it is expected to provide dramatically improved mobility services. Although current ADA paratransit provides demand-responsive services, in most cases, customers must make advance reservations to use the service. A national survey conducted in 2017 revealed that more than one-third of respondents reported that scheduling the reservation for ADA paratransit was a barrier to using the service (Bezyak et al. 2017). By removing the reservation process, the automated MOD services operating as an alternative for ADA paratransit may allow people with disabilities to use the service whenever and wherever they need. Automated MOD services could allow transit agencies to quickly ramp up supply in the event of increased demand, without having to keep additional drivers on standby. Automated MOD paratransit services may not only improve mobility for people with dis- abilities but also increase transit agencies’ financial sustainability. According to FTA, the rider- ship of ADA paratransit service has continued to grow, increasing from 45 million in 2000 to 67 million in 2008 (Thatcher et al. 2013). (More recent data are not available because reduced and rural reporters to the NTD do not need to separate ADA paratransit trips from other demand- responsive trips.) The demand for paratransit service is expected to increase as the population above 65 years increases to 20 percent of the nation’s population by 2030 (Kaufman et al. 2016). For transit agencies, this increasing demand is a critical issue because of paratransit’s high oper- ating costs per passenger trip. When compared to other transit modes, the costs per unlinked passenger trip and passenger mile are highest for demand-response/paratransit services. Given the high operating cost per passenger of current ADA paratransit systems, HAVs could reduce paratransit labor costs while increasing operating efficiency (Machek et al. 2018). However, ser- vice for people with disabilities may still need a staff person or robotic assistant on board to assist passengers when boarding and alighting, provide help carrying luggage or packages, or secure a passenger’s wheelchair (Lutin 2018; Peirce et al. 2019). The capability to remotely monitor and

Transit Vehicle Automation Use Cases 33   communicate with the onboard passengers may also be necessary. All these needs may dampen the expected cost savings of automated MOD for paratransit. Another frequently mentioned application of automated MOD is to fill first-/last-mile gaps in travel to facilitate multimodality and provide door-to-door trips between a fixed-route transit stop and user-specified locations (Machek et al. 2018; Yap et al. 2016). First-/last-mile auto- mated MOD service can be operated as fully fixed routes or various forms of deviated routes with an expanded catchment area. Since lack of service at the first/last mile can decrease access to fixed-route service (Wang and Odoni 2016), this type of automated MOD service is likely to make public transit more attractive by providing more flexible transport options to riders (Scheltes and de Almeida Correia 2017). Potential cost savings of any automated MOD services depend on reducing labor costs, assuming additional technology costs for HAVs are modest (Machek et al. 2018; Peirce et al. 2019). To provide on-demand shared-ride service more efficiently, transit agencies may need to look for various types of service models. For instance, through partnerships with private companies, transit agencies could facilitate the deployment of shared automated MOD services. Automated MOD service operated by a transit agency may be seen in all types of service areas, including large and small urban as well as rural; however, the particular type of service (e.g., first/last mile or an alternative to ADA paratransit) deployed in each area may vary depending on local needs. At least in the near term, the most likely operational model for public transit agencies provid- ing automated MOD service will be through contracts with private automated MOD providers who will provide at least the vehicles and technology infrastructure needed to support and main- tain automated MOD service. As such, transit-agency-operated automated MOD will likely not significantly impact vehicle maintenance requirements and tasks (although private providers will need their qualified technicians). However, automated MOD may impact daily operations of ADA paratransit and other demand-response services. If operators or onboard attendants are no longer needed for at least a portion of demand-response trips, some operator positions may be eliminated. Other operator positions may be converted to an onboard attendant role, assisting riders who need additional help, but with attendants potentially earning lower wages (USDOT 2021). Alternatively, humans could provide on-call assistance at the end points of passenger trips, meeting HAVs at the origins and destinations for passengers that require additional assistance. (However, the cost-effectiveness of this service model is unknown.) Automated MOD service would also significantly change other roles and functions in the demand-response area of transit agencies, including operator/attendant training, trip reserva- tions and scheduling, customer service, and street supervision and dispatching. 3.4.4 Implementation Horizon and Considerations In early 2019, GM announced that it would make automated MOD service available in several big cities in 2019/2020 through its AV division cruise automation (Hughes-Cromwick and Dickens 2019). However, in July 2019, GM removed the timeline from its AV plans, noting that the development may take a while and needs to be done right to allow for rapid scaling when ready (Sakelaris 2019). In a similar message, Waymo’s CEO, John Krafcik, said that the transition to fully autonomous vehicles will take a long time (Abuelsamid 2018). To the research team’s knowl- edge, all pilot projects that have been implemented across the United States have had a safety driver on board to take control of the vehicle if necessary (see also Stocker and Shaheen 2019). Most original timeline predictions by private companies have proven too aggressive; there- fore, the potential timeline for rollout of privately operated autonomous MOD service is highly uncertain.

34 The Impacts of Vehicle Automation on the Public Transportation Workforce While automated MOD development is proceeding at a rapid pace in the technology and auto sectors, public transit agencies are just beginning to explore this technology. A potential strategy would be for transit agencies to engage in a variety of potential partnerships with private-sector firms. For example, in 2018, Waymo and Valley Metro in Phoenix initiated a partnership to use Waymo vehicles to provide first-/last-mile travel to the agency’s light rail system. The timeline for the adoption of automated MOD depends to a large degree on whether transit agencies seek partnerships or attempt to initiate such services on their own. The FTA’s STAR Plan indicates that the demonstration research projects on automated MOD for ADA paratransit would be finalized by federal fiscal year 2022 (Machek et al. 2018). Because automated ADA paratransit service would focus on providing rides to people with disabilities, it may require specialized equipment or other design features for vehicles that are not yet fully developed. However, if a human operator is required to help passengers, transit agencies may need to develop new training programs for their employees to prepare for a role change with a special focus on customer service. Because automated MOD services specifically for people with dis- abilities have their own set of additional complexities (beyond auto- mated driving), and because ADA paratransit service has very specific requirements for passenger assistance and passengers may often have unique needs that would be difficult to be attended to by an HAV, the research team found that automated MOD service for non-ADA para- transit demand-responsive service would be the most likely near-term adoption of this use case. Given that demand-responsive service is one of the most common modes across both rural and urban areas, the research team applied this use case to all transit agencies that offered some form of demand-responsive service. 3.4.5 Planning and Policy Decisions Several planning and policy decisions will need to be made that will have implications on how automated MOD will affect the transit workforce, including: • Will automated MOD likely be provided as turnkey operations in which private firms provide the vehicles, technology, and O&M staff? Or will transit agencies eventually take over parts or all of automated MOD operations? • Will dedicated positions for human operators/attendants be needed, and if so, at what position-to-vehicle ratio? • Will the reduced time associated with demand-response scheduling and dispatching of auto- mated MOD service be reallocated to new tasks, allowing for more time to perform current manual tasks, or will the number of hours needed for scheduling and dispatching activities be reduced (potentially reducing staffing requirements)? • Will transit agencies mainly use automated MOD to replace current services, create new services, or both? 3.5 Use Case #5: Automated Local Bus Service Local bus service is the regular operation of transit buses along a route, stopping at fixed bus stops according to a published timetable. A primary challenge of local bus service is that timetables can be difficult to maintain in the event of traffic congestion, breakdowns, on- or off-bus incidents, road blockages, bad weather, or fluctuations in passenger loads. For workforce effect estimates under the partial adoption scenario, the research team applied the automated MOD use case to all transit agencies that provided demand-responsive service but applied the use case only to non-ADA paratransit demand-responsive service.

Transit Vehicle Automation Use Cases 35   Automating local bus service could provide benefits in terms of more reliable bus opera- tions, improved safety, and lower labor costs. Assuming labor costs are reduced, transit agencies with local bus service could operate more service (e.g., more frequency on cur- rent routes) at a lower operating cost—improving service quality and potentially increasing ridership. Automating full-sized, full-speed buses to travel under all roadway and environmental conditions will be challenging. In May 2019, the Automated Bus Consortium was formed (Green Car Congress 2019). The consortium’s founding members include the Dallas Area Rapid Transit; Foothill Transit; Long Beach Transit; Los Angeles County Metropolitan Transportation Authority; MetroLINK (Moline, Illinois); Metropolitan Atlanta Rapid Tran- sit Authority; Michigan Department of Transportation/Michigan’s mobility initiative, PlanetM; Minnesota Department of Transportation/Rochester Public Transit; Pinellas Suncoast Transit Authority; and Virginia Department of Rail and Public Transportation/Hampton Roads Transit. (Members of the consortium may vary over time.) The collaboration was designed to investigate the feasibility of implementing pilot automated bus projects across the United States. 3.5.1 Supporting Technology Automated bus demonstrations have relied on various technologies, including systems with predefined fixed points, either physically embedded in the infrastructure (e.g., magnets or RFID tags) or marked digitally with GPS and geospatial maps. Recent automated bus demonstrations and developments have tended to use the baseline technology architecture discussed in Section 2.2. To provide fully automated local bus transit (i.e., without an opera- tor or attendant on board), buses would also need the additional supporting technologies discussed in Section 2.3. 3.5.2 Current Deployments In the United States, automated transit bus demonstrations have used predefined routes and lanes with known geometries and limited exposure to other vehicles. These environments include bus-only road shoulders, dedicated bus lanes, high-occupancy vehicle (HOV) lanes, and transit stations with boarding platforms (Nasser et al. 2018). There are other automated local bus pilots and demonstrations planned; however, most are very limited in their scope and are nowhere near the capabilities required for the full-speed, multi-corridor operations that would be needed to have truly automated local bus service (FTA 2021). Volvo Buses and Nanyang Technological University (NTU) in Singapore have demonstrated the world’s first 40-foot autonomous electric bus. The Volvo Bus began trials on the NTU campus. There need to be more demonstrations that evaluate how automated buses operate on routes with mixed traffic, pedestrians, bicyclists, vehicles, and other obstacles, and how well auto- mated buses can navigate a wider range of slowly changing road surface conditions (e.g., crowns, potholes, speed bumps) and road articles (e.g., signs and trees). 3.5.3 Potential Operational Impacts and Workforce Effects Transit agencies are looking to bus transit automation to reduce overall operating costs and increase safety. Although there is still a large degree of uncertainty about when and how automated local bus transit will be implemented, the use case could be applied to a wide variety of bus transit services, as demonstrated by the Automated Bus Consortium’s (2019) candidate pilot projects. In addition to two BRT routes and two projects focused on bus

36 The Impacts of Vehicle Automation on the Public Transportation Workforce automation for yard operations, the pilot projects include other types of traditional fixed- route bus transit services: • Urban shuttle-type service operating between an airport and a light rail station. • Urban shuttle-type services that connect university campuses to a park-and-ride lot, an Amtrak® station, or student housing. • Rural shuttle-type service that connects key residential and business areas to a manufacturing plant. • Express bus service utilizing freeway high-occupancy toll (HOT)/HOV lanes. • Mixed-freeway and local road service connecting a major employer/office park with an urban downtown. • Local bus service operating on urban arterials. Although these pilot projects are not yet in operation, the wide variety of applications for automated local bus transit demonstrates the very broad potential for full-size automated buses to operate across service types (e.g., shuttle, local, or express), street/roadway types (e.g., local roads, urban arterials, mixed-flow freeways, or HOV/HOT lanes), and land-use patterns (e.g., rural areas, suburban areas, urban areas, and airports). Especially in the beginning of automated local bus deployment and testing, the same chal- lenges faced by traditional HAVs will be faced by automated local buses, and initial deployments are more likely to be limited to simpler applications with more dedicated right of way, fewer left turns, and less interactions with other road users. Expected impacts from automating local bus service are related to reducing operating costs, improving safety, and improving headways and time point adherence. Local bus automa- tion may reduce labor expenses if a human operator is not required; however, the likelihood of removing a human operator is highly uncertain (USDOT 2021). If some cost savings are achieved through this use case, transit agencies could increase service levels, helping to attract ridership. On the other hand, if transit agencies keep humans on board, automated local bus transit can eliminate the stress of driving for long hours, potentially increasing operational safety and driver retention—with reduced operating cost advantages. Also, automated local buses will likely require increased maintenance and technician qual- ifications and may create a need for additional maintenance staff to keep all automation- supporting systems in full working order and to troubleshoot vehicles in service. Automated local bus transit service would also significantly change other roles and functions in the transit agency, including operator/attendant training, fare collection, planning and sched- uling, customer service, and street supervision and dispatching. It is likely that buses capable of full automation in revenue service will also be capable of automation while in non-revenue service, including during yard movements, further impacting transit staff who perform service- readying tasks, like cleaners and fuelers. 3.5.4 Implementation Horizon and Considerations The timeline for the full implementation of the L5 use case is more long-term than near- term for many reasons. The operator for any automated local bus service would most likely be a transit agency, which is often risk-averse (USDOT 2021). There is a much higher level of local bus service at urban transit agencies than at rural transit agencies, although some rural agencies do operate a local bus service. Due to the likely higher cost to purchase an automated bus and the uncertainty of whether an automated local bus service would produce cost savings, the research team assumed that larger transit agencies would be the most likely to implement this use case. Larger agencies often have dedicated sources of funding and more flexibility to

Transit Vehicle Automation Use Cases 37   experiment with automated transit services than rural transit agencies, and per-route ridership at larger transit agencies is typically higher, allowing the costs of implementing automated local bus service to be amortized across more passengers. Also, the research team assumed that automated local bus service could be used to replace any current non-BRT services, including the MB, TB, and CB modes. 3.5.5 Planning and Policy Decisions Several key planning and policy decisions will need to be made that will have implications on how automated local bus transit will affect the transit workforce, including: • Will dedicated positions for human operators/attendants be needed, and if so, at what position-to-vehicle ratio? • Will transit agencies mainly use automated local bus transit to replace current services, create new services, or both? For workforce effect estimates under the partial adoption scenario, the research team applied the automated local bus transit use case to all transit agencies that provided fixed-route bus service (including the motor bus, trolley bus, and commuter bus modes) but assumed that larger agencies would be more likely to implement this use case.