Using shared automated vehicles to deliver public transport

Shared automated vehicles (SAVs) will affect mode choice in several ways: door-to-door (D2D) services may be the most popular choice for SAV travelers, thanks to relatively low travel times and access costs, as well as ride-sharing opportunities to split costs. First-mile last-mile (FMLM) SAV services that connect travelers to transit stations (including airports) make use of existing transit investments while avoiding parking at or walking to and from those stations. Larger (and mid-size) SAVs may also replace traditional buses by adding frequency and lowering service costs on popular bus routes. Motivated by SAVs’ many potential benefits, this dissertation specifies traveler choices for different US settings, simulates fleets of SAVs in concert with other modes, and analyzes results for travelers, their networks, and their regions. This dissertation first micro-simulates operations of an SAV-based transit route to understand how traffic conditions and all travelers’ costs (those in SAVs and in thousands of private vehicles) are affected by vehicle sizes, use of pull-out bays, and presence of signals along a 6.5-mile corridor. Using SUMO (Simulation of Urban MObility) software, results show that use of smaller, but more frequent SAVs lowers passenger wait times but can increase travel times for background travelers. Use of smaller SAVs does not significantly affect background traffic conditions due to their shorter dwell times (during passenger pickup and dropoff). Replacing 40-seat bus-type SAVs with 5- and 10-seat SAVs reduced total corridor transportation costs by 4% (reflecting all travelers’ values of time, plus vehicle operating and ownership costs), under various demand levels and SAV seat-occupancy factors. The next microsimulation assigned and tracked SAVs delivering FMLM service to 5 commuter rail stations in central Austin neighborhoods. Thanks to SUMO, results include traveler mode choices, walk distances and routes, wait times, vehicle idling and bunching details, vis-à-vis train arrival times. With rail service headways of 15 minutes, simulations predict dramatic increases in train use (by roughly a factor of 10, at those stations). Variations in train headways and SAV fleet sizes illustrate how door-to-door travel remains the key predictor of travelers’ mode choices. 4-seat SAVs perform similarly to 6-seat SAVs but cost less to provide. A dynamic ride-sharing (DRS) (vehicle-to-passenger assignment) algorithm tightly coordinated with train arrivals delivers 87% of travelers to their stations in time to catch the next train, while uncoordinated assignments deliver just 57% of travelers in time. Finally, the POLARIS mesoscopic simulation is leveraged to integrate three SAV services across the 20-county Chicago region. When SAVs with DRS serve only D2D trips, at just $0.50 per mile, with 1 SAV for every 40 residents, they attract 15% of trips (and ownership is predicted to fall from 0.66 to 0.37 household vehicles per capita), with a 15-minute average travel time and 4.6-mile average person-trip distance. Adding FMLM service (to about 54,000 train and bus stops) increases Chicagoans’ transit mode split from 5.4% to 6.3%, with the same SAV fleet serving 12% more person-trip requests per day and driving 4.2% more SAV-miles traveled). Most FMLM person-trip distances are under 2 miles, with rail-station connections dominating (rather than those to bus stations). In terms of social welfare impacts (using normalized differences in logsums of mode and destination choice utilities), suburban area residents appear to benefit most from SAVs’ arrival, followed by those in urban neighborhoods, especially for D2D trip-making. Overall, US communities and transit systems seem poised to benefit from SAV services.