Shared autonomous vehicle system designs for major metro areas : an examination of geofencing, ride-sharing, stop-location, and drivetrain decisions




Gurumurthy, Krishna Murthy

Journal Title

Journal ISSN

Volume Title



With autonomous vehicles (AVs) in the testing phase, researchers and planners must resort to simulation techniques to explore possible futures regarding shared and automated mobility. An agent-based discrete-event transport simulator, POLARIS, is used throughout this dissertation to simulate passenger and freight travel in regions with a shared AV (SAV) mobility option. Using this framework, three broad areas of SAV design are explored – geographical restrictions via geofences, operational modifications via pickup-dropoff (PUDO) stops, and the use of all-electric drivetrains – using the case studies of Bloomington and Chicago in Illinois.

Constraining SAVs within pre-defined geofences indicate that empty vehicle-miles-traveled (eVMT) can be curtailed when service areas in Chicago have a balanced mix of trip generators and attractors. Geofences also help lower spatial response times, system-wide VMT across all modes, and ensure uniform access to SAVs. Dynamic ride-sharing (DRS) is useful in lowering VMT and percent eVMT that arises from sprawling when demand is low, whereas savings from DRS is highest when all demand within smaller regions are served by SAVs. Various PUDO spacings and trip-demand densities with fixed mode shares in Bloomington reveal that greater PUDO spacing or distances between stops and higher levels of SAV use or trip demand increases AVO by up to 11% per 4-seater SAV, on average. Aggregation also decreases SAV VMT (by up to 20%) compared to door-to-door SAV fleet operations without DRS or PUDOs. Higher SAV mode shares in a forecasted scenario for Chicago outline total VMT savings of about 2 to 3%, even though SAV VMT increases from serving more trips. A quarter-mile PUDO spacing is recommended in downtown regions to keep walking trips short, but further analysis on congestion spillback is warranted.

A variety of electric SAV (SAEV) fleet designs and charging strategies show that a mixed fleet of short (100-mi) and long (250-mi) range SAEVs performs better than a homogenous short-range fleet, with lower empty vehicle miles traveled (eVMT) and lower idling time. Charging and service priority policies are both required, but at different times of the day to accommodate slow Level 2 chargers, but is not as important for DCFC charging stations. SAEVs can stay in place longer (1 hr versus 15 min) to keep eVMT low, but only if long-range SAEVs are in the fleet and the region is small. SAEVs in large regions like Chicago are exposed to location-specific trip requests when idling in place, and need to have high average state of charge (SoC) across the fleet to serve all incoming requests. Smart siting of EVCS and availability of fast chargers remain key to minimizing fleet size and keeping response times low.


LCSH Subject Headings