Highlights • Shared ride services with autonomous vehicles result in lower user waiting times. • Sharing also results in lower operational costs for operators than no sharing. • Trade-offs examined from perspective of operators, users and policymakers. • Benefits accrue only when demand pool is sufficiently large. • Role for public sector to incentivize shared ride offering and use.

Abstract This paper presents a quantitative analysis of the operations of shared-ride automated mobility-on-demand services (SRAMODS). The study identifies (i) operational benefits of SRAMODS including improved service quality and/or lower operational costs relative to automated mobility-on-demand services (AMODS) without shared rides; and (ii) challenges associated with operating SRAMODS. The study employs an agent-based stochastic dynamic simulation framework to model the operational problems of AMODS. The agents include automated vehicles (AVs), on-demand user requests, and a central AV fleet controller that can dynamically change the plans (i.e. routes and AV-user assignments) of AVs in real-time using optimization-based control policies. The agent-based simulation tool and AV fleet control policies are used to test the operational performance of AMODS under a variety of scenarios. The first set of scenarios vary user demand and a parameter constraining the maximum user detour distance. Results indicate that even with a small maximum user detour distance parameter value, allowing shared rides significantly improves the operational efficiency of the AV fleet, where the efficiency gains stem from economies of demand density and network effects. The second set of scenarios vary the mean and coefficient of variation of the curbside pickup time parameter; i.e. how long an AV must wait curbside at a user’s pickup location before the user gets inside the AV. Results indicate that increases in mean curbside pickup time significantly degrade operational performance in terms of user in-vehicle travel time and user wait time. The study quantifies the total system (user plus fleet controller) cost as a function of mean curbside pickup time. Finally, the paper provides an extensive discussion of the implications of the quantitative analysis for public-sector transportation planners and policy-makers as well as for mobility service providers.