Ride-hailing markets with mixed autonomy: how will path choice modes of idle vehicles affect repositioning schemes?

In recent years, several companies have attempted to apply autonomous vehicles to on-demand ride-hailing services. The on-demand ride-hailing market is expected to present a scenario of shared autonomous vehicles (SAVs) mixed with human-driven vehicles (HVs) soon. Operational differences between SAVs and HVs present significant impacts on road traffic and raise challenges for ride-hailing platforms in terms of scheduling and managing SAVs. To derive the optimal repositioning scheme of SAVs, we propose a bi-level model considering interactions between road traffic operation and ride-hailing market conditions in the mixed-fleet scenario. A profit-maximization model is established at the upper level for the SAV repositioning schemes under market equilibrium conditions. With the mixed-fleet size and market indicators derived from the upper level, we develop network equilibrium modes at the lower level in which idle SAVs follow the system optimality (SO) and the rest follow the user equilibrium (UE). These models differentiate not only the types of vehicles, but also their occupancy status to guarantee passengers’ experience. By solving the complex bi-level model with a tailored algorithm, we show the advantages of idle SAVs following the SO principle in terms of market operations and user experience. Our results demonstrate that by following specific SAV management strategies, the platform can achieve a balance among system efficiency, passenger experience, and market fairness. This provides new insights into the scheduling of SAVs.