An optimized driver repositioning strategy in ridesplitting with earning estimates: a two-layer dynamic model and control

Abstract

Positioning of ride-sourcing drivers may improve vacant travel times, waiting times and matching opportunities. Herein, we develop a ride-sourcing fare optimizer which, in parallel with a revenue estimator, aids drivers’ repositioning decisions to minimize customer abandonments in a system offering ride-hailing (solo) and ridesplitting (shared) rides. A Markov chain estimates near-future individual revenues and forms a second layer in a MFD-based model to predict system conditions. We applied the proposed model in a simulation of the central business district of Shenzhen, China. Our results show that repositioning with fare control can decrease the number of abandonments by 98%. In the other hand, controlling fares decreased travelling speeds in the busiest periods of each area of the system, without entering a hyper-congested regime. These findings expand the literature on fare optimization including ridesplitting operations and providing a tool to estimate near-future earnings.