Relocation incentives for ride-sourcing drivers with path-oriented revenue forecasting based on a Markov Chain model

Abstract

Proper positioning of ride-sourcing drivers may improve vacant travel times, waiting times, and matching opportunities. Herein, we evaluate the potential repositioning response of drivers when provided guidance based on estimates of their earnings in a system offering ride-hailing (solo) and ridesplitting (shared) rides. Therefore, we develop a strategy that enumerates the best regional repositioning destination based on the expected number of requests a driver will encounter during the forecast horizon. A mixed continuous-discrete time Markov Chain (MDCTMC) is developed to predict a driver’s activities and the associated revenues. In summary, the developed strategy provides a group of drivers with individualized near-future revenue estimates guiding drivers toward repositioning decisions that are more likely to maximize their earnings. Our main findings indicate that if the operator selects only a fraction of active drivers to provide guidance, these are likely to expect higher earnings than those without guidance. We also show that it manages to decrease the number of unserved requests compared to several state-of-art benchmarks while increasing vehicle occupancy and decreasing the deadheading.