Spatio-Temporal Hierarchical Adaptive Dispatching for Ridesharing Systems

Abstract

Nowadays, ridesharing has become one of the most popular services offered by online ride-hailing platforms (e.g., Uber and Didi Chuxing). Existing ridesharing platforms adopt the strategy that dispatches orders over the entire city at a uniform time interval. However, the uneven spatio-temporal order distributions in realworld ridesharing systems indicate that such an approach is suboptimal in practice. Thus, in this paper, we exploit adaptive dispatching intervals to boost the platform's profit under a guarantee of the maximum passenger waiting time. Specifically, we propose a hierarchical approach, which generates clusters of geographical areas suitable to share the same dispatching intervals, and then makes online decisions of selecting the appropriate time instances for order dispatch within each spatial cluster. Technically, we prove the impossibility of designing constant-competitive-ratio algorithms for the online adaptive interval problem, and propose online algorithms under partial or even zero future order knowledge that significantly improve the platform's profit over existing approaches. We conduct extensive experiments with a large-scale ridesharing order dataset, which contains all of the over 3.5 million ridesharing orders in Beijing, China, received by Didi Chuxing from October 1st to October 31st, 2018. The experimental results demonstrate that our proposed algorithms outperform existing approaches.

A Dynamic Matching Time Strategy Based on Multi-Agent Reinforcement Learning in Ride-Hailing

“…Multi-Region UNIFORM (MRUNIFORM). The UNIF-ORM algorithm [16] is a commonly used comparison algorithm, which will do the matching for every n time slots. We modify the UNIFORM algorithm to increase its dynamic matching property, with a half probability of matching at the current time slot and a half probability of not making a match, which is called the MRUNIFORM.…”