Ordis: A Dynamic Order-Dispatch Algorithm for Ridehailing and Ridesharing in a Large Region

Abstract

Online ridesourcing framework is often designed with a goal to achieve optimized matching between driver and passenger for not only minimizing passenger’s travel time but also maximizing driver’s revenue incomes. However, traditional order dispatching that materializes this matching is typically conducted on per pair of passenger and driver basis in a relatively small area, which often lead to suboptimal global matching rate. In this paper, we model the order dispatching as a distributed matching problem for a wide area based on a developed multi-queue model, each queue in charge of a region service. Based on this model, we propose a dispatching algorithm, called Ordis, by applying the network-flow theory to the multi-queues with an attempt to maximize the overall service revenue while improving the service efficiency and user experience for both ridehailing and ridesharing across multiple regions. To validate the effectiveness of the proposed approach, we developed an online taxi-hailing simulation system and compared it with some commonly used dispatch algorithms based on real trace data. We found that the proposed method is superior to those compared methods.