Order-Dispatching Strategy Induced by Optimal Transport Plan for an Online Ride-Hailing System

Abstract

Order-dispatching strategy is fundamental for online ride-hailing systems in three major ways. First, it should be capable of effectively matching hundreds of orders to thousands of vehicles in less than 1 min, whereas both orders and vehicles are geographically widely distributed. Second, it should decide how to fairly allocate profit among participators. Third, it should unify interests of present and future. Conflicts always exist within the three objectives. We observe that the system is dynamically balanced during fine-tuned time intervals. Based on this observation, we propose a novel dispatching strategy capable of juggling the three objectives. By applying optimal transport theory, we demonstrate that, under appropriate presumptions, this dispatching strategy is ruled by the optimal transport plan realizing the Wasserstein distance between distributions of orders and vehicles. Furthermore, we develop a kit of methodological and algorithmic tools which has substantial and sensible advantages in characterizing and optimizing the system.