Satisficing Approach to On-Demand Ride Matching

Online ride-hailing platforms have developed into an integral part of the transportation infrastructure in many countries. The primary task of a ride-hailing platform is to match trip requests to drivers in real time. Although both passengers and drivers prefer a prompt pickup to initiate the trips, it is often difficult to find a nearby driver for every passenger. If the driver is far from the pickup point, the passenger may cancel the trip while the driver is heading toward the pickup point. For the platform to be profitable, the trip cancellation rate must be maintained at a low level. We propose a computationally efficient data-driven approach to ride matching, in which a pickup time target is imposed on each trip request and an optimization problem is formulated to maximize the joint probability of all the pickup times meeting the targets. By adjusting pickup time targets individually, this approach may assign more high-value trip requests to nearby drivers, thus boosting the platform’s revenue while maintaining a low cancellation rate. In numerical experiments, the proposed approach outperforms several ride-matching policies used in practice.