Real-Time Dispatching with Local Search Improvement for Dynamic Ride-Sharing

Abstract

Dynamic ride-sharing services such as UberPool or MOIA are becoming increasingly popular as they offer a cheap and flexible mode of transportation and reduce traffic compared to traditional taxi and ride-hailing services. One key optimization problem when operating ride-sharing services is the assignment of trip requests to vehicles to maximize the service rate while minimizing operational costs. In this work, we propose a real-time dispatching algorithm capable of quickly processing incoming trip requests. This dispatching algorithm is combined with a local search that aims to improve the current routing plan. Both algorithms are embedded into a planning and simulation framework for dynamic ride-sharing and evaluated through simulation studies on real-world datasets from Hamburg, New York City, and Chengdu. The results show that the local search improvement phase can improve the request acceptance rate as well as vehicle travel times. We achieve an average reduction of the request rejection rate by 1.62% points and a decrease in vehicle travel time per served request of 6.5%. We also study the influence of pre-booked rides and show that the local search yields even larger benefits when part of the trip requests are known in advance.