Learning While Repositioning in On-Demand Vehicle Sharing Networks

Abstract

We consider the vehicle repositioning problem for a vehicle sharing service with a fixed number of vehicles distributed across multiple locations in the network. The vehicle sharing service is on-demand and one-way, which means that customers can pick up one vehicle from any location without prior reservation and drop off the vehicle at any location in the network after the usage. Due to uncertainty in both customer arrivals and vehicle returns, the service provider needs to periodically reposition the vehicles to match the supply with the demand while minimizing the total costs of repositioning labor and lost sales. The repositioning problem is critical in the successful management of on-demand one-way vehicle sharing services, and it is challenging both analytically and computationally. The optimal repositioning policy under a general n-location network is intractable without knowing the optimal value function. For n-location networks and censored demands, we develop an online learning method to dynamically reposition vehicles without knowing the demand distribution a priori. We study the performance of our algorithm by comparing it with the best base-stock policy. The best base-stock policy is a generalization of the popular inventory control policy to the vehicle repositioning problem, and we establish its asymptotic optimality in two different limiting regimes. We prove that the T-period aggregate regret of our algorithm is bounded sublinearly by O(T^{\frac{n}{ n+1}}(\log T)^{\frac{1}{ n+1}}) for all n. We illustrate our learning algorithm by applying it to the Car2Go dataset.

Keywords: censored demand, vehicle sharing, vehicle repositioning, online learning, regret analysis