Fleet Repositioning for Vehicle Sharing Systems: Asymptotic Optimality of the Balanced Myopic Policy

54 Pages Posted: 4 Feb 2021 Last revised: 25 Jan 2024

See all articles by Yihang Yang

Yihang Yang

The Laboratory for AI-Powered Financial Technologies Limited; City University of Hong Kong

Yimin Yu

City University of Hong Kong - College of Business

Qian Wang

City University of Hong Kong (CityU)

Junming Liu

City University of Hong Kong (CityU) - Department of Information Systems

Date Written: January 9, 2021

Abstract

We investigate the fleet repositioning problem aimed at dynamically optimizing vehicle distributions to maximize long-run average social welfare in a free-floating vehicle-sharing system.
We model the problem as a Markov decision process under the ex ante committed decision scheme. Surprisingly, we show the optimality of a balanced myopic policy for the average reward setting, which maintains the vehicle distribution at a level that best matches the vehicle supply and trip demand. The balanced myopic policy can overcome the curse of dimensionality and significantly improve computational efficiency. However, it uses less information than the benchmark under the ex post decision scheme, which might result in potential performance loss. To evaluate such performance loss, we construct the balanced myopic lower bound and upper bound for the optimal average rewards under both decision schemes, which yield the max performance gap
between them. For a vehicle sharing system with size θ, we demonstrate that the magnitude of relative performance loss is O(1/(√θ)), i.e., the balanced myopic policy is asymptotically optimal under the regime of large system size. We also consider the seasonal demand setting and show that a generalized balanced myopic policy is optimal. Our results shed light on how to design effective heuristics for improving the efficiency of vehicle sharing systems.
Our results suggest a simple and effective solution procedure for fleet repositioning with performance guarantee. We quantify the operational value and benchmark it against myopic solutions through numerical experiments and a counterfactual case study of a real-world vehicle-sharing system with 5 regions.

Keywords: vehicle sharing system; fleet repositioning; Markov decision process; balanced myopic policy

JEL Classification: R41

Suggested Citation

Yang, Yihang and Yu, Yimin and Wang, Qian and Liu, Junming, Fleet Repositioning for Vehicle Sharing Systems: Asymptotic Optimality of the Balanced Myopic Policy (January 9, 2021). Available at SSRN: https://ssrn.com/abstract=3763049 or http://dx.doi.org/10.2139/ssrn.3763049

Yihang Yang

The Laboratory for AI-Powered Financial Technologies Limited ( email )

Units 1101-1102 & 1121-1123, Building 19W
Science Park West Avenue, Hong Kong Science Park
Hong Kong
Hong Kong

City University of Hong Kong ( email )

Units 1101-1102 & 1121-1123, Building 19W
Science Park West Avenue, Hong Kong Science Park
Hong Kong
Hong Kong

Yimin Yu (Contact Author)

City University of Hong Kong - College of Business ( email )

83 Tat Chee Avenue
Kowloon
Hong Kong

Qian Wang

City University of Hong Kong (CityU) ( email )

83 Tat Chee Avenue
Kowloon
Hong Kong

Junming Liu

City University of Hong Kong (CityU) - Department of Information Systems ( email )

83 Tat Chee Avenue
Kowloon
Hong Kong

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
211
Abstract Views
1,171
Rank
269,585
PlumX Metrics