Title: Reinforcement learning-based dynamic pricing research on ride-hailing platforms

Authors: Yue Bai; Wei Zhang

Addresses: School of Traffic and Transportation, Lanzhou Jiaotong University, Lanzhou, 730070, China ' School of Traffic and Transportation, Lanzhou Jiaotong University, Lanzhou, 730070, China

Abstract: Due to the significant spatiotemporal dynamics and the complex supply-demand relationship in ride-hailing operations, this study proposes a multi-module dynamic pricing model for ride-hailing (MDPM-RH) platforms, which is capable of real-time sensing of regional supply-demand fluctuations and constructing differentiated spatiotemporal price coefficients to enable dynamic pricing. The model considers the vacant vehicle dispatching and driver-passenger matching problems, and corresponding sub models were constructed separately. The dynamic pricing problem is formulated as a Markov decision process (MDP), and solved using the soft actor-critic (SAC) algorithm to maximise long-term profits. The experiments based on real ride-hailing order data in Haikou City show that the dynamic pricing method proposed in this study can effectively improve the platform profit. Compared with static pricing, the dynamic pricing strategy achieved a weekly profit improvement of ¥129,751.16, at the same time, the order response rate increased by 11.88%, and the driver's average time to take orders decreased by 0.64 min.

Keywords: dynamic pricing; ride-hailing; reinforcement learning; SAC; soft actor-critic.

DOI: 10.1504/IJVSMT.2026.155795

International Journal of Vehicle Systems Modelling and Testing, 2026 Vol.20 No.3, pp.265 - 290

Received: 29 May 2025
Accepted: 24 Sep 2025
Published online: 14 Aug 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article