Pattern transfer learning for reinforcement learning in order dispatching

Order dispatch is one of the central problems to ridesharing platforms. Recently, value-based reinforcement learning algorithms have shown promising performance to solve this task. However, in real-world applications, the demand-supply system is typically nonstationary over time, posing challenges to reutilizing data generated in different time periods to learn the value function. In this work, motivated by the fact that the relative relationship between the values of some states is largely stable across various environments, we propose a pattern transfer learning framework for value-based reinforcement learning in the order dispatch problem. Our method efficiently captures the value patterns by incorporating a concordance penalty. The superior performance of the proposed method is supported by experiments.