Research on Delay Adjustment Strategy for Metro Trains Based on Expected Sarsa Algorithm

Train delays have become increasingly common in metro systems; this paper proposes a novel method based on the Expectation Sarsa algorithm to deal with train timetable rescheduling. In the algorithm, the state space is significantly reduced, leading to improved computation speed and solution quality. The algorithm also combines the strengths of Sarsa and qualification tracking, exhibiting high learning capability and broad applicability. The integration of qualification tracking effectively reduces the steps required to identify the optimal strategy, enhancing learning efficiency. Finally, a case study is conducted on Shenzhen Metro Line 14. The results demonstrate the convergence of the algorithm, its rapid calculation speed, and real-time adjustment capabilities. In comparison to alternative approaches such as no adjustment, manual adjustment, and first-in-first-out (FIFO), the proposed method notably reduces the average total train delay while ensuring more uniform train spacing. Moreover, when compared to the traditional Sarsa algorithm, the optimization algorithm presented in this paper exhibits faster convergence within a short timeframe.