Timetable Rescheduling for High-Speed Trains Based on a Multi-Agent Game Reinforcement Learning Method

As the high-speed rail network grows, unexpected disruptions affecting train operations have become increasingly common. Consequently, efficiently adjusting train schedules to minimize delays has become a critical challenge in high-speed rail timetable rescheduling (TTR). To address this issue, this paper proposes a Multi-Agent Game Deep Reinforcement Learning (MAGDRL) approach to tackle the TTR problem during a complete blockage. Initially, the paper defines the MAGDRL state and action spaces, incorporating spatiotemporal distribution information, and develops an effective reward feedback system. Next, the Nash-Q learning algorithm ensures convergence to the Nash equilibrium strategy. Additionally, an approximate Nash equilibrium solving method is applied to enhance the computational efficiency of determining the Nash equilibrium strategy. Finally, simulation results based on the Beijing-Shanghai high-speed railway demonstrate that the proposed method effectively handles timetable rescheduling across multiple disruption scenarios involving complete blockages of segments.