Optimization of train real-time dynamic scheduling under unexpected events in high-speed railway based on MAPPO

With the expansion of high-speed railway networks and the increase in traffic density, delay propagation caused by unexpected events such as equipment failures, natural disasters, and safety incidents has become more pronounced. Traditional approaches, including mixed-integer programming and fireworks algorithm-based optimization, perform well under stable conditions but are less suitable for dynamic situations with multiple interacting trains. To address this issue, this paper proposes a dynamic scheduling framework based on multi-agents proximal policy optimization, with the goal of reducing train delays. In this framework, each train is treated as an independent agent within a distributed decision structure and makes decisions based on local observations described by 11-dimensional state features. A composite rewards mechanism is designed by combining punctuality incentives, safety headway constraints, speed control rules, and rewards and penalties related to minimum dwell time. A centralized training with decentralized execution strategy is adopted to improve coordination among agents. Results from case studies show that the method can effectively mitigate delay propagation under disruption conditions. The average station punctuality rate reaches 56.3%, the average terminal punctuality rate reaches 91.6%, and key trains achieve 100% terminal punctuality, indicating that the approach is applicable in complex railway systems.