A deep reinforcement learning framework for integrated optimization of train scheduling, train control and passenger flow assignment

As the dynamic operating components of urban rail systems, train flow, passenger flow, and energy flow do not exist independently but rather interact with each other. Therefore, optimizing different subsystems within an integrated framework can enhance the overall efficiency of urban rail transit systems. This paper considers the integrated optimization of multi-train scheduling with virtual coupling, passenger flow assignment, and train control, aiming to reduce traction energy consumption and average passenger waiting times. A hierarchical optimization framework is designed, and a deep reinforcement learning algorithm is employed based on the framework, enabling interaction between different layers to achieve integrated optimization. The effectiveness of the method is validated using simulation data based on real scenarios from Chongqing Metro Lines 1, 2, 6, and the Loop Line. Compared to individual optimizations or simple combinations, the integrated optimization provides a better comprehensive solution.