A Markov decision process model for coupled intelligent signaling and train operation systems in high density rail transit

This study addresses the inefficiency in signal control for high-density rail transit systems, where complex interactions between intelligent signaling and train operations compromise both safety and operational stability. An advanced signal control model is developed based on a Markov Decision Process (MDP) framework enhanced with reinforcement learning (RL) to optimize rail operations. The model mathematically characterizes the rail environment through state space (track signals and train queues), action space, and a reward function focused on delay reduction. A Deep Q-Network (DQN) algorithm with prioritized experience replay enables the intelligent control system to continuously learn and adapt signal strategies under dynamic conditions. Experimental results demonstrate the model’s superiority over conventional approaches such as stepwise prediction and multi-agent reinforcement learning (MARL), achieving significant improvements across key metrics: 18–22% reduction in passenger wait times, 12–15% shorter journey durations, 8–10% lower fuel consumption, and 25–30% higher system rewards. These findings confirm the model’s effectiveness in managing high-density rail operations while maintaining safety standards, offering a robust solution to the challenges of modern rail traffic management. The adaptive nature of this approach ensures optimal performance under varying traffic conditions, marking a substantial advancement in intelligent transportation systems.