Multi-Objective Control of Urban Railway Speed with Deep Reinforcement Learning

Efficient and sustainable urban railway systems are important for addressing the transportation demands of densely populated areas. This paper explores the application of deep reinforcement learning (RL) to enhance real-time control in urban railway operations in Japan, focusing on optimizing train speed, energy consumption, and compliance with dynamic operational conditions. Unlike traditional approaches that often target predictable scenarios, this work extends RL methodologies to handle real-world complexities. A RL-based framework is developed that integrates vehicle longitudinal control strategies to manage train dynamics adaptively, thus ensuring efficient operation and safety. Simulation results demonstrate that our RL agent can effectively balance multiple objectives, such as improving energy efficiency while minimizing running time and maintaining optimal operational parameters. This study not only confirms the potential of RL in improving the performance and sustainability of railway systems but also opens new avenues for future research in intelligent transportation systems. The paper concludes by discussing the implications of our findings and proposing directions for further exploration in adaptive railway management technologies.