Deep reinforcement learning approach for routing and scheduling of trains at railway station

Deep reinforcement learning approach for routing and scheduling of trains at railway station

연구 분야: Infrastructure

학회: CODS-COMAD '24: Proceedings of the 8th International Conference on Data Science and Management of Data (12th ACM IKDD CODS and 30th COMAD)

Scheduling of trains at a railway station involves assigning traversal path for trains entering a station network, and scheduling their entry, exit, and reversal activities. The schedule must ensure conflict-free train movements in accordance with operational guidelines and minimize the total time trains spend at the station. This task is challenging due to the complexity of the decisions involved, as there are numerous interconnected tracks to consider. This paper introduces a deep reinforcement learning approach to schedule trains at a railway station and terminus having route-lock and route-release interlocking settings. We test our approach on two real world stations of Indian Railways. Experimental results demonstrate that the trained reinforcement learning agent performs better than or as good as benchmark dispatching rules under various traffic conditions for two different configuration of railway stations.