Semi-Hierarchical RL Approach for Railway Vehicle Rescheduling

A new paper on arXiv (2605.10257) introduces a semi-hierarchical deep reinforcement learning (RL) approach to the Vehicle Rescheduling Problem (VRSP) in railway traffic management. The method separates dispatching from routing using dedicated action and observation spaces, allowing policies to specialize in distinct decision scopes. This addresses the challenge of managing disruptions under rising traffic density and infrastructure limits, where traditional Operational Research (OR) methods often underperform and human expertise remains prevalent. The proposed RL formulation aims to improve scalability and performance in dense rail networks, bridging a gap where existing RL approaches struggle.