Guest lecture: Multi-Agent Reinforcement Learning for Decision Support in Railway Operations

Railway operations are characterized by dense traffic, tight safety constraints, and complex trade-offs under time pressure. This talk presents how multi-agent reinforcement learning (MARL) can be used as a decision-support approach for railway dispatching, combining large-scale simulation with learning-based optimization. The focus lies on practical insights, evaluation in realistic scenarios, and the role of human–AI teaming in future railway operations.

Dr. Roman Liessner  is an AI Innovation Manager at Deutsche Bahn, working at the intersection of railway operations, optimization, and artificial intelligence. His work focuses on simulation-based optimization and reinforcement learning for traffic management and dispatching. He is involved in applied research projects, including the AI4REALNET project, with a focus on human-centered AI in safety-critical transport systems.