Defining Railway Traffic Conflicts and Optimising Their Resolution: A Machine Learning Perspective

This paper reports on the initial phase of research into automated traffic conflict resolution for suburban railway operations. It defines railway traffic conflicts, categorising types such as catch-up, crossing, and proximity, and establishes optimisation criteria focused on punctuality, efficiency, safety, and passenger satisfaction. Promising machine learning approaches are reviewed, including supervised learning for conflict prediction, reinforcement learning for adaptive resolution, and unsupervised methods for identifying conflict-prone scenarios. The study concludes by proposing a simulation framework for empirical evaluation, providing a foundation for AI-driven advancements in railway traffic management.