Modeling the Line Interruption Issue in a Railway Network

Rail line interruptions are rare but very costly events, as they require a complete re-definition not only of the timetable of the trains, but also of their path, with major variations at least in the hit area. To the best of our knowledge, the literature is rich of documentation on timetable re-scheduling in case of delays and/or disruption of train lines, but without considering path deviations. The Flatland initiative has published a 2D railway-world toolkit that allows developers devise and experiment with different solutions to deal with the train re-scheduling issue particularly, but not exclusively, through reinforcement learning (RL) with multi-agent path finding (MAPF). While the approach looks very promising to deal with the complexity of the issue, the platform still has some limitations in terms of the modeling of a realistic railway network scenario. This article proposes the integration of some key features in order to make Flatland a valuable platform for training RL agents for supporting decision making in real-world train re-scheduling.