Railway systems are extremely sensitive to disruptions, which disturb the regular running of the system and cause delays. Without fast and efficient counter actions, these delays may propagate through the entire system and may cause a lot of damage. Today the handling and resolution of these disruptions is usually done manually by human operators. Scheduling problems for static scenarios have been well investigated. The first automatic dispatching systems for railway networks have been developed as basic prototypes in the last years. A central problem is detecting and resolving occupation conflicts (e.g., if two trains simultaneously intend to use the same track section); in the worst case such conflicts may lead to blocking and even deadlocking of resources.
This dissertation project aims at developing fast algorithms for certain problems and subproblems arising in disruption handling. Here a crucial question is how to deal with missing or with uncertain data. In the ideal case, the decision makers would have full information all the time, on every detail of the railway system. Then current speed and location of trains must be continuously available. Unfortunately, in the real world the available information is often unreliable—it may only be available at certain discrete time points, it may be faulty, or it may be missing altogether. On the one hand, this adds uncertainty features to the resulting optimisation problems, and on the other hand it adds real-time features (as the information arrives over time and has to be processed and used immediately). Another important issue is the robustness of chosen decisions: Sometimes if new information becomes available, it might be good to undo an older decision and to switch to another (better) one; and with even more information available, one might want to switch again back to the first decision; and so on. Then in the worst case this may lead to fluctuating and alternating decisions, and of course this has to be avoided.
The first stages of the dissertation project will concentrate on the problems that arise from uncertainty and missing information. During later stages, other relevant issues will be addressed such as how to take traffic flows and connections of passengers into account (if their travel routes are known), and how to optimise the circulation of trains and carriages.