An Optimization-Based Decision Support Tool for Incremental Train Timetabling

We consider the typical workflow of a route planner in the context of short-term train timetabling, that is, the incremental process of adjusting a timetable for the next day or up to the next year. This process usually alternates between (1) making rough modifications to an existing timetable (e.g., shifting the departure of a train by half an hour) and then (2) making small adjustments to regain feasibility (e.g., reduce or increase the dwell time of some trains in some stations). The most time-consuming element of this process is related to the second step, that is to manually eliminate all conflicts that may arise after a timetable has been modified. In this work, we propose a mixed-integer programming model tailored to solve precisely this problem, that is to find a conflict-free timetable that is as close as possible to a given one. Previous related work mostly focused on creating complex models to produce “optimal” timetables from scratch, which ultimately resulted in little to no practical applications. By using a simpler model, and by trusting route planners in steering the process towards a timetable with the desired qualities, we can get closer to handle real-life instances. The model has been integrated in a user interface that was tested and validated by Norwegian route planners to plan the yearly timetable of a busy railway line in Norway.