Enhancing recoverable robust train routing in railway stations

Abstract

Busy railway stations with multiple intersecting lines often form critical points in railway timetabling, where delays from one line can propagate through the network. Increases in railway traffic as well as plans for integrated clock-face timetables thus require careful planning of platform assignments and routes through the station to limit the effect of delays and disruptions on the overall service quality. We present a new method for the long-term train routing problem in the station context. The goal is to find a microscopically feasible, robust train routing that is compatible with a predetermined network timetable. The method consists of a combined MILP model: A multicommodity flow formulation ensures the feasibility of the train routing by prohibiting conflicts between chosen routes. To ensure robustness to small initial delays, we directly include a scenario-based delay propagation model and minimize the delay propagation in the worst case, thus extending previous work on recoverable robustness. We discuss modelling choices and computational performance, present results on a case study and give an outlook on possible applications for long-term infrastructure planning.