A model for Robust Rolling Stock Scheduling

Abstract

Major disruptions render the schedules of public transport operators infeasible. The majority of the recently developed algorithms for updating these schedules assume the duration of the disruption is known when they occur. However, in practice, this is generally untrue. This paper compares three different models for Rolling Stock Rescheduling under uncertainty: an optimistic approach, a strict-robust model inspired by the definition of disruption of Ben Tal et al, and a Light-Robustness approach that aims to provide a middle way between the two. The models are evaluated on a realistic case study of the Netherlands Railways. Initial results indicate that building robustness against different disruption durations is worthwhile when alternative scenarios are associated with a sizable probability mass. The best approach depends on the probability distribution over the different scenarios.