A Robust Rolling Stock Rescheduling Approach

Railway operators all over the world transport millions of passengers on a daily basis. These railway operators schedule the timetable and associated resources during an extensive planning period. The most common resources are the rolling stock and the crew schedule. All three schedules are often an optimized trade-off between passenger service, efficiency and robustness. During actual operations, railway operators are faced with all kinds of unforeseen incidents having a negative effect on the railway schedules, which are named disruptions. For example, a system malfunction, an accident, or the complete blockage of a track segment by a fallen tree. In such cases, the timetable, the rolling stock schedule, and the crew schedule might be rendered infeasible. Effective disruption management is currently an active research area in Operations Research. This has lead to many papers on algorithmic tools for rescheduling the timetable, rolling stock and crew schedule (see Cacchiani et al. (2014)). The focus in this paper is solely on rescheduling the rolling stock (RS) schedule, scheduling the timetable and crew is outside the scope. Many researchers (e.g. Nielsen (2011), Wagenaar (2016)) have already conducted research on this topic. However, all these models assume the duration of a disruption to be known at the time of rescheduling. Unfortunately, in practice the duration of a disruption is often uncertain. Assuming that the disruption duration is known may lead to new problems if the realized duration of the disruption differs from what was assumed. Not considering the variation in disruption duration will result in a situation in which the rescheduling must be performed all over again. Some decisions could turn out to be very costly in terms of i) operating costs, ii) passenger service, and iii) possible communication problems. In practice, an estimation on the disruption duration is made as soon as a disruption occurs. This estimation gives, among other information, a minimum and maximum duration. The main contribution of this paper is a model that is able to reschedule the RS schedule in a robust way. This means that the RS schedule created for the minimum disruption duration requires no or small additional changes in case the disruption duration turns out to be longer. A fully robust solution might come at an high expense. To that end, a second contribution is that we have developed a model to create a semi-robust rolling stock circulation, which includes a given level of robustness required.