Recoverable robustness for railway rolling stock planning

In this paper we explore the possibility of applying the notions of Recoverable Robustness and Price of Recoverability (introduced by [5]) to railway rolling stock planning, being interested in recoverability measures that can be computed in practice, thereby evaluating the robustness of rolling stock schedules. In order to lower bound the Price of Recoverability for any set of recovery algorithms, we consider an "optimal" recovery algorithm and propose a Benders decomposition approach to assess the Price of Recoverability for this "optimal" algorithm. We evaluate the approach on real-life rolling stock planning problems of NS, the main operator of passenger trains in the Netherlands. The preliminary results show that, thanks to Benders decomposition, our lower bound can be computed within relatively short time for our case study.