Rolling Stock Scheduling in an integrated planning process

Abstract

The Nederlandse Spoorwegen (NS) is the largest passenger railway operator in the Netherlands and operates a fleet of more than 700 train units. One of the problems they encounter in their planning process is the Rolling Stock Scheduling Problem, which involves the assignment of compositions to a set of pre-existing trips. The goal is to find composition assignments that minimize the number of passengers that cannot be seated, while also taking into account other costs. In this thesis, we will show a way to approach this problem using a local search heuristic. An additional challenge in this is the planning of night movements to allow for a more efficient usage of the available train units. We propose a flow approach to solve this subproblem, which we then expand to an MIP that can be solved during steps of the local search. We compare our results with those of TAM, a solver that is currently used to find solutions to the problem using a MIP. Using this we are able to find solutions that are 5-10\% off the TAM benchmark in most scenarios.