Rolling-stock maintenance scheduling in an uncertain environment

This paper introduces a constraint programming approach to the problem of optimising maintenance scheduling of rolling-stock in a maintenance depot. The maintenance scheduling problem considered here differs from that most widely studied in the literature in that we consider the problem at a more granular level. In particular, a time-variable fleet demand profile is incorporated in the objective function, and depot resource capacity restrictions are enforced at an hourly level. Due to the level of uncertainty inherent in the problem inputs, both proactive and reactive features are incorporated into the proposed method. The prototype scheduler is empirically shown to be able to adapt to unexpected maintenance events. Additionally, we show how the prototype can be used as a tool for identifying problem bottlenecks and as a simulation tool for performing cost-analysis of possible changes to depot infrastructure and resources.