Decision support for scheduling security crews at Netherlands Railways

We address the problem of scheduling work of security guards operating on trains and stations, and explain how operations research is suitable for solving a problem that adds new challenges to classical crew scheduling. Planning the work of these security guards is challenging because it requires dealing not only with the complexity inherent to crew scheduling problems but also with an optimisation goal that itself is defined more in qualitative than quantitative terms. Our contribution to handling these challenges is summarised as follows: we describe the problem and model it as a shortest path problem and a set covering problem with additional constraints; we adapt an existing heuristic that is based on Lagrangian relaxation, subgradient optimisation, column generation and greedy heuristics to be applicable to the problem (for instance, we add an improvement step based on local search); we implement the resulting solution method, resulting in a software prototype named TUTIS; we test TUTIS with a real problem instance supplied by Netherlands Railways. Experimental results lead security experts to believe that the prototype not only presents meaningful results in terms of scheduling work, but also can help decide the way security guards will be deployed in the future.