Scheduling mechanics on a Shunting Yard: skills, synchronization and train movements

Abstract

When trains are not in use, they reside on shunting yards. On these shunting yards, the trains move around following a predetermined shunting plan to ensure an efficient use of the shunting yards. The trains have to be checked and small repairs have to be carried out by mechanics working at the shunting yard. Obviously, tasks cannot be carried out when a train is moving. Therefore, trains can have multiple moments at different times and locations where they are stationary and a mechanic is allowed to work on them. As some tasks require certain skills, there are restrictions on which mechanics can do which tasks. Some of these tasks require two mechanics to work on at the same time, and therefore the time when these mechanics start working on this job has to be synchronized. At the moment, scheduling is done ad hoc by the mechanics at the start of their shifts. The manual planning takes time out of a possibly busy schedule while not always resulting in a desirable, robust solution. Therefore, in this thesis, we propose a local search approach for scheduling mechanics on a shunting yard subject to these constraints. We aim to create robust schedules for the mechanics which can be generated before their shifts start and are easy to update if a disruption in the shunting plan were to occur. Our method is able to generate good results on a variety of realistic problem instances.