Recoverable robustness for scheduling with precedence constraints

In this paper we apply the concept of recoverable robustness by network buffering in a real-world context for the first time. In particular, we consider the problem of assigning trains to platforms in a large railway station, and extend the mathematical formulation of the "nominal'' problem with variables and constraints that allow us to limit the propagation of delays that may occur at some point in the schedule. Our experimental results show that the resulting solutions have the same quality in terms of the nominal objective function, whereas delay propagation is significantly reduced.