Heuristics for Delay Management in Utrecht, Gelderland and Noord-Brabant

Since standard integer programming techniques do not succeed in tackling large-scale real-world in- stances, I apply several heuristics for delay management. The first heuristic I implement is the Waiting Time Rule (WTR), where the decision whether or not to delay a connecting train, in order to maintain a connection, is based on a predetermined threshold, varying between 0 and 5 minutes. In contrast to the WTR, the Ratio of Transferring Passengers (RTP), which is the second heuristic I implement, actually takes into account the number of passengers that plan to use a connection. This could be beneficial for the results since now the usage of a specific connection is considered in the decision. Finally, the third heuristic is based on the classical delay management model without passenger rerouting from Schobel (2007). This model can be solved with the assumption that passengers who miss a connection will wait for an estimate of additional delay (D) or for the timetable cycle time (the time after which the timetable repeats itself). For each heuristic, the goal is to find the parameter value for which the heuristic performs best (dmax for the WTR, ρmin for the RTP and D for the classical model). I evaluate the outcomes for these heuristics separately for the case where only long-distance trains and corresponding stations are taken into account and for the case where all trains and stations are considered. To this end, I execute numerical experiments on real-world instances from Netherlands Railways. I discuss the results and eval- uate and compare the performance of the different heuristics. Finally, some improvements and directions for further research are listed.