Rules of thumb: practical online-strategies for delay management

Abstract

The delay management problem asks how to react to exogenous delays in public railway traffic such that the overall passenger delay is minimized. Such source delays occur in the operational business of public transit and easily make the scheduled timetable infeasible. The delay management problem is a real-time problem further complicated by its online nature. Source delays are not known in advance, hence decisions have to be taken quickly and without exactly knowing the future. This work focuses on online delay management. We enhance established offline models and gain a generic model that is able to cover complex realistic memoryless delay scenarios. We introduce and experimentally evaluate online strategies for delay management that are practical, easily applicable, and robust. Our experiments show that the most promising approach is based on simulation and a learning strategy which is able to deal very well with the wait-depart decisions. Finally, by analyzing the solutions found, we gain interesting insights in the structure of good delay management strategies for real-world railway data.