Passenger delay models for rail networks

Abstract

Evaluation and forecasts of punctuality for railway systems have traditionally - if at all - been computed for trains, not for passengers. Little research has been carried out concerning passengers' route choice and assignment in schedule-based networks with delayed services. The article presents a schedule-based route choice model, where passengers' route choice depends on the planned timetable. In the realized (delayed) timetable, passengers may follow the same route as planned (pessimistic principle), i.e. they are nonadaptive. Using an optimal path search algorithm assumes that the passengers have complete knowledge on all present and future delays in the system (optimistic principle). The true choice must be somewhere between these extremes. In the model presented in the article, it is therefore assumed that passengers start reconsidering their route given they experience a certain delay threshold. If the threshold is large, the route choice approaches the pessimistic principle, and if it is small it approaches the optimistic. The model is now in operation and run each night on the realized timetable for the Copenhagen suburban rail network for the day before. This is used to measure passenger delays in the existing network. The article presents some results of this, including the importance of the threshold. The second application of the model links it with a rail simulation model enabling it to forecast passenger delays related to future timetables. The final part of the article presents this application. The article shows that the new model enables more accurate estimates of door-to-door passenger delays than prior methods.