A multi-level framework for generating train schedules in highly utilised networks

Abstract

This paper proposes a comprehensive multi-level framework for railway scheduling, which starts with a high-level commercial description of intended train services and aims to generates a conflict-free detailed schedule as the final outcome. The approach consists of three description levels and corresponding interfaces that enable a structured decomposition of the scheduling problem into sub-problems of manageable size. The starting point is a formal structure for describing the service intention including periodicity information. As typical timetables are neither entirely periodic nor aperiodic, a projection scheme is used to create an augmented periodic problem. This augmented periodic timetabling problem is solved first globally on an aggregated topology and simplified safety model, and subsequently refined locally by considering all details of the infrastructure and train dynamics. In particular, a new model is used that solves the local problem in station areas more efficiently and faster than previous model, which is crucial for the interaction of the different levels. Finally, the generated periodic conflict-free schedule is rolled out over the complete day to create a production plan fulfilling all requirements specified in the service intention. The presented approach is finally tested and illustrated on a real-world case study for the train services currently in place in the Lucerne region, a highly utilised part of railway network in central Switzerland. There, it is possible to perform the complete procedure in less than 10 minutes. The generated solution is optimal in each step of the multi-level approach and has a comparable quality with the operated schedule.