Dynamic Graph Generation and Dynamic Rolling Horizon Techniques in Large Scale Train Timetabling

Abstract

The aim of the train timetabling problem is to find a conflict free timetable for a set of passenger and freight trains along their routes in an infrastructure network. Several constraints like station capacities and train dependent running and headway times have to be satisfied. In this work we deal with large scale instances of the aperiodic train timetabling problem for the German railway network. The problem is modelled in a classical way via time discretised networks, its Lagrange-dual is solved by a bundle method. In order to handle the enormous number of variables and constraints dynamic graph generation and dynamic rolling horizon techniques are employed.