Skip to main content
To KTH's start page To KTH's start page

Improving Timetable Robustness and Punctuality of Railway Traffic

A Combined Simulation-Optimization Approach for Nonperiodic Timetabling on Double-Track Lines

Time: Tue 2022-12-06 09.00

Location: Kollegiesalen, Brinellvägen 8, Stockholm

Video link:

Language: English

Subject area: Transport Science, Transport Systems

Doctoral student: Johan Högdahl , Transportplanering

Opponent: Professor Rob Goverde, Delft University of Technology

Supervisor: Docent Markus Bohlin, Transportplanering; Docent Oskar Fröidh, Järnvägsgruppen, JVG, Transportplanering

Export to calendar

QC 202211-14


To evaluate the robustness of a timetable against minor delays, it can be simulated. A natural question following a simulation is how the results can be used to improve the timetable, which has received limited attention in the literature. This thesis therefore aims to investigate how the combination of simulation and optimization can be used to improve robustness and punctuality of railway timetables, measures that are of great importance. In the thesis, I propose two-step methods based on first simulating a timetable and then optimizing it. I propose models to predict how delays and punctuality change when adjusting the timetable based on the simulation results. In turn, these models are used in the objective function to determine optimal adjustments for a given timetable. This approach is based on exact optimization distinguishing it from previous methods to create robust timetables using combined simulation and optimization.

The proposed methods were evaluated in simulation experiments on the Swedish Southern and Western Main Line, which are two highly utilized lines connecting the Swedish capital Stockholm with Gothenburg and Malmö (second and third largest cities in Sweden). The results indicate that the proposed methods improve punctuality and robustness in simulation at the cost of slightly longer travel times. Compared with two optimization-based methods from the literature and two optimization-based methods based on simple strategies that in principle can be implemented manually, the proposed methods result in either comparable or better robustness and punctuality. Furthermore, the results suggest (1) the methods are effective even if only minor adjustments are allowed; (2) they can be used on problem instances with a large number of trains; and (3) they are robust against under or overestimating the delays in the simulation step.