Automatic timetable conflict resolution with genetic algorithms

Posted: 11 August 2017 | , | No comments yet

In order to support the timetable planners at Swiss Federal Railways (SBB) in preparation of their daily timetables or construction site timetables, SBB IT has developed a special tool. The tool utilises genetic algorithms – a process borrowed from artificial intelligence – to generate new timetables, all of which are automatically checked to ensure their feasibility. For Global Railway Review, SBB colleagues Jürg Balsiger and Dirk Abels provide more details about these genetic algorithms.

Automatic timetable conflict resolution with genetic algorithms

Railway operations are underpinned by a valid, conflict-free timetable to which, thankfully, no major changes are made from day-to-day. However, changes and adjustments are a normal part of everyday operations and must be implemented in such a way that the timetable remains feasible. The reasons for making changes include scheduled deviations due to building measures or additional train paths.

When it comes to rescheduling the timetable, planners are currently reliant on the most powerful calculator known to the modern world – their brains. Rescheduling the timetable for a specific area requires a lot of experience given the many potential valid timetables there are to choose from. Timetable planners are responsible for finding a good solution; in other words, for defining a feasible timetable while keeping additional problems such as delays to a minimum. The aim is to revert to the standard timetable as quickly as possible following a line interruption. However, timetable planners cannot guarantee that their chosen timetable is the best of all potential solutions in the event of a disruption. The Infrastructure IT Solution Centre at SBB has therefore developed a fully automatic application to help timetable planners calculate a feasible, valid timetable at the push of a button in the event of disruptions or track/resource allocation conflicts.

The rest of this article is restricted to logged-in members. Login or subscribe free to read it.

Leave a Reply

Your email address will not be published. Required fields are marked *

Send this to a friend