Our AI team is first runner-up!

Solving disruptions in networks with Machine Learning

By leveraging Reinforcement Learning, a sub-domain of Machine Learning and Artificial Intelligence, Netcetera reached the second place out of 700 participants in the second round of the “Flatland” competition by the Swiss Federal Railways (SBB), Deutsche Bahn (DB) and French railways (SNCF). With our AI model, the trains learned to manage disruptions and to coordinate automatically causing minimal delays in large train networks.

The Flatland challenge competition in 2020 by SBB, DB and SNCF addresses issues of train scheduling and rescheduling and the need to resolve disruptions and delays. It tackles a key problem in the transportation world: How to efficiently manage dense traffic on complex railway networks? Out of over 2’000 submissions from over 700 participants from over 51 countries, we reached the outstanding second place with our model based on a Reinforcement Learning approach.

Similar to last year’s competition, our model was challenged in different grid world environments (Flatlands) simulating unforeseen disruptions and solved them with a dynamic and fully automatic rescheduling of trains. We are proud to be the first runner-up in this challenge, coming in with an excellent result. Our AI model proved to be very effective and efficient: it solved unforeseen incidents quickly, without disrupting larger parts of the network. This will help to increase the transportation capacity of the network without having to physically expand it.

Now, Netceteras AI experts are ready for the next challenge!

Flatland simulator: Manage and maintain railway traffic in complex networks as efficiently as possible using AI techniques developed by our Netcetera experts. https://i.imgur.com/Pc9aH4P.gif

Talk to our expert:

Ramon Grunder

Head of Delivery Digital Enterprise

More stories

On this topic