
Capacity and Traffic Management System
The optimised railway operation of the future
Every day, over 40,000 local, long-distance and freight train journeys are facilitated by the German rail network. The Deutsche Bahn network alone covers around 33,000kms of track: supporting trains and stations with a wide range of characteristics. That’s why rail operations need forward-looking, 24/7 dispatching and control.
Today, local dispatchers in Deutsche Bahn's operation centers make many individual decisions every day. With Digital Rail for Germany, we want to largely automate these. Decisions then become easier: whether about waiting times, train sequences, track changes or reroutes. They can both be calculated more efficiently and implemented more quickly, so traffic can continue flowing smoothly throughout the network.
Digital Rail for Germany will also automate timetable planning in a similar way – at the same time as construction site planning, which, until now, has been a parallel process. Today, it takes several months and stages to create timetables. In future, an integrated system will distribute network capacity more efficiently, taking into account the needs of all customers and stakeholders.
Implementation
To make artificial intelligence (AI) suitable for capacity and traffic management, we created a simulation environment that digitally represents the railway world. This allows us to train AI algorithms in a realistic setting.
AI gains experience through so-called ‘reinforcement learning’. This type of learning is based on offering rewards when it successfully solves journey planning and dispatching problems. This is how algorithms gradually learn to master a variety of situations – similar to the way humans learn. It is then possible to develop AI algorithms that can react flexibly to complex and unknown challenges, and weigh up different scenarios. As this type of algorithm is scalable and can be parallelised on mainframe computers, the resulting system will be able to plan and control the entire rail network and its traffic.
The following video ‘reinforcement learning’ learning as the fundamental AI method of the capacity and traffic management system of the future.