SD
SIGNAL+DRAHT | Issue 01-02/2025

Towards a holistic traffic management system with Multi-Agent AI

February 2025 | Stefan Schneider, Anne Roets, Ciprian Stirbu, Farhad Safaei, Faten Ghriss, Jan Wülfing, Mehmet Güral, Nima Siboni, Roman Liessner, Thomas Hustache, Thomas Lecat, Umashankar Deekshith, Valerii Markin, Victor Le, Wissam Bejjani,Michael Küpper,Irene Sturm

An intelligent traffic management system constitutes a core component for highly automated railway network operations and plays the role of a “planning brain”. The approach presented here uses the Multi-Agent Reinforcement Learning (MARL) method, a subfield of Artificial Intelligence (AI). AI acts on a railway simulation and creates highly detailed schedules for train operations. Significantly shorter computing time is required compared to classic optimisation methods. The approach therefore promises to be applicable to large network areas...