Simulation of track infrastructure and train traffic

Challenge

• In order to be able to generate decision recommendations for train scheduling, reinforcement learning is to be applied.
• As a basis for the training of RL models, a realistic, track-focused simulation of train traffic is to be created. SUMO is to be used as the simulation software
• Track plan data and timetable must be translated into a format that can be used for the simulation. The resulting track graph should represent the actual location and position of the operating points

Solution

• Understanding structure and documentation of the data formats used
• Evaluation of the realism of the simulation based on simulated times and distances using target data
• Modular software structure to allow the use of different data sources (e.g. timetable) or simulation tools

Result

• Realistic simulation that can be used to generate states for reinforcement learning
• Python package for parsing the infrastructure description and generating the track graph for the simulation
• Python package for parsing the RailML timetable and generating route files for the simulation