A dynamic data-driven approach for rail transport system simulation

Public rail transport systems concern infrastructure and control strategies with long life spans. While many rail system simulations aim at planning and design, this paper proposes a dynamic data-driven approach to improve the adaptability of the model, hence promoting an extended use of the simulation model. In the proposed approach, the simulation study uses real data streams for automatic model calibration at run-time. For situations that cannot be automated, expert interference can be supported by interactive processes. Different model calibration schemes can be applied to several replications simultaneously to assess the schemes and to determine the parameter values that best match the most recent situation. The model can be fed with data derived from different scenarios, from decision variations or from real-time measurements to accomplish accurate and automated model calibration. This provides a foundation for the use of simulation for railway controller training tools and real-time rail monitoring systems.