Several algorithms have been proposed for optimizing rail traffic management in case of perturbation. However, their assessment often overlooks modeling assumptions. Realistic evaluations are essential to ensure theoretical solutions match practical feasibility and performance.
Our research introduces a modular framework for evaluating the performance of rail traffic management algorithms in a realistic context. It integrates three main contributions and relies on an API. 
The first two contributions consist in the integration of a passenger simulation layer into the commercial rail traffic simulator OpenTrack, together with the implementation of a stochastic passenger route choice model. Trains are realistically simulated and real historical distributions of traffic perturbations are replicated. This integration allows the direct assessment of how traffic disturbances and control strategies affect passenger flows, platform crowding, waiting times, and missed connections within a detailed microscopic traffic environment.
The route choice model selects paths for passengers through Monte Carlo simulations, considering the utility function associated by passengers to the possible alternatives, estimated using observed choices. When train delays increase between subsequent stations, the API triggers passenger rerouting: depending on the current and planned traffic evolution, passengers decide whether to stick to their original plan or choose a different option. This mimics real-time decisions, assuming passengers access current traffic information. 
The third contribution is the periodic exchange between the simulation and the optimization algorithm assessed, in close-loop (Quaglietta et al., 2016). 
At each control cycle, the simulator provides the optimizer with the current state of the rail system and passenger flows, and the optimizer returns updated routing and scheduling decisions to be applied in the subsequent simulation step. The closed loop is fully demand-sensitive, as control decisions directly interact with passenger behavior and observed flows. To the best of our knowledge, this is the first demand-sensitive closed-loop rail traffic management framework implemented with the commercial simulator OpenTrack.
We validate the proposed framework using different passenger and traffic scenarios in Copenhagen rail suburban network over a three-hour simulation. The scenario includes about 243 trains and 15910 passengers. The network features a radial topology with multiple branches converging on a central shared section, representing the main capacity constraint.
To showcase the application, we consider different variants of an optimization algorithm (Pellegrini et al., 2014), in its classic version or augmented with a passenger consideration module (Pascariu et al., 2025}, and we analyze the impact of different levels of passenger route choice flexibility to assess the effects of potential discrepancies between simulated passenger behavior and the demand assumptions embedded in the optimization process.