AI-driven dynamic timetable optimization for the Casablanca-Mohammedia-Rabat commuter railway

This study aims to investigate how artificial intelligence can enhance the resilience and efficiency of railway timetables in disruption-prone commuter corridors. Specifically, it focuses on Moroccan railway networks connecting Casablanca, Mohammedia, and Rabat, where recurrent delays and congestion compromise service reliability. The research seeks to determine how integrating predictive delay modeling with adaptive passenger behavior can reduce secondary delays, alleviate congestion, and maintain timetable stability under operational disturbances.

A unified, simulation-based framework was developed, combining 3 interlinked modules: (1) machine learning-based predictive delay forecasting, (2) agent-based modeling of passenger adaptive behavior, and (3) dynamic timetable reoptimization using a rolling-horizon heuristic approach. The framework operates as a closed-loop system, where predicted delays and simulated passenger responses continuously inform real-time timetable adjustments. Empirical validation was conducted using operational data from Moroccan commuter trains, with scenario-based analysis comparing baseline, prediction-only, and fully integrated interventions.

Results show that the fully integrated framework significantly improves operational performance. Average train delays were reduced by 46%, total passenger waiting time decreased by 43%, and congestion intensity was nearly halved, while timetable stability remained high at 95%. The study also demonstrates that passenger behavior plays a critical role in delay propagation, and that combining predictive forecasting with adaptive control strategies prevents the nonlinear amplification of secondary delays that traditional train-centric models fail to address.

This research advances the field of railway operations by presenting a passenger-centered, AI-driven timetable reoptimization framework that integrates predictive analytics and behavioral simulation within a dynamic feedback loop. Unlike conventional models, it captures emergent congestion patterns, anticipates disruptions proactively, and provides actionable operational strategies without requiring major infrastructure expansion. The study offers a novel methodological contribution with practical implications for enhancing commuter railway resilience in high-density, disruption-prone contexts.