Simulation-Driven Railway Delay Prediction: An Imitation Learning Approach
Clément Elliker, Jesse Read, Sonia Vanier, Albert Bifet
TL;DR
<3-5 sentence high-level summary> Frames train delay forecasting as a stochastic simulation task and introduces Drift-Corrected Imitation Learning (DCIL), a self-supervised extension of DAgger that mitigates covariate shift without an external oracle. By combining macroscopic, event-driven dynamics with data-driven representations and enabling Monte Carlo uncertainty via trajectory rollouts, DCIL delivers robust delay forecasts up to 30 minutes ahead. On a large Infrabel dataset, DCIL consistently outperforms traditional regression and Behavioral Cloning, with Transformer-based models achieving the best overall accuracy and simpler MLPs also surpassing larger models trained with regression. This work demonstrates the practical potential of imitation-learning-based simulation for scalable, uncertainty-aware railway delay prediction and opens avenues for integration with decision-support tools in railway operations.
Abstract
Reliable prediction of train delays is essential for enhancing the robustness and efficiency of railway transportation systems. In this work, we reframe delay forecasting as a stochastic simulation task, modeling state-transition dynamics through imitation learning. We introduce Drift-Corrected Imitation Learning (DCIL), a novel self-supervised algorithm that extends DAgger by incorporating distance-based drift correction, thereby mitigating covariate shift during rollouts without requiring access to an external oracle or adversarial schemes. Our approach synthesizes the dynamical fidelity of event-driven models with the representational capacity of data-driven methods, enabling uncertainty-aware forecasting via Monte Carlo simulation. We evaluate DCIL using a comprehensive real-world dataset from \textsc{Infrabel}, the Belgian railway infrastructure manager, which encompasses over three million train movements. Our results, focused on predictions up to 30 minutes ahead, demonstrate superior predictive performance of DCIL over traditional regression models and behavioral cloning on deep learning architectures, highlighting its effectiveness in capturing the sequential and uncertain nature of delay propagation in large-scale networks.