Delay Propagation on a Suburban Railway Network

EasyChair Preprint 2517

Abstract

Understanding and predicting delays is a central task for any railway system, but it is made much more difficult by interactions between trains. We propose a new model for the delay propagation phenomenon, which takes into account infrastructure constraints without assuming knowledge of the resource conflicts linking trains with one another. Our approach relies on a set of hidden variables called the network jam, structured as a Dynamic Bayesian Network to enable spatial propagation. We also present a statistical analysis of the minimax estimation error, a variational inference method and numerical tests on simulated and real data.

Keyphrases: Dynamic Bayesian Network, Railway Operations, Train delay prediction, variational inference

@booklet{EasyChair:2517,
  author    = {Guillaume Dalle and Yohann De Castro and Axel Parmentier},
  title     = {Delay Propagation on a Suburban Railway Network},
  howpublished = {EasyChair Preprint 2517},
  year      = {EasyChair, 2020}}