Uncertainty in resource reopening times complicates train holding and rerouting decisions when closures coincide with arrival delays and processing‑time extensions. We propose a rolling‑horizon distributionally robust model that updates predictions of the remaining closure duration as observations of disruption resolution become available. The model jointly optimizes train routes, sequences, and timings, minimizing the worst‑case expected total completion delay over an ambiguity set of scenario probabilities. The solution algorithm, GNN‑LBBD, employs logic‑based Benders decomposition to coordinate routing and resource scheduling. Conditional cuts are derived from resource‑use patterns, and a graph neural network guides the order in which resource‑relaxation subproblems are solved. Experiments on 28 instances for a five‑track station show a mean total completion delay of 61.13 train‑minutes, 16.63% lower than tabu search and 20.06% lower than variable neighborhood search, with computation time comparable to tabu search. Information comparisons on six representative instances demonstrate that using observations of disruption resolution reduces mean total completion delay by about 21.7%. Under the same information conditions, scenario‑based and conditional‑mean predictions yield similar overall results. The benefit of information is more pronounced in instances with alternative routes, suggesting a relationship among observation usage, station resource structure, and the scope for schedule adjustments.
Review