DEVELOPMENT OF A DELAY PROPAGATION MODEL FOR THE METROPOLITAN RAILWAY NETWORK

Abstract

 This study targets the Tokyo metropolitan railway network. We construct a timetable-based spatiotemporal network and estimate passengers’ route choice and demand assignment using a Recursive Logit model. We then represent, in an integrated manner, both exogenous delay injection into selected trains and endogenous delay amplification and propagation caused by crowding-induced extensions of dwell and running times, train-following constraints, boarding denials, missed transfers, and partial recovery, and iteratively couple demand assignment with train operations. By handling delay generation and propagation together with route re-selection within a single computational framework, we capture the interaction between demand reallocation and the deterioration of crowding conditions under delays. The results show that localized delays can cascade to following trains, spread to other lines via transfers and through-services, and reshape the spatial distribution of crowding; under the analysis time-window constraint, some trips may also become infeasible. The estimated losses are dominated by increased waiting time, indicating that delay impacts appear not only as longer in-vehicle travel times but also as additional time spent waiting for connections and subsequent train services.