Controlling the propagation of passenger disruption impacts in multi-level public transport networks

The passenger impact of a disruption on the train network can propagate over the multi-level public transport (PT) network, via the transfer hub to the urban PT network. Hence, an optimal holding control decision for urban services at the transfer location should account for the impact of a disruption on another PT network level. Modelling framework We first quantify the passenger impacts of disruption propagation resulting from an exogenous train network disruption to the urban PT network level. Thereafter, we develop a rule-based controller for holding urban PT services while taking into account predicted passenger delays and rerouting from the train network level caused by the train network disruption. This means that in this study a control decision is triggered by services which are not subject to this same control decision. Scenario design We quantify the total passenger welfare for three different scenarios, expressed as the generalized travel time over all passengers: -Scenario 1: undisrupted train network; no urban control intervention; -Scenario 2: train network disruption; no urban control intervention; -Scenario 3: train network disruption; urban control intervention. Control problem description The applied control strategy entails the decision whether to hold urban PT runs at multi-level transfer stops for a certain holding time in case a disruption occurs on the train network. The predicted welfare impacts on four different passenger segments are incorporated in this holding decision: (i) Upstream boarding and downstream alighting (through) passengers; (ii) Downstream boarding passengers; (iii) Reverse downstream boarding passengers; (iv) Transferring passengers at holding location. A passenger-oriented decision rule is applied for the controller, where predicted costs of the control decision are deducted from the predicted control benefits for all passenger segments, aimed at minimizing passenger travel costs on the urban network. Holding results in a direct extension of in-vehicle time at the holding stop of passengers who board upstream the holding location and alight downstream the holding location, and a waiting time extension for downstream boarding passengers, corrected for turnaround buffer time for reverse downstream boarding passengers. Besides, holding reduces waiting time for passengers transferring at the holding location, compared to having to wait for the next service. The holding strategy also affects the different passenger segments in terms of perceived in-vehicle time due to changed crowding levels. Due to the non-linear nature of perceived in-vehicle time as function of crowding, we quantify crowding effects over all passenger segments simultaneously. Application We apply our methodology to the multi-level PT network of The Hague, the Netherlands. BusMezzo, an agent-based dynamic simulation model for PT operations and passenger assignment, is used as evaluation tool.