A Mesoscopic Transit Assignment Model Including Real-Time Predictive Information on Crowding

This paper presents a mesoscopic transit assignment model suitable for real-time prediction of on-board passenger numbers in transit networks with real-time individual predictive information on travel time components and also including on-board crowding. The path choice modelling framework is based on the reproduction of a travel strategy using random utility models that simulate both choices of departure time at origin and first access stop, and en-route choices of vehicle to board at stops. Such choices are based on attributes anticipated through a learning mechanism, which considers previous experiences and provides real-time predictive information. Within-day dynamic network loading considers vehicle capacity constraints, which allows the explicit modelling of fail-to-board events. Finally, results of an application on a real-size test network show the ability of the model to capture effects of providing individual predicted information on vehicle crowding.