Although the traffic and transit assignment processes are intertwined, the interactions between them are usually ignored in practice, especially for large-scale networks. In this paper, we build a simulation-based traffic and transit assignment model that preserves the interactions between the two assignment processes for the large-scale network of the Greater Toronto Area during the morning peak. This traffic assignment model is dynamic, user-equilibrium seeking, and includes surface transit routes. It utilizes the congested travel times, determined by the dynamic traffic assignment, rather than using predefined timetables. Unlike the static transit assignment models, the proposed transit model distinguishes between different intervals within the morning peak by using the accurate demand, transit schedule, and time-based road level-of-service. The traffic and transit assignment models are calibrated against actual field observations. The resulting dynamic model is suitable for testing different demand management strategies that impose dynamic changes on multiple modes simultaneously.
Graph Neural Network for Large-Scale Network Localization
