Comparison of Simulation-based Dynamic Traffic Assignment Approaches for Planning and Operations Management

Transportation practice is increasingly recognizing the need for dynamic models for realistic scenario analyses and policy decision-making. Dynamic Traffic Assignment (DTA) theory and tools are now thought to be sufficiently advanced to meet this need. However, DTA has historically been associated with mesoscopic models, motivated by the perceived computational intractability of high-fidelity microscopic models in an equilibrium context. In this paper, we use a real-world project to demonstrate that current hardware and efficient software allow for large-scale DTA under the more desirable microscopic traffic simulation framework. Such a tool is a more natural fit for most dynamic analyses, including evacuation and emissions modeling. A calibrated peak-period microscopic model for 500 square miles of the Greater Phoenix region of Arizona is presented to illustrate that DTA no longer need compromise modeling accuracy purely on running time considerations. We also qualitatively discuss the potential loss in accuracy arising from accepting the more common assumptions inherent in mesoscopic approaches.