Activity-based models (ABM) are a desirable travel demand modeling paradigm due to their fine behavioral resolution in both the spatial and temporal dimensions. Motivating such detailed analyses is the recognition that travel is simply a by-product of individuals’ need to fulfill daily activities. For such models to realize their potential, the supply side must effectively support the fidelity on the demand side. Key is the ability to evaluate dynamic network performance and congestion patterns on a fine temporal basis while being sensitive to changes in demographics, activity patterns and network supply. Dynamic traffic assignment (DTA) has the modeling capability to provide this fidelity.

DTA and ABM must be compatible at several levels in order to achieve a productive integration. The DTA should handle individual trips (rather than aggregate trip matrices) and operate on a network representation sufficiently dense so that the ABM’s activity locations are well-connected. The DTA’s temporal resolution must also mimic the demand side (e.g. trip departures, activity durations, etc.) in order to minimize aggregation error. Further, its components must replicate important phenomena such as route choice, response to traveler information, and capacity reductions due to supply disruptions. Finally, the integrated system must possess practical running times so that planners might use the tool for real-world scenario analyses.

In this paper, we present real-life examples integrating an ABM with a high-fidelity, simulation-based DTA model. The case of Jacksonville, Florida is used to illustrate a framework for facilitating the exchange of information between demand modeled with the DAYSIM ABM, and supply modeled through a large-scale DTA based on traffic micro-simulation. Demand output from DAYSIM is imported into the simulation model directly, as a table of activity sequences. Activity locations are tagged to the nearest road segments to generate trip origins and destinations. High-resolution network performance measures from the simulation model could then be fed back to DAYSIM to inform the next iteration of activity choices. Practical concerns such as convergence and the impact of stochasticity (on both demand and supply sides) are discussed. A similar study based on the Burlington, Vermont network is also presented.