Integrating activity-based models (ABMs) with simulation-based dynamic traffic assignment (DTA) have gained attention from transportation planning agencies seeking tools to address the arising planning challenges as well as transportation policies such as road pricing. Optimal paths with least generalized cost are needed to route travelers at the DTA level, while at the ABM level, only the least generalized cost information is needed (without fully specified paths). Thus, rerunning (executing) the least generalized cost path-finding algorithm at each iteration of ABM and DTA does not seem to be efficient, especially for large-scale networks. Furthermore, storing the dynamic travel cost skims for multiclass users as an alternative approach is not efficient either in regard to memory requirements. In this study, the aim was to estimate the least generalized cost so as to be used in destination and mode choice models at the ABM level. A heuristic approach was developed to use the simulated vehicle trajectories that were assigned to the optimal paths in the DTA level to estimate different cost measures, including distance, time, and monetary cost associated with the least generalized cost path for any given combination of the origin, destination, and departure time (ODT) and value of time. The proposed approximation method presented in this study used vehicle trajectories, aligned with the origin–destination direction and located in a specific boundary shaping an ellipse around the origin and destination zones at a certain time window, to estimate travel costs for the given ODT and user class. Numerical results for two real-world networks suggest the applicability of the method in large-scale networks in addition to its lower computational burden, including solution time and memory requirements, relative to other alternative approaches.
A dynamic traffic assignment model for highly congested urban networks
