A Gap Function Vehicle-Based Solution Procedure for Consistent and Robust Simulation-Based Dynamic Traffic Assignment

This research proposes a gap function vehicle-based (GFV) gradient-like procedure for solving the simulation-based dynamic traffic assignment (SBDTA) problem in a computationally efficient manner. In contrast to the method of successive averages (MSA) based approach, for each iteration and each origin-destination-departure triplet, the amount of vehicles to be updated with a new path depends on the gap function value - current solution’s proximity to the time-dependent user-equilibrium (TDUE) condition - to implement both gradient-like search direction and step size methods. Vehicles with longer travel time are prioritized to be selected for path update. The proposed approach allows for faster convergence, compared with the MSA-based approach since each triple has an individual search direction and step size. When vehicles are loaded with previously solved baseline TDUE solution for alternative scenario analysis, the solution appears to be more consistent than the MSA-based approach as the proposed algorithm avoids over adjustment of flow for triplets that are not significantly affected by the capacity change in the alternative scenario. This results in preservation of consistent and robust assignment results. The results from numerical testing of two networks demonstrate the advantageous performance of the proposed algorithm over MSA-based approaches from both the convergence and solution consistency standpoints.