Utilizing a Static-based Initial Feasible Solution to Expedite the Convergence of Dynamic Traffic Assignment Problems

Dynamic Traffic Assignment (DTA) is rapidly emerging as the tool of choice for both regional planning and innovative operational traffic analysis. However, existing DTA techniques require extremely high computational resources and time to provide a meaningful solution and an acceptable convergence. One way to decrease the required computational resources and time is by providing the DTA model with an initial set of feasible paths and flows (warm-start solution). This paper introduces the concept of static warm-start procedures to expedite the convergence of DTA methodologies. The proposed techniques are based on the solution of static traffic assignment models for an equivalent static network. Approaches differ in the way in which an initial path set and the corresponding dynamic flows are derived from the solution of a static UE model. The warm-start solution is used to replace the initialization steps of a Cell Transmission Model-based (CTM-based) DTA model at a much lower computational costs. The warm-start techniques were tested on two real networks, Sioux Falls and Anaheim. The overall trends, which are expected to be similar for different DTA implementations, are very encouraging. Initial gaps two times smaller than the gap produced by a typical initialization of the DTA model in 2% of the time were observed, along with considerably lower gaps at convergence. This suggests that warm start solution approaches are feasible, and may lead to better and faster solutions for large networks, which is a crucial step towards facilitating the utilization of DTA models by metropolitan planning organizations (MPOs).