Abstract
While static traffic assignment model cannot adequately depict real network, dynamic traffic assignment model is too complex and requires detailed Origin-Destination (OD) data. As an effective combination, the semi-dynamic model with flow propagation is used. In semi-dynamic traffic assignment model with endogenous traffic congestion, not only is the static equilibrium reached in each period, but flow propagation also is considered. Some of the travellers cannot exit the link and are considered as residual flow of that link which is propagated to the next period. And, the present period experiences the remove of residual flow on subsequent links. The problem is that if the residual flow for flow propagation is eliminated, not only the travel cost changes but also the inflow changes via network equilibrium. Sensitivity analysis is used to solve this problem. The semi-dynamic traffic assignment model with flow propagation based on sensitivity analysis was proposed with route-based approach. However, because of route enumeration in the sensitivity analysis formulations, applying the proposed method to very large-scale network has been in difficulty. In fact, the extremely large alternative routes connecting an OD pair in a real network can cause overloaded calculation. In this study, thus, we will upgrade this method to solve this problem to increase the applicability of the semi-dynamic Stochastic User Equilibrium traffic assignment model. This research will apply, therefore, link-based procedures which perform only link and node variables. This method is developed from STOCH3 algorithm and Depth First Search (DFS) algorithm. The goal of our research not only reduces computation time but also reduces the need of computer memory and increases applicability of the semi-dynamic traffic assignment model. The effectiveness of the calculation method will be illustrated by an application to Kanazawa City road network.
