| dc.description.abstract | Dynamic user equilibrium problem explored travelers’ choice behavior involving temporal dimension consideration, which has been tackled for years. Contrast to link flow as the decision variable in static traffic assignment, three variables are identified in the dynamic user equilibrium problem, i.e., link inflow, number of vehicles and exit flow. However, by concept of flow propagation, these three flow variables can be transformed each other. For example, the number of vehicles and exit flows can be deemed as subsequent states of link inflows. With only one decision variable adopted, the structure of dynamic user equilibrium model can be significantly simplified.
Based on the results that have been developed so far, this study attempts to strengthen the applicability of the dynamic user equilibrium model by taking three more issues into study. The first issue is to relax the assumption of single class of travelers, thus yielding a multiclass or multicriteria dynamic user equilibrium model.
The second issue is to retrospect the link capacity requirement from the practical point of view. In the past, either link inflow or exit flow was taken for link capacity constraint in the dynamic user equilibrium problem. Here, we take the number of vehicle (or equivalently link queue length) as the link capacity constraint instead, and solve it accordingly by the penalty and augmented Lagrangian methods. The results show that under the assumption that time dependent network constructed by the actual link travel times, the outputs obtained from the link capacity constraints formed by link inflow, exit flow or the number of vehicles are basically comparable.
The last issue is to deal with link capacity reliability analysis. The framework is mainly borrowed from Chen et al. (2002), which involves sensitivity analysis embedding a generalized inverse approach, uncertainly/risk analysis within a Monte Carlo simulation. This systematic procedure is general in that the reliability on connectivity and travel times is constructed as special cases. The findings associated with the three issues are summarized and few remarks are given in the end to conclude the research. | en_US |