瞬时动态用户最优决策点逆序分配方法研究

Instantaneous Dynamic User Optimal Assignment in Reverse Order of Decision-nodes

以基于瞬时路段阻抗的变分不等式模型的对角化算法为研究对象,指出若在一次迭代中完成所有时段的分配,其结果不能直接用于下一次迭代对应时段的流入率调整,对其产生的原因进行了分析,并根据瞬时动态用户最优网络的特性,阐述了时段可分离性和决策点可分离性,提出了基于时段推进的决策点逆序分配方法。结合与变分不等式模型等价的非线性规划模型,分析了该方法的原理并给出了逆序分配的步骤和示例。结果表明,该方法的核心是可用Frank-Wolfe算法求解的在各决策点上进行的各自紧后路段间的分配,决策点紧后路段之后的子路径最小瞬时阻抗的确定和一些决策点组合的同步分配都可以提高分配的速度。

The object was to solve the problem imbedding in Ren＇s diagonalization algorithm of variational inequality model based on instantaneous link travel times. The analysis showed that if the assignment was carried on for all time intervals at one iteration, the results could not be used in the inflow-rate adjustments at the next iteration. According to the separabilities of both time intervals and decision-nodes derived from the inherence of instantaneous dynamic user optimal network, an approach of assignment in reverse order of decision-nodes was proposed. The principle of this method was analyzed based on the equivalent non-linear programming of the variational inequality model. The results indicate that the kernel of the approach can be regarded as distributed assignment at each decision-node among its adjacent links by using Frank-Wolfe algorithm, and that the assignment speed can be improved