非平衡交通分配的拟Frank-Wolfe迭代算法

提出了基于最短路动态生成的一种新的非平衡交通分配迭代算法.在每轮迭代中,将按全有全无方法在当前最短路上分配的交通量与前一轮迭代所得到的交通量加权组合,而各O-D对的加权系数则依据Logit原则来确定.和Frank-Wolfe算法不同,不必通过一维搜索确定加权系数.同时又避免了Logit方法要求枚举所有路径的困难.本文还证明了算法的收敛性,而计算实例显示,由本算法所得结果与平衡交通分配非常接近,因而它是一个高效而可靠的交通分配算法,适用于大、中型道路交通网络的交通分配计算.

A Quasi-Frank-Wolfe Algorithm for Non-equilibrium Traffic Assignment Problem

We present a new algorithm for non-equilibrium traffic assignment problems based on the dynamic generating of the shortest paths between every O-D pair.In each iteration,the traffic volumes on each path are updated by weighted summing the traffic volume assigned to current shortest paths and the volumes obtained in early iteration.By contrast to Frank-Wolfe algorithm,it doesn′t need the line search procedure to obtain the weighing coefficients.It is superior to Logit method by avoiding the paths numeration.Both the convergence result and the numerical example show this algorithm is effective and efficient.