Solving the logit-based stochastic user equilibrium using modified projected conjugate gradient method via convex model
Solving the logit-based stochastic user equilibrium using modified projected conjugate gradient method via convex model
复制标题
利用凸模型修正投影共轭梯度法求解基于 Logit 的随机用户均衡
DOI：
--
复制
发表时间：
2019
影响因子：
0.2
通讯作者：
Naihua Xiu
中科院分区：
文献类型：
--
作者：
Chao Zhang;Qian Zhang;Naihua Xiu
本文将经典的随机用户均衡（SUE）转化为一个凸模型，并提出了改进的投影共轭梯度（mPCG）方法来求解基于逻辑的SUE。我们证明了mPCG方法收敛于凸模型的唯一全局最小值。数值。在中等规模的Sioux Falls网络上进行了实验，将mPCG方法与求解凸模型的投影梯度法（PG）和求解非凸模型的逐次平均法（MSA）进行了比较。数值结果表明，所提出的凸模型mPCG方法优于MSA和PG方法。
In this paper, we convert the classic stochastic user equilibrium (SUE) into a convex model and propose the modified projected conjugate gradient (mPCG) method to solve the logit-based SUE. We show that the mPCG method converges to the unique global minimizer of the convex model. Numerical.experiments on Sioux Falls network of moderate size are done to compare the mPCG method solving the proposed convex model, with the projected gradient (PG) method for the convex model and the method of the successive averages (MSA) for the nonconvex model. The numerical results indicate that the proposed mPCG method for the convex model outperforms the MSA and the PG method.