论文部分内容阅读
根据随机用户均衡问题的特点构造一种基于BFGS校正公式和Armijo线搜索的截断拟牛顿法。介绍截断拟牛顿方程的构造过程及其算法的具体步骤;针对随机用户均衡模型的特点给出算法的收敛性和两个需注意的问题,并将此算法应用于一个路网。数值算例分析表明:所构造算法在迭代次数和误差方面均优于截断牛顿法,改进截断拟牛顿法可以避免二阶Hessian矩阵的计算,还可以用于某些Hessian矩阵不正定问题的求解。
According to the characteristics of stochastic user equilibrium problem, a truncated quasi-Newton method based on BFGS correction formula and Armijo line search is constructed. The construction process of truncated Quasi-Newton equation and the specific steps of its algorithm are introduced. The convergence of the algorithm and the two problems to be noticed are given according to the characteristics of stochastic user equilibrium model. The algorithm is applied to a road network. The numerical example shows that the constructed algorithm outperforms truncated Newton method in terms of iteration number and error. Improved truncated quasi-Newton method can avoid the computation of second-order Hessian matrix and can also be used to solve some indefinite Hessian matrix uncertainties.