共轭投影梯度滤子算法及其全局收敛性

摘要

A novel SQP-filter method is presented by combing the concept of conjugate projection gradient. The algorithm doesn't solve a QP sub-problem by the introduction of filter. With the idea of conjugate projection gradient. this method is effective to avoid the restoration algorithm in the general filter algorithms. In some conditions, its global convergence is proved.%提出了一种基于共轭投影梯度思想的SQP-滤子算法,滤子技术的引入使得该算法不需要求解二次规划子问题.借助于共轭投影梯度思想能有效避免常规滤子算法中的恢复算法,简化了算法结构.在一定条件下,证明该算法具有全局收敛性.