摘要
对于非线性约束的优化问题.最近给出的各种SQP算法均采用罚函数技巧以保证算法的全局收敛性,因而都必须小心地调整惩罚参数。本文给出一个不依赖于惩罚参数、每步迭代的校正矩阵也不需正定且仍具有全局收敛性的SQP方法,而且罚函数形式简单、具有和约束函数同阶的光滑性.
In order to guarantee the global convergence, the existing SQP methods for the nonlin-ear constrainted optimization problems all use the technique of penalty functions,so they must correct the penalty parameter carefully. In this paper, we present a SQP method which not only does not depend on the penalty parameter, but also possesses global con-vergence property. Moreover, the penalty functions are of simple from and of the same smooth order as the constraint functions.
基金
广西大学科学基金资助课题
