Author's: Zhijun Luo and Guohua Chen
Pages: [109] - [120]
Received Date: January 3, 2009
Submitted by:
In this paper, a new modified SQP method is proposed to solve the inequality con- strained optimization. This algorithm starts from an arbitrary initial point and adjusts penalty parameter automatically. A descent direction is obtained by solving only one modified QP subproblem. In order to avoid Maratos effect, a height-order correction direction is computed by an explicit formula. Thanks to the height-order technique, under mild conditions without strict complementarity, the global and local superlinear convergence properties are obtained.
inequality constrained optimization, SQP algorithm, global convergence, superlinear convergence.