SQP技术与广义投影相结合的次可行方向法

本文建立非线性不等式约束优化的一个新算法，分析和证明了算法的整体收敛性和超线性收敛性。其技巧在于将广义投影和ＳＱＰ技术结合使用。

A SUBFEASIBLE DIRECITON METHOD COMBINING SQP TECHNIQUE WITH GENERALIZED PROJECTION

In this paper, a new algorithm for nonlinearly inequality constrained optimization ispresented. Its skill is to combine generalized projection with SQP technique. This algo-rithm possesses the following important advantages: (1) The starting-point is arbitraryand penalty functions and parameters are not required. The other iterated points will be-come feasibly descent whenever some iterated point has got into the feasible region, so it iscalled subfeasible direction method. (2) Only one quadratic programming is to be solved ateach iteration, and then the search direction will be generated by making a simple correc-tion for its K-T points with the help of generalized projection. (3) It does not use the aux-iliary remedy of solving linear programming. Hence this algorithm is very simple and com-pact, needs only a small amount of computation, but possesses the global and superlinearconvergence.