Sequential Quadratic Programming Based on IPM for Constrained Nonlinear Programming

The field of constrained nonlinear programming (NLP) has been principally challenging to various gradient based optimization techniques. The sequential quadratic programming algorithm (SQP) that uses active set strategy in solving quadratic programming (QP) subproblems proves to be efficient in locating the points of local optima. However, its efficient determination of the optimal active set heavily relies on the initial guess of the starting point. This remains a serious drawback to both primal and dual active set approaches especially for NLPs with several inequality constraints. Thus, we propose a sequential quadratic programming algorithm (SQP/IPM) which uses an infeasible interior point method (IIPM) for the determination of descent directions. We propose using quadratic search algorithm for effective minimization of merit functions. Our test results reveal that SQP/IPM algorithm is efficient and promising.