Problem Solving Constrained SQP Adaptive Filter Algorithm

SQP algorithm is one of the most effective methods for nonlinear programming and used widely. The method generally involves four aspects in processing. They are the positivity definiteness of the Hessian matrix, the compatibility of QP sub-problem, the feasibility of the initial points, and Maratos effects. The SQP algorithm is usually combined with filter to overcome the difficulty of selecting a suitable penalty factor. In this paper, we make some modifications on the traditional SQP algorithm. First, we use the information of the function value to approximate the Hessian matrix. Second in order to to modify QP sub-problems,we increase the tolerance of sub-problem. Third, we revise the acceptance criteria of the filter by adapting technique. In addition, we process QP sub-problems, which is not feasible, and present the restoration process in detail. The result of the experiment data suggests that the new algorithm not only has an efficient convergence,but also feasible.