A Smoothing Active-Set Newton Method for Constrained Optimization

摘要

In this paper, a smoothing Newton-type method for solving constrained optimization is presented. This method uses smoothing Newton method to solve the KKT system of the original problems. A kind of active-set technology is introduced to identify the inequality constraints which does not satisfies the strict complementarity conditions. The proposed method has the global convergence to the KKT point. Under some regularity assumption, the degenerate indices can be identified correctly near the solution, and local super linear convergence is obtained without strict complementarity.