A robust SQP method based on a smoothing approximate penalty function for inequality constrained optimization. (Q1607579)

The general inequality constrained nonlinear programming problem is considered where both the objective function and the constraints can be nonlinear. The paper presents a solution algorithm based on a sequential quadratic programming approach. The merit function it uses to determine the steplength is easy to compute thus it is quite suitable for practical computations. The algorithm automatically adjusts penalty parameters. It can also efficiently cope with the possible inconsistency of the quadratic search problem.