A globally convergent SQP method for semi-infinite nonlinear optimization

摘要

A new approach for semi-infinite programming problems is presented, which belongs to the class of successive quadratic programming (SQP) methods with trust region technique. The proposed algorithm employs the exact L∞ penalty function as a criterion function and incorporates an appropriate scheme for estimating active constraints. It is proved that the algorithm is globally convergent under some assumptions. Numerical experiments show that the algorithm is very promising in practice.