A method for solving inequality constrained minimization problems is described. The algorithm is based on a primal-dual interior point approach, with a line search globalization strategy. A quasi-Newton technique (BFGS) with limited memory storage is used to approximate the second derivatives of the functions. The method is especially intended for solving problems with a large number of variables with bound constraints and a medium number of general inequality constraints. Some numerical experiments are presented to validate our approach.
Rapport de recherche du LACO 2003-08, LACO, Universite de Limoges, Faculte des Sciences et Techniques, 123, avenue Albert Thomas, 87060 Limoges (FRANCE)