A NEW EMBEDDING FOR THE AUGMENTED LAGRANGE METHOD

Abstract

Several algorithms such as penalti, barrier, Augmented Lagrangean and parametrical approaches are used in the solution of non-linear optimization problems. One of these approach construct of each optimization problem (P) or the results of some iterative algorithms. For this parametrical problem a necessary condition for a good behavior of the continuation or to define jumps is that the parametrical problem is JJT – regular. In this work we propose an embedding for the Augmented lagrangean Method, using the ideas of Bertsekas for this kind of algorithms and we proof that for almost every parameter, fixed the original optimization problem, the constructed parametric problem is JJT – regular. Some numerical examples to illustrate the solution are presented