Abstract: In this work we propose a primal-dual algorithm for the
solution of equality constrained optimization problems with
bounded variables. The core of the method is a local
algorithm, wich relies on a truncated scheme for the
computation of a search direction in the product space of
the problem variables and of the multipliers associated with
the equality constraints. The local algorithm is globalized
by means of a mixed penalty augmented lagrangian merit
function, where the bound constraints are treated by an
exact penalty approach and the equality constraints are
treated by an exact augmented Lagrangian approach. The
resulting algorithm is globally and superlinearly convergent
under mild assumptions. Due to the theoretical properties of
the exact merit function, the algorithm can be fruitfully
employed to search for
global solutions of constrained problems.