On the Solution to the Subproblems of a Globally Convergent SQP Algorithm for Nonlinear Programming

This work introduces a bound-constrained-based strategy for dealing with the quadratic subproblems of the sequentialquadratic programming (SQP) algorithm proposed by Gomes, Maciel and Martínez (1999). Two approaches for choosing the Hessian of the quadratic model are suggested. Numerical experiments are presented that illustrate the use of the proposed ideas.