A generalized gradient method for nonlinear constraints and its convergent properties

Although many effective algorithms of the gradient projection type have been proposed for nonlinear programming with nonlinear constraints, these algorithms must require a regular assumption (i.e. the gradient vectors of the active constraints set at any feasible point are linearly independent). In this paper, a generalized gradient projection method is presented for nonlinear programming with nonlinear constraints under degeneracy, and its convergent properties are discussed.