1: \begin{abstract}

2: A polyhedral active set algorithm PASA is developed for solving a

3: nonlinear optimization problem whose feasible set is a polyhedron.

4: Phase one of the algorithm is the gradient projection method,

5: while phase two is any algorithm for solving a linearly constrained

6: optimization problem.

7: Rules are provided for branching between the two phases.

8: Global convergence to a stationary point is established,

9: while asymptotically PASA performs

10: only phase two when either a nondegeneracy assumption holds,

11: or the active constraints are linearly independent and a strong

12: second-order sufficient optimality condition holds.

13: \end{abstract}

14: