Towards computational complexity certification for constrained MPC based on Lagrange Relaxation and the fast gradient method

This paper discusses the certification of Nesterov's fast gradient method for problems with a strongly convex quadratic objective and a feasible set given as the intersection of a parametrized affine set and a convex set. For this, we derive a lower iteration bound for the solution of the dual problem that is obtained from a partial Lagrange Relaxation and propose a new constant step-size rule that we prove to be optimal under mild assumptions. Finally, we apply the certification procedure to a constrained MPC problem and show that the new step-size rule improves performance significantly.