Reducing the computational effort of MPC with closed-loop optimal sequences of affine laws