A modified maximum principle for optimum control of systems with bounded phase space coordinates