Abstract: A model predictive control process which approximates a constraint set as an ellipsoid to directly address actuator limitations, to provide a robust controller structure, and reduce computation demands. A bounded polytope defining the constraint set is approximated as an ellipsoid in an off-line computation. The ellipsoid is scaled on line via bisection to determine the ellipsoid that optimally fits the polytope, calculating a set of control moves for manipulating device actuators. Preferably, an unconstrained solution is first computed and tested to see if any constraints are violated. If none are violated, it is accepted, if one or more are violated, the scaling is reconducted as a constrained solution in iterative fashion until the solution yields an active constraint.