Model predictive control using segregated disturbance feedback

This paper proposes a new control parametrization under the Model Predictive Control frame-work for constrained linear discrete-time systems with bounded disturbances. In an effort to minimize conservatism, the proposed parametrization takes the form of a piecewise affine disturbance feedback and is a generalization of linear disturbance feedback that has appeared in recent literature. Numerical computations and stability properties of the resulting MPC problem using the proposed parametrization are discussed. When the disturbance belongs to an absolute set and the problem data satisfy reasonable assumptions, the associated finite-horizon optimization can be computed efficiently and exactly. The generality of absolute set for disturbance modelling is also discussed. The advantage of the proposed parametrization over linear disturbance feedback is illustrated via numerical examples. I