Constrained multivariable cautious stable predictive control

Necessary and sufficient conditions for the stability of predicted input/output trajectories form the basis of predictive control algorithms with guaranteed stability; the proof of stability derives from the fact that the resulting optimal cost of performance behaves as a stable Lyapunov function. While the single-input single-output case is in the open literature, this paper extends these concepts to the multivariable case and develops an algorithm which utilizes the largest possible class of stable predictions for a given number of degrees of freedom.