Handling hard constraints in infinite horizon predictive control

One of the strengths of predicted control is its ability to handle constraints on line. However, general stability results only exist for a special class of predictive controllers, those with endpoint constraints which themselves can restrict the degrees of freedom for performance or those which use infinite horizons. Here we present two techniques that can be used for online handling of hard constraints over an infinite horizon. The first is based on an extension of admissible sets to handle a nonlinear predictive control law and the second, which is more efficient, is based on the computation of constraint horizon beyond which constraints cannot be violated.