Switching supervisory predictive control of input saturated plants

A predictive control scheme based on a blend of receding-horizon plus switching-control ideas is presented for regulating linear plants subject to input saturation constraints. While only local asymptotic regulation can be ensured in general, semi-global exponential regulation is achieved for linear plants which are free of exponentially unstable modes. The overall scheme is based on a fixed finite set of control effort weights so as to ensure stability and constraints fulfilment when they are used for LQ control and the state of the plant belongs to a suitable bounded region of the state space. A logic is then used to select online the most appropriate control weight according to a state set-membership criterion.