A numerically robust state-space approach to stable-predictive control strategies (Q1129658)

The stable general predictive control (SGPC) transfer function framework is extended to a state-space representation which has significant computational and numerical advantages in the case of an open-loop unstable plant. The key idea is prestabilization of the plant before the predictions (for the SGPC algorithms) are computed. Prestabilization ensures stable and bounded predictions which improves numerical conditioning hence the use of matrices with very large elements associated with unstable predictions is avoided. Prestabilization is later removed through optimization. It is shown that the proposed alternative to a SGPC algorithmic formulation for state-space control laws does not suffer from poor conditioning and meets most terminal constraints imposed by the SGPC approach.