ABSTRACT
The closed-loop paradigm (CLP) is an alternative mechanism for implementing Model Predictive Control algorithms which can have advantages. The CLP was originally proposed as part of an algorithm stable generalised predictive control but can be abbreviated hereafter as stable predictive control to take account of other strategies. The beauty of the CLP approach is that the d.o.f. can be set up as perturbations about the most desirable performance, that is, the performance that arises from unconstrained control. An equivalent of No terminal control for unstable processes was developed independently by several authors, some using the CLP approach with transfer functions and some using the open-loop paradigm approach with state-space models. The advantage of the CLP is that it uses stable predictions in mode 1 and mode 2 and hence the prediction matrices do not contain coefficients with large variations in magnitude. This helps avoid issues with round-off errors as well as reduces prediction mismatch.
