Parameter robust control design based on parametric identification quality

In this paper, we deal with the problem of designing a feedback controller for a linear discrete time plant that is liable to perturbations of a physical parameter vector. The new robustness measure is based on the link between the poorest quality of the closed loop parameter identification and the performance and stability robustness for regulators. Reversely, this means that the better the robust control, the worse the closed loop identification quality. Our first objective is not to identify unknown parameters but to derive an analytical expression of the covariance matrix of a bayesian estimator. We obtain then the controller gain matrix which would minimize this quality if the identification algorithm was implemented. Important relationships between PRLQG method and this new technique called PRCBI are obtained and it is shown that the PRCBI is a generalization of the PRLQG method. Furthermore a flexible structure application shows that the best results are achieved using the PRCBI method.