LQG self-tuning controller with dynamic control and estimator weightings

The problem of the convergence of LQG self-tuning controllers using dynamic weighting factors is considered. The effects of these factors on the identification algorithm performance are discussed and a convergence proof is given using the concept of the margin of strictly positive realness (SPR). A general LQG self-tuning algorithm is then proposed which has the superior properties of efficient information processing, improved parameter convergence rate and a tuning mechanism to limit the input variations at the initial stage of the self-tuning process. The algorithm is also suitable for estimating fast parameter variations. The technique is shown to be applicable to multivariable LQG self-tuning control by using the margin matrix of SPR. Simulation results are presented to show the performance of the algorithms