LQG adaptive control of ARMAX plants by LS multistep predictors

An adaptive LQG (Linear-Quadratic-Gaussian) self-tuning controller for ARMAX (AutoRegressive Moving Average) plants is developed. Its feedback updating scheme is based upon an RLS (Recursive Least Squares) parameter estimation of T+1 independent output predictors, T??n, n being the plant order. The seemingly heavy numerical burden is actually limited by the common updating gain shared by all estimates in the independent predictors. The proposed regulator turns out to be a refinement of the MUSMAR self tuner that has proved to posses attractive properties of robustness.