Approximate Identification in view of LQG Feedback Design

Due to the modelling error a model-based controller generally works better with the model than with the modelled plant. This difference between the performances can be made small by selecting a model that is accurate at the closed loop relevant frequencies. In this paper it is shown that an iterative approach of identification and control design can lead to a model that is much better suited for feedback design than a model resulting from a plain open loop identification. In this iteration each identification is performed such that a certain closed loop criterion function is minimised. Each control design step employs the latest identified model to construct an LQG compensator. The performance requirements are gradually increased during the iteration.