A parametrization for the closed-loop identification of nonlinear time-varying systems (Q674926)

The paper is devoted to a parametrization for the closed-loop identification of nonlinear time-varying systems, generalizing the parametrization idea to nonlinear plants, and treating identification of nonlinear plants. Closed-loop, as opposed to open-loop, identification is considered. The starting point is the assumption that one knows a controller that stabilizes the unknown plant to be identified. The characterization of the model set in a noise-free setting is considered. Characterizations on the basis of both left- and right-coprime factorization descriptions of the controller are presented. A novelty is the characterization of the plant model set in terms of the left-coprime factors of the controller. Further, the two characterizations are modified to incorporate the measurement noise term. It is demonstrated how the models can be exploited to meet identification objectives under a high signal-to-noise ratio assumption. Considering the closed-loop identification of nonlinear plants stabilized by linear, albeit time-varying feedback, the main focus is on deriving factorization-based structures that help convert the underlying closed-loop identification problem to one that is essentially one of open-loop identification.