Dynamic Systems Identification

Abstract

AbstractChapter 6 focuses on the identification of dynamic systems, both linear and nonlinear. The selected model structure of linear dynamic systems and, in particular, the structure of the noise model appear to be of crucial importance for specific applications and the estimation methods to be used. In this chapter, it is stressed that both the linear and nonlinear model structures can be formulated in terms of (nonlinear) regression equations, which allow a unification of estimation problems. Special attention is paid to subspace identification for the direct estimation of the entries of the matrices A, B, C, and D in a discrete-time, linear state-space model formulation, to the identification of discrete-time linear parameter-varying models of nonlinear or time-varying systems, to the use of orthogonal basis functions for efficient calculation, and to closed-loop identification in LTI control system configurations.KeywordsPrediction ErrorOutput ErrorInfinite Impulse ResponseHankel MatrixInterpretable Model StructureThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.