Literature Review of the 'Errors-In-Variables' Approach to the Problem of System Identification.

摘要

The papers reviewed here deal with various aspects of the problem of system identification when all the observed variables are contaminated by noise errors. Theoretical issues of identifiability and practical methods for developing algorithms to estimate parameters of a dynamic system are considered. The total least squares approach and the Koopmans-Levin method are among the methods discussed. Both these methods are strongly based on matrix singular value decomposition, which is a computationally robust numerical technique. A book, which is a compilation of papers originating from a workshop on singular value decomposition and signal processing is included in the references. System identification, Parameter estimation, Noise modulation, Dynamical systems, Algorithms.