A Recursive Method for System Identification Using Adaptive Orthogonalization

概要

論文の詳細を見る
A fast recursive method has been proposed to identify the dynamics of a linear system. The characteristics of the method was shown by computer simulation as well as qualitative investigation.There are some robust recursive methods for single input, single output, linear system identification using the vector model of discrete weighting function representation, such as the learning identification method. Although this method is stable against noise, it is not so fast at decreasing rate of identification error especially when the input is not white.This defect could have been improved by Gauss-Jacobi method or Schmidt's orthogonalization etc., if it were perfectly noise-free. However, Schmidt's orthogonalization alone would make the identification process unstable in practical situation. This problem in adopting orthogonalization was settled by the proposed recursive identification method with adaptive nonlinear error-correcting term.
公益社団法人計測自動制御学会の論文