MODIFIED RECURSIVE LEAST-SQUARES ALGORITHM FOR PARAMETER-IDENTIFICATION

Abstract

A modified recursive least-squares algorithm is proposed for estimating the parameters of a linear system, including ill-conditioned situations. The proposed algorithm is obtained by minimizing the modified cost function in which the initial guess in the cost function of an ordinary recursive least-squares (RLS) algorithm is replaced by the latest estimate. Analysis of the convergence shows that the proposed algorithm has a better convergence rate than the ordinary RLS algorithm. The extension of the algorithm to the time-varying case has also been considered.