Identificação adaptativa de sistemas atraves de minimos quadrados lineares utilizando fatoração QR e janela movel de dados

AUTOR(ES)
DATA DE PUBLICAÇÃO

1993

RESUMO

This work deals with the numerical solution of linear least-squares problems for adaptive identification of time-varying systems. Traditional recursive algorithms using normal equations and exponential data weighting are analysed and their numerical drawbacks for solution of ill-conditioned least-squares problems are pointed out. It is proposed a new recursive algorithm using orthogonal QR factorizations with sliding window on the data, numerically stable, being particularly recommended for solution of ill-conditioned problems. Significative simulation results are presented, illustrating the performance of the following algorithms identifying an ARMA model : normal equations with exponential data weighting, QR factorization with exponential data weighting, normal equations with sliding data window and the proposed algorithm using QR factorizations with sliding data window

ASSUNTO(S)

estimativa de parametro algoritmos minimos quadrados fatoração (matematica)

ACESSO AO ARTIGO

Documentos Relacionados