An on line identification method using instrumental variables and deconvolution

A method is proposed in this paper for on line estimation of the ARMA model parameters of a system, whose input is white unmeasurable, stationary, or non-stationary noise. The method is based on the instrumental variable principle for recursive parameter estimation and on fixed-lag smoothing of the equivalent state -space model for deconvolution. The consistency and asymptotic behaviour of this approach are discussed and a simulation study is given, which supports the theoretical results.