Convergence analysis of some decentralized parameter estimators

Parameter estimators based upon standard estimation techniques, viz, least squares (LS) methods, instrumental variable methods, etc. may occasionally have difficulties with systems having a somewhat ill-conditioned nature, e.g., stiff systems. The robustness of the estimator can, however, be significantly improved by employing some kind of decentralized estimation techniques. The method presented here utilizes certain moving average filters and a system of parallel estimators, where the moving average filters are used to prefilter the input-output data. The form of these filters depends upon the estimated model, and the decentralized estimator will therefore usually have to be employed in a bootstrap fashion. The convergence properties of this bootstrap estimator are analysed, and necessary and sufficient conditions for local convergence of some decentralized estimators based upon LS or IV techniques are derived.

Abstract licence: CC BY

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.