Uncertainty evaluation for estimates from recursive projection algorithms

摘要

AbstractIn this paper we consider parameter estimation of linear systems described byyi=a −Tiθ +ei, where the ith measurementyiis linearly dependent on the parameter vector θ ε ℝ−pthrough the regressor vectora −Tiε ℝ−pand the measurement erroreiis unknown but bounded.Some properties of previously presented algorithms for recursive parameter identification in the unknown but bounded error (UBBE) context are discussed. In particular it is analysed how different levels of information on the error structure can influence the choice of the identification algorithms and the possibility of evaluating the reliability of the estimates. Attention is also focused on the influence that forgetting schemes have on the estimates and on their co