Using S-estimators in parameter identification

This paper describes an improved S-estimator using the fast-S formulation and the standard Least Squares identification, aiming at realtime usage in a parameter identification setting. The improvement step uses a recursive version of the Least Squares algorithm. Experiments with simulated and real data are presented. The new algorithm is shown to have the outlier rejection properties of state of the art S-estimators while used in the identification context.