A modified EW-RLS algorithm for systems with bounded disturbances

Abstract

This paper presents a new estimation algorithm, which is essentially a modification of the exponentially weighted recursive least squares algorithm (EW-RLS), for systems with bounded disturbances. Assuming knowledge of a disturbance upper bound, the algorithm's derivation is performed by minimizing a cost function weighted by two factors: one is fixed by the user and exponentially weights the arriving information, the other is time-varying and data-dependent. The purpose of the latter is to ensure algorithm stability, faced with the bounded disturbances. From this proposal results a very simple algorithm which completely depends on the significant information and stops when the data are meaningless. It is also shown that if the signals are persistently exciting, the convergence of the algorithm is exponentially fast. An application of the algorithm to the problem of estimating the friction forces in servo-mechanisms (under structural bounded errors) is also presented. © 1990.