Accelerating the rate of convergence for LMS-like on-line identification and adaptation algorithms. Part 1: Basic ideas

Abstract

In the paper a modification enabling acceleration of the rate of convergence for LMS-like on-line identification and adaptation algorithms is proposed. This is based on an artificial decaying of initial conditions in recursive identification as well as adaptation algorithms. The decaying is done using a set of the most recent measurements. Properties of the algorithms with the proposed modification are compared with non-accelerated identification and adaptation algorithms in simulations of a practical adaptive control system.