SELF-TUNING CONTROLLERS BASED ON A FIXED LENGTH DATA WINDOW.

Abstract

Self-tuning controllers (STC) are designed based on constant but unknown plants. However, because the controller contains an estimation algorithm and continuously adjusts its parameters, it can be used to control slowly time-varying plants. In this paper, a STC for a fixed length data window is proposed. Unlike the conventional least squares estimation algorithms, the proposed STC updates the UD factors of an extended information matrix and then computes the current parameter estimate by inverting the U-factor of the extended information matrix. The number of AO required is approximately 1. 6 times that for the one using an exponential forgetting factor. This is an approximately 20% improvement over conventional estimation algorithms.