Multi-model parameter estimation scheme for self-tuning controllers

Abstract

A multi-model parameter estimation scheme (MMPES) based on a modification of the gaussian sum approach is proposed and its application in self-tuning it-incremental controllers is presented. As shown in the simulation examples, the MMPES gives a fast retuning of the estimated parameters when there is a parameter change; and by preserving the optimal model, it avoids the corruption of the estimated parameters when there is a load change. Thus it is more robust than existing techniques such as the variable forgetting factors. The only penalty in the MMPES is the increase in computing load, yet the proposed MMPES requires substantially less computing effort than that proposed by Andersson (1985).