LINEAR REGRESSION APPLIED TO SYSTEM IDENTIFICATION FOR ADAPTIVE CONTROL SYSTEMS

Abstract

The purpose of this paper is to describe a method of process identification using a linear regression technique and to indicate how this method may be applied to adaptive control systems. The unknown system parameters are considered as additional state variables. Estimates of system parameters as well as the system state are made from noise-contaminated data. Differential equations with random forcing functions describing the parameter variations are adjoined to the system of differential equations describing the process. The estimation of the error is assumed to propagate linearly about the current estimates, which are updated as new data are received.