Model reference adaptive control using only input and output measurements

Abstract

A new method of designing model reference adaptive controllers using only input and output measurements is presented. The adjustable parameters are all contained in the control input signal to the process. State variable filters are used to avoid taking pure derivatives of the process output while the hyperstability theorem ensures global stability of the parameter adaptations. The control input can also contain a term to diminish the effect of external load disturbance. The natural extensions to multivariable and certain non-linear systems are then presented. The performance of the adaptive controller in the presence of noise can be further improved by the use of the instrumental variable approach together with an adjustable series compensator. The theoretical development is substantiated by extensive simulation results.