ON THE ROBUSTNESS OF DISCRETE-TIME INDIRECT ADAPTIVE (LINEAR) CONTROLLERS

Remarkable research activity has been devoted to the design of certainty equivalence-based adaptive controllers which perform well in those ubiquitous non-idealities as bounded disturbances, time varying parameters and some classes of unmodeled dynamics. The long-standing issue is the estimated model admissibility condition; i.e. the underlying linear control law should stabilize the estimated model. The motivation of this paper is to propose a general framework for robustly designing the adaptive linear controllers, irrespective of the underlying control law. More specifically a new solution to the problem of the estimated model admissibility using an ad-hoc modification of the control law is given. Such a modification consists in adding an internal impulse exciting sequence, while freezing the controller parameters, whenever the estimated model admissibility is lost.