Robust adaptive regulation without persistent excitation

This paper presents a globally convergent adaptive regulator for minimum or nonminimum phase systems subject to bounded disturbances. The control strategy is designed for a particular input-output representation obtained from the state space representation of the system. The leading coefficient of the new representation is the product of the observability and controllability matrices of the system. The controller scheme uses a Least Squares identification algorithm with dead zone. The dead zone is chosen to obtain convergence properties on the estimates and on the "covariance matrix" as well. This allows the definition of modified estimates which secure well-conditioned matrices in the adaptive control law. Explicit bounds on the plant output are given.