On the convergence of a model reference adaptive control algorithm with unknown high frequency gain (Q802514)

The model reference adaptive scheme proposed by \textit{A. S. Morse} [IEEE Trans. Autom. Control AC-25, 433-439 (1980; Zbl 0438.93042)] is examined for exponential convergence of parameter estimates. It is revealed that the high frequency gain estimate is not exponentially stable because of the loss of persistent excitation when the relative degree of the plant is greater than one. It is known [see the author, Automatica 21, 247-258 (1985); the author and \textit{R. M. Johnstone}, When will adaptive systems really adapt? The robustness issue, Proc. 2nd Conf. on Control Eng. I.E., Aust. 1982, p. 59-66 (1982)] that adaptive algorithms without exponential convergence exhibit unacceptable behaviour in the presence of noise and modelling inadequacies. Whether or not the lack of exponential stability applies to other algorithms is an open problem. However, the authors expect that the conclusion is general for plants with relative degree greater than one.