Model reference adaptive control using a low-order controller

In the model reference adaptive control problem, the goal is to force the error between the plant output and the reference model out put asymptotically to zero. The classical assumptions on a single-input-single-output (SISO) plant is that it is minimum phase, and that the plant relative degree, the sign of the high-frequency gain, and an upper bound on the plant order are known. Here we consider a modified problem in which the objective is weakened slightly to that of requiring that the asymptotic value of the error be less than a (arbitrarily small) pre-specified constant. Using recent results on the design of generalized holds for model reference tracking, here we present a new switching adaptive controller of dimension two which achieves this new objective for every minimum phase SISO system; no structural information is required. Copyright (C) 2001 John Wiley & Sons, Ltd.