Adaptive control algorithms for disturbance rejection (Q1176077)

The paper considers several variants of adaptive control algorithms based upon the internal model principle. No new results are described, but rather a comparison of different approaches is given and some simulation experiments are discussed. It is a useful paper for those interested in the internal model principle and its application in adaptive control formulations. The idea behind this paper and several others in this field is that disturbances may, at least approximately, be modeled by the solution to a homogeneous difference difference equation of the form \(D(q^{- 1})d(k)=0\). For disturbances which are bounded and non-vanishing, it is obvious that at least some of the roots of \(D(\cdot)\) must lie on the unit circle in the complex plane. Since the ideal control law involves the filter \(D(q^{-1})^{-1}\), the existence of these ``marginally stable'' roots can cause difficulties in implementation since large control values may be obtained. In the adaptive laws described in this paper, the system parameters and the filter \(D(q^{-1})\) are estimated on-line, and then a control is chosen based upon these estimates. Because of the aforementioned difficulty caused by division by \(D\), a projection is typically employed to ensure that \(D^{-1}\) is a stable filter. Stability of the algorithm is not ensured by this projection, since the estimates vary with time. The simulations described in Section 4 are useful, if brief. A significant gap in this section of the paper is that the authors do not describe the behavior of the control sequence, which is probably sensitive to disturbances and the dynamics of the reference input.