Robustness analysis of identification and adaptive control for stochastic systems (Q1094367)

This paper examines the question of parameter identification and robust tracking in the presence of unmodelled dynamics. A key assumption, in addition to the usual persistence of excitation assumptions, is that the unmodelled dynamics are essentially deterministic with respect to any current measurement. In addition, in order for the results to be of value, the unmodelled dynamics must represent a ``small'' perturbation to a linear model for the observed process. The paper uses previously obtained error estimates for ``exact'' models, to obtain error bounds on estimation and tracking errors in separate analyses. The main conclusion in each case is that the error bounds are proportional to an estimate on the size of the unmodelled dynamics, so that as this estimate tends to zero, the effects of the unmodelled dynamics disappear. This result is intuitively simple, but certainly warrants proof. The paper is precise and well-written.