Error Bounds for Model-Plant Mismatch in Identifiers and Adaptive Observers.

摘要

The stability properties of six different adaptive schemes with respect to model error are analyzed. Bounds on parameter identification and state errors are established. All adaptive schemes considered are robust in the sense that the error is of order of the speed ratio micron between the modeled slow phenomena vs. the neglected fact. The dependence of the error on the input signal is shown to be crucial. The bounds obtained indicate possibilities for reducing the error by a proper choice of the input signal. (Author)