9308917 Smith This work addresses the problem of modeling systems in a robust control framework. Analysis and synthesis in this framework are relatively well developed. However, the lack of a compatible modeling and identification theory is slowing the practical application of robust control. The problem to be addressed is the following: How accurately must a system, or component of a system, be modeled in order to maintain stability and achieve a specified closed loop performance objective? This problem is generalized to the case where other system components themselves are uncertain (to within known bounds). This immediately leads to the problem of connecting the required accuracy with the taking of experimental data. An iterative identification/modeling approach is proposed for both experimental and theoretical study. The models considered are standard in the robust control framework: linear, time-invariant with Hoo norm bounded perturbations. Closed loop performance is specified as a general weighted Hoo criterion. Progress on these problems will provide a more direct connection between the rich theoretical area of Hoo robust control and its application to practical problems. This will also lead to a better understanding of what types of physical uncertainties can be usefully modeled as linear fractional perturbations. ***
