An ordinal optimization framework for robust design

We consider systems with parameter uncertainty and suggest an ordinal optimization framework for robust design. To illustrate the approach we focus attention on the problem of robust stability margin design. Our formulation exploits frequency domain properties and polynomial value set ideas. A parameterization of stable polynomials is used to characterize the controller search space. Robust control design is then carried out by: 1) employing easily computable stability margin lower bounds to estimate the performance and 2) using ordinal optimization to deal with the lack of known structure in the search space. Preliminary numerical results are provided to illustrate the framework.