The design of a feedforward compensator for robust H-infinity or Ha performance under structured uncertainty is considered. For Linear time-invariant uncertainty, a convex method based on linear matrix inequalities (LMI＇s) across the frequency variable is given. For nonlinear or time-varying perturbations and H-infinity performance, the design problem is reduced exactly to a state-space LMI, and extensions to H-2 performance are discussed. An example illustrates the application of these techniques to two-degree-of-freedom control design.