Abstract : This report describes accomplishments in developing methods of system identification for robust control design. The starting point is an a priori plant description containing both parametric and nonparametric uncertainty. The identification methods are developed under differing a priori assumptions on the parametric and nonparametric parts of the model set. For example, when a bound on the nonparametric part is known, it is shown that the parameters in the parametric part of the model are contained in either an ellipsoid or hyperboloid, depending on the data. Computational methods are very similar to standard least-squares methods and can be computed in a batch or recursive manner. The parameter set membership description is used for robust control design via a mini-max optimization problem. Other approaches explored include high-order ARX models which produce purely parametric uncertainty under standard statistical assumptions on the disturbances. A learning scheme is also investigated where the control and identification are iteratively coupled by the closed-loop.