Robust control under parametric uncertainty via primal-dual convex analysis

A numerical method is proposed for optimal robust control synthesis. The method applies to the case when the coefficients of the characteristic polynomial depend linearly on the uncertain parameters. A primal/dual pair of infinite-dimensional convex problems is solved by successive finite-dimensional approximations. The primal/dual pair has no duality gap, and both upper and lower bounds produced by the approximations converge monotonically to the optimal value.