The problem of designing a globally optimal full-order output-feedback controller for polytopic uncertain systems is known to be a non-convex NP-hard optimization problem, that can be represented as a bilinear matrix inequality optimization problem for most design objectives. In this paper a new approach is proposed to the design of locally optimal controllers. It is iterative by nature, and starting from any initial feasible controller it performs local optimization over a suitably defined non-convex function at each iteration. The approach features the properties of computational efficiency, guaranteed convergence to a local optimum, and applicability to a very wide range of problems. Furthermore, a fast (but conservative) LMI-based procedure for computing an initially feasible controller is also presented. The complete approach is demonstrated on a model of one joint of a real-life space robotic manipulator. (C) 2004 Elsevier Ltd. All rights reserved.