Randomized algorithms for reduced order H/sub /spl infin// controller design

In this paper, we study randomized algorithms for reduced order H/sub /spl infin// controller design. First, full-order optimal H/sub /spl infin// controllers are randomly generated using a finite dimensional parameterization of all /spl gamma/-optimal controllers and a uniform matrix sample generation algorithm. Then, the minimum closed loop performance achieved using standard model order reduction techniques is estimated in probability. The results are illustrated on a F-8 aircraft controller design example.