Augmented gradient flows for on-line robust pole assignment via state and output feedback.