Design of Stable Suboptimal LQG Controllers

One of the most attractive features of the LQG design method is that, under mild conditions, stability of the closed loop is automatically guaranteed. The optimal controller has the structure of observer-estimated state feedback where the observer is known to be asymptotically stable. The controller itself, however, may be unstable. In case of sensor failure, when the system is actually an open-loop one, instability of the controller leads to an unbounded response. In this paper, three design methods, aimed at achieving a stable, suboptimal controller, are suggested.