A linear-quadratic optimal control approach is presented for designing linear time-invariant decentralized controllers to solve the robust servomechanism problem. It is assumed that not only the plant and the measurements but also the controllers may be subject to certain disturbances. A physically meaningful quadratic cost functional is defined, making it possible to assign desired weights to the errors, to their derivatives, to the system effort, and to the controller efforts. It is shown that the resulting controlled system has many properties like asymptotic tracking, stability, and robustness. The solution does not require much control effort at high frequencies, and the system is guaranteed not to have high frequency oscillations.
展开▼