Robustness of Discrete Linear-Quadratic Controllers

It is important that model-based controllers exhibit robustness to deviations between the model used to design the controller and the true process. For certain types of disturbances minimum variance controllers are not robust. However, in this paper it is shown that minimum variance controllers subject to a constraint on the variance of the manipulated variable (linear-quadratic controllers) exhibit excellent robustness for even small amounts of constraint. The manner in which these linear-quadratic controllers achieve robustness is different from that usually used in Internal Model Control (IMC). Furthermore, it is shown that for the same level of robustness to model deviations, linear-quadratic controllers usually exhibit much better performance than simple Internal Model Controllers.