In many cases, control system design is formulated as minimization of some prescribed closed loop performance reflecting a control requirement. Since the closed loop performance is a function of a controlled plant and a controller, the model identification and controller design must interact with each other. This motivates us to consider the model identification and controller design simultaneously. However, most of the previous joint design methods are not applicable to unstable plants because the identified plant model is usually used as the design parameter, i.e. a weighting function in the controller design. On the other hand, we have analyzed the internal structure of 2DOF control system using coprime factorization on
RH∞ and shown that two free parameters,
K and
Q ∈
RH∞, specify tracking performance and feedback performance, respectively. Also, Tay et al. have proposed a parameterization of the plant dynamics by switching the role of the controlled plant and controller, and introduced a free parameter
R ∈
RH∞. In this paper, we propose a new joint design strategy based on the identification of
R and the design of
Q. Since the identified plant parameter
R is always stable, the proposed joint design strategy can be applied to wider class than the conventional joint design methods. Moreover, it is known that
R is well approximated as a less order model than the plant itself. This leads to the advantage in designing
Q with less order. Finally, the usefulness of the proposed method is verified by some simulations and experiments.
抄録全体を表示