Abstract
In general, in multivariable adaptive control, direct method involves much more parameters to be adjusted than indirect method. It is often the case, however, that some of controller parameters are constant or even of zero value because of the intrinsic feature of the plant structure. That is, the number of unknown controller parameters may be considerably reduced by utilizing effectively a priori knowledge on the plant structure.
In this paper, adaptive control of water levels of a two-tank system is considered. By analyzing the system, some of controller parameters could be set to zeros, which in turn enable us to decrease the amount of computation required in each step. Another emphasis is put on the way of dealing with noises contained in output signals. It is confirmed that these noises cause oscillatory plant input signals with high frequency and large amplitude which do harm to input equipments. The noise problem could be overcome by adjusting the poles of observer as well as zeros of interactor.
These two ideas concerning with utilizing a priori knowledge of plant structure and with overcoming the noise problem are very universal and may be applied to many other plants.
