Feedforward, correlated disturbances and identification

Feedforward is an important method to counteract the effect of measurable disturbances. There are ways to design optimal feedforward controllers, which minimize the output variance caused by the disturbances. In case the measurable disturbances are correlated with some unmeasurable ones, the performance will be degraded. It is shown how the underlying model of the disturbances can be rewritten into an equivalent form, so that the standard design rule for an optimal feedforward controller can be applied. It is also shown how this equivalent form is obtained directly if the system is modelled by using system identification. (C) 1999 Elsevier Science Ltd. All rights reserved.