1993 Volume 29 Issue 1 Pages 39-45
In a discrete time control for multi-input multi-output systems, when the response speed of output differs greatly each other, it is better to use different sampling periods for each output in order to improve the robustness of the control system and decrease the effect of the S/N ratio. This paper proposes a multivariable generalized predictive control method with multirate sampling model. The sampling and holding actions are done in a single short sampling period for all loops, while the plant modeling is carried out by multirate sampling. First, the single-rate sampling model and the multirate sampling model of the plant are derived. Next, the predicted future value of each output is expressed using the single-rate and the multirate sampling model, respectively. Then the cost function is so modified that sampling interval for prediction of each output is set differently corresponding to the response speed of each output in order to ensure enough prediction interval using a small costing horizon. By using this cost function, the control law is determined. The dynamics from the reference trajectory to the plant output of the proposed method using the multirate sampling model is shown clearly to be the same as that using the single-rate sampling model. The simulation results illustrate that the proposed method using the multirate sampling model is superior to the method using the single-rate sampling model in terns of the effect of the observation noise.
