A Multivariable Generalized Self- Tuning Controller with Feedforward

This paper presents a multivariable generalized minimum variance feedforward controller which minimizes a cost function incorporating the system input output vectors, measurable disturbance vector, load offset and time-varying reference signal to allow the best possible disturbance rejection. The stability properties of this controller are also explored in detail. The self-tuning feedforward controller based on the optimal control law proposed adopts the techniques of updating the weighting polynomial matrices of the cost function on line and eliminating the tracking error and the load offset without an integrator. It can be applied to a class of multivariable systems to obtain a simple self-tuning control algorithm which realizes adaptive decoupling control and specifically handles the MIMO system with a different delay associated with each output. The successful application to a multivariable electric heated furnace demonstrates the effectiveness of the adaptive control scheme presented.