A multivariable self-tuning controller with integral action (Q1095090)

An algorithm is presented which is of the self-tuning type and it is applied to CARIMA models (controlled autoregressive integrating moving average). The reason that this model is used instead of the ARMAX frequently used model is that the system is made robust in disturbances which apply to the plant and are of constant load. The main feature of this algorithm is that it performs an integral action on every component of the disturbance. This allows for compensation, due to the presence of constant load in the error. The drawback of this method is that, unfortunately, here the system delay structure should be known a priori. Thus the system interactor matrix should be known. There are some interesting approaches in the controls literature for the reduction of this ``a priori'' needed information. They are given by \textit{L. Dugard}, \textit{G. C. Goodwin} and \textit{X. Xianya} [ibid. 20, 701-709 (1984; Zbl 0547.93077)]. The algorithm is applied to minimum phase systems and with a modification to non-minimum phase systems. Finally, it should be noted that it is a very useful algorithm, also for the case of adaptive control of nonlinear systems, which exhibit disturbances of constant load, but the details need further research.