Nhận dạng các đối tượng mimo ổn định với giới hạn bất định cho trước

Abstract

The modern control theory is mainly dealing with MIMO systems. In this case the dimension of the state vector as well as transfer function matrix may be high. In contrast to the large scale systems where the problem could be  solved via the separation of the primary system in the subsystems, the MIMO control theory takes the system as an unitary object in consideration. Of course, a MIMO system is really more complicated than the SISO one. The identification problem of the MIMO systems involves many difficulties and that is  the reason for why  the adaptive algorithms have been applied to the MIMO systems  only in a small measure up to now. In contrast, the robust control approach is seeming to be an useful tool for solving the control task with model uncertainties. Alone the number of  publications on this issue in the past can prove this assertion right enough [12].

However, it is to see the  fact that the available results of the MIMO system identification are very restricted. There are no proper facilities to overcome the difficulties expressed in [1]. The 4 SID algorithm is recently attracting attension but the use of this algorithm requires the stipulations which could be  not perpetually fulfilled [8]. The  present paper proposes an approach for complete the MIMO system identification with given uncertainty bounds. At first, some models of MIMO plants are described in Section 2 in order to look for the way solving the identification problem. Section 3 repeats the conventional identification method developed by Landau for  SISO system and later completed by other researchers. Section 4 deals with the improvement of the algorithm of Section 3 in order to fit for the MIMO systems. Finally, in Section 5, an multi-approach is provided for solving the MIMO system identification problem. At the first stage the identification is done to  receive the raw model. Then,  the tracking quality of the system with a robust controller designed on the raw model will be checked. The uncertainty bound analysis gives the information to adjust the raw model.