抄録
This paper treats the identification problem of multivariable bilinear systems. The identification procedure is the following: at first, the external description model is derived from the zero-state response of the systems, then the weighted sequence matrices in the model are estimated through the least mean-square method. Next, the internal description model is realized by using the previously estimated matrices.
However, since we must use the estimated weighted sequence matrices instead of the true weighted sequence matrices in the case of solving the realization problem, it is generally difficult to realize the external description model via an equivalent internal description model. Therefore, both a method to overcome this difficulty and an actual algorithm of minimal bilinear realizations are proposed in this paper. The realization algorithm in this paper is simpler than the algorithm given by A. Isidori.
The main feature of an identification method in this paper will be that we need not use the state-variables of the identified bilinear system in the case of identifing an bilinear system.
