System identification using neural networks

A recent result by S.R. Chu et al. (1989) shows the use of a Hopfield neural network to identify the A and B matrices of a linear system using state measurements. The authors extend this result to the identification of the complete (A, B, C, D) description using output measurements. They develop a convenient state-space representation for the identification, and describe how the identification problem can be mapped onto the Hopfield network. A computer simulation to illustrate the technique is presented