Remarks on System Identification Using a Quaternion-valued Neural Network with Application to Optimal Predictive Control

In this study, a method of identifying the direct transfer function of a target plant using quaternion-valued neural networks (QVNNs) was proposed. Its possibility for controlling nonlinear system applications was also explored. The QVNN-based identifier predicted the future plant output in the model predictive control approach. The resulting prediction was used to define an objective function for deriving an optimal control input that facilitates the convergence of the plant output to the reference output. Computational experiments of the optimal predictive control using the QVNN-based identifier for a tracking task of a discrete-time nonlinear plant were performed. Experimental results confirm the feasibility and effectiveness of the QVNN in the control application of nonlinear systems.