Remarks on a Nonlinear Model Predictive Control Using an Echo State Network

In this study, we explored the possibility of using reservoir computing; especially echo state networks (ESNs) to control systems. We used the ESNs to obtain a prediction model to achieve optimal model predictive control (MPC) of nonlinear systems. The model’s effectiveness was investigated by comparing the control performance with that of the MPC using the radial basis function network (RBFN) as the prediction model. We conducted computational training and testing experiments of the ESN-based nonlinear MPC (NMPC) for testing the tracking task of the discrete-time nonlinear plant. The experimental results show that although the ESN-based NMPC accurately performs the control task, it does not outperform the RBFN-based NMPC. This study demonstrates the potential application of ESNs for controlling nonlinear systems.