Self-organizing quasi-linear ARX RBFN modeling for identification and control of nonlinear systems

The quasi-linear ARX radial basis function network (QARX-RBFN) model has shown good approximation ability and usefulness in nonlinear system identification and control. It owns an ARX-like structure, easy design, good generalization and strong tolerance to input noise. However, the QARX-RBFN model still needs to improve the prediction accuracy by optimizing its structure. In this paper, a novel self-organizing QARX-RBFN (SOQARX-RBFN) model is proposed to solve this problem. The proposed SOQARX-RBFN model consists of simultaneously network construction and parameter optimization. It offers two important advantages. Firstly, the hidden neurons in the SOQARX-RBFN model can be added or removed, based on the neuron activity and mutual information (MI), to achieve the appropriate network complexity and maintain overall computational efficiency for identification. Secondly, the model performance can be significantly improved through the structure optimization. Additionally, the convergence of the SOQARX-RBFN model is analyzed, and the proposed approach is applied to identify and control the nonlinear dynamical systems. Mathematical system simulations are carried out to demonstrate the effectiveness of the proposed method.