Regulation of nonlinear plants using radial basis function neural networks

A large class of nonlinear discrete systems with accessible states can be controlled through feedback linearization. This paper develops a practical algorithm for state feedback control design using radial basis function neural networks, which are trained using dynamic backpropagation. Linear least-squares is coupled with a Gram-Schmidt orthogonalization procedure to perform size reduction of the neural networks. An example of regulating a nonlinear plant is included to illustrate the effectiveness of the proposed algorithm.