Nonlinear dynamic system identification using radial basis function networks

A radial basis function (RBF) network is used to approximate a continuous nonlinear dynamic system described by a set of nonlinear state equations. The learning laws to adjust the network weight parameters are derived using a Lyapunov function, such that the convergence is guaranteed. Simulations show that the method is simple and extremely effective. The robustness of the approach, with respect to the parameters and implementation, is considered in the paper.