Abstract: A stable adaptive control approach using dynamic neural networks (DNN's) has been developed for a class of multi-input multi-output (MIMO) sampled-data nonlinear systems with unknown dynamic nonlinearities. Unlike static NN's (SNN's) to approximate nonlinear components in the dynamic system, DNN's are used to approximate the whole dynamic system. The system control law is composed of the dynamic inversion of the DNN system, adaptive compensation and NN variable structure control (VSC) components. The NN variable structure control is used to guarantee the stability of the controlled system and improve the system dynamic performance. The proof of complete stability and tracking error convergence is given by using Lyapunov stability theory, and the learning algorithm for the DNN system is obtained thereby. Simulations for a two-link manipulator show that the stable adaptive control approach using DNN's has a better dynamic performance than that using SNN's.