Improving tracking control for robots using neural networks

Tracking control of robots in joint space is studied in this paper. A new control algorithm is proposed based on the well-known computed torque method and a feedforward compensating controller. The function of the feedforward controller, which is realized using an RBF neural network, is to provide high tracking accuracy of robot path-following performance. The RBF neural network is trained offline or online to identify robot modelling error, and convergence of the neural network training is guaranteed based on Lyapunov theory. Through simulations, the proposed scheme is shown to be able to achieve much better tracking performance.