Trajectory Control of Robot Manipulators Using a Neural Network Controller

Abstract

In this article, we presented dynamic trajectory tracking control of industrial robot manipulators using a PD controller and a neural network controller. Some different kinds strucutres of neural network control systems were discuessed. The neural network controller was designed as a three layers feed-forward network. The learning law of weights of the neural network was derived using a simplified dynamic model of the robot and back propagation approach. Dynamic trajectory tracking control simulations and experiments were carried out using an industrial manipulator AdeptOne XL robot. The results showed the effectiveness and usefulness of the proposed control method. From the simulations and experiments, it was seen that according the increase of learning times the neural network controller took over of the PD controller on playing the role in generating actuating force/torque required by the dynamic trajectory. It also was clarified that the learning effect of the neural network has some limitation, i.e. after some specified time of learning, trajectory tracking accuracy remains unchanged.