A NN controller and tracking error bound for robotic manipulators

In this paper, a robust neural network control scheme is proposed for robot tracking tasks. The neural network is trained online and the weight tuning algorithm has a small dead zone to overcome bounded disturbances. Under this proposed control scheme, it is shown that the tracking error bound is completely determined by the neural network approximation error bound, disturbance bound, as well as the control design parameter. The tracking error bound does not depend on the weight estimation errors. A two-link manipulator is used to illustrate the performance of the control scheme