A neural network compensator for uncertainties of robotic manipulators

The authors propose neural networks which do not learn inverse dynamic models but compensate nonlinearities of robotic manipulators with the computed torque method. A comparison of the performance of these networks with that of the conventional adaptive scheme in compensating the unmodeled effects was carried out. As a result, the adaptive capability of the neural network controller with respect to the unstructured effects is shown, although the conventional scheme had no capability to reduce the unmodeled effects. Furthermore, a learning method of the neural network compensator with true teaching signals is shown. The tracking error of the robotic manipulator was greatly reduced.