Three hybrid control schemes consisting of neural network and adaptive controllers for robot manipulators

Three hybrid control schemes that consist of neural networks and adaptive controllers are proposed. They are simple in implementation in the sense that no system dynamic model is required and yet are as accurate as a conventional controller that fully utilizes dynamic models. Two of the three schemes have online learning capability in the neural network part to enhance their performances. The three schemes are implemented onto a PUMA-type robot manipulator and are compared with Slotine´s adaptive robust controller (1987) for their performances