The number of movements required to reach a specific target with a manipulator arm is quite large, and the dynamic functions of the robot are generally complex non-linear and time-varying. As a result, conventional controllers are unable to efficiently control movements under different conditions, speed, weight . . . . The aim of our work is to design a non-linear system based on a neuro-fuzzy controller network using supervised learning, to perform trajectory tracking by a manipulator robot. Identifications of the controller network structure are performed using the controller (ANFIS), with new parameters and synaptic weights automatically adapted and adjusted. Finally, the robustness of the proposed approach has been tested, taking into account parameter variations.
