A type of neural networks sliding mode control in the robot manipulators

For the typical non-linear robot control system, this paper proposes a class of sliding mode variable structure control method (SMVSC) based on RBF neural network, and the controller´s output is added a low-pass filter. A two-joint robot is the research object, This method apply a powerful learning and processing power of RBF network as the sliding mode controller for uncertain part, and resolves the chattering problems of sliding mode control through the filter. A two-joint robot trajectory tracking control is simulated and studied. Simulation results show that the design of RBF neural sliding mode reaching law control system do not only effectively inhibit the chattering phenomenon, but also be good stability, control accuracy and achieve simple.