Abstract
In this paper, the nonlinear three-layer Neural Network using a dynamic back-propagation method which based on the output signals of the hybrid controller and past output information of the plant is proposed for the Cartesian position/force control of multi-DOF robots. The generalized delta law is used as a learning law to adjust the weights of the Neural Network at every sampling time in order to adapt to an unknown environment.
Experiments were done with a 3-DOF Direct-Drive Planar Robot Manipulator. The proposed Neural Network Controller controls the position and the angle of the end-effector as well as the applied force to an unknown environment. The experimental results show the effectiveness and flexibility of the proposed Neural Network Position/Force Controller for Multi-DOF Robots.
