Visual control of a robotic manipulator using neural networks

A control strategy and scheme for a robotic manipulator using a vision system to position and orientate the end effector is described. The proposed system directly integrates visual data into the control process without subdividing the process into determination of the workpiece position and orientation, and inverse kinematic calculation. Neural networks are used to learn reproduction of the nonlinear relationship between image data and control signals for the changes in joint angles required to achieve the desired position and orientation. The validity and effectiveness of the proposed control scheme are verified by computer simulations and experimental results