A visual guided robot using CAD models and fuzzy controller for industrial manipulation

A visual positioning method is proposed to use the CAD model of an object to determine its position and orientation in real-time. Vision oriented features are generated off-line from the CAD model of the object by taking into account the camera intrinsic parameters. The proposed method can successfully identify the pose of the object even though some part of the object cannot be seen by the camera during manipulation. In addition to the CAD model based visual positioning method, a fuzzy controller is adopted to visually guide an EPSON robot for industrial manipulation and/or assembly. The controller can successfully stabilize and smoothly move the robot to the target object in experiments.