Abstract
Recently, visual feedback control has become important, in order for the mechanical system to recognize dynamic environ- ment and then to achieve their purpose task in real time. The purpose of this research is to sense the state of the 2 dimensional inverted pendulum by stereo vision of universal camera and to stabilize it by 2 link manipulator. To compensate the long sampling rate and lens distortion of CCD cameras, a multi-rate control is applied to manipulator servo control, and a neural networks is trained the relationship between a world coordinate and image coordinates. Also, the image processing is included the pendulum's dynamics to avoid the misrecognition. We verify the effect of the controller and the visual feedback with universal cameras through the experiment of stabilizing the pendulum.
