The design of an automatic vision switching system

Abstract

The importance of Robotics has been recognized by the manufacturing industry. The introduction of Computer Vision has greatly increased the versatility and application domain of robots such as the interaction between the robot and its changing environment. From previously developed vision position correction methods and currently used visual servo-feedback techniques, computer vision systems present many possibilities in improving the quality and productivity of an industrial product. But computer vision systems also have limits in dealing with some robotic tasks such as Storage Battery Cap Installation. Random geometric distortion exists, which results from the characteristics of battery material or from the positioning error of fixtures. Robotic teaching-playback method can not guarantee the assembly precision when deviation happens. A vision system introduced into the robotic system allows it to respond to an uncertain environment. Conventional vision correction methods only drive robots to the final position without considering robotic trajectory control. They reduce the robotic operation speed compared with stable conditions. Visual servo-feedback techniques can generate the optimal path and assembly precision, but the calculation of visual servo control algorithm is time consuming. It also reduces robotic execution time. In addition, random assembly tasks sometimes do not need vision correction. So the justification of this robotic vision system is improved robotic assembly accuracy that does not affect other robot performance. In this thesis a new robotic vision control system, Automatic Vision Switching System (AVSS), was designed. It was derived from the concept of Just-in-Time and dealt with the installation of storage battery caps. It demonstrated a vision system that functions to correct just when it is needed. Image processing techniques for solving the battery hole’s centroid values were included in AVSS design. The experiment was carried out based on an AdeptOne robot, a Vision-EZ system, a camera with eye-in-hand configuration, and a battery model. The experimental results showed that the AVSS can integrate a vision system with a robotic system efficiently. The AVSS identified the deviation of battery holes’ locations and generated the corrected values in real time. The control accuracy, operation speed and optimal path of robotic manipulator were obtained. The AVSS can also be used in other industrial tasks where random variation is present.