Visual Control of a Heavy-Duty Robot Manipulator

Authors: Sadettin KAPUCU, Sedat BAYSEÇ

Abstract: Robots can be made more flexible and adaptable by incorporating sensory information into the feedback loop. Visionbased robots can adapt quickly to the emerging requirements of an unknown task and can react appropriately to changes in the environment Because image identitication involves massive arithmetic and data analysis, the traditional approach to using visual feedback has been to separate the problems of obtaining information about the object identity and position according to the camera images from the manipulator control. In this paper, the amount of visual information required is reduced to a minimum so as to speed up image analysis, so that it can be done on-line and on-duty. A spot of definitive color is put onto the object to be idenfitied and manipulated, and the control computer is programmed to search for this unique and definitive color spot in the image captured, relieving the system of the necessity for intricate image idenfitication. The difference between the locations of the spot in stereo images produced by two static, parallel-axis cameras reveals the distance of the spot from the cameras. Three such points detine the location and angular orientation of the object in space. The position of the characteristic spot is determined with respect to the manipulator coordinate frame and, via inverse kinematics, the robot is guided to reach the spot Experimental results of the manipulator reaching a single spot showing the duration of visual for vision analysis and manipulator motion are presented.

Keywords: Visual control, Stereo vision, Robot Camera calibration