Using Available Potentials of Vision-Based Robots by Converting Passive Watching into Active Seeing

An approach to make use of available potentials of uncalibrated vision-guided robots is introduced. It allows the robot to manipulate differently shaped objects that may be located anywhere in its workspace, even if they are not visible in the initial fields of view of the cameras. Key points of the approach are the conversion of passive "watching" into active "seeing", so that the fields of camera view scan the robot's whole workspace, and a direct transition from image coordinates to motion control commands, so that the need for a calibration of the robot and of the vision system is eliminated. Vol ker Graefe Institute of Measurement Science Bundeswehr University Munich entations and positions anywhere in the robot's work space and need not be visi.'I ble in the initial fields of view of the cameras. None of the coefficients characterizing the arm and the cameras are known (completely uncalibrated system). The arm. The arm's 5 DOF correspond to the 5 joints J , to Js. Joint J5 refers to the gripper's rotation around its 3:: Te rOEmr symmetry axis. Joint J I allows the arm to ment'of the cameras be rotated around a vertical axis. The and their fields of remaining joints J2, J3 and J4 allow the 'Iew. gripper to be moved within a certain C X ~ iedtion of a vertical plane, the work