Methods for Hand-Eye Coordination of a serial Robot from partial Observations

Abstract

Precise object manipulation by a robot requires precise knowledge of the position of the robot endeffector relative to the object. By the so-called eye-to-hand coordination, both the position of the object and the position of the robot relative to the camera are determined. In practice, usually the position of the robot base to camera is calibrated in advanced and the position of the robot endeffector relative to the base is calculated by forward kinematics with joint angle confgurations. For the robots working in the human environment, they are constructed with lightweight in order to increase security, which achieves lower stiffness than industrial robots. Thus, the reached position of robot-effector deviates from its commanded position. The work of this thsis is to develop a method based on the image processing to minimize deviations and thus to estimate the real position of the robot endeffector in real time. Thus, the robot end-effector can be guaranteed to precisely grip the target object.