On-line motion-feedforward pose recognition invariant For dynamic hand-eye motion

This paper presents a pose measurement method of a 3D object detected by hand-eye cameras. We propose a motion-feedforward (MFF) method to improve visual recognition dynamics, which become worse by disturbing hand-eye motion during visual servoing of the robot manipulator. The MFF method is used to compensate the fictional target motions in the camera view Induced by the end-effctor's motion, so the pose recognition can be assumed invariant for dynamic hand-eye motion. The effectiveness of the proposed method is confirmed by simulations of object's 3D pose recognition being affected by dynamical oscillations of hand-eye cameras.