This paper presents a method for visually guided object grasping and alignment using a camera that is not mounted onto the robot. The method involves real-time estimation of the image Jacobian and a 3-D projective representation of a grasp or alignment. The method is robust against discrepancies in camera parameters and is capable of performing in complex environments. The performance of the visual servoing algorithm and the expected precision of the grasping are analyzed.
Publication date: 22 Nov 2023
Project Page: https://ieeexplore.ieee.org/document/704605
Paper: https://arxiv.org/pdf/2311.12660
