Research on Automatic Assembly Method for Multihole Shaft Fitting Using Mirror-Imaging-Based Pose Recognition and Robot Visual Servoing

Visual occlusion caused by the end-effector is a major challenge in robotic automated assembly. To address this problem, this paper proposes a mirror-assisted visual servoing method for high-precision printed circuit board (PCB) assembly under severe occlusion. A planar mirror is introduced to create a virtual observation viewpoint, enabling occlusion-free 6-DOF pose estimation of the grasped PCB based on mirror imaging geometry and a PnP algorithm. A hybrid visual servoing strategy is then adopted, combining mirror-based coarse positioning with imagebased fine alignment. Experiments conducted on a robotic platform with a Basler camera demonstrate that the final position and orientation errors are approximately 0.1 mm, meeting the required accuracy. The proposed method provides a reliable and low-cost solution for precision assembly in vision-constrained environments.