A High-Precision Assembly System of 3C Parts Based on 6D Pose Estimation and Visual Servoing

Abstract

Automated assembly of 3C (Computer, Communication, and Consumer Electronics) parts has attracted increasing attention from the manufacturing industry in recent years. Employing various types of vision sensors to assist robots in completing high-precision assembly tasks has become one of the most effective methods. There appears to be such a complicated situation in the actual manufacturing assembly line. The two parts to be assembled simultaneously have pose uncertainty when one part clamped at the end-effector after robotic grasping needs to be aligned with another part randomly placed on the workbench. This paper presents an automatic high-precision assembly system based on 6D pose estimation and visual servoing by using multiple cameras for the solution. A pose compensation strategy and real-time visual servoing approaches are implemented to tackle pose errors from grasping and to align the target to be assembled with a random pose. Our assembly strategy is composed of three stages. In the first stage, a high-resolution monocular camera is employed to estimate the 6D pose of the part which is clamped by the gripper and calculate the grasping error relative to the standard pose with the eye-hand system. After that, a visual servoing controller is employed sequentially to locate and align the part on the platform with an RGB-D camera. Once the servoing finished, correct the clamped parts to the standard post through robot motion compensation to eliminate grasping errors. The proposed approach is verified by experiments, which can be used to accomplish automated assembly tasks even if two parts to be assembled both exist in uncertain poses. The proposed assembly system appears to be flexible and efficient for high-precision assembly, such as 3C products.