<![CDATA[The assembly of fastening components traditionally relies on labour-intensive human-machine collaboration, which incurs high costs. Existing methods often assume fixed positions or use markers for guidance, requiring extra effort to place and maintain them. This study aims to develop an intelligent control system for a vision-equipped robotic arm to autonomously assemble fastening components in industrial settings, enhancing flexibility and reducing labour costs. The system integrates object detection with edge and ellipse detection, alongside filtering techniques, to accurately locate the centres of the fastening components.  The key contribution is the system's ability to perform autonomous assembly without predefined positions, enhancing flexibility in varied environments. YOLOv8 is employed to detect the bolt and nut, followed by edge and ellipse detection to pinpoint centre coordinates. A depth camera and kinematic calculations enable accurate 3D positioning for pick-and-place tasks. Experimental results demonstrate the system’s high effectiveness, with less than 1% of targets undetected. Based on experiments conducted in randomly arranged conditions, the system demonstrated high effectiveness, achieving over 99% detection accuracy. It achieved an 87% average success rate for picking fastening components ranging from sizes M6 to M18, and a 90% success rate for precise placement. Additionally, the system demonstrated robustness across various component sizes, with a minor increase in orientation errors for smaller components, attributed to depth estimation challenges. Future work could explore alternative depth data collection methods to improve accuracy. These results confirm the reliability of the system in flexible assembly tasks, demonstrating its potential to reduce costs by minimising manual involvement in industrial settings.]]>
Copyrights © 2025
