About this Abstract |
| Meeting |
2026 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2026)
|
| Symposium
|
2026 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2026)
|
| Presentation Title |
Real-Time Object Detection for Autonomous Robotic 3D Printing |
| Author(s) |
Agya Aboagye-Otchere, Ahmed Bendaouia, Jianzhi Li |
| On-Site Speaker (Planned) |
Agya Aboagye-Otchere |
| Abstract Scope |
Autonomous robotic 3D printing is a growing area within the additive manufacturing field. Robotics and artificial intelligence (AI) in manufacturing provide a platform for increasing autonomy within these systems. This paper aims to develop a robust computer vision system for real-time object detection during 3D printing with a six-degree-of-freedom (6-DOF) robot under varying light conditions. The models are trained using a custom dataset for part identification and a public fused deposition modeling (FDM) dataset for defect detection. The vision system is designed to provide information that can be used by the robot motion-planning system for autonomous adjustments. The custom three-light YOLO26n model achieved a precision of 0.8985, recall of 0.9023, mAP50 of 0.9532, and mAP50-95 of 0.8618. Pre-training on the custom three-light dataset before training on the FDM defect data increased recall from 0.9775 to 0.9900 and mAP50-95 from 0.9884 to 0.9949. On the Jetson AGX Orin, the custom three-light model achieved a forward-pass inference latency of 41.62 ms per 640 ×640 image, corresponding to 24.03 frames per second (FPS). This framework showcases the ability to autonomously inspect sectional printing of objects in real time and supports further development toward closed-loop robotic additive manufacturing. |
| Proceedings Inclusion? |
Planned: Post-meeting proceedings |