AI-Guided Robotic Welding (ROS 2)

Manual welding setup is slow and inconsistent. The client needed a system that could identify the workpiece and suggest optimal weld parameters automatically.

• Planned the robot motions with ROS 2 and MoveIt 2.
• Trained a YOLO model to recognise the metal and locate the weld positions.
• Integrated the 3D camera to detect the weld, generated the path and built the full pipeline.

• The robot recognises the metal workpiece and finds the weld locations automatically, with suggested welding settings.
• A 3D camera is integrated to detect the weld seam and generate the robot path.
• One pipeline runs from detection to a planned, ready-to-weld path.

Reduced setup time per workpiece and improved consistency of weld seam detection. The operator UI enables human-in-the-loop verification before every weld.

Welding environments are harsh for vision systems — smoke, sparks, and reflective metal surfaces. The model needed robust training data and real-time inference at the edge. Integrating AI suggestions with MoveIt 2 motion planning required careful coordinate frame management.