Abstract
This paper proposes an AI-based welding robot 3D vision module and manipulator control system for welding automation. With conventional welding robots, welding is performed only at the designated welding points through robot teaching; however, the difficulty with this method is that the position of the welding point changes due to the tolerance (gap) of the welding object. To solve this problem, we automatically recognized the welding points using a deep learning-based object detection model called YOLOv4. The time required for labeling was thus reduced through a deep learning labeling tool that uses an existing learned weights file. To control the robot with the recognized welding point, a hand/eye calibration method calculating the homogeneous transformation matrix between the robot and the camera was adopted. Additionally, to correct the final robot control error caused due to calibration and camera depth errors, the control precision at the welding point was improved through interpolation method. The system has been verified by conducting sufficient number of experiments in a non-destructive formwork welding environment.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call