A novel laser vision-based method for robotic curved welding seam tracking

Abstract

Abstract Robotic welding technology is widely used in industries such as shipbuilding and construction, where complex curved welds are common. The ‘teaching-playback’ mode cannot cope with the processing and assembly errors in curved weld seams, and the weld seam localization accuracy of 3D vision is insufficient, therefore, this paper proposes a laser vision-based robot curve tracking welding method. First, 3D vision is used to localize the workpiece, and the weld position obtained through point cloud feature extraction serves as the initial welding path for subsequent operations; Then, the global weld planning strategy is constructed by combining the tracking welding model of the laser vision sensor, to ensure that the weld will not go beyond the detection range of the sensor in the process of weld seam error compensation; finally, a fusion method of real-time position and global pose is proposed to ensure the stability of the weld error compensation process. Experimental results show that the proposed method effectively enables automated tracking welding of complex curves, ensuring smooth welding trajectories. The tracking accuracy of the weld can reach 0.4 mm. Additionally, the method does not require manual programming or teaching, significantly improving both welding efficiency and accuracy.