面向高反光窄焊缝场景的人工引导协作机器人焊接轨迹修正方法

Laser welding is widely applied across industries. However, traditional manual teaching or offline programming lacks effective improvements for batch workpiece shape variations. Manual corrections are time-consuming and labor-intensive. In complex welding scenarios like high-reflectivity narrow seams, noise and instability hinder accurate trajectory corrections, affecting quality. A non-rigid registration-based method for correcting welding trajectories on high-reflectivity narrow seams is proposed. Firstly, a positioning method based on dynamic ROI prediction was proposed, which obtains partial weld position points of the actual workpiece through a line laser sensor in a manually guided collaborative manner. Secondly, an optimized WTo-CPD algorithm registers the dense trajectory point set from offline programming to the target point set, creating a new welding trajectory. Finally, the experimental results show that with random errors of 0–0.3 mm, the convergence speed of the WTo-CPD improves by an average of 27.16% and 40.50% compared to Nonrigid-CPD and Bayesian-CPD. The average error is around 0.02 mm and the maximum error is less than 0.21 mm, ensuring the welding quality.