An integrated calibration and 3D reconstruction method for humanoid welding robots

Abstract

Welding, a key process in traditional manufacturing and repair, is evolving towards intelligent automation, with humanoid welding robots (HWR) playing a central role in the next generation of welding technologies. This paper addresses the challenge of rapid coordinate system unification and area identification for HWR by proposing an integrated calibration and three-dimensional (3D) reconstruction method. A multi-camera, multi-IMU, dual robotic arm calibration system is first introduced to precisely define the robot’s coordinate system and enhance calibration accuracy. A novel multi-parameter error optimization method is then proposed, significantly improving calibration precision. This represents the first application of this method to multi-camera-IMU and hand-eye calibration scenarios in the context of welding. Furthermore, the integration of multi-view 3D reconstruction technology, driven by the fusion of the head binocular camera and the RGB-D camera on the left robotic arm, substantially enhances the accuracy of welding area reconstruction. Experimental validation demonstrates the feasibility and effectiveness of the method, achieving an absolute 3D model error of 1.09 ± 0.23 mm, with a 0.99% error rate. Future work will focus on integrating these methods with dual-arm collaborative welding path planning to further advance intelligent welding manufacturing using HWR.