Virtual reality based programming of human-like torch operation for robotic welding

Abstract

In complex welding tasks, skilled manual welders often outperform welding robots, primarily due to their expertise in torch manipulation. To address this, a robotic welding teaching system was developed to assist welders in controlling the torch. This system utilizes human–robot interaction to track the welder's movements, significantly enhancing the welding robot's ability to handle intricate weld seams. A virtual welding robot module, created in Unity3D with virtual reality (VR) technology, closely replicates the real robot. This module is integrated with a human–robot interaction interface and a welder operation module. Motion mapping strategies were devised to transfer the welder's movements from the handle to the welding torch, including "static", "dynamic", "velocity" and "acceleration" strategies. These strategies were tested across four trajectories: linear, arc, sinusoidal, and spatial intersection curves. The results revealed the superiority of the "dynamic" strategy. Further evaluations of the teaching system's performance—specifically its "trajectory accuracy", "trajectory delay" and "time delay"—were conducted for straight lines, arcs, and flip motions. The test results showed that, within the operating speed range of 1 to 40 mm/s and 10 to 40°/s, the system's time delay is less than 0.12 s, with actual trajectory errors remaining below 0.06 mm and 0.1°. These performance metrics demonstrate that the system effectively meets the requirements for precise tracking of both the weld seam and torch manipulation.