Clothoid-based CAD Model Compensation for Precise Welding in Manufacturing Processes

In an industry production line, a crucial task has always been welding, which requires millimeter precision to meet the high-quality standards of the products. In recent years, collaborative robots (cobots) have been integrated into robotic welding cells to guarantee production quality. A human operator can program cobots to follow any desired welding trajectory by moving the manipulator around the workpiece. However, this approach can be physically demanding and lead to poor accuracy. It is also time-consuming since it has to be repeated for each workpiece. This paper presents a novel motion planning approach for welding robots that combines welding workpiece CAD data with optical-tracked learning-by-demonstration. With the proposed approach, it is possible to automatically extract the motion primitives from the CAD model and map them into the workpiece by identifying a set of waypoints using a pointer and a motion capture system. As a result, the process becomes less time-consuming and more straightforward for the operator. We validated the approach in a laboratory setting, achieving an average error of 1 mm in the positioning and a time reduction of 75%.