DemoTeach—robot programming with MoCap system for foundry applications

Abstract In the field of casting manufacturing, the post-processing of residual structures and burrs remains a labour-intensive and expertise-driven process, especially for small- and medium-volume production. This paper introduces DemoTeach, a novel semi-automatic (human-in-the-loop) robot teaching method designed to streamline and automate the fettling of complex casting parts using motion capture (MoCap)–based demonstrations. The system enables skilled workers to intuitively guide a dummy tool along casting burrs, generating task-relevant robot trajectories without requiring programming knowledge or extensive 3D data processing. The approach integrates task generation, path planning, and robot control and execution with Cartesian impedance control. A comparative analysis with a reference method based on 3D scanning and point cloud segmentation highlights DemoTeach’s advantages in simplicity, flexibility, and efficiency. Experiments conducted on aluminium casting using a Kuka iiwa 14 R820 (7-axis) demonstrate the system’s ability to reproduce high-fidelity robot trajectories, with end-effector force estimation confirming accurate surface contact during execution. Further opportunities for enhancement are discussed, including quantitative benchmarking, dummy tool adaptation strategies, and the use of augmented reality for real-time feedback and teaching visualization. The results suggest that DemoTeach offers a practical and future-oriented solution for enabling human-guided automation in industrial casting applications.