XR-enhanced DemoTeach: an intuitive digital twin-based robot programming for foundry applications

Abstract In casting cleaning operations, such as the removal of gating and venting structures, remain predominantly manual despite their unattractive working conditions. Although automated approaches based on part scanning have been explored, their algorithmic complexity and limited robustness often hinder practical industrial deployment. Programming by Demonstration methods, such as DemoTeach, address this issue by allowing workers to demonstrate machining operations using a tracked dummy tool, from which robot trajectories are automatically generated. However, conventional DemoTeach lacks transparency during demonstration and provides no feedback on robot reachability or potential collisions. To address these challenges, this work introduces an Extended Reality-enhanced DemoTeach, a digital twin-based framework that augments programming-by-demonstration with immersive feedback. The system provides visualization of tool trajectories, inverse kinematics validation, and collision detection during the demonstration phase. By embedding XR as a validation and guidance layer, the framework transforms the demonstration process from an offline system into a transparent and adaptive workflow. The concept has been implemented and validated in a representative industrial scenario of gating and venting removal from castings, with quantitative evaluation of system accuracy and user experience. Results confirm the feasibility and practical benefits of XR-enhanced DemoTeach, highlighting its potential to improve robustness, safety, and efficiency in small- and medium-batch casting production.