Search for a command to run...
Industry 5.0 drives the evolution of manufacturing by demanding enhanced efficiency, versatility, and safety in increasingly complex digital environments. This paper aims to address critical challenges in the welding industry, where the lack of effective, interactive guidance during robotic operations and real-time monitoring has hindered productivity and safety. In this work, we introduce WELD-E, an innovative, end-to-end solution that integrates Artificial Intelligence and Extended Reality technologies to facilitate intuitive human-robot collaboration. WELD-E leverages a multimodal interface combining voice commands, spatial awareness, and automated decision-making, supported by a Microsoft HoloLens-based XR environment and a ROS2-driven robotic manipulator. At its core the system incorporates deep learning models, including Automatic Speech Recognition, Neural Machine Translation, a domain-specific Welding Large Language Model and a Visual Language model to refine and execute welding commands with high precision and safety. The proposed framework allows multilingual, context-aware interactions while improving the overall operational efficiency and safety in hazardous industrial settings. Through rigorous validation in simulated real-world scenarios, WELD-E demonstrates its potential to significantly reduce setup times, training requirements, and error rates, while promoting a safer and more responsive welding environment. This work paves the way for broader adoption of AI-driven, XR-enhanced solutions in modern manufacturing, aligning with the transformative objectives of Industry 5.0.