Federated Semi-Supervised Digital Twin for Enhanced Human-Machine Interaction in Industry 5.0

The integration of collaborative machines in Industry 5.0 has necessitated advanced safety and interaction protocols to enhance productivity and ensure human safety. Traditional safety measures are often bulky, inflexible, and costly, limiting true human-robot collaboration. This research introduces a federated semi-supervised Digital Twin framework designed to address these challenges by leveraging synthetic data generated via Digital Twin technology and real-world data, facilitated by ROS2-based communication. The framework employs Federated Semi-Supervised Learning to enhance the detection and classification of human-machine interactions. Evaluations demonstrate that the proposed model significantly outperforms existing models such as Yolov8 and Mask-R CNN in terms of mean Average Precision (mAP) of 91.87% and accuracy of 98.10% while being more resource-efficient. This study underscores the potential of combining federated learning with Digital Twin technology to create adaptable, efficient, and scalable solutions for smart manufacturing, thereby advancing the field of human-robot collaboration.