Toward Improving Human Training by Combining Wearable Full-Body IoT Sensors and Machine Learning.

Area of Science:

• Industrial Engineering
• Artificial Intelligence
• Human-Computer Interaction

Background:

• Traditional training for physical labor is often one-size-fits-all.
• Lack of objective, data-driven methods for assessing skill proficiency.
• Need for personalized and efficient upskilling in manual workforces.

Purpose of the Study:

• To introduce DigitalUpSkilling, an IoT and AI framework for personalized training in labor-intensive jobs.
• To develop a system for recognizing work activities and assessing skill proficiency using digital twins.
• To demonstrate the framework's efficacy in a real-world meat processing environment.

Main Methods:

• Utilizing wearable IoT sensors to capture kinematic data of workers.
• Synthesizing worker digital twins from sensor data.
• Implementing generative adversarial network (GAN) and machine learning (ML) models for work activity recognition.
• Developing ML models for skill proficiency evaluation.

Main Results:

• Achieved 99% accuracy in recognizing specific work activities in meat processing.
• Successfully evaluated worker proficiency by comparing kinematic data.
• Demonstrated the feasibility of using digital twins for skill assessment.

Conclusions:

• DigitalUpSkilling provides a novel, data-driven approach to personalized worker training.
• The framework enables accurate recognition of work activities and objective skill evaluation.
• Lays the foundation for next-generation digital training solutions in physical labor sectors.