Human Digital Twin for Real-Time Physical Fatigue Estimation in Human-Robot Collaboration

Abstract

Human-robot collaboration is a vital approach in manufacturing, integrating the capabilities of both humans and robots effectively. In recent years, the well-being of manufacturing workers has received increasing attention with the development of manufacturing systems. However, the perception of human characteristics, such as physical fatigue, and the integration of these characteristics with human-robot manufacturing systems, remain relatively limited. The lack of awareness regarding human physical fatigue may negatively impact workers' health and, in severe cases, lead to musculoskeletal disorders. To overcome this bottleneck, this paper presents a human digital twin method for real-time fatigue estimation in a manufacturing scenario. Firstly, we adopt a human muscle force estimation method to simulate the upper limb muscle activity of humans during assembly activities. Secondly, an IK-BiLSTM-AM based surrogate model is used to accelerate the process of estimating the muscle state. Lastly, we adopt a muscle force-fatigue model for real-time muscle fatigue assessment. This scheme is validated through a proof-of-concept experiment in a manufacturing activity dataset. The findings highlight the efficiency and resilience of the suggested approach.