A human digital twin approach for fatigue-aware task planning in human-robot collaborative assembly

Human-robot collaboration (HRC) has emerged as a pivotal paradigm in manufacturing, integrating the strengths of both human and robot capabilities. Neglecting human physical fatigue may adversely affect worker health and, in extreme cases, may lead to musculoskeletal disorders. However, human fatigue has rarely been considered for decision-making in HRC manufacturing systems. Integrating adaptive decision-making to optimise human fatigue in HRC manufacturing systems is crucial. Nonetheless, real-time perception and estimation of human fatigue and decision-making informed by human fatigue face considerable challenges. To address these challenges, this paper introduces a human digital twin method, a bidirectional communication system for physical fatigue assessment and reduction in human-robot collaborative assembly tasks. The methodology encompasses an IK-BiLSTM-AM-based surrogate model, which consists of inverse kinematics analysis (IK), bidirectional long short-term memory (BiLSTM), and attention mechanism (AM), for real-time muscle force estimation integrated with a muscle force-fatigue model for muscle fatigue assessment. An And-Or graph and optimisation model-based HRC task planner is also developed to alleviate physical fatigue via task allocation. The efficacy of this approach has been validated through proof-of-concept assembly experiments involving multiple subjects. The results show that the IK-BiLSTM-AM model achieves a minimum of 8% greater accuracy in muscle force estimation than the baseline methods. The 12-subject assessment results indicate that the task planner effectively reduces the physical fatigue of workers while performing collaborative assembly tasks.