A fatigue-adaptive hill model for online torque prediction

Modelling the muscles during human movement provides critical support for optimizing exercise training, developing intelligent rehabilitation systems, and advancing artificial prostheses. Nevertheless, a multitude of factors, such as intrinsic muscle properties, contraction history, and most notably, muscle fatigue, have significant impacts on modelling outcomes. In this paper, we proposed an adaptive Hill-type musculoskeletal model that can be modified on the basis of the EMG fatigue index. By establishing a robust relationship between the EMG fatigue index and the specific parameter intrinsic to the Hill-type musculoskeletal model, we enabled the continuous monitoring of the fatigue index throughout the movement process, allowing the model to adapt effectively and optimize its performance. The proposed model was validated through a continuous isometric contraction task performed by three healthy subjects. The results demonstrated that the proposed model could maintain relatively high torque prediction accuracy, even when the force period and force amplitude changed. This feature endows the model with enhanced practical applicability and reliability in real-world scenarios.