Personalized human musculoskeletal models are essential for simulating realistic interactions in human-robot systems, particularly for safety-critical applications like robotic exoskeletons and co-transportation. A major challenge in creating these models is identifying subject-specific muscle and bone parameters, especially because internal biomechanical variables such as joint torques are difficult to measure directly in living subjects. The proposed Diff-MSM addresses this challenge by using an end-to-end automatic differentiation technique that links measurable muscle activation to joint torque and observable motion, eliminating the need for direct joint torque measurements. Extensive simulations demonstrate that Diff-MSM significantly outperforms existing methods, achieving highly accurate muscle parameter estimations with minimal error, even when starting from approximate initial guesses. This advancement not only enhances the fidelity of musculoskeletal models but also opens new possibilities for real-time muscle health monitoring and personalized rehabilitation protocols. Furthermore, the technique’s ability to accurately identify parameters could improve sports science applications by enabling better assessment and optimization of muscle function. Overall, Diff-MSM represents a promising tool for advancing biomechanical modeling and its practical applications in healthcare and human-robot interaction. Future work may focus on validating the model in clinical settings and integrating it with real-time monitoring systems to further enhance its utility.
👉 Pročitaj original: arXiv AI Papers
