Framework enables shared muscle control across different body models

A generalist policy for muscle-driven control across individual musculoskeletal models

Robotics

Summary

Controlling muscles in computer models that represent different human bodies is hard because everyone’s body is unique. The authors created MyoAcross, a tool that learns general ways to control muscles that work across many different individual body models without needing extra adjustment. Tested on 31 new body models and 22 shape-estimated models, it improved tracking accuracy and worked even when there were asymmetries like shorter limbs. They also showed simulated muscle forces matched real muscle activity and forces measured from 120 people. This means MyoAcross could help compare movements across many unique body types using a single control approach.

What this means in practice

  • For biomedical engineers: Develop personalized rehabilitation simulations that reuse a single muscle control policy across patients with different body shapes without extra tuning.
  • For game developers: Create character animations driven by muscle models that adapt automatically to varied virtual body shapes, improving realism and ease of animation.$Commercial implications: Enables development of muscle-driven character animation systems adaptable to different avatars, improving realism and reducing manual tuning effort.

Authors

Bozhao Wang, Yinglei Zhu, Zhengqing Li, Zhiyuan Zhang, Weiming Kang, Yiheng He, XinJun Liu, Senhao Hou, Huichan Zhao, Xiaoqiang Tang

Abstract

Differences in musculoskeletal morphology challenge policy reuse across individual models. Here we present MyoAcross, a framework that combines individual musculoskeletal model construction, morphology-conditioned policy learning and virtual model augmentation to enable muscle-driven reference-motion tracking without individual fine-tuning. Across 31 held-out individual models, MyoAcross achieves 99.24% completion and reduces tracking error by 22.60% relative to an unconditioned multilayer perceptron. Policy reuse extends to models of 22 held-out individuals constructed from single-image shape estimates, measured height and body mass, and, with asymmetry augmentation, to models with unilateral lower-limb shortening. Comparisons with measurements from 120 additional individuals show agreement in vertical ground reaction force waveforms and channel-dependent agreement between simulated muscle activations and electromyography. These results establish a common control framework for comparative simulations of movement across individual musculoskeletal models.
View PDFOpen arXiv