Scaling Whole-Body Human Musculoskeletal Behavior Emulation for Specificity and Diversity

The embodied learning of human motor control requires whole-body neuro-actuated musculoskeletal dynamics, while the internal muscle-driven processes underlying movement remain largely inaccessible to direct measurement. Computational modeling offers an alternative, but inverse dynamics methods struggle to resolve redundant control from observed kinematics in the high-dimensional, over-actuated system. Forward imitation approaches based on deep reinforcement learning remain limited in tracking accuracy and learning efficiency for anatomically detailed human models due to the curse of dimensionality in both control and reward design. Here we introduce a large-scale musculoskeletal computation framework for biomechanically grounded whole-body motion reproduction. By integrating parallel simulation with adversarial reward aggregation and value-guided flow exploration, the MS-Emulator framework overcomes key optimization bottlenecks in high-dimensional reinforcement learning for musculoskeletal control. Using MS-Human-700, a standard whole-body human musculoskeletal model actuated by 700 muscle-tendon units, we trained controllers to effiently achieve diverse reference motions. We show that the framework can also explore the solution space of muscle coordination, identifying distinct musculoskeletal control patterns that converge to nearly identical external kinematic and mechanical measurements. Our large scale experiment demonstrates that simulated muscle-control solutions encompass substantial variation observed in human measurements and that the size of this solution space depends on movement dynamics. This work establishes a tractable route to analyzing the specificity and diversity underlying human embodied control of movement.