Reflex-Informed Neuromuscular Reinforcement Learning for Muscle-Driven Locomotion
A reflex controller handles the base walking behavior, while reinforcement learning adjusts four residual parameters for stability and adaptation.
The framework targets muscle-driven locomotion that is both physiologically plausible and resilient to changes in strength or outside disturbances. Its learned policy modulates reflex gains and thresholds tied to hip swing, knee support, and ankle propulsion. The authors report improved kinematic accuracy, dynamic consistency, bilateral symmetry, and stride-to-stride consistency in nominal walking. They also say the policy stayed robust under muscle weakness and perturbations without retraining. ArXiv · AI/CL/LG's note
The framework targets muscle-driven locomotion that is both physiologically plausible and resilient to changes in strength or outside disturbances. Its learned policy modulates reflex gains and thresholds tied to hip swing, knee support, and ankle propulsion. The authors report improved kinematic accuracy, dynamic consistency, bilateral symmetry, and stride-to-stride consistency in nominal walking. They also say the policy stayed robust under muscle weakness and perturbations without retraining. ArXiv · AI/CL/LG's note
score 4