<ul><li>Recent advances in learning-based controllers for legged robots have been limited in human-centric environments due to safety concerns.</li><li>Position-based control methods struggle with compliance and adaptability in unseen environments, potentially leading to unsafe behaviors.</li><li>Inspired by animal movements, torque-based policies offer direct and precise control, enabling safer and more adaptable behaviors.</li><li>SATA, a bio-inspired framework, addresses challenges in learning torque-based policies, achieving zero-shot sim-to-real transfer and demonstrating compliance and safety in various environments.</li></ul>

SATA: Safe and Adaptive Torque-Based Locomotion Policies Inspired by Animal Learning

Discover more