US researchers build fall-safe biped robots to advance real-world reinforcement learning
SMRTR summary
University of Illinois researchers developed innovative bipedal robots using a "HybridLeg" design that combines serial and parallel linkage mechanisms, creating more agile and fall-resistant humanoid robots. The 1.84-meter tall, 29-kilogram robots feature a protective lantern-shaped cover and concentrate 10 of their 12 motors near the pelvis to reduce leg weight and improve dynamic walking. This design enables autonomous recovery from falls and self-reset capabilities, advancing real-world reinforcement learning experiments for humanoid locomotion.
SMRTR provides this summary for quick context. The original article belongs to Interesting Engineering.
Read the original article