Researchers Build HybridLeg Biped For Safe Reinforcement Learning

Researchers at the University of Illinois KIMLAB recently unveiled HybridLeg, an untethered bipedal robot platform designed to advance real-world reinforcement learning. The 1.84 m, 29 kg robot uses a five-bar hybrid leg with 12 motors (10 near the pelvis), multimodal fall detection, and a protective mechanical cover enabling autonomous self-reset after falls. The platform supports longer, safer RL trials and improved dynamic walking validation.