Helix Walking Controller: Blind, Robust, and Superhuman

Figure’s new Helix walking controller looks confident and grabs you immediately. Trained with reinforcement learning, the robot strides over clutter and traps without relying on cameras, trusting only its body sense to stay balanced. That choice matters: without vision, glare, darkness, and occlusion cannot spook it; every misstep becomes data that makes the gait smarter. You can see the payoff in the rhythm—foot placement, recoveries, and a steady core that shrugs off nudges. It feels less like a demo and more like a team pushing toward real-world reliability. For anyone dreaming about robots at home, in warehouses, or outdoors, robust walking is the foundation. If legs work anywhere, everything else gets easier: navigation modules, manipulation, and safety across 2 layers of decision making. I am impressed by how the cadence looks, and curious how it scales when vision returns.

If you enjoy breakthroughs like this, please like, subscribe, and hit the notification bell so you never miss the next step.