This robot has got some moves! 🕺🏼 Helix is @Figure_robot new walking controller, part of its broader Vision-Language-Action (VLA) architecture. It fuses perception, language, and action to coordinate whole-body movement. What’s new: stability and reliability. @adcock_brett https://t.co/p8EshboaMc
Boston Dynamics and Toyota Research Institute are advancing humanoid robotics with Large Behavior Models (LBMs) for Atlas. https://t.co/wiy8dgOEnT
Figure 02 robot walking over obstacles using a reinforcement learning controller, operating blindfolded (without camera inputs). https://t.co/X2fVtSmO7Y
Figure AI has released video footage of its Figure 02 humanoid robot walking over obstacles using a new locomotion system called the Helix walking controller. According to the company’s description in the video posts, Helix is trained with reinforcement learning and enables the robot to walk without input from cameras, relying instead on internal sensors to maintain balance and avoid hazards. The demonstration shows the system achieving what the firm describes as “near-superhuman” performance in stability and reliability. Helix forms part of Figure AI’s broader Vision-Language-Action architecture, which integrates perception, language understanding and motor control to coordinate whole-body movement.