Figure AI has introduced natural walking to its Figure 02 humanoid robot. The company said its end-to-end neural network has been trained with reinforcement learning to bring the natural movement to the humanoid. Figure AI trained its reinforcement learning controller in “high-fidelity simulations, running thousands of virtual humanoids with varied parameters and scenarios,” allowing the company to collect years’ worth of simulated demonstrations in a few hours, the company said. The company said the natural walking controller learned in simulation using reinforced learning allowed the fleet of humanoid robots to “quickly learn robust, proprioceptive locomotion strategies and enables rapid engineering iteration cycles.” “These initial results are exciting, but we believe they only hint at the full potential of our technology,” Figure AI said. “We’re committed to extending our learned policy to handle every human-like scenario the robot might face in the real world.” The update follows Figure AI’s unveiling of its new vision-language action AI model Helix in March. The company said Helix is the first model to directly control a robot’s upper body using natural language.

Full report : Figure AI used reinforcement learning in simulations to develop the humanoid’s latest human-like walking ability.