A major advancement in humanoid robotics has been achieved by Figure AI with the introduction of natural walking capabilities to its latest humanoid robot, Figure 02. This innovation stems from an end-to-end neural network trained using reinforcement learning, enabling more human-like movements. The company utilized high-fidelity simulations to accelerate learning processes and develop robust locomotion strategies. This development builds on their earlier unveiling of Helix, a vision-language action AI model that controls robotic upper body movements via natural language commands. Together, these advancements signify significant progress towards creating robots capable of handling real-world scenarios effectively.
Figure AI leveraged sophisticated reinforcement learning techniques to teach its humanoid robots how to walk naturally. Through high-fidelity simulations, the company was able to generate extensive data sets within short periods, simulating years' worth of movement patterns. These simulations involved thousands of virtual humanoids operating under diverse parameters and conditions, ensuring comprehensive training. As a result, the humanoid fleet quickly learned stable and proprioceptive locomotion methods, allowing for rapid engineering iteration cycles. This approach not only enhances efficiency but also paves the way for further technological breakthroughs.
Beyond walking, Figure AI continues pushing boundaries with Helix, an innovative AI model introduced earlier this year. Helix distinguishes itself through its ability to control robotic upper bodies directly using natural language instructions. Unlike traditional systems requiring lengthy training sessions or complex manual programming, Helix operates dynamically, adapting instantly to various tasks. Its strong object generalization capability enables it to interact effectively with unfamiliar household items simply by following verbal directions. This combination of advanced walking mechanics and versatile upper-body control represents a leap forward in humanoid robotics technology.
The journey toward fully autonomous and adaptable humanoid robots is still underway, yet early achievements are promising. Figure AI remains dedicated to expanding their learned policies so that robots can adeptly manage any human-like situation encountered in practical environments. By integrating refined walking abilities with intuitive upper-body manipulation, they aim to produce machines capable of seamlessly interacting within everyday settings, thus transforming industries ranging from healthcare to manufacturing.
