Figure founder and CEO Brett Adcock recently unveiled a new machine-learning model for humanoid robots, allowing them to complete complex household tasks.
The news arrives weeks after Adcock announced the firm’s decision to step away from OpenAI collaboration. The company has now centered itself around Helix – a “generalist” Vision-Language-Action (VLA) model. VLAs leverage perception, language understanding, and learned control to process information and accomplish tasks.
“Helix displays strong object generalization, being able to pick up thousands of novel household items with varying shapes, sizes, colors, and material properties never encountered before in training, simply by asking in natural language.” Figure stated.
Once it receives a language voice prompt, the humanoid robot is designed to visually analyze the environment and perform the requested action. Figure provided examples of this in action with prompts including “Hand the bag of cookies to the robot on your right” or, “Receive the bag of cookies from the robot on your left and place it in the open drawer.” These examples showcase Helix’s multi-robot collaboration ability or capability to control two robots at once.
Figure displays the effectiveness of its VLM by showing its 02 humanoid robots functioning in a home environment, which is notoriously difficult for robots as it lacks the consistency and structure of factories or warehouses. The layout, function, and tools within each room in a home vary too much for manual programming. Figure believes that working within this complexity will ultimately open up new opportunities for a broader range of robotic capabilities in various settings.
“Helix thinks like a human… and to bring robots into homes, we need a step change in capabilities. Helix can generalize to virtually any household item,” Adcock said in a social media post. “We’ve been working on this project for over a year, aiming to solve general robotics. Like a human, Helix understands speech, reasons through problems, and can grasp any object – all without needing training or code. In testing, Helix can grab almost any household object.”
These robots have undergone hundreds of hours of repetitive training in order to complete highly variable tasks quickly and efficiently. Helix utilizes a distinctive approach by using a single set of neural network weights to learn various behaviors, which eliminates the need for task-specific fine-tuning and streamlines the learning process.
According to Figure, Helix introduces a novel approach to upper-body manipulation control. It offers high-rate continuous control of the entire humanoid upper body, which includes the wrists, torso, head, and individual fingers. This level of control allows for more nuanced movements and actions.
“This represents a transformative step forward in how Figure scales humanoid robot behaviors—one that we believe will be pivotal as our robots increasingly assist in everyday home environments,” Figure concluded.
