Boston Dynamics and Toyota Research Institute (TRI) are collaborating to push the boundaries of humanoid robotics. The partnership, announced Wednesday, aims to combine TRI’s expertise in large-scale behavioral modeling (LBM) with Boston Dynamics’ Atlas robot, potentially accelerating the development of general-purpose humanoid robots.
The large-scale behavioral model behaves similarly to the large-scale language models that power well-known platforms such as ChatGPT. However, LBM allows robots to learn various physical tasks by observing human behavior rather than generating text. TRI and Boston Dynamics are working together to train Atlas robots to perform a wide range of increasingly complex tasks.
“Recent advances in AI and machine learning have great potential to advance physical intelligence,” said Gil Pratt, CEO of TRI and chief scientist at Toyota. He said the opportunity to implement TRI’s AI on Boston Dynamics’ hardware is “game-changing” for both organizations.
This partnership brings together two leading robotics companies. Boston Dynamics, now owned by Hyundai, has a track record of developing agile and capable robots. TRI, backed by automotive giant Toyota, is pushing the limits of robot control using AI.
Robert Player, CEO of Boston Dynamics, emphasized the timing of this partnership, saying, “There has never been a more exciting time for the robotics industry.”
The immediate goal is to use Atlas for narrowly focused industrial tasks, with an eye toward broader use cases in human-robot interaction, such as elderly care and household assistance. However, the team has not committed to a specific schedule or budget for commercial deployment.
This partnership comes as interest in humanoid robots increases across the technology industry. Other companies are also vying to enter this futuristic market, including Tesla with its Optimus robot, Uptronic’s Apollo, and Figure’s Figure 02.
The research collaboration will be conducted in Boston and will be led by Scott Quindersma of Boston Dynamics and Russ Tedrake of TRI. Their work involves collecting data about Atlas’ performance across a variety of tasks and using that information to train more advanced AI models.
While the technology is exciting, many questions remain about the real-world impact of increasingly capable humanoid robots. Alongside technology development, the research team plans to investigate human-robot interaction and safety considerations.