a day ago
Humanoid robot able to do complex tasks with little code added
Aug. 20 (UPI) -- A humanoid robot can now perform complex tasks with a large behavior model without needing hand programming for each task.
Boston Dynamics and Toyota Research Institute announced this breakthrough Wednesday in a press release.
In a video jointly released by the two organizations, Atlas performs a long, continuous sequence of complex tasks that require combining object manipulation with locomotion. By adopting LBMs, new capabilities that would have been laboriously hand-programmed in the past can now be added quickly and without writing a single new line of code.
The video shows Atlas using whole-body movements, such as walking, crouching, and lifting, to accomplish a series of packing, sorting, and organizing tasks. Throughout the sequences, researchers interject unexpected physical challenges mid-task, such as closing the lid of a box and sliding it across the floor, requiring Atlas to self-adjust in response.
Humanoid robots that have demonstrated this capability before typically separate the low-level walking and balancing control from the control of the arms for manipulation. But a single large behavior model has direct control of Atlas, treating the hands and feet almost identically.
This breakthrough is the result of the October 2024 joint research partnership between Boston Dynamics and TRI, which was designed to use their combined strengths and expertise to speed the development of smart robots.
The project, co-led by Scott Kuindersma and Russ Tedrake, is conducting research to answer fundamental questions about humanoid robots and large behavior models, to advance the field's understanding of large models for whole-body control, including advanced manipulation and dynamic behaviors.
"This work provides a glimpse into how we're thinking about building general-purpose robots that will transform how we live and work," Kuindersma, vice president of Robotics Research at Boston Dynamics, said in a statement. "Training a single neural network to perform many long-horizon manipulation tasks will lead to better generalization, and highly capable robots like Atlas present the fewest barriers to data collection for tasks requiring whole-body precision, dexterity, and strength."
Humanoids can help in many ways.
"One of the main value propositions of humanoids is that they can achieve a huge variety of tasks directly in existing environments, but the previous approaches to programming these tasks simply could not scale to meet this challenge," Tedrake, senior vice president of Large Behavior Models at Toyota Research Institute, said in a statement. "Large behavior models address this opportunity in a fundamentally new way -- skills are added quickly via demonstrations from humans, and as the LBMs get stronger, they require less and less demonstrations to achieve more and more robust behaviors."
Humanoid robots have been an ongoing challenge. The World Humanoid Robot Games highlighted this, showing that tumbling robots is the norm.
The games featured more than 500 robots from 16 countries, including Japan, the United States and Germany, competing in 26 events.
One robot was disqualified from a 1,500-meter race when its head fell off in mid-stride.
"Keeping [the head] balanced while in movement is the biggest challenge for us," said Wang Ziyi, 19, a member of the Beijing Union University team.
