AIToday
RoboticsRobotics & Automation NewsPublished: Nov 23, 2026, 01:00 JST

Boston Dynamics puts wheels on Atlas for reinforcement learning tests

Send on LINE
Boston Dynamics puts wheels on Atlas for reinforcement learning tests

3 Key Points

  1. What happened

    Boston Dynamics released video of its Atlas humanoid rolling on wheels attached via custom adapters and scooter rims, sliding sideways and landing a 360-degree jump, as part of a reinforcement learning research project.

  2. Why it matters

    The wheeled setup let researchers test Atlas's motors across a wide range of speeds and torque demands, and see how accurately behaviors learned in simulation transferred to physical hardware, with findings already being applied to the standard configuration.

WHO IT HITSRobot hardware and controls engineers working on humanoid platforms should take note: the project shows how simulation-trained policies can be used to probe actuator limits and transfer learned behaviors to physical robots, and Boston Dynamics says the findings are already being applied to Atlas's standard configuration.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Context & Analysis

Boston Dynamics frames the wheeled Atlas as a deliberate test of how well behaviors learned in simulation hold up on physical hardware. By swapping the robot's feet for scooter rims attached via custom adapters, researchers could push the existing ankle actuators across a wide range of speeds and torque demands — conditions that are harder to explore with feet alone.

The project combines traditional motor control engineering with whole-body reinforcement learning, trained across millions of simulated scenarios. The body notes that some movements, including sliding maneuvers and the 360-degree jump, originated from human motion capture recordings that the robot adapted to its wheeled configuration. It also highlights the risk of mismatches between simulated and physical actuators, which can produce movements that cannot be reliably reproduced in the real world.

Boston Dynamics says the findings are already feeding back into Atlas's standard configuration, with work aimed at improving the robot's ability to respond to actuator failures and other unexpected operating conditions. That suggests the wheeled experiment is less about a new form factor than about hardening the control systems that would support reliable humanoid operations in real-world environments.

FAQ
Is Boston Dynamics making a wheeled version of Atlas?
No. A controls engineer at Boston Dynamics, Xiaoyu (Andy) Wu, said a wheeled Atlas isn't a product but a research testbed, and the ultimate objective is more robust control systems, not a wheeled humanoid.
How were the wheels attached to Atlas?
Engineers temporarily replaced Atlas's feet with wheels, using custom adapters to attach commercially available scooter rims to the robot's existing ankle actuators.
How did Boston Dynamics train the movements?
The project used millions of simulated scenarios and whole-body reinforcement learning before transferring behaviors to the physical robot. Some movements, including sliding maneuvers and the 360-degree jump, were based on human motion capture recordings.
Robotics & Automation NewsRead Original Article
Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleSamsung, SK hynix, Micron set robot charging standard for chip fabs