RoboBoom
Boston Dynamics· November 9, 2026· 2:51

Atlas on a Roll | Classical Motor Controls and Whole-Body RL | Boston Dynamics

Andy Woo, a controls engineer on the Boston Dynamics Atlas team, discusses motor control, wheeled maneuvers, and reinforcement learning for Atlas.

In this talk

  1. 0:13

    A motion-capture 360

    Woo says the 360 jump was adapted from a human motion-capture recording. The team applied its trajectory-targeting training pipeline to a library of motion trajectories, and the move worked on the first attempt.

  2. 0:32

    Testing actuation limits

    As an Atlas controls engineer, Woo works to make motors and actuators behave reliably while testing how they perform in demanding situations. He says the wheels were a way to push the actuation hardware, rather than the main goal of the project.

  3. 1:21

    Wheels in side dashes

    Woo describes side dashes that use a jump and lateral slide. He was surprised that training produced a strategy in which Atlas rolls on its wheels during the maneuver.

  4. 1:46

    Learning to handle failures

    Woo says he is exploring how the low-level actuation system can work with Atlas's whole-body reinforcement learning stack. His next project focuses on how the policy might respond when actuators fail and whether reinforcement learning can produce useful responses.

  5. 2:16

    Simulation to hardware

    Woo says the behavior transferred from simulation to the physical robot with little adjustment. He enjoys changing simulation settings and seeing how those changes affect Atlas on hardware, and says adding wheels made the work more enjoyable.

Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.

More talks