• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI
    Reaction

    How humanoid robots' control techniques could help wheeled robots

    A post argues that reinforcement-learning whole-body control techniques developed for humanoids could also benefit wheeled robots, particularly in safety, reliability and adaptation.

    XW
    CP
    KH
    7 Sources, ,

    TLDR

    The post predicts that wheeled robots will benefit from reinforcement-learning whole-body control techniques developed for humanoids, citing safety, reliability and adaptation.

    Combined views

    20.6K

    7 Sources, first seen 18h ago

    158 likes

    Combined views

    20.6K

    7 Sources, first seen 18h ago

    158 likes
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    18h ago
    first seen 18h ago
    13 comments
    49 saves
    20 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Featured Source
    13 comments
    49 saves
    20 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    7 Sources

    @chris_j_paxtonEven wheeled robots will benefit from the RL whole body control techniques developed for humanoids - for safety, reliability, adaptation
    @xiaolonw@chris_j_paxton What does RL add here compared to doing IK?
    @kun_h____Why RL for System 0? Wheeled semi-humanoids are sometimes treated as a “simple” control problem because they don’t need to walk. I think that misses most of the challenge, unless you’re happy with a robot that moves like a turtle. RL gives us a dynamics-aware controller that can track aggressively and precisely while staying compliant and robust to infeasible or adversarial commands. Training it on human trajectories also gives us a useful prior over natural whole-body postures, which lets us keep the tracking objective surprisingly simple. In our teleop benchmarks, this translated directly into faster task completion.

    7 Sources

    @chris_j_paxtonEven wheeled robots will benefit from the RL whole body control techniques developed for humanoids - for safety, reliability, adaptation
    @xiaolonw@chris_j_paxton What does RL add here compared to doing IK?
    @kun_h____Why RL for System 0? Wheeled semi-humanoids are sometimes treated as a “simple” control problem because they don’t need to walk. I think that misses most of the challenge, unless you’re happy with a robot that moves like a turtle. RL gives us a dynamics-aware controller that can track aggressively and precisely while staying compliant and robust to infeasible or adversarial commands. Training it on human trajectories also gives us a useful prior over natural whole-body postures, which lets us keep the tracking objective surprisingly simple. In our teleop benchmarks, this translated directly into faster task completion.