Reproduce “Push the Limit: Learning High-Speed Sprinting and Aggressive Cornering on a Full-Scale Humanoid” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Huaxing Huang, Ruidong Li, Jinchao Han, Ziyang Dong, Wenhao Cui
The authors present a humanoid controller for sprinting, reference-guided motion, and high-speed cornering. It encodes 10 sensor-history frames with multi-head attention, uses travel-direction feedback for turns, and routes motion-style feedback by commanded skill. In Gazebo, its heading-reward policy turns at 6.5 m/s with a 24.2 m radius, versus 146.7 m with yaw-rate reward; separately, a weight-reduced robot averaged 7.14 m/s over 100 m.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.