Reproduce “Terrain-Aware Quadrupedal Locomotion via Reinforcement Learning” 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
Haojie Shi, Qingxu Zhu, Lei Han, Wanchao Chi, Tingguang Li, Max Q.-H. Meng
The authors present a reinforcement-learning controller that uses proprioception and a height map around the feet to adapt a cyclic gait. In simulation, it traveled 4.954 ± 0.070 m on stepping stones, versus 1.539 ± 0.856 m without perception. In physical trials, Max crossed gaps up to 25.5 cm using motion-capture-derived terrain maps; the method provides no safety guarantees.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.