Reproduce “Quadruped Locomotion on Non-Rigid Terrain using 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
Taehei Kim, Sung-Hee Lee
The authors train a simulated quadruped to walk on flat elastic tiles that sink. The policy plans each step using recent foot-position history and foot velocity; removing one phase of history prevented learning, while removing velocity yielded balancing without forward motion. A 55 cm-base Laikago walked on tiles sinking up to 5 cm in simulation, not on a physical robot.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.