Reproduce “Reinforcement Learning from Real-Robot Interaction without High-Quality State Estimates” 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
Ryan Gao, Ashvin Anilkumar, Nicholas Corrado, William Cong, Josiah Hanna
RSSM–RLPD gives a walking NAO robot predictive ball-state memory during dribbling. It updates its estimate when the ball is visible and predicts while it is out of view, using onboard sensing and online experience. It scored five evaluation configurations within four minutes per goal, versus 80% for GRU-RLPD and 0% for feed-forward RLPD and behavior cloning in this task.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.