Reproduce “ODE Policy: Robot Learning via Action Dynamics, Not Actions” 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
Jindou Jia, Jingliang Li, Kuangji Zuo, Tuo An, Yuxuan Hu, Mengfei Zhao, Bofan Lyu, Jiaqi Bai and 4 more
ODE policy generates commands by predicting how an action changes and integrating those changes into chunks, each continuing from the last commanded action. This makes the commanded trajectory continuous, though the robot may not track it exactly. On LIBERO, it reaches 79.2% after one epoch, versus at most 43.2% for six baselines; authors report highest success on 14 real-robot tasks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.