Reproduce “ELA: End-to-End Latent Action Policy Learning from Video Demonstrations” 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
Cheng-Chun Hsu, Peter Stone
ELA trains a robot imitation policy to use human videos by predicting an intermediate motion representation and decoding it into robot controls. It shapes that representation to be predictable from observations and convertible to actions. Across five real-world tasks, ELA-DP averages 0.70, versus 0.48 for DP, but scores 0.30 versus 0.35 on open-oven, with 20 trials per task.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.