Reproduce “Retiming Robot Demonstrations with a Learned Speed Prior” 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
Seungyong Moon, Deokjae Lee, Hansang Lee, Munkhtselmeg Battsengel, Youngin Kim, Seonghyeon Yun, Sungjin Jeon, SongEun Kim and 2 more
The authors learn a speed prior that transfers timing from handheld demonstrations to robot paths without changing those paths, then use a controller to execute the new pace. On 120 RoboLab tasks, the policy reaches 2.38× mean speed and 60.8% success, versus 2.48× and 53.0% for uniform acceleration; the speeds are close, not identical, and real-robot execution depends on the controller.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.