Reproduce “RESETTLE: Robotic Recovery through Disagreement-Triggered Retrieval and Efficient Corrective Control” 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
Yuxin Chen, Senqiao Yang, Zixuan Wang, Jinhui Ye, Changsheng Lu, Pengguang Chen, Shu Liu, Zhuotao Tian and 1 more
RESETTLE detects instability in a frozen policy by comparing independently sampled action chunks under the same observation and instruction, triggering after two consecutive threshold exceedances. It retrieves a same-task demonstration and combines a bounded state-servo motion with a guarded visual residual for a corrective action. On LIBERO-Plus, FastWAM’s case-weighted success rises from 49.3% to 58.0%; disagreement is not a failure detector.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.