Reproduce “Sequential Recovery Updates for Learned Controllers under Successful-Trajectory Preservation” 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
Zongtan Li, Tianming Liu, Ye Zhao
The authors present a humanoid control method that adds recovery skills while constraining when new skills may intervene. It learns entry rules, excludes states on successful recorded trajectories, and admits skills after autonomous recovery and replay checks. Across four simulation panels, updates add 62 recoveries among 192 perturbations with no base-success losses; removing exclusion adds 33 but disrupts protected behavior.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.