Reproduce “Demonstration-Free Skill Composition for Navigation and Manipulation” 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
Madeleine Nouri, Gelareh Mohammadi, Francisco Cruz
The authors evaluate skill composition in simulated navigation and manipulation. They train policies for local goals and hand off between them; navigation uses manual plans and manipulation scripted controls. On FourRooms layouts, success is 0.555 versus 0.020 for flat DQN; manipulation success is 0.858 versus 0.914 for flat SAC. These results do not establish real-robot performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.