Reproduce “SCOUT: Planning under Occlusion via Object-Centric World Model Rollouts” 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
Amir Hossain Raj, Xuesu Xiao
SCOUT plans shelf rearrangements when targets and access routes are hidden by clutter. It estimates object-level hidden geometry from sparse RGB-D observations, uses a language model to propose candidate moves, and evaluates predicted outcomes with a learned world model and search. In simulation, mean success was 80.8%, versus 72.8% for ParticleFormer; this applies to a controlled initial-view setup, not real homes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.