Reproduce “ScheNav: Schematic-Guided Object Goal Navigation” 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
Guichen Huang, Chenchen Jing, Che Sun, Yuwei Wu, Yunde Jia
ScheNav lets agents use rough schematic guidance alongside observations to navigate to object goals. Its learned SchemaField predicts distance and corridor fields, while rule-based decisions select frontiers and approach detected goals. Sequence training and cross-frame consistency support coherent decisions across viewpoints. On UniGoal’s instance-level HM3D benchmark, it achieves 61.4% success and 36.3% SPL across 1,000 episodes, but baselines receive textual descriptions rather than schematic guidance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.