Reproduce “ScaleNav: Scalable Internet Video Pretraining for Language-Conditioned Local 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
Yihui Yang, Yuanhong Yu, Jiawei Wang, Zhongrui Yu, Jia Shi, Kun Zhan, Ruizhen Hu, Xiaowei Zhou and 2 more
ScaleNav enables robots to navigate to initially visible objects by learning visual approach dynamics from action-free Internet videos, then aligning them with executable actions using labeled simulation data. Its shared model predicts action chunks for closed-loop execution. It reaches 94.9% success on HabitatGS-LN and 83.3% success across 60 real-robot trials in three indoor environments, but broader robustness is untested.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.