Reproduce “Commonsense Reasoning for Legged Robot Adaptation with Vision-Language Models” 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
Annie S. Chen, Alec M. Lessing, Andy Tang, Govind Chada, Laura Smith, Sergey Levine, Chelsea Finn
VLM-Predictive Control (VLM-PC) helps a quadruped navigate obstacle courses by selecting among six preconfigured locomotion skills. A vision-language model uses camera images and interaction history to assess progress, plan several skills ahead, execute the first, and replan. Across five real-world courses, it achieved 64% success versus 36% for a history-informed, single-skill-at-a-time variant, on one robot and fixed skills.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.