Reproduce “"Hi AirStar, Guide Me to the Badminton Court."” 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
Ziqin Wang, Jinyu Chen, Xiangyi Zheng, Qinan Liao, Linjiang Huang, Si Liu
Beihang University researchers developed AirStar, an intelligent aerial assistant that integrates Large Language Models as a cognitive core for Unmanned Aerial Vehicles. It enables natural interaction via voice and gestures, performs vision-and-language navigation across different scales, and offers advanced capabilities like cross-modal question answering and target tracking.
Key insights
- Large Language Models can effectively serve as a high-level planner and reasoning engine for embodied robots, particularly UAVs, in dynamic 3D environments.
- A modular "Tool Factory" architecture, encompassing navigation and skill libraries, enables extensibility and seamless integration of new robotic capabilities for general-purpose agents.
- Combining geospatial knowledge with visual reasoning provides robust vision-and-language navigation across varying scales in complex outdoor settings, addressing multi-scale challenges.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.