Robotics-STAR-Lab

Robotics-STAR-Lab / OnFly

Public

OnFly: Onboard Zero-Shot Aerial Vision-Language Navigation toward Safety and Efficiency

robotics-star.comOnFly
44
2
100% credibility
Found Mar 22, 2026 at 44 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
AI Summary

OnFly is an academic research project for enabling drones to navigate safely and efficiently using onboard vision and language understanding, with code forthcoming.

How It Works

1
🔍 Discover OnFly

You find this cool drone project while browsing for smart flying tech ideas.

2
👀 Read the overview

You learn how it makes drones navigate safely by understanding sights and instructions.

3
🎥 Watch the demo video

You get thrilled watching real drones fly smartly and avoid dangers onboard.

4
🌐 Visit project page

You check the full website for extra videos, details, and behind-the-scenes info.

5
📄 Explore the paper

You grab the research paper to see the smart ideas that power this tech.

⭐ Get inspired

You're excited for the code release soon and can share or cite this breakthrough in drone smarts.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 44 to 44 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is OnFly?

OnFly enables onboard zero-shot aerial vision-language navigation, letting drones follow natural language instructions like "fly toward the red building" without prior training on specific environments. It prioritizes safety and efficiency by processing vision-language models directly on the drone's hardware, solving the problem of brittle, map-dependent nav stacks in dynamic outdoor settings. Code is coming soon, building on arXiv research with demos on Bilibili.

Why is it gaining traction?

It stands out with sub-10g lightweight models like OnFly X3D for real-time onboard inference, beating sim-only alternatives in canyon-like terrains via zero-shot generalization. Developers dig the efficiency gains—safety toward obstacles without heavy sim-to-real tuning—and the project page's OnFly Canyon/OnFly 7 benchmarks. Early buzz from robotics labs hooks those chasing deployable VLN.

Who should use this?

Aerial robotics engineers building autonomous drones for search-and-rescue or delivery, tired of offline planners. Researchers in vision-language nav prototyping onboard systems for sub-10cm precision. Teams onboarding OnFly for efficiency in unstructured environments like canyons.

Verdict

Hold off—1.0% credibility score, 44 stars, and "code coming soon" mean it's pre-release research, not production-ready. Watch for the drop if zero-shot aerial nav fits your stack; pair with the paper meanwhile.

(178 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.