Skip to content

Jeffuz/AERIS

Repository files navigation

AERIS: Autonomous Emergency Response & Intelligence System 🚁🆘

Inspiration 🌪️📉

During disasters, over 60% of victims lose phone service, cutting them off from emergency help.
Rescue teams are forced to search blindly, wasting precious hours while lives hang in the balance.
We realized search and rescue needs to be faster, smarter, and autonomous.
That’s why we built AERIS — AI drones that find victims and guide rescuers automatically.

What it does 🎯

  • 🚁 Locates victims needing urgent care using autonomous drone navigation.
  • 🚑 Dispatches first responders directly to the victims' coordinates.

How we built it 🛠️

  • 🎨 Front end: Next.js, Tailwind CSS, Shadcn/ui, Leaflet, Butterfly for a smooth, responsive command dashboard.
  • 🎥 Live Video Streaming: Dockerized NGINX RTMP server + FFmpeg ingests the DJI drone feed
  • 🧠 AI agents: Dain for intelligent mission control orchestration, including hazard detection and hospital coordination.
  • 🚀 CV Pipeline Pipeline: FastAPI server exposing REST endpoints to ingest frames, run YOLOv8 object detection, invoke Groq LLaMA for victim triage, and persist results to Supabase
  • 🗄️ Database: Supabase for real-time detection/event logging
  • 👁️ Object detection: Ultralytics YOLOv8 custom model trained on composed disaster datasets (humans, fire, smoke, flood, debris, collapse) for live inference
  • 📞 Emergency Response Automation: Dain agents locate the nearest hospitals based on pinned coordinates, list phone numbers and distances, and automate emergency calls via Twilio API

Challenges we ran into 🧩

  • 🔄 Getting hospitals and dispatchers synced with Supabase without massive lag.
  • 🧠 Designing complex systems combining computer vision, AI agents, and real-time routing without everything breaking.

Accomplishments that we're proud of 🏆

  • 🖥️ Displaying urgent victim and hazard data in a clean, intuitive way.
  • 🚑 Getting first responders moving toward critical victims faster than traditional search teams.

What we learned 📚

  • 🤯 AI agents are wildly hard to coordinate when they actually need to act in the real world.
  • 🧹 Always design a stupid simple MVP first, then layer complexity only after basic success.

What's next for AERIS 🚀

  • 🤖 Fully autonomous dispatching without human coordination.
  • 🔍 Smarter triage: better hazard detection, victim urgency evaluation, and threat prioritization.

About

Autonomous Emergency Response & Intelligence System

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors