ARIA Robotics

AI-powered disaster response and field reconnaissance system for rapid situational awareness. ARIA Robotics combines computer vision, GPS tracking, zone scoring, and web dashboards to help identify affected areas, assess damage severity, and guide dispatch decisions in real time.

This repository contains a full-stack disaster intelligence workflow for both Raspberry Pi deployments and laptop-based experimentation, along with ESP32 hardware firmware for vehicle/sensor control and monitoring.

Why this project exists

During disaster response operations, teams need fast answers to questions such as:

• Where is the damage concentrated?
• Which zones are highest priority?
• Are there people, fires, smoke, debris, or blocked roads?
• Can response teams be routed quickly with the right equipment?

ARIA Robotics addresses this by combining:

• image and video analysis
• multi-model AI inference
• GPS and geolocation data
• live map / zone prioritization
• emergency action recommendations
• Raspberry Pi and embedded hardware integration

Key features

• Multi-model disaster scene analysis
  • xView2-style damage assessment
  • YOLOv8 fire and smoke detection
  • LADI-v2 classification analysis
  • Roboflow-based ambulance/vest detection
  • Groq Vision scene verification and summarization
• Zone scoring engine
  • weighted region severity scoring
  • green / yellow / red classification
  • suggested emergency response action
• GPS-aware dispatch workflow
  • live coordinate reporting
  • browser geolocation fallback
  • support for field deployments
• Web dashboard
  • Flask-based UI
  • image/video upload processing
  • live feed processing mode
  • zone tracking and status API endpoints
• Embedded hardware support
  • ESP32 firmware for monitoring, sensors, and vehicle systems
  • motor / sensor control sketches for field robotics

Project architecture

The repo is organized into a few primary areas:

• Pi/ - Raspberry Pi deployment version of the Flask application
• Laptop_Only/ - desktop/laptop development version with local model support
• root-level .ino files - ESP32 and Arduino sketches for vehicle control, dashboard, calibration, monitoring, and GPS-linked functionality
• templates/ - UI templates used by the app

Repository layout

ARIA_Robotics/
├── Pi/
│   ├── app.py
│   ├── main.py
│   ├── requirements.txt
│   └── templates/
├── Laptop_Only/
│   ├── main.py
│   └── templates/
├── templates/
├── README.md
├── gps_auto_return.py
├── auto_return.py.txt
├── calibration.ino
├── car-dashboard.ino
├── car_with_telegram.ino
├── esp32s3-car.ino
├── final-car.ino
├── final-monitor-system.ino
├── monitor_sys.ino
├── mpu&hmc.ino
├── rth.ino
├── v1.py
├── v2.py
├── wihtout-mag.ino
├── without-dust.ino
└── ...

How it works

1. A user uploads an image or video, or pushes a live camera frame.
2. The backend runs multiple visual inference models in parallel.
3. Detected classes are mapped into disaster categories such as:
   • building damage
   • road damage
   • fire
   • smoke
   • flood
   • debris
4. A weighted zone score is computed from the detections.
5. The system suggests a response action based on the resulting severity level.
6. The output is displayed via a dashboard and saved as a result image/video.

Tech stack

• Python
• Flask
• OpenCV
• NumPy
• Requests
• YOLO-based vision workflows
• Roboflow integrations
• Groq API
• ESP32/Arduino firmware
• GPS and sensor hardware integration

Raspberry Pi deployment

The Pi/ folder contains the main application used for deployment on a Raspberry Pi.

Requirements

From the Pi/ directory, install dependencies:

cd Pi
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Environment variables

Create a .env file in the Pi/ directory with values similar to:

ROBOFLOW_API_KEY=your_key_here
GROQ_API_KEY=your_key_here
COLAB_MODEL_1_URL=https://your-colab-endpoint.example
COLAB_MODEL_2_URL=https://your-colab-endpoint.example
COLAB_MODEL_3_URL=https://your-colab-endpoint.example
GPS_SERIAL_PORT=/dev/serial0
GPS_BAUD_RATE=9600

Run the app

cd Pi
python3 main.py

The app serves on:

http://0.0.0.0:8000

Laptop / development version

The Laptop_Only/ folder is intended for local experimentation and testing on a desktop or laptop.

This version includes support for a local YOLOv8 fire/smoke model path and browser-based geolocation fallback.

Run it from the Laptop_Only/ directory:

cd Laptop_Only
python3 main.py

Hardware and embedded systems

The root-level Arduino/ESP32 files are used for:

• vehicle control
• calibration routines
• motor and sensor integration
• telemetry dashboards
• monitoring systems
• GPS return-to-home / autonomous movement logic

The repository also includes a hardware wiring guide in the original README content, which is relevant for physical integration and pin mapping.

Typical usage scenarios

• disaster reconnaissance after earthquakes or floods
• wildfire perimeter monitoring
• road blockage and infrastructure assessment
• search-and-rescue support
• field deployment with GPS-tagged impact zones
• rapid response dispatch prioritization

Notes and caveats

• Some external integrations (Roboflow, Groq, Colab-hosted models) require valid API keys and endpoint URLs.
• The project is designed for research, prototyping, and field deployment experiments rather than as a turnkey production platform.
• Local model paths and hardware wiring may need adjustment depending on your environment and device configuration.
• The repository currently contains both prototype code and embedded firmware, so the exact setup depends on which deployment mode you want to use.

Future improvements

Potential enhancements include:

• map UI with real-time markers
• persistent zone database storage
• improved telemetry and historical event tracking
• stronger production deployment structure
• better packaging and Docker support
• more robust model orchestration and failover logic

License

This project does not appear to include a dedicated license file in the repository root. If you plan to publish, distribute, or commercialize this project, it is recommended to add an explicit license such as MIT, Apache 2.0, or GPL.

Contributing

Contributions are welcome if you are improving:

• inference quality
• dashboard behavior
• field hardware reliability
• deployment scripts
• documentation

Contact / project origin

This repo is maintained under the username EgoisticCoder and is intended as an AI-driven disaster response robotics and reconnaissance project.

---

If you want, I can also make this README more polished for GitHub by adding:

• badges
• a screenshot section
• a demo GIF/video section
• a more technical installation guide
• a shorter “project pitch” version for public sharing

I can also generate a version tailored specifically for a Raspberry Pi deployment or a clean landing-page style README.