ESC

🚁 Project RAAHAT

Rapid Aerial AI for Hazard Assessment & Tracking
Developed by Team Sutra for the Smart India Hackathon

Project RAAHAT is an Edge-AI powered autonomous UAV reconnaissance system designed for rapid disaster search-and-rescue (SAR) operations.

Built for the Smart India Hackathon, the platform utilizes a Software-in-the-Loop (SITL) architecture to simulate drone flight physics, perform entirely offline on-device AI inference, and mathematically map 2D video pixel detections into precise 3D real-world GPS coordinates for emergency responders — backed by a validated custom hardware payload designed for real-world deployment.

📑 Table of Contents

• System Architecture
• Hardware Architecture
• Directory Structure
• Mathematical Core: Pixel-to-GPS
• Installation & Prerequisites
• Operational Guide (Usage)
• Dashboard Features (ARES-C2)

🏗️ System Architecture

To ensure safety, rapid iteration, and hardware-readiness, Project RAAHAT is built on 4 foundational pillars:

1. SITL Simulation (Webots): Simulates a DJI Mavic 2 Pro with physically accurate IMU, GPS, and optical sensors. The Python-based controller acts as the flight computer, running Boustrophedon (lawnmower) autonomous search patterns via PID stabilization.
2. Edge-AI Computer Vision: Utilizes YOLOv8n optimized for edge GPUs alongside ByteTrack. This ensures real-time victim detection while preventing duplicate counting of the same survivor across consecutive video frames.
3. Hardware Payload & Power System: A custom-designed carrier PCB integrating the compute, sensing, and communication stack onto the physical airframe.
4. ARES-C2 Ground Control Station: A monolithic, Python-native Streamlit dashboard styled to MIL-STD-1472H defense ergonomics. It provides responders with live telemetry, offline map rendering, and geo-triage audit logs.

🔧 Hardware Architecture

Project RAAHAT's hardware stack is engineered so all AI inference happens onboard, ensuring functionality even when network connectivity is entirely unavailable at a disaster site.

Compute & Sensing

Power System

Custom Carrier PCB & Physical Integration

• KiCad Verification: Fully designed project schematic verified with zero Electrical Rules Check (ERC) errors.
• RF Optimization: Includes a Pi-filter LC matching network for optimal LoRa RF output on 868 MHz.
• Physical Assembly: Sourced accurate .STEP CAD files for the airframe, Jetson Orin Nano, and FPV camera module to verify 30.5 × 30.5 mm standoff mounting feasibility and dual payload bay clearance.

Data Flow (Hardware → Software Bridge)

[RGB + Thermal Cameras] ──▶ Jetson Orin Nano (YOLOv8 Inference)
                                      │
[Pixhawk GPS/IMU] ─────────▶ MAVLink  │
                             Telemetry│
                                      ▼
                        [Compact JSON Alert] 
                                      │
                                    (UART)
                                      │
                                      ▼
                          [LoRa Transmitter (SX1276)]
                                      │
                                (868 MHz RF)
                                      │
                                      ▼
                          [Ground Base Station]
                                      │
                                      ▼
                             [ARES-C2 Dashboard]

SIH_DRONE_PROJECT/
├── webots/                     # SITL Simulation Environment
│   ├── controllers/
│   │   └── autonomous_patrol/