सुपर्ण — Autonomous Surveillance Path Optimization

A bio-inspired fixed-wing UAS for persistent day/night battlefield reconnaissance.
Inspired by the Common Swift (Apus apus)—a bird that sustains unbroken flight for 10 months—SUPARNA encodes aerodynamic constraints directly into the path planning algorithm.

---

📑 Table of Contents

• ⚡ The Problem: The High Cost of Hover
• 🦅 The SUPARNA Solution
• ⚙️ System Specifications
• 💻 The Repository: PCCE Software Stack
• 🚀 Quick Start (Simulation)
• 📁 Repository Structure

Persistent Endurance Sea level battery-only ops

Continuous Forward Flight Observation via loiter patterns

Runway-Independent Controlled spiral descent

---

⚡ The Problem: The High Cost of Hover

Every rotorcraft ISR drone in Indian tactical service wastes 70–80% of its energy on hovering — the observation mission runs on the leftovers. This fundamentally restricts operations to 25–45 minute sorties, requires 6–8 battery swaps per 4-hour operation, and leads to operational collapse above 3,000m AMSL (like in Ladakh).

This is a physics problem. No software update can fix it.

❌ Target-Centric (Traditional) Stop and hover over points of interest. Extreme energy drain fighting gravity. Rotor aerodynamic collapse at high altitudes. ~1 km² coverage per sortie.

✅ Motion-Centric (SUPARNA) Continuous circular loiter patterns. Thrust entirely forward-directed. +38% more coverage per joule. ~3 km² coverage per sortie at ≥95% density.

"SUPARNA converts energy directly into coverage — not hover. Every joule translates into ground observed."

---

🦅 The SUPARNA Solution

SUPARNA tackles the challenge at the airframe level. Derived from the Common Swift, the platform is designed so that hover is structurally impossible.

Key Innovations

👉 Interactive Feature Grid

🧬 PCCE Planner By encoding forward-flight-only as a hard physical limitation, the algorithm collapses the search space, allowing O(1) Dubins transits and rapid Greedy set covers.

🎯 Loiter-to-Land Landing inside the observation circle. The drone spirals down 3-5m per loop until belly touchdown. Any loiter zone is a potential recovery site. No parachute. No runway.

🏔️ High-Altitude At 4,000m AMSL (like in Ladakh), traditional rotorcraft endurance plummets. SUPARNA delivers 2.65 hours at 4,000m, fundamentally changing tactical surveillance.

---

⚙️ System Specifications

Parameter	Specification
🪽 Airframe Structure	210cm wingspan, CFRP fixed-wing, Common Swift crescent planform
🚀 MTOW & Power	3.5 kg MTOW 6S4P Samsung 21700-50E (432 Wh, 1.66 kg)
⏱️ Endurance	3.25 hr (Sea Level) 2.65 hr (4,000m AMSL)
🎥 Payload	Dual EO/IR on 2-axis gimbal. 4K EO (day) + FLIR Lepton 3.5 (LWIR, night)
🧠 Flight Control	ArduPlane on Cube Orange+ (EKF3, Dual IMU) + RPi CM4 companion
📡 Comms (Triple-Link)	900MHz FHSS primary, 433MHz LoRa fallback, 868MHz RC override
🔇 Acoustic Signature	<48 dB at 150m AGL (Covert ISR)

---

💻 The Repository: PCCE Software Stack

This repository hosts the Physics-Constrained Coverage Engine (PCCE) — a standalone, platform-agnostic sovereign path planner validated in full 3D simulation.

graph TD;
    A[Grid Map, Obstacles, Start Pos] --> B[Obstacle Inflation];
    B --> C[A* Pathfinder + Simplification];
    A --> D[Coverage Planner: Greedy Set Cover];
    D --> E[Transition Planner: Dubins Curves];
    C --> F[Reactive Avoidance: Bug2 + 7-Ray Scan];
    E --> G[Mission Path & Loiter Sequence];
    G --> H[Simulation & Energy Log];

Loading

The algorithm yields optimal coverage efficiently via:

• Greedy Set Cover: Energy-weighted coverage grouping Score = Coverage ÷ Energy
• Dubins Curves: O(1) query for 6 curve archetypes (LSL, LSR, RSL, RSR, RLR, LRL) providing shortest provably flyable non-holonomic transition distances.
• Bug2 Avoidance: 7-ray raycasting for reactive navigation (NORMAL → AVOID → RECOVER).

---

🚀 Quick Start (Simulation)

You can run the simulated drone and PCCE path visualizer out-of-the-box.

Tip

Action In-Simulation: Watch how the drone strictly avoids hovering, opting instead for Dubins curve transitions and continuous circular sweeps across the terrain.

(Replace this placeholder segment with a GIF/Video recording of main.py running)

# Clone
git clone https://github.com/404Avinash/suparna_beta.git
cd suparna_beta

# Install dependencies
pip install -r requirements.txt

# Run interactive simulation
python main.py

Controls

Key	Action
SPACE	Pause / Resume
+ / -	Speed up / down
R	Reset mission
ESC	Exit

To generate a robust LAC Border mission profile or randomly generated terrain for web visualization (Three.js compatible mission.json):

python export_mission.py --map lac --seed 42

---

📁 Repository Structure

📦 suparna_beta
 ┣ 📂 assets                      # Hero banners and media
 ┣ 📜 main.py                     # Simulator Entry point
 ┣ 📜 export_mission.py           # Generate mission & export JSON
 ┣ 📜 PROJECT_ARCHITECTURE.md     # Detailed software architecture & algorithms
 ┗ 📂 src
   ┣ 📂 core                      # Geometric utils, Dubins algorithms
   ┣ 📂 planners                  # Coverage, transition, reactive edge planners
   ┗ 📂 simulation                # Pygame visualization & drone state-machine

---

Tech Stack

---

Built within India.

Made with 🦅 by Avinash Jha