🚁 SAVTEK Swarm Core

Decentralized, GPS-Denied Swarm Intelligence Library for UAVs

🎯 Project Overview

SAVTEK Swarm Core is a modular C++ library designed to control heterogeneous drone swarms in communication-denied and GPS-denied environments. It utilizes a peer-to-peer architecture with custom encrypted telemetry and decentralized decision-making algorithms.

Unlike centralized systems, SAVTEK nodes operate autonomously, relying on local neighbor perception and a robust Leader-Follower architecture with Virtual Point formation control.

---

📊 Performance Metrics

Recent field tests demonstrates robust formation stability and precise target acquisition.

Key Achievements

• Formation Integrity: 84% maintained during high-speed maneuvers.
• Precision: 0.5m final target accuracy.
• Stability: 5.5m ± 2.4m average separation, effectively preventing collisions.

---

🚀 Technical Highlights

1. Robust Navigation (Global → Body Frame)

We solved the critical "drifting" issue by implementing a robust coordinate transformation pipeline.

• Problem: Drones applied global frame forces (North/East) directly to body frame motors (Forward/Right), causing incorrect movement when yawed.
• Solution: Implemented real-time yaw extraction from quaternions and a rotation matrix transform:

  // Global Force (North-East) -> Body Velocity (Forward-Left)
  Eigen::Vector3d body_vel;
  body_vel.x() = global_force.x() * cos(yaw) + global_force.y() * sin(yaw);
  body_vel.y() = -global_force.x() * sin(yaw) + global_force.y() * cos(yaw);

2. Virtual Point Formation

Follower drones do not just chase the leader; they chase a dynamic "Virtual Point" relative to the leader's heading.

• Leader: Prioritizes target acquisition.
• Followers: Calculate a unique offset position (Semi-circle or V-formation) based on the leader's current heading.
• Result: Smooth, organic turns where the formation "swings" behind the leader naturally.

3. Scalable Boids Algorithm

Collision avoidance and flocking behavior are handled by a customized Reynolds' Boids implementation optimized with a Spatial Hash Grid for O(n) performance.

---

🛠️ Usage

Installation

# Clone the repository
git clone https://github.com/SAVTEK/savtek_swarm_core.git

# Install dependencies
rosdep install --from-paths src --ignore-src -r -y

# Build
colcon build --packages-select savtek_swarm_core

Running the Simulation

source install/setup.bash
ros2 launch savtek_swarm_core swarm_demo.launch.py

Sending a Command

# Send the swarm to a new target location (x=80, y=20)
ros2 topic pub /swarm/target geometry_msgs/msg/PointStamped "{header: {frame_id: 'world'}, point: {x: 80.0, y: 20.0, z: 5.0}}" --once

---

🏗️ Architecture

• Decentralized Logic: No central server; every drone is a sovereign node.
• SafeSky Protocol: Encrypted (AES-256) inter-drone telemetry using UDP multicast.
• Hardware Agnostic: HAL interface supporting SITL (Gazebo), PX4, and DJI Enterprise.
• GPS-Denied Navigation: Structure ready for VIO and Optical Flow integration.

---

© 2026 SAVTEK Defense Technologies. All rights reserved.