Autonomous Multi-Drone Swarm for Disaster Search & Rescue

An autonomous multi-drone swarm system for disaster search-and-rescue operations, simulated in PX4 + Gazebo + ROS 2 Humble. Instead of relying on a single UAV to search a large disaster area, multiple drones coordinate autonomously: a commander drone manages the mission while swarm members divide the search area, detect potential victims, and dynamically reassign tasks if a drone fails or loses communication.

Overview

• Commander: 1x Omnicopter, manages the mission and coordinates swarm members
• Swarm members: 5x Holybro X500 drones, divide and cover the search area
• Coordination: ROS 2 handles inter-drone communication and task distribution
• Key feature: Dynamic task reassignment — if a drone fails or loses comms, its unfinished search region is automatically reassigned to another available drone
• Goal: Rapidly cover large disaster areas, detect survivors, handle drone failures, and complete the mission without manual control of every drone

Architecture

                    ┌─────────────────────┐
                    │   Omnicopter         │
                    │   (Commander)         │
                    │   ROS 2 swarm_commander│
                    └──────────┬───────────┘
                               │
                 uXRCE-DDS Agent (port 8888)
                               │
        ┌──────────┬──────────┼──────────┬──────────┐
        │          │          │          │          │
     X500 #1    X500 #2    X500 #3    X500 #4    X500 #5
    (px4_1)     (px4_2)    (px4_3)    (px4_4)    (px4_5)

Each PX4 SITL instance runs in Gazebo and exposes its fmu/in and fmu/out topics over a namespaced uXRCE-DDS session (/px4_1 … /px4_5, /commander). The swarm_commander ROS 2 node manages arming, offboard mode, mission state, and task reassignment for all six vehicles.

Stack

Component	Version/Tool
Flight stack	PX4 Autopilot (SITL)
Simulator	Gazebo
Middleware	ROS 2 Humble
Bridge	Micro XRCE-DDS Agent
Message types	px4_msgs
Commander airframe	Omnicopter
Swarm airframe	Holybro X500

Repository structure

swarm_ws/
├── src/
│   ├── swarm_commander.py   # ROS 2 node: mission state machine, per-drone control, task reassignment
│   └── swarm_sim.py         # Orchestrates PX4 SITL + Gazebo + DDS agent launch for all 6 vehicles
├── .gitignore
└── README.md

Setup

Prerequisites

• Ubuntu 22.04
• ROS 2 Humble
• PX4 Autopilot (with Gazebo simulation support)
• Micro XRCE-DDS Agent
• px4_msgs built in this workspace

Build

cd ~/swarm_ws
colcon build
source install/setup.bash

Run the full simulation

python3 swarm_sim.py

This will, in order:

1. Check the PX4/Gazebo environment and px4_msgs
2. Start the Micro XRCE-DDS Agent
3. Launch the Omnicopter commander (spawns Gazebo)
4. Launch the 5 X500 slave drones (attach to the same Gazebo instance)
5. Start the ROS 2 swarm_commander node
6. Run the mission and print live swarm status

Ground control (optional but recommended)

Run QGroundControl alongside the simulation — it auto-connects to all 6 vehicles' MAVLink ports (14550–14555) and is useful for monitoring arming state, position, and manually intervening if needed.

Mission state machine

The commander steps every vehicle through: INITIALIZE → PRE_OFFBOARD → OFFBOARD → ARM → TAKEOFF → SEARCH → (FAILURE/REDISTRIBUTE if triggered) → RETURN.

Status

Actively in development — current focus is verifying reliable multi-vehicle arming and stable DDS communication across the commander and all 5 slave drones ahead of Review 2.

License

(Add a license, e.g. MIT, if you want this open for reuse.)