DroneOS

Open-source autonomous drone control framework

DRONEOSAI Fleet Command
AI fleet orchestration for PX4 Autopilot + ROS 2 · powered by OpenClaw
DRONE FLEET
drone1IDLE92%
drone2FLYING64%
drone3RTL71%
INCIDENT DISPATCH FEED
1 active · 8 total
P1medical emergencyRESOLVED
P3property damageRESOLVED
P1structure fireDISPATCHED
P2suspicious activityRESOLVED
P1traffic accidentRESOLVED
P3noise complaintRESOLVED
P1medical emergencyRESOLVED
P2suspicious activityRESOLVED
ACTIVE
drone2 FLYING SAFE ALT 50.0m XY (80, -40)SIM: Gazebo Baylands480p
ADAsonnet
>command...
MAP
D1
D2
D3
🔥
CONTROLSONLINE
ALT 50m
AI LOG
02:19:00AI DISPATCH ONLINE
DroneOS Console v1.0 — type "help" for commands
>arm, offboard, pos 0 0 -15, land, state, help

The DroneOS Command Center — real-time fleet management with AI-powered dispatch.

DroneOS is a framework for autonomous drone control built on ROS 2 and PX4 Autopilot. It provides the developer-friendly abstraction layer between raw flight controller topics and high-level AI-driven operations — from single drone commands to multi-drone fleet coordination.

AI Fleet Dispatch

Autonomous emergency response — AI receives incidents, evaluates the fleet, selects the best drone, and dispatches it. Incident to airborne in ~10 seconds.

Cloud-to-Drone Control

Commands from anywhere over Tailscale VPN. Real-time telemetry, camera feeds, and fleet status streamed to the ground station via WebSocket.

Simulation-in-the-Loop

Full PX4 SITL environment with Gazebo. Test AI decision-making and fleet coordination in simulation before flying real hardware.

Components

drone_core

C++ library and ROS 2 node providing high-level drone control APIs for arming, takeoff, landing, and position control.

drone_control.py

Python library + CLI for remote drone operations over WebSocket. Works from anywhere that can reach rosbridge — including cloud VPS over Tailscale VPN.

AI Fleet Dispatch

Autonomous emergency response system. AI receives incidents, evaluates fleet status, selects the best drone, and flies it to the scene in ~10 seconds.

Ground Station UI

React/TypeScript web interface with real-time telemetry, camera feeds, interactive map, and built-in AI chat for natural language drone control.

camera_ros

ROS 2 camera node using libcamera for Raspberry Pi camera modules. Streams video to the ground station via web_video_server.

Micro-XRCE-DDS-Agent

Bridges PX4 flight controller (running XRCE-DDS client) to the ROS 2 DDS network over serial or UDP.

Architecture

The system splits across two machines: a cloud VPS running the AI fleet commander, dispatch services, and ground station UI — and a drone-side machine running PX4, the DDS bridge, and the drone SDK. Communication happens over Tailscale VPN with ~98ms round-trip latency.

Cloud (VPS)

AI agent (Ada), dispatch service, bridge, ground station UI, rosbridge relay. Orchestrates the fleet.

Drone-Side

PX4 SITL or real flight controller, Micro-XRCE-DDS Agent, drone_core SDK, rosbridge, camera streaming.

Measured Performance

Incident-to-dispatch: ~10-12s. AI decision time: ~3-4s. Drone command execution: ~2.5s over VPN.

Get Started

DroneOS runs on the companion computer (Raspberry Pi), not the flight controller. The GitHub README includes setup instructions for both SITL development and real drone deployment.