Roadmap
What we have built, what is working now, and what comes next.
Completed
DroneOS SDK
2025Custom C++ drone control library built on ROS 2 and PX4 Autopilot. High-level APIs for flight control, state tracking, and multi-drone coordination. CLI and Python API for programmatic access. Docker-based deployment for simulation and real hardware.
Real Hardware Deployment
2025Deployed on custom X500 airframe with Pixhawk flight controller and Raspberry Pi 5 companion computer. Cloud-to-drone communication over 4G via Tailscale VPN. Production Docker containers on the companion computer.
Autonomous AI Dispatch (Simulation)
2026Integrated OpenClaw AI agent for autonomous fleet command in PX4 SITL. AI receives emergency incidents, evaluates live fleet status, selects the optimal drone, and flies it with no human pilot. Incident to airborne in about 10 seconds, with multi-drone coordination and a real-time web command center.
Next, Simulation
Object Detection in Simulation
Connect the object detection pipeline to simulated camera feeds. Drone sees something in Gazebo, detection runs, agent reacts. Infrastructure exists, but it still needs workflow integration.
Multi-Drone Coordination
Move beyond basic multi-drone dispatch to obstacle avoidance and airspace deconfliction. Requires virtual sensors and coordination logic after single-drone object detection is working.
Next, Framework
Modular Plugin Architecture
Restructure DroneOS into swappable modules for flight stack adapters, perception backends, and AI integrations.
Rust Rewrite
Rewrite the core SDK in Rust for safety, performance, and more modern tooling around a flight-critical codebase.
Multi-Flight-Stack Support
Abstract the flight stack layer so DroneOS can support PX4 today and additional autopilots later without rewriting the control layer.
Developer Experience
Expand documentation, quickstarts, and tooling so developers outside the project can get value without deep ROS 2 and PX4 knowledge.
Next, Hardware
4G Reliability for AI Dispatch
Validate the full AI dispatch loop over cellular on real hardware, not just basic drone control over Tailscale VPN.
Object Detection on Real Hardware
Take the Coral Edge TPU setup beyond example inference into real model tuning and live integration with the drone agent.
Obstacle Avoidance Sensors
Add dedicated sensors for obstacle detection in real-world flight, likely through a mix of range, lidar, and computer vision approaches.