The Foundation Behind PX4 Autopilot, Pixhawk, and the Future of Open Source Drones

Drones, such as quadcopters, are still relatively new technology. But we didn’t get to this point because we made great strides in electric motor or battery technology; drones were finally able to take off because of improved sensors and control processing. For a decade now, Dronecode Foundation has been at the forefront of the open source side of that technology, maintaining well-known projects like PX4 Autopilot and Pixhawk. The Dronecode Foundation 2024 Annual Report highlights achievements of the past 10 years and gives us a forecast for future development.

Stable flight is a tech marvel

Sensors and control are crucial for multirotor drones, because they can only achieve stable flight through the precise balance of all rotors.

An airplane is naturally stable and only needs engines for forward thrust in order to maintain lift. That’s why a glider doesn’t require any power at all for stable flight. Even helicopters are very stable, in comparison to multirotor drones. A helicopter’s main rotors provide all of the lift, while the tail rotors resist rotation caused by the torque of the main rotors. A trained helicopter pilot can maintain stable flight entirely through manual control.

But it would be almost impossible for human to maintain stable quadcopter flight if they had to control each motor manually. Just to hover, they would have to provide constant and precise adjustments to all four motors, all at the same time.

For that reason, a flight controller is a necessity. The pilot tells the flight controller how they want the drone to move, and the flight controller handles the tiny motor speed adjustments necessary to achieve that.

Because the drone’s center of gravity may change and wind conditions are unpredictable, the flight controller must contain sensors capable of detecting orientation. If the drone is supposed to hover and the flight controller detects motion, it can adjust the speeds of the appropriate motors to compensate.

Flight controllers have to be powerful and sophisticated to pull that off, and Dronecode Foundation, its members, and project maintainers have mastered their design for the open source market, as well as the software to make use of them.

Dronecode Foundation projects

If you’ve ever looked into building your own drone, then you’ve come across PX4 Autopilot and Pixhawk. PX4 Autopilot is software for flight controllers, which performs the flight tasks described above — and much more. Pixhawk provides open standards for flight controller hardware, which can run PX4 Autopilot software. Both Pixhawk and PX4 Autopilot are Dronecode Foundation projects.

Other notable Dronecode Foundation projects include QGroundControl and MAVLink/MAVSDK.

As the name suggests, QGroundControl is ground control station software. It can run on Windows, Linux, macOS, iOS, and Android. It has all kinds of great features, including map-based mission planning with a 3D viewer and management of multiple vehicles.

QGroundControl communicates with Pixhawk flight controllers running PX4 Autopilot via MAVLink, which is a lightweight messaging suite. MAVLink offers several specific protocols to suit different needs. It works for everything from simple flight commands to video streaming.

Recent development

Drone Foundation’s 2024 Annual Report touches on recent advancements made by those members and there are a lot of impressive new features to talk about.

PX4 Autopilot

A great deal of recent development for PX4 Autopilot has gone into improving ROS 2 integration and performance. The original ROS (Robot Operating System) may be more approachable, but ROS 2 has much greater potential for customization and complex expansions. In particular, ROS 2 emphasizes clean communication between software and hardware modules, and PX4 Autopilot takes advantage of that to facilitate the construction of drones with robotic capabilities beyond simple flight.

PX4 Autopilot now has enhanced support for Gazebo simulations. That is important for both testing and giving drones the ability to better solve spatial problems.

Additional improvements include refined navigation algorithms, a new slow flight mode for precision movements, and even an experimental “throw mode.” That gives pilots the ability to launch a drone by throwing it into the air — not just a videogenic gimmick, but a useful strategy when a suitable take-off area isn’t available.

Pixhawk

The Pixhawk team has also been hard at work with the development of the new Pixhawk FMUv6X-RT open standard. That is the first Pixhawk flight controller to contain an NXP-based microcontroller (the NXP i.MX RT1176). That MCU increases the real-time computing power of the flight controller, compared to the STM32 microcontrollers used in previous generations.

The Pixhawk FMUv6X-RT has a host of sensors, including redundant IMUs (Inertial Measurement Units), a compass, pressure sensors, dual GPS units, and a high-accuracy barometric pressure sensor. Holybro, an official Pixhawk supplier, sells the Pixhawk 6X-RT based on the new standard. But, of course, this is an open standard and other suppliers can sell their own, too.

QGroundControl

QGroundControl Version 4.4, which is out now, made some great quality-of-life improvements. The two most exciting are gimbal control, via MAVLink’s Gimbal Protocol (v2), and terrain following.

The gimbal control feature works with both PX4 Autopilot and ArduPilot systems, giving pilots better gimbal management during flight. That’s great for those who use their drones for cinematography and for pilots who use FPV (First-Person View) goggles.

Terrain following takes advantage of Auterion’s new server, which provides detailed data on local geography. That can be a tremendous boon for navigation, as it lets drones fly at low altitudes safely.

MAVSDK and MAVLink

MAVSDK’s contributors focused on incremental improvement over the last year, including redesigns to the Camera Plugin and Gimbal Plugin. They also enhanced connection and data communication handling, for better reliability, throughput, and configurability.

One notable new introduction is the Events Plugin, which may better suit some control strategies. Event-based functions make more sense in many situations, compared to other models like publisher-subscriber communication, so flexibility of this new plugin to handle events is a nice addition.

The focus of MAVLink contributors has largely been on standardization and consolidation of repositories, to streamline development and deployment. Those can easily become overwhelming, so these efforts are very worthwhile.

But there are new features and improvements, too. Fuel/battery status information, camera thermal detection, and wind estimations are notable examples.

The future of open source drones

The most exciting insights we can glean from the Dronecode Foundation’s 2024 Annual Report are hints about future development and what we can expect to see in the open source drone scene as it continues to mature.

The report authors are, understandably, careful about making any promises. But we can suss out some clear trends about the direction of upcoming development.

It seems that we’ll see a great deal of development in three key areas: enhancement of computer vision and AI, expansion of autonomous navigation capabilities, and more sophisticated swarm management.

Each of those, in its own way, has the potential to dramatically expand on the utility of drones — particularly for open source builders.

Drone shows, for example, require very careful swarm management and a great deal of current implementations rely on propriety systems. The community would certainly welcome open source alternatives. But swarm management is important for more than just entertainment. Combine that with more robust autonomous navigation and you have the perfect technology for search-and-rescue efforts, as well as inspection of infrastructure and assets.

Similarly, artificial intelligence could be a kind of “magic bullet” if integrated with care. Those search-and-rescue efforts, for instance, would benefit from artificial intelligence that has the ability to recognize promising locations and identify victims who need attention.

A well-implemented artificial intelligence model could also improve a drone’s flight capabilities by letting it predict, rather than react to, conditions. It could even pick out and capture the most breathtaking imagery of a location.

If you're in the area, Dronecode Foundation has community meetup in Philadelphia coming up on March 26th, 2025. Other events will come up in the future, so be sure to keep an eye out.

Conclusion

Dronecode Foundation just celebrated its 10th anniversary and the 2024 Annual Report proves how much it and its members accomplished in that time.

If you’re a drone enthusiast, you have Dronecode Foundation, its members, and especially those member’s contributors to thank for all the great open source technology we have today. They’ve achieved a lot in the last year and many exciting new features are just around the corner.

To see all the details, be sure to read Dronecode Foundation’s 2024 Annual Report.