An AutoPilot in the Clouds - part 1


In our RC world, autopilots are typically some kind of software/hardware combination that is physically located on the craft flying/driving/sailing etc.. This has some disadvantages since autopilots are expensive and singular, meaning that they only know them selves and have no interaction capabilities with other autopilots.

As an example, a friend once asked me what it would take to make a drone that could analyse a large field of crops taking samples of soil were the vegetation was less eager to grow. My reply was “Two drones!” - one fixed wing to fly over the field, doing a flir/nir analysis - and one copter type to fly out, land, take samples and return them.

The operation had to be coordinated somehow - so the copter would know were to land etc…

In short - some problems are not solved using one single type of drone, but rather a combination of drones with different capabilities.

In the dungeons of my cellar, pilled with winter clothes, umbrellas, broken china and old lamps, I’ve been working on a universal autopilot system, that has a central computer system that is capable of guiding several drones and coordinate their activities. The main software can be executed on a very small computer ie. a raspberry pi or similar, or a laptop. The drones run a modified version of multiwii 2.2., that sends very compact telemetry data several times each second as well as being able to receive commands from the autopilot on the ground.

The idea is to keep the autopilot on the ground (or on a mother drone), and keep track, issuing flight directions using short and efficient telemetry data between drone and main computer. Several crafts can be controlled by the software at once, making multi drone missions possible and coordinating that drones do not collide.

Each drone uses a XRF ( radio module to send back telemetry as well as receive orders on direction, speed, altitude etc… The XRF modules are great, very easy to use and setup and have a range up to 1.8 miles. Most serial transceivers are however usable.  


Each drone uses a multiwii board for stabilisation and serial communication, and a gps module so the central computer knows were the drones are at all times.


The autopilot software (running on the central computer) is controlled by a state machine, meaning that a system is making sure that everything happens in the correct order. For instance, that auto land is only possible after takeoff. Or that takeoff is not possible if the preflight sequence was unsuccessful or GPS lock has not been achieved. The state machine also makes sure that transitions between states are carried out correctly - ie. going from autopilot to manual flight makes the state machine automatically pause the autopilot function and set the craft on a straight path before control is set to manual.

The system is easy to extend for different craft types as well as different hardware.

The guidance part is based on the USAV project described some decades ago (so it seems) here.

Currently I have a simple system working, with one fixed wing, called the UTV-3, based on the Z84 from Zeta Science.


I’ve implemented some simple maneuvers as auto takeoff, auto land and simple waypoint navigation. Simple stuff that I use to test the system with. Once everything is stable, I’ll be implementing the multi drone hub, needed to coordinate the works of several drones. Although I work a lot in my dungeon, I’m also lucky to live close to a field for RC enthusiasts were practical testing can be done. Videos to come.

E-mail me when people leave their comments –

You need to be a member of diydrones to add comments!

Join diydrones


  • Currently using 2x18650 LG 3400 mAh's. They give me around 40 minutes when flying economically

  • What kind of flight time did you get?

  • Just noticed this. I think this is an important step for real time networking. It doesn't matter if the "brain" is on the ground or not. The real technology is synchronization. Putting it into perspective, the Apollo spacecraft was navigated by mission control, yes that's right on the ground! Now imagine doing that with a cluster of drones. Reminds me of the movie "Battle of LA". Jesper's a clever guy keep going buddy.

  • Thanks UAS_Pilot! I certainly will :-)

  • I to think this is a very interesting project... Good for you on trying to create something original.
    These aircraft would be the true definition of a "dron"... A dumb aircraft with a ground computer controlling it.

    Keep us informed on your progress.
  • Thank you Hugues :-) I know there is a tendency here to favour the APM and the likes, for obvious reasons. But we should not forget the DIY-part in DIYDrones and the fun in experimenting and turning concepts upside down.  

  • MR60

    Great project. Nice to see innovative ways. I don't get the negative criticism of previous posters focusing narrow minded on the cost of a fc...

    I see useful applications of this, such as the upcoming swarm "intelligence" where every individual in the swarm is as basic as can be, running simple rules. Inspired by bees, ants, termites,...

  • @Bojan: If the drone looses the contact with ground control - the failsafe functionality of the multiwii takes over and returns the drone to were it started.

    Let's say you own a big field and you use a system like this in order to do frequent analysis of crops, take soil samples etc... Likelyhood of someone jamming you = 0%. Accidents? You could just as easily experience mechanical failure (which I think is far more likely) and the result would be the same, some injured corn or a scared cow :-)

    @Tobias: Or simply use more radios :-) The current system can handle 4-5 drones - enough for me.

  • Besides safety, scalability is another issue here: If you add more drones, it is likely that your radio link gets saturated; and you would need to care about real-time capable medium access strategies.

  • You are missing the point. While experimenting with autopilot software it's nice to have the big brains safely on the ground. I've personally lost two planes (and all hardware) while developing autopilot software.

    You say that autopilots don't require a lot of processor power but to me - just flying on a predetermined waypoint mission, is pointless. Wouldn't realtime analysis, adaptive mission planing, image stitching etc., be a great thing to add? That do require lot's of processing power - and that kind of power is still expensive.

    But economy is a minor point. The purpose with this is to present a different approach to autopilots - one that allows drones to interact and solve problems together. This is not my invention - you've hopefully seen the videos of drones building structures by placing blocks on top of each other. This requires a central governing system and this is my take on how to build a system that allows these kind of things. 

    You put way too much into a very small part of this post (three words I believe).

    Putting drones to good use is what really proves their worth, and some uses require drones to interact.

This reply was deleted.