Servers by jDrones

Can a PixHawk work in a real airplane?


Greetings everyone. Noob alert!! I’m new to this forum and VERY new to this subject matter so please forgive my lack of knowledge. Feels right now like I am drinking from a fire hose and getting older doesn’t seem to help either ;-). My expertise with R/C aircraft ended shortly after college (30+ years ago!) but my love and fascination with airplanes and flying never has. I’m a tinkerer at heart and over the years have built and flown 2 experimental airplanes. I love the innovation spirit that is present within the Experimental Aircraft community which is also very obviously present within your group. Very cool and interesting stuff!! Unfortunately some of it is way above my head right now.

I’ll try to briefly explain what I’m trying to accomplish. I’ve attached 2 servos to small trim tabs along the trailing edges of one aileron and the elevator. The tabs have enough authority to control pitch and roll but I have plenty of authority with the normal flight controls to override any undesirable inputs( plus I can kill power to the servos as a safety feature as well). I’ve invested some time writing my own software (self taught programmer) to drive the servos with good results in the pitch axis, not so much on the roll axis. So I’m trying to go about this in a different direction now.

Q1: Can I hook up a PixHawk to make it work? Will it work under the vibrations of a real airframe?

Q2: I don’t understand the system well enough but I would like to not have to have a transmitter involved. How else could I command the autopilot pitch/roll in real time?

Q3: I would like more functionality that a wing leveler. Ideally I would like to be able to tell the PixHawk to fly a heading or a course to a waypoint (lat/lon) plus give it commands to climb descent to a given altitude. Can that be done real time via a serial port or only through the Mission Planner? Any other options?

FYI: I have a Lenovo ThinkCentre M73 running Win7 installed in the airplane running and EFIS program that I’ve cobbled up over the years.

I hope this topic isn’t out of bounds in your forum but if it is I understand. TIA for your help/inputs.



1 Like
(tridge) #2

While it may be technically possible for ArduPilot to be used with manned aircraft we strongly advise and request you not to do it. ArduPilot is not designed with the sorts of hardware and software considerations that are suitable for the risks associated with manned aircraft.
Please stick to using ArduPilot for unmanned model aircraft. For manned aircraft we can’t help, sorry.
Cheers, Tridge


Tridge, while I understand your reluctance in using ArduPilot in manned aircraft you are missing the point that it is for Experimental use only. It doesn’t mean or imply that it will be used in a certificated aircraft or in actual instrument conditions. It is for educational purposes only. Building a better mouse trap does not happen unless you think outside the box, wouldn’t you agree? . I would argue that it’s safer having the pilot in situ as opposed to on the ground relying on a telemetry link. The risk of a crash is always present with a UAV, not so much with a manned aircraft. I hope you reconsider your position.


(Ric) #4

How could you expect any other answer?
You are in a hobby forum, asking advise on using hobby/DIY level equipt in a full-sized aircraft with humans aboard!

(tridge) #5

@CharliEcho, it being experimental doesn’t really change things from our perspective. I don’t ever want to hear that someone was badly injured or killed in an aircraft flown with ArduPilot, no matter the certification state. We can’t stop you from doing it, but we have decided within the dev team that we won’t do anything to support or encourage it.


It’s a shame you all feel that way. Apparently you know little of eaa and what it’s all about. Most of us started out as hobbyists and took our passions to the next level. I will pursue my ideas elsewhere, sorry to have troubled you.


(Todd Peterson) #7

It could be done. But I would have a safety mechanism in place to disengage and set the controls to a safe position if there are any problems and have a mechanical backup to override the computerized system. Probably should be triple redundant for any flight critical operation. Hairly stuff.

(Todd Peterson) #8

I have submitted an entry to the EAA Founders Innovation Prize contest, using a drone board as my basis. It is a supplemental alert system. Based upon the SPRacing F3 Mini board and an auxiliary system for sensing airspeed ambient temp/humdity and AOA. It runs CleanFlight.


That sounds really cool Todd. Is this a concept to be presented or an actual working prototype on a real aircraft?
I wish you well in the contest.

(Todd Peterson) #10



I have a partially complete prototype. I hope to complete it by EAA AirVenture. The top 5 get to be on stage for presentation of the prizes. I hope to fly it on an aircraft by then to demonstrate the functionality.

(davefolts) #11

Technically possible Carlos…but will you get enough deflection of controls with just the tabs? I’m an EAAer and droner who is in flight training…and Friday I used an autopilot for the first time. Nice!
Isolating the Pixhawk from vibration is important, and an interface with an fms is needed I’d think to drive manned aircraft servos which are $750 a piece right! You will also need to spend a lot of time testing your parameters so you get proper control deflection and no pitch or roll inputs that exceeds your aircrafts limits at Va. What you should do is set up the HIL sim in mission planner and test the pixhawk setup on your plane or a model. Btw…you need to find a way to calibrate the pixhawk gyros and accelerometers while its in the plane…with uav installs we just roll the whole system around. The pixhawk is performing the AHRS function in your scenario…so calibration is ultra critical!
Another consideration is electronic noise from the alternator and magnetic interference from the aircraft structure. If you had a wood plane, this tough problem could be ameliorated. Tons of testing and experimenting.

(marcmerlin) #12

@CharliEcho: I know EAA, I’m actually going to Oshkosh/Airventure for the 5th time this year, but I also know Ardupilot and Tridge and some of the rest of the team.
Long story short, pixhawk and ardupilot can probably do an ok job to pilot a full size aircraft in most cases, but it’s just not designed to be used for anything life critical, and if you hit a nasty bug like this, instead of having a pilot of foam crashed upside down in a field, you potentially end up with someone dead:

Yes, I understand you say you’ll have enough control authority to override the auto pilot, but basically the ardupilot team does not know this and does not want to be involved if anything goes wrong.

So this is a long way to say again, that indeed you can do this at your own risk, but understand that when you build an experimental aircraft you can verify that all the parts will stay together and not fly off, but with a very complex autopilot, you just cannot, so you’re trusting a lot of other people to have written code that is hopefully not going to kill you.
This is why they don’t want that responsibility.
Tridge is ok feeling just a bit bad if my plane crashes twice due to a software problem :slight_smile: but not if someone get injured or killed because the AP went wrong, and then something else went wrong too, leading to a bad end result.

Hope this makes sense. If you proceed anyway, make sure you indeed have a very failsafe low level disconnect to take over if anything goes wrong.

(garlandk) #13

Look up the Snowbird UAV Project. :slight_smile:

(Ric) #14

@marcmerlin sums it up well: you are dealing with an open-source package that should never be used for any "life critical" application.

(Martin Rüedi) #15

I found that this rather old thread matches my question, so I append the question here:

What if ArduPilot and a modified GCS is used as a so called ultralight EFIS?

There are two good reasons, why the hesitation might not be warranted:

  1. An EFIS just presents flight data. It is only an instrument. It is not active and acts not as an autopilot.

  2. This aircraft class has no requirements for instruments at all and whatever is used does not need to be certified or meet any kind of airworthiness standard (in the US FAR part 103).

I develop a GCS (FlightZoomer) and for the second time I was asked for an EFIS version of my GCS. When I see some of the existing products, that target this market, I think that ArduPilot could compete well:

Often people are using iOS or Android tablets with app-only solutions:

Or there is hardware involved to get better raw sensor data: with this app

So I think a Pixhawk with an airspeed sensor and ArduPilot as a flight data provisioning tool would match or even beat the precision of the best competing products in this area.

How does the community rate this idea?