HI, first post here 
In the near future, I will be working on a system that will integrate 3 pixhawks on a single drone. All the data would be compared and voted upon which to send out. I believe the term was fly-by-wire? So, just wondering if anyone has seen anything similar, or maybe even using two pixhawks, so i can have some sort of a base line. Thanks.
yeah… https://www.google.com/search?q=ardupilot+redundancy+site:diydrones.com
mostly https://diydrones.com/profiles/blogs/705844:BlogPost:2188566
IMHO redundancy is the key, but not separate boards, 3 or 5 sensors and a good software to manage data.
Well, you put three autopilot into the system, then your single point of failure will be the management and switching circuity. Overall, your redundancy will not be better than with one flight controller without the switching… edge540T is right, redundancy already built in with three sets of sensors a good EKF algorithm to fuse them.
3/2 voting logic is much more simple and reliable than an embedded computer with a large software stack, so triple redundancy can increase your reliability. I would use dissimilar hardware and software - one apm, one paparazzi running on differen boards - you just have to ensure bumpless transfer between FCCs - possibly by using the same control architecture and gains. EMC will be the issue to work on…
Voting on what ? Motor PWM ? Even with identical FC’s there will be differences of actual motor PWM, so you need a quite window which considered OK. If you have redundant sensor sets per FC (Which is the actual case with the Cube) then multiple FC’s will provide redundancy against catastrophic FC failure, but then you have a single point of failure with more like the same MTBF, the switching logic.
And from the mathematical point, if you take out the MTBF of the switching logic, unless you put three totally independent FC’s with own power supply and all, then due the interdependence your total MTBF won’t be significantly better than with only one FC.
As always, I trow these numbers when the concept of redundancy arise:
Pilot fault 70%
Power failure 20 %
Mechanical failure 9%
Do the maths 
1 Like
To have real flight controller redundancy is a very complex task. i do not think that the efford is worth for.
But, since some authorities want to see a redundant system, we developed a simple backup redundancy system with a pixhawk running arducopter and a small F3, in our case with Inav (now with Chibios on F4 it would be worth thinking about an F4 with AC as backup).
To keep things simple and save, the switch mechanism is pure machanical (relais). At the moment, a RasPi listens to the Mavlink telemetry of both FCs. In case of no more mavlink pakets from main FC it switches to the secondary FC. Also the switch has a normal PWM jnput for switching.
At the moment I’m working on a new version with a Teensy instead of the Pi, and a smaller board. Feel free to ask for more details if you are interested. Board layout und software will be open.
You can see a video of a switch flight with some wind on youtube: https://www.youtube.com/watch?v=sV9anBD0_Og
1 Like
Of course it only makes sense if you think about few hours of flight time and UAS worth 100K or more, but there are turnkey solutions to do that: https://www.micropilot.com/products-mp21283x.htm
Also worth looking at the way how this is done properly: https://ieeexplore.ieee.org/document/495891/
1 Like
some replies already, cool 
So yeah, I’m just trying to get more information for when I start this. Thanks guys.
1 Like