ArduPilot cannot be used to fly aircraft carrying human passengers. It is not certified software for such a purpose. The following test information from was from a flight test conducted under strict safety mitigation requirements for data gathering purposes and obtaining a grant of waiver to develop, test and fly an unmanned helicopter over 55 lbs gross takeoff weight in the U.S. The aviation authority is the FAA Flight Standards District Office, Minneapolis, MN. The grant of waiver was denied based on the test data.
We did successfully hover a Mosquito XET helicopter with a modified version of ArduPilot. But I was flying the helicopter (in the seat) while my assistant tried to hover it manually via RC control. We used S-tec autopilot servos with clutches that I could over-ride with the cyclic. We only had cyclic control hooked up to the RC, I controlled collective, torque and engine throttle from the seat.
That being said, some warnings:
- The Mosquito has a teetering rotor system similar to a 206B.
- We only used acro, did not let the autopilot actually fly it with any sort of stabilization.
- I had to make significant corrections from the seat to hold it in an acceptable hover.
- We did it primarily to get some logging information on what would happen if the autopilot tried to fly it.
- My conclusion was, that while my assistant (who is an expert RC pilot) could hold it in a pseudo hover using visual reference, the rotor system was too disconnected from frame attitude for the autopilot to work. It would not be able to hover it.
- In hover, the logs showed the frame attitude (that the autopilot measures) having significant delay from actual control inputs - the rotor system does all the work and the frame stays pretty much static.
- If we would’ve tried to use a mode like Stabilize, I concluded the autopilot would start “chasing it” and result in a dynamic rollover because the autopilot does know what the rotor system is doing, as it “sees” no response in frame attitude from a control input, so it would tend to over-control.
In flight, where the rotor system is a little less disconnected I think it would be able fly it and perform basic roll leveling or stabilization. But not in the pitch attitude. However, if it can’t pick it up into hover, it’s a non-starter to get it into forward flight. The control inputs to hover a teetering or hinged rotor system are very minute - your arm is resting on your leg and the movements on the cyclic are usually no more than 5 or 6 mm in any direction.
The Mosquito is quite a bit heavier and larger than your model, obviously. But if you can, I would suggest using a rigid rotor system, then it will work fine if the input shaping is adjusted properly for the response of a larger size helicopter like you are building.