Sorry to dig this up, I understand OA is no longer under development but wanted to collate my findings for anyone else interested on this topic from the last few months of research I've done on it. There are 8-9 different posts raising the same question but for different reasons and I have yet to find a means of implementing this in OA or A3.
I can not stress enough how powerful this simple command would be for the community, in particular for flight; a setAction command wouldn't just enable vehicle failures to be simulated, it'd allow the community to produce complex scripted aerobatic maneuvers, smooth banking turns(for UAV, Chopper gunner, AC130), loops, attack profiles, evasive maneuvers; the possibilities are just about endless.. hell why not an entirely new scripted flight model for AI with dynamic decision making/planning.. I've already produced a self-sustaining flight controller and PID for control surfaces, there are just no means of properly implementing it that I could find.
setPitchBank really doesn't cut it for aircraft control mainly because it doesn't utilize the physx engine/flight dynamics even with a relative velocity controller; the resultant effect is jerky and unrealistic as it simply sets the vehicle to a new vector on each loop which also conflicts with the stock AI controller.
The functionality was present in VBS2 as a setAction command allowing for direct control of user/ai input (or rather 'output'). e.g. _this setAction ["rudder", (-1 to 1)]...
..and was also present in ToH as setActualCollectiveRTD, so it's clearly possible within the engine, and as far as I can see, wouldn't be unrealistically difficult for developers to implement in A2 or A3 - probably just a matter of exposing the command.
In my desperate attempt to gain any kind of control over direct flight inputs via scripting, and at great length, I manually scripted the rudder/aileron/elevator animation phases, however in the end this didn't influence the aircraft's movement.
Other things I tried but to no avail included:
â€¢ setVariable on action strings
â€¢ _this action "rudder", "forwards", (non-A2 player actions but set inputActions)
â€¢ control surface control with animation control source "user"
â€¢ scripted waypoints (only a 2D coordinate so no bank pitch control) - temperamental at best
â€¢ AI look-ahead-distances
â€¢ actionKeys and DIK codes
The only thing left I can think of (short of developer intervention) is some kind of string manipulation; BIS_fnc_Help might be of use in finding string functions if anyone's interested.
There are also many posts I found about controller input related issues that could easily be addressed with direct input control, but instead require 3rd party joystick emulation programs such as PPJoy and GlovePIE to resolve.
These only serve to control player input, not AI, and are fairly limited in this respect.
If anyone has found anything or can shed some light on this I'd be incredibly grateful.
If by some miracle a Dev notices this and were to deem it a worth while investment for A2 or A3, It'd make my freakin' year, and I know many others too!
ARMA is an otherwise highly diverse game with incredible potential for scripting and community made mods/input.
I honestly feel that implementing the above would add a great deal of control for mod makers and the community in all mission scenarios.
EDIT: I have posted this as an issue/request on ARMA 3 Community Issue Tracker, please +1 if you agree!