Controlling a Quad copter

[Curveball Anders]

Ok, I managed to get a Quad to land and work (-ish) at Duna.

But it was 100% trail and error (mainly error).

Getting the blade angle was by adding the angle to a control, adding the aero overlay and testing.

I think that the idea is to control the lift by assigning the RPM limiters to a control, which works, but a complete nightmare.

So Ye Gurus how know this stuff,  how is it done properly?

 

[AHHans]

Well, for the time being I've given up trying to get a propeller powered vehicle to work on Duna. (But maybe I should try again with the ducted fans, allegedly they work better than the normal props.)

The quadcopter I made for Kerbin (current version here, with most of the instructions here) is controlled in the way that the RPM of the rotors is kept constant and the lift is controlled by collective pitch angle of the blades. Because in a quadcopter the blades are always in the same general direction from the COM of the craft (i.e. the blades of the front left rotor are always to the front and left of the COM during the full rotation of the fan, which is in contrast to the blades of a classical helicopter), the blades can be used as control surfaces without the logic of KSP freaking out. So they can be used for pitch and yaw control (control direction is up). The difference between the two designs is how collective and cyclic (well, control) pitch is implemented: the old version needed extra servos, but since 1.8 the deflection angle and the authority limiter are separate on the blades so no extra servos are needed anymore. Roll control is done by changing the RPM of the clockwise and counterclockwise rotation fans via a KAL.

[Fierce Wolf]

Im just gonna say: I wish SQUAD would test a quadcopter (or rather, a squadcopter, amairite?) in Duna so they can get ALL the bugs with rotors.

[Curveball Anders]

I've ignored the cyclic totally and cheats by using a reaction wheel.

I've got the rotors and blades counter rotating in diagonal pairs so the torque cancels out.

I tried using pitch to control lift but it becomes very hard to control with any precision, it's easier but laggier by using the rpm limiter.

My main problem is the control the horizontal motion but it might be that I'm just stupid and not reading the visual ques correctly.

So I can lift off, I can fly in any direction at quite a good lick but it's very hard to hover, and thereby very hard to land without falling over

 

[AHHans]

11 minutes ago, Curveball Anders said:

I've ignored the cyclic totally and cheats by using a reaction wheel.

Well, my design was also kind of a self-imposed challenge to design ad quadcopter without reaction wheels. After I thought that this should be possible.

14 minutes ago, Curveball Anders said:

I tried using pitch to control lift but it becomes very hard to control with any precision, it's easier but laggier by using the rpm limiter.

With the first version - that used extra servos for collective pitch - I could increase precision by reducing the range of the servos. I also always keep a PAW of the blades open so that I can see the actual value. I guess I just prefer the fast and fairly direct control over the climb-/sink-rate over the less finicky control.

21 minutes ago, Curveball Anders said:

My main problem is the control the horizontal motion but it might be that I'm just stupid and not reading the visual ques correctly.

With the control direction being straight up, I can set SAS to "surface - radial-out" which will keep the quadcopter steady, then I can use pitch and yaw control inputs to nullify my horizontal motion (center the pro- or retrograde marker on the navball). And once it is zeroed out, it stays zeroed out.  [Looking at your screenshot] Well, I added a probe-core that supplies "radial-hold" in SAS. You might want to give it a try yourself.

One drawback of my design is that the control direction has to be straight up! If you want to change that then you would need to change the setting on the blades (for which direction they act as control surfaces) and the action group that is currently bound to roll control. That makes long-distance navigation rather cumbersome...

[IkranMakto]

Well, I think it's time to post my quad copters. I'll do it later on Spacecraft exchange. 

Best way to do it is using one KAL-1000 for each axis. X, Y, Z и Lift. Respectively bind to pitch, roll, yaw and throttle axis in action groups. 

Forvard left rotor - clockwise, forward right - counter clockwise. 

Rear left - counter clockwise, rear right - clockwise. 

Motor speed - constant, as your craft needs. 

KAL-1000 should change rotor's deploy angle in inverse:

For pitch - rear and forward (from - 10 to +20 degrees for examples). 

For roll - left and right.

For yaw - rotors which spin clockwise and counter clockwise. 

As result blades will have mean angle between all 4 axes. 

[IkranMakto]

Done. It's quad copter, but complicated :p Of course no reaction wheels
https://kerbalx.com/IkranMakto/Dragon-Gunship

[BigFatStupidHead]

We had quite a good discussion a while back about this over HERE. This was before shrouds came out, so they may make a world of difference.