UAV and Bank Angle Detection

[Gary Warner]

Question.

If I have a plane in a coordinated flight condition using the slip/skid control project and the AOA project (to maintain a constant AOA in a bank), can I just read the static (apparent) vertical G forces to compute the angle of bank? I know it won't say which direction (left or right) but won't I be able to read the angle with a bit of math?

I found this (see picture). But I don't know how to get from "Cos(b) = .263" to "B = 74.7 degrees". How do I get the solution for B?

Gary
--

[MX]

...I don't know how to get from "Cos(b) = .263" to "B = 74.7 degrees". How do I get the solution for B?...

Take the arccosine of 0.263. On a calculator it might be labeled invcos or cos-1. (-1 is superscript)

MX

[Majortomski]

Yep, you could detect it with that, but what are you going to do with the data? If it's feeding your autopilot, what is going to happen when the banked aircraft hits turbulence? Would this initiate a false bank command?

[LukeZ]

Maybe I'm missing something in the formula above, but isn't the more interesting number LF? How do you compute that?

[Gary Warner]

Maybe I'm missing something in the formula above, but isn't the more interesting number LF? How do you compute that?

LF is Load Force (in G's). Easy to get from the accelerometer.

Gary
--

[Gary Warner]

Take the arccosine of 0.263. On a calculator it might be labeled invcos or cos-1. (-1 is superscript)

MX

Thanks. That's what I was looing for.

Now I can make a look-up table for rough bank angle readings:

LF - Bank Angle

1.0000 = 0
1.0038 = 5
1.0154 = 10
1.0352 = 15
1.0641 = 20
1.1033 = 25
1.1547 = 30
1.2207 = 35
1.3054 = 40
1.4142 = 45
1.5557 = 50
1.7434 = 55
2.0000 = 60
2.3662 = 65
2.9238 = 70
3.8637 = 75
5.7587 = 80
11.473 = 85

[Gary Warner]

Yep, you could detect it with that, but what are you going to do with the data? If it's feeding your autopilot, what is going to happen when the banked aircraft hits turbulence? Would this initiate a false bank command?

Yes, no. The readings will be averaged, with weighting. In other words, the newest data would be factored by 10. the second newest data would be factored by 9, etc down to the oldest data factored by 1. This is a linear example. Other factoring curves might be used. The data would also be factored in from a GPS (rate of heading change data).

Gary
--

[Gary Warner]

I know my question sounded funny for a UAV. But today is slow and it's a 'thinking' day.

It's the third step in my getting a stable airplane for thermal flying (sailplanes).

First was the AOA feedback system. Second was the rudder slip/skid project. Thirdly is controlling bank angle.

Mode of operation:

Establish best sink alpha with the AOA. Maintain that best alpha at all times. Elevator stick ‘selects’ the desired AOA.

Keep turns coordinated with slip/skid correction. Rudder stick ‘selects’ skid/slid desired (or centered for no slip/skid)

Input the desired bank angle with the transmitter. Have the plane hold that angle indefinitely. Aileron stick ‘selects’ the bank angle.

When performing a turn, a change would be needed in 'flying style' with the aileron input. I'd 'hold' the stick over throughout the turn, since the plane will hold that bank angle. I will not re-center the aileron stick until I wanted to stop the bank. Kind of like turning a car.

Some day, the final step will be to add GPS for navigation. That would be the final leap to autonomous flight.

Gary
--

[Arp]

Interesting projects. :)

My loosely related sidenote:

I am currently installing a *somewhat* similar system (for performing maneuvers while maintaining a fixed AOA or bank angle) on my gliders. My system, however, does not employ accelerometers.

Instead it functions by tilting an FMA stabilizer in relation to the plane.

[sesat]

LF is Load Force (in G's). Easy to get from the accelerometer.

I disagree. To be precise, I still disagree. Accelerometers measures the result of net force acting on the sensorts. You can't get LF from accelerometers without making assumptions about either bank angle or body acceleration.

edit: the coordinated turn is a special case where we can make assumptions about both bank angle and body accelerations, ie that they will remain constant throughout the turn. This is why the slip eliminator works, and it only works because the rudder correction is in the same direction for both deviation in bank angle and sideslip.

Ram.

[Gary Warner]

I disagree. To be precise, I still disagree. Acelerometers measures the result of net force acting on the sensorts. You can't get LF from accelerometers without making assumptions about either bank angle or body acceleration.


Ram.

:)

Yes, you made that clear in the slip/skid project. In this example, I only want to read what ever the apparent G forces are, in relation to the vertical axis of the plane. I don't care how they get there, I only want to read them and use the data. The data will be 100% useful in coordinated flight, regardless of how the sensor derives it's given output. That's why I've addressed the coordinated flight issue early on.

What might be confusing is I'm taking a somewhat different tact to flying autonomously. Old school autonomous control has proven to be filled with at least as many short comings as my approach, requiring much time, code and technique to over come the problems. I aim to minimize the headaches involved in autonomous flight. I don't want to cut butter with a chainsaw.

Gary
--

[sesat]

I agree the data will be 100% useful in coordinated flight. How do you intend to enforce coordinated flight, if the control routines don't deal with uncoordinated flight?

I'm intrigued by your approach that is clearly different from mine.

Ram.

[Gary Warner]

I agree the data will be 100% useful in coordinated flight. How do you intend to enforce coordinated flight, if the control routines don't deal with uncoordinated flight?

I'm intrigued by your approach that is clearly different from mine.

Ram.

Maintaining AOA and zero slip/skid.

Gary
--

[Gary Warner]

Also, about my approach...

Notice how I deal with all three axis of flight without any of the control systems talking to each other? A UAV must be able to maintain stability on all three axis. I'm addressing the stabilty of each axis individualy. This approach keeps the stability separate from the navigation too. They would only come together in a controlled navigation system (GPS).

Kind of like... well, if a real plane is stable, you don't need a pilot. You only need a pilot to decide how, where and when to 'destabilize' the airplane. Once the pilot finishes 'destabilizing the plane, the plane tends to stabilize its self.

So the way I see it, I keep the two functions (stability and navigation) as far apart as possible. They would only talk to each other as a function of navigation, not a function of ‘aerodynamic’ control. I know that sounds like bull crud, but that's the tangent I'm taking with my approach.

Gary
--

[Gary Warner]

Interesting projects. :)

My loosely related sidenote:

I am currently installing a *somewhat* similar system (for performing maneuvers while maintaining a fixed AOA or bank angle) on one of my gliders. My system, however, does not employ accelerometers.

Instead it functions by tilting an FMA IR stabilizer in relation to the plane (the stabilizer is mounted on top of servo-tiltable sensor mast).

Currently, it only controls pitch. After enough testing, I also hope to include roll. I hope it helps me safely obtain more experience in flying this (otherwise rather difficult) glider.

That's amazing. Does it work well? Oscillations?

[Comatose]

Roll is a little bit tricky, but I've flown a model in full feedback control mode with angular inputs from the TX and an accelerometer as the only feedback. It works really nicely until a big gust of wind comes up....

I had pre and post filtering going. Worked okayish, works a lot better with a gyro in addition though. I certainly wouldn't have trusted it by itself without me flying (and able to turn it off remotely)

[Gary Warner]

Roll is a little bit tricky, but I've flown a model in full feedback control mode with angular inputs from the TX and an accelerometer as the only feedback. It works really nicely until a big gust of wind comes up....

I had pre and post filtering going. Worked okayish, works a lot better with a gyro in addition though. I certainly wouldn't have trusted it by itself without me flying (and able to turn it off remotely)

I have to agree. I could see the using the gyro to detect and stabilize the roll rate while the accelerometer would set the course roll angle.

Mmmmm... got me thinking again.

Gary
--

[Arp]

Does it work well? Oscillations?
Your suspicion is correct. When I initially chose a too weak servo, I noticed already indoors that the sensor mast would start wobbling in wind.

Thus I installed a stronger servo. Since there hasn't been proper weather yet (snow just melted, but unreasonably cold weather still tends to freeze fingers)... I really cannot tell how it behaves in flight.

But when I've tried provoking it to behave unreasonably:
-- by moving all other servos to try creating noise in the servo system
-- by doing that while the motor is running
-- by doing it all while the Tx is without antenna (naturally also no propeller on the motor)
-- by testing how it responds to IR warmth

...I have not noticed it suffering interference, nor undertaking anything apparently self-destructive (quite oppositely, it seems to properly perform its stabilizing task). So when suitable weather appears during a weekend, I will go testing it.