I am flying Control Line models.
(link for example on YT: Control Line Handle Talk First Maneuvers - YouTube)
And now I wanted to develope controller for eletric version :).
Maybe a little explanation, what is the problem.
When model is flying on level, there should be for ex. 60% of power.
When I climb, I want to have 100%, and when diving, power should be reduced for ex. to 50%.
Of course, there are ready to fly controllers, but cost about $300, so I wanted to try to make my own ;). I know, that it will be a tough problem to solve, but maybe we could do it together. I plan to make it open source -> it is for hobby!
I read a little about it, bought MPU6050, read accelerations, but it was 5min to do it . But what now?
From everything what I read until now, it looks like a 1 year project for me... Programming is not a problem, but what to do next in general .
NOT to start with Euler angles, becouse there will be problem with flying straight up.
There will be a lot of dynamic moves (so a lot of acclererations) - is there a way, to get rid of them, and stay only with gravity vector?
Is Atmega328 and MPU6050 enough for a such task? Models flies around 80km/h, so sampling must be pretty high to keep up with plane moves...
Actually no. Idea is to control only power of the motor regarding to actual position. You force actual orientation (move your handle up and down), and want to react the engine properly.
So, here is the one, which I know. http://184.108.40.206/timers_a/timers_a.htm
It costs about 300Euro.
As you see, it is really basic system. PCB are so small, that there is no chance, to fit some big uC. Maybe just some STM?
Maybe the problem is, that it is a wrong approach. System is pretty constrained - it uses only a surface of hemisphere. We can even think about it, as flying in 2D space. I dont need absolute POSITION, but ORIENTATION.
I know, that the problem is, that flight is pretty dynamic, a lot of centrifugal forces during turns in any position.
Again - not that way . Angle sensor is in flying model. You control the model through handle. You cant control motor. It goes on constant power, OR using this device, it automatically changes the power according to actual orientation
I fly diesel, nitro and electric Both RC and CL (not so skilled to try combat, but that's what I am looking for!). Right now, I don't really need that controller. But every guy from top10 on any CL cup, uses such controller, so in future it will be needed. Right now I am training, but I wanted to try to build that controller on my own.
And the real reason is that when you flies nitro/diesel engine, it automatically regulates the power, going lean(more rpm) on ascent and with too much nitro on descent (less rpm). That's where you have lower speed on level flight. And just like flying RC, you add power when going up, and lower it when flying level . I know, that for 50 years everyone was using nitro/diesel engine, and there was no such a problem.
Actually yes, I am considering it as my solution, but I have not seen an example with some dynamic system. I see, that there is a chance to get only the gravity vector (that is all that I need). But I am afraid, that it works rather statically, than dynamically .
If the model is flying at 80 km/h, and assuming the length of the control line is 10 m (radius of circle), then the inward directed, centripetal acceleration v^2/r is about 49 m/s^2 or about 5 g.
To get model orientation with respect to down from the accelerometer, you need to separate out that 5 g component from the downward acceleration due to gravity, 1 g.
If the accelerometer is carefully oriented in the model so that one axis (e.g. Z) is down and another axis (e.g. Y) is directed parallel to the control line, most of that subtraction is already done for you. The tilt angle in radians would be given by atan2(ax,az).
The rate gyros could make the tilt measurement, but only for a short while as they drift. An absolute (non-rate) gyro would be best.
Yes, that's true. The problems I see, are accelerations during sharp turns. What is the result of DMP from 6050 - it is just filtered, mean accelerations value - it doesnt do any sensor fusion, right? I mean, using gyroscopes to get better acceleration values?