Originally Posted by ukionz
I was so boring waiting the postman to send my Arkbird. Searching the web for more info......not much on this thing, so go to Chinese site to learn more.
Ok, found this, the Arkbird designer share his thought on design & finetune of this system. Again it is Simplified Chinese site (Google translate is your friend)
I wouldn't go off topic, if linking Chinese site here is bad, let me know.....I will stop doing this.
Google Translate translation (good luck!):
Depth analysis of flight control (balancer) software technology Share Arkbird flight control design experience
Arkbird flight control test flight commissioning starting in 2010, initially by Jiangnan Big Brother, the exchange of queens and discussion help (when doing open source version of the infrared), they were pretty good, to express my sincere gratitude to them, and would like to share flight control design, debugging experience, test data. For a deeper understanding of flight control some help.
A. The difference and importance of the software
Three foreign open source flight control, looks almost, but the performance and prices vary greatly. Why?
Read open source, the KK Flight Control, only about 200 lines, MK has more than 4,000 lines of software is the real key to distinguish between the flight control.
The hardware system board as long as 10 days, the software debugging ranging from six months to more than a year,
It is no exaggeration to say that the hardware, peripherals bottle of software is the bottle of wine. And most users smoke and mirrors, the fly can not see the difference between the control software.
Software adjust the year? Exaggerated, right? To debug it? Please read on:
II. There is no good algorithm, sensors and consequently undetectable
The Throttle larger sensors become miserable. Different filtering algorithm, shown in blue, red, yellow, brown a few lines (four values of the great difference between the average, the value of the average, 16 average, Kalman filtering).
Even if the sensor selection, different engineers selected algorithm is different, very different. Seismic performance, will be completely different.
Do UAV will the famous Kalman algorithm, Arkbird use a "simplified" version of the Kalman algorithm. The algorithm benefits from a confusion of values, to find the true value.
In order to carry out the algorithm, you want to debug months.
III. Interfere with the acceleration of the impact
Flight Control gravity measured vertically, sometimes very tricky. Such as sitting in a roller coaster car acceleration was "pushing back" turn was "thrown out", hang a pendant on the roller coaster, the pendant will be thrown upside. Gravity determination, it will be very accurate.
Interference acceleration up to 3-4 times gravity. Very scary, If you do not do software estimation and compensation, gravity will be completely drowned out most of the flight control, acceleration health value "a display, but this value is not calculating the standard, the same flight, A fly control "health value" to 0, but the B Flight Control "health value" 50, which is also related with the engineers to design, debug.
Of course, there are a lot of flight control is not to compensate.
This mold Friends of flight shots, you can see more than 139km / h, the health value of only 5.
IV. Attitude solution
Just said easy to inaccurate sensor, then this is not an accurate measurement value calculation of flight attitude
Euler angles, quaternion algorithm (divided into first, second, some UAVs with 16 order), different attitude solution approach, the accuracy is also different, take a look at the following figure the goodness of fit of the two curves. These differences are dynamic error means that the ground can not see, only fly to see that:
In order to do more accurate, but also commissioning several months.
V. PID - flight control algorithms "
Baidu Library Search "PID" paper, the results of 1,453,197. PID variety. Simple differential forward fuzzy control, neural network control, adaptive (self-learning) control ........ I fuzzy control of a soft spot for different designers will select a different algorithm.
The PID regulator needs an engineer long-term experience. Usually to measure the system response curve analysis system poles (shock point), engaged in UAV brothers also do matlab simulation, draw a lot of response plans, pole analysis. Write a lot of papers.
Then adjust the tone, and finally one day, found the tone very stable, like this picture below, the fuzzy control curve than "change the PID control curve small shocks, more stable.
Upgrade point differential forward (to avoid the control value mutation), increase the variety of intelligent control, limiter like.
Spend so much time tune this what use is it? Good tune, take a look at the beautiful the large Xinjiang's four-axis "snaps" foreigners playing table tennis with four axes, to build a house with four axes, is quite stable, with like still locked up in the air.
Six. About spiral stall
Breeze +2212 motor +2200 mah battery, quite good flying, I believe very few people can fly to spiral out.
But when Arkbird tune PID in 11 years in July, actually be able to breeze fly spiral. The following video, it can be seen that the spiral is "very strange" state. And snails so probably not the reason of the aircraft itself.
Check data said spiral flow separation, the rudder is invalid. For example, get threw poker. If poker "cut wind throw, you can fly very far; poker" airflow separation, then it will be like a piece of paper falling.
So in PID control algorithm to increase the way to prevent flow separation (slow control), and solved the problem.
Was asked Arkbird why Placing airspeed gauge, of course, not because of our technology is bad, not do this.
But for UAV and large aircraft, the airspeed slow very dangerous, easy to stall, increase the airspeed gauge can effectively prevent large aircraft stall. For small model aircraft, and a little bit of throttle is not easy "flow separation" or "stall". Optimized control algorithm, not the one case Arkbird user feedback spiral, stall airspeed gauge to temporarily little role.
VII. Other software details
The above said the same hardware, different software caused a big difference, or even several times the difference. In addition, there are more details, such as elevator - aileron mixed control (turn lever), throttle - elevator mixing (fuel rod) ...
These software effort, might do for a few months, users not see that.
KK flight control 200 line program, the same as it was able to tune Detuo controlled flight, the foreign Niubi flight control, also off control flight (even playing table tennis),
So maybe, flight control, good or bad, to see the flight video, is not a true hand contrast, might not objective.
Finally, sharing just mentioned, sensor vibration test data, a total of 30 documents, consisting of various sensor test data. Help you DIY flight control.
DIY flight control, and more than we imagined. A few days ago, a flight control DIY group, nearly 200 people. I believe each person to do the flight control, all love flying, love technology.