This is a discussion re the ArduCopter-3.1 AutoTune feature. A place for people to report successes and failures, ask questions and get answers.
The AutoTune wiki page is here.
You need to be a member of diydrones to add comments!
This reply was deleted.
Replies
Hi Leonard,
Thank you for the clarification! Here are pictures of the rig including CG & distance. CG seems to off by .5-1mm.
I put a tiny (awful quality) stationary camera on the frame so we can get a look at the wobble. After watching the video, it looks like even when the copter is hovering there is a slight wobble (oscilation?). In the video, I hover at low alt in stab for a few seconds, then shoot up & down a couple times. Towards the end I have it Loiter. In all the aforementioned scenarios, you'll see just about the entire time it looks like the APM is trying to correct roll/pitch, especially when losing alt, but surprisingly also when gaining alt straight up.
Thanks again for taking the time to look at this. I've spent probably the last 4 weeks trying to properly tune my Y6. Any help is greatly appreciated. It's not easy (or I am terrible at it). Log of the above flight is attached as well with motors logging enabled.
2013-11-25 17-47 107.log
What are the required conditions for Autotune. I've been trying to find a day with little wind over the last 2 or 3 weeks, but can't seem to get anything good when I can get out and tune.
This link shows the wind conditions in my area. Average is around 16km/h, but goes as low as 3 or 4, but as high as about 30km/h http://wind.willyweather.com.au/wa/peel/westfield.html
Obviously I need it to be as calm as possible, but I can't go and tune at 11pm or 2 in the morning, or even 11am cos im at work.
Damn work, always interrupting my hobby time.
Well, some prop replacements and it flies again, I feel confident that mechanical or electrical failure was not involved. Alt hold and loiter solid (but not perfect, that’s what auto-tune is for …). I did a bit of (ultimately destructive) testing by flying in stabilize, twitching roll to see how far I could get away with. Wanted to reproduce the problem and I sure did … Turned out after looking at tlogs and dataflash that a fast 25 degrees roll and roll back was ok, but I hit the deck at 30-35 degree . Of course in manual I can’t really judge how fast I was rolling back … And if it is out of tune to start with, there’s no knowing how much is too much … Turns out the copter landed perfectly upside down, no damage. Flew happily after …
Question for Leonard or devs … Isn’t AT inducing 15 degrees max? If so, I don’t understand why the copter flipped on the first try. I saw that roll in was flat on the original failed try, I was not touching the sticks, yet is that expected? About the auto-tune process: does it start with stock or existing P and sets D and I to 0, to then increase P, check for response, then deal with D then I? Wondering if a lower P to start with, in my case, would prevent excessive roll … Also, what is the scoop with transmitter trim, does AT take this into consideration? (I have some on the roll axis … ).
For what it’s worth: “Kamikaze” copter that failed AT is reasonably well-built, though, despite it’s name. 25mm CF arms, 10” center plates. A tad under 2.5kg/5pounds as flown (I typically use it with extra small Go-pro gimbal) , 660mm motor to motor. A tad over powered as flown, throttle hover in mid 400. Flew with stock PIDs.
I really want to make it work … Auto-tune kicks ass and rocks! (Even if I didn’t get it to work so far on that one … ) Separates and clearly makes APM superior to the one size fits all (won’t work imho) closed source WKM (I fly that too) ...
Well, I'm not sure if it's 100% relevant to this, but I tried to auto-tune my hexcopter today and it crashed miserably before I even started: I took off in stabilize, but a few seconds later one of the motors went full throttle and the hexcopter just flipped and went down. No major harm done. The main problem is I don't know what happened. The only thing wrong in the log is apparently a software crash judging by the nonsense in the last log message. The log doesn't show a high motor toward the end, so it was probably not instability - just a crash. Is the latest version (3.1 rc7) known to have bugs? If anyone is inclined to look through the log, please let me known if anything looks amiss.
2013-11-25 15-40 1.log
Houston, we have a problem ...
Oops ... (: Well, I played with autotune on my dedicated "kamikaze" quad and ... it didn't go well. That's ok, that's what that copter is for and I learn more from it than my working others, except that .... I can't figure out what went wrong on this one.
Flown this copter many times on rc5 without issues (alt hold, loiter, rtl, etc ..). After the first two AT induced 15-20 deg rolls, the third one appears to have led to a roll of 90degree+, from which I was unable to recover. Have looked at the logs, can't find an explanation. The only thing I can conclude is maybe excessive vibration? (yet the quad hovers at less than .1g vibration). I can't also rule out mechanical failure of course but should add that motors are fine after the crash, props were broken but not loose, etc ...
GPS, baro, etc ... data seems fine.
Will continues delving into logs but I must be missing something obvious ?
2013-11-25 15-12 7.log
I just completed my AutoTune on my Y6 and the tail motors pulse a bit and the roll is very sensitive. I have attached the logs and here is the details.
3dr Y6
APM 2.6 with ext Mag and GPS
4s 5000Mah
rctimer 2830/750kv
Top CF 10x4.5
Bottom CF 11x4.7
Any guidance on removing the pulsing and lowering the sensitivity on roll would be appreciated.
2013-11-25 12-01-18.rlog
2013-11-25 12-01-18.tlog
2013-11-25 12-27 2.log
Hello Leonardthall,
I installed 3.1.RC7, I resetting the EEPROM and with the default values I run Autotune, everything ok about 4 minutes with no wind, but with the parameters of autotune I like oscillations in the post, I declined STAB P 5.25 autotune value STAB P 4, but the fluctuations are the same lower intensity, I wanted to change the code as I had advised you:
"go to config.h on line 786 You Should see this:
# define STABILIZE_RATE_LIMIT 18000
Change it to this:
# Define STABILIZE_RATE_LIMIT 9000"
but I have not found the line to be replaced, I was obliged what should I change?
thanks Fabio
I've autotuned my 3.4kg quad a couple of times now, and for general flight the autotuned PIDs make it handle well (after reducing stab p from 8 to 3.2), but I find descending at max rate (1.5m/s) while stationary in alt-hold/loiter is still terrible. It's very bumpy and pretty much ruins any video I'm shooting.
Specs are 2814-11 motors w APC 12x4.5MR props on 4s, on a UAP2 frame.
I've been running autotune without the gimbal (1kg less), and then putting the gimbal back on. I ended up with the following values:
RATE_PIT_P,0.175
RATE_PIT_D,0.015
Where should I start to try and improve the performance during descent?
Hi Randy and Leonard - thanks to you both for your ongoing support and also to the rest of the dedicated developers who devote so much of their time to support when they could be out enjoying flying themselves.
Leonard - I finally got my altitude drift sorted out in Alt hold and Loiter.
I did another compassmot even though no electronics had moved on my frame since my magpie hit. I think it was the new motors that draw a bit more current that brought my compassmot % up. I moved the GPS/Mag up about 50mm and all became sweet - I had tried all sorts up to this point.
I then had my first opportunity to do an auto tune - went extremely well, very impressive.
I now have a very responsive quad that holds well in all modes and a joy to fly.
Well done Leonard for your efforts.
Regards
John (from work)
I gave in and tried Autotune today with rc7, I had spent a lot of time tuning my quad to fly how I personally like it and it was flying as well as I felt I could expect but I was interested to see what values AT came up with.
Autotune completed in about 4 minutes.
My 2c's:
- I prefer a high I values as to me it feels more locked-in and a little more resistant to disturbance, a personal preference.
- My folding I-quad is a little longer in the fore/aft plane than the left/right plane, accounting for I presume, the variance between the autotuned Roll and Pitch RATE_P values.
- I had tried a higher D value but got some small high frequency oscillations so had settled on 0.011, quad seemed fine with the new values.
- I had chose a much lower STB_x_P value as it makes the quad a lot less nervous to my inputs and much better for any kind of video'ing or FPV, again a personal choice. I have gone back to my values. (A much better choice for beginners too!)
-
51
-
52
-
53
-
54
-
55
of 55 Next