Hi
I had my first opportunity to try Auto Tune in 3.1 (stable) today - the weather in the UK has been abysmal.
Following the Arducopter wiki very carefully I performed all of the steps. The APM 2.5 quad did exactly as described in the wiki and seen in the video. First the roll and then the pitch “dance” was performed. The air had a very gentle breeze and so I had to re-adjust the quad’s position in the field a few times, not by any means an excessive amount of times. And within about 4 minutes the Auto Tune stopped. I landed with Auto Tune switched on, disarmed to save the parameters, then switched it off and flew the aircraft for a short while. It didn’t seem any different.
Upon returning home I checked the screen shot of the Before and After (Arducopter PIDs) and they we identical. The Auto Tune did not take effect.
I’ve attached a log for review (not sure what I’m looking at in this respect). Can anyone tell me why the Auto Tune has apparently not worked? I hope to go back to the field tomorrow to try again - this time I will take my groundstation laptop with me.
We looked a this on the UK multi-rotor forums, and its failing to complete on the pitch axis, but I’m not so familiar with the autotune code and the likely reasons for failing (beyond the fact it looks like its ending up with a minimum value for D probably).
The other thread is here: multi-rotor.co.uk/index.php?topic=5667.0
Is Leonard still interested in autotune failures?
I believe it will fail if during tuning if it hits the minimum allowed gains but still doesn’t see a “successful” response to the twitch. So while changing gains it’s looking for no overshoot (while tuning P) or a characteristic little stall in the rate (if tuning D).
As for what to do about it…I’m not the expert so I’m not sure but I’ve heard that Y6s rarely do well and you should remove any extra wobbly bits on the frame like a camera gimbal.
Thank you Randy, I am flying a quad (DJI 450 clone) with 980KV SunnySky motors and no “wobbly bits”.
This morning I went out and re-ran the Auto Tune while connected to a GCS for telemetry. The air was relatively still, similar to yesterday’s attempt. Although this time I only had to manually adjust position twice, whereas I had to intervene more yesterday.
The Auto Tune succeeded this time and the values returned are significantly different to how I had them manually.
I think this topic can be closed as it may just be an aberration on that first run. But evidently Auto Tune does work on my aircraft.
If you would like the success log posted here do let me know and I will do so.
Thank again.