Interesting (Extreme?) autotune results ...

3691189965?profile=originalNever seen anything like this .... D 20 times default, I 3.5 times default, P over twice default, Stab pitch and roll  three times ...

Copter flew fine on stock PIDs, also flies fine with these new parameters from autotune, although I now see an occasional twitch in pos hold. Copter is a fairly standard 4kg quad on 6S with 15" props, very sturdy, neither over nor under powered ...

Anyone experienced something similar?

You need to be a member of diydrones to add comments!

Join diydrones

Email me when people reply –

Replies

  • 3701962458?profile=original3701962411?profile=originalMakes sense (Re: twitches more visible on autotuned pids).

    I took a closer look at logs to try to get a clue on what's going on, must admit I am in the dark ...

    Second  pic shows desired roll vs roll (first part of auto-tune). The flat green at the beginning does not makes sense, I suspect that's an artefact.

    First  pic is a close up of latter rolls. I'd think somewhere around 56 are where the optimum settings are. Need to understand Leonard's scheme better ...

    Just for grins, I may try the whole thing again with a "regular" 3DR GPS instead of m8. Better be the same ... :)

    • Developer

      Hi John,

      It is hard to say too much about your tune without looking at the logs. I have developed autotune further for the next release. It will check a few more edge cases and hopefully be a little more robust to some of the more interesting setups.... maybe....

      Looking at the graphs above it looks like you have a good tune and I wouldn't be concerned about your results. As usual back the stab  P off to soften the response but don't change the Rate settings.

      You are correct about the response around 56 being what you are looking for. Autotune increases Stab P until it gets an overshoot then backs it off by 75%, hopefully back to about where the 56 point is.

      The new autotune has some improved tests so it is able to stop when it starts to overshoot rather than look for a 10% overshoot then back off.

      The sharper tune is probably not creating any twitching, it is more likely that the softer tune is hiding twitching that has always been there. We have been slowly polishing up the controller designs, especially the navigation controllers, to make them as smooth as possible. There is still work to be done...

    • btw, you are a bold man flying master on what I suspect is an expensive rig!

      Not sure I am quite ready for that  -:)

      • T3

        I always fly the latest beta or master on my own machines. Never had any real problems :-) 

        But you better wait a while with testing a new build if there is a warning on drones discuss due to some fundamental changes in the code. 

        Yes, would be perfect if you could test the 3DR GPS with the autotune settings and see if the twitches are still visible.

  • Do you see the twitches on both stock and autotuned PID's?

    I only see them on the auto-tuned settings ...

    I've progressively backed out from the autotuned settings, to try to find a middle of the road sort of balance. Went from 12 to 8 for stab roll and pitch, then from 0.35 to 0.2 for rate P and I, finally decreased D from 0.02 to 0.008. Am amazed at how little difference these rather large adjusments make ...

    I need to do some more tests and may end up reverting to stock PIDs, as  what I am after is the best possible stability (film), I don't care about aggressive/tight control ..

    Are you ending up flying with stock PID's?

    • T3

      The twitches are much more visible when flying the autotuned PIDs. I did the same and fly PIDs somewhere between  the stock and the autotune ones. I am looking forward to AC 3.3. I flew master a while ago and it was less twitchy.

  • T3

    I have similar results (however I and D are a little lower on my 3kg, 4S, 15'' setups). The copters fly fine with stock PIDs and very aggressive on Autotune. But most interestingly it twitches from time to time. Some of these twitches are related to GPS timing jitter. 

  • Edit: 5 times default.

This reply was deleted.

Activity

DIY Robocars via Twitter
RT @TinkerGen_: "The Tinkergen MARK ($199) is my new favorite starter robocar. It’s got everything — computer vision, deep learning, sensor…
yesterday
DIY Robocars via Twitter
yesterday
DIY Robocars via Twitter
RT @roboton_io: Join our FREE Sumo Competition 🤖🏆 👉 https://roboton.io/ranking/vsc2020 #sumo #robot #edtech #competition #games4ed https://t.co/WOx…
Nov 16
DIY Drones via Twitter
First impressions of Tinkergen MARK robocar https://ift.tt/36IeZHc
Nov 16
DIY Robocars via Twitter
Our review of the @TinkerGen_ MARK robocar, which is the best on the market right now https://diyrobocars.com/2020/11/15/first-impressions-of-tinkergen-mark-robocar/ https://t.co/ENIlU5SfZ2
Nov 15
DIY Robocars via Twitter
RT @Ingmar_Stapel: I have now explained the OpenBot project in great detail on my blog with 12 articles step by step. I hope you enjoy read…
Nov 15
DIY Robocars via Twitter
RT @DAVGtech: This is a must attend. Click the link, follow link to read the story, sign up. #chaos2020 #digitalconnection #digitalworld ht…
Nov 15
DIY Robocars via Twitter
RT @a1k0n: Got a new chassis for outdoor races (hobbyking Quantum Vandal) but I totally didn't expect that it might cause problems for my g…
Nov 11
DIY Drones via Twitter
First impressions of the Intel OpenBot https://ift.tt/36qkVV4
Nov 10
DIY Robocars via Twitter
Nov 9
DIY Robocars via Twitter
Excellent use of cardboard instead of 3D printing! https://twitter.com/Ingmar_Stapel/status/1324960595318333441
Nov 7
DIY Robocars via Twitter
RT @chr1sa: We've got a record 50 teams competing in this month's @DIYRobocars @donkey_car virtual AI car race. Starting today at 10:00am…
Nov 7
DIY Robocars via Twitter
Nov 6
DIY Robocars via Twitter
RT @a1k0n: Car's view, using a fisheye camera. The ceiling light tracking algorithm gave me some ideas to improve ConeSLAM, and having grou…
Nov 5
DIY Robocars via Twitter
RT @a1k0n: To get ground truth I measured the rug, found the pixel coordinates of its corners, calibrated my phone camera with my standard…
Nov 5
DIY Robocars via Twitter
RT @a1k0n: @DIYRobocars is back in December, but outside. Time to reinvestigate ConeSLAM! I rigged up a quick and dirty ground-truth captur…
Nov 5
More…