Guys,

I need some help in setting up yaw PID's. My TBS Discovery with APM2.6 has a poor yaw performance. It seems not very locked and what is most annoying it is slowly oscillating which is most visible in onboard video footage.

To get to the root of the problem I did some log analysis and found that actual_yaw is indeed tracking nav_yaw but leaves some +-3 degrees slow oscillations. Please have a look at this log screen and my current PID's. Any ideas where to start or how to systematically set up the yaw PID's?

Thanks,

Michael

Views: 19465

Attachments:

Reply to This

Replies to This Discussion

If my understanding of github is right, Autotune for Yaw is scheduled for 3.2 

https://github.com/diydrones/ardupilot/issues/874

Tuning my Rate Yaw P up to 0.55 I had better results.

Will keep with Leonards advice and try to rise it up to 1

I have no idea how Yaw-Oscillation by too high values will look like.

Has anybody experienced that?

After having some yaw oscillation issues with Discovery Pro and Pixhawk, I found this thread. Very helpful, although the settings here weren't quite right for me either. 

I'm very split on this one... not sure whether it's best to try and find a 3-axis gimbal for a GoPro or stick with trying to fix the yaw movement. The problem here is that the only readily available 3-axis GoPro gimbal that I'm aware of is the Zenmuse H3-3D which, as far as I know, requires a DJI FC in order to function properly with control over tilt and pitch. Haven't yet heard of anyone hacking it for other FCs.

I would LOVE for 3.2 to have autotune for Yaw, but it wasn't in the RC2 as far as I can tell. 

After trying to tune yesterday, albeit in pretty windy conditions, I think I got the best results with my Stab-Yaw around 5.0 and Rate-Yaw-P also around 5.0, and Rate-Yaw-I at .01. 

Anyone else have any further luck? Any word on Autotune Yaw really ending up in 3.2?

I would like to know too, I ended crashing my F450 and I'm building a TBS Discovery with a pixhawk on it

Sadly, it looks like Yaw Autotune has been pushed to AC 3.3, so we'll be waiting a bit longer for that. 

The frustrating part is that while Autotune is not included for Yaw, it's also very glossed-over in the wiki. I see that lots of people don't use any D-term for Yaw Rate, but I can't figure out why. No one actually explains this.

My auto tuned numbers are much higher than anyone else's I've found on the same frame and motors. Stab-Roll/Pitch are just over 10 and Stab-Yaw is about 6. I'd upload all of my numbers, but seeing how far off they are from everyone else, I don't know if it would be at all helpful?

This is a very frustrating topic and this thread exemplifies the problems with the Pixhawk and APM in general right now. Sure there's autotune for pitch and roll, but you're left completely on your own for Yaw. That would be fine IF the defaults were good... which they're not. So sure there's got to be some great instructional videos and easy-to-read documentation on how to tune Yaw, right?? Yeah, not so much.

I really want to love the Pixhawk and APM platform but the difficulty to tune Yaw, (as exemplified by this very thread!) is why the platform has a long way to go. And yeah, autotune for yaw has slipped to 3.3, so who knows when we'll see that.

I wouldn't be so frustrated if there were good, accurate, and up-to-date documentation and tutorials for how to tune, but there isn't. Most of what's out there is either out of date and thus no longer very applicable, or very confusing and thus not helpful. There needs to be a great video tutorial series called "Tell me how to tune APM like I'm 5". I'm not 5, in fact I'm a pretty quick study and very technology savvy, but so much of the documentation and videos assume that you already understand things. It's as if the platform is built for other experts and developers. There's got to be a better way. Sorry for the rant, but I've been working on this problem for months and the only people that seem to be helpful are other people who are stuck in the exact same rut that I'm in, and none of us have really dug the way out. 

That all being said, I've actually found 3 (three!!!) non-DJI 3-axis gimbals for the GoPro that would potentially work on anything outside of a DJI FC:

DYS Smart3 Gimbal from HobbyKing

Arris CM3000 Gimbal from Hobby Wing

FY-G3 Ultra Gimbal from GetFPV

So, hope springs eternal I guess. If you can't tune it out, throw more money at the problem and get a better gimbal. Sigh... 

Yeh, Arducopter is the only flight controller with auto tune that doesn't support yaw. How crappy is that!

I wanted to get yaw in there but I spent my time working on why auto tune didn't work on small and high powered copters. I realize that helping get their primary flight control working isn't as important as making it easy for you to tune yaw. I am sorry, I made a mistake.

I told work about this situation but they wouldn't pay my salary if I stayed home to fix my mistake. And my kids they keep crying if I don't feed them......

So Yeh it will be delayed. Thanks everyone for their patience. We will get there.

Hey hey, back the truck up. No need to get defensive. I was in no way attacking anyone with my little rant, nor was I trying to imply that you were in any way at fault. I'm frustrated. It happens. There's been a lot of folks asking for advice without that many answer, so what else am I supposed to do?

I think you're getting me wrong here. I'm a fan of Pixhawk and of developers like yourself and what you're doing. I want it to succeed and do well and I encourage others to use it. I'm also a business person and from a business standpoint, it can be a tough sell sometimes. The easy answer is, go try something else if you don't like APM... well, I don't want to do that. I've invested a lot of time learning the system and I'd like to think that I'll get to the point where I really understand it deeply.

Forget Yaw Autotune and ignore that part for a second. Since, well, it's never had it before and still doesn't, solets live in the now I guess. So how do I figure out how to tune Yaw on my own? Where would I go to find that out? It's certainly not in the wiki and there's not a ton of info in the forums.

I'm not saying this to attack anyone or to throw blame anywhere, I simply want it (the platform, the entire system) to be better.

Yeah, Yaw Autotune didn't get into 3.2, and that was probably Randy's call and seeing as how late 3.2 is, it was a completely logical decision. But that still doesn't change the fact that I would love it if someone explained APM tuning in a super simple and easy to understand way. Do I use autotune as it is now? Sure I do! But I also wish I knew not only how it arrives at the numbers it does, and how I could arrive at similar numbers through my own testing.

I was venting. It happens. Don't get defensive though, that doesn't help anyone. I get that there are lots of very talented developers spending their free time on the platform. But again, realize that my criticism comes not from a standpoint that I want it to fail, but rather that I want it to succeed. Being approachable, easing that initial learning curve to make the platform more inviting, can go a long way to making the entire platform flourish.

Does Yaw Rate-D do anything? I've heard yes some places and no other places. Some places just don't mention it, which makes it a bit more confusing. I've also now read that Yaw Rate-I has a max of some number, but I don't know why that is. How does Yaw Rate-P and Stab Yaw relate to each other exactly?

I think tuning is really important. I'm not sitting on my hands counting the hours until yaw autotune arrives, I've been working with what I have (just like all the other folks in this thread) and trying to get that working with the tools we have (3.1.5). But there's so much confusion about tuning Yaw (see this thread...), that it's frustrating for us that are counting on APM to give us good results assuming we put in the effort.

I don't give much advice here because I haven't had the time to explore the yaw tuning issues and most of the community us in the same boat.

My default yaw tune has been to double the defaults and make the D term 0.004. But other than that you will need to talk to others for more detailed ideas on how to tune yaw.

Personally I would start by tuning in the same way as roll and pitch. D up till you get oscillation then half it. Repeat for P then set I =0.1 P (roll and pitch use a ratio of 1 )

Thanks for that. I appreciate your candidness. I'll take that as a good place to start. 

Interesting that you do recommend paying special attention to the D term. It's not even mentioned in the wiki, which is why I was confused whether it's relevant or not (others have said it's ignored and doesn't matter what number you put in there, etc.). But as it is, the wiki really glosses over Yaw tuning in general.

Again, thank you for the advice. Hoping the rain clears up soon so I can get back out and continue testing soon.

I used leonard PIDs which I think improved my TBS Discovery yaw movement a lot, they're at 0.4, 0.04 and 0.004 and this is what I get, could I improve it even more?

Attachments:

Esteban, those are your Rate Yaw PIDs, but what is your Stabilize Yaw-P value? 

Reply to Discussion

RSS

© 2019   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service