Jason is travelling this week, so I'll take the helm for the next software release post.
UPDATE: the motor remapping thing was confusing everyone, so we took that out and returned to the regular motor mapping. That means that APM 2 users with Hexas and Octos should wait for the next version. APM 1 users should be fine with any frame.
NOTE: Hexa and Octo users: there have been motor mapping changes that may affect you. Please don't upgrade until we can update the documentation to reflect the changes. This should happen by the end of the day today (Feb 1).
ArduCopter 2.3 is now available in the Mission Planner. This is the next revision of the ArduCopter 2.2B6 code, which is perhaps the most tested code we've ever released (1288 comments in the thread!) and certainly in my experience the best code, too.
The default PIDs are optimized for a 3DR/Jdrones quad with 850 motors and 10" props. If you're using more powerful motors/props, start by turning down Rate Roll P (default is 0.14, so start by turning it down to 0.1. In general tune PIDs in 25% steps).
Now that we've got solid code out there, we can turn to collecting suggested gains for standard frames, and a better guide to how to tune PIDs for your unique setups.
Here are Jason's note on the latest changes (mostly from 2.2B6)
A dampening term called STAB_D has been refined. A D term for all of the Rate based control loops has been added based on Igor's work. Landing for Baro and Sonar has been refined based on JLN's work. A slightly new approach to Loiter and Navigation is being used to try and linearize the pitch and roll for rate control. It tends to use lower gains, yet has a more assertive response in the air.
STAB_D : This is the gyro accretion dampener. This can remove small wobbles during sharp changes in angle commands. Making this too high can have a negative effect in performance and add a memory effect that can cause temporary loss in control. The in flight tuning is ranged so you are just below that effect.
If you haven't noticed before the control loops are in two stages. The first is a PI stage that converts some sort of position or angle error into a desired rate. These generally do not need to be tuned. They are more of a user preference on how fast you want the copter to perform a motion.
The second stage is the actual PID loop that needs to be tuned for the copter. This converts the desired rate into a motor command of some sort. I added a D term based on Igor's recommendation to the PI's for each rate controller. These should show up soon in the mission planner for the release. I cannot give you a concrete answer for how to tune the D terms, because they each depend on their function such as alt hold or loiter, etc.
Still, the absolute most important term is always the Rate_P term for each loop. Start tuning here.
The default PIDs are in the what flies great for a stock jDrones/3DR Quad with the purple motors in X mode.
Note the Mission Planner does not yet highlight these D terms on the main tuning page (it will soon), but you can find them and modify them on in the Parameters list.
Autolanding should now work well (see video above) and the Tri servo issue is now resolved.
The code should now compile with Arduino 1.0 (thank, Randy!), but remember that you need to use the "relaxpatch" version of Arduino in our downloads section.
[Update: we've reverted the below. See update at the top of the post]
Important for Octo users:
We've changed some of the motor orders for some more exotic airframes. We'll be updating the docs on the Wiki in a day or two to reflect this. Pat Hickey explains:
As before, the hexa plus APM2 motor setup has changed from the ordering [1, 2, 3, 4, 5, 6] to [ 5, 6, 1, 2, 3, 4 ].
The Octa V layout for APM2 is:6 42 58 13 7Motors 1 through 4 spin clockwise, and 5 through 8 spin counterclockwise.
And RATE roll & pitch P at zero? That seems scary to me!
How is that even possible???
Denny, can you explain *exactly* what the timing setting does? I've always wondered.
I gave the 2.3 version a try today, but it really did not work at all for me, maybe something has happened to my hw, it just seems wacko!
First I just downloaded by MP, edited my params (no erase/reset, think that was done automatically), and tried to handheld. It was not stable at all and tried to flip forward in pitch, no resistance to stabilize. Saved a version of the params.
Then reloaded and did erase/reset in CLI, went through config and recalibrated radio level etc. Tried a handhold in stabilize and this was worse than ever. Really hard doward pitch, and after some time abrupt changes in motorspeeds and inclination as if it were flying an auto mission, almost lost the copter.
I diffed the params and noticed that there were a couple of diffs that puzzled me. In the first try the RATE_RLL_P and RATE_PIT_P were set to 0 ???
In the second try (after going through a full setup) they were both set to 0.14, but that made the test even worse!
Dont know what to search for here, but I enclose the log if anyone cares to have a look.
PS, a documentation for the logs would have been fine too!
Hope this has nothing to do with motor reversals, I have a quad in x.
you describe my experience.
puzzled as well and not so hapy about the reorganized attitude controller.
Pancakes eaten. More flights performed. Spirit is good and performance still promising.
My quad now flies very well in stable mode, better than ever before. It is smooth, predictable and stable. Simply enjoyable to fly. ACRO mode does not feel well for hand flying, but I have come to think that is not so significant, now I just regard it as a sub-controller used by STABLE.
I have done some loiter flights in my garden. Dark outside, -7 degree Celsius, some week wind close to ground, a little more 30 m up above tree-tops. After some tuning (maybe not final) it performs well. It would stay put within a few meters laterally and some meter vertically for an entire flight, without me intervening. And it responds harmonically on lateral stick inputs, moving it laterally. Fairly good with vertical commands too, even if that was not so thoroughly tested.
During R&P rate D tuning, I noticed that at hover throttle rate D parameter could be 0,02 or maybe more without oscillation. But when increasing to full throttle, there would be subtle fast oscillation. So I left the rate D at 0,01.
I think this can be explained by propeller thrust being a non linear function of ESC input value. The curve gets significantly steeper at higher RPM, and so the system loop gain becomes higher at full throttle. In future versions maybe this non-linear transfer function can be accommodated in the algorithms somehow (suggestion!).
Tomorrow I hope to try some RTL, simple mode and AUTO, on a mission. My first mission! Exciting.
By the way, this is the first time I seriously considered letting it go on a mission. Previous FW version did not give me confidence enough, but this one does. That´s a BIG leap forward! :-)
Providing my parameter file.
the initial wobbling should go away when setting RATE P & I to tero.
just having the d-term with the pidt1 element active is fine.
Which part? the copter being wacko :)
Or the diffs in RATE_P?
Geir: what board/frame are you flying? We've tested on as many airframes as we know how and they all work great, right out of the box, with the stock gains. Just this morning I loaded it out of the MP onto three quads (APM 1, 2, X and +) and they were all great, with no changes.
You can see all the default gains here.
Are you sure Rate Roll P was 0?? I don't see how that can be.
Question for some PID gurus.... Does the rate integral gain help compensate for CG imbalances like the stab integral gain? The attitude tuning wiki says that using the rate integral gain in favor of the stab integral gain may perform better, but I need to understand if it will effectively compensate for CG issues.
Yes it is strange, I have std Jdrones quad in X frame. I am quite sure the Rate Roll/pitch was 0, I saved the params and diffed them afterwards. Tomas commented the same earlier in the thread!