I just installed 3.1 RC-7 on a tricopter and tried out Autotune and Drift mode. I thought I'd post the results as there don't seem to be a lot of tri's running APMs.
The tricopter in question is a resurrected GetFPV Delrin Tricopter kit. By "resurrected" I mean so beat up it was time for a complete rebuild. The body is mostly original but the yaw servo mechanism was modeled after RCExplorer's
Here are the tricopter's specs: Weight with HD cam, no FPV, and 4S 3300 mAh battery, 1,240 grams. Distance from center to motors, 370 mm. APM 2.5 clone with GPS, external magnetometer and PM, Motors/props, FM4008 620KV/APC 11x4.7. ESCs, Turnigy Plush 18 amp flashed with BHeli firmware (one ESC provides power to the yaw servo).
After installing RC-7 I tried Autotune. It was rather disconcerting to see the tri twitching all over the sky, as that in my experience is immediately followed by something very bad. But not so this time. After Autotune the thing flew great (it was flying OK before on 3.01, but much better after).
In looking at the PIDs, Autotune made some substantial changes. Here they are in a Before Autotune/After Autotune format:
Roll P: 0.095/0.075
Roll I: 0.100/0.075
Roll D: 0.006/0.006
Pitch P: 0.095/0.045
Pitch I: 0.100/0.045
Pitch D: 0.006/0.010
Stabilize Roll: 4.5/5.25
Stabilize Pitch: 4.5/5.25
As fantastic as the Autotune was, Drift mode was a disappointment. Pitching forward for forward motion was OK, but asking for roll was....exciting. It yawed with the roll, as it should, but there were wild roll oscillations. It felt similar to an extremely high P value or like it was far too sensitive to roll inputs. Very overcontrolling.
Looking at the logs verified my observations (see the image below). I plotted the Roll in against the Roll out. While in AltHold the Roll out tracked the Roll in just fine, as it should. But as you can see when the tricopter was put into Drift mode the Roll out was strongly oscillating around the Roll in. Like I said, exciting.
And I do know how Drift mode is supposed to be as I also installed RC-7 on my QAV500. Autotune worked magnificently on it too, and so did Drift mode. It was a complete blast to fly the QAV around maniacally in Drift with no problems at all. So maybe Drift, in its current iteration, is not so hot for tricopters. But Autotune, yeah!
Replies
An update to this post......After a lot of attempts at tuning I finally decided my problem might be that the tricopter's arms were too flexible. They are square, 10 mm Poplar wood, with a span between the body and heavy motors of just over 300 mm.
To the rear arm I epoxied 0.8 mm skins of G-10 fiberglass along all surfaces. This had the effect of making the arm exceptionally stiff both laterally and torsionally. Flight testing showed much improved handling, including elimination of an annoying slow yaw wobble I had been trying to get rid of. Testing of the Drift mode flight showed it to be very good, comparable to what I experienced with my QAV500 quad. Apparently arm stiffness is very important to tricopter handling and successful Drift mode flight. I didn't bother stiffening the two front arms as doing just the rear was adequate to get rid of all problems.
So....Drift Mode, Yay!