Today I did a lot of testing and tuning with the new unreleased beta version of ArduCopter code, known as "2.9.2ß" (V3 i think for the official exit), quad with APM V2.5 and U-Blox GPS by 3DRobotics.
The inertial control developed by Leonard and Randy, with the support of Jason, is finally giving the results that for over a year all we are waiting.
To be perfectly sure I've repeated this test at least 10 times in three different flight, with the same great performance.
As you can see from this image and video the max radius in Loiter is now not over one meter in windless/gusty winds and max two meters in strong crosswind, near the position hold performance of the best commercial product, DJI Wookong-M.
We are trying to improve the waypoint navigation but now we're here, the next release will be really a milestone.
We still have to solve a few problems related to the inertial control x/y/z, but I am sure that the excellent results achieved so far are to mean that we are on the right track.
I attach the telemetry log.
DJI, here we come...
Ciao!
Marco
-= AC Dev Team =-
EDIT: in my video you can see the version "3" of the code, nothing secret of course, is the current master branch on GitHub called at the moment "2.9.1b-DEV".
Jason Short on the dev list has proposed to move version from 2.9.1 to 3.x for the massive changes inside the code (ap_hal, PX4 support and other), and following its proposal I have appointed as such in my videos, that's all.
Comments
You're doing something wrong and I think you know what, let's see if you can find what ... :-)
FYI Marco. I tried compiling this morning but came up with a lot exceptions (compile errors). Looks pretty much like in a WIP (work in progress) state.. And thanks for the warning..
I think I'll just wait for the official/stable release and leave the testing to fearless leaders like you. (LOL ;o)
All the best.
Daniel
@Andreas: be careful to try these versions are not considered stable, is at your own risk, you are warned.
This beta uses "ap_hal" library, many parts of the code have been rewritten to have full compatibility with PX4, so it is possible that there may be bugs, even though all the ones I've found have already been reported.
@Cliff-E: I measured visually the range during Loiter, it is obvious that the log indeed shows that detects gps, not the actual real radius.
WAAS is only North America. South American countries that can see the horizon of NoA can get a weak WAAS signal, but it will be less accurate, say 2-3meters.The closer you are to a WAAS GC, the better. EU has something similar in system capability... then there's GLONASS.
We did walk around tests with a non-flashed MTK (AC2.7/APM1.4) under WAAS and got down to 1.1m in resolution, but it did drift around over time (around every 30-40secs) +- a meter. In the air, having a moving antenna, it's a different story (+-3m on average), likely a result of a small patch antenna and [pretty much absent] ground plane.
Yes, GPS logs can be a little misleading and I would agree measuring the actual ground truth (a laser?) would be a 'for sure' measurement. Heck I found GoogleMaps/Earth, Bing, JPL/NASA, and Mapquest data to be slightly off of actual lat-longs at the sub meter level (there's such thing as bad orthorectification). Of course, then we get into DTED and other factors.
Nonetheless, having used APMs extensively with it's behaviors (and quirks), looks very good Marco.
Fantastic work!
Could not resist to give. 2.9.2b a try.
Stabilize worked. Loiter seemed pretty good. I tried to nudge the copter away from myself. But then discovered that even full stick input did not move the copter at all. Switched back into stabilize and landed. I don't have the guts to try it again in my backyard. That was rather scary...
How well does it worth with PX4? Is there any point in investing in an APM 2.5?
Thanks.
Yes, is the same.
Hi Marco.
I must be missing something, as when opened (master branch downloaded file) in Arduino, the Ardcopter.pde still have the "THISFIRMWARE" identified as "ArduCopter V2.9.1b-dev". Is this nonetheless equivalent to 2.9.2ß?
Cheers.
Daniel
-
10
-
11
-
12
-
13
-
14
of 15 Next