Hi,
We did a XAUV Talon flight with latest Arduplane release.
The flight started fine in auto take off with a catapult method.
Once in flight the plane started to make a sinusoidal trajectory as if the center of gravity was off, although the CG had been perfectly tuned before take-off. For that part, I suspect the battery moved during the catapult launch, so that is easily explained (we also see clearly in the log a difference between desired pitch and pitch).
What is more strange and something I can't explain is that , once flipped into auto mode, the plane started to loose altitude from 70m to 40m (between line 550 and 580) when the waypoint should have brought it to 100m high ! (see picture below).
The plane was immediately flipped back to FBWA to increase "manually" its altitude back up (from line 580).
After a few moments, still in FBWA the plane starts to climb alone without any reason (pitch remains flat near zero), as shown in picture (from line 640):
- First question : What can cause the plane to loose altitude in auto mode when it should climb ?and reversely why does it start climbing in FBWA with no pitch commanding to go up ? Could it be a CG balance problem putting the nose up ? Strong winds ? Bad parameters ?
Then mission planner displayed these EKF errors that I do not know how to interpret,
- Second question : what mean these EKF errors ? (see picture)
Error pos horiz variance
Error velocity variance
Attached the log via "wetransfer" as it exceeds the 7Mb limit:
Thanks for helping understand what went (or is) wrong with the altitude weird behaviour (question 1) and interpretation of the EKF errors (question 2),
rgs
Hugues
Replies
For update, we did another test flight which was succesful.
The main differences with the problematic flight were:
-AccelZ way too high. Pixhawk was remounted properly to reduce vibrations
-Recalibration of Accel
-Recalibration of compasses
-Recentering of masses to have the CG where it should be on a XUAV Talon airframe
We have to tune still a bit the airspeed ratio because there is still in average a a difference between the GPS and pitot tube speeds.
Still have no clue how to explain the EKF errors from the problematic flight. I can only speculate it had to do with vibrations ? If a dev knows better about that, some explanations would be nice.
thx
@Hugues,
do you apply catapult from this video ?
https://m.youtube.com/watch?v=UqjzJBJtc-A
I am just building Drone Test Lab within Open Fabrication Lab via Open Technology Park.
Many computers, large monitors to play back log files in 3D Flight Simulator
and full problem analysis.
Now I need EKF manual with help file, as complete as possible, by developer/s and third parties.
Need full vocabulary of ArduCopter, ArduPlane, EFK terms, abbreviations with help file to let me forward flight analysis to research team members not involved directly with ArduPlane, ArduCopter or DIYDrones, to save their time spent on deciphering every term, abbreviation used.
Your log file plots resemble me early GPS plots made of a cloud of random geolocations,
log file values jump up, down at random.
Could you describe sensors installed and sensor clocks, min-max output value range ?
Could you test your GPS outdoor, placing your plane still at one place and having GPS geolocation recorded to log file for 1-3 minutes ?
Could you conduct the same test with sensors, moving your plane manually around, to know what data could you expect to get recorded and later played back and plotted (within range) to collect as much sensor output data recorded into log file ?
Open Technology Park
OpenFabLab
darius
manta103g@gmail.com
Yes that is the catapult that is used.
The sensors used are the 3DR GPS+Compass (external module) & of course Pixhawk with its own internal sensors.
@Hugues,
problem already solved
are you still need my help ?
I can provide you with step-by-step clarification to problems with your drone.
Open Technology Park (OpenTechnologyPark)
Open Fabrication Lab (OpenFabLab)
Peer To Drone Crash Investigators
manta103g@gmail.com
@Hugues,
I would start from Yaw chart since your catapult doesn't seem to affect sensors badly, not generating huge acceleration to make them out of range.
Since my PC is still off-line, I get unlocked modem tommorrow,
could you attach GPS track over Google Maps with data stamps
to let us match Yaw against heading read from compass/ GPS.
I am looking for an option to chart logfile data in multi windows in one web page.
Not sure how to convert .bin .log file into mySQL or another csv format for easy viewing, charting in HTML5.0.
Since Pitch and DesPitch charts match and are both flat until auto mode is switched on, altitude may get lost due to Pitch and DesPitch ups and downs around "0" value since more energy at fixed throtle is lost to follow fluctuations in DesPitch around "0" value and your Plane (not Quad) as badly as predicted ( low speed, wind gusts or throttle not properly calibrated).
Pls read
"
Patrick Hanley "Aerodynamics in Plain English" on DIY Drones
"
PS
I was busy with my tablet went off due to battery charging problems.
Charger is hot but battery is still 71% after 3 hours of charging.
Do I really need a new 10 inch Android tablet ?
The EKF horz and velocity variences mean that the gyros and accellerometer 'dead reconing' positional data is not matching up to the data obtained from the barometer and gps sensors. From playing your logs, I'm guessing its due to the noise and jumpiness that is visible in your HUD. To me you're either getting way too many vibrations transferred into the APM, or you've got a bad sensor somewhere. Since its two errors, I'm betting vibrations.
As for the altitude hard to tell. It didn't seem that out of the ordinary playing the logs, but sometimes its hard to tell. I'd get the battery mounting situation and vibrations figured out and try again.
Thx a lot.
About altitude weird descent in auto mode, I made this snapshot showing alt-throttle-pitch:
Does it fly well in Cruise mode?
If you haven't used that mode yet, you should verify flight in that mode before trying the auto modes. The auto modes are more complex than the Cruise mode, and the Cruise mode requires a lot more sensors than FBWA. Get Cruise working and flying stable - then start troubleshooting Auto mode. If Cruise mode flies well, then that should help lead to a narrower list of possibilities of why Auto is doing what it is.
Never tried cruise mode. Auto mode was working perfect with previous release. I wonder what parameters changed or new parameters came with the latest arduplane release.
Did you save older params to compare?