I've put together a test stand for my 450 class quadcopter running APM 2.6/Arducopter 3.2. After a recent crash where my quad suddenly fell out of the sky, I want to run some tests on the stand, keeping most variables the same as if it were in flight. I'm pretty sure the crash was due to battery cell failure but I want to run some tests on the stand using the same battery (which still balance charges okay).
My quad is zip tied solidly to the stand, i.e. it cannot move. This presents an apparent problem because the motors won't all run at the same speed in Stabilize flight mode. How can I achieve this kind of test? Do I need to attach the ship to a rod on a ball joint? I've seen a couple of videos on YouTube where people did this, but I would rather have my ship solidly mounted. Have any of you done this?
thanks
You need to be a member of diydrones to add comments!
Replies
Here's a compilation of my tests so far. I think have the PIDs for ROLL dialed in pretty well, but I don't really know. I'll do some outdoor tests next to see if I've made improvements. In this video, I display the PID values for each test.
To repeat some points:
1) During the initial tests, the quad couldn't even remain on the stand without tilting over to one side. It was using PID settings from Arducopter's AutoTune and it SEEMED to fly pretty well. Then again, I also had an unexplained crash.
2) I've noticed that AutoTune always puts the same values for P and I. Why?
3) I'm still not clear why you should set D. Some people say its important, others say to leave it alone. Both groups have ships that fly really well.
4) During test 4, I applied maximum left stick and the quad started twitching a bit. I can't tell if that's because P is too high or because the quad is fighting against the tether. When I lowered P and I a little for test 5, the twitching seemed to be less as well.
Thoughts?
Really interesting Robert.
Set I outdoors when dropping through prop wash. It will take a while.
Leaving P at .1, increase D until you see that quiver when D was at .011. Then back off D 10% or so.
Then increase P until it quivers again. Then back it off 10% or so.
Then fly it outside. If it quivers, back off P more.
Then fly it in wind. If it quivers, back off P more.
Then the only issue is to decouple roll and pitch PIDs and see if they can be tuned even finer. Right now the ship rolls side to side over the air bubble but not fore and aft, telling me that the PIDs should be different.
Thanks Forrest! Roll and Pitch have ALWAYS been decoupled because the weight on my quad is distributed fore and aft instead of in the center. I discovered this early on when my quad was a completely stock QAV400.
Pitch (all tests, from 20150203 to current):
P = 0.1750
I = 0.1750
D = 0.0165
thanks ... now your results make sense. was trying to figure out how the longer/heavier axis could be more stable than roll using the same low PIDs. well done.
Did you disable and disconnect GPS during your indoor tests ?
Darius - Stable and Acro do not rely on GPS. So do you see any reason to disable GPS? As a precaution, i'd make sure that all modes on the Tx were set to stable so that an accidental switch change doesn't send the ship flying in another direction.
Forrest,
GPS is serialized data transmission.
GPS is a process consuming processor's power
If not disabled, disconnected, process is pinging GPS hardware to get fix.
Getting GPS fix in indoor environment is highly problematic, since signal from satellites is reflected by walls, roof.
So the processor and GPS daemon process is busy pinging for GPS fix.
It is not smart to get slow processor overloaded with byprocesses
responding to Robert I get a reply from you.
Are you Robert's twin brother ?
yes ... we are twins. thank you.
understand your concern, but in our experience, having the GPS search is not enough of an issue to interfere with tuning.
Hey Darius,
First, I appreciate your input. I'm on this forum to share ideas and work to collectively figure things out that I would have a much harder time figuring out on my own. This is an open thread. Everyone is free to comment.
I've done a lot of programming, mostly for desktop and web applications, but I'm not familiar with the code for Arducopter. Even so, I can't imagine that updating the GPS state requires so many CPU cycles to affect performance. I'm using STABILIZE flight mode, which does not use GPS. I realize the external GPS unit is still pinging the controller because sometimes it will acquire a GPS signal while I'm in the middle of a test. I have my FPV equipment connected so I can constantly see the GPS state on my OSD.
The whole point of my test stand is to fly the quad exactly the way it would be flying outdoors, with gimbal, GoPro, FPV gear, etc. It may or may not be the best way to tune PIDs, but the stand has several purposes. I'm still experimenting with all of it.
If you have some real data showing that the GPS process can overload the processor, please share it with the community by starting a thread about this issue.
Finally, Forrest is not my twin but he's a friend outside of this forum.
Hi Robert,
it looks like I have got a solution to your stand problems.
got this thread from your recent discussion
--
Permalink Reply by Randy on March 5, 2013 at 6:27pm
The critical time is when the red-blue lights are flashing when you arm. At that point it's calibrating the gyro center and it will keep trying to calibrate if it senses too much movement. My guess is that if the boat is rocking too much it will take a long time to calibrate but once it does, i think it'll work
---
Since intelligence , logistics, wire frame behind autopilot has not been disclosed so try to guess approach is the only way to get to know how it works.
GPS is not an issue but active gpsdemon process can invoke a number of child processes to keep processor busy.
Intelligent drone can sense wind strength and wind direction.
If this is the case, intelligent drone can sense your tethering as strong, downwards wind, pushing your drone down to the ground requiring urgent action.
"
The whole point of my test stand is to fly the quad exactly the way it would be flying outdoors,
"
Exactly the case, but tethering mode can be sensed by an intelligent drone as above.
Ok, you can turn debug mode on to save listed processes to sd_ram with shell script, from time to time.
Good example of how an application or hadware add-on can freeze a processor is my Android tablet.
Default web browser freezes 10 inch Android 4.2.2 at random.
My touch screen pings frequently, making cursor of the connected PC mouse and PC keyboard to jump over the screen at random to invoke screenshot icon from time to time.
In theory I should disable touch screen process since I don't use it at all -10 inch Android tablet is PC mouse and PC keyboard operated only
but sometimes, touch screen stops to generate pings to cursor for a long time and this makes me happy.
So since we don't get access to wire frame behind logics, intelligence of the autopilot (sensors connected) better live on the safer side and have
sensors disabled and disconnected (GPS, gyro ? )
Read again this thread
http://diydrones.com/forum/topics/launching-from-a-moving-boat-can-...
Launching from a moving boat vs. launching from your flexible stand
I am happy developer with Nokia meamo tablet, Google Maps, Navteq Location Based Service (non-GPS indoor navigation), 3D Navigation, TAWS in drones, Coffee&Power, Yahoo Pipes, MIT Media Lab Object Programming, 3D LCD LED Printer
and tens, hundreds of small or not small IT projects world-wide (Natural Language Processing Princeton/ China recently)
If you know programming well just switch your ardupilot into debug mode on and save traffic to sd_ram, save sensors data for off-line analysis and processing, visualization, as you already did it with some
charts coming with your video.
darius
Tethered indoor flight tests can be simulated in my wind tunnel, pushing strong downwards wind against a drone.
There is a bug, a number of bugs in ardu pilot software but ardu drone software development looks to be closed project, not open to the public and third party developers.
-
1
-
2
-
3
-
4
-
5
of 6 Next