Arducopter Crash - Help Needed - Discussions

Yesterday I had a really bad crash with my Arducopter. I'm hoping that someone can point me in the direction of what happened so it won't happen again.

I toke off in stabilize mode then shifted to altitude hold and then position hold. That all seemed to be working fine. Then in the APM flight data window I right clicked on the map making the drone go to that location. Again everything seemed fine. I noticed that APM said the quadcopter was now in guided mode which also seames ok. I did a couple of "Fly to here" with different alitudes with almost the same location. Again all was good and the quadcopter did as expected. Then I clicked on the map and to make it go maybe 50 meters away against the wind direction. And here things started to go bad. First it seamed like it didn't really get the command because it did not seem to change position very much. So I tried giving it a new location again by clicking on the map. Then it started to go it small circles, then bigger circles. At this point I realized that it was going out of control. So made it go into RTL mode from my transmitter. For about a second it seamed like it changed heading correctly but then it started a very big spiral while descending very fast. In panic I put it in loiter mode to do something. But it just continued the spiral and crashed very hard. I probably should have done some manual control, but my experience piloting a quadcopter amounts to about 10 minutes. To my great regret I found out that I had forgotten to record the flight on the mounted GoPro camera:(... The film would have been epic:).

I really hoping that someone could help me understand what happened. I have attached the log og kmz file. I would really appreciate anyone could look through the files and enlighten me:).

Sadly I think only the ESC's, Ardupilot Mega, GPS and 3DR Radio survived.

Flight.kmz

Flight.log
Here is a bit information about my setup:

Ardupilot Mega 2.5 running the latest version of APM arducopter version is 2.9.1 if I remember correctly about 3 weeks ago it was the newest.

Wind was about 10-12 kts. Maybe a bit high but I didn't have any problems controlling in stabilise mode.

Frame is BumbleBee Carbon fiber - No PID tuning done.

Temperature was about 0 degrees celsius.

Here is a screen shot from Google Earth. I removed the part of the flight where everything seams well. Everything is in the attached KML file if needed..

I searched a bit on google and found this post: http://diydrones.com/profiles/blogs/arducopter-2-0-38-flight-test-wind-effect . He seems to have the same flight paths leading to almost crashed. Fortunately he was a much better pilot and was able to recover through simple mode manual control. Could the wind alone have caused this behaviour. To me it seems that the will alone should not cause the Arducopter to spiral to the ground?

I hope you can help me out:)! Have a great day everybody!

Ok, so first off, it's not a mechanical failure. This is clear because when you plot Roll-In vs Roll and Pitch-In vs Pitch they all track well which means the copter frame was properly doing what the controller was asking it to do.

I'm pretty sure this crash is caused by compass interference and a weakness in the software.

Spirals are caused by the copter being seriously confused about which way it's heading. normally you'll just get a curved path if it's off by up to 45 degrees but beyond that you will get a spiral.

What's most likely is that you've got a lot of compass interference from the power wires below the APM, the ESCs or motors so while the motors are at a moderate level your compass interference isn't too bad, it basically holds a heading ok but once you punch the throttle up to full it becomes more than 45 degrees off and you spiral away. So as you headed into the wind and it tried harder and harder to reach the point it leaned over more and more. The "angle boost" would rev up the motor to try and maintain altitude and the yaw heading would have gotten even worse making it miss the target more and more and just spiral off.

The software weakness is that we always prioritise roll/pitch over throttle even when the roll/pitch request is too large for us to maintain altitude. This is what we need to fix to pull us out of these situations or at least stop the copter from hitting the ground.

By the way, we're planning to add a method of compensating for motor interference on the compass in 2.9.2. Until then, I guess once you've rebuilt your copter, you might want to put the APM further from any sources of interference and then perhaps tie it down and try reving up the engines to see how much it affects the headings. We can also see this very directly if you've got a tlog from the flight.

I'm really sorry to hear about your crash. :-(

Replies

Rich Gopstein March 11, 2013 at 10:02am

Check this image out... Look familiar? I posted the details to the 2.9.1 release posting.

QuadCopterCrash.jpg

https://storage.ning.com/topology/rest/1.0/file/get/3692648882?profile=original
Fabio March 3, 2013 at 4:57am

Hello Ken,
I am very sorry for your crash, so I have a Bumblebee with a APM2,
the frame is built very well, but at least for me there are resonances during turns due to flexing propellers that create vibrations that resonate with the tubular carbon, creating vibrations to the structure. These can hear acoustically sound like blowing into a bottle. I believe that with the 2.9.1 firmware can cause problems.

If you hear I found this site to find spare parts.
http://www.hobbylord.com/en/

greetings
Markus Järve February 28, 2013 at 12:42am

Compass calibration problem is pretty standard. This is exactly the flight path that comes with compass problems.

Also, The wind makes it sound like High currents to motors could have influenced the compass.

EDIT: Or exactly what Søren Kuula said.
Søren Kuula February 27, 2013 at 3:55am

Hi,

Sorry to hear about yr loss. In my Mikrokopter days we would have blamed it on the sunspots :)

Spirals seem more to be a compass problem. Say the controller finds (GPS) it is a little too far north, and the nose is to the east by the compass. The fix is to move a little right. Now if the nose is really pointing to the south, the move to the right will cause the copter to go to the west. Shortly after it is too far to the northwest, and with some more error distance than before.. next reaction is with an error direction again and worsening the result again.. Repeat that thought experiment in small enough steps, and you get a spiral.

Compass errors can be caused by magnetic fields from high current wires running nearby. That is, battery to ESC wires. I am not certain that this is really your problem but I would recommend that you check it when you have an intact copter again. Compare in different headings and with different throttle levels against a known correct heading. The props need to be on so the motors will consume some current so watch out! Tie the copter to some heavy object.

Anyway if something automatic goes bad again you have to cut it out and save the aircraft manually. RTH is just a special case of waypoint flight so if the latter stops working, so will RTH. If you can't already, you have to learn to fly the copter in any attitude, like nose-in and sideways.

Regards

Soren
Developer

Randy February 26, 2013 at 6:47am

By the way - regarding us always prioritising roll/pitch over throttle...it sounds dumb in this situation but it's the right thing to do in other situations. For example, if you lose part of a prop or have a weak motor, you're coming down no matter what so it's best to come down flat. So really what we need to add is a way to properly determine what situation we're in.

Anyway, we can sort this bit out..
Developer

Randy February 26, 2013 at 6:45am

Ok, so first off, it's not a mechanical failure. This is clear because when you plot Roll-In vs Roll and Pitch-In vs Pitch they all track well which means the copter frame was properly doing what the controller was asking it to do.

I'm pretty sure this crash is caused by compass interference and a weakness in the software.

Spirals are caused by the copter being seriously confused about which way it's heading. normally you'll just get a curved path if it's off by up to 45 degrees but beyond that you will get a spiral.

What's most likely is that you've got a lot of compass interference from the power wires below the APM, the ESCs or motors so while the motors are at a moderate level your compass interference isn't too bad, it basically holds a heading ok but once you punch the throttle up to full it becomes more than 45 degrees off and you spiral away. So as you headed into the wind and it tried harder and harder to reach the point it leaned over more and more. The "angle boost" would rev up the motor to try and maintain altitude and the yaw heading would have gotten even worse making it miss the target more and more and just spiral off.

The software weakness is that we always prioritise roll/pitch over throttle even when the roll/pitch request is too large for us to maintain altitude. This is what we need to fix to pull us out of these situations or at least stop the copter from hitting the ground.

By the way, we're planning to add a method of compensating for motor interference on the compass in 2.9.2. Until then, I guess once you've rebuilt your copter, you might want to put the APM further from any sources of interference and then perhaps tie it down and try reving up the engines to see how much it affects the headings. We can also see this very directly if you've got a tlog from the flight.

I'm really sorry to hear about your crash. :-(
Developer

Marco Robustini February 26, 2013 at 2:39am

Hi Ken, why you haven't tried to recover the "bad Guided" mode switching to Stabilize and not RTL?
Switch to RTL when the quad for some reason is not working well with the automatic control is a bad choice.
I've always recover my quad from possible "strange situation" with Stabilize or Acro mode.
As in RC helicopters, if you lose the "tail" must be groped by activating the throttle hold/autorotation.
Can you post here your parameters?
Ken Falk February 25, 2013 at 10:23pm

Just checked and I can verify that I was running 2.9.1 version of Arducopter
Mike Boland February 25, 2013 at 11:11am

The compass values are in the header of the flash log, although to my uneducated eye they look within reason.

I have found that switching to an unguided mode is the only way out of that spiral, stabilise or acro.

This spiral of death has cropped up 5 or 6 times in the past few weeks, so there is SOMETHING going on in the code, in my view, that is not obvious, but has to do with guided control.

In all cases the copter exceeds max angle and max speed and continues until it hits the ground.

I have seen no reply's from developers to crash stories outside the '2.9 released' thread.

You may have more luck posting there.
Ken Falk February 24, 2013 at 2:19pm

Chris gave me this reply which might shed some light on the issue:

"It looks like your compass wasn't calibrated yet and needed some more manual flight to learn your frame/location".

However I'm not sure if you can see in the logs that the compass wasn't calibrated or if this is common knowledge of the slightly more enlighten people:).

of 2

This reply was deleted.