Today I was test flying my new GoPro gimbal setup and on a perfectly normal autonomous take off it all went crazy.
I've done around 10 perfect auto missions in the past from take off to landing. But today as soon as the drone left the ground it was flying sideways as fast as it could, narrowly missing a lot of cars!! I put the drone into Loiter mode but it was still trying to pull sideways as hard as it could, a few seconds later it crashed into a tree :-/ Luckily with no serious damage
See this Youtube video from the gimbal of the whole flight : http://youtu.be/Bstx2x7Xtkg
I've attached the flight plan text file and the flight log from the unfortunate flight.
Does anyone know what could cause the drone to fly sideways as hard as that on a takeoff? The first command is simply a take off to 30m command.
Thanks for any help,
Replies
I had a similar issue on my hex, which was caused by a faulty ESC, on the side it flew towards.
It appeared to spin up OK on take-off, but must have been creating lower thrust than the rest, flying sideways gathering speed until it crashed and broke lots.
The APM has no mechanism to detect such problems and react - on a hex it should be able to power down the opposite motor to compensate, but no :(
did you had any flights after? I had strange situation where my quad went and did a crazy took off and hit a tree without any of my commands(in RTL for brief moment and testing failsafe). Then i thought i could have easily put the compass the wrong way but i could not figure out coz after that crash my quad was in pieces.
Second crazy crash after just after two flights, then i figure out that my IMU decided to work poorly. When i move my quad, the virtual horizon it moved slowly. This took me like a week to figure what actually cause my quad to go mad. I bought a new IMU and replaced, wala worked fine soon after that.
This was set of events and i really dont know what exactly led to my IMU failure but i know for sure that My regulator was fine and apm did not give out any errors or warnings.
I hope you find what caused it because i know the feeling of nearly missing somebody else property.
Hi Hansaka,
Yes, I did two perfectly normal flights after fixing a couple of mounts that were damaged in the crash.
I avoided autonomous take off's mind. I took off in Stabilise mode then switched to Auto while airborne. This worked perfectly.
I do have Learning set for the compass, I don't know if it's possible that the compass got temporarily messed up by this.
Well, I think you had an accurate GPS lock right from the start, judging from some of the images in the video and I think where you were standing. Look for GPS messages that have a 3 in the first columns, which indicate a 3D lock.
One possibility is having the compass set up incorrectly. Meaning, it doesn't actually point north. In that case though, I would have expected the quad to start drifting slowly and increasingly gain speed as it got further away from the reference point. That didn't really happen here, so not likely. But good thing to check. Is your compass reading good?
Do you have a geofence set up? If a geofence is set up, it could cause a reaction like this if it thinks home is elsewhere. so you need to figure out and remember how home was set and where you were at the time. Such events should be logged and traceable in the log though as "out of geofence" or whatever it's called.
Prior to take off when CURR was 0 (no current yet to motors so no physical takeoff), if the last 3 readings from IMU lines are accelerator readings, then you have acceleration x/y readings going beyond 1 as if there were an acceleration that way. Could be a vibrational issue. If that happens, then the down-vector of the IMU gets calculated all wrong and it thinks it's tilted to one side but in fact is just vertical still. The IMU loop takes precedence over all, so you should expect a pretty steady flyaway to one side, which I think is what I say.
Steps to take:
Thanks Gerard,
I've just had a look through the logs, and the X accel immediately before take off is vibrating less than 1 centred around 0, however Y accel is vibrating the same amount but centred around -1. Come to think of it the drone was on a slight slope, but this has never caused any problems taking off in Stabalize mode!
Here's a screen shot of the logs showing the values you talked about:
Many thanks,
Pete
Log ss on takeoff.gif
Ok, we're just ruling things out here. One thing I noticed is that as soon as the AP finishes the parameter communication is that it turns into AUTO mode, which means you were already into AUTO when it was turned on? From an operational perspective, I really recommend having your tx switch to STABILIZED (so no LOITER) when on the ground. Then as a pre-flight check I usually take off for 5 seconds to see if the IMU is ok and land before the flight commences. This rules out issues with motors and IMU and is very important.
Other than that, running out of ideas here. It's late. I hope someone else has fresher ideas.