Hello Everyone-
I've had two flyaways so far because of what appears to be random gyro offset drift. During the first flyaway the drone almost crashed into a house before I was able to crash land it and in the second flyaway it crash landed on the pavement and broke two legs, propellers and the camera gimbal. I posted about the first one (including logs) here:
http://diydrones.com/forum/topics/iris-highly-erratic-behaviour-midflight?xg_source=activity
The logs for the second one are on Google drive:
In each case, I almost completely lost control of the roll/pitch. The quad listed to one side and just ran off. I crash landed the quad by completely cutting off the throttle.
When I plot the gyro offsets I can see there is a jump in the offset prior to the bad flightswhile for the good ones, the offset remains the same. I've seen other threads about this topic as well. This seems to me a very serious issue as it happens randomly and makes me very scared to fly again Has this been fully diagnosed and fixed in AC 3.3?
Replies
Ankur-
Good that you have a ticket open. 3DR will want the logs plus a few other readings from the Pixhawk that require following their explicit instructions, then you send the packet of data to them. That plot you posted looks similar to what I sent in to tech support.
In my case, I had an identical anomaly.... threads are located on this forum. Sad to say, there seem to be some Pixhawk IMU problems in a run of the units, and there have been a number of Iris crashes related to anomalies similar to yours. (defective components in the Pixhawks?)
My Iris was in warranty, and was replaced. You will likely end up replacing the Pixhawk. There are instructions on how to replace the legs... get your soldering iron ready.
As well, there seem to be heat-related anomalies that the user community is contending with.
You will get a lot of help.
I just feel bad for you feeling so scared to fly with a bird that gets a lot of flights every day and seems to have that workaround to keep the drift in check (which I agree shouldn't be necessary, also!).
I guess it's because I'm not as technical, my sympathies go out to all the uncertain and scared folks and I'm looking for a fix for the heart while the smart guys consider the most efficient change to the pre-arm check code, etc.
Hope that helps and thanks for your contribution of detailed issue reporting which should be useful for developer types. On that note, have you posted this on an arducopter / Pixhawk group?
Cheers
FWIW, here's another google group thread about this exact same issue:
https://groups.google.com/forum/#!msg/drones-discuss/fnWYM48pGys/3ohG6D96pFUJ
This is a hardware problem with the sensor on board the pixhawk. The gyro offset can suddenly drift by a large amount and since the offset is calculated before arming, the flight controller (DCM/EKF doesn't matter) is unable to cope with a large change in offset. 3DR is replacing by Iris+ that crashed because of this issue, but I don't think that it will be solved in a fundamental way. The same thing can happen again at the most inconvenient time..
Customer service reps while trying to be helpful are clueless about the true nature of the issue. 3DR is fully focused on the Solo and sort of abandoning Iris+ customers..
I'm just scared of flying the Iris again, as I know this can happen again and I don't want to crash into a house or hurt someone. Really unfortunate. Been trying the DJI Phantom 3 as it doesn't have issues with fly aways, but really missing the mission planning features possible with Pixhawk based quads..what a dilemma.
I know the Pixhawk 2 used on the Solo uses 3 IMU's, so hopefully is a bit more resilient, however I feel it's a bit unfair on me to have to shell out another 1.5K on a new copter.
Not sure what you mean by mandatory. Of course it is not mandatory to do so. However it is perfectly reasonable to fly the Iris, land it, let it sit for a few minutes without disconnecting the battery and then fly again. If it has been sitting for a few minutes, it automatically disarms. Ideally, it should check if all is good before arming again and fail to arm if something is amiss.
Additionally, who's to say this can't happen mid flight? That would be even more dangerous as the quad could be flying a mission and far from the pilot.
Thanks Dave! My Iris is only about 3 months old, so hopefully under warranty as well. I'm just surprised that this wasn't caught during flight testing of the product before release. There are a few google group threads about this topic too where users have provided explicit instructions on how to repro the issue. It appears that if you fly the Iris (so the gyro offsets have been calculated during pre-arm) and then let the Iris set with the battery plugged in, the gyro offsets can drift. This appears to happen randomly. When you fly again, it appears to use the wrong offsets and also attaches more weight to the gyro readings during sensor fusion which leads to bad behaviour. Scary as it can lead to completely uncontrolled flights..
I had a couple of missions I had scheduled for this weekend that will now have to wait.. Hoping for a speedy resolution of this issue and a proper fix so that this doesn't happen again. I really want 3DR to succeed and these flyaways are getting the company some bad press..
You seem to have a leg-up on understanding the Pixhawk...that helps. I suspect its more than one issue 3DR is dealing with on the Pixhawk. I'm fairly certain heat is a player too.
I also have a support ticket open about this, but wanted to hear the community's opinion about this issue. Thanks!