log file attached.....running AC3.3.1 on Pix, was in loiter then guided mode just hovering at 5m for a while, all good, then sent the land command thru GCS and it just accelerated wildly to the right eventually ramming into my car and then into the ground, GPS masts broken, broken arm and motor mount, and nice gash on my car's door....
what the heck is Err: 0-44 mean? never seen that before but it happned just before I issued the land cmd, I dont see any mechanical issues and thou it appears there is some EKF issue right after the Err: 0-44 I dont see any specific EKF error
Randy or Leornard? please help! thanks
Replies
Interestingly looking at the param you have set for the copter you have for glitch protection
The defaults in the code for Copter3.3.1 are link
EKF_GLITCH_ACCEL, 150
EKF_GLITCH_RAD, 25
And the master (3.4) they have been lowered to 100 and 15 respectively.
Would be interesting to play back the recorded data and see if the new params trigger glitch detection in the EKF.
Do you have an idea why your params are set larger than the defaults?
PS: I'll have more of look later, run out of time for now!
thanks Bill, glad you are helping.
I recall changing those parameters in july I think when I had a GPS glitch and by recommendations that the larger GLITCH radius would tolerate a larger GPS discrepancy thus handling a longer glitch, and the ACCELL was default at 350 back then and it was suggested to bring it down to 250 to reduce sudden jerks in acceleration during position changes using 2 GPS (I use 2 GPS)
how do i play back the data using different parameters? that sounds very cool....I use SITL but no expert with it
I remember before EKF , when INS was used there was GPS glitch protection. If the GPS had a sudden change that couldn't be real due to outrageous accelerations that would be required, the loiter controller would just reset to the new wrong position. This was introduced to prevent this exact type of fly away condition. This was dropped when EKF was introduced. I thought the EKF was supposed to handle GPS glitches making the simple glitch protection no longer necessary ? I have seen some pretty serious GPS glitches while connected on the bench. Lucky for me I never had one in the air. I have been landing in RTL or Loiter mode mostly within ~ 10 meters of me. If GPS glitches are back I will have to rethink my landing procedure and only use Alt Hold or Stabilize.
After further research I found that glitch protection is in EKF and is more configurable. See Bill B's explination below and the EKF-Glitch params. Ivan's problem was a second GPS and a switch over at a bad time. You would think that a GPS switch would be the same thing as a glitch, but I don't know. It may not be. Maybe EKF needs to address this issue. Don't use a 2nd gps until this is fixed.
An information from discussion: http://diydrones.com/forum/topics/copter-3-3-beta-testing
may be an answer to your crash...
Warning #2: an issue has been found in Copter-3.3.1 which can lead to hard landings in RTL or AUTO if the WPNAV_SPEED_DN is set too high (i.e. >400 or 4m/s) and/or the WPNAV_ACCEL_Z is set too low (i.e. <100 or 1m/s/s). We hope to resolve this in Copter-3.3.2-rc1.
Don't use Loiter because that does use GPS, use Altitude Hold or Stabilize.
As the copter lowers to the ground the GPS signal coming from the satellite bounces off your roof line and will cause position error. If the error is towards your house the copter will correct for that by moving away from your house.
When this glitch happens it's very fast and I have seen my GPS position change by several hundred meters in less than a second. The copter will fly horizontally very fast in the opposite direction of the error.
I was only 1 meter from the ground in Stabilize in front of my garage just playing around when the battery failsafe kicked in unexpectedly and my copter went dead left into my car which was 10 meters away.
Hi Ivan,
All this happened just as you lost 3 satellites (20 to 17) so I assume your gps moved off. I can also see that the copter is reacting because it thinks it has moved, not because it is trying to move. So from the controlers perspective it was trying to stay stationary but the GPS was telling it that it was moving.
How close were you to the car when you started to land and how far did it move in total?
is this reaction a result of the 'new' EKF filters in 3.3x?, or has that always existed in the algorithm?
It is surprising to me, as I have a M8N on one of my PixHawks, and since reading your post on Thursday, I have been watching the GPS status on my bench. It does fluctuate Sat locks. . . dropping from 1 or 2 in a 5 minute interval, to an occasional drop of 3 or 4 Sats.
Will we possibly see this reaction in 'land', 'loiter', and 'RTL' modes below, say, 30' AGL?
Should we be, as suggested below, landing only in non-GPS modes?
i am not convinced is a GPS issue honestly...my M8N fluctuates between 15-20 sats in the air and it has never been an issue, I didnt see any different behaviour when this crash happened, my EKF filter is also on the "relaxed" setting, but I didnt even get an EKF error
i for one wont be landing auto for a long time until this is all sorted out!
Copter was about 25ft away from the car at the time, the copter was suppose to land about 15ft away from me when it suddenly veered off towards the vehicle for like 25ft, and I believe was still descending at the land descent rate while veering off
-
1
-
2
-
3
of 3 Next