Skip to main content

Unexpected Drone Crash

Comments

13 comments

  • Jim McAndrew

    Sorry to hear that you had a crash. At the end of the day, the sensors on DJI's consumer drones are not accurate. The barometric altimeter and the GPS can be off by as much as 4-5 meters in either direction.

    We created an entire user interface module to help uses who are doing inspection based missions that demand 1-2 meter accuracy, and it can be turned on in the plan settings (Growth plans only):

    Even still, this UI requires the operator to be "on top of it" and manually put in corrections for both the altitude and position in real-time as they see the errors adding up.

    1
  • Larry Fortna

    I can understand there can be deviations and such, but what occurred on both the 24th and 25th with the same plan was an unexpected drop in altitude, not just a deviation.  Yes, it could explain the near miss on the 24th compared to the crash into the bush on the 25th, but the drone dropped to 0' on both days at the same marker which was not a part of the plan.   I was at 50' and only issued a POI change at marker 2.  There should have been no change in altitude until marker 3.  Did you watch the entire video?

    https://app.dronelink.com/.../plan/CoeLdXHKu0nCQmP2zdMe

    0
  • Jim McAndrew

    Are you talking about this marker?

    0
  • Jim McAndrew

    It is hard to tell exactly from your video:

    But the drone appears to be within the previously stated position tolerance to marker 3:

    As for the altitude, given that the marker is set to 8ft, it seems like that could be anywhere from -8ft to 24ft given the +-5m accuracy of the altimeter.

    0
  • Larry Fortna

    Marker 3 was where it was supposed to have dropped from 50' to 8'.  However, the problem occured at Marker 2 which was simply a Point of Interest marker to face the drone toward the landing zone.

    On the 24th, the drone did the same thing except it was off by 1 foot in distance from the bush and did not crash into the bush.  So at Marker 2 it dropped toward 0' and the bottom sensors kicked in at 3' and lifted the drone where it continued on path to Marker 3 at 8' as intended.

    I revised the plan this morning and moved the path away from the bush and added a destination component instead of relying on markers.  I flew this plan successfully and everything worked exactly as expected.

    0
  • Larry Fortna

    If you watch the video again, go to the 3 minute mark and pay attention to the altitude of drone as it approaches marker two.  It is easy to spot on the 24th because it missed the bush by a foot.

    0
  • Raffaello Di Martino IZ0QWM

    I think it may be an altitude sensor issue.
    The altitude sensor works with the barometer but also with ultrasound if you get very close to the ground.
    The ultrasonic sensor may not have worked well above the hedge.
    In fact at minute 3:17 it is noted that despite being more or less at the same height from the ground, the sensor indicates 6.2ft instead of 3.3ft of the previous day.

    This difference made Dronelink think that it was necessary to lower the drone again and that perhaps marker 3 had not yet arrived.
    Now I don't know if Dronelink double checks between location and altitude and which one it prioritizes, but either way, it seems like an obvious sensor problem.

    On the 24th it is likely that fortunately the hedge was overcome immediately and the drone found the asphalt where the ultrasonic sensor works much better and so it made it rise again.

    Care must be taken to get too close to stepped or very rarefied structures (tall grass, hedges, accumulations of stones, rocks with spikes, etc.)
    The ultrasonic sensor works by emitting waves that are gradually wider which, if they are not well reflected, can give very high values.

    Ah I want to add that even a quick rotation on itself can give problems to sensors, such as IMU, GPS and sometimes altimeter.

    0
  • Larry Fortna

    I would be inclined to agree with you except for the fact that the drone should not have descended from 50' to 8' until it reached marker 3. 

    The actual decent came about at marker 2 which was a simple POI marker to turn the drone towards its landing zone.  It was previously on a 270 degree heading along the path.

    You will notice on the 24th, the drone descended until the altitude sensor responded and lifted the drone and allowed it to continue to marker 3 where it actually lifted to the 8' height.

    On the 25th, it would have done the same thing except the GPS was off just a bit.  

    On the 24th and 25th the airdata logs also indicate the altitude was ~26.9' at marker 2 in the sensor section, which I suspect was as it was descending.  

    All this indicates to me that dronelink misinterpreted something when negotiating the POI at marker 2.

     

    0
  • Raffaello Di Martino IZ0QWM

    It dropped to 8ft right at the hedge. This is why I asked what Dronelink gives priority to: the altitude read by the sensor or the GPS position?
    Attention, even the GPS can be wrong, so all points may have been shifted by a few meters.

    If I can give you some advice, do not let the drone descend automatically to a height of less than 30m.

    https://developer.dji.com/onboard-sdk/documentation/guides/component-guide-altitude.html

     

     

    0
  • Arizona Wyldwest

    Not sure if I’m fully understand all the issues. The only thing I do know is when you set markers to change elevation it always descends slowly before the lower set altitude so it will be at the height when it gets to it. Also may you try changing the interpolation and see if that helps. Then save the KML file and pre run the mission and look at it on Google Earth Pro. Works really well and will at least give you an idea. The altitudes are definitely not perfect as the drones altimeter isn’t perfect. As Jim said, they just aren’t that accurate. Sometimes when I take off from my driveway and then land when finished, it can indicate a plus or minus of altitude of about 3 feet sometimes more. 8 feet is fairly close to the ground and could be off as he also stated. Flying that low and near any obstacle is going to be more difficult. Hope any of this helps you. Still learning a lot when using this app which I believe is amazing for my personal purposes. 

    Good luck Larry. Enjoy.

    Mike.....

    0
  • Jim McAndrew

    The native app uses the same exact code as the mission previewer, which renders this path:

    If there was an error it is not with how the code it interpreting the markers because the preview would show that error. The only explanation I can see is sensor errors in the real world, and the only way to prove that is to fly this exact mission again and compare the results (possibly somewhere else that has less obstacles, and while being ready to take over at a moment's notice). If it performs well then you know the sensors were lying on the day of the incident. I can try flying this mission myself but it won't be until the end of next week.

    0
  • Larry Fortna

    I modified the plan to use a destination component at the end rather than depend on just a path with markers.  That seems to work just fine.  I'm not going to worry about it, but if it happens again I will follow up on it.

    0
  • Arizona Wyldwest

    Ok. Makers have several options you can change. Good luck and enjoy. 

    0

Please sign in to leave a comment.