The drone was doing a fine job till it decided to suicide into Frosty going much faster than it was even supposed to and a lot closer than it was supposed to. Don't turn off the avoidance system like I did.

I did not know waypoints were for such fine precise control.  Impressive what was being done right up to the attack on Frosty.

TonyPHX Posted at 12-22 13:04
I did not know waypoints were for such fine precise control.  Impressive what was being done right up to the attack on Frosty.

It's surprisingly accurate but it obviously can bug out. It wasn't supposed to even get close to frosty. It was supposed to stop in the video where it speeds up for some reason. But I have a no crash video that matches the path from the crash video pretty well. It's a very nice feature they added.
Damaged the front motors and next time I'm putitng on the prop guards.

So, it's not the drone or waypoints. It IS operator error. Next tile maintain control of the drone, yes you can override waypoints.

Waypoints are only as accurate as consumer GPS.
With perfect GPS reception, that might still have precision of +/- 1-3 metres.
If GPS reception is poor because of trees, buildings etc blocking the skyview, GPS could be even less accurate.
Accuracy can also be affected by inaccurate plotting of the waypoints.
Using waypoints for flights closer than 5 metres to obstacles is asking for trouble.

seriously i think you could have flown that flight yourself and obtained the same/better results.

Mobilehomer Posted at 12-22 13:34
So, it's not the drone or waypoints. It IS operator error. Next tile maintain control of the drone, yes you can override waypoints.

Nope, it flew the path multiple times with no issue. This ran it decided it was going to speed run into Frosty. As others noted the GPS probably drifted and the drone was probalby trying to recover.

The Saint Posted at 12-22 14:07
seriously i think you could have flown that flight yourself and obtained the same/better results.

But not if you are trying to do a day to night transition nor does it let you try out new functionality.

Labroides Posted at 12-22 14:01
Waypoints are only as accurate as consumer GPS.
With perfect GPS reception, that might still have precision of +/- 1-3 metres.
If GPS reception is poor because of trees, buildings etc blocking the skyview, GPS could be even less accurate.

I'm guessing when it happened the GPS did drift and the drone was trying to correct and even with WAAS the precession isn't going to be better than 5 meters. Which is why it was not supposed to get close, nor move fast, and to give time for pause button. Which did not work or work fast enough. I hit it well before it suicided, and it kept going. When I review the flight log it also records rudder settings I never gave which also don't match the flight path.
It very likely just bugged out.

It is the classic, sometimes S&&T just happens.  

djiuser_CgYhBqmxP83j Posted at 12-22 14:29
I'm guessing when it happened the GPS did drift and the drone was trying to correct and even with WAAS the precession isn't going to be better than 5 meters. Which is why it was not supposed to get close, nor move fast, and to give time for pause button. Which did not work or work fast enough. I hit it well before it suicided, and it kept going. When I review the flight log it also records rudder settings I never gave which also don't match the flight path.
It very likely just bugged out.

I'm guessing when it happened the GPS did drift and the drone was trying to correct and even with WAAS the precession isn't going to be better than 5 meters.
No need to guess.
The recorded flight data would show the level of GPS reliability during the incident.

Which is why it was not supposed to get close, nor move fast, and to give time for pause button. Which did not work or work fast enough. I hit it well before it suicided, and it kept going.
It would be interesting to see how good/bad GPS reliability was.
With really bad GPS reliability, the drone has no brakes and can't stop.

When I review the flight log it also records rudder settings I never gave which also don't match the flight path.
That's another thing that would be interesting to see.
I haven't seen a DJI flight data from the new waypoint implementation, but with Litchi waypoint flying or other DJI automated flight modes like RTH or POI, the flight data only shows actual joystick inputs and the waypoint control inputs never show.

It very likely just bugged out.
Drones are mechanical and don't just "bug out".
There are always underlying causes for whatever they do.

Labroides Posted at 12-22 15:16
I'm guessing when it happened the GPS did drift and the drone was trying to correct and even with WAAS the precession isn't going to be better than 5 meters.
No need to guess.
The recorded flight data would show the level of GPS reliability during the incident.

The log is cut off before the crash. It record the pause (which in the video seems to be right before it his frosty)  late and then nothing about the crash when normally it does.

The GPS signal was strong (20-18 satellites at the time). It can only record a drift if it knows there is a drift and the logs do not appear to record the guessed accuracy of the gps location in the logs. So yes, it's a mystery if there was gps drift or not.

There are no rudder inputs in waypoint mode (shouldn't be but you would with say a 3rd party if SDK was out) till the end and then while flying at frosty it shows it going slightly left with rudder strafe and down altitude. I suspect airdata is having a hard time processing rudder data because none is shown in waypoint mode.

Drones are mechanical and software and waypoint mode is purely software flight. It is not without possibility there is a software bug in waypoint I managed to hit.

djiuser_CgYhBqmxP83j Posted at 12-22 15:51
The log is cut off before the crash. It record the pause (which in the video seems to be right before it his frosty)  late and then nothing about the crash when normally it does.

The GPS signal was strong (20-18 satellites at the time). It can only record a drift if it knows there is a drift and the logs do not appear to record the guessed accuracy of the gps location in the logs. So yes, it's a mystery if there was gps drift or not.

the logs do not appear to record the guessed accuracy of the gps location in the logs. So yes, it's a mystery if there was gps drift or not.
There is much more information in the recorded flight data than most people know or see.
The GPS reliability as judged by the Flight controller is assessed and recorded for each 1/10th of a second during each flight.
Post a link to the Airdata summary and I can dig into it and check.


There are no rudder inputs in waypoint mode (shouldn't be but you would with say a 3rd party if SDK was out) till the end and then while flying at frosty it shows it going slightly left with rudder strafe and down altitude.

It's normal for there to be no joystick info showing for any kind of automated flight.

If your fliight data shows joystick input, someone or something was physically moving joysticks.
Could the joysticks have been accidentally moved, maybe caught on a strap or clothing etc?

I suspect airdata is having a hard time processing rudder data because none is shown in waypoint mode.
I'm a little confused because you previously said: When I review the flight log it also records rudder settings I never gave which also don't match the flight path.
Airdata has no trouble processing rudder data, it simply shows exactly what the app recorded.

If you post a link to that Airdata summary it might help clear up some of the confusion.
Make sure you tick the share box so the actual data is available rather than just the summary.

Interesting to see.
Thanks for the tip.