Glitch with Active Track Sent Drone Toward Trees

[tcope]

Active Track is done with the DJI Go app software. As such, it goes by what is on the device screen. You mention that the video was breaking up. This is a very good reason why subject tracking was lost. It is possible that it got "confused" just prior to that and thought that the subject was moving in another direction, perhaps toward the trees.

DJI's software for Active Track is _not_ very good.

 

[Basspig]

In the case of loss of lock, the drone should hover, not veer off at high speed in a direction perpendicular to where it had been heading. Had a person or someone's property been in the path of that sudden change in course, I could be looking at serious legal trouble.

 

[Fly Dawg]

In the case of loss of lock, the drone should hover, not veer off at high speed in a direction perpendicular to where it had been heading.

First, take this with a grain of salt, because I am not disputing your story at all. But, unless I am looking at this data incorrectly ( And that is entirely possible, because I do not know how long the tracker continues searching after loss of lock) It appears that your hard turn was several seconds before the loss of active track. This could be normal due to latency, and other factors, but I wanted to clarify that with a second opinion. That said, according to the graph below the hard right you mentioned is 7 seconds before the loss of active track which is slightly different than what you explained, so if @sar104 has the chance, I would just like a clarification.

 

[Basspig]

That data is correct, except that I made a hard left not a right. I walked out of my driveway and into the road, making a 90 degree left turn at the intersection to the road and walked about 7 seconds down the road until it veered to the left suddenly.

 

[Fly Dawg]

About halfway down the road, the P4P suddenly veered to the left,

I don't see a severe left veer, only the hard right from the RC, which I assume is in response to that. I would think that you would see that veer in the Roll Data, there is a very fast spike to the left, but only around 300ms-to 500ms, approximately.

 

[Basspig]

Oh, yes, that's correct. I thought you were looking at the aircraft flight path, not the RC. Yes, my response was a hard right in an attempt to save the drone from crashing into the pine trees.

Here's the video from the controller:

Dropbox - 2018_06_10_13_25_21.mp4

 

[tcope]

In the case of loss of lock, the drone should hover, not veer off at high speed in a direction perpendicular to where it had been heading. Had a person or someone's property been in the path of that sudden change in course, I could be looking at serious legal trouble.

Which might be one reason why AT won't start until the drone is above a certain height. Starting with the P4, it also maintains altitude up until the VPS sensors work.

 

[Fly Dawg]

@tcope Good Point....VPS was active for the entire flight....as per below.....

 

[tcope]

Here's the video from the controller:

Dropbox - 2018_06_10_13_25_21.mp4

WOW... you one lucky SOB!

 

[Basspig]

Yeah, lucky nobody else was around and that no one was injured and no property was damaged.
Did you notice that the video is breaking up frequently? I get better video than this from 1km away when in manual flight mode!

 

[Fly Dawg]

Doing a frame by frame on the video, this could be purely coincidental, but may also be a cause of the loss of lock. In the image below, notice the landing gear are extended far in to the frame. Coincidentally or not, this was within only a few frames of when it began moving toward the trees. Also, I looked at the frame rate and unless you changed it in the edit, this was shot at 24fps. That could also have an effect on the tracking. You might try a higher frame rate and see if that improves it

. I would think that it would be more efficient at 30fps.

 

[Basspig]

I use 23.976fps for several reasons: Blu-ray authoring (BD doesn't support 30P or 60P) and my dual Xeon E5-2667 workstation struggles with anything over 24P frame rates in 4K. And finally, when you have only 100mb/s CODEC, the fewer frames you can encode, the better the quality of each frame.

I did notice that sudden yaw where the gimbal kept the shot straight but the landing gear came into the frame. I think that's where something went wrong first.

 

[Russ43Phantom]

I find processor speed (e.g. my Magic Air tracks better than my P4P), lighting level, contrast, and lack of sudden turns are all factors in successful visual lock ons.

 

[Basspig]

I haven't yet figured out what was the cause of this glitch, but I can believe insufficient processor speed could be a contributing factor.
Someone else just posted an identical incident this morning in the DJI forum. I saw the video--same thing happened, with the landing gear coming into the frame and the drone veering off to the left at high speed.

 

[pe9999]

most active tracking algorithms convert the image to BW before computing the features. I believe DJI is no exception.

 

[cneedelman]

Glitches happen.
Using automated flight modes in a tree rich environment is not a good idea.
Whenever you need to quickly cancel any automated flight, simply flick the flight mode switch to atti and back to P-GPS again and it's cancelled and you have full control.

Great tip that I wasn’t aware of. Thanks!

 

[Liveddica]

Doing a frame by frame on the video, this could be purely coincidental, but may also be a cause of the loss of lock. In the image below, notice the landing gear are extended far in to the frame. Coincidentally or not, this was within only a few frames of when it began moving toward the trees. Also, I looked at the frame rate and unless you changed it in the edit, this was shot at 24fps. That could also have an effect on the tracking. You might try a higher frame rate and see if that improves itView attachment 100068 . I would think that it would be more efficient at 30fps.

Been reading this whole post ( just to educate myself if something weird were to happen during AT) and I have to say I've learned a lot from you Fly Dawg..Solid Info. Keep up the great work!

Hope BassPig figured out the issue

 

[Basspig]

I still don't know what caused this incident. I would have to chalk it up to loss of lock, but the craft should have hovered. Why it veered left at high speed is the more disturbing aspect.