Follow Me mode fly-aways

Joined
Jun 24, 2015
Messages
1,478
Reaction score
566
Age
75
Location
Arroyo Grande, California
Lately there have been reported incidents of a fly-away occurring in FollowMe mode. I'm starting this thread to serve as a focal point to talk about these incidents.

You may want to look at the analysis of these FollowMe mode incidents.
Aircraft lost control in follow me mode
Testing Follow me-mode, fly away and crash :( and the next 20 or 30 posts
{God is Good, all the time}"follow me" is garbage! (dji is not)

The results of @Balabonov 's flight will be in the next post.

There appears to be two causes of these incidents; 1) a software glitch that presents when FollowMe mode is intiated, and 2) the tablet's GPS is not quite precise enough sometimes. Another "cause" is that maybe DJI and/or the forum has not done a good enough job of explaing how FollowMe mode works and the pitfalls. Disclaimer: I have not personally used FollowMe mode.

This thread is devoted to the technical analysis of FollowMe mode fly-aways. To do that we need the participation of members who have experienced this type of incident. This can be made difficult if these members are worried about being criticized for mistakes, either real or perceived. So, please, no intimidating, threatening or otherwise judgmental remarks about a pilot's skill, judgement, or intentions.
 
@Balabonov

Before I get into the details the summary is that what you experienced was "normal" behavior. But, I question 1) does DJI provide a good enough explanation of how Follow Me works and the pitfalls, and 2) should Follow Me be offered at all given the inherent limitations of the tablet's GPS.

I've not personally used FollowMe mode.

From your description and the .DAT the incident occurred at time 692.2
upload_2016-3-31_8-7-8.png

The P3 had been placed in FollowMe mode and the tablet started moving towards the P3 as indicated by the change in tablet longitude. The P3 start moving and reached a velocity of 9 M/sec. There was a acceleration event at time 697 indicating it had crashed into the tree.

This annotated Google Earth image shows the locations of the P3 and the tablet.
capture-20160331-061244.jpg

The red line is the P3 and the green line is the tablet. When the tablet started to move south the P3 attempted to move to a location that would allow it to follow the tablet. Is this how FollowMe mode is supposed to work? I analyzed another FollowMe mode incident where the P3 maneuvered to follow the tablet.
 
Last edited:
Interesting. This isn't how I would expect Follow me to work. With the original system it would follow the controller, not altogether very well, but at least it did it.

Here's a test that I did while using a bike and an empty (almost) industrial estate one evening last year with Follow Me when it first came out.

Not sure how the new system works in comparison to the old one. I would expect it to recognise the co-ordinates of the tablet, but also the orientation of it as well. From the way the old system works and comparing it to @BudWalker analysis, it looks like the tablet only moved a short distance while the aircraft moved quite a long way away.

Unless the craft was trying to get behind the tablet, I don't know how this could have happened.

Anyway, here's my experience with the original version of follow me - which incidentally, I haven't used since because I thought it was very limited.

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.
 
  • Like
Reactions: pomonabill220
Interesting. This isn't how I would expect Follow me to work. With the original system it would follow the controller, not altogether very well, but at least it did it.

Here's a test that I did while using a bike and an empty (almost) industrial estate one evening last year with Follow Me when it first came out.

Not sure how the new system works in comparison to the old one. I would expect it to recognise the co-ordinates of the tablet, but also the orientation of it as well. From the way the old system works and comparing it to @BudWalker analysis, it looks like the tablet only moved a short distance while the aircraft moved quite a long way away.

Unless the craft was trying to get behind the tablet, I don't know how this could have happened.

Anyway, here's my experience with the original version of follow me - which incidentally, I haven't used since because I thought it was very limited.

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.
Good video. It would seem that if you cant trust the system to work on a consistent basis, it's worthless. I'm surprised the quad missed all the light poles. Now your test is in an area that I would not have tried because of all the obstacles. Have you tried it out in the open?
 
Good video. It would seem that if you cant trust the system to work on a consistent basis, it's worthless. I'm surprised the quad missed all the light poles. Now your test is in an area that I would not have tried because of all the obstacles. Have you tried it out in the open?

Just running around in a field. This was done in September last year when the new modes came along with 1.5 I think? I thought it would be good to follow a mountain bike along a track (we have some world class MTB facilities around where I live) and so this was the test. But it didn't work so I never used it again.

I haven't tried the new system on 1.7.
 
Unless the craft was trying to get behind the tablet, I don't know how this could have happened.
My take exactly. It's called Follow Me, so it would make sense that it wouldn't move away from anyone as they walked towards it. I don't think it's designed to stay a certain distance from you in any direction, so it must try to move behind you so it can continue to follow you. If so, that sure seems dangerous since it doesn't really know what is in the way if it tries to move behind you. I think I'll test this the next time I go out flying, unless someone already knows with certainty.
 
My take exactly. It's called Follow Me, so it would make sense that it wouldn't move away from anyone as they walked towards it. I don't think it's designed to stay a certain distance from you in any direction, so it must try to move behind you so it can continue to follow you. If so, that sure seems dangerous since it doesn't really know what is in the way if it tries to move behind you. I think I'll test this the next time I go out flying, unless someone already knows with certainty.

Aha! I don't agree!

If you look at the video above, the phantom will try and maintain position even if the user is moving forward towards the unit. For my use, the feature was no good as the calculations couldn't be performed fast enough for the unit to keep up.However, if I had been moving away from the phantom, I think the result would have been the same, but the field of view would be increased as the object (me) would still be in the background. Does that make sense?
 
  • Like
Reactions: pomonabill220
Some of these incidents occurred when the tablet moved towards the P3. I'm taking a look to see if the P3 orientation has any effect on the P3's decision to follow by staying in front or maneuvering to follow from behind.

@Balabonov stated that the P3 turned away from the tablet after initiating FollowMe. But, then the P3 turned back towards the tablet before the fly-away. Here is the plot that shows that
upload_2016-4-1_7-57-41.png

The P3 to tablet bearing was about 300 degrees until the fly-away occurred. Initially the P3 was facing the tablet. Then at 682.5 it started turning away from the tablet. At 686.5 it started turning back towards the tablet and was facing the tablet when the change in position (and fly-away) started at 688.7.

This is what you'd want from FollowMe. I.e., the P3 determines the bearing to the tablet and then faces that direction. Then, you'd expect the P3 to follow by staying in that relative position while facing the tablet. This isn't what happened in @Balabonov 's flight.

@happydays is there any chance you have the .DAT from that flight you showed us?
 
Some of these incidents occurred when the tablet moved towards the P3. I'm taking a look to see if the P3 orientation has any effect on the P3's decision to follow by staying in front or maneuvering to follow from behind.

@Balabonov stated that the P3 turned away from the tablet after initiating FollowMe. But, then the P3 turned back towards the tablet before the fly-away. Here is the plot that shows that
View attachment 49149
The P3 to tablet bearing was about 300 degrees until the fly-away occurred. Initially the P3 was facing the tablet. Then at 682.5 it started turning away from the tablet. At 686.5 it started turning back towards the tablet and was facing the tablet when the change in position (and fly-away) started at 688.7.

This is what you'd want from FollowMe. I.e., the P3 determines the bearing to the tablet and then faces that direction. Then, you'd expect the P3 to follow by staying in that relative position while facing the tablet. This isn't what happened in @Balabonov 's flight.

@happydays is there any chance you have the .DAT from that flight you showed us?
Don't think so @BudWalker it was last September, but I'll look at the Dat files that I have downloaded.
 
@Balabonov

Before I get into the details the summary is that what you experienced was "normal" behavior. But, I question 1) does DJI provide a good enough explanation of how Follow Me works and the pitfalls, and 2) should Follow Me be offered at all given the inherent limitations of the tablet's GPS.

I've not personally used FollowMe mode.

From your description and the .DAT the incident occurred at time 692.2
View attachment 49058
The P3 had been placed in FollowMe mode and the tablet started moving towards the P3 as indicated by the change in tablet longitude. The P3 start moving and reached a velocity of 9 M/sec. There was a acceleration event at time 697 indicating it had crashed into the tree.

This annotated Google Earth image shows the locations of the P3 and the tablet.
View attachment 49059
The red line is the P3 and the green line is the tablet. When the tablet started to move south the P3 attempted to move to a location that would allow it to follow the tablet. Is this how FollowMe mode is supposed to work? I analyzed another FollowMe mode incident where the P3 maneuvered to follow the tablet.

Is that a corrugated steel roof? If so, you could have gotten what's called multi-path GPS error. What would be happening is that the tablet is getting confused on where it's at (in relation to the satellites) because some of the sat signals are bouncing off the roof.

S
 
  • Like
Reactions: pomonabill220
Is that a corrugated steel roof? If so, you could have gotten what's called multi-path GPS error. What would be happening is that the tablet is getting confused on where it's at (in relation to the satellites) because some of the sat signals are bouncing off the roof.

S
The analysis is based on the GPS reported location, not the actual location. I.e., any GPS errors are not relevant since the behavior doesn't seem correct based on what the P3 thinks is the location of both the P3 and tablet.
 
The analysis is based on the GPS reported location, not the actual location. I.e., any GPS errors are not relevant since the behavior doesn't seem correct based on what the P3 thinks is the location of both the P3 and tablet.
Bud:

You're popping way more synapses on this problem than I am! o_O (thumbsup)

Sagebrush
 
Aha! I don't agree!

If you look at the video above, the phantom will try and maintain position even if the user is moving forward towards the unit. For my use, the feature was no good as the calculations couldn't be performed fast enough for the unit to keep up.However, if I had been moving away from the phantom, I think the result would have been the same, but the field of view would be increased as the object (me) would still be in the background. Does that make sense?
I agree with you, on my first two Follow Me tests (during the same flight) I walked towards the P3A and it behaved as I supposed it should do, i e it backed as I walked towards it, the camera always facing me. However, on the third flight it just flow away when I initiated Follow Me.
 
  • Like
Reactions: happydays
Lately there have been reported incidents of a fly-away occurring in FollowMe mode. I'm starting this thread to serve as a focal point to talk about these incidents.

You may want to look at the analysis of these FollowMe mode incidents.
Aircraft lost control in follow me mode
Testing Follow me-mode, fly away and crash :( and the next 20 or 30 posts
{God is Good, all the time}"follow me" is garbage! (dji is not)

The results of @Balabonov 's flight will be in the next post.

There appears to be two causes of these incidents; 1) a software glitch that presents when FollowMe mode is intiated, and 2) the tablet's GPS is not quite precise enough sometimes. Another "cause" is that maybe DJI and/or the forum has not done a good enough job of explaing how FollowMe mode works and the pitfalls. Disclaimer: I have not personally used FollowMe mode.

This thread is devoted to the technical analysis of FollowMe mode fly-aways. To do that we need the participation of members who have experienced this type of incident. This can be made difficult if these members are worried about being criticized for mistakes, either real or perceived. So, please, no intimidating, threatening or otherwise judgmental remarks about a pilot's skill, judgement, or intentions.
I will support this with as much information I can from my fly-away accident when initiating Follow Me mode. Just tell me if you need more.

My P3A was sent to DJI in Europe two weeks ago. I've had some mail contact with them, they have received my diagrams and they have mailed me that they are analysing the incident. They have the .DAT-file and the Flight Record so I believe they have everything they need. I will keep you informed. Of course I hope they will take it on warranty.
 
To add to the list, cdnstudman, a popular Phantom youtuber, crashed and lost his phantom due to a follow me mode flyway.

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.
 
I believe the DJI Go "Follow me" is very limited and will only keep you in frame if you are walking.
Autopilot/Airspace is much more designed for "follow me"

I totally agree. I would never use the DJI's or even Lichi's follow mode. I have used Autopilot FOLLOW mode with an Airspace object many times (actually 53 Follow missions) without any problems (iPhone 6 attached to a working Search Dog - probably one of the most challenging Follow applications I can think of especially because I don't want the quad to be too far away - 30' high and 50' distance with position absolute 180 degrees). Occasionally you see minor unexpected corrective movements due to variations/errors in the GPS positioning. However I found Autopilot to be extremely "smart" in recognizing unrealistic changes in both location and altitude. I have grown so comfortable with the Follow mode in Autopilot that I leave the RC unattended with the Quad following me for the whole K-9 search event (typically 15+ minutes covering 20+ acres of terrain).
 
  • Like
Reactions: Tenly
that's the last time i use follow me


Sent from my iPhone using PhantomPilots mobile app
 
To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.
I've tried the follow me mode a couple of times. This video shows the best result. I sped the playback up to 2X in Movie Maker, since in the follow me mode the quad only flies about walking speed. Another time, I did almost have a fly away when trying out a new tablet. I quickly switched back to GPS and it went into hover. I was in a parking lot near the store and hadn't calibrated the compass, which might have caused it. I tried it again without calibrating and it followed me.
 
  • Like
Reactions: Romans109

Recent Posts

Members online

Forum statistics

Threads
143,066
Messages
1,467,352
Members
104,933
Latest member
mactechnic