Serious bug in the Mavic's "Home Point Me" function - caused loss of drone!

[marklyn]

I understand the OP's situation. I've taken my MP out many times and practiced, practiced and practiced manual flying and other "common" maneuvers as to become better experienced. I also have used the pause button in certain situations when something was happening that I either didn't understand why or was apprehensive about the drone being too close to an object (tree, etc.). But, I do understand that in a moment your mind may need a moment to 'catch up' with the actions and you take an extra moment to understand (or panic). I've had several of those instances, some of them I had the presence of mind to pause and some of them I was trying to understand if what was happening was a momentary minor glitch or more serious.

 

[locoman]

I understand too. I don't know how may times my brain and finger are not on the same wave length. I think left and my fingers push right. I am lucky my bird is still flying.

 

[Mavic 57]

"Home point recorded - Check it on the map". I love those words and I do as it tells me, so far after 320 flights, 500+ kms, no RTH failures.

I'm not complacent and I always prepare for the unexpected. Just last week the RTH was fighting the sun as an obstruction. It was stuck at 120 metres altitude with the battery running down, when it went into auto landing, I managed to coax it home.

Flying in Norway the sun is low, so I should have turned off the front sensors. Lesson learnt.

 

[locoman]

I can relate.
Shortly after I get my MP and was doing a RTH, I kept getting the avoidance warning. it would go about 10' and then stop. Battery was going down, down, down. Then Remembered reading about flying into the setting sun. So then pointed the camera down and it continued to fly home. Didn't think of it at the time but could have turned it around and flew it home backwards.

 

[kodiak1120]

Sigh...

Folks, it's possible to keep two different thoughts in your mind at the same time:

1. There's a bug in the Mavic's software that can cause it to mishandle the "Home Point Me" function, and
2. When this (unexpectedly) happened to me, I panicked a bit, and didn't react as well as I could. I could have avoided losing my drone.

I'm really not interested in hearing more about point 2. I know this. Please don't bother responding unless you're willing to address point 1 (or unless you happened to find my drone

You didn't know that you were among much better pilots that instantly react to any situation in the best possible way so as to avoid any accidents or issues 100% of the time?

People need to lighten up on this forum. The pilot should never have been in this position in the first place. This is 100% DJI's fault, especially if it is a known issue. NEVER should the drone use the prior RTH location and NEVER should the pilot have to remember some arcane system of force-closing apps and powering down their device in between flights.

 

[Mavic 57]

I can relate.
Shortly after I get my MP and was doing a RTH, I kept getting the avoidance warning. it would go about 10' and then stop. Battery was going down, down, down. Then Remembered reading about flying into the setting sun. So then pointed the camera down and it continued to fly home. Didn't think of it at the time but could have turned it around and flew it home backwards.

Its the front sensors not the camera, turning the camera down should not help, flying backwards, I agree would, as the sensors would not be seeing the sun and not get confused.

 

[locoman]

I know that turning the camera down shouldn't have made a difference so maybe the sun set far enough by the time I got done screwing around that it wasn't effecting the sensors. there are a lot of trees between the MP and the setting sun to block the light.

 

[finlayson]

I posted twice in the DJI "Mavic" forum, but my post didn't show up (perhaps because my post included the word "bug" in the subject line - who knows?). But I started another thread - titled "Why didn't my posts show up?" - and that thread appeared OK So I posted my bug report in that thread.

I now have an open support request with DJI. I'll keep this thread updated with the outcome.

 

[Dragonfly]

There is no bug, your Mavic is back where you originally started your bike ride.
When using the follow me mode and you hit the GPS marker it takes the GPS co~ordinates from your phone and saves to the drone. When you move off the drone may and will follow the RC but RTH is now set to last position of RC GPS position when you TAPPED RC RTH. You now have ridden 500 or maybe 1k further on and you hit RTH then Mavic will go back to where you started unless you you again hit the RC Home to set and send the new GPS co~ordinates to Mavic then it will return to RC as long as you dont move. If you move, Mavic will land where you just set new RTH setting.
So for RC RTH to work you have manually keep updating your RC GPS position and saving to Mavic by tapping the RC RTH icon as you move and your phone or tablet must have GPS lock or Mavic will not know where RC is to RTH and will go to the last co~ordinates uploaded when you tapped that screen icon.

 

[finlayson]

There is no bug

Sorry, but you're completely wrong. Perhaps you're missing the fact that I did a "Home Point Me" operation immediately before doing RtH??

"Home Point Me" is intended to set the Mavic's 'home point' to where the Remote Controller (i.e., "Me") is currently located. As shown by my Flight Record (note the screen shots in my earlier posting), my iPhone - and thus my RC - knew exactly where it was. It should have sent that location to my Mavic. But it didn't. Instead it sent the location that happened to be the home point of my *previous* flight. That's unquestionably a bug.

Thank you for playing

 

[royster]

That is messed up, I;ve also moved off my original "home point recorded location, by about 1/4 mile & hit "Home point me" which I assumed meant bird will fly to my new location, on hitting RTH, I watched my monitor as my bird turned to face me, & continued turning to face the lake & took off like a rocket! I did hit pause & using my little red bird flew her home to me, I didn't think much of this at the time, but, your right, finlayson, its gotta be a bug or a could it be the GPS in my monitor is out of wack?

 

[dwallersv]

The MP should never update your current location with an previously recorded home point and make you think you recorded your current location correctly. That is the bug. Why would anyone need that?

Without entering the fray over what happened and who's at fault, just want to correct some misconceptions about how the system works.

The Home Point is sensed and stored entirely by the Mavic, with no interaction from the optional attached smart device. If you start up your MP and the RC, without a phone or tablet attached, wait for the lights to indicate GPS lock, then a minute or so more, the home point will be set. Now, fly around a bit with the RC (do the usual 20' pause to record a Precision Landing image), go out far and high enough for RTH to work (at least 20m distance, I think), and hit RTH on the RC.

The MP will come back and land within inches of take-off, just like normal.

There's a Home Point API in the DJI SDK that can be used to change the HP after it's been set, and this is how GO4 (and 3rd party apps) implement some of the dynamic HP features. However, unless you explicitly send a new HP to the MP from an app, there's no stored previous HP that gets randomly loaded by the MP itself.

 

[dwallersv]

Interesting... This suggests that the bug is probably in the "DJI GO 4" app, rather than in the Mavic's firmware. However, in my case - according to the "Flight Record" - the app had the correct GPS location when I did "Home Point Me", so it's strange that it would give an old 'home point' to the Mavic.

Actually, this sort of inconsistency is exactly the kind of thing you'd expect with a bug -- the displayed location is correct, but what gets sent to the MP is something else -- different code that wasn't changed when they recoded something else. Seen it a million times.

Not saying that's exactly what's screwed up here, just pointing out that the correct display and wrong HP in terms of what the MP thinks is actually consistent with a bug of some sort, in GO4.

 

[Robbyg]

This is a known issue that crept in one of the updates and has never been fixed. The fact that the OP is getting so much heat is really sad. Most of us need to step back and remember the first time we got our Mavics and how intimidating all the options were.

It is expected that when one uses a feature that it will work as described. To expect a pilot to have enough experience to get out of a unexpected bug induced problem is a bonus, not a required skill.


Rob

 

[finlayson]

There's a Home Point API in the DJI SDK that can be used to change the HP after it's been set, and this is how GO4 (and 3rd party apps) implement some of the dynamic HP features. However, unless you explicitly send a new HP to the MP from an app, there's no stored previous HP that gets randomly loaded by the MP itself.

Correct. I think there's general agreement here that the bug is probably in the smartphone app (DJI GO 4 on iPhone) and/or controller - not the Mavic Pro itself. The Mavic Pro is being told an incorrect location - despite the fact that the iPhone (and thus DJI GO 4) should know where it really is.

 

[finlayson]

It's still inconclusive that this is a Mavic's bug. To prove that it's a bug, it must be reproduced, with an independent GPS logging app running on the phone to see what coords are fed to Mavic by the phone OS.

Yes, and I hope that DJI does this sort of testing (in the off chance that they're following this thread

There could be a minor bug in DJI app with displaying the blue dot while the phone lost GPS lock. It's still a possibility that when Mavic was commanded "Home Point Me", it asked the phone, "what are your coords?" and the phone replied with the previous day's coords.

I don't think that the "Home Point Me" function is something that the Mavic (i.e., the aircraft) knows about; instead, it seems to be a feature of the "DJI GO 4" app. The app (via the RC) is apparently telling the Mavic the new 'home point' - but in this case this new 'home point' is incorrect. The general consensus on this thread is that the bug is probably in the DJI GO 4 app; for some reason it is choosing an old 'home point' - rather than the current position - to give to the Mavic.

Another possibility is that setting the "Home Point Me" to known previous flight's home point when the phone lost GPS is not a bug, it's a "feature" - it was intended, and the logic here could be that if the phone lost GPS (or turned off/crashed), what does Mavic do now? It has to RTH somewhere, but where? So as a safety fallback in case RTH becomes unknown, it returns to previous flight's RTH position, assuming that in majority of cases, the dronist flies in the same location. (e.g. mostly takes off from their backyard)

Wow - that's a real stretch I can imagine a fallback to the current flight's starting home point possibly being seen as being a 'feature'. But in this case we saw the previous flight's starting home point being used.

But, in any case, this is all speculation. Only DJI will know for sure what's going on. (Again, I hope they're following this thread

 

[finlayson]

// the bug is probably in the DJI GO 4 app; for some reason it is choosing an old 'home point' - rather than the current position - to give to the Mavic.

Is there an evidence that the phone reacquired sats before "Home Point Me" was selected?

Yes. Once again, note these two screen shots from the 'Flight Record', taken 1 second before, and 1 second after, I did "Home Point Me":

And was "Home Point Me" used on the previous flight?

No.

 

[kodiak1120]

No.

This picture ends the debate about bug v. pilot error, IMO. The screen shot shows "Home point set to remote controller position," but the home point shown on the map is not the location of the RC. That shows the system did not operate as intended and that this is clearly a bug.

 

[kodiak1120]

// The screen shot shows "Home point set to remote controller position," but the home point shown on the map is not the location of the RC. That shows the system did not operate as intended and that this is clearly a bug.

Not necessarily. Can the player be stepped with each recorded point, which is every 0.1s? If not, the screenshot not necessarily represents the exact data point that was recorded as new home position. I.e. if the blue dot and "H" were the same at that frame, but that frame was skipped in the player because it moves only in 1s increments, and blue dot jumped from "H" to the current position in this frame that was used for the screenshot. When GPS is lost or unreliable, the phone can spit out all kind of nonsense coords, everyone observes once in a while how maps "fly" to some old location and then back to correct location. We need to see the log that contains coords of the phone and see phone's coords at exact moment when home point was updated.

Wut?

 

[Nomand Eloc]

Its the front sensors not the camera, turning the camera down should not help, flying backwards, I agree would, as the sensors would not be seeing the sun and not get confused.

The better idea is to just flip the Mavic into sports mode. This will disable to front sensors much quicker than digging in a menu.