Edited: Modified title as the direction of this thread is trending towards general concern around unexpected autonomous responses of the Mavic under different scenarios and how the lack of ATTI mode may impede the pilot's ability to regain control.

Edited:  Originally thought this looked like firmware.  Community corrections to my hypothesis follow below.

Just found this video on youtube.  Seems like a firmware bug - when the drone descends below its initial takeoff point, a negative altitude causes an unsigned integer overrun making it look like instead of being negative it is now above the maximum allowed altitude, causing the drone to force a descent.

Great example of possible gotchas of being the first to receive a new model, assuming my interpretation of events is correct.

He is under a balcony without proper GPS signal.  This situation is not related to a firmware problem, but is 100% user error.

Uhh, no, more importantly, he is on a balcony where the RTH "ground floor" is set to 0, and he is flying it down below the balcony. It's not a bug, it is most definitely user error.

Thank you for clarifying.  I did notice that the flight mode was bouncing between GPS and non GPS modes.

So what if I did the same thing inside my house.  No GPS, take off from top of second story stairs and start to fly down the stairs.  Will the max altitude failsafe be triggered in that case?  

Should this be considered pilot error in this case?  

Is it NEVER safe to fly below your takeoff point without having first established GPS?

With no ability to manually select ATTI mode in the Mavic, how could the pilot recover from this error?

qMvQTfjUrm4z Posted at 2016-10-19 05:16
Thank you for clarifying.  I did notice that the flight mode was bouncing between GPS and non GPS mo ...

Yes, pilot error.
You should wait until you have GPS lock to take off. He was probably in beginner mode and the altitude is limited.
If you are not in beginner mode then you would not have an altitude problem
Also, if you ever do fly from a balcony and get GPS lock, make sure your RTH is set accordingly otherwise if it goes into RTH mode and you don't cancel it when signal is regained you may have an accident.Also, you could edit the thread title as it's not a firmware issue.

Re-reading the replies so far, I guess I still don't understand fully how the lack of GPS signal contributes to this issue.

Why does the Mavic believe it has exceeded the maximum altitude?  The video does not seem to suggest that the drone is more than 400 feet off the ground.

Not trying to be argumentative - newbie trying to understand the aircraft behavior so I don't repeat mistakes like the one above.

Beginner Mode!  Now that makes sense, DJI Ken

When DJI hands these new products out for review they should at least give them to people who know what there doing.

Chances are it says Max Altitude because they just don't have an equivalent programmed in for the notification strings for Min Altitude. Also, in a way, Min Altitude wouldn't make sense. Best message might be "Exceeded Altitude Limitations" to cover for both directions, but that would be too long for the screen perhaps.

It should not be flown in a negative altitude from starting altitude unless the initial RTH has a nice GPS lock to properly set the starting altitude. It would seem to me that this guy did not have a proper GPS lock before initially flying. Also, did it get calibrated? And lastly, is there magnetic interference for compass and GPS due to excessive concrete in the area?

I do certainly agree that not having a manual ATTI mode does bring some concerns, especially for situations like this where the Mavic could get confused from improper calibration or other issues.

I dont understand why beginnermode would lower the quad, it is because beginnermode relies on vision if no GPS and it forces its way down till it sees "the floor"?

I was also second thinking beginner mode.  Is 30m the default max height?  If so, it doesn't look like this was violated, although it's hard to tell for sure (we know it's more than 21m, but looks like at full descent, he's only 3-4 m above the ground, which would be less than 30m at balcony height.

It wouldn't.  I hesitated replying, but after reading TECHNICAL SUPPORT KEN's posting, I'm a bit worried.  

Doesn't the VPS system have a maximum altitude?  

Beginner mode allows for a higher altitude that he EVER achieved even after 7 GPS locks & "in-flight (GPS)" in this flight.

GPS acquisition clearly takes forever.  This is entirely possible with smaller GPS transponder(s).  

Even after GPS lock & the green bar with the in-flight mode states that, it then drifts.  Yes, yes - I see where it goes from 6 to 7 to 6 to 7 sats & accordingly kicks down to ATTI.  BUT, prior to calling the flight quits, it's "GPS" & still drifting.  

Perhaps watch the video 4 or 5 times like I did to get every last bit of analysis out of the video since that's all we have to go by.

I'VE BEEN MEANING TO RESOLVE THIS:
Please describe, in detail and to an adult, how a Phantom 4 or Mavic (something with VPS) would perform transitioning from inside flight to outside AND THEN back to inside flight.  Except use a GARAGE as an inside starting point (so we're on ground level).  The Mavic is of even greater concern since it lacks manual reversion (ATTI) for weird situations.

Nees Posted at 2016-10-19 06:12
I dont understand why beginnermode would lower the quad, it is because beginnermode relies on vision ...

That was just my guess, first and foremost you should always take off in GPS mode.
Maybe also, with no GPS lock when he flew off the balcony it descened until it's in VPS range
If the user synced his flight records and contacted DJI then we may know more of exactly what happened,
Maybe also, with no GPS lock when he flew off the balcony it descened until it's in VPS range

Forgot to add in that an indicated NEGATIVE ALTITUDE is normal...  It is not indicating MSL or AGL, but ALTITUDE FROM T/O POINT.  I've seen this numerous times when taking off from a mountain.

Note that the Mavic still has 5 meters to the sidewalk in the OP video thumbnail.



ALSO, I'm thinking this could happen to a TON of people.  You're on vacation in a hotel & you've got a tiny, awesome little 4k aerial camera rig...  C'mon, you know you were gonna fly this thing right out your hotel room prior to seeing THIS VIDEO!!!  


So I suppose this would NOT be suitable for inspecting elevator shafts, flying indoors in a LARGE/TALL building, or something with split stairwells (basically everything but a house)?

But it clearly looks like an uncontrolled descent.  The pilot is moving the throttle up and down and the drone is maintaining a constant descent and indicating max altitude exceeded.

Oh I'm not arguing that - I fully agree that this is not the correct action from the 24 computing cores inside the vehicle.  It "should" stay put in space upon instantly realizing it's too high to SEE (for calculating "CLEARANCE" as indicated on the RC). Why does it stop at 4.9 meters AGL??

So what's gonna happen when it flies over a giant sinkhole (for example) AND has horrible GPS lock suddenly/intermittently???  Would switching into sport mode help remedy the situation?  I'm genuinely asking this.

DJI-Ken Posted at 2016-10-19 08:25
Yes, pilot error.
You should wait until you have GPS lock to take off. He was probably in beginner  ...

Well, I have seen other flights from people indoors without any GPS lock.
How can this be a pilot error?

stratosHD Posted at 2016-10-18 18:26
Well, I have seen other flights from people indoors without any GPS lock.
How can this be a pilot  ...

See page 49 of the user manual:
"Height is restricted to 26 feet (8 meters)when the GPS signal is weak and Downward Vision System is activated. Height is restricted to 164 feet (50 meters) when the GPS signal is weak and Downward Vision System is inactivated."

stratosHD Posted at 2016-10-19 07:26
Well, I have seen other flights from people indoors without any GPS lock.
How can this be a pilot  ...

But indoors, you are within VPS range, flying off a balcony without GPS maybe caused it to descend to VPS range.

DarkEnergy Posted at 2016-10-19 07:36
See page 49 of the user manual:
"Height is restricted to 26 feet (8 meters)when the GPS signal is ...

Now THIS reply makes sense to me, and is also consistent with pilot error.  

So the lesson here is "don't fly off a balcony with low GPS and with VPS activated"

Also, "There are multiple altitude caps - FAA, geo fence, beginner mode, weak GPS with VPS, weak GPS no VPS.  Seems like the altitude warning and forced descent could be triggered for any / all of them".

Here's an example where a table of all the override scenarios would be highly useful for pilots.  

One still thing still nags.  Why did it seemingly descend below 8m?

DarkEnergy Posted at 2016-10-19 01:36
See page 49 of the user manual:
"Height is restricted to 26 feet (8 meters)when the GPS signal is ...

Wow wow wow! I do not agree on that. 8m or 50m compared to what? Till it notices something is below it? A tree? A building, a car? So when your drone is up somewhere and gets bad GPS, it will start to drop without control? I can fly a miniquad indoors at fullspeed. I sure want to be able to control it when it looses gps. What if an obstable is in front of it after it started lowering, how will you ever be able to get it back to you, if GPS never comes back, because you can't make it to go up anymore to get over the obstacle... Good bye drone?

DRONE-flies-YOU Posted at 2016-10-19 10:10
Oh I'm not arguing that - I fully agree that this is not the correct action from the 24 computing co ...

Sports mode will save it in the sink hole provided you act quickly, other wise it drop down until it finds 5m above ground level and you may lose control connection.

Height is restricted to 164 feet (50 meters) when the GPS signal is weak and Downward Vision System is inactivated."

Now the question is... is the 50m in sport mode relative to the ground or take off point?? I assume take off point as VPS is off and cannot measure height. If its take off point, then you should be safe to get out of the 200m deep sink hole.

DarkEnergy Posted at 2016-10-18 16:36
See page 49 of the user manual:
"Height is restricted to 26 feet (8 meters)when the GPS signal is ...

Woah. That's scary if that info from the manual is correct.  There is so much incorrect/misinfo out there I think i'll just have to wait for it to show up, but if one of the DJI-* people has an answer, that'd be cool too

On the other models, when GPS cut out or there was a problem, i'd just flip in atti and carry on.  So does that mean on a mavic, if you lose GPS, it's going to drop ether to 8 or 50 meters depending if VPS is on?  For example, sometimes we fly around a cliff/rock face and fly up and check out all the little nooks and stuff on it.  The GPS  can screw up when we get close on the P3p because the cliff is so high, i'll usually just either 1) back away until gps comes back or 2) go into att, and hold it still to keep filming.  

So, if GPS screws up now, there is no "ATT" equivalent? it'll just start descending?  I'm wrong right? right?

DING DING DING!!  TELL HIM WHAT HE'S WON, BOB!   

That's what I'm getting at EXACTLY.  You 'could' quickly go through the menus & disable VPS right quick to get AT LEAST 50 meters back out of the thing (in video scenario). But yes, that P-S-A switch on P4 is what Mavic NEEDS.  I suppose that extra 'click' to A mode wasn't in the budget, hehe.

BTW, the guy in the video has his sticks setup in MODE 1.

Also, look at the Mavic status light at the very beginning of the video.  I would say that is blinking fast green.  I think there's still a typo in the manual, as it mentions "Green Flashes Slowly" and "Green Flashes Twice" with the same descriptive text, but it seems to imply that both are GPS modes.  "Yellow Flashes Slowly" is supposed to indicate no GPS.  So why is it showing contradictory information between the status light (GPS safe to fly) and the controller (Opti mode - not enough GPS)?

Granted, we can't make too many assumptions based on a not officially released manual, but once again, a lot of variables in play that make it hard for me to chalk this one up as 100% pilot error.

qMvQTfjUrm4z Posted at 2016-10-18 18:23
Also, look at the Mavic status light at the very beginning of the video.  I would say that is blinki ...

Look at page 48 of Phantom 4 user manual.
AND I absolutely loathe how the information on page 12 is not correct AND they've copy/pasted this into the Mavic manual.

So I guess I'll close out my thoughts on this with the following:

1) Ability for pilot to manually take control (ATTI) is a critical missing feature

2) The (as yet to be released, copy/pasted from Phantom 4) manual has an error in it.  The maximum altitude for non GPS with vision system enabled is 5m, NOT 8m.  All behavior seen in the video (with the exception of the conflict between the status light and the controller status) can be explained by this assumption (I think).

Hang on, let me get this straight..

If it loses GPS signal it will descend, over-riding your control, until it can see the ground?

And somebody thought this was a good idea?

I'm trying to think of any scenario where descending would increase the chance of getting a better GPS signal...? Nope.

And clearly there is some kind of bug when going below the take-off point triggers a warning of exceeding the max altitude. At 2.24 it is very clearly shows MAX altitude has been reached, not minimum. I don't think that can be explained away as anything other than a bug, and one that needs fixing ASAP.

It's not the end of the world; such bugs are expected and it's why I haven't ordered mine yet, to let such things be discovered and resolved. It may also be why there is a delay shipping them, if such a basic error has been discovered (and it has, because it's right there, on Youtube).

Lets all take a deep breath and give them time to fix this stuff, rather than shipping with such a glaring bug.

   Maybe of Equal and just as big a concern is,  flying from balconies anyway, even if you were to get a solid GPS lock is asking for disaster, with a machine with NO manual mode,  and so much concrete and steel around, as no matter where you fly after taking off, if the craft goes into failsafe mode,  surely it will climb to its predetermined height over take off, fly back to its take off point and descend,   now surely this puts the craft straight into the wall of the building 4 floors up !  unless you can regain control somewhere before and the VPS cuts in ?

i am just thinking of what the electronics will do in these situations?
Also the drifting around has to be expected with such a low GPS count and buildings near by,  thats standard for all my non DJI crafts, i have seen vertical drifting and horizontal drifts,   by many metres.

I am also concerned if this happens over water, where the VPS has issues anyway right?  You know & it's got that $#@! moisture sensor that I'm going to have anxiety about hahaha!!

We've established launching from a balcony isn't the BEST idea.  But the "what if" scenarios are many in this video besides the whole balcony thing.

Agreed that hopefully these situations are analyzed, as that is USUALLY why TEST units are TESTED.  Meanwhile, I'll limit my risk by keeping good situational awareness of structures and obstacles that may lower my GPS view.  And DO NOT start a flight without good GPS locks.  I believe this video really showed all of the worst case scenarios in one.


Hmmm, and good point Mike on the low GPS count & hover accuracy.  That's 1 concern possibly explained lol.

bigglyguy Posted at 2016-10-19 01:46
Hang on, let me get this straight..

If it loses GPS signal it will descend, over-riding your contro ...

It does make sense in a way. If you lose GPS, the only thing that will keep it steady is the VPS. VPS only works up to a specific height.  
Imagine you are at 100 meters and the GPS fails and you also have strong winds. This is what everyone will call a flyaway since it will just drift with the wind.  
If it descends until VPS works again, at least it will only drift until it properly sees the ground.

Whether or not this is a good solution is another question, but at least it can be justified.

The Mavic in the video is engineering simple with bata firmware.
Mass production won't have this issue.Please don't worry about it.

The behaviour does seem to make sense from an enforcement point of view.

Imagine that if someone disables GPS intentionally within NFZ he would would be limited to 8m, and not be able to do too much damage. Wouldn't be surprised if there is a maximum distance from home point either.

If that person is flying indoors, 8m should be plenty.

taneks@gmail.co Posted at 2016-10-19 04:42
The behaviour does seem to make sense from an enforcement point of view.

Imagine that if someone di ...

There actually isn't one for the maximum distance.

JulianX4 Posted at 2016-10-19 15:15
It does make sense in a way. If you lose GPS, the only thing that will keep it steady is the VPS.  ...

I could buy that, except when I consider the situation where it would happen, such as with tree cover, going down to explore a valley and other such scenarios, it would mean even greater loss of control and signal.

"Hey, I've lost GPS signal, so I'm going to descend into this valley, in among the trees and stuff, maybe you can find me later?", doesn't strike me as at all useful?

Just to be clear, we're talking GPS signal, not control signal, BUT IT OVER-RIDES YOUR CONTROL SIGNAL AND DESCENDS, that's the bit I have a real problem with.

An urgent error message, about having lost GPS positioning, asking "Do you wish to descend for visual positioning?", fine. I can decide if that's appropriate or not.

But over-riding my control and descending anyway, out of sight and losing control and video signal, where it could be 200 or 300 meters away? A kilometer away?

Then just running out of battery and landing wherever it happens to be at the time, in the trees, in a river, on someone's roof? That's nutz.

The video is in Chinese, so I'm not sure how the pilot managed to get it to come back up again? But my impression with a quick view earlier was that the Mavic was ignoring his control stick commands and descending against his will.

Edit:

The Mavic in the video is engineering simple with bata firmware. Mass production won't have this issue.Please don't worry about it.

Ah, OK, that's what I wanted to hear

Really reconsidering getting the mavic after learning about no manual ATTI mode. I like to have full control if th drone, and the mavic does not give you that. I have personally seen where ATTI mode outdoors in high winds was the only thing I had saved someone's drone from crashing. Really hope DJI changes this, otherwise I'll have to get the phantom 4.

wait there's no atti mode ? how are we supposed to recover from flyaways then ?

DJI Mindy Posted at 2016-10-19 00:54
The Mavic in the video is engineering simple with bata firmware.
Mass production won't have this iss ...

Agreed Mindy, that it may be unfair to draw conclusions about the firmware quality or behavior prior to final release of the drone and it's associated final documentation.

BUT - that is why I re-titled this thread.  In my opinion (which seems to be shared by others on this thread), the real issue under discussion is lack of manual ATTI mode.  Without it, the pilot is severely limited in their ability to recover from issues - whether they be due to pilot error, to firmware bugs, to lack of GPS, magnetic interference, signal interference, or other operating conditions, or to partial hardware failure, ALL of which are known to happen not infrequently.

qMvQTfjUrm4z Posted at 2016-10-20 00:46
Agreed Mindy, that it may be unfair to draw conclusions about the firmware quality or behavior pri ...

It's been requested to have manual ATTI mode.