Dear fellow Mavic-pilots and DJI-forum-representatives!

After two months of training to understand and to handle all the functions Mavic PRO offers as correct as possible (still no in depths-manual for DJI GO !) and to know her limits I forced my Mavic into ATTI-mode yesterday indoors since due to heavy fog outside one of the rooms covered with reflective uniform black tiles went dark enough to allow a change into ATTI-mode. (I was passing this room for indoor - training multiple times in TRIPOD-mode but was always lit enough to leave OPTI-mode engaged).

What surprised me and caught me unaware (and led to the destruction of one propeller :-) ) was the sudden and intense veer/drift of my Mavic  when entering ATTI-mode in/from a stand-still hovering position out of OPTI-mode. (It hit a curtain before I could bring it fully down to ground so it dropped about 20 cm without any damage but one prop. I am still waiting for my prop-guards to be delivered for further ATTI-mode learning but foolishly couldn´t await yesterday to test)

As far as I understand ATTI-mode it means that GPS/OA/VPS (= all "position-stabilizing-aids" of GPS/OPTI-mode are OFF) and Mavic is prone to shift/veer/drift by wind and keeps vertical position/flight-level just by barometric-input.

So the question I ask myself now is: why - under perfect conditions without wind, barometric changes, lighting changes - does Mavic drift instantly and (for my inexperience) uncontrollable when entering ATTI-mode ?  If no external influences drive Mavic into a certain direction (like wind), why does it not "sit in the air" when there is no stick-input ? Manual just says that simply GPS/OA/VPS are shut off/not available and Mavic is prone to wind-drift .......................

Is there no internal GYRO-sensor (like in every smartphone) that tells Mavic to hover "on the spot" when there is no GPS/OA/VPS/RC-stick-input ?

Or is it, that in ATTI-mode the four propeller-rpm`s are not synchronized anymore for stationary stability and my Mavic drifted because of slightly different non-synchronized propeller rpm´s ?

Any other Mavic-internal electronics/processor influence that might be the cause ? Radio/Signal-interference by WiFi ? Compass-interference by steel-armed walls ?

I really would like to UNDERSTAND and KNOW the external and internal factors that influence flight-behaviour of my Mavic in ATTI-mode.
Thanks in advance for any answer that can give me a deeper insight (and maybe spare me one of my 7 replacement props left )

The mavic, has an IMU which is basically your gyro, baro flight controller. I'm guessing that this isnt calibrated properly so it registers horizontal as shifted slightly so when your mavic enters ATTI mode, the IMU tried to keep it level and makes it veer off.

I'd calibrate the IMU if i were you, then test again ... might be worth while waiting for the prop guards though

Danny-B- Posted at 2017-1-29 03:27
The mavic, has an IMU which is basically your gyro, baro flight controller. I'm guessing that this isnt calibrated properly so it registers horizontal as shifted slightly so when your mavic enters ATTI mode, the IMU tried to keep it level and makes it veer off.

I'd calibrate the IMU if i were you, then test again ... might be worth while waiting for the prop guards though

Thanks, I´ll try that (but will wait for my prop-guards now )

Your experience with ATTI mode sounds a lot like what I've read from many other users who've had their Mavics unexpectedly switch to ATTI mode. ie. The Mavic drifts suddenly and quickly, and quite often doesn't seem to respond to stick inputs. In fact it happened to me early on, and I crashed my Mavic. Now, being quite inexperienced with quads in general and ATTI mode in particular, I was left to believe that the Mavic behaved exactly as it was supposed to in ATTI mode. However, I've now read so many describe very unstable behavior in ATTI mode. I really suspect that the Mavic's ATTI mode is highly unstable, but there's no easy way for us to test it properly because there's no manual ATTI switch.

Unfortunately this is also my experience of atti mode. I instigated an auto take off , then atti mode kicked in and the mavic veered very quickly off to the left and I felt that I had no control over it at all. Dumped itself in a ploughed field after 9 seconds of uncontrolled flight. Just broke a pro so was very lucky on that occasion.

There is probably no manual switch because of this reason I suspect.  Maybe a switch will arrive when it is stable?

There is 100% no manualy switchable function just because of this.
All of my self built crafts will hover just fine on the spot given no external influence,most of them you wont even notice the change from gps to atti.
Smaller crafts are ofcourse more easly affected by the ellements, but as you proved to yourself, the mavic just plain sucks in atti.
Sometimes its due to  the build, but the mavic is preaty ballanced and has no extreme prop geometry.
Preaty confident this can be sorted by softwate.

Mrimstad Posted at 2017-1-29 05:41
There is 100% no manualy switchable function just because of this.
All of my self built crafts will hover just fine on the spot given no external influence,most of them you wont even notice the change from gps to atti.
Smaller crafts are ofcourse more easly affected by the ellements, but as you proved to yourself, the mavic just plain sucks in atti.

Tx for Your expertise, I learn more and more ............

Hi, definetely There is a problem with Mavic's software regarding instantly drift on Atti mode. I crashed my Mavic and it is totally unusable now. Dji doesn't accept that This is their problem. I am experienced drone pilot have 1 Phantom 3 and.

https://forum.dji.com/thread-82184-1-1.html See my thread

Ejder Bingül Posted at 2017-1-29 07:13
https://forum.dji.com/thread-82184-1-1.html See my thread

Read Your post already when You posted it ...........  sorry for the experience You had.
This is why I wanted to test ATTI-mode INDOORS with AMPLE space to exclude drafts from wind and propeller-airflow (room is absolutely empty, so no turbulences or vortex-effects). If THAT would have been the case the Mavic should have been moving away from the curtain and not into it. The downblow of the airflow would build up a higher pressure area below and again upwards on a wall/curtain and would "push" it away from it ...............

So, still the question has to be answered: does the Mavic hav a 3-axis inertia-gyro built in like a smartphone or fitness-tracker ? And if yes, why doesn´t it make Mavic just hover when there are no forces acting on her from outside in ATTI-mode?

This is from my limited experience of Drones (Please ignore this post if you find it irrelevant to your question), but some experience with r/c electric helicopters. It used to be critical that the heli was properly balanced, and that the gyros were calibrated to maintain a stable hover. This would include balancing props and checking for any damage or imperfections in the main and tail rotors. Given the complexity of propulsion from a multi rotor drone, I assume what flight controls would normally be controlled autonomously are instantly handed over to the pilot in ATTI mode. This should not be taken de facto, but could offer an explanation as to why ATTI mode seems so unpredictable.

rick39 Posted at 2017-1-29 08:12
This is from my limited experience of Drones (Please ignore this post if you find it irrelevant to your question), but some experience with r/c electric helicopters. It used to be critical that the heli was properly balanced, and that the gyros were calibrated to maintain a stable hover. This would include balancing props and checking for any damage or imperfections in the main and tail rotors. Given the complexity of propulsion from a multi rotor drone, I assume what flight controls would normally be controlled autonomously are instantly handed over to the pilot in ATTI mode. This should not be taken de facto, but could offer an explanation as to why ATTI mode seems so unpredictable.

The thing is, every explanation I've read about ATTI mode says that the drone should basically maintain a level attitude. So wind will cause it to drift, but it should at least remain level and relatively stable. The Mavic has (or should have) enough internal sensors to at least know if it's flying level or not. You can even see a nice demonstration of ATTI mode on a Phantom here:

As you can see, that Phantom in ATTI mode behaves nothing like what most people are experiencing with the Mavic in ATTI mode. The guy even gives the Phantom a good push, and you can see it instantly correct it's attitude even though it continues to drift in the direction it was pushed.

Having seen that video, this is how I was expecting ATTI mode to be like on the Mavic.

Jason Lane Posted at 2017-1-29 08:42
The thing is, every explanation I've read about ATTI mode says that the drone should basically maintain a level attitude. So wind will cause it to drift, but it should at least remain level and relatively stable. The Mavic has (or should have) enough internal sensors to at least know if it's flying level or not. You can even see a nice demonstration of ATTI mode on a Phantom here: https://youtu.be/yEH32qiTu1U?t=2m7s

As you can see, that Phantom in ATTI mode behaves nothing like what most people are experiencing with the Mavic in ATTI mode. The guy even gives the Phantom a good push, and you can see it instantly correct it's attitude even though it continues to drift in the direction it was pushed.

Wow, thanks for posting this video. This is exactly what I was expecting as behaviour in ATTI-mode.

I really wonder why no DJI-rep here (e.g. Ken, Paladin) step into the discussion here for deeper explanations and insights why Mavic can´t do what Phantom obviously can ?

It's assuming indeed but yes, this could be a good reason that there is no atti and dji keeps it shut. Not capable (YET, hopefully).

MacIak Posted at 2017-1-29 13:29
It's assuming indeed but yes, this could be a good reason that there is no atti and dji keeps it shut. Not capable (YET, hopefully).

At least they could explain technically which sensor-array is active in ATTI-mode (3-axis-gyro if at all ?, compass ?)

Angelo26 Posted at 2017-1-29 09:37
Wow, thanks for posting this video. This is exactly what I was expecting as behaviour in ATTI-mode.

I really wonder why no DJI-rep here (e.g. Ken, Paladin) step into the discussion here for deeper explanations and insights why Mavic can´t do what Phantom obviously can ?

They are on Chinese New Year holidays as far as I know

I don't have the mavic yet, but I was expecting that behavior. I feel the problem with the ATTI mode is that is not something you can select, simply. When the aircraft gets into the ATTI-GPS loop it gets crazy trying to get back quickly to GPS, and the constant switch in modes reflects that dummy pattern that we've seen from some reports.

If you could just hold ATTI when it switches, and wait to receive a proper positioning so you could get safelly back to GPS mode, probably most of the problems would be sorted out.

CuaC Posted at 2017-1-30 01:11
They are on Chinese New Year holidays as far as I know

I don't have the mavic yet, but I was expecting that behavior. I feel the problem with the ATTI mode is that is not something you can select, simply. When the aircraft gets into the ATTI-GPS loop it gets crazy trying to get back quickly to GPS, and the constant switch in modes reflects that dummy pattern that we've seen from some reports.

Question is, if this could be managed by software/firmware: i.e. Mavic stays stable based on 3-axis-gyro-input (??) in ATTI-mode and seperately searches for GPS-Sats and ONLY if GPS-sat-signal is stable for e.g. 10 seconds once again GPS-mode steps in.

Anyway, they managed somehow in Phantom .......................

CuaC Posted at 2017-1-30 01:11
They are on Chinese New Year holidays as far as I know

I don't have the mavic yet, but I was expecting that behavior. I feel the problem with the ATTI mode is that is not something you can select, simply. When the aircraft gets into the ATTI-GPS loop it gets crazy trying to get back quickly to GPS, and the constant switch in modes reflects that dummy pattern that we've seen from some reports.

I don't really believe that's what the problem is. I'm pretty sure the Mavic's programming would only switch back to GPS when there's a good solid lock. I don't think it would switch back and forth so easily.

That is easy to check.
Stick a 4-5 layers thick slab of aluminium tape on the back to block out gps.
If it flyes bad then you know.

Atti should use gyro and baro only for stabilization. No position hold.
So people who cant tell the difference from drifting and uncontrolled movement should think twice beefore posting crashes.

Now i will admit, i got ballsy..
I have long experience with both multirotors and rc heli's.
But i still dont usaly fire upp inside.
Tried in a werrry tight spot in my livingroom and made sure i had vps off. also no sats innside here so atti properly engaged.
It flew as expected, the downwash pushing it arround but perfectly controlable.
Not a werry scientific and long test but even so.
Might help to prove that this is an isolated problem?

Jason Lane Posted at 2017-1-30 02:46
I don't really believe that's what the problem is. I'm pretty sure the Mavic's programming would only switch back to GPS when there's a good solid lock. I don't think it would switch back and forth so easily.

Check all the reports here with people having some surprises when flying indoors as the GPS signal bounces and the drone starts getting/losing positioning...

Angelo26 Posted at 2017-1-30 01:58
Question is, if this could be managed by software/firmware: i.e. Mavic stays stable based on 3-axis-gyro-input (??) in ATTI-mode and seperately searches for GPS-Sats and ONLY if GPS-sat-signal is stable for e.g. 10 seconds once again GPS-mode steps in.

Anyway, they managed somehow in Phantom .......................

Yup... early adopters in the end, I guess they'll figure out in the near future how to improve everything.

Those that have experienced the sudden and extreme drifting when the MP switches to ATTI mode, did you calibrate the IMU beforehand?  Just wondering if there is any correlation between the sudden drifting and a lack of a good IMU calibration?

FlashGJ Posted at 2017-1-30 05:40
Those that have experienced the sudden and extreme drifting when the MP switches to ATTI mode, did you calibrate the IMU beforehand?  Just wondering if there is any correlation between the sudden drifting and a lack of a good IMU calibration?

We're repeatedly told by DJI reps not to do an IMU calibration unless the app tells us to. They're supposed to calibrated at the factory with precision equipment, so I've trusted in DJI's ability to properly calibrate these aircraft at the factory and I've never calibrated my IMU.

Jason Lane Posted at 2017-1-30 06:02
We're repeatedly told by DJI reps not to do an IMU calibration unless the app tells us to. They're supposed to calibrated at the factory with precision equipment, so I've trusted in DJI's ability to properly calibrate these aircraft at the factory and I've never calibrated my IMU.

I personally did not trust that the IMU could handle the shipping process without the possibility of it being knocked out. Upon receiving mine I did a full calibration and the two times that I have flown in ATTI mode, it was solid as a rock with no drifting (no wind both times). Maybe there is absolutely no correlation, but it has got me wondering if there is.

FlashGJ Posted at 2017-1-30 06:19
I personally did not trust that the IMU could handle the shipping process without the possibility of it being knocked out. Upon receiving mine I did a full calibration and the two times that I have flown in ATTI mode, it was solid as a rock with no drifting (no wind both times). Maybe there is absolutely no correlation, but it has got me wondering if there is.

That's an interesting data point. Thanks for your input. You may have a point. I guess I never thought about any of the IMU components being the type of thing that could get "knocked out," but I really don't know much about the mechanics/internals of an IMU unit. You could be right about rough handling knocking things loose.

This is yet another reason why we should be able to select ATTI mode manually... So we can check its stability under controlled conditions and make adjustments (eg. IMU calibration) if necessary.

One problem with the barometric transducer indoors is that the pressure inside a building is influenced by HVAC equipment cycling on and off.  

Griffith Posted at 2017-1-30 06:51
One problem with the barometric transducer indoors is that the pressure inside a building is influenced by HVAC equipment cycling on and off.

But I assume a barometric transducer is only used for measuring height. Is that correct? Or are you saying that the Mavic uses this in some way to also measure for a level attitude?

In all these accounts of ATTI craziness, it's not maintaining height that's the problem. The Mavic seems to shoot off horizontally very quickly and suddenly.

You are correct, the baro sensor normally gives information related to height., and should not cause the craft to veer laterally.  That being said, who the hell knows what computations are taking place (if any) in ATTI mode between compass and visual and baro and ultrasound and whatever other sensors may be lurking.  I get the feeling ATTI mode is not just a fly-by-stick system,  but includes some sort of stability control  system.

Having experience writing control software,  this situation reminds me of "poor initialization", a condition in which an non-(or improperly) initialized control system behaves unpredictably on startup until it has sufficient accurate feedback (recent history) to stabilize.   Like everything (pitch, yaw, height,  direction) starts at zero (or some erroneous value), regardless of the actual current conditions at the moment just before ATTI switches on.

Since I don't know a thing about the Mavik control systems, you might want to consider this commentary as just the ramblings of an old engineer :-)

Jason Lane Posted at 2017-1-30 06:47
That's an interesting data point. Thanks for your input. You may have a point. I guess I never thought about any of the IMU components being the type of thing that could get "knocked out," but I really don't know much about the mechanics/internals of an IMU unit. You could be right about rough handling knocking things loose.

This is yet another reason why we should be able to select ATTI mode manually... So we can check its stability under controlled conditions and make adjustments (eg. IMU calibration) if necessary.

I am not sure if they can actually get "knocked out" either, I am definitely no expert. But months ago while I was reading and studying and waiting for my MP to arrive, I remember reading something about the possibility of IMUs being inaccurate if the drone has been handled roughly. How much it would take?, or if it is actually true?, I have no idea. I decided then to calibrate everything at arrival, and it has flown flawlessly from day one and I have never calibrated a thing since that first time. Possibly just luck with getting a good unit, I have read many accounts of people that have never calibrated a thing, not even the compass, and never had a speck of trouble either. So who really knows what is best!!

Mrimstad Posted at 2017-1-30 03:38
That is easy to check.
Stick a 4-5 layers thick slab of aluminium tape on the back to block out gps.
If it flyes bad then you know.

"Atti should use gyro and baro only for stabilization. No position hold."


Interesting point: I considered a 3-axis-gyro to be responsible for holding the position as a feedback-loop when no stick-input/wind-influence occurs. At least it should drift only SLOWLY when no wind-drift pushes it.


Anyway: I just received my prop-guards from HongKong and will start some testing indoors creating "perfect conditions" without fear to ruin another prop ..............  

Up until now I've read posts like this and wondered how frequently the forced switch to ATTI really occurs...until today.
I was filming in a rural area, away from likely  interference when suddenly I got an ominous, recurring beeping on the RC, but (bizarrely) no feedback as to what was causing it on the GO app.
Then it stopped, so, for a brief moment I thought all was well.
Seconds later GPS was gone and i was in ATTI mode, with the MP quickly veering away. While I have some experienceI flying drones without stable positioning systems, (e.g. GPS or  VPS), the fact that the Mavic was suddenly veering away so quickly was disconcerting. I had LOS, but decided that if I didn't move quickly I risked losing it, so I decided to try and force it down in the nearby, ploughed field ASAP. I had to hop a couple fences and was worried as the light was starting to fade that I'd not find it. After a few minutes of searching, I tried starting the motors, thinking I might hear the Mavic.  Despite the props being blocked, I got lucky and could hear it about 20 metres away.
Fortunately the only thing damaged was my faith in the MP keeping a reliable connection to GPS and being able to autonomously hold a position. All this occurred on a flight within about a 200 mètres. I've flown further out in the past, but this makes me a lot less confident in the machine.

Good theme. It would be helpful to hear answer from the technical staff

If i remember correctly, somewhere in the youtube i saw advertising or like first user's clip that it is very cool to fly indoors with Mavic.  It was recommended to do it. That was one of the first clips about Mavic drone .

I can also add in a data point to all this.  Brand new Mavic, less than a week old.  Have done no calibrations since it never asked.  Seconds after takeoff, it losses GPS, enters ATTI, and starts veering like crazy.  Gained altitude right away so as not to hit things, and even though I'm trying to control it, I've having a very hard time.  For a few seconds, it appeared as if I would just have to wait for it to crash somewhere.

Thankfully it regained GPS mode, but my heart was pounding and I landed right away.  This ATTI mode is absolutely useless.  Given that is has a compass and IMU, I see not reason why it should drift like crazy when so close to the ground with little wind.  If all it did was keep the motors at the exact same speed as before losing GPS, I doubt it would drift as far as it does.  The ATTI mode really does cause it to behave erratically.

That is perfectly normal.
Atti mode uses baro and gyro to stabilize the craft so it will stay level and not tipp arround.
And to some extent keep the hight. It will drift arround if you fly low due to the ground effect.
And the fact this thing weights nothing!
Just the nature of flying rotarywings of any kind.
The fact many who fly indoors get  cought off guard, and doing fast large control inputs doesnt help things with an agile craft like this either.

Control scheme for the mavic stock can be a bit touchy for indoor flight in atti aswell so maybe go inn and soften it some
with some expo.

Now i have tried to find out if the continuies switching from gps to atti might be an issue. But i cant.
Even with a 6mm slab af alu on the back it still finds enough sat strenght to keep it in gps hold.

I just now realise i have also gone arround this all wrong, and that the op actualy wants to know whats going on.
Il try to be short.

You craft has an IMU, well actualy it has two. But thats beside the point.
The IMU consists of a range of sensors, gyro's, barometers,compas and accelerometers in some cases (dont know about the mavic)
And the GPS, the GPS is though in most cases counted as an external sensor, as it is not a part of the IMU package.
Even so, the IMU controls the level flight of the craft. It will meassure, calculate and send relevant information to the speedcontrollers in order to keep the craft level.
So a multirotor is in practise just a floating cushion of air, even the smalest touch will send it gliding away like its on ice.
Needless to say the smaler and lighter the more effected. Also the mavic has a rather large rotor disc area for its size, so it can feel even floatier.
Now this motion might be so minute even a propper accelerometer wont pick it upp. And so the craft will have no idea it is actualy gliding away.
This is where GPS and VPS comes in and seccures POSTION HOLD.

Not trying to be a hothead or seem like i know everything, just trying to explain atleast from my experience what would be the logical
buildupp. I have made my own flightcontrollers and i feel confident dji cant have taken such a differenet aproach than everyone else.
That being said. If you read all that, and can have that in back off your head. and still feel something is wrong when your flying and given situation  happens.
Then please document it video as best as possible so the rest of uss can have a look.

Mrimstad Posted at 2017-1-31 17:30
That is perfectly normal.
Atti mode uses baro and gyro to stabilize the craft so it will stay level and not tipp arround.
And to some extent keep the hight. It will drift arround if you fly low due to the ground effect.

"Even so, the IMU controls the level flight of the craft. It will meassure, calculate and send relevant information to the speedcontrollers in order to keep the craft level."

That's what we expect, and that's quite clearly demonstrated with the Phantom in the YouTube video I posted above. But that's definitely not the behavior that many people are seeing when the Mavic switches to ATTI mode. The Mavic seems to veer off in one direction or another under its own power. It doesn't feel like it's just being pushed by some external force like wind, ground effect, prop wash, etc.

Jason Lane Posted at 2017-1-31 17:57
"Even so, the IMU controls the level flight of the craft. It will meassure, calculate and send relevant information to the speedcontrollers in order to keep the craft level."

That's what we expect, and that's quite clearly demonstrated with the Phantom in the YouTube video I posted above. But that's definitely not the behavior that many people are seeing when the Mavic switches to ATTI mode. The Mavic seems to veer off in one direction or another under its own power. It doesn't feel like it's just being pushed by some external force like wind, ground effect, prop wash, etc.

Then if that is the case i will agree there might be an issue.
I have not seen these symptoms, So a GOOD video might help out alot.
The ones i ahve seen on yt clearly is not good enough.

This is an interesting thread that I'll be keeping an eye on. I too lost a prop when I was flying indoors in ATTI mode. The mavic did not seem to take stick input and ended up hitting some furniture before I was able to get it to land.

I am torn between thinking there is an issue with ATTI mode and with thinking that I was just very low to the ground, indoors and the ground effect made it act weird and uncontrollable. I don't know what is the truth though.

Mrimstad Posted at 2017-1-31 18:07
Then if that is the case i will agree there might be an issue.
I have not seen these symptoms, So a GOOD video might help out alot.
The ones i ahve seen on yt clearly is not good enough.

Well I can share my video that I also shared in early Nov when I crashed my Mavic. I'm not saying it's a GOOD video, but it's what I have.

I was very new to quads and the Mavic, and I made the mistake of flying it in ATTI mode and at night. I know that wasn't the smartest move, and I eventually put it down to pilot error, accepted that and paid for the repairs. (No need to remind me that it was a bad move). However, I always maintained that the Mavic did not feel like it was just drifting, and it didn't seem to be following my commands. But being a novice pilot, I couldn't really say that with much authority. The video is short, and doesn't show much, but the one thing it does show is a definite bank angle as the Mavic flew sideways and up until it crashes into a tree.

So as you said, I would have expected it to maintain level flight even if there was a wind, but it doesn't look that way in the video. The part where it starts flying sideways without my input is at about 0:25

Not sure what you'll make of it, but there you have it: