Charles Adams Posted at 2017-10-1 08:56
I'll try again with different words.  My test of these circumstances with my P3S failed.  The reason is that the P3S set in manual ATTI mode still knows its location.  It only changes flight behavior (it lets the elements have influence over the flight), but it is still able to identify and map its coordinates.  Therefore the P3S in manual ATTI mode cannot be used to perform this test.

The only way to perform this test is as Hallmark suggested, taping a magnet on to disrupt the compass.  I'm not that ambitious.  But it does remind me of a movie quote that I'll alter slightly:

"All I ask for is a frikkin drone with laser beams!"

Charles Adams Posted at 2017-10-1 08:58
The only way to perform this test is as Hallmark suggested, taping a magnet on to disrupt the compass.  I'm not that ambitious.  But it does remind me of a movie quote that I'll alter slightly:

"All I ask for is a frikkin drone with laser beams!"

First order of business, in my opinion, is the GPS. ATTI mode means that both GPS and vision are not working or their data is not being read.
If GPS data still exists but gets thrown away when it contradicts the compass, then the algorithm should be fixed. GPS is typically much more reliable than a compass, and a moving AC can identify its orientation based on the change in coordinates. If you're looking for a fail-safe arrangement, that's the ultimate one.
If GPS gets lost and does not get re-acquired, even after the AC has moved, then something is extremely questionable with the performance of the GPS itself, or with its driver in the AC's formware, or with its electromagnetic shielding. Think about it - any low-end smartphone nowadays can acquire and maintain GPS reception, or reacquire it when it gets momentarily lost - and that's in much tougher conditions: at ground level, typically near many more sources of magnetic interference, and seeing a smaller portion of the sky (hence, fewer satellites) than a drone in mid-air.
For me, that's the part that needs to be looked at first, because it doesn't add up.

Straightforward answer - there are no inbuilt algorithms or flight autonomy in the event of such a rare situation (I've never lost GPS during flight). You would simply control the drone manually via visual line of sight using the control sticks until GPS reconnects during the manual flight to your position. The only in-built component that would remain operational is the barometer if flying above the downward VPS limitations which I believe is >8m AGL . If you have flown too far and lost VLOS then it is likely you would lose the drone which is why most aviation authorities stipulate the need to maintain VLOS at all times in many drone codes. To be clear, this would not result in a true flyaway though - it would just be far more difficult to fly as the elements play a huge part in the flight controls - especially wind speeds. Flyaways are where the signal is lost and the drone 'connects' to an overpowering signal such as a telephone mast and you have no control over the drone. This is why I am an advocate for DJI to include an ATTI mode option in both the Spark and Mavic - I'm amazed it has been removed.

ACW Posted at 2017-10-1 09:49
Straightforward answer - there are no inbuilt algorithms or flight autonomy in the event of such a rare situation (I've never lost GPS during flight). You would simply control the drone manually via visual line of sight using the control sticks until GPS reconnects during the manual flight to your position. The only in-built component that would remain operational is the barometer if flying above the downward VPS limitations which I believe is >8m AGL . If you have flown too far and lost VLOS then it is likely you would lose the drone which is why most aviation authorities stipulate the need to maintain VLOS at all times in many drone codes. To be clear, this would not result in a true flyaway though - it would just be far more difficult to fly as the elements play a huge part in the flight controls - especially wind speeds. Flyaways are where the signal is lost and the drone 'connects' to an overpowering signal such as a telephone mast and you have no control over the drone. This is why I am an advocate for DJI to include an ATTI mode option in both the Spark and Mavic - I'm amazed it has been removed.

I would say that 99% of the Spark flyaways result in this very same situation that I am addressing here: suddenly mid-flight there is a compass error and the AC switches to ATTI mode (thus ignoring the GPS). It starts drifting away because of the wind. The pilots don't know how to control it (because of lack of experience). Finally the lose connection because of the distance and weak Wifi signal (especially in CE region) and that's it. The Spark is lost - because no failsafe mechanisms seems to kick in while in ATTI mode. The drone flies away.

This is why I also think that in the situation when the signal to RC is lost the drone should automatically descend until it reaches an altitude where the VPS can be turned on.
This would increase the chances of finding your drone as the radius would be limited.

ciesnik Posted at 2017-10-1 22:53
I would say that 99% of the Spark flyaways result in this very same situation that I am addressing here: suddenly mid-flight there is a compass error and the AC switches to ATTI mode (thus ignoring the GPS). It starts drifting away because of the wind. The pilots don't know how to control it (because of lack of experience). Finally the lose connection because of the distance and weak Wifi signal (especially in CE region) and that's it. The Spark is lost - because no failsafe mechanisms seems to kick in while in ATTI mode. The drone flies away.

This is why I also think that in the situation when the signal to RC is lost the drone should automatically descend until it reaches an altitude where the VPS can be turned on.

but once again I ask, like I asked somebody else -  why do you accept the throwing away of GPS data as God-given fate?
Your suggestion means that the drone will descend randomly at the spot it happens to be at, hundreds of meters away from you, perhaps onto inhospitable territory e.g. water, treetops, or rocky edges. Even if it lands safely, it's pretty far away from you and you could have a hard time finding it.
My suggestion for a fail-safe mechanism for compass error is - if GPS still receives and is able to determine location, don't ignore it and don't go into ATTI mode at all. It doesn't make sense. Instead, use GPS to regain orientation after the AC has moved, and get back home, or close to home. That should also bring it closer to the pilot, so the RC can reacquire connectivity.

djiuser_Kf4iPA3 Posted at 2017-10-1 23:28
but once again I ask, like I asked somebody else -  why do you accept the throwing away of GPS data as God-given fate?
Your suggestion means that the drone will descend randomly at the spot it happens to be at, hundreds of meters away from you, perhaps onto inhospitable territory e.g. water, treetops, or rocky edges. Even if it lands safely, it's pretty far away from you and you could have a hard time finding it.
My suggestion for a fail-safe mechanism for compass error is - if GPS still receives and is able to determine location, don't ignore it and don't go into ATTI mode at all. It doesn't make sense. Instead, use GPS to regain orientation after the AC has moved, and get back home, or close to home. That should also bring it closer to the pilot, so the RC can reacquire connectivity.

Good point and I wouldn't want my drone to descend to 8m AGL away from my line of sight - at that altitude somebody at ground level could quite easily knock it out of the sky. The key is to ensure a strong GPS signal with a homepoint lock on the map BEFORE takeoff and to maintain VLOS. Perhaps this is why I have never lost a drone in >1000 flights...

djiuser_Kf4iPA3 Posted at 2017-10-1 23:28
but once again I ask, like I asked somebody else -  why do you accept the throwing away of GPS data as God-given fate?
Your suggestion means that the drone will descend randomly at the spot it happens to be at, hundreds of meters away from you, perhaps onto inhospitable territory e.g. water, treetops, or rocky edges. Even if it lands safely, it's pretty far away from you and you could have a hard time finding it.
My suggestion for a fail-safe mechanism for compass error is - if GPS still receives and is able to determine location, don't ignore it and don't go into ATTI mode at all. It doesn't make sense. Instead, use GPS to regain orientation after the AC has moved, and get back home, or close to home. That should also bring it closer to the pilot, so the RC can reacquire connectivity.

It is not possible for the drone to fly somewhere without a working compass. GPS alone tells the drone only where it actually is. But the AC has no idea where it is heading to. And if it does not know where it is heading to it cannot know where left and right of the drone is.
Even if you have multiple GPS points over time as it is drifting away: the drone can not determine in which direction it should move to compensate the drift without knowing where its head is pointing to.

ciesnik Posted at 2017-10-1 23:58
It is not possible for the drone to fly somewhere without a working compass. GPS alone tells the drone only where it actually is. But the AC has no idea where it is heading to. And if it does not know where it is heading to it cannot know where left and right of the drone is.
Even if you have multiple GPS points over time as it is drifting away: the drone can not determine in which direction it should move to compensate the drift without knowing where its head is pointing to.

Not true.
Once you have multiple coordinates (and you have them when you move), a simple calculation (vector difference between the different coordinates) will tell you your heading.
If you have it in your firmware, you can know your heading without a compass.

djiuser_Kf4iPA3 Posted at 2017-10-2 00:02
Not true.
Once you have multiple coordinates (and you have them when you move), a simple calculation (vector difference between the different coordinates) will tell you your heading.
If you have it in your firmware, you can know your heading without a compass.

This is only true for a car or for a vehicle where the direction of moving is known: a car usually only moves forward so you can assume where the heading is once you have multiple coordinates.
But an AC can drift or move sideways, backwards or any direction. Therefore you cannot assume any heading with multiple coordinates.

ciesnik Posted at 2017-10-2 00:07
This is only true for a car or for a vehicle where the direction of moving is known: a car usually only moves forward so you can assume where the heading is once you have multiple coordinates.
But an AC can drift or move sideways, backwards or any direction. Therefore you cannot assume any heading with multiple coordinates.

Yes you can, if you don't just let yourself "drift". The AC can initiate a short flight straight on, where its nose is heading, whichever direction that is, while collecting GPS data, to gain an initial reference direction. Then it knows not only its direction, but also the orientation of its nose - so it can yaw toward the right direction and fly.

djiuser_Kf4iPA3 Posted at 2017-10-2 00:12
Yes you can, if you don't just let yourself "drift". The AC can initiate a short flight straight on, where its nose is heading, whichever direction that is, while collecting GPS data, to gain an initial reference direction. Then it knows not only its direction, but also the orientation of its nose - so it can yaw toward the right direction and fly.

This sounds easy only in theory. In practice you have winds changing direction or speed, gusts, turbulences, which makes this method very unreliable.

I still think for the Spark you have only 2 options in this situation (on signal loss):
1. sink until VPS is able to kick in
or 2. do nothing until battery dies

In both situations the chance of losing the AC is high. But, for the first option the chances are slightly better that you will be able to find it again or re-gain signal as it would not fly away too far.

ciesnik Posted at 2017-10-2 00:24
This sounds easy only in theory. In practice you have winds changing direction or speed, turbulences, which makes this method very unreliable.

I still think for the Spark you have only 2 options in this situation (on signal loss):

Yes, you have winds that may drift you away while you're flying straight or yawing, and it may introduce error into your calculation. But the algo I'm suggesting doesn't mean you do it once and stop. After the AC has corrected itself once and started flying in the corrected direction, it should keep collecting GPS data and correct itself further. Do this recursively, and the error will decrease over time.
The AC may zigzag a little at first, but the zigging will decrease as it corrects itself further, and it will get to where it's supposed to get - or at least close enough for the pilot to regain control and LOS.

ciesnik Posted at 2017-10-2 00:24
This sounds easy only in theory. In practice you have winds changing direction or speed, gusts, turbulences, which makes this method very unreliable.

I still think for the Spark you have only 2 options in this situation (on signal loss):

I'm always more worried about not doing harm to people around than than about losing  AC. So 1-st option to descend to VPS height would be more preferable for me.

fans04bc4ba7 Posted at 2017-10-2 01:59
I'm always more worried about not doing harm to people around than than about losing  AC. So 1-st option to descend to VPS height would be more preferable for me.

I refer you to link below, Aircraft flying in Atti mode under 16 feet, he has no VPS because he is in Atti mode, this has been made clear to OP by dji Susan so it’s a bit ridiculous to keep on about somehow being saved by VPS when you don’t have it.
As it stands flying in Atti mode keep Aircraft in VLOS and try to fly to a safe landing area, there are no other options or hidden failsafe, you have more chance of regaining gps by increasing altitude in Atti https://forum.dji.com/forum.php? ... &fromuid=260008

Charles Adams Posted at 2017-10-1 08:58
The only way to perform this test is as Hallmark suggested, taping a magnet on to disrupt the compass.  I'm not that ambitious.  But it does remind me of a movie quote that I'll alter slightly:

"All I ask for is a frikkin drone with laser beams!"

So I can confirm what happens when spark goes into Atti mode and drifts (not radio loss) it will continue to drift without operator inputs until it reaches critical battery and will then enter forced landing. I discovered this through looking at log in the link below. This is a real situation and it was played out to the end.
Yes I know it’s not the answer to the question the OP posed but frankly I think the result would be exactly the same . if radio as well as gps were lost, I think it’s completely pie in the sky (pardon the pun] to think that there is some algorithm in the spark that dji is keeping secret that will allow spark to land after 3 seconds if gps and radio were lost. If that was the case when gps is lost we would just be able to turn off radio and spark would just land.

Below link may help others who are not aware of the conclusion of a so called fly away. So my thinking is when spark goes into Atti mode that pilot should try to keep it in VLOS if possible over a clear area not to high allowing that critical battery at say 12% would be enough for forced landing from his/her altitude.

https://forum.dji.com/forum.php? ... &fromuid=260008

hallmark007 Posted at 2017-10-2 07:15
So I can confirm what happens when spark goes into Atti mode and drifts (not radio loss) it will continue to drift without operator inputs until it reaches critical battery and will then enter forced landing. I discovered this through looking at log in the link below. This is a real situation and it was played out to the end.
Yes I know it’s not the answer to the question the OP posed but frankly I think the result would be exactly the same . if radio as well as gps were lost, I think it’s completely pie in the sky (pardon the pun] to think that there is some algorithm in the spark that dji is keeping secret that will allow spark to land after 3 seconds if gps and radio were lost. If that was the case when gps is lost we would just be able to turn off radio and spark would just land.

What is the mechanical definition of a "forced landing"?  The craft descends until it senses a surface and then attempts to delicately land upon that surface?  The manual doesn't contain the term "forced landing".  I'm curious how the craft determines "down" without gps (barometer?).

Charles Adams Posted at 2017-10-2 07:38
What is the mechanical definition of a "forced landing"?  The craft descends until it senses a surface and then attempts to delicately land upon that surface?  The manual doesn't contain the term "forced landing".  I'm curious how the craft determines "down" without gps (barometer?).

You have to differentiate between phantomhelp terminology and dji’s dji may not use forced landing.
But it does tell you in the manual, when there is only sufficient battery power left to land Aircraft will just land, it will hover at 0.5 metre for a few seconds and just land,

I don’t have an answer regarding how would be nice to know.

hallmark007 Posted at 2017-10-2 07:48
You have to differentiate between phantomhelp terminology and dji’s dji may not use forced landing.
But it does tell you in the manual, when there is only sufficient battery power left to land Aircraft will just land, it will hover at 0.5 metre for a few seconds and just land,

Based on this evidence/case it appears that the drone won't just float until it runs out of power and crash.

Charles Adams Posted at 2017-10-2 08:39
Based on this evidence/case it appears that the drone won't just float until it runs out of power and crash.

It will crash if there are obstacles in the way, but yes it looks like it will just land as per the case I showed in the link, at least we know it’s not going to drop on somebody’s head, if the wind stops here I might just go out and try my P4 pro in Atti mode , I should be able to control until critical battery and have a big enough field to try it in.

hallmark007 Posted at 2017-10-2 08:51
It will crash if there are obstacles in the way, but yes it looks like it will just land as per the case I showed in the link, at least we know it’s not going to drop on somebody’s head, if the wind stops here I might just go out and try my P4 pro in Atti mode , I should be able to control until critical battery and have a big enough field to try it in.

That would be an awesome test, but one thing to factor into your test...  as with my P3S test, I think that your P4 will still be "GPS aware" even in atti-mode.  I noted when flying my P3S in atti mode that GPS was still being tracked accurately on the map.  Flight didn't benefit from GPS, but the craft was still "aware".

What I'm curious about (I did not test this because I also was contending with winds) is if the craft (when set purposefully in atti mode) will still attempt RTH functionality upon critical battery levels.  In the test case you pointed to, that atti mode was "forced" and not user selected...  the craft really had no GPS info.  In your tests (where you are manually selecting atti-mode) I'm curious if it will use the GPS data that I believe it retains.

Charles Adams Posted at 2017-10-2 09:09
That would be an awesome test, but one thing to factor into your test...  as with my P3S test, I think that your P4 will still be "GPS aware" even in atti-mode.  I noted when flying my P3S in atti mode that GPS was still being tracked accurately on the map.  Flight didn't benefit from GPS, but the craft was still "aware".

What I'm curious about (I did not test this because I also was contending with winds) is if the craft (when set purposefully in atti mode) will still attempt RTH functionality upon critical battery levels.  In the test case you pointed to, that atti mode was "forced" and not user selected...  the craft really had no GPS info.  In your tests (where you are manually selecting atti-mode) I'm curious if it will use the GPS data that I believe it retains.

Your right it will still retain gps ie coordinates as it did in the link I just posted, if there is no gps at all you won’t see any coordinates.

hallmark007 Posted at 2017-10-2 07:15
So I can confirm what happens when spark goes into Atti mode and drifts (not radio loss) it will continue to drift without operator inputs until it reaches critical battery and will then enter forced landing. I discovered this through looking at log in the link below. This is a real situation and it was played out to the end.
Yes I know it’s not the answer to the question the OP posed but frankly I think the result would be exactly the same . if radio as well as gps were lost, I think it’s completely pie in the sky (pardon the pun] to think that there is some algorithm in the spark that dji is keeping secret that will allow spark to land after 3 seconds if gps and radio were lost. If that was the case when gps is lost we would just be able to turn off radio and spark would just land.

I am not thinking that there is "some" algorithm: we know that there is one. You observed that the aircraft goes into forced landing when reaching critical battery level even in ATTI mode.
Why should the AC then not simply do the same 3 seconds after signal is lost to RC?

ciesnik Posted at 2017-10-3 01:22
I am not thinking that there is "some" algorithm: we know that there is one. You observed that the aircraft goes into forced landing when reaching critical battery level even in ATTI mode.
Why should the AC then not simply do the same 3 seconds after signal is lost to RC?

Why should it not. Because it doesn’t that’s the simple answer. You started the thread telling us you taught that AC after loosing gps and radio the AC would land after 3 seconds, this doesn’t happen , can it be programmed to happen , I’m not qualified to say it can, maybe this can be the topic of your next thread, but as for this thread I think it’s fairly easy to work out what actually happens.

hallmark007 Posted at 2017-10-3 01:50
Why should it not. Because it doesn’t that’s the simple answer. You started the thread telling us you taught that AC after loosing gps and radio the AC would land after 3 seconds, this doesn’t happen , can it be programmed to happen , I’m not qualified to say it can, maybe this can be the topic of your next thread, but as for this thread I think it’s fairly easy to work out what actually happens.

How can you say? As far as I read about your tests you did not test this very case: ATTI mode and RC signal lost, or did you?

ciesnik Posted at 2017-10-3 06:08
How can you say? As far as I read about your tests you did not test this very case: ATTI mode and RC signal lost, or did you?

Glad you asked, just did that today. Using my P4Pro in Atti mode 60ft alt, 80 from homepoint, turned off RC , after 8 seconds AC drifts to almost 300ft wind 5mph 2m/s , so turned back on radio made signal AC now at 360ft still drifting in Atti mode , so regained control in Atti flew back closer to 100 ft from home still at 60ft alt no low battery warning at 25% which is my low battery warning, at 16% beeping and red critical battery aircraft landed almost straight down into grass .
I couldn’t leave it without RC for more than 8 seconds because AC going away from me pretty quickly although almost no wind, but I can tell you it doesn’t have any algorithm which will begin landing after 3 seconds and I’m convinced it would not have landed until critical battery so even at less than 30% battery when turned off RC it would be a long way away before it reached critical battery.
As I said in earlier post you could tape a magnet to your own craft which should force Atti because of compass and gps data conflict if it does you could try it out on your spark, but I’m reasonably confident that I’m aware of the conclusion of end of flight with no gps, and although I didn’t get to the end of flight with no radio and no gps for obvious reasons I won’t be depending on my AC landing after 3 seconds without gps and radio

Thanks for this test! This is pretty interesting and probably explains most of the fly away/lost spark cases.
So the conclusion is that it probably would ran out of battery and fall from the sky? Unbelievable that DJI did not think of this situation and include any fail safe for this case. :O

ciesnik Posted at 2017-10-3 09:09
Thanks for this test! This is pretty interesting and probably explains most of the fly away/lost spark cases.
So the conclusion is that it probably would ran out of battery and fall from the sky? Unbelievable that DJI did not think of this situation and include any fail safe for this case. :O

No thats totally wrong, it gets to critical battery and lands. and thats the failsafe. it won't fall out of the sky so the likelihood of it causing damage to people or property are minimum to none.

Ardenno Posted at 2017-9-29 13:18
RTH is not possible in ATTI mode because is no GPS and Spark don't know where he is... I think after losing RC signal Spark in ATTi would fly same direction when flew in the moment when signal was lost. Probably till battery is critical landing or crashing. Please correct me if its different.

After loosing RC and GPS, I think it will land only.
I think thats how the "Drone-Gun" works:
Kill RC signals and GPS connection.
Probably kill compass functionality, too.

But I think without RC and GPS is only one thing to do, landing.

If I remember that right, there was a request to keep the drone in air, as long as possible to gain connection and signal back.
But right now, it goes direct down.

hallmark007 Posted at 2017-9-30 07:58
Just tape a magnet to the body of your spark, this will throw off your compass , risky particularly when we already know there is no failsafe landing in Atti mode.

I tried to start my Spark in a steal-concrete area. I got compass trouble, moved it 2m onto the grass.
It looked like, no more compass problems, started it, watched it some seconds, started a Pano and saw my Spark sailing with the wind "Magnetic filed Interference (Tipo in DJI App) Exit P-GPS Mode" I was lucky, it was already over water, but rising the altitude got it back in Compass and GPS Mode.

hallmark007 Posted at 2017-10-3 09:15
No thats totally wrong, it gets to critical battery and lands. and thats the failsafe. it won't fall out of the sky so the likelihood of it causing damage to people or property are minimum to none.

Ok, either way the answer to my initial question is: it keeps on flying after RC signal loss in ATTI mode until critical battery level.

This means, once you have a compass error mid-flight and the drone keeps drifting away  - if you do not manage to gain control over it, instead it flies away too far so that you will lose signal: it will be impossible to recover your Spark as it will keep on flying away.

ciesnik Posted at 2017-10-3 10:39
Ok, either way the answer to my initial question is: it keeps on flying after RC signal loss in ATTI mode until critical battery level.

This means, once you have a compass error mid-flight and the drone keeps drifting away  - if you do not manage to gain control over it, instead it flies away too far so that you will lose signal: it will be impossible to recover your Spark as it will keep on flying away.

If you fly in good conditions, good environment VLOS proper preflight checks, then you will much reduce the chances of getting in to that situation. I have only read of one incident here of that scenario happening and this was using mobile phone with WiFi, so yes it can happen .

ciesnik Posted at 2017-10-3 10:39
Ok, either way the answer to my initial question is: it keeps on flying after RC signal loss in ATTI mode until critical battery level.

This means, once you have a compass error mid-flight and the drone keeps drifting away  - if you do not manage to gain control over it, instead it flies away too far so that you will lose signal: it will be impossible to recover your Spark as it will keep on flying away.

I'd say "highly unlikely" instead of "impossible", but remote at best.  What you describe is what I consider to be the only true "fly away" scenario a pilot could face.  Every other scenario I can think of has opportunities for the system to autonomously return to home or allows the pilot to influence the flight.