Flayaway theory

[SteveMann]

I believe that the Flight Limits in the firmware may be at fault for some flyaways.

From the DJI manual:

"If the aircraft is flying out of the max radius in Ready to Fly (non-GPS) status, it will fly back within the limits range automatically if 6 or more GPS satellites have been found."

"Non-GPS status"? Does that mean ATTI mode? So if I fly in either GPS or ATTI mode and penetrate the horizontal limit, the Phantom will return to the limit.

Here is my theory, let me know if you see a flaw in it.
What happens if for whatever reason the Phantom thinks it is outside the horizontal limits? Maybe a brief GPS loss, scrambled data, or a GPS error that tells the Phantom that it is really a hundred miles thataway. It will then make a beeline to get back to it's radius limit. And since the limits have priority, your remote can't regain control and RTH is what the Phantom is trying to do. A hundred miles thataway.

If I am correct, then the only way to stop a runaway from this cause is to switch to Manual Mode where the GPS is truly offline and the Phantom won't try to RTH or return to the fence. Or disable the horizontal limits. (Is this even possible?)

If this is a possible error, then maybe the firmware needs a change. The limit radius should not be a hard wall, but a zone. The horizontal limit zone should be no more than the Phantom could penetrate under normal circumstances plus a little tailwind, maybe 30 Meters. If the Phantom enters the zone then it should return to the limit as it does now. But even if you fly directly into the zone with a little tailwind and inertia, you will not be able to penetrate the zone because the Main Controller will redirect it back to the fence. If the Phantom thinks it is outside the limit zone, then it should flash a GPS warning and not try to return to the fence.

Any thoughts?

 

[N017RW]

Applying your scenario...

Only if the home point can be erroneously set/reset through some kind of 'bug'.

Because if it were to become 'confused' as to it's current location due to GPS error/bug as you have proposed the home point remains the same. The return path back to the limit is still a straight line to or towards the [original?] home point.

or...maybe I don't fully understand

my $0.02

 

[Hughie]

Applying your scenario...

Only if the home point can be erroneously set/reset through some kind of 'bug'.

Because if it were to become 'confused' as to it's current location due to GPS error/bug as you have proposed the home point remains the same. The return path back to the limit is still a straight line to or towards the [original?] home point.

or...maybe I don't fully understand

my $0.02

Interesting thoughts chaps.


What if when the flight starts, the quad is inside the limit but has a home point somewhere else (say 1 mile away) due to a bad home lock or whatever. During the flight, the quad happens to be flown outside the fence with respect to what it thinks is the home point, and it then does RTH (1 mile away)

I agree this has to start with a inaccurate home point.

Therefore if the quad "believes" the home point is right, why would it not fly there.

EDIT: I think there is a Youtube video somewhere where this is discussed and one recommendation is to set the fence to an extremely long way away, so that this event never occurs. That is of course at the expense of proper fence limiting when you have a good homepoint.

EDIT+1: This link https://www.youtube.com/watch?v=_ge3GuhEDRM
The whole theory is based on the homepoint not being set properly in the first place.

 

[Buckaye]

I'm not sure if this helps or confuses the matter - but I have noticed something rather odd when powering on my iPad Ground station BEFORE the Phantom 2 achieves home lock.

If I click on the edit home point - the app zooms out to this Island in the middle of the ocean (I live in Florida) - it's off the east coast - and I've never seen a name of the island but, if I am remembering correctly, it's in the area of San Salvador in the bahamas (24.083917, -74.479733). To be clear... I have NO IDEA if that is the island - because I usually just waited for home lock to be achieved and the home point would zoom back to where I am located in Central Florida.

So - it seems to me (at least when using the ipad ground station - it that there is SOME parameter set as a home point... what I can't tell is if it is:

A. Just 0,0 which is off the coast of Africa in the "general" direction the GS seems to zoom to (maybe it's limited by how much map is loaded by the GS app and picks the point closest to 0,0 it can find)

B. If there is an actual default home point set in the bahamas (unquestionably unlikely)

C. It's just random - I find this unlikely as well - because it always goes to the East and always ends up around some strange island.

D. It's a fixed distance in some fixed direction and just happens to be an island - again - I doubt this... why would you bother programming it this way.

If I had to choose - I would pick "A" in some variation of trying to get to 0,0

What would be interesting to see is if people who experience flyaways could tell us if they were in the direction of 0,0 - That would tell us (perhaps) that it is trying to get to a "home point" that wasn't set and was possibly defaulted to 0,0?

I dunno - just occurred to me reading all of this.

 

[wattage]

Not too sure about this. Most of the flyaways I've seen on the web show a Phantom behaving erratically, flying at scarily steep angles of tilt or roll at abnormally fast speeds. A Phantom having an inaccurate home point it thinks is outside its limits and trying to return to home should exhibit a smooth controlled flight towards this fictional home point, not like the 'joyriding teenager' footage that accompanies most flyaways on YouTube.

I think it's something to do with the compass data getting corrupted or hit with interference. Didn't DJI upgrade the compass module recently by putting it inside an anti-static enclosure, without an explanation of why they deemed this necessary?

 

[Buk]

Steve not sure if this impacts your theory for some flyaways, but being in Phantom mode versus Naza-M mode the Phantom responds differently to the limits. I'm not sure I'm using the manual you are, but in Phantom mode being in Ready To Fly (Non-GPS status) there is no horizontal limit. Quoting the the manual I'm reviewing it states only the height limit works, "In Ready to Fly status, height, distance limits works together to restrict the flight. In Ready to Fly (non-GPS) status, only height limit works and the flying height restricted to be not over 120m."

Question: When you fly outward and hit a horizontal limit, right and left flight is still possible. You just follow the circumference of the circle. If you are beyond the limit would you be able to fly right and left while being "pulled" towards home and travel in a spiral curve toward home??

 

[Narrator]

On the GPS fault issue....

I was away onsite the other day, looking at Google Earth on my phone. I told GE to find my current position. It pinpointed me about 2km away from where I actually was. I'm not sure if this is a GE issue or a GPS fault, and am wondering if that would happen to the Phantom.

On the Limits Setting, Distance tab...

If I don't make any changes, does that mean no limits are set?
And does zero (0) on the horizontal distance mean no horizontal limit?

EDIT: The manual I'm using is v1.6 and Steve's quote is on p30

 

[msinger]

Or disable the horizontal limits. (Is this even possible?)

Wouldn't this be sufficient?

 

[Hughie]

Or disable the horizontal limits. (Is this even possible?)

Wouldn't this be sufficient?

That depends on whether the data type used to store this in the phantom is large enough. If you can read the same number back later from the phantom then there is a good chance it would be ok.

 

[msinger]

Hughie, excellent point! Hopefully the developers capped those at the appropriate max values.

 

[ianwood]

There is no evidence to support the theory of a home point bug, or at least I have seen none. I'd like to see an OSD screen or app screen showing the Phantom is 10,000m away with 97% battery remaining. Or a picture of the OSD/app with a large distance number and the Phantom sitting right there.

The Phantom requires 7 SVs to record a home point. That's a PDOP of 4 or better and more likely 3.5 or better. Those kind of PDOP values translate into a small margin of error, nothing of the nature that would lead to a fly away.

If that Youtube video is the same one I am thinking of (can't see it as Gogo blocks Youtube), it's irresponsible and should be taken down.

 

[msinger]

The Phantom requires 7 SVs to record a home point

What if you change locations and take off before the home point is marked in the new location? Would the Phantom still remember the previous home point?

 

[PhantomFanatic]

Applying your scenario...

Only if the home point can be erroneously set/reset through some kind of 'bug'.

Because if it were to become 'confused' as to it's current location due to GPS error/bug as you have proposed the home point remains the same. The return path back to the limit is still a straight line to or towards the [original?] home point.

or...maybe I don't fully understand

my $0.02

I agree. As long as the home point is still set, it shouldn't matter if the GPS signal gets scrambled for whatever reason. It still knows how to fly back to the home point, even if it has no clue where it is.

But, I might have missed something. However, I do commend you for starting this brain storming session.

 

[SteveMann]

EDIT+1: This link https://www.youtube.com/watch?v=_ge3GuhEDRM
The whole theory is based on the homepoint not being set properly in the first place.

The author of this video doesn't understand how GPS works. If your home location is wrong when you start flying, it will immediately head to where it thinks the fence is located. There won't be an opportunity to fly around to "calibrate" the GPS.

My theory is based on the current position data being corrupted, but a corruption of the home point could so it also. widening the limit radius to a very high number would help reduce this type of flyaway.

 

[Fyod]

On the GPS fault issue....

I was away onsite the other day, looking at Google Earth on my phone. I told GE to find my current position. It pinpointed me about 2km away from where I actually was. I'm not sure if this is a GE issue or a GPS fault, and am wondering if that would happen to the Phantom.

On the Limits Setting, Distance tab...

If I don't make any changes, does that mean no limits are set?
And does zero (0) on the horizontal distance mean no horizontal limit?

EDIT: The manual I'm using is v1.6 and Steve's quote is on p30

Phones are much different.
Because it takes a while to pinpoint accurate GPS positioning, phones use Wifi, GSM, GPS, usually in that order to mark your position. That is why "hopping" sometimes occurs.
Wifi - If any network is in reach, your position is matched with a database of SSIDs with previous GPS-SSID matches. This position is usually within 30 feet accuracy, but works in buildings too.
GSM - Your position is triangulated based on 3 or more surrounding GSM towers. If only one is available, or it takes the others a while, you may see your position near where the first tower is.
GPS - Basically works like on the Phantom.

There certainly needs to be a default value for home position. A value in the memory that is either 0, the last known position, or some weird default.
This value should be overwritten every time homelock is established.
An error could occur either if the default location overwrite failed or a mistake happened in the value. The best way to rule this out would be to register home position twice, compare the values, flash green only if both values match, otherwise retry acquiring the homelock position.
It is not uncommon for corrupt data happening on high data rate CAN networks. Car manf. usually secure this by repeating the message through the network over and over and over again.

 

[SteveMann]

I think it's something to do with the compass data getting corrupted or hit with interference. Didn't DJI upgrade the compass module recently by putting it inside an anti-static enclosure, without an explanation of why they deemed this necessary?

Bad compass data wouldn't cause a fly-away. If you told your Phantom to go home and it flies in a straight line, then your compass is OK. If it flies in an arc to home, then your compass needs recalibrating. No compass data would still go home, but it would be a spiral route.

I am completely mysterified what they mean by "anti static" enclosure for the new compass since the compass is not affected by static.

 

[SteveMann]

There is no evidence to support the theory of a home point bug, or at least I have seen none. I'd like to see an OSD screen or app screen showing the Phantom is 10,000m away with 97% battery remaining. Or a picture of the OSD/app with a large distance number and the Phantom sitting right there.

I have asked some victims of fly-aways if they happened to notice their distance to home, but only one recalled and it was only 180 meters, so my theory was not proven.

 

[SteveMann]

I agree. As long as the home point is still set, it shouldn't matter if the GPS signal gets scrambled for whatever reason. It still knows how to fly back to the home point, even if it has no clue where it is.

But, I might have missed something. However, I do commend you for starting this brain storming session.

No, if the current location is screwed, the Phantom has no accurate starting point to fly home. Calculating the heading to fly to a waypoint requires two positions - the current position and the destination position. Look at it this way, if you are in Chicago and the Phantom GPS thinks it is in Florida due to messed up GPS data, it will calculate a new heading to get home, which will generally be north. It doesn't matter where the Phantom really is if it thinks it is in Florida - it will fly north to get to Chicago.

Here's the formula: θ = atan2( sin(Δλ).cos(φ2), cos(φ1).sin(φ2) − sin(φ1).cos(φ2).cos(Δλ) )
Ouch, my brain hurts.

 

[SteveMann]

Or disable the horizontal limits. (Is this even possible?)

Wouldn't this be sufficient?

That is almost 19,000 miles, but I would worry that the firmware programmer didn't anticipate an unusually large number and you may overflow the data into another parameter. I wonder if zero would effectively turn off the radius limit?

 

[Zinnware]

I once witnessed a fly-away when my brother took off on his second flight after a quick battery change. (Phantom 2 -non-vision version 1.08) He had just lifted off and put it into a hover when It took off at high speed over a lake. He was not hitting any controls at the moment when it took off. I know this because I thought it was odd that he was so bold to fly that fast over the lake on his second flight. I asked him twice if he was hitting any controls and he confirmed that he was not hitting any controls. I quickly asked him to pass over the controller and switched it from GPS mode to ATTI mode and brought it back safely and landed it.

His fly-away was caused by not following a good start-up procedure and not getting a good home lock before taking off. I believe that the only way to recover from a fly-away where your controller still has a connection is to get it off GPS mode and into ATTI mode. In my brother's situation, putting it into HL (Home Lock) would not help since the Phantom obviously would not have a good home lock.