Please select Into the mobile phone version | Continue to access the computer ver.
software bug causing flyaway / rogue RTH at 5-6 satellites coverage
12Next >
7984 43 2014-12-26
Uploading and Loding Picture ...(0/1)
o(^-^)o
ciprianboboc
lvl.3

United States
Offline

This is my aprox 10th flight. I've experienced a flyaway on my 3rd flight (aprox 10 days ago) and, at that moment I was not sure what was going on.
Today I've experienced the same problem twice. But this time, I had NAZA mode on and I was able to recover by flipping to ATTI mode.

I did compass calibration and took off after I made sure that I had home lock on (multiple green flashes). I had 6 satellites coverage when I took off.

I was in GPS mode, with 6 satellite GPS signal and noticed that once of sudden my Phantom decided to perform a straight flight toward me ignoring all my input. I switched to ATTI mode and regained control.
I switched back to GPS mode and after another 30 seconds, I had the same exact experience.

My flight was in a park, in a rather open space, but surrounded by some tall trees. The phone didn't say "return to home" and there were no yellow flashes to inform me on a planned return to home.
After those 2 flyaways (or rather "rogue" RTH), I eventually pushed the Phantom at a higher altitude and got better GPS coverage (7 and eventually 8). During the rest of my flight (over 15 minutes), I haven't experienced anything similar. Therefore, I doubt there was a hardware problem or radio interference.

I'm quite confident this is a software bug or even maybe a "feature". The only problem is that, the drone going by itself in one direction is a good receipe for a potential crash. My RTH altitude is set at 220 feet and my Phantom was at aprox 80 feet. It should've at least go up (to 220 feet) and not in one straight direction. I understand that droping to 5 satellites, it should change the Phantom in ATTI mode and not trigger a flyway (or a "wannabe" return to home)
I'm a software developer and this really looks like a software bug to me. It's up to DJI to take into account my report (even work with me on this issue if they want) or just simply ignore it. In the end,  I can still return my drone to the seller and get a refund as long as it doesn't have even a scratch on it (and it's still in its return interval).

All the firmware on my Phantom Vision 2+ and remote is up to date.

Main Controller 3.08
GPS 6.0.1.4
Receiver-5.8G  1.0.1.3
P330CB 1.0.2.10
Battery  2.6.0.0
Zen IMU 1.0.0.6
2014-12-26
Use props
ciprianboboc
lvl.3

United States
Offline

I need to add a few more details:
- my Phantom was aprox 180 feet away from me, direct line of sight - absolutely no obstructions between my RC and my Phantom
- KPI index today is 4 - therefore it's quite possible to have some GPS errors even at 6 satellites coverage

2014-12-26
Use props
ciprianboboc
lvl.3

United States
Offline

ciprianboboc@gm Posted at 2014-12-27 07:01
I need to add a few more details:
- my Phantom was aprox 180 feet away from me, direct line of sight ...

I've also sent an email to support and will wait to see if I get any help / response.
2014-12-26
Use props
tonyphantom147
lvl.4

United Kingdom
Offline

Correct, The manual recommends 6 sat locks minimum. I wait for 9 minimum, however I have read people are generally waiting for 7 sats.
Think you have answered your own question, GPS is not good where you fly. If it flys fine in ATTI then there is your proof. Interesting you didn't use IOC home lock, I find that much easier to come home if I loose orientation. I installed extra GPS copper shielding with good results.
2014-12-26
Use props
ciprianboboc
lvl.3

United States
Offline

My point is that, no matter what GPS coverage I have (even 0), the Phantom should not fly in one direction by itself without answering to my commands. The expected result should be an automatic switch to ATTI mode by itself (or manual mode if there are GPS issues) and the quad to respond to the user input. And it would be nice to get some sort of notification in the phone app.

Also, if I think more about this, it's quite possible that, in case of GPS issues, the quad will see itself at some distance from where it was before and try to go there to "maintain" position. Therefore, not a rogue RTH but the keep position routine. However, the software can be fixed to determine cases like this and not to try to make corrections when the distance from the previous known location exceeds a specific limit (the wind can push the quad only so much between 2 GPS readings and the interval of GPS readings should be known to the Phantom software).
2014-12-26
Use props
kenargo
First Officer
  • >>>
United States
Offline

I think the only method to have 100% control is to have a 'manual' switch setting.  I can understand why DJI doesn't fail-over the Phantom to manual control as a last resort but it would be nice to have something in-between, even if that means the Phantom detect inaccurate information from a sensor and changes to the best hover/lower maneuver it has given the state of the machine.
2014-12-26
Use props
jeepinocala
lvl.4

United States
Offline

ciprianboboc@gm Posted at 2014-12-27 08:34
My point is that, no matter what GPS coverage I have (even 0), the Phantom should not fly in one dir ...

The problem I see is you knew you had only 6 sats that is the bare minimum that is directed from dji so you sort of started off in a bad direction. If you are not getting sats do you think you could have had some bad interference?
2014-12-26
Use props
ljhellmich
Second Officer

United States
Offline

ciprianboboc@gm Posted at 2014-12-27 08:34
My point is that, no matter what GPS coverage I have (even 0), the Phantom should not fly in one dir ...

Was it actually pitched forward coming back to you in powered flight, or was it possibly drifting with the wind as if in ATTI mode?

Jamie
2014-12-26
Use props
ciprianboboc
lvl.3

United States
Offline

ljhellmich@yaho Posted at 2014-12-27 08:58
Was it actually pitched forward coming back to you in powered flight, or was it possibly drifting w ...

It was pitching forward and it was going against the wind direction - I was able to determine the wind direction by looking at the quad drifting when I was flying in ATTI mode. The wind speed was 3 mph  at ground level, and I was at aprox 80 feet altitude it a place shielded by some trees).
2014-12-26
Use props
ljhellmich
Second Officer

United States
Offline

ciprianboboc@gm Posted at 2014-12-27 09:08
It was pitching forward and it was going against the wind direction - I was able to determine the  ...

Hmm...  Other than the lack of climbing to RTH Altitude, it does resemble RTH behavior.  Good luck with getting it figured out, I'll stay tuned.
2014-12-26
Use props
ciprianboboc
lvl.3

United States
Offline

jeepinocala Posted at 2014-12-27 08:55
The problem I see is you knew you had only 6 sats that is the bare minimum that is directed from d ...

I don't think there was any wifi interference. The Phantom was immediately responsive after I switched to ATTI mode. I was also away from any high voltage lines, houses, away from obstacles, etc.
And "starting in a bad direction" is relative. I was able to observe what was going on and recover easy. "Bad" is if you have a crash, if you lose your Phantom, etc.
2014-12-26
Use props
Tahoe_Ed
Captain
Flight distance : 2605 ft
  • >>>
United States
Offline

ciprianboboc@gm Posted at 2014-12-27 09:21
I don't think there was any wifi interference. The Phantom was immediately responsive after I swit ...

I think you made the best of a difficult situation and did not crash your Phantom.  Congratulations on that.  6 Satellites and 4k solar activity is not the most ideal flying situation.  I have friends that will not fly if that is the case.  These are not military or commercial components with redundancy.  You don't say how many satellites you had when this was occurring.  If you have that information, I would be curious.  If not I understand there was a lot going on right then that had your attention.  
2014-12-26
Use props
droneflyers.com
First Officer
Flight distance : 60709 ft
  • >>>
United States
Offline

GPS mode is unreliable in a number of situations - solar storms, confined areas, areas with other interference, etc....

It's more of a hardware deficiency than a software bug, IMHO.

You may be a software developer, but the NAZA logic is the result of many 100's of thousands of man-hours as well as a lot of flight time and aeronautic engineers....not to say it can't be improved, but just that it's not as simple as squashing a particular bug.

I intend to buy a new Vision+ relatively soon and I am going to replace/shield the GPS antenna before I even start with it. You can see some of the problem - and possible solutions - here.
http://www.usdronerepair.com/dji-repairs-and-upgrades.html

Sure, DJI should upgrade the GPS and shielding....hopefully they will as they go. At the same time, they are under price pressure on everything they sell and make. So it's always a compromise.

I suspect part of the eventual answer may be to add in some cell tower triangulation like our phones use....to confirm and double check GPS.

2014-12-26
Use props
ciprianboboc
lvl.3

United States
Offline

Tahoe_Ed Posted at 2014-12-27 09:31
I think you made the best of a difficult situation and did not crash your Phantom.  Congratulation ...

Hello Ed, I said in my original post: "I was in GPS mode, with 6 satellite GPS signal and noticed that once of sudden my Phantom decided to perform a straight flight toward me ignoring all my input"
2014-12-26
Use props
Tahoe_Ed
Captain
Flight distance : 2605 ft
  • >>>
United States
Offline

My question was did that satellite number ever change?  Did you start or stop recording when this occurred?  
2014-12-26
Use props
ciprianboboc
lvl.3

United States
Offline

droneflyers.com Posted at 2014-12-27 09:35
GPS mode is unreliable in a number of situations - solar storms, confined areas, areas with other in ...

Hi Craig, I've read your Phantom book and I respect your knowledge on this subject.
My GPS cable is shielded, indeed. There were 8 satellites available at the time when I was flying. As I said, after increasing the quad altitude I got 8 satellite reception and have no issue anymore.

It's quite possible that a recent software update could introduce a software bug. I understand that NAZA logic is the result of a lot of research but that doesn't automatically translate into software which can't be fixed. If you want to fix something, you need first to understand what's going on. I hope a report like mine can help DJI developers.
Let's not forget that software companies like Microsoft or Apple will ask the users if they want to share their app info (logs) so they can determine later what was going on. Sadly, this is not the case of DJI. Eventually they'll get there - collect logs from users, reiterate and improve.
And it's quite possible that a recent "feature" or update in the flight logic may trigger all kind of issues.
It's a product made by humans and it will be improved by humans, no matter how complicated looks like. People were able to land on moon, Mars, comets and other space objects - and sorry, but NAZA mode is very simple by comparison.

The fact that I bought a P2V+ says that I considered it a good product. And I made this purchase after some decent research.
2014-12-26
Use props
ciprianboboc
lvl.3

United States
Offline

Tahoe_Ed Posted at 2014-12-27 10:01
My question was did that satellite number ever change?  Did you start or stop recording when this oc ...


I haven't noticed the satellite number to change (drop) until or during those 2 incidents (in that flight session). I'm not 100% sure if there was a change because I usually fly by looking at the quad. The first incident occurred during my first minute of flight time and I'm sure that at that moment the phone was indicating a 6 satellite coverage.  As I said, eventually the GPS coverage increased when I increased the quad altitude. A few minutes later I got the quad in the same location (same altitude or even lower) and the GPS coverage was at 7 (but in a few minutes the satellites location change so that doesn't say too much).
I was not in record mode and I even didn't have a SD card in my Phantom. The battery was fully charged. The RC extender was fully charged, too.

Also, in previous flights, I haven't noticed any interference with GPS reception when I start/stop recording. I know other people experienced that, but my quad doesn't seem to have that issue.
2014-12-26
Use props
Daninho
Second Officer
Flight distance : 70203 ft
Germany
Offline

Why everybody blames the GPS for Fly aways? If GPS is the reason then the phantom would work in atti mode but it isnt like that. I think the Sensors in the flight controller went crazy from time to time. Thats why after a hit or hard landing the phantom goes nuts. Maybe a stuck sensor or it gets decalibrated.  
2014-12-27
Use props
ciprianboboc
lvl.3

United States
Offline

Daninho Posted at 2014-12-27 19:41
Why everybody blames the GPS for Fly aways? If GPS is the reason then the phantom would work in atti ...

In my case described here, my Phantom worked fine in ATTI mode. And fortunately I haven't had a crash or hard landing yet.
When you write software depending on GPS input, you have to account for possible errors - the satellites are moving across the sky, the quad is also moving and, once of a sudden, can get bad readings, etc. Compass readings can have a lot of errors too. By comparison, gyro and accelerometer readings are very reliable. I have written software using readings for these kind of sensors (on phones) and my opinion is based on that experience.
What bothers me is not the fact that there are some GPS errors, but because, in suboptimal conditions, the quad is sent into a totally "autonomous" flight in some random direction ignoring user input. It seems that switching to ATTI mode recovers from that bad state, but I'm not 100% sure of it.
2014-12-27
Use props
Daninho
Second Officer
Flight distance : 70203 ft
Germany
Offline

ciprianboboc@gm Posted at 2014-12-28 00:12
In my case described here, my Phantom worked fine in ATTI mode. And fortunately I haven't had a cr ...

there were fly aways in ATT mode as well, so this cant be the reason
2014-12-28
Use props
ciprianboboc
lvl.3

United States
Offline

Daninho Posted at 2014-12-28 21:25
there were fly aways in ATT mode as well, so this cant be the reason

I'm talking about my own experience in this thread and I'm not saying that a software bug may be the only cause for all the flyaways or crashes... A flyaway in ATTI mode can be triggered by hardware failure, pilot error or other reasons.
For now, in my case, I rule out hardware issues (except maybe a weak GPS antenna), because I had other flights without a glitch after these incidents.

If I were a quad manufacturer, I would try to collect data logs from users to understand better what was going on in cases of accidents. As a software developer you need to get a repro to fix a bug (and get as much data as possible - software and hardware version, datalogs, etc). A diagnostic software may prevent hardware issues, before they occur. By collecting data when the quad is in a GPS mode in hover, you can measure how well the 4  motors are performing, if there are some sensor errors, etc. I know a team in charge of developing "preventive maintenance" for ATMs - that is they were analyzing data to figure out when ATMs may have problems in the future, in order to minimize maintenance costs and reduce the number of service trips for their support engineers.
2014-12-28
Use props
michael.haws
lvl.3

United States
Offline

If you can reproduce this problem at will then It may be worth it to get an iosd mark ii and use it's black box functionalilty.
2014-12-28
Use props
ciprianboboc
lvl.3

United States
Offline

michael.haws@ya Posted at 2014-12-29 10:39
If you can reproduce this problem at will then It may be worth it to get an iosd mark ii and use it' ...

My idea is to actually fly using an iOS app I'm writing to log the maincontroller data (satellite count, home location, drone location, altitude,  velocity, power level, etc). All that info is exposed by DJI SDK.
2014-12-28
Use props
michael.haws
lvl.3

United States
Offline

ciprianboboc@gm Posted at 2014-12-29 10:50
My idea is to actually fly using an iOS app I'm writing to log the maincontroller data (satellite  ...

Excellent idea,   probably could sell the finished product to a ton of dji customers.
2014-12-28
Use props
rwskinner
lvl.1

United States
Offline

You would actually think DJI would log this detailed information so they could find and fix the issues, if there are any.
Why can we get data logging from everyone except DJI, thru aftermarket apps?
2014-12-28
Use props
michael.haws
lvl.3

United States
Offline

rwskinner@ccwip Posted at 2014-12-29 11:02
You would actually think DJI would log this detailed information so they could find and fix the issu ...

You can get data logging from DJI using their dji iosd mark ii unit.
2014-12-28
Use props
ciprianboboc
lvl.3

United States
Offline

michael.haws@ya Posted at 2014-12-29 10:55
Excellent idea,   probably could sell the finished product to a ton of dji customers.

Actually if I finish such app, it would be a free app, according to my agreement with DJI when I got access to DJI SDK.
The app is actually easy to write, but some time needs to be spent to "polish" it and eventually submit it and get it approved on Apple Store. DJI SDK exposes the data I'm interested in, and after the app collects the data, it can send it into an email. I have the "send the email" code in place and also, DJI has a good sample that exposes all that data...
2014-12-29
Use props
michael.haws
lvl.3

United States
Offline

ciprianboboc@gm Posted at 2014-12-29 18:25
Actually if I finish such app, it would be a free app, according to my agreement with DJI when I g ...

It will be extremely interesting to see if it sheds any  light on your flyaway issue.    Keep us posted
2014-12-29
Use props
michael.haws
lvl.3

United States
Offline

ciprianboboc@gm Posted at 2014-12-29 10:50
My idea is to actually fly using an iOS app I'm writing to log the maincontroller data (satellite  ...

I'm betting the app you build will only  work for "vision" versions of the Phantom  and not the Phantom 2  (non-vision) ... or will it work for both????
2014-12-29
Use props
ciprianboboc
lvl.3

United States
Offline

michael.haws@ya Posted at 2014-12-30 03:01
I'm betting the app you build will only  work for "vision" versions of the Phantom  and not the Ph ...

The DJI SDK documentation doesn't say if main controller data is available on non vision models. I've asked this question on the SDK forum, maybe I'll get an answer there.
It may not work - because there's no iOS app for non vision models, and I don't think Phantom 2 has a wifi module inside. While the data may be available inside the quad, probably there's no way to connect to it and get it through an ad-hoc wifi network.
2014-12-29
Use props
michael.haws
lvl.3

United States
Offline

ciprianboboc@gm Posted at 2014-12-30 03:33
The DJI SDK documentation doesn't say if main controller data is available on non vision models. I ...

I suspect you would have to communicate with the dji 2.4g data link via the BTU unit in order to get the data.
2014-12-29
Use props
ciprianboboc
lvl.3

United States
Offline


I was able to get some data from my Phantom using my log application.

The data looks now like this:
timestamp(seconds)        satellites        home lat        home long        quad lat        quad long        altitude        battery        isFlying
1419895848        6        42.268176        -100.135908        42.268176        -100.135908        5.355148        3        0
1419895848        6        42.268176        -100.135909        42.268176        -100.135909        5.310593        3        0
[...]

More data is in the attached pdf file. This forum doesn't allow me to attach csv files, therefore I had to convert it to pdf.
I can probably gather even more data from the Phantom.

Just by looking at the data, I've found some interesting facts (they apply to my P2V+ with the latest set of updates):
- even if we have dynamic home point set, the quad doesn't seem to set a home point until it gets 6 satellites reception
- I'm not so sure about those reports of return to a wrong home if you take off with less than 6 satellites reception: in that case the home point is set to -180, -180 (and not a previous location)
- RTH seems to be triggered indeed if we don't close the motors and switch off the remote; but it works as expected: yellow flashes for a few seconds (probably 10) before RTH is triggered
- by just putting my hand over the GPS, it droped to 0 coverage; after that it came to 4-5 satellites coverage
- there is position data (as expected) when there is 3 satellites coverage; but that data probably is not precise enough - I need to analyze the data more to see what kind of precision is there

Eventually I need to try to repro my problem and see if there's a "jump" in GPS data (as expected). I haven't really been today in the park where I usually fly.

myFile.csv.pdf

39.22 KB, Down times: 24

2014-12-29
Use props
ciprianboboc
lvl.3

United States
Offline

Here's a result of getting this Phantom telemetry info (exported by my log app into a map):
https://www.google.com/maps/d/vi ... ji92ro.kL_4lkax_aAQ
Clicking on the markers gives more info (drone location, home point location, time stamp, altitude, etc).



I'm not very eager to fly with low GPS coverage to try to repro the issue I've noticed a few days ago. I got an answer from DJI support on my request and I need more (technical) info from them. We'll see where we get with that.

My opinion is that a logging feature can be useful for multiple scenarios:
- analyze flight patterns (I can imagine setting flight restrictions as predefined polygon shapes; in my case I don't want to go over the park limits and end up over private property. Right now, I can find when I'm at edge by pointing the camera down; but usually I go forward with the camera pointed forward so I can "see" potential obstacles)
- get logs data in case that something doesn't work as expected (a jump in GPS info can pinpoint possible software issues)
- get an improved "radar" functionality; while DJI Vision app displays the direction of the Phantom in a 2D space, that info can actually be rendered in a 3D space (taking into account the z axis, as well - basically display the flight vector)

Do you have any suggestions?
2015-1-1
Use props
michael.haws
lvl.3

United States
Offline

ciprianboboc@gm Posted at 2015-1-2 05:46
Here's a result of getting this Phantom telemetry info (exported by my log app into a map):
https:/ ...

Suggestions for what?   additional logging functionality?????
2015-1-1
Use props
ciprianboboc
lvl.3

United States
Offline

michael.haws@ya Posted at 2015-1-2 06:32
Suggestions for what?   additional logging functionality?????

I mean suggestions for other scenarios...
2015-1-1
Use props
lcarpenter
lvl.1

United States
Offline

Tahoe_Ed Posted at 2014-12-27 09:31
I think you made the best of a difficult situation and did not crash your Phantom.  Congratulation ...

looking for tahoe ed?  Is this you?  Had a fly away trying to get it fixed....please reply...
2015-1-8
Use props
Tahoe_Ed
Captain
Flight distance : 2605 ft
  • >>>
United States
Offline

lcarpenter@7q10 Posted at 2015-1-9 15:24
looking for tahoe ed?  Is this you?  Had a fly away trying to get it fixed....please reply...

I am here what do you need?  I have been kind of busy with CES so I apologize for my apparent absense.
2015-1-9
Use props
lcarpenter.emai
lvl.1

United States
Offline

Tahoe_Ed Posted at 2015-1-10 00:25
I am here what do you need?  I have been kind of busy with CES so I apologize for my apparent absen ...

Hi there.  I've been in LA with sick relatives and just back.  Would any of you still buy the DJI?  I loved it and yet I'm afraid that the next one may just fly to freedome too.  My sad story ends in that I took the bird to LA with me to get to an authrozied dealer to fix it as the damage wasn't too bad and the entire case was stolen from my car!  My drone hates me....  Good news is that I had insurance for theft....
2015-1-22
Use props
bjr981s
Captain
Flight distance : 139698 ft
  • >>>
Australia
Offline

guymacdonald@ip Posted at 2015-1-10 11:08
What is it about less than 6 satellites that the Phantom software cannot handle?

Why cannot the Ph ...

The number of satellites is in a way fairly irrelevant.

On a high Quality GPS system. (As I have installed on my Motor Yacht) the receiver is capably of receiving a specific number of satellite channels.

In my case 16. And it will often have a lock on 16 satellites. but it does not use them all. It determines the best 8 satellites that are in a position in the sky to give the most accurate GPS location.

The Display on my Autopilot system will show which satellites are used in the lock and which have a signal being received but are not used in the position fix as they are in a poor position in the sky.

The GPS accuracy is displayed in a HDOP (Horizontal Accuracy and VDOP (Vertical Accuracy) the lower the number the better.

I don't believe that the Phantom GPS does Sat selection. Hence it operates by a minimum number of satellites and assumes that 6 will give a good fix regardless of the position of the satellites in the sky.

Remember this is not a simple triangulation exercise as it must operate in 2 dimensions Horizontal and Vertical.

As an example imagine if it locked only 6 satellites and this six were in the same position in the sky.

The accuracy would be pretty low.

So I can understand why they have a minimum number of 6.

I also notice on the aspire that it has a different GPS receiver and the software does not display the number of sats but the accuracy of the GPS positioning.
The abnormal behaviour on the OPs return to home may be due to losing Vertical accuracy in the fix. So it would return to the X,Y, coordinates for the home position but as it could not determine its vertical position it would hold at that altitude return home and then land.



2015-1-23
Use props
gnixon2015
First Officer

United States
Offline

bjr is exactly right, they sacrificed 'minimum accuracy' for 'simplicity of design' from a coding complexity standpoint and i can understand why.  
2015-1-23
Use props
12Next >
Advanced
You need to log in before you can reply Login | Register now

Credit Rules