Is so called 'flyaway' possible at all?

[Andy9]

Very often we can read about the drones that flies away and not respond to commands.
But according to technology the drone will responds in two ways. Loosing the signal it will switch to the return to home mode or hoover in place, depends on how it was preset.
Loosing the GPS connection it will switch to the ATTI mode. I don't know if there is a third option related to the compass error. But in such case switching to ATTI should resolve the issue.
All those 'flyaways' mostly are related to unrecognised ATTI mode and the drone is simply blown away by the wind. And this mainly happened when the operator has not visual contact.
The only 'real' flyaway is maybe a serious radio signal interference but this is rare I believe.
Am I right?

 

[Deleted member 62848]

Sounds logical to me unless someone has a gravity pull gun like star trek
It might be like when the bike throttle when slides stick open on full noise,,RTH off to China,lol

 

[PhantomWetSuits]

Very often we can read about the drones that flies away and not respond to commands.
But according to technology the drone will responds in two ways. Loosing the signal it will switch to the return to home mode or hoover in place, depends on how it was preset.
Loosing the GPS connection it will switch to the ATTI mode. I don't know if there is a third option related to the compass error. But in such case switching to ATTI should resolve the issue.
All those 'flyaways' mostly are related to unrecognised ATTI mode and the drone is simply blown away by the wind. And this mainly happened when the operator has not visual contact.
The only 'real' flyaway is maybe a serious radio signal interference but this is rare I believe.
Am I right?

There is another : One that is caused my Magnetic Interference that does switch it into Atti mode but also Yaw Errors making impossible to control the drone in Atti Mode:
The saving grace when this happend to me twice , DJI replaced both my Phantoms with 100% coupons . So looking for that Yaw Error is critical in Atti Mode.

Phantomrain.org
Gear to fly your Phantom in the Rain
Coal

 

[PhantomFandom]

I don't know if there is a third option related to the compass error. But in such case switching to ATTI should resolve the issue.

No, switching to ATTI mode will not help if you have a true compass issue either where the compass does not work (rare) or where you took off with significant magnetic interference. You will have zero control over the direction of the drone if it cannot get an accurate compass reading. It is a "fly away" only in the sense that the pilot could not control it at that point. However not truly a hardware/software failure "fly away" since it would be pilot error during the launch.

 

[quaddamage]

The only 'real' flyaway is maybe a serious radio signal interference but this is rare I believe.

These drones are a complex network of processors and sensors. many, many things may go wrong. But I'm pretty sure over 90% of fly away cases is as you said - just a switch to ATTI mode. Though there may also be many reasons for the switch to ATTI.

For possible reasons for the very rare cases of non-ATTI flyaways:
- Flight Controller may crash, re-start and continue working on a corrupted state - trying to adjust to invalid position
- GPS may start sending invalid data, which FC will use to adjust the flight
- A packet received from RC might be corrupted and contain a command which changes one of many FC flight options, or overwrite part of the current flight state
- One of the non-essential components may hang and start flooding internal network within the drone, making new packets from RC impossible to propagate
- IMU may malfunction, telling the FC that drone moves in certain direction, or moves faster than allowed limit; the drone will try to compensate that and start really moving.
- Sticks on RC may develop a bias and report non-zero position even if not touched

That's just a few from the top of my head.

There are precautions for most of the above, which try to detect such things and avoid bad behavior - often by just switching to ATTI. But the detection isn't perfect.
For example, corrupted GPS data - the drone will ignore a packet telling it it's suddenly on the other side of the world. But if the position is within certain tolerance, it will be accepted, even if invalid.
Or the RC packets - they do have a checksum, which can detect if the message was corrupted. But there's a chance that the packet will get corrupted in a way which passes checksum verification.
Or sticks - the RC will detect the bias when it's being turned on. But if the bias develops after the RC is enabled - the RC will use that biased data.

Also, these precautions were introduced over time - in Ph1/Ph2, some of them do not exist - early phantoms could really decide to fly back to China.

 

[Andy9]

Probably at compass error you get the message about that at startup.
The compass data is telling to IMU at which direction the drone is flying. But isn't there any logical circuit in the IMU that tells which two rotors are at the side where the camera is facing to and which ones are the opposite. If you put the stick forward it isn't necessary to have a compass data to fly in the direction of front motors whatever direction it would be. This would be logical and therefore as long as you have the signal you should control the drone - if you see it! If not, then you have an unsolvable problem. Only navigating is not possible as the drone even if it flies forward it would not know in which direction forward is heading to.
This all is maybe wrong and I would like to clear it if somebody can.

 

[PhantomFandom]

Probably at compass error you get the message about that at startup.
The compass data is telling to IMU at which direction the drone is flying. But isn't there any logical circuit in the IMU that tells which two rotors are at the side where the camera is facing to and which ones are the opposite. If you put the stick forward it isn't necessary to have a compass data to fly in the direction of front motors whatever direction it would be.

No it is not that simple. Since every movement of a quadcopter involves complex calculations of motor speeds, it's not just a matter of "going forward". The FC is constantly comparing input from all the sensors to come up with its current spatial orientation, direction, velocity, altitude, etc... If the compass is incorrectly telling the FC which direction it is facing, it disagrees with the other IMU instruments (gyro, accelerometer), it is forced to try and correct the situation. This "correction" actually makes the situation worse since the compass is not reliable and you end up having no control over the drone.

 

[Meta4]

"Flyaway" usually just means: I lost my drone (or lost control of my drone) and don't understand why.

The only 'real' flyaway is maybe a serious radio signal interference but this is rare I believe.

At worst, signal interference will only swamp your control signal and nitiate RTH.
Interference won't fly your drone away.

 

[ProPhotogPaul]

Those of you who think that a flyaway isn't possible have just never experienced it. I have. On a brand new P4P2. Twice it began flying away and up with all indicators showing normal. After the first time I tried again in a different locality. Same thing. Repeatedly, I would try to get it to come back and it would for a few feet and then it would ignore the command and continue flying up and away. ATTI mode did not change it. RTH was ignored. With my heart in my throat, I continued to nurse it downward which again, it would do for a few feet and then start climbing again. For about 10 minutes I kept doing that and was eventually able get it to the ground. DJI looked at the flight data and decided to replace the drone and controller. The new drone never did it again.

 

[John Locke]

Just to clarify, if you lose GPS signal, such as flying in a narrow canyon, flying through a tunnel, or under a wide cement bridge, the craft automatically goes into ATTI mode. This is announced visually in the Go4 app, (upper left corner, which is normally green) it will not announce audibly (which is an app flaw IMO). If the craft goes into ATTI mode due to loss of GPS, there is no need to switch to ATTI to control the drone, it's already in ATTI. IMO, it's best to leave the craft in P mode, so when GPS signal is reacquired, automatically the flight controls will go back to normal, in P mode. What is necessary when the craft auto-switches into ATTI mode is the SKILL to fly in ATTI mode via FPV, Or, if you have VLOS of the craft, you need the skill set to fly in ATTI via VLOS. Assuming you have a video feed to fly FPV, you do have ATTI control, and you can easily fly out of your situation to reacquire GPS signal. Controlling in ATTI isn't difficult, if you know what to expect. Slip and slide is what I call it. It's important to practice ATTI mode in an open field so you understand how wind affects the craft, and the fact you don't have automatic brakes, OA sensors, or XY fixation when you have your hands off the sticks. Fortunately you have Z stability, as the altitude is controlled with a barometer sensor, not GPS. Learning ATTI mode flight skills is important IMO. Those with Mavics don't have an ATTI option switch to facilitate practice, so that's problem, unless you hack the firmware to acquire ATTI mode capability.

 

[John Locke]

Those of you who think that a flyaway isn't possible have just never experienced it. I have. On a brand new P4P2. Twice it began flying away and up with all indicators showing normal. After the first time I tried again in a different locality. Same thing. Repeatedly, I would try to get it to come back and it would for a few feet and then it would ignore the command and continue flying up and away. ATTI mode did not change it. RTH was ignored. With my heart in my throat, I continued to nurse it downward which again, it would do for a few feet and then start climbing again. For about 10 minutes I kept doing that and was eventually able get it to the ground. DJI looked at the flight data and decided to replace the drone and controller. The new drone never did it again.

DJI doesn't test fly their craft before packaging. Your situation is a case of a manufacturing flaw, or a possible component failure in manufacturing, and that can happen. The first flight is important to test things out. It's rare to see an actual flyaway in DJI craft, assuming it's manufactured correct. Anything is possible for component failure though, but it's pretty rare IMO.

I am often flying over places I could never retrieve the craft if it went down. In 5yrs I've been lucky, I haven't had that happen. But if it does, I'm OK with it. I've enjoyed the craft enough to justify buying another one.

 

[PhantomFandom]

Those of you who think that a flyaway isn't possible have just never experienced it. I have.

I never said it isn't possible. Just that it is rare and that in most instances where a pilot reports a fly-away, it ends up being pilot error. That is why the flight logs are important, both here to get the full story and with DJI in order to make a warranty claim.

 

[PhantomWetSuits]

Those of you who think that a flyaway isn't possible have just never experienced it. I have. On a brand new P4P2. Twice it began flying away and up with all indicators showing normal. After the first time I tried again in a different locality. Same thing. Repeatedly, I would try to get it to come back and it would for a few feet and then it would ignore the command and continue flying up and away. ATTI mode did not change it. RTH was ignored. With my heart in my throat, I continued to nurse it downward which again, it would do for a few feet and then start climbing again. For about 10 minutes I kept doing that and was eventually able get it to the ground. DJI looked at the flight data and decided to replace the drone and controller. The new drone never did it again.

One thing DJI does look for is if you did try to Control the Drone in Atti Mode , If they see you moved the sticks and that nothing happened to to Yaw Errors , this than changes the perspective of DJI and your likely to be in Favor of DJI as we were.

Thus the reason why we got Two new Phantoms from DJI , care of the Yaw Errors. So Pilot error can be determined pretty quickly when they see that no stick movement was involved and the drone just blew away...in Atti mode because the Pilot did not understand what is needed to control the drone in the wind without the help of GPS.

But you our correct, Unless you have Experienced it as I have twice. It will most likely be blamed on Pilot error for having flown in magnetic interference in the first place.
But the reality is Atti mode is not flawless.

 

[David_Cambridge]

I agree that this is a complex situation as the flight controller is charged with integrating many different inputs successfully (GPS position, inertial guidance info of both velocity and acceleration in 3 dimensions for both, altitude, and compass heading), as well as user input from the rc.

Uncontrolled flight is possible when this solution is not successful due to conflicts between the inputs. Looking at the inputs:

GPS errors - overwhelmingly just no position considered valid and automatic switch to ATTI mode. Actual reported position errors are vanishingly rare.

Inertial guidance system errors - vanishingly rare unless an actual hardware failure of the MEMS chips that power it.

Altitude errors - common. The altimeter is based on changes in air pressure from the pressure recorded at takeoff, and varies a bit as time changes. This cannot create a fly-away situation, but can lead to crashes from pilot choices made reading the reported altitude.

Compass errors - this is a common and well documented problem which generally results in an erratic flight followed by a crash. It results when the compass setting at drone power-up is wrong due to some external magnetic field close enough to the drone’s compass to report magnetic north being in a different direction than it actually is. When the drone takes off and leaves that field distortion, the data it feeds the flight control system is wrong, either a little bit or a lot. If it is alot, the conflict between what it is telling the flight controller the drone is doing, and the data from the other sensors cannot be resolved, and the drone crashes.

Another condition can cause what feels like a flyaway is the loss of control signal from the rc to the ac. This occurs after the video downlink fails. Loosing the video downlink is a clear early warning of pending control signal loss.

If ignored, when the control signal is lost, several things can happen:

If the drone has GPS, it will execute the user programmed RTH until control signal is regained, at which point the pilot can choose to cancel RTH and go back to flying the drone manually. This works only if RTH is set to return to home. If it is set to hover, it will just stay there until it runabout of battery and lands.

If GPS is not valid, it switches to ATTI mode. At this point, the drone does nothing as best as it can. It will drift with the wind and wait for commands.

All of these conditions can give rise to the perception of a fly away. Being the pilot in command brings a lot of responsibility.

 

[Bone4Drone]

These drones are a complex network of processors and sensors. many, many things may go wrong. But I'm pretty sure over 90% of fly away cases is as you said - just a switch to ATTI mode. Though there may also be many reasons for the switch to ATTI.

For possible reasons for the very rare cases of non-ATTI flyaways:
- Flight Controller may crash, re-start and continue working on a corrupted state - trying to adjust to invalid position
- GPS may start sending invalid data, which FC will use to adjust the flight
- A packet received from RC might be corrupted and contain a command which changes one of many FC flight options, or overwrite part of the current flight state
- One of the non-essential components may hang and start flooding internal network within the drone, making new packets from RC impossible to propagate
- IMU may malfunction, telling the FC that drone moves in certain direction, or moves faster than allowed limit; the drone will try to compensate that and start really moving.
- Sticks on RC may develop a bias and report non-zero position even if not touched

That's just a few from the top of my head.

There are precautions for most of the above, which try to detect such things and avoid bad behavior - often by just switching to ATTI. But the detection isn't perfect.
For example, corrupted GPS data - the drone will ignore a packet telling it it's suddenly on the other side of the world. But if the position is within certain tolerance, it will be accepted, even if invalid.
Or the RC packets - they do have a checksum, which can detect if the message was corrupted. But there's a chance that the packet will get corrupted in a way which passes checksum verification.
Or sticks - the RC will detect the bias when it's being turned on. But if the bias develops after the RC is enabled - the RC will use that biased data.

Also, these precautions were introduced over time - in Ph1/Ph2, some of them do not exist - early phantoms could really decide to fly back to China.

Gosh, why was I fretting about COVID-19 when there was so much to worry about with my drone heading for China! After reading this post, I took some fishing line and dangled my P4P from my ceiling. Now I just take my controller in hand and visualize I'm flying over Niagara Falls, Mt. Rushmore, etc. while watching video recordings by others on my computer. This is great!! Watching Apollo mission vids, so far, I'm the only pilot to ever land his DJI on the moon!

 

[3rdof5]

I've never had a flyaway but I have seen my P2 take off completely on its own. It only happens in GPS mode. Atti works perfectly normal. But if i do a motor start in GPS mode I dont dare turn my back on it. Suddenly the motors will speed up to the point it takes off. I suspect a faulty IMU which is thinking its falling when sitting on the ground with motors running and it tries to compensate by climbing. Makes no sense, it shouldnt be possible, but it does it. My P3 has never misbehaved.

 

[makneyse]

I've had two fly-aways and they invoke mini heart attacks. Both with P4PV2.0s. One could be attributed to pilot error as I turned the drone on in a downtown area on a sidewalk and momentarily received a compass error. The error went away so I went ahead and took off, hovered for a few seconds and then if flew away from me, directly into a building and then crashing to the ground. I literally had three seconds and my even though I swear I pulled down on my throttle, the data log says I did the opposite. Surprisingly (and luckily) DJI replaced it.

And just a couple weeks ago I was flying over a baseball field I've flown numerous times and had no errors to speak of. I was flying around a scoreboard in the outfield and the drone started flying directly for it. I immediately switched out of P mode and the drone stopped in its tracks and began to drift away from the scoreboard which was the direction the wind was blowing. I have no idea why that instance happened but am so glad my eyes were on the drone when it happened and that I had learned from my previous mistake (and this time knew to flip the switch first).

Looking back at some of the flights I've had (400 ft in the air next to skyscrapers, for example) it worries me that fly-aways are a real possibility. I also now know that when flying in a RF crowded environment where GPS blockage is also possible, it is best to fly in Atti mode and have VOs for an added layer of safety.

And outside of fly-aways these birds do other strange things. One day I had a drone RTH twice (with two different batteries) when there was plenty of juice left and I was standing next to it. And one evening my drone refused to land (again with 2 different batteries). It just hovered a foot above the ground. And I had lots of time to try every possible solution I could think of.

As people said, a lot can go wrong. I get it that a lot of the time it may be pilot error but many times it is not. Or at the very least, the pilot might not know what the solution is for something they've never experienced before.

 

[3rdof5]

I've had two fly-aways and they invoke mini heart attacks. Both with P4PV2.0s. One could be attributed to pilot error as I turned the drone on in a downtown area on a sidewalk and momentarily received a compass error. The error went away so I went ahead and took off, hovered for a few seconds and then if flew away from me, directly into a building and then crashing to the ground. I literally had three seconds and my even though I swear I pulled down on my throttle, the data log says I did the opposite. Surprisingly (and luckily) DJI replaced it.

And just a couple weeks ago I was flying over a baseball field I've flown numerous times and had no errors to speak of. I was flying around a scoreboard in the outfield and the drone started flying directly for it. I immediately switched out of P mode and the drone stopped in its tracks and began to drift away from the scoreboard which was the direction the wind was blowing. I have no idea why that instance happened but am so glad my eyes were on the drone when it happened and that I had learned from my previous mistake (and this time knew to flip the switch first).

Looking back at some of the flights I've had (400 ft in the air next to skyscrapers, for example) it worries me that fly-aways are a real possibility. I also now know that when flying in a RF crowded environment where GPS blockage is also possible, it is best to fly in Atti mode and have VOs for an added layer of safety.

And outside of fly-aways these birds do other strange things. One day I had a drone RTH twice (with two different batteries) when there was plenty of juice left and I was standing next to it. And one evening my drone refused to land (again with 2 different batteries). It just hovered a foot above the ground. And I had lots of time to try every possible solution I could think of.

As people said, a lot can go wrong. I get it that a lot of the time it may be pilot error but many times it is not. Or at the very least, the pilot might not know what the solution is for something they've never experienced before.

Well said, talking about having to be experienced in Atti mode, I bought a P3P as my first real quad and I'm ashamed to admit I never practiced Atti mode with it, GPS never failed. However, I later aquired an abused P2V+ that only flies with any stability at all in Atti mode. GPS mode it has a mind of it's own and cannot be trusted. So then when I learned Atti mode and I cannot agree more that all pilots need to learn it. I now love to fly my P3 in Atti mode just to have fun.