Recovered from a flyaway

[SkunkWerx's]

I am in agreement with You and Alan 100% , while were still on the subject of "TRY Away" lollll , I really like the name you gave it !!....
From all that I read on this matter -- Up is the only way to go until Full control is regained . All that experienced this issue say they lost the
Horizontal Stick control , Then there are people that say you still have Horizontal Control but it's not that noticeable because of wind ?
This is what I have a problem understanding , The Mavic is capable of flying 40mph so surly it should have enough power to fly with noticeable horizontal control ? I've never experienced The "TRY Away" / ATTI mode while I fly, but I'm fully aware that it can happen .
As far as a correction to be made with a large jump in GPS location, the Mavic may not even have one, and is the Million dollar question
now that you mention it -- I've been trying to obtain information like this for a while and it would have to do with the FW -- DJI is pretty tight
lipped about this kind of information -- If they were to give this info to the owners, a lot of the Mavic's ongoing problems would become
nonexistent -- even a block diagram or flow chart for some parts of the system?
Please feel free to speak of your thoughts about the Bird going into ATTI with a large number of satellites with no GPS
I too am interested in this matter along with some others of course. This is a very interesting and complicated subject especially when there's no Documentation on it -- One idea that comes to my mind is --> Maybe it's operation is like a hand held GPS where it shows you
that you have say 10 satellite's in range it really only needs 3 to triangulate for your location the other satellite's get used if you loose 1 of out of the 3 that are needed -- Now I'm Thinking -- (My Brains Working OT) You smell any wood over by you ?? loll Oh almost forgotten the Mavic also uses the GPS for it's Speed telemetry maybe that's why DJI recommends you have a minimum of 6 satellite's showing on the controller

 

[dewster]

Having the ability to disable GPS and fly attitude mode would be a great help DJI.

 

[geigy]

Having the ability to disable GPS and fly attitude mode would be a great help DJI.

The phantoms do that - DJI may have taken that option away because so many crashes and fly-aways are the result of accidentally setting ATTI mode, then having the wind or poor piloting result in losing the bird. Perhaps that is why it is not available for the Mavic. Until everything works perfectly every time, flying with sticks alone will remain a good skill to have.

Regarding my theory about the software behind "try-aways", it's fairly simple. Once the Mavic detects a sudden large increase in position error, it goes into ATTI mode, but continues to move, as best it can, toward the location it thinks it should be anyway, and ignores the sticks until that move is accomplished. The number of GPS satellites remained at its original number (12 I believe), but the bars went to zero, indicating in effect that the Mavic does not trust GPS at the moment. This remains a theory at the moment.

You are right it would be nice for DJI to make some of this logic available to the public, so that we can better predict what our drones will do.

I have attached a log file of my try-away that may show this behavior. On my iPad the GPS location makes a sudden jump, sits there ignoring pilot input, and then responds again after the GPS bars are restored. On android (running DJI Go 4 on the same file), the GPS location makes a smooth, slow transition from the old to the new position, reflecting the actual physical path that I remember the Mavic taking.

 

[AlanTheBeast]

I agree with Alan that the drone should filter out large GPS changes, but it's a tough call. They obviously filter out small GPS errors with standard low pass filter techniques, but what about a large jump in GPS location? It's an interesting question. I have some thoughts on this: where the bird goes into ATTI mode with a large number of satellites but no GPS bars.

I don't know if DJI filter the GPS position or leave that to the GPS sensor. (They usually have various filtering modes depending on application or speed).

IMU data alone could confirm if a step change in GPS is too large[1]. If too large, then converge slowly from the old position (before the step change) towards the new solution over some reasonable time (say 100 m per minute).

I'm guessing the IMU used in the control loops is not used at all in the navigation package which seems dependent on the mag compass (of all things) for heading reference. Again, once underway the heading gyro should be used instead with the mag compass and GPS to discipline the gyro.

It would be wonderful if someone at DJI got into Kalman filters. There are many sensors on board and their characteristics are likely well modelled. A well designed Kalman filter could surely improve the navigation system (that does not mean " more accurate ").

[1] The reports of DJI products dropping GPS even in benign conditions troubles me a lot. They use GPS sensors destined for mobile phones and such. These receivers are made to be extremely sensitive so they will work in a coat pocket or in a car, etc. Such sensitivity is not required in an airborne product. While it may help a lot in tree'd area clutter, it also increases sensitivity to multipath and interference sources which of course are most prevalent in urban settings.

A GPS on a drone is (for most of the flight) in very benign conditions and there is no reason it should drop out. I suspect that DJI's GPS implementation is buggy. (Reports of it showing full bars but saying GPS out pop up here and there...)