5 Ways to Prevent Flyaways

[Logger]

...I wonder if the forward and rear compasses disagree does the Mavic use the IMU and GPS data to decide which one is correct?
Rob

Yes I think that sounds perfectly reasonable. The IMU and GPS are still going to always need the compass for correct orientation regardless on startup.

 

[LeMur]

Moreover, there are TWO compasses

 

[triumphrider]

There's only one way to prevent a fly away, well other then not fly it. Just do what I do and tie a 4 mile long string to your MP.
Problem solved.

 

[finalfantasyfans]

Can't believe they miss one important point: Don't fly if it is too windy.

 

[AlanTheBeast]

Hey Alan I don't see how your newsflash refutes the need for a compass to determine orientation. How exactly do you propose the GPS derives the drones orientation? For arguments sake the drone is travelling east at 10ms. Can you explain to me how the GPS alone could tell if it is flying sideways forward or backwards. I feel you are confusing orientation (derived from a compass) with direction of travel through space. The three directions you speak of, are relative to the previous GPS position and nothing to with the the aircraft's heading, which the GPS cannot know unless a compass is involved.

Quite surprised if your really think DJI engineers have a minuscule understanding of sensor "products" and integration BTW

I do. Since they are not taking full advantage of the marvellous sensor suite on board. And other reasons***.

I never said that there was no need for a magnetic compass (BTW)**. Read completely before replying, please.

Other than at (and very near) 0 speed, GPS gives very accurate direction of movement information - as you say this is independent of body orientation and thus at powerup/takeoff the drone needs the mag compass for orientation.

This drone also has a very nifty IMU. 2 of them. The IMU has accelerometers and gyros.

So they have all the ingredients to ignore the magnetic compass* if they would integrate GPS and IMU. Frankly, if they're using the IMU to stabilize the vehicle and gimbal as well as they do, using the magnetic compass as a primary heading source is pretty strange as it is nowhere as accurate as the IMU gyros are, especially over short periods (many minutes).

Indeed commercial airline and military GPS' output range-rate data (relative to each satellite) to the INS (by whatever name) for the INS to correct its internal state along with velocity / velocity-dot data; not to mention position. (INS even take range data from DME as a position correction source (at least in the pre-GPS era) - but I digress).

*Of course, at takeoff, GPS is not at all useful for orientation, but within a few seconds of flight in any direction, use of true (or computed mag) heading referenced to the IMU body angles would be quite trivial. Indeed from there a very accurate calibration of the mag compass could follow so that it remains a viable _backup_ and not primary heading source after takeoff.

I'd venture that the compartmentalization of their dynamics software and their navigation software makes the integration I describe a larger task than simply coding the equations. They have to pass more data between modules and this means a harder effort and more te$ting. But, if they are to move the technology forward they should be doing it at this fundamental level too - magnetic heading is not a good thing, especially when used close to the ground and in an electrically noisy environment such as drone motor control with currents being switched from 0 to high amps at high frequency. (I suspect a lot of s/w level filtering for that).

Another product of this venture would be accurate wind direction information by using power to the rotors to estimate the airspeed vector. From there the wind direction and speed is trivial. They could do this now if they wanted.

**As a final (and not realistic) notion, they could eliminate the mag compass altogether. But that would require a long power on sequence before takeoff in order to measure earth's rotation locally and hence the state of the IMU and thence the body heading of the drone. Not realistic because in such a small IMU this could take 10 minutes or so to get a rough (±5° heading) which would be adequate for takeoff until the GPS/IMU solution for heading was computed. Battery drain would not be acceptable.

You say: <<The three directions you speak of, are relative to the previous GPS position>>
As a nitpick, GPS does not calculate velocity based on position change but on the Doppler shift of each satellite channel tracker. It then computes the velocities in ECEF coordinates and finally converts them to E,N,U at the local position. The direction of travel is then computed from those vectors.

*** DJI are very good drone designers. But they don't do "aviation" very well. They use the wrong terms for a long list of things; the red/green lights on the craft are plain wrong; They need, IMO, to pivot to correct aviation terms for everything. Indeed while I'm a very metric guy, I believe they should face the user solely in feet, knots, nautical miles and correct terms for "altitude" (ASL) and "absolute altitude" (AGL) rather than their hodgepodge of not well cooked definitions. Their documentation ("manual" - cough) is very crappy and imprecise and ambiguous in many areas.

 

[AlanTheBeast]

Moreover, there are TWO compasses

A man with two watches does not know the correct time. - Chinese proverb.

 

[AlanTheBeast]

There's only one way to prevent a fly away, well other then not fly it. Just do what I do and tie a 4 mile long string to your MP.
Problem solved.

Works for me, but my string is only 2 km long.

 

[AlanTheBeast]

Actually the Canadian dictatorship just limited my string to 508 metres assuming I can still see the Mavic Pro. (I can't see the MP beyond 300 metres or so and only if I tracked it visually going outbound against a dull cloudy sky).