EHEA
lvl.4
Flight distance : 363960 ft
Germany
Offline
|
JussiSL Posted at 8-4 04:12
Thank you for the very nice analysis. I can't see it from the picture but both IMU and GPS were at the same position at the start? Then somehow the IMU position jumped about 30 meters to NW. If the flight controller program was more clever it would understand that it's not possible to move so fast. A full manual control could be a nice feature. I have another drone with Naze32 controller. It has no GPS and it doesn't calculate the speed. It just levels the drone. It doesn't drift very much in calm weather.
Yes at the start, the positions match, but that is by definition. The IMU can only sense acceleration and rotation (and velocity from doing math on the former values), not location, so the initial starting point in the plots is taken from GPS.
The divergence between IMU and GPS does not happen in one big jump, but rather gradually while the drone is flying away, that is shown in the graph I posted earlier about the differences of the gpsVelocity and the IMUvelocity. This coincides with the period when the drone got only 8 or 9, sometimes 10 sats, and vanishes when the drone receives > 11 sats, so I rather suspect that the somewhat inaccurate GPS values were bouncing around quite a bit during this incident, which led the drone to "think" it was moving at crazy speeds for short momenets (up to 43 mph!) while actually it wasn't.
From an engineering point of view, this can be a rather challenging situation with a lot of coupling and bad feedback in the control loop.
Think about it: the GPS positions bounce around so the drone thinks there is a high velocity it needs to counter-act. The result is that the drone is tilting a lot to correct for this. The tilt will change the orientation of the GPS antenna ... oh .oh... this can lead to a different set of satellites being used when there are already few available ==> another jump in the (inaccurate locations) ==> the drone again thinks it needs to correct for sone spurious velocity change ==> changes the tilt, changes the GPS antenna orientation .... if this is not dampened correctly, it's easy to believe that this can lead to "surprising" drone movements.
|
|