This is a new thread for discussion about what we can do to get a very precise Loiter.
The initial thread is here :
My proposition was to use Kalman filtering to merge GPS and acceleration data, but according to Roberto, it is not possible to do that with the actual hardware.
Anyway first we need to test if a dual integration, to get position from accelerometer data, is possible without too much drift.
If we can get something like less than 1 meter drift during 10 secondes from the dual accelerometer integration, it should be possible to mix this solution with the GPS position, to get a nice final result.
Second, Roberto, are you sure that the 2560 chip is not powerfull enough to run a Kalman filter ? After all 10 Hz refresh rate should be enough on the kalman filter side, because GPS data is 10 Hz.
Next, we could use a simplified filtering, and APM 2.0 should allow to get 30 - 40 % more power thanks to motion fusion code that should be able to move in the MPU6000 DSP processor.
Replies
One key thing if we are to go down this road:
I think it will become necessary to consider the location of the APM board relative to the CG of the aircraft. If the board is not on the CG, then any rotational accelerations will also appear as linear accelerations. The acceleration due to rotational acceleration will need to be subtracted from the linear acceleration in order to achieve an accurate INS.
You can see other systems account for this by asking the operator to input distances, and sometimes even a calibration routine?
I always wondered why the APM did not require the relative location to be input. Now I know why. We only ever looked at angular velocity to perform stabilization. But have not yet used linear accelerations to achieve INS dead-reckoning. (is that correct?)
Hi Oliver,
so i discuss today with Randy about the possibility to use as starting point his work on optical flow control.
So Randy confirm that the output is only x and y vector of speed and not other output from the sensor and the result of his work is good but have the limit of altitude we can try to develop a INS as in the video of Marg demostration using our prefer platform. I can use my MP32F4 + MPU6000 so i think that i have all the power need for develop a good INS.
Before to integrate it we can doing some test and demostration of functionality then we can add it to APM and doing a test as it is a optical flow control.
So other work that i'm doing is try to put on shared document a visual schema of ACopter so will be more simple for us how and where we can upgrade the code and algorithm .
@Jason @Chris what do you think about this kind of approach ? @Jason in your development are you using some kind of tools for obtain automatically doc as the document that i produce manually ? Could be very usefull for DEV Team and other people
Best
Roberto