Hello,
The vertical filter used for rotorcraft is based on a kinematic
model not a dynamic model. It means that the input of the filter's
propagate function is actually the accelerometer measurement,
instead of the acceleration estimated by the a dynamic model. The
reason is that the dynamic model is usually unknown, and it makes
the filter independent of it. If you want to tune the noise
associated with the accelerometer, you have to look at the system
noise.
Gautier
Le 08/12/2012 01:21, Michal Podhradsky
a écrit :
Hello guys,
I am currently adjusting the VFF filter (z_dd, z_d, z estimation)
to our rotorcraft airframe (i.e. changing the process and
measurement noise covariance). As the filter is currently
implemented, in the propagate phase (vff_propagate) the measured
acceleration is propagated to all three states with zero noise
(line https://github.com/paparazzi/paparazzi/blob/v4.0/sw/airborne/subsystems/ins/vf_float.c#L86).
Is there an assumption that the IMU accelerometer is always
correct?
If I understand it correctly, the barometer (z measurement) and
accelerometer(z_dd measurement), and optionally other sensors
(sonar etc.) should be updated at different frequencies. In the
current master branch ( https://github.com/paparazzi/paparazzi/blob/master/sw/airborne/subsystems/ins/vf_extended_float.c
) are already separate updates for baro (offset) and gps/sonar/etc
(no offset) sensors, but the acceleration is still propagated
directly.
Do you know why the filter was implemented in such a way?
Thanks a lot!
Michal
_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel
|