[Paparazzi-devel] Inertial navigation theory

[Filipe Varela]

Hi all

I'm new to this list to please bear with me if im saying something stupid 
but: is kalman filtering really necessary when using inertial sensors?
An accelerometer's y (pitch) output is pretty accurate when the platform 
is idle (ie not moving) right? But this signal is subject to interference 
when the platform moves in a direction that is perpendicular to that of 
the measurement axis. Picture a little air bubble inside a mini water 
container. If you push it straight ahead, the bubble moves back like it 
would move front if you applied it a positive (front up) pitch. Inertial 
reaction. So, you could determine the ammount of interference caused by 
the linear movement and you could even determine it's instant acceleration 
and add (positive pitch) or subtract(negative pitch) that to the sensor's 
output value and then use simple trig to determine an angle...

Is this nonsense? Is it that difficult to determine the inertial force 
(then displacement, then speed, then acceleration) that the linear 
movement causes on the accelerometer? Unless this is what kalman filtering 
is all about which im not sure it is. (im a newbie to this)

If im right, you could do the same process on the x axis (roll). Then with 
a little effort and a single z axis gyro you could apply a precise 
correction of both y and x outputs. This would require a 30 euro dual axis 
mems accelerometer and a 45 euro z axis mems gyro.

I'm looking forward to getting some feedback from you guys...
Thanks for reading,
Filipe