[Paparazzi-devel] Attitude Sensing: RE: Paparazzi-devel Digest, Vol 43, Issue 11

[David]

Maybe the Tiny Arm7 at 60ghz is not an ideal platform for filters.
In my mind it's 2007 (almost 2008)...so dream it and create it.
We have the resources as far as technology all around us.
Surely if it's filters then code up the filter now then see if it's  
too slow.
Others would know better than I but it seems a Classix attached to a  
Gumstix should be able to handle the filter processing?
I haven't a classix or gumstix so I can't comment on how fast that is.

Shielding, Faraday cages, out of the question or no? Seems simple way  
to protect against EMP and "noise".

-David

Christophe said: "
Hi David,

in order to fly in clouds (sometimes people want to do for instance  
atmospheric measurements in the boundary layer) we are trying to by- 
pass the state-estimation problem. Since the small sensors give just  
a little to much noise, and processing is just a little bit too slow  
we are trying to find controllers that use direct sensor-actuator  
feedback only.

The state estimation becomes exponentially more difficult with  
smaller platforms:

1) low weight means less weight for shielding, damped mounting, ->  
more noise
2) small size means everything is packed closer together -> more  
interference problems
3) small vehicles respons quicker -> need faster update rates (>100Hz  
to integrate gyros)
4) small vehicles have bigger turn rates (>500deg/sec) -> need for  
extra measurement resolution (-> 16bits ! / 24 bits ?)
5) low power/size imposes high processing contraints: fixed-point  
processing... low mem: with 3 Watts of power and 120 grams: the  
filtering is no problem at all... but at 5gram 100 miliWatt... its a  
little more difficult :-)

My experience is that filters that work perfectly on 5kg rotorcraft  
become almost useless in MAV (unless you only fly straight forward at  
constant speed)

So: With a little help from the aerodynamics (relatively stable a/c),  
this direct gyroscope to actuator feedback seems to be very usefull:  
The (big) EasyStar for instance flies very well on 2 gyroscopes and 1  
differential pressure sensor. So only 3 sensors (just a little more  
expensive than the IR sensors). With this configuration you can fly  
through any kind of smoke, cloud, or in between buildings (which I do  
not recommend). The pitch loop is a little less accurate than with  
the IR, so I do not recommend flying extremely low (a few meters  
above ground), but the processing even fits in a 8-bit  
microcontroller. (Unlike any 17 state extended kalman filter)

Hope this helps,

Christophe"