Re: [Paparazzi-devel] IMU's

[Christophe De Wagter]

About the discussions of AHRS filters: 

DCM or quaternions or euler angles are all fully equivalent notations for attitude and can be converted from any notation to any other. The advantage of euler is its intuitive (for small angles) interpretation, its disadvantage is the non-linearity at 90 deg pitch, the advantage of quaternions is the reduced memory size (e.g. when used in a full kalman filter it is better to add 4 states than 6/9 states). The DCM attitude notation is needed in all attitude cases at least as an intermediate step to convert vectors and is used in all cases (euler->DCM = [cos.cos sin.cos ....] and quaternion->DCM= [q0*q1/2, q1*q3/2, ..].

Now indeed without ground-speed information extremely large errors will occur. This can be provided with (slow and delayed) GPS or approximated with an airspeed sensor. This secretly assumes that the plane is pointing in the direction of motion and therefore is only valid for fixedwing. All fixedwing attitude algorithms use this: that is DIY'DCM based <subsystem name="ahrs" type="dcm_float"/> as well as <subsystem name="ahrs" type="ic"/> as well as textbook extended kalman filters. The difference only lies in filter gains and computational power: e.g. the IC (integer complementary) filter does the same as DCM but with 30 times less computational effort for the autopilot (but still needs some fixedwing integration update).

However: all the above mentioned filters do not have a 3D velocity and change in velocity feedback making them not useful for helicopters and give trouble when applying very long longitudinal accelerations (large airspeeds). The 3D speed feedback is possible and quite easy to implement but makes the filter much less stable, which is why this is not applied as a default. 

So Hector Garcia certainly has a point about the performance but as dr Warmers says in practice it will work nicely on slow fixedwing aircraft with not too much wind (much defined as in: constantly clearly blown away from its normal trajectory with visible yaw and pitch oscillations).

Sincerely

-Christophe 

On Wed, Aug 10, 2011 at 8:29 AM, Christophe De Wagter <address@hidden> wrote:

Work is on its way to include Magnetometer in order to get the auto-takeoff right (as it does not work so well yet since the filter needs a bit of time to get the heading right).

Please note that http://www.sparkfun.com/products/10252 is not compatible at this moment as the driver needs to be changed a bit for the talk-through-i2c. It was already made and will hopefully be added to the master soon.

-Christophe 

On Wed, Aug 10, 2011 at 5:34 AM, David Conger <address@hidden> wrote:

FWIW: This one: http://www.sparkfun.com/products/10121 also works since fixed-wing doesn't use MAG.
-David

On Aug 9, 2011, at 3:00 AM, address@hidden wrote:

> Hi Guys,
>
> i m also flying a tiny with IMU (TWOG+Aspirin)  didn t had any problems with the DCM code so far
> it really performs quite good, at least in slow planes like the twinstar.
> if someone want s to use the aspirin wit a tiny/twog there have to be soldered three little wires on the back of the standart carrier board to change it to i2c - not a problem
> i have also seen the sensor stick from sparcfun http://www.sparkfun.com/products/10724
> with the dcm code and it worked right out of the box, just check that the sensor orientation is configured correctly
> have a nice week everyone
>
> Tobi
> ___________________________________________________________
> Schon gehört? WEB.DE hat einen genialen Phishing-Filter in die
> Toolbar eingebaut! http://produkte.web.de/go/toolbar
>
> _______________________________________________
> Paparazzi-devel mailing list
> address@hidden
> https://lists.nongnu.org/mailman/listinfo/paparazzi-devel


_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel