Problems Reading Accelerometer Device

Werner Almesberger werner at
Wed Nov 26 15:40:54 CET 2008

Michael Tansella wrote:
> I'm a student and in my research project I have to do inertial navigtion and 
> therefor I need this.

Let's hope this navigation is only in a very limited area.

Accelerometers in general aren't good at tracking position, and
I'd be surprised if ours were an exception.

> What would you recommend me to do,
> is this a problem of the current kernel and will perhaps be fixed in the next 
> one, so should I only wait?

I don't know if anyone is currently working on the acceleration
sensors. There is generally interest in getting them to behave,
both inside Openmoko and in the community, so I think you won't
be alone in your quest.

I also don't know how bad the situation really is. There has been
some work on gesture recognition, so perhaps someone who has
gathered some experience there could comment.

If you wanted to make sure that you understand exactly what kind
of distortions you get (e.g., whether the samples are unevenly
spaced, if samples are missing, if just the timestamps are wrong,
etc.), a good test may be to mount the GTA02 on a device that
rotates it vertically at a constant velocity (perhaps some old
turntable would do), and see how far the data you get deviates
from a sine wave. You could still communicate with the GTA02
through WLAN.

> Do you think that I can help in any way to solve this problem without deeper 
> kernel and arm development skills?

You may consider this as an opportunity to acquire such skills :-)

It's hard to tell how much it takes to make this work for your
application. Maybe you can just filter/reconstruct the data
yourself. Maybe you have to fix things in the kernel. Maybe
some of the logic needs to be redesigned.

- Werner

More information about the devel mailing list