Since you presumably know the robot effector position from it's
feedback system, why not just use that information directly?
If you are mounted on some part of the robot, you just need to
apply the appropriate scaling to the kinematic equations giving you
position.

I'm not clear one whether the robot is moving the platform on purpose
or because of a poor design choice. e.g. it may be possible to
stabilize your platform using parallelogram linkages to the robot frame.
R

"Robert B." wrote:
>
> I've been working on a microchip-based system to stabilize a sensor platform
> using an ADXL202 and an angular rate sensor.  The platform is moving around
> in all directions (it's mounted on a robot) so the linear ADXL can't really
> be trusted to give a consistent or even accurate reading.  It has been off
> by as much as 80 degrees in the limited testing I've done.  Most of the
> correction is currently coming from the integrating the rate sensor, but
> it's subject to too much drift and really needs the corrections from a
> linear accelerometer.  Does anybody know a work-around to this problem?
>
> Right now it's integrating at about 60Hz, and checking tilt on the ADXL
> every 100ms, then correcting 1/10th the offset indicated by the ADXL to
> create a crude time delay filter.  This combination (so far) gives the best
> stability at low levels of tilt, but at high rates of tilting (<150
> deg/second) the position vector takes too long to catch up to where it
> should be.
>
> If you need or want more details let me know.

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics