Experience with dead-reckoning navigation?

Hello, Has anybody got any experience with gryo and accelerometer sensors? I'm wondering how fast the errors accumulate when calculating relative-position by dead-reckoning. For example, if a rover equipped with one of these sensors sets off to do a lap of a room, how near could it get to its starting point after (say) a minute?

I don't have any experience with your parts, but I was recently looking into the issue - which isn't technically "dead reckoning", but more "inertial navigation":


Note the graph at the top - note how poor the accuracy is compared to say, D/GPS...

With that said, I am sure those numbers have to do with actual vehicles (airplanes and/or missles/rockets - most likely) in real-world conditions over large distances. A robot in a room with a perfectly flat floor (and little wheel slipage) would likely fair better. But as soon as you start adding other needs (carpeted floor, irregular surfaces, etc) - navigational accuracy likely goes down quickly...

Yes, sorry, I used completely the wrong term. Inertial positioning is what I should have said.

What I have in mind is double-integrating the 3D acceleration to get position, assuming the sensor does not rotate. Then I'd like to remove the rotation constraint by incorporating readings from a gyro.

The specs for reasonably-priced sensors seem amazingly good, but of course the errors will grow enormously once integrated over time. The 650m/hr figure on Wikipedia is discouraging.