Ok, i started doing some tests with my IMU module. It turns out that, while it is has very low noise while standing still, is suffer greatly from the robot vibration and movements.. Some research was in the need :)
I ended up to a couple of sites, one is this, a blog post of a guy who managed to have a near perfect direction sensing. He's using an Arduino and also released the source code! Yayy :P Reading around his site, i ended up to another blog which inspired him, that is Tom Pycke's. It has such goodies as a practical tutorial to kalman filters, explanation on how to use gyroscope and accelerometer and much more, including a C implementation of the program.
Both sites are for robotic air vehicle, but the theory works for balancing robots too :)
Anyway my understanding of the kalman filtering just jumped from 2% to 25%. This weekend i hope to make some real test. I think i'll write a simple application to visually display data (just like in the video).
Well that's enought for now :P
NOTE: the guy in the video is NOT ME! just embedding one of the site video to show the wonderful results
https://www.youtube.com/watch?v=1DDEH2n2_hY