Do I understand it right, that the X-series do not contain magnetometer (so only 6dof)? I need that to calculate odometry from IMU, and currently the yaw has a horrible drift.
The actuators do not contain magnetometers. Even if they did they would be mostly measuring magnetic effects from the motor and gears. Overall, forward-integrating IMU data to get pose estimates doesn't work well without other sources to fuse in, like wheel odometry, vision, lidar, etc.
Out of curiosity are you trying to integrate based on the quaternion orientation estimates from the actuators, or are you working with the raw gyro readings? If you can use the quaternion orientation from the actuators that should work better.
Lastly, we have found benefits from averaging the estimates from multiple actuators, to reduce drift and noise. So for example if you have 4 actuators on the frame of a mobile base, you can average the estimates from all 4 actuators, which should cut noise and drift down by roughly 1/2.
Hope that helps,
And thanks for the averaging tip, my thesis is based on this idea