I am using 3-space sensors (usually LX Embedded) for a robot. We need a good sense for where "down" is. We've had some issues that seem exacerbated by sensor positioning and I was wondering if someone here might be able to point us in the direct direction (pun intended).
Specifically, since the sensor can never be perfectly aligned with the center of mass of the solid body, any rotations (about the center of mass) induce accelerations at and measured by the accelerometers. Theoretically, this should not be a problem.
In practice however, we've observed that simple linear moves result in a changing orientation or "down" estimate. We've also observed that for the same rotation about the center of mass, the sensors further away from the center of rotation report larger rotations than the sensors near the center of rotation.
Our current guess is that we need some tuning on the accelerometer confidence matrices but, again, that is just a guess. Any pointers on this would be greatly appreciated.