Hi Mark,
thank you for your answer. With the caveat of my line of thinking being entirely wrong somewhere, I see the following issues with remapping: First, to correctly remap linear accelerations, one would probably need angular accelerations, which are not measured directly. So the remapping is either approximated, or the angular accelerations are derived from angular velocities, which might introduce an additional delay. Second, the remapping might introduce correlations of the noise on the measured accelerations to the noise of the gyroscope measurements, which violates an assumption of our estimator.
To us, it seems that the cleanest option would be to make use of the unmapped measurements.
Would that justify asking one of the designers whether she was so kind to determine the location of the accelerometers/gyroscopes with respect to the point of percussion provided in the datasheet (from CAD files, from your calibration procedure, or any other method)?
Thanks, Joern