You asked about a comment I made in the DCM algorithm, concerning the determination of foward acceleration in connection with the compensation of the accelerometers for acceleration.
Since then, we have learned a lot about this topic, I really should update the document.
We found that you can ignore the forward acceleration, you can still get pretty good performance.
The reason I was not keen on the GPS information was that there is generally a time lag between change of speed and when GPS reports it.
When I was debugging another problem, I thought that maybe the forward acceleration was the culprit, so I did put in the computation of forward acceleration as the time rate of change of GPS velocity, but it did not seem to make any difference one way or another.