"Should we start a new thread, "shrinking a concave polygon"? - NA
.... Yes! That would be great! .... but it must cover both a mix of convex and concave.
--
"You don't need to profile the IMU drift, just estimate the combination of maximum IMU drift and GPS measurement noise" - NA
Here is were I keep losing you! Why do we need to account for IMU drift at all? You give me the impression that you will be compensating for drift over a long time (>2-3 seconds). The maximum IMU drift would only be for one second max! ..... negligible.
Also, you give me the impression that the GSP is the more accurate position authority. The GPS could be off 3-4 feet at any given time and for a long time (several seconds). Where as the IMU , given a fix to start, is much more accurate in the under one second output frame.
IE: IMU calculates the next position very accurately for the next second after fix, GPS outputs for this reading as well but has errored by 3 feet!
This is why I wanted to use the IMU out to simply "bound the future GPS error swing" .... ... by simply taking the distance (halve way) between the two (IMU and GPS) best guesses..... halving any error .... and use this for the working position.