Make use of yaw/heading/bearing and no-sideslip nonholonomic constraint to calculate speed
Mont and I have talked about this based on Sinan's work. We could likely improve our calculatations of both distance and speed if we could make use of the fact that the boat has to have a velocity vector aligned with the long axis of the boat (assumes no slide slip, a nonholonomic constraint). This could effectively let us remove the effects of a large amount of the GPS position error in the lateral boat direction. If we can do a simple adjustment to the current algorithm we can potentially include this in the paper, but if we need elaborate changes we may hold off.
Here is a gallery of the available smartphone produce data related to yaw/heading/bearing for each trial to start the discussion:
https://photos.app.goo.gl/CveqDyjEAndp9t179
And an example:
FYI @monthubbard