I think the general principle in all cases is that a gyro gives you an acceleration-related signal, which you integrate once to get velocity, and integrate again to get distance i.e. position.
The implementation is normally complex due to the need to make small corrections for all kinds of things (e.g. temperature related drift) and then you have to do some more maths (old well established stuff though) to work out where you are in terms of lat/long on the earth's surface.
The current holy grail is to make something useful with solid state gyros but they aren't good enough by orders of magnitude. You might get 10 minutes of navigation out of them
