Apparently the US Navy has this kind of system for positioning of submarines to correct inertial navigation system errors. Allegedly they also have very good maps of large swaths of the ocean floor to make it useful.

Very accurately measuring gravitational anomalies is important for this too.

Plate tectonics would've been accepted much earlier if the seafloor magnetism maps hadn't been classified.

I imagine the earth isn't particularly static on the mm level.

True - buildings and exposed rock probably stay still, leaf litter and trees definitely don’t. You’d need to find enough of a stable reference to run your correlation - probably more useful in an urban area.

On a millimeter scale I expect weather/seasons will cause movement of the ground as the soil swells with water from rain, or water freezing.

Taller buildings can flex up to a few feet in each direction actually!

I believe tectonic plates generally move about 15mm a year - which is pretty fast when you think about it!

> I wonder if you could make a completely offline location service by correlating the results of a local 3d scan of the ground to the global model?

It's possible, see https://news.ycombinator.com/item?id=39507267