They are often called "quantum compasses" although they are complete inertial measurement units which have an accelerometer, magnetometer (a compass is a magnetometer), and a gyroscope.

The subtle parts of the magnetic field of the Earth shift a lot related to the movement of metal through the mantle. NOAA and the DoD periodically create new maps of the magnetic field, but they have a lot of error and noise.

Sure, I’m not saying it is a GPS replacement or anything.

But, a measurement of absolute position is a fundamentally different thing from trying to figure out a position by integrating from an IMU, right? The latter inevitably accumulates errors. The former should not accumulate error. The measurements might be extremely noisy, but noisy and error accumulating are different from a signal processing point of view.

The magnetometer's reference data is so inaccurate for positioning and shifts so much that error will still accumulate.

Magnetometers have been used in IMUs for over 50 years. Their shortcomings have little to do with the precision of the instrument.

(I think you meant to say "The former should not accumulate error" instead of the latter.)

Good catch, thanks

NOAA did the mapping already, making this like some kind of geoguessr for robots.

I wonder how much those anomalies change over time and how updates would get rolled out without breaking that "unhackable" aspect of it

There's a slew of detail missing.

The "fine detail" magnetic field changes throughout the day (like the tide) and the underlying pattern itself shifts such that a new "map" is issued every five years which is actually a spherical harmonic equation with a depth of 24 (IIRC) terms at least.

(Addendum: degree 12, my bad .. (or degree 790 if you're aiming a missile)

https://en.wikipedia.org/wiki/World_Magnetic_Model or

https://en.wikipedia.org/wiki/International_Geomagnetic_Refe...

choose your model adventure, these are 'coarse' global models with five year epochs, see also:

    The Enhanced Magnetic Model (EMM) is a sister product of the NGDC featuring a much higher amount of data to degree and order 790, giving a wavelength of 51 km as opposed to the 3000 km of WMM.

)

Ideally you also want nine sensors, one for each XYZ axis at three spread out vehicle locations: typically tail + each wing tip on an aircraft to get that mag flux differential across the craft.

It also helps to have ground base station that can record the diurnal flux (daily changes as the earth turns) from a fixed position to subtract from the nine readings on a moving vehicle.

Oh, yeah, heading also plays a role, so you need a Kalman filter formed from a butterfly wing motion pattern in order to subtract the field induced on a the craft by motion that varies with heading.

It's fun stuff.

From what I have read the magnetic mapping it not enough for a position. It needs additional information to limit the possible positions. So the technology could be fairly accurate on a train or car. On a ship or airplane with an imperfect current position and course assumption it may not be very accurate and even drift over time.

All magnetic compasses are magnetometers, but not all magnetometers are compasses. Some magnetometers output absolute magnetic field strength (scalar) while vector magnetometers output heading and strength of a field.