How maintenance intensive are they, though? Moving parts in or near vacuum tend to be a bit iffy (e.g. lubrication), especially for multi-year (or multi-decade) robotic missions.

There are non-moving part designs that utilise fibre-optics or laser rings. I've only seen either the physical or laser ring gyros. Not sure about which ones they'd use in space though.

Grandparent is talking about gyrocompasses, not ordinary gryoscopes though. I haven't heard of a laser-ring gyrocompass. Not sure it if that is even possible.

The gyrocompass orientation should, in principle, be obtainable by properly integrating gyroscope readings (and accelerometer? not sure if required).

You still need a way of determining the initial position/pose

I don't think so. The gyrocompass relies on gyroscopic precession. The precession itself can be obtained by integration, which gives orientation. Of course, the precision of this process may well be very low or unusable.

you don't need an initial heading measurement but you still need to know where you are

It would suffice to know the radius of the planet and it's rate of rotation.

Using only a compass and the knowledge that you were somewhere on the planet earth, could you determine the shortest route from your locations to Chicago?

Ah I thought you were referring to the gyrocompass functionality. Yes, for general navigation you need an initial position to determine your position from inertial measurement alone (as I said, you can determine orientation and latitude, but not longitude).

The laser gyroscopes do exist. They use lasers, plus the Sagnac effect, to measure rotations around on axis by interferometry.

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

I would assume that once you establish direction you can just put some 3 beacons and have a local navigation system based on that.

Maybe you would know, but are MEMS gyroscopes of accurate quality to output true north data?

Specifically referring to Invensense 9150 9dof (3 accel, 3 magnetometer, 3 gyro)

Not for very long at decent accuracy. My of the hip guess is minutes. The zero rate output is +/-20 deg/s for the part you listed with RMS noise of 0.02 deg/s.

Hmm.

A group of us at our local hackerspace is building an autonomous robot. We already have that 9DoF sensor as well as a GPS and barometer.

I'm able to accurately calculate where North(magnetic) is, as well as down, and roughly what altitude I am, and synchronize them to sporadic GPS locks. Depending how useful, this could be good at providing North(true). I'll see if any sort of calibration would be useful (considering I know my lat/lon and can calculate the expected rotational vector).

I'm investigating using ORB-SLAM with ROS to also provide accurate locations of the localization.

The whole idea is that I can use this as yet another piece in a probability-location detection as well as map other things quickly and accurately.

Yeah using a compass and a magnetic field model using GPS position should let you estimate the bias from the gyro.