NOTES
So how do you keep the accelerometers in the desired orientation? It's a multiple step process, with each step correcting for the limitations of the previous one.
Step 1: Isolate the accelerometers from ship motion - using gimbals
Step 2: Use gyroscopes to measure or correct leftover undesired changes in orientation - in the ESGN, these changes are measured with two 2-degree of freedom gyroscopes, and the changes are not corrected - as long as the system knows the physical orientation of the accelerometers relative to inertial space, navigation calculations will be done correctly.
Since the system is now dependent on gyro accuracy, step 3 is to account for gyro errors. This is done by modeling the gyro errors with as many parameters as necessary, calibrating the stable parameters, updating the less stable parameters as often as necessary, and developing a method to measure errors occasionally so that the updating can occur - this is lots of work.