As mentioned here [1] they use atom interferometry to build extremely precise sensors of acceleration and rotation. This is everything needed to build an inertial navigation system (INS). This is completely independent of any external system. My guess is that this is what makes is so interesting for subs.
Probably the biggest issue with INSs is accuracy of the raw IMU measurements. The acceleration must be integrated twice. That means any error in the measurement gets exponentially larger with time. Rotation velocity measurements are integrated only once (some IMUs measure absolute orientation but it is not the case here) so the error does not explode as fast. Is is the combination with translation that is problematic. If you rotate around vertical axis by 90deg +/- 1% and then travel for 1000km +/- 1% the rotation accuracy will cause bigger error in final location.
To have an INS capable of long time operation without any other means of drift correction one has to nail many other things.
- Gravitational constant varies and you have to have an accurate map of it. We might not have a good enough one. I don't know. The subs could collect their measurements and perform a large scale optimization to increase the accuracy of this map.
- The rotation of earth gets added to the rotation speed measurements. The rotation is not constant (hence the leap seconds from now and then). We have to be able to extrapolate it well enough.
- Sampling rate of the quantum IMU has to be high enough. How usefull is the most precise measurement if you can measure only once in 10 minutes? Traditional IMUs sample at 1 kHz and more.
- Does the new quantum gyroscope add enough value to the accuracy of the final IMU compared to Fibre-optic gyroscopes? They are incredibly accurate as they are based on laser interferometry and then can sample at high frequency. Plus they are an established technology which partially benefits from economy of scale [5] because they are used in airplanes [6].
- The measurements of both rotation velocity and acceleration have to be measured in 3 axes. Are the relative orientations and translations of those axes stable enough as the device moves?
- Is the noise present in the measurements corelated to something or is it truly random? The corelation could mean an systematic error to the state estimate.
- You will likely also have to have an atomic clock on board labeling each IMU measurement as soon as it is made. But this is not a problem nowadays as one can have a small Rubidium-based atomic clock in a package starting to resemble a chip [2] [3] [4].
- And of course you will need enough computing power to calculate everything. I'm not sure how modern hardware they are allowed on board of nuclear-carrying subs. This is slight problem for things like Mars rover as due to the testing requirements the main chip can be easily 30 years old when the rocket is launched [7].
Also, the situation is not as bad as we can utilize the sensor fusion and correct lots of errors caused by IMU measurements. From the top of my head one can obviously measure water pressure which estimates depth (with some error). Sea bed topology could be other type of measurement but my guess is that this measurement is too crude compared to the accuracy of the quantum-based INS.
Probably the biggest issue with INSs is accuracy of the raw IMU measurements. The acceleration must be integrated twice. That means any error in the measurement gets exponentially larger with time. Rotation velocity measurements are integrated only once (some IMUs measure absolute orientation but it is not the case here) so the error does not explode as fast. Is is the combination with translation that is problematic. If you rotate around vertical axis by 90deg +/- 1% and then travel for 1000km +/- 1% the rotation accuracy will cause bigger error in final location.
To have an INS capable of long time operation without any other means of drift correction one has to nail many other things.
- Gravitational constant varies and you have to have an accurate map of it. We might not have a good enough one. I don't know. The subs could collect their measurements and perform a large scale optimization to increase the accuracy of this map.
- The rotation of earth gets added to the rotation speed measurements. The rotation is not constant (hence the leap seconds from now and then). We have to be able to extrapolate it well enough.
- Sampling rate of the quantum IMU has to be high enough. How usefull is the most precise measurement if you can measure only once in 10 minutes? Traditional IMUs sample at 1 kHz and more.
- Does the new quantum gyroscope add enough value to the accuracy of the final IMU compared to Fibre-optic gyroscopes? They are incredibly accurate as they are based on laser interferometry and then can sample at high frequency. Plus they are an established technology which partially benefits from economy of scale [5] because they are used in airplanes [6].
- The measurements of both rotation velocity and acceleration have to be measured in 3 axes. Are the relative orientations and translations of those axes stable enough as the device moves?
- Is the noise present in the measurements corelated to something or is it truly random? The corelation could mean an systematic error to the state estimate.
- You will likely also have to have an atomic clock on board labeling each IMU measurement as soon as it is made. But this is not a problem nowadays as one can have a small Rubidium-based atomic clock in a package starting to resemble a chip [2] [3] [4].
- And of course you will need enough computing power to calculate everything. I'm not sure how modern hardware they are allowed on board of nuclear-carrying subs. This is slight problem for things like Mars rover as due to the testing requirements the main chip can be easily 30 years old when the rocket is launched [7].
Also, the situation is not as bad as we can utilize the sensor fusion and correct lots of errors caused by IMU measurements. From the top of my head one can obviously measure water pressure which estimates depth (with some error). Sea bed topology could be other type of measurement but my guess is that this measurement is too crude compared to the accuracy of the quantum-based INS.
[1] https://www.imperial.ac.uk/centre-for-cold-matter/research/q...
[2] https://eu.mouser.com/new/microchip/microchip-macsa5x-atomic...
[3] https://www.sparkfun.com/products/14830
[4] https://www.accubeat.com/nano-atomic-clock
[5] https://emcore.com/wp-content/uploads/2016/12/EMCORE-Commerc...
[6] https://www.airbus.com/en/products-services/space/equipment/...
[7] https://bigthink.com/hard-science/perseverance-rover-brain/