They don’t use anything for their latency. I’ve looked into the space network they have, what protocols they use and what they do about the massive delays… Just a little bit, I don’t know it super well or anything, but from my understanding, the “network” they use is more about assurance than delivery speed.
The publicly available version of what they use is called delay tolerant networking, which essentially uses a mesh of nodes that may, or may not be able to communicate to other nodes at any given time. As messages are sent, they are relayed from node to node as connectivity allows until they get to a base station for final delivery. It’s a bit like the mail system, but instead of large centralized sorting facilities, you only have local post offices. The message is sent from one office to another until it leapfrogs it’s way to the destination. It can wait at one post office indefinitely until a path opens up to the next one.
In the case of delay tolerant networking, it basically sends it along to the next station in the mesh, and that station will confirm the delivery of the information, which is when the sender can remove the message from its buffer.
Ideally, the nodes should have some type of non-volatile memory (like nvram) to store pending deliveries, so nodes don’t waste power trying to keep the information in their volatile memory (RAM).
Terrestrialy, we use DTN for tracking stuff like the movement of animals in large and unserviceable areas (where mobile networks like LTE, don’t exist), such as deserts and undeveloped forests. As the trackers on the animals come within range of another tracked animal, updates occur, and when either gets near enough to a base station to upload the information, then the updates are sent out to the records systems.
Don’t ask me how the logic works to figure out when to push data one way or another. I haven’t gotten that deep into the protocol yet.
Anyways, for NASA, the information is sent to satellites, which relays to the rover eventually. In NASA’s case, they can directly transmit, from Earth, using microwave arrays, to the satellites in orbit around Mars if we want.
I’m not sure on the specifics of how they have their version of DTN setup, so I’m only speculating at best.
They don’t mitigate latency, they simply account for it, and work with that as part of the problem.
I think when the hype dies down in a few years, we’ll settle into a couple of useful applications for ML/AI, and a lot will be just thrown out.
I have no idea what will be kept and what will be tossed but I’m betting there will be more tossed than kept.