If you were going to make a 100% robotic factory/refinery/etc. you would design it for robots. The problem with the Roomba is that houses are designed for people, not robots. A robotic factory could avoid a lot of problems by not needing to design for people.

While I disagree with your benchmark, I don't think we're going to be putting robotic factories in space in 3-4 years. Have we even built a 100% robotic factory on earth yet? Plus, I don't think that anyone has built an autonomous robot for zero-g yet (I'm not counting satellites and Canada-arm).

> Plus, I don't think that anyone has built an autonomous robot for zero-g yet (I'm not counting satellites and Canada-arm).

https://www.theverge.com/2017/7/17/15981250/japan-space-came...

"Japan’s space agency has for the first time released photos and videos taken on the International Space Station by its resident robot drone..."

But that factory won't do any good unless it gets input, and can put its output somewhere. The input would supposedly be from the local regolith which has a lot of the same problems that the example roomba would face.

Now a factory on Mars that takes in sunlight and atmosphere, and fills a tank with rocket fuel, could be possible. But in that case I would call it a specialized machine, not a factory or a robot.

I don't think we have to go that far, though.

Humans operating a remote robot on the Moon is a feasible concept, but there's a lot of remaining work on telepresence to be done. Haptic feedback, VR, robotic finger dexterity.

Full autonomy isn't necessary with the Moon's time lag. It'll be required if you want to do the robotic factory approach on Mars or the asteroid belt or something.

A Roomba is consumer-grade technology, though. There's a whole different paradigm of manufacturing and production used for i.e. defence, space/aerospace, high-level particle physics at LLNL and so forth. Not really a good metric. Ever read about the National Reconnaissance Office?

I was using "Roomba" as an illustration device. I necessarily wouldn't expect that the technology I described would be available at a consumer price point, but many of those items on my list should be doable with the same technological (AI) advances needed to make off-planet robotics work well (especially ones that are more distance such as on Mars, which need to work more autonomously).

For example, the rovers on Mars, as much as they have accomplished, really haven't done much more in all the time they were there than humans did physically on the Moon 50 years ago.