So basically we study cat robots. That's the job. We put them in environments, we watch what they do, we write it down. Sometimes it's interesting, sometimes it's really interesting, occasionally something happens that makes everyone stop and stare for a bit. We're looking into that last one.
Whatever we figure out gets fed back into how the units are built and how they behave. It's a whole cycle. It works pretty well.
We run the units through scenarios, record everything, and compare it to what we expected. Usually it lines up. Sometimes it doesn't, and that's actually the more useful outcome honestly.
If a unit starts doing something it shouldn't, we pull it aside and figure out what happened. No big deal, just part of the process.
The units are cat-shaped robots. That's intentional. The feline form factor turns out to be really good for a lot of what we're testing — movement, balance, sensory response, that kind of thing. It also makes them easier to work with, which is a bonus.
We try to keep the movement feeling natural. It matters more than you'd think.
We build the environments modularly so we can swap things around between runs. Keeps the results honest — if the unit's seen the exact same room before, that affects things. A little variability goes a long way.
We also monitor the environments themselves to make sure they're not drifting. That's happened before. It wasn't great.
We're looking at a few things:
If something goes off the rails we isolate it, figure out the cause, and reset if needed. We've got protocols for it. They're not complicated, they just require people to actually follow them, which — mostly happens.
We do think about this stuff, for the record. The units are evaluated safely, nothing gets pushed past reasonable limits without oversight, and we review everything regularly. It's not just a checkbox. We actually care how this goes.
Anyway. That's research. Hope you found it interesting.