Brian Gerkey

And then you can run all that in simulation because that's also an important part of being able to test these systems.

Testing on hardware is time consuming and expensive.

So we want to make it super easy to be able to test in simulation.

So you can check out, make sure everything works before you deploy it.

But then you can hit deploy and we take your application.

It's all containerized and ready.

We package it up and it gets deployed onto your industrial PC or IPC.

This is a computer that will sit next to your robot and it'll eventually go into the factory for your customer.

And then instead of talking to the simulator components, it talks to the physical components.

And at that point, you might have to make some tweaks.

There might be some changes you need to make now that you've gone from simulation to hardware because the two are not exactly the same.

But if we've done our job right, the simulation got you a lot of the way there and then you can deploy it onto the robot.

And along the way, you mentioned the vision model.

This is one of those algorithms that I mentioned earlier.

Perception is an area that has just been absolutely revolutionized by modern AI.

I'd say almost any perception system.

And by perception, I mean, you know, you're going to take some usually camera data in and then try to answer some question about that camera data, like what's going on or.

Where is the object that I'm supposed to pick up?

Or where should I pick up the object if I've got two fingers?

Where should I put them in the world in order to pick that object up?