The Swiss Are Making Some Very Good Robots. Really

Come on, buddy. Should we go for a walk? Let's go for a walk. Come on. Just a guy and his robot having a nice stroll. Oh, we're gonna go on the crosswalk. Are you ready? Don't get run over. There you go. Oh, you want to go? All right, we can go, buddy. Got to watch out for the bikers. Don't be afraid, he's just a real dog. Come on, relax. You had a big day, didn't you?

Switzerland has its charms. There are pretty buildings and cityscapes that speak to the long-term benefits of neutrality and housing other people's money. Sometimes being boring and punctual and following rules has its benefits. Which is to say, "Funiculalala." But I am not in the great city of Zurich for melted cheese and obedience. I am here for robots. And the best robots are built at the local public university, ETH. I mean, look at this shit.

Am I allowed to be over here? This is like a nightmare, a dog with a robot arm. That's like a spider.

That's a spider. Unlike many European universities that lack capitalist bloodlust, ETH does actually try to spin out technology from its labs and form startups. And I am here to learn about one such company called Gravis Robotics.

This is you guys, right?

And that's us.

It was started in 2022 by 9— yes, 9 people, including this man, Dominik Judd.

So we at Gravis Robotics, we automate heavy construction machines. So we turn an off-the-shelf excavator into an autonomous robot.

And why would people want to do that?

Is it mostly just for safety, just for reducing costs.

And there's a lot of challenges in the industry today. So one of them is labor shortage. The companies don't find the people anymore to drive these machines. It's also that if you look at productivity in the construction industry, that stagnates quite a bit, whereas in other industries like manufacturing, we've seen crazy growth in productivity through automation and robots. And that hasn't come to construction yet.

To see the autonomous excavators in action would require a quarry. And by a stroke of luck, Dominik had one.

We're in a quarry, obviously.

Yeah, so this is an active gravel quarry. This is where we build up most of the new machines and test what we develop.

One point of all this is the same equipment on each one, and they're 3 totally different vehicles.

That is our— story, right? We developed one retrofit kit that can go on any machine. And the wheel loader is completely different from the excavators, but it needs the same sensors, needs the same computers, the same networking equipment. It just sits on the roof, um, there's the two sensor heads.

So part of it is just the sensors and everything on the top, and then you must have to make some kind of connection into the internals.

That varies quite a bit between machines. Now here they all have an electronic pilot stage, This means that there's an electronic valve somewhere in the machine that controls the motion of the machine. Whereas in old-school machines, you don't have this. When you move the joystick in an old-school machine, you directly move physically a valve, a hydraulic valve. So there's no way of interacting with that machine from a computer.

How long does it take to put the kit on?

This only takes a couple days. So we either work together with the manufacturers.

Yeah.

Or we do it ourselves.

All right, I'll follow you. Try not to get squashed.

So there's the rack here in the in the front. So you, you see one camera and a LiDAR on it. Also the computers and parts of the networking equipment is in there. Then this is paired with the sensor heads that we have back there, right?

Okay. And so camera, LiDAR, and then camera and double camera per mast and LiDAR. So there's what, 3 cameras?

5. 5 cameras because there's another mast on the other side.

Okay.

And this is giving you like a 360 view, 360° coverage of LiDAR and camera. Okay.

And then there's sensors on the arms.

So for example, there in front of the light, there's one sensor there. They monitor the angle of the individual links.

Okay.

And with that, the machine knows the posture of the arm. There's a couple of pressure sensors as well that we install in the hydraulics for getting additional measurements. Otherwise, yeah, nothing else.

As often tends to be the case, it's software that makes the hardware magic.

Here, this is our tablet. On the top, what you can see is our image view. It's actually the front view camera overlaid with our augmented reality features. In the middle, it's the full 3D view, and the colors indicate blue too high, green would be on grade, and red is too low. And then the thin blue line below here is the target that we want to dig to. So what I can also do here is I can very quickly draw a trench. So for example, with the free tab, I just mark the far corner, a near corner, and then the width of it and simply press start and the machine would start to dig this trench.

Here we go. I've never seen a robot dig a hole before.

What you could hear quite nicely is the bucket doesn't stall. It doesn't get stuck. But still, there's quite a lot of force. Putting enough force into it but not stalling the machine is the art.

Do you call this a robot? I mean, it's like an autonomous system attached to a regular machine.

For me, it's very much a robot. Our system is just the sensors and the brain. The body is something we buy. It's just a machine off the shelf.

Yeah.

So now it has finished digging and what it's trying to do is just a very detailed, fine pass of the bottom of the trench. And that is going to ensure the accuracy. It's going to now move back again a little bit. And then continue digging.

That was cool. Yeah, it had a little finesse there at the end. Yeah.

Gravis is anything but alone when it comes to construction bots. There's Built Robotics in the US and Komatsu in Japan. And there's Lumina, which apparently makes this thing that is clearly meant to re-sculpt Mars. Beyond all that, there are bots to lay bricks, spray orchards, and pick rocks. So, like, the robots are coming.

So this is the teleop zone.

Yes, so we have here our teleoperation stand. This is the same tablet that we had inside the cabin. So now what we can do is we can also here now set up an autonomous task, or you can take over manually for remote operation.

So the same idea, somebody might start out there and then and come back here to do some other job. Okay, cool. And I get to try to drive it?

You get to drive it.

Okay, okay. Am I allowed to try to dig now?

Yeah, you can. You push your left hand forward while also pushing the right hand forward in order to go out and down. Yeah, stop, stop. Pull back and pull back with the left hand at the same time.

Oh, you didn't tell me that All right, okay.

Just straight down, right there, right? And now if you go with your hand now to this side, you're going to start closing the bucket. And that's also— that's a very simple digging motion, just with, with the bucket joint.

So I got my scoop, but now I have to go put it somewhere. Yeah, where do I want to put it?

So now you can extend the arm by pushing the left hand forward, and then also with the right hand to the right, you then start uncurling the bucket and you dump it.

That was pretty fun.

These machines are a lot of fun.

In the vast spaces between ambition and execution, Lies the architecture of tomorrow. Puma. The future remembers everything. Even the ideas too quiet to whisper. Core memory meets brass.

After a couple of hours at the quarry, I grew tired of thinking about construction workers losing their jobs. I needed a lighter side of robotics. And so I returned to ETH to be mesmerized by this. This contraption comes out of the lab run by Raffaello de Andrea. Raff got rich selling Kiva systems to Amazon in 2012. But what's more interesting about Raff is that he's perhaps the world's foremost robot artist. He's made wonderful, weird installations and has a company called Verity based in Zurich that puts on fantastic drone shows. Raff's students have embraced his bot-meets-art philosophy.

What is the game called?

The Labyrinth game?

Yes. You can buy them on Amazon, most of the toy stores. It's very widely available.

Okay, so that's computer vision.

Yes. It's really very simple. We just have the camera and we have two motors, and that's kind of the system itself. And then we just have this third component of this reload mechanism so it can learn autonomously.

So the first time you set it on there, it knows Nothing?

It knows nothing. And the learning is very gradual. So it learns how to go past this section. And once it knows how to do this, it kind of proceeds to the next one. And then it really learns in a very incremental fashion. When you just watch it learn, you really see how it gets better and better.

How long did it take to get good?

At the start, it was about 5 to 6 hours until it gets to like this proficiency. We got it down to about 1 to 1.5 hours. So it only needs to practice 1.5 hours without knowing anything, and then it kind of gets to here.

And the human record is 15.1 seconds?

I think so.

And this has gone—

About 13.8, if I remember correctly.

Are the humans upset that you've built this?

Um, I don't know. So on one of the videos, there's a comment saying, "AI does it better." I guess it just shows how far these algorithms have come. It's not gonna take too long until those robots really can do all the tasks that humans can do.

Thomas and his companions, Aswin and Antonio, are the elder statesmen of Raf's lab.

For me, it's the attraction to being here is first Raf. He's a scientist and he's an artist. And all of us have that in common, I would say, in some ways. So Thomas's master project was letting the four-legged robot dance for music.

That's right.

Thomas makes amazing photography. Antonio is an amazing dancer.

Well known.

Right?

These guys are training up a new crew of young roboticists. And on the day I visited, their students were hard at work on something that would make Raff quite proud. They were building robotic swans that can dance on the water.

This is the innards of the swans.

Yes, and these are the people that built it. So that's like the swan jet nozzle that has a water stream coming out of it. We can just put a pump at the bottom and just push the water up. We said, now we want to work on building robots and capacity for autonomy on water.

As day turned to night, the students went to a nearby lake to test the RoboSwans. It was an adventure that went on for several hours and culminated in this. And, in its more idealized form, is quite something to behold. We are almost certainly heading into a great age of robotics. The combination of AI and better hardware is making a raft of new things possible. The US, China, and Japan are leading the charge in this new era. But thanks to its engineering pedigree and ample government support, Switzerland too will likely be a major player shaping the sci-fi future to come. This is no small feat for a tiny country locked inside of a continent that seems to abhor progress. So good on you, Switzerland. Respect. Be like the robot. Hit that like and subscribe.

Oh!

Oh!