Building Safe, Scalable Robots at GM With Péter Fankhauser

Transcript

Péter Fankhauser (00:00)

Ultimately, I'll call it nobody wants a robot and you cannot you know, they the customers are looking for a solution

 

Brian Heater (00:07)

I don't know if I'd go that far. don't know if I'd say nobody wants a robot, but...

 

Péter Fankhauser (00:12)

Our customers, a robot is just another thing that they need to pay for. It's another move. It's an expense, it's also a risk. Something is moving. So it's an investment. It's a tool, but that tool needs to serve a purpose and that is solving a problem.

 

Brian Heater (00:38)

Hello and welcome to another episode of automated. name is Brian heater. am the managing editor of the association for advancing automation. We have a another great conversation for you this week as we go international to speak with Peter Fankhauser, the co founder and CEO of the ETH Zurich spin-off ANYbotics. We'll dive a little bit into the European robotics community and a whole lot more. So stay tuned for that. And if you're enjoying the show, don't forget to like and subscribe and we will see you on the other side.

 

AD:

When you're coordinating complex multi-axis systems at millisecond precision, your controller is either your bottleneck or your advantage. Sanyo Denki Sandmotion C300 supports up to 32 axes with Ethernet Cat and one millisecond control cycles. Code to space programming with flexibility engineers expect. Built for reliable uptime and real machines backed by Sanyo Denki motors, drives and absolute encoders. Complete reliable motion control from a single source. Sand Motion by San Yodanke, precision in motion. Learn more at sanyodanke.com slash america.

 

Brian Heater:

I know obviously this isn't my place. ? I'm not in marketing, but yesterday I ? was thinking about some potential names for what antibiotics might call a humanoid robot. I know I was like, obviously you're not here to make any big announcements, certainly not at the top of the podcast. That's the kind of thing that maybe like 45 minutes in when nobody's really paying attention anymore. But I landed on one. I feel like there's only one thing you could possibly call. a humanoid robot. I'm sure you've had these conversations. I don't know how serious they've been, but I want to know for on the same page about this. ? Have you have you have you had these like top level discussions and do know what you would call a humanoid robot?

 

Péter Fankhauser (02:41)

We certainly play around with the any, and we have any male, we have any drive, we have all of these, even have an any soccer club. One idea could be, you know, not genderizing our robots, but any woman would be pretty cool.

 

Brian Heater (02:54)

Okay, so yours is gendered. I'm going with anybody. I feel like that's the only possible choice.

 

Péter Fankhauser (03:01)

I like that choice. That's better. Yeah

 

Brian Heater (03:04)

I don't know. To me, maybe that is a sign. Maybe you're not as far down on the road as this is. As maybe I might have suspected. This is something that obviously, like everybody is exploring some form of mobile manipulation by dexterous manipulation. When it's really in the news like this right now and when obviously a lot of funding has been being thrown at humanoid robots, is that something that all of sudden that you are taking a look at?

 

Péter Fankhauser (03:33)

Oh yeah. In general, by dexterous manipulation, general manipulation. It's a big topic. And for us, we are a four-legged robotics inspection company. For now, that's, need to be very focused, but our mission and vision, we call it creating a workforce of autonomous robots. And so starting to interact with the world and changing, manipulating the world. And especially where we work in the industry where it's also about maintenance and eventually repair. Of course, the two arms or multiple arms are. You know, a great solution. Um, it doesn't don't need to come in a humanoid fact form, meaning that two legs and the head and two arms, we can also strap two arms on a robot. for us, we're not just running blindly behind copying the human form. Well, I often observe now, you know, with all the human rights companies for us is the function, what it needs to define to form and the beauty of robotics that you can optimize the shape quite easily. You know, don't need hundreds of millions of years of evolution. then again, if you want to change your body in robotics within a few months, we can build a new prototype and every couple of years you can build a new hardware, but still being driven by the same fundamental technologies. But I'll leave it at that for the moment.

 

Brian Heater (04:46)

I partially ask because I know certainly here in the States, I know that obviously a lot of the companies that have backed you, like Qualcomm, for example, are US based. obviously there's a lot of trend chasing happening, AIs in the news. So all of a sudden everybody's got AI in their pitch deck. Humanoids right now, everybody's of pushing that in as well. backing up to your last raise or even moving forward to a future raise when you're having these conversations with investors, whether there is some external pressure to kind of push towards some of those trends.

 

Péter Fankhauser (05:28)

Ultimately, we're doing the right thing for the business, but we're not in a state where just, you know, we want to raise funds on hype. want to raise funds on merits and business. And I think what we have been able to achieve over the last couple of years is to really build a solid business, showing tractions in actual commercial operational deployments with customers. Our robots, you know, collect hundreds of thousands of autonomous inspection points a month. So we're starting to see, you know, the business outcomes. And this is the foundations of on what we're building our company, the usage and revenues we're generating with real customers and the scaling customers. So that's the foundations. And then of course, on top, you discuss what's next. And there's several things that we're excited about that I'm sure we'll talk about. One is, you you have all this data of the robots, how do you make use of that? So generating, going from data to insights, eventually recommendations, not only being a robot, but being a full solution that helps our customers. And then the second is certainly how do you expand to generate even more tapping to new value pockets? And this means also different types of platforms. So yes, we are already building the next generation of robot, right? Anymolex fundamentally very similar and it's going to come out next year, but then we look beyond that. So humanoid is one form, but strapping wheels on it, magnetic wheels, drones. So that discussion, cannot disclose too much, but yes, absolutely. We're looking at expanding the portfolio different types of robots but for very similar customers in the industry.

 

Brian Heater (07:01)

I actually wanted to talk a bit about productizing because obviously, you know, in terms of form factor, I'm sure that you're sick and tired of getting compared to Boss of Dynamics, but you know, Quadruped probably comes up a lot. But I know that, you know, in my conversations with Mark Raybird, their switch to productization wasn't the most natural for him, right? I mean, he's he's obviously he's an MIT guy. He's a research guy. He had been doing that for a long time. They bring Rob on, Rob becomes the commercial guy. And I'm wondering, know, over the course, obviously, NABox has been around for a long time. You you've been doing this all the way back to 2009. Whether you were a research company first or whether the plan was always to move in this direction of actually selling robots.

 

Péter Fankhauser (07:51)

Yeah, that's a great question. And for us, was, we came out of ETH Zurich, so the MIT in Europe, one of the leading robotics lab. And when we founded the company, was already clear that we cannot just continue what we're doing. We cannot be a research company. That's in Switzerland. We couldn't fund that. And so from the beginning, we said, Hey, we need to be out there. We need to, even in very early phases, we need to ship robots, test them to POCs. I myself. Stopped, you know, loved programming. have a PhD in robotics in motion control and planning. I stopped the day that essentially, you know, I started with the company and was clear, Hey, we need to be out there, sell marketing, field engineering. And so I spent, you know, many, many weeks and months with customers. I'm offshore certified. So that became also a cultural element of the company. Yes. High tech. We're pioneers and we're building, but it always needs to have a real impact. Right. and we take responsibility for our technology. There's a lot of cultural elements that we baked in from the beginning of the foundation of Anybotics.

 

Brian Heater (08:56)

There's a lot of things I want to get to in there, but one of them is you said you're offshore certified. Does that mean you've been hanging out on oil rigs?

 

Péter Fankhauser (09:02)

Yeah, sure, wind, but I could also go oil rigs. So that means you can transport with helicopter transfer, know, all the emergency exits, fire fighting, all that stuff that you need to do.

 

Brian Heater (09:12)

There's like a bunch of training involved in that, assume. What did you have to go through in order to get that kind of certification?

 

Péter Fankhauser (09:14)

That's correct So does the basic health checks, emergency response, health, firefighting, the coolest one that I enjoyed is the helicopter transfer. You have to assume that the helicopter will need to land on water. So you need to, it's a helicopter escape training. So they put you in a fake helicopter upside down, you sink, you have to release the vest, the breathing, open the windows, climb out. So it's an intense, a very cool, hopefully you'll never use it, but an interesting experience.

 

Brian Heater (09:49)

Yeah, exactly. Hopefully you'll never use it. I mean, is this absolutely necessary for the CEO of this company to have this kind of certification?

 

Péter Fankhauser (09:56)

Well, maybe not today, but in the beginning, yes, you built the company on knowledge on what customers need, the language, the problems, the specifications that was crucial. I asked the founders and multiple people that is at the company to be out there and learn. yes, offshore, especially oil and gas, but also wind is one of our core industries. So it's good to have that in your back pocket and having actually been there.

 

Brian Heater (10:21)

That's interesting. So it's twofold. it's one sort of like knowing what they have to go through the process to get certified. But two, it sounds like you actually ? early on when you were sort of, guess, looking at use cases that you actually spent quite a bit of time on some of these platforms.

 

Péter Fankhauser (10:41)

Yes, 100 % because, know, ultimately I'll call it, nobody wants a robot and you cannot, know, they, the customers are looking for a solution.

 

Brian Heater (10:50)

I don't know if I go that far. don't know if I say nobody wants a robot, but...

 

Péter Fankhauser (10:55)

Our customers, a robot is just another thing that they need to pay for. It's another move. It's an expense, but it's also a risk. Something is moving. So it's an investment. It's a tool, but that tool needs a certain purpose and that is solving a problem. And so what is that specific problem you're solving? And the robot itself does not solve that problem. So it's all about what do do with the robot, the use cases, the applications, the data you're gathering and how you're helping the customers. to run their facilities more safely, high productivity, environmentally friendly. So whatever they're solving, you need to be able to draw that connection from your basic technology to their needs. In between, that is a lot of innovation and to be able to do that, you need to speak the customer language and have gone, walked in their boots and have had their experience. So yes, that has been fundamental for us and we have a significant team of customer success, product managers. They are constantly out there with customers because we're not done in learning, right? This is a long, long journey and we're still ? learning along.

 

Brian Heater (11:58)

Do you feel like there's one particularly surprising thing that you learned in the process of being out there in those rigs?

 

Péter Fankhauser (12:04)

Well, you know, as a optimistic, maybe naive engineer, you think the world all figured it out and it's highly automated, digital twin, everything is connected with IOT. And then you go and look and it's often not always, but often very, very rudimentary. So they have cases where they ask us, Hey, can the robot push this button because it's not automated and we need to restart here. And it's an automating will be such a hassle or they use pen and paper. They put it in a folder never to be looked at again for data that they're collecting, isolation of data, teams not speaking to each other. And then of course, know, pencil whipping is a topic, people not even doing the checks of inspection. So there's so many things that are just in theory not necessary, but it's just a human factor. ? Constraints of hiring people, constraints of accessibility, safety. The machines often tell you what's wrong, but the relevant data is often not collected. and then also not connected to the relevant stakeholders. So yes, it's quite eye opening once you spend time at these facilities, how rudimentary things are often still run.

 

Brian Heater (13:10)

Yeah, that's interesting because I tend to think of like inspection and a lot of what the robot's doing as being kind of like a fail safe in case anything goes wrong. But the humans are there to occupy the space ? in order to I mean, they need to be there like something needs to be there every single day. It sounds like in order to just make sure that operations are working as expected.

 

Péter Fankhauser (13:36)

Yes, mean, these facilities, are built often and still are built with, you know, with continuous human presence in mind. And yes, you can automate a lot, IoT, DCS, Scala system, all of that exists, but still you cannot place for every single problem in every location, the sensor, right? So you still have people either being ignored and you have reactive maintenance, you run into failures, or you do very scheduled maintenance out of fear, you overdo it. So it's a big investment. Or you start collecting data either with humans, which is very cumbersome distance, or now with the availability of robots, you can automate that part. And it just provides you a 360 perspective on the outside of the facility, on the infrastructure. Imagine, know, leakages, gases, you don't know where they're to come from, liquids coming out, overheating equipment, pumps, generators, cables, transformers. There's so many things we can talk about, things that can go wrong. Ultimately, it's the peace of mind knowing early enough that you have signals of the machines that are about to break so can you react properly.

 

Brian Heater (14:43)

I haven't actually contextualized it like this. When I think about greenfield versus brownfield solutions, I'm thinking about factories and warehouses mostly, right? I mean, that's usually what we're talking about when we talk about those. But I wonder if there's a sense in which something like an offshore wind or offshore oil, you can kind of look at one of these robots as being a brownfield solution where you're sort of slotting it into existing processes. Do you feel like that's a case where like, eventually those systems are going to become fully automated?

 

Péter Fankhauser (15:18)

So Brownfield, so existing facilities, it's a no-brainer, right? Those facilities are built some decades ago, but still need to run for many, many years, decades. And also you want to expand the lifetime of that asset. Second is Greenfield. Also then we observe even the most modern, let's say offshore wind facilities. We had a case where some customers complained they had 40 % false positives from their built-in automation. And in that case, you don't know, is the sensor broken? Is it actually a problem? You don't know what exactly the problem is. So you're sending out people and you always need to send out the crew with some tools without really knowing what spare parts to bring, et cetera. So it's very, very cost ineffective. So having the robot in those cases also as a backup and cover to see, get a situational awareness before you send in the person for safety, but also cost, reason and transport makes a ton of sense. And now what we observe what's happening is mobile robots in all shapes and forms will be a core citizens also of the Greenfield platforms. And we're actually speaking now to the builders of the Greenfield facilities who say, now with robots in mind, I can build my platform differently. So I don't need to build all the extra safety precautions. ? I don't need to build the survival, the people are able to live there, the ? water desalination, all those expenses that make sure that people can go there and you build it with robots in mind. So more confined spaces, more narrow, less safety, and ultimately more cost effective because you can deploy robots.

 

Brian Heater (16:54)

I mean, I assume that like you still have to to a certain extent build it so that people can get there people still need to be the ultimate

 

Péter Fankhauser (17:02)

I agree. Yeah.

 

Brian Heater (17:05)

Yeah. Can can animal at this point press that button if need be?

 

Péter Fankhauser (17:11)

No, that button, we were working obviously on manipulation solutions that the frequency of ask for button press is not that frequent that we have said, let's build a custom solution for that. We're really focused on data collection these days, visual, thermal, acoustic, partial discharge. There's a lot of data collection that we're focused on and not only collecting, but also then interpreting the data, analyzing it, because you know, imagine collecting all those thousands of points. There's no human. wants to look at that data. So what operators want is to clearly flag them problems, anomalies. The green stuff, the green lights, that goes into archive, good to know and good for historic purposes to dig deeper. Orange, the amber lights, they go like, we might ignore that because it's not critical yet. Tell me what's red, what's on alert status. And this what we're consistently flagging and helping the operators take the decision in action.

 

Brian Heater (18:08)

Something I wanted to back up to that you had mentioned earlier that jumped out at me as well is you said, when we were talking about productizing, said, ? I think something along the lines of in Switzerland, ? we basically like we can't just be like a research facility, which is interesting to me because like, that's something certainly that I would say about an organization here in the States, you know, and a lot of that is up to kind of really The pressures of of of late capitalism is it's are you are you experiencing like similar pressures there? You feel like it's it's difficult to just sort of like survive and thrive in Zurich as a pure ? research organization.

 

Péter Fankhauser (18:50)

Since we're not a research organization, cannot comment on it. I don't see it.

 

Brian Heater (18:54)

But there was a point, I mean, you made it clear that there was a point when it was like, you know, we kind of need to make it sell products in order to really go forward.

 

Péter Fankhauser (19:03)

But it's also what we wanted, honestly, right? It was financially something that you have to do. Otherwise, you're at university, right? Or in academia or a research center that's state funded.

 

Brian Heater (19:13)

It's fine, that's a whole different...

 

Péter Fankhauser (19:16)

Also my motivation, the motivation of the team, you know, we always said, let's be useful. Let's take the robots outside into the real world. Let's do cool stuff and let's make people, you know, successful. The moment of building a robot is very, very cool. And you see it work, but the real success for us, what we strive towards is when customers use it on a daily basis and then give you two thumbs up and say, this is awesome. I want even more robot. That's, that's why we're doing it. And that means you're becoming a product or even a solutions company. Right. Because you know, robots today, it's not that commoditized that you can say everybody knows how to use a four legged inspection robot. Sure. No, we're going quite deep advising customers how to use it, which assets, how to optimize the reliability processes, how to make sense of the data. this also, you know, part of the sales, but also after sales advising the companies how to use robot. It's very rewarding because we help them essentially in their operations. as almost a traditional company in consultancy, but the difference is we're actually using robots to get the job done. And this is very cool for us because we often say we have, you know, 400 employees, 200 of them are human. know, and the non-human part, the robot part grows every week. So we're shipping a lot of robots and that makes us very proud. The robots are part of our team as well, and we deploy them with our customers.

 

Brian Heater (20:41)

So you come to ETH around what, 2006, 2007?

 

Péter Fankhauser (20:47)

2007 right from ?engineering

 

Brian Heater (20:49)

You come as a mechanical engineer and this is something this is so this is my humanities creative writing major part of me something that ? that I need to ask you and feel very silly asking but ? how different of a major how different of a focus ultimately really is robotics because it strikes me there's like probably a fair bit of overlap between mechanical engineering and robotics.

 

Péter Fankhauser (21:19)

yeah. So mechanical engineering, of course, there's a lot of traditional mechanical engineering, but you have electronics and a lot of software control systems. So I very soon moved into a lot of programming. did that as a kid. And so for me, the control systems planning, the software, that became the core part of my focus. Then also in the masters and then eventually also in the PhD. And robotics is core part of the mechanical engineering department.

 

Brian Heater (21:48)

I mean, it sounds like you were being, I guess, I don't know. Obviously, this this predates. ? company, but it also sounds like you were being very pragmatic in your choice of a major as far as like, what can I do? What can lead to a career from here as well? Right. And and maybe no. OK, no, not at that point in your college career. OK, I'm just curious from the standpoint of, know, like going back to 2007, I'm just trying to think of, you know, the state of robotics at the time and whether it seemed then I know it. It was and it wasn't all that long ago, but at that point, whether it seemed like feasible that you could make a career in robotics.

 

Péter Fankhauser (22:31)

No, probably not. And I didn't do it for the career, right? I go where the creativity, where you can create things and where your energy takes you. I think for a lot of the people in the lab and co-founders, this is how we found each other. And for me specifically is I love creating things to be creative and building things. And robotics back in the days was so cool for me because it was so brand new. was completely under the radar, essentially. but it was such a vast open field. It's this big, wide canvas you can start drawing and painting on. And so I was fascinated by the sheer opportunity to do that. And then of course, I'm very happy that now it's becoming more and more the center of attention of, and being able to build a company around it is a massive privilege, but I didn't think that was down the line an opportunity.

 

Brian Heater (23:27)

How do you or can you, and obviously this isn't a problem specific to robotics, this is certainly a problem in any creative field and a lot of jobs in general, but when that is your passion and when you're so sort of excited and invigorated by the newness of it, like a decade down the line or so when you've made that shift, when you're not hands-on, when… As you said, you're not really kind of like a roboticist in that same way anymore. Does the romanticism necessarily go away? And when you're effectively kind of focused on the same product and because it's a business are moving in a lot of ways at a much slower pace.

 

Péter Fankhauser (24:12)

We are still in a market that's very, very quickly evolving. So for us, more than half, two thirds of R &D work on the next gen topic, that next software, next gen platform. So there's no end in sight and we're moving very, very quick and we have to, right? This feels like the Nokia era of robotics where yes, you have first devices, but you're going to look back on those devices in a couple of years and you'll smile and go like, that's old school stuff. The world is moving very, very quickly and it's still very exhilarating and exciting to move in that space. But you're right. I'm not hands on programming or designing the robots anymore, but energizing the team thinking of where do we play? What is the right investment in tech? Still thinking about the right product to build. So that's still quite hands on. Although of course you're not doing the designs and the programming yourself.

 

Brian Heater (25:07)

Yeah, one of the things that I really appreciated just being in and around robotics people is that there just really does seem to be because it's robotics as a career. It's both big and it's small, right? Like everybody to a certain extent does know everybody. And there is. to an extent some competition, but it seems like people like largely are cheering each other on and are interested when they see progress from other companies. And I assume again that you're probably sick, especially from people in the US of the boss dynamics comparison, but you had to be sort of watching their progress from afar with some interest.

 

Péter Fankhauser (25:52)

Now, course, so 100 % that I'm, know, Boston MX is a great company. I'm very close to their leadership team. We often, you know, speak and compare notes on what do we see.

 

Brian Heater (26:03)

Mark seems to be, ? the other Mark seems to be a big fan of yours.

 

Péter Fankhauser (26:08)

Yeah, no, great guy and close exchange, but also with X robotics and others, right? We need to build the market together, which means customer needs optionality. And while Bostonomics has a fantastic brand, right? And many more resources than we do. We are, I would say more focused. are more vertically integrated. So according to specifics for customers, often we are clear in the value proposition but we're also working on a smaller market. So whenever we are direct, we have a very fair chance to win, even if we're a smaller company. So it's interesting as the constraints of your brand visibility and financials also make you creative how you can still play the game with a bigger company.

 

Brian Heater (26:57)

I partially ask because it's really interesting and I advise everyone who's interested in the field and maybe who doesn't know a lot about the company to go take a look at this timeline that you have up on your site like for 2009 until now. And we've got a very good editor on this program. She's going to be tossing up some of the earlier robots are out here. it is interesting to see what in biology they would call like bit of like convergent evolution happening, right? Like these two different product lines kind of happening in parallel, you know, there's a reason why quadrupeds make a lot of sense in a lot of cases. And there's a reason why, you know, kind of certain aspects of them end up being somewhat similar to one another.

 

Péter Fankhauser (27:41)

Yes, and this is all driven by the environment and the tasks that a robot needs to operate in. so for us, for industrial work, and especially if you think to be able to go anywhere where humans go means you have a certain leg length to be able to climb steep stairs. You have to need certain compute. need to be autonomous. You need to carry your sensors. That determines the shape and the size of your robot. Then there's lot of ? differences in the details. Ultimately, the size and the form factors become very, similar. That's true.

 

Brian Heater (28:15)

What do you feel like in terms of like, again, being really focused on the industrial side of things like from a hardware perspective, what do feel like defines your robot versus other four legged robots out there?

 

Péter Fankhauser (28:26)

Yeah. So there's a lot of four-deck robots. Almost every day you see somebody also building one. So what we do differently is we're really focused on the industrial work, meaning that it's a ruggedized unit. It's less of a toy, but also when you touch it, right, it's a sturdier robot, more powerful, a little bit heavier, but also completely water and dust proof. This is a key differentiator. Our robots are being used in very harsh environment. Dusty corrosive environments, outdoors, offshore, you can imagine it, right? So the IP67 completely water and dustproof is a key differentiator. Then certifications is a big topic. So the safety light, the e-stops, these are things that are not very attractive on YouTube, but being able to send in a robot into environments where there's explosive gases is unique and by law mandatory. So that's why we also built an externalization robot, AnimalX. It's the world's first and only robot that can actually do that. It's a must for the oil and gas and chemicals industry.

 

Brian Heater (29:25)

I detected a little bit of shade there when you said not very attractive on YouTube. You throw in a little bit of was there a little bit salt in your voice there when you said that.

 

Péter Fankhauser (29:33)

Well, no, it's, you know, we're not hunting behind just YouTube clicks. That's not one of our KPIs, which I see now a lot of human rights company do public attention. We want the attention of industrial operators and what gets them excited is knowing that they can safely deploy the robot at their facilities, which are obviously, you know, hazardous environments. Explosive gases in there is no joke, right? And the last thing you want is being responsible for an explosion.

 

Brian Heater (30:00)

I say this, this was actually something that I want to talk about because we're at an interesting point right now. I wanted to get your perspective on this. know, one of the things that's really happened again, sorry getting back to Boston IMAX, but something that they did really well in the early days was a really cornered YouTube, right? That's a big part of their success. In more recent years, especially with humanoids, we've seen the line really then start to blur between, you know, I just wrote a piece recently and I said, kind of at the point where it's hard to tell a robotics announcement from a ? car advertisement. ? You know, I get it. I get that there's a lot of money going into and coming out of these systems right now. But I'm worried that ? especially with the humanoid gap right now, that there's a lot of false expectations, that there's a lot of kind of like bending of the truth happening. as a company that one both has this like connection to ETH has this this research backgrounds, but two has a cool looking robot and does like to put out cool looking robot videos. How do you balance the two?

 

Péter Fankhauser (31:10)

There's an advantage because if you have a cool-looking robot, opens doors with customers, with investors, employees.

 

Brian Heater (31:17)

I mean, you're not quite open doors yet, but maybe in the future.

 

Péter Fankhauser (31:21)

But yes, it's your entry ticket, let's say it opens the doors for the discussion, which is great. But then you have to be very careful because if you just go in and say, Hey, I have this AI driven robot that can do everything. That just creates wrong expectations. So you want to open the right doors and have the right discussion. So we're very selective in which industries we're focused on. And we'd rather say no to a customer is where we believe we cannot solve the problem. And so it is an entry ticket, but also you need to use it wisely without power. And for us ultimately where in our storytelling and where we focus, the robot is not the hero. And I know in Loftman and videos, it's all about the robot, robot, robot. But actually, who is the hero of the story? It's the people whose life we make easy, right? Where we provide additional safety to the operators, the reliability engineers. So we tend also marketing communications to shift much, much more on the impact we're having with those people.

 

Brian Heater (32:15)

Is there a certain amount of like disclosure? You know, I don't know that there are like truth in advertising law, obviously, again, be different in the US versus the EU, but maybe standards for just like disclosure of what robots really can and can't do. I people to read my stories to have more realistic expectations. And is there something that the people who produce robots do to help set those expectations?

 

Péter Fankhauser (32:39)

Honestly, for us, that's not even a question, right? We show this stuff from real-world deployment. But yes, I understand. I think people are starting to be very skeptical, even too skeptical. Now, the pendulum is moving. Interesting. They claim, this video is AI generated and fake. ? Ultimately, show stuff that you can solve. And the trick with robots is if there's still a human in the loop, it's much, much easier than solving it autonomously. even manipulation, even just navigating a robot through the forest, right? For autonomy, it's actually not that easy to find the right path. As a human, you can see if the legs are stuck, you'll get back and forth. So you're helping the system. That's a remote control tool. Also very cool, but the usage of it, people assume if you don't put the human in the video, they assume it's autonomous. So their transparency, you can fake it for a while, I guess, but ultimately once you start shipping to customers, these things will be very… So leading through transparency, think, is obviously very important and creates trust.

 

Brian Heater (33:43)

The pendulum slipping back is an interesting point too, because that's another interesting avenue of this conversation. I don't want people to fully discourage teleoperation, because there's a place for teleoperation with your robot too, I assume as well, when we're talking about these systems being in remote locations.

 

Péter Fankhauser (34:02)

Yes, there's no shame in helping your robot fleet out. And that's the path to, you know, ever growing fleet. There will always be a long tail of situations where the robot can get stuck when he's up and that's completely fine. You know, with autonomous driving, same thing. And so just recently I walked up, I saw one of my colleagues pretty much hanging in the chair like this and teleoperating one of the robots and looked at the screen and it's completely offshore. Just the helicopter flew away, really nasty environment. going upstairs and he sits in the warm office. And the reason was why he teleoperated with, they had big construction sites going on and he just wanted to make sure that it's all safe, right? The robot doesn't get lost or does anything it shouldn't or trade off the path it's supposed to. So yes, they told us it's constructions for a couple of missions with the teleop and went back to autonomous. No problem. That's the way it should go.

 

Brian Heater (34:54)

One of the things you mentioned when talking to clients was ? obviously like only choosing clients based on problems you feel like you could really realistically solve, which makes a lot of sense. I wonder with robots like this, especially a lot of the news stories that we've seen over the past several years, ? at a certain point, do ethical considerations enter into those questions as well?

 

Péter Fankhauser (35:21)

In what sense? for us, we don't really military, example, weaponized. Okay. Yeah. So for us in the beginning, that question came up because you're still so broad figuring out what, what actually you want to solve. ? we said to founders, it was important to us to set some guidelines. said no active warfare or weaponized applications for robot. And we held truth to that also published with our friends at Boston dynamics and others, you know, the

 

Brian Heater (35:25)

Exactly.

 

Péter Fankhauser (35:49)

The responsibility, you know, the white paper or the letter ? about weaponization, because roboticists carry a responsibility for what they're building. And so we keep up to that truth. The topic recently hasn't come up so much because we're seeing great traction and progress in the industrials. And also our product tends to be much, much more optimized for that user group. So there's less of distractions of trying to do something else.

 

Brian Heater (36:14)

Maybe this is just like my outsider bias, but I wonder how much like being in Switzerland plays a role in maybe not wanting to be involved in warfare as well.

 

Péter Fankhauser (36:22)

Certainly it plays a role from our upbringing. Yes, but I also understand that I don't condemn other robotics companies that move in that space. The world is changing very quickly. ? Ultimately, it's your decision, but then you should also stick to it. Just doing it opportunistically, I don't think that's right approach. If you do it, then go all in.

 

Brian Heater (36:43)

Something I heard you mention in a recent interview was this idea, think the way you phrased it was pioneering reinforcement learning for legged robotics. And again, I'm not entirely sure where the timeline is versus other legged robots, but what do you mean specifically? What were the breakthroughs that were happening there? How much of that was happening on the ETH side? is that a continued relationship between that research and what happens on the commercial side of things.

 

Péter Fankhauser (37:14)

Yes. So we came out of ETH Zurich, ? ETH and also two professors, Marco Uttler and Roland Sigurd, are both co-founders of the company. Also Marco serves on the board. So there's a very close relationship and we sponsor PhDs at ETH. So there's a very close relationship pushing the frontiers of robotics. Now to the story of reinforcement learning, and I personally don't take credit for that. It was really at ETH, the research was there.

 

Brian Heater (37:41)

You can if you want to. I won't tell anybody.

 

Péter Fankhauser (37:43)

Okay. I'll tell you what I take credit for, but in 2018 and 19, we saw first progress of deep reinforcement learning to robot training itself and simulation. And then we were able to transfer that successfully to one of the robots. And so this was pioneer work, one of the first ones in the world to successfully do that and deployed on a physical robot. And what we take credit for though is recognizing quite early on that this mix of AI and robotics was very powerful. So after the first experiment, I saw it, we said, we stop all the traditional model based control at the team. We shifted the entire team, sounds big, was five people in the control space over onto reinforcement learning. And within six months, we had the first robots walking in a way that we couldn't deploy to customers. And I'm very proud of the achievement there because traditionally model-based control, where also Boston Amics has been leading for many, many years. Big team is hard to achieve. And we very soon met or even outperformed what you can achieve with model-based control with a very small team, thanks to AI. So this is really an S curve of the next technology coming in. And we wrote that from the beginning.

 

Brian Heater (39:03)

You alluded to this a little bit before, you know, to a certain extent, obviously the team's much larger now, but to a certain extent, early on, were there advantages to being a really small team?

 

Péter Fankhauser (39:16)

There's always advances to being a small team. You're, you know, decision-making information, all of the things just don't slow you down. But eventually you need the muscle ? to, to, just the head count to, to, do more. Right. It's funny, the more people you are, the less you can actually do, the more you need to focus, slows you down. But yes, you need just, you know, for, for, for sales development. There's so many functions that we cover. We are full stack, meaning we're building the hardware software And the software is not only the robot software, but the cloud software analytics as well, the production quality management. So it just adds up and you need specialists for all of these roles and the company becomes bigger and bigger. Ultimately right now at 200 people, we have found a pretty good groove on where we're at. So we're not looking to double every year. We're rather focused on hiring the right people, efficiencies, maintaining the burn within limits.

 

Brian Heater (40:10)

I always want to know why people like go full stack, you know, and don't don't end up focus on just like a specific part. Was it was it just that the more you explored different aspects, the more you felt like they weren't going to quite do the job that you needed them to do?

 

Péter Fankhauser (40:28)

In the beginning, really had to, there was nothing off the shelf that you could buy even to the level of batteries, know, or motors. We looked around and looked at all the industrial solution. was just nothing there. So we said, look, instead of here, ? trying something unoptimized from the industry, let's build it ourselves. We're engineers, we'll figure it out. Of course, you're naive in the beginning. And so it takes more time that you think, but ultimately a robot is a very integrated, lightweight, also thermally managed system. So a lot of optimization needs to happen. But what's happening now is that the industry is maturing and also thanks to Humanoids, more components and subsystems starts to be available that are optimized for lightweight mobile robots. So whenever somebody can build something that we need, we're not going to do it, right? So we're not doing motors anymore, no batteries, all of these things were completely outsourcing So full stack doesn't mean for us that we have to own every single piece and design it, but be responsible for the value delivery and servicing towards the customer. And so this is for us, you know, we're moving up the value chain and we're very happy to buy any component that others do better than we do. And we integrate them into a robot.

 

Brian Heater (41:43)

Yeah, so right now, in terms of the different parts of the stack that you're creating, you're created them exclusively for your own robot.

 

Péter Fankhauser (41:50)

Yes. So in the beginning, actually sold ? a few hundreds of the motors that we have on the robot, which is a specialized design. have a patent on it. But quite soon we realized either you're a robotics company or a motor company, you cannot be both. So we decided let's go full steam towards robotics.

 

Brian Heater (42:09)

Obviously, again, lot of people get into this world. They, you know, see a show like Star Wars ? fall in love with robots. It doesn't surprise me that a lot of people also had the ambition to be astronauts early on in their careers. ? I believe that you were one of those people. And it seems to me like maybe to a certain extent that being on a. ? a wind rig or oil rig out in the ocean is maybe the closest you can get to being on the ISS. Is that fair to say?

 

Péter Fankhauser (42:44)

Yes, that's fair to say. I still have ambitions to fly helicopters, maybe learn how to fly a plane. There's more things I can do if I have the time. But yes, that was a very close experience. And I think the experience is some adventure, but also depending on technology to survive. When you're out there, you know, right, this is all artificial stuff. You depend on the power, on the water, the food and just the infrastructure out there. You couldn't survive the ocean by yourself. So that's probably the closest you can get to when you're in space, which is a very similar feeling.

 

Brian Heater (43:17)

Is there a part of you like in the back of your head that realizes that, we're doing these really difficult environments. Maybe at some point down the line, like we can start sending these things outside of the atmosphere.

 

Péter Fankhauser (43:32)

I'd love that if the market is big enough, we'll certainly go after. know, space exploration and all of that, once it becomes, you know, ? feasible and there's actually a market, we're very happy to do that. We had actually at ETA some of our robots that did some cool experiments of space hoppers that instead of wheels, they jumped on the planet because it's more efficient instead of digging yourself through the sand. For the moment, we're focused on solving their real world problems here on Earth.

 

Brian Heater (44:00)

Peter, thank you so much for taking the time.

 

Péter Fankhauser (44:03)

My great pleasure, thanks a Brian.

 

Brian Heater (44:07)

Thank you so much to Peter for the great conversation. Thanks to you as always for tuning in. Don't forget to like, subscribe, rate and review. Please subscribe to the automated newsletter over at automated.fm. We have a packed schedule coming up on the automated podcast. We've got interviews with the CEO of Fonda Robotics, president of Matic Robotics, current CEO of iRobot, PlusOne, ServoBotics. A whole load of folks coming up. It's going to be great. Very excited to bring them all to you. So stick around because we are going to be back just about this time next week with another episode of Automated.