Latest / Autonomy Insiders / Hands-Free Charging for Robotaxi Fleets: The Hidden Layer That Makes Robotaxis Profitable

Transcript

Hands-Free Charging for Robotaxi Fleets: The Hidden Layer That Makes Robotaxis Profitable · Autonomy Insiders

• Daniel Abreu Marques: Welcome to Autonomy Insiders. I'm your host Daniel. And today we're talking about a problem that most people in the autonomous driving world don't think about enough. So what happens when the robotaxi arrives and the depot needs charge? So our guest today is Crijn Bouman He's the CEO and co-founder of Rocsys a Dutch company building autonomous robotic charging systems. So today we are going to talk about why charging infrastructure might be the real bottleneck for autonomous mobility. Today with Crijn Bouman Welcome to Autonomy Insiders, the show where global industry leaders in autonomous driving unpack their real-world insights. where Rocsys fits in the AV value chain and what Crijn thinks about the future of autonomous driving in Europe. So Crijn, so happy to have you on the show today. Crijn Bouman: Thank you very much for having me. Daniel Abreu Marques: So to start, does Rocsys build and what problem does it solve? Crijn Bouman: Well, so Rocsys is known generally for our hands-free charging solutions. ⁓ They're based robotics, but zooming a bit out, we basically focus on the problem that a vehicle has no driver to take care of the ⁓ servicing task to the vehicle running, that problem needs to solved, right? So obviously in the case of vehicles, there is no driver. But still, vehicles need to be charged, need to be ⁓ need to be cleaned on the interior, etc. And we make sure that that gets done with robotics. product is actually the Rocsys platform, which consists of robotics AI-based software and software to with fleet management systems. And it's basically a physical AI platform, sort of the operating system for AV service depots the platform makes sure ⁓ those jobs get done and the fleet keeps running, ⁓ starting with ⁓ Daniel Abreu Marques: Yeah, and I think a lot of people hear robotic charging and think, OK, now it's a convenience feature or a nice to have gadget, but you have argued that it's a must have for autonomous vehicles. So could you walk us through why and also maybe what breaks in an autonomous fleet if the charging stays manual? Crijn Bouman: Yeah. So, that case, course, autonomy stops at the depot, right? So maybe me take you back to the origin story of Rocsys and how we came across this idea. Basically, so previous company acquired by ABB and I stayed quite some time with ABB. And we were charging infrastructure. And in the last year, I had an experience which in the end led to Rocsys. So this was 2017, we had a customer and they bought quite a few chargers and they were very secretive about what they were doing with the chargers. And I had to go to San Francisco on a business trip and I decided to see if I could pay them a visit. And it turned out to be one of the early robotaxi companies. in 2017, and at that time they didn't have a permit to test on public roads. So they basically rented a gigantic warehouse and built a test track inside the warehouse. the vehicles would run themselves on test track. And at the end of the testing, the vehicles would park themselves in the corner and a person would walk over to plug in. And ⁓ that of illustrates already ⁓ the idea that, you know, the whole problem. So I started to realize, okay, first of all, autonomous vehicles are going to happen this decade. And second of all, When you take out a driver from a vehicle, the interface to the built world is broken. So you really want to scale like mass market autonomous mobility, there needs to be a company that fixes the interface and that company became Rocsys When you take out a driver from a vehicle, the interface to the built world is broken. So, know, you can think of the charging interface, but also like going into a parking garage, going into a car, which like all the interfaces in the build world are built from the assumption that there is a driver behind the wheel. So if you really want to scale like mass market autonomous mobility, need, there needs to be a company that fixes the interface and that company became Rocsys and the interface stays manual, ⁓ really scale. If the interface stays manual, this really can't scale. amount of manpower you would need in every city, in every geography is just very, very large. That's one aspect. And the other thing is that if you to build this service, it needs to be really consistent. So you want to have the same performance everywhere, in every city, every region. And... That is difficult to achieve with all kinds of different setup, labor setups, et cetera, in different regions. ⁓ it's very crucial for an AV operator to have a consistent experience there because it defines your customer experience, it defines your revenue. So you to have something that is super scalable across geographies and that is very consistent, delivers a consistent value. You can build a service on That's where automation comes in, basically. Automating the more, you know, the repetitive tasks that need to happen ⁓ need to happen in a very high degree of consistency. it's absolutely not a nice to have. It's crucial for scale. Daniel Abreu Marques: Yeah, and when people think about autonomous vehicles, they often think also just at Waymo or at RoboTaxis per se. But there are much more applications for AVs, also, for example, like in ports or logistic yards, autonomous trucks. Rocsys focused only on RoboTaxis so far, or is the opportunity actually much broader than Crijn Bouman: Yeah, much broader. We are of course very focused on robotaxi at the moment. It's a booming industry. The moment now so but the opportunity is much broader. We have been active the ports and distribution space for the last six years where we are supporting some of heavy duty fleets that are running often behind the fence in distribution chain, for example in container terminals or logistic yards in ⁓ ⁓ the retail big book store kind of domain. that market is there and it's a huge market, very interesting. But we see Robotaxi as really like the first mainstream autonomous vehicle application. But have to realize it's really the first. So Robotaxi comes first, then there is like autonomous delivery, there will be automated valet parking, there will be autonomous trucking, you know, and in the end, like personally owned autonomous vehicles. And all of these vehicles require a different kind of service infrastructure. It's very similar. probably move to a concept with more like a shared large depots, which are used for all these modalities of vehicles. So ⁓ much, much broader than a robot actually, but it's the first mainstream and it's happening today, right? Daniel Abreu Marques: Yeah, makes sense. can you give us also some sense of where Rocsys is today? So how many sites are live right now and how many robots are in the field? Crijn Bouman: Yeah, so we started to roll out our solutions approximately four years ago. Over the last year, we have become a trusted name in the ports and distribution space. In total, we have over 30 enterprise customers and there are a couple of dozen sites where our systems are installed. So often very critical operations. So one I can speak publicly about is APM terminals, part of Shipper Maersk. Marsk, company, their complete $1.2 billion ⁓ terminal in Rotterdam, the new one is going to be powered by our systems, all automated vehicles. Very We're running in multiple terminals in the US, etc. And over last ⁓ the platform itself become sort of the core of the proposition, running it live 24 and now we are using that platform enter the robotaxi market basically. ⁓ Very exciting times. Daniel Abreu Marques: Absolutely. just recently you also announced the Rocsys M1, ⁓ you described as the world's first multi-bay hands-free charging system for robotaxi. So it's like a single overhead mounted unit that serves up to 10 bays via rail system. So it's a ⁓ fundamentally architecture from like a one robot per bay. Can you walk us through the engineering and also the economic logic? So what was with the one robot per Bay approach and what made you commit to this multi-bay Crijn Bouman: Yeah, I mean, there's nothing wrong with the one robot per bay. But looking at the problem actually for the robotaxi companies, we spent a lot of time thinking like really first principles thinking like how is this market going to evolve? What is it the customers actually need? What is the form factor that makes sense? So in the end, our product is a platform and the form factor of the robotic system are basically designed to the application. So most important design consideration is that, or realization was that first of all, all a robotaxi fleets over time will become mixed fleets. if it's, know, one type of vehicle with multiple generations, or if it's like actually different vehicles in the service So if you think of like, you know, all a robotaxi fleets over time will become mixed fleets. if it's, one type of vehicle with multiple generations, or if it's like actually different vehicles in the service If you book an Uber ride, a normal ride hail app, there's the economy version. There is high-end version. There is like the big van, which can transport seven people. There will be all kinds of modalities. And in order to be competitive for a robotaxi company, you need to offer different modalities of vehicles and you have the generations of your own vehicle. So fundamentally, robotaxi fleets will become mixed fleets, which also means that the depot needs to serve a mixed vehicle There will be all kinds of modalities. And in order to be competitive for a robotaxi company, need to offer different modalities of vehicles and you have the generations of your own vehicle. which also means that the depot needs to serve a mixed vehicle fleet. And so now the proper, of course you don't want to limit ⁓ your customer in how select the vehicle. And if you look at all those vehicles, the charge port for example, can be in very different locations. So it can be in the front, it can be in the left, can be in the right hand side. And we don't want to limit our customers there. So ⁓ basically looked for a way to enable ⁓ plugging in on all sides of the vehicle. And with the overhead mounted gantry, we can access left, right, front. And if you look at all those vehicles, the charge port for example, can be in very locations. And we don't want to limit our customers there. And with the overhead mounted gantry, we can access left, right, front. we can access basically all sides of the vehicle. So that's a huge benefit for the customers. They are not limited to select any vehicle partner they want. The robot will just take care of it. You install it once. A depot is typically an investment for 10 years or more. So you install it once and then you can basically do whatever you want on the fleet side. So that is a huge value. And the second most important was space efficiency. So ⁓ robotaxi hubs as you have seen, ⁓ So that's a huge benefit for the customers. are very large, like 30, 40, 50 up to 80, 100 parking bays basically. And these properties are in the middle of a city or close to a city. And that land is very expensive. You can't find it easily. So it's really, really expensive. It's a very scarce asset to have that piece of land. So you cannot add something that makes your space use less efficient. For example, if you would have a stationary robot next to every parking bay, you need to reserve the space for that stationary robot. But also there are still operators walking around. So you need to reserve also the walkway for the operator. So if you look at that, your site gets much, much bigger. you know, in the same ⁓ of land, you can only serve, ⁓ for example, 30 % less vehicles. So it's really not attractive. So with the overhead mounted system, ⁓ zero extra space required. It's literally above the parking. bays which is there can be retrofitted. So I would say those are the two main things mixed fleet and ground space used for placement no extra space required in the depot. Daniel Abreu Marques: Yeah, that makes a lot of sense and probably also that direct economic consequences. I think I also read in your press release that you claimed that up to 75 % higher operational efficiency from the existing staff, 1.7 million in annual saving in a 50 bay depot and a 99.9 % plug-in success rate. So very specific numbers. Crijn Bouman: Yes. Daniel Abreu Marques: Could you break this down for us? Where exactly does the 1.7 million come from? Is it like pure labor reduction or are there other contributors like vehicle uptime, reduced damage, faster turnaround, and so on? Crijn Bouman: Yeah, ⁓ so what we have done in the last three years is starting to focus on robotaxis We with building like ⁓ a elaborate ⁓ modeling how depots could work. So we've built an in-house kind of modeling tool to model robotaxi operations with a lot of our own experiments, like how, ⁓ you know, what are the most efficient routines in a depot? What is crucial? How long does a certain task take in what sequence makes most sense? So we tried that. A lot of these things we tried at our office modeled it into our modeling. So one of the fundamental realizations was that for example, charging is two times walk time. So you there to plug in, but you also have to walk back to plug out. if charging sounds like a small task, but it's actually two times walk time. So it's a very significant task. So we looked at like, how can you create most throughput at the depot? Because that in the end is very important to the customer, more throughput. And then ⁓ is of defines the rhythm of the depot. It's the longest so it defines the rhythm of the depot. So therefore you have to give charging priority over other tasks. So that means if somebody is doing something else. and a new vehicle comes in, they have to stop doing what they're doing, walk over and plug in. So that creates also lot of inefficiency. So if you model that all into our system, we saw that, know, if you take out the charging, an operator can serve about 75 % more ⁓ parking bays than would do now. So that's actually where the whole efficiency equation comes. The same can serve a larger depot. that's in end translating to the savings we mentioned in our press release Daniel Abreu Marques: And maybe let's stay on the financial side. So the same day you had the M1 launch, you also announced the Series A extension. So you raised $13 million Series A led by Capricorn Partners. What is the plan for this $13 million? How will you deploy it? Crijn Bouman: Yeah, so it will be very generally used to expand the company with a specific focus also to expand further into the robotaxi market, ⁓ which happening now. ⁓ And the logistics market, are ⁓ relatively well now. So expansion across Europe and US. Daniel Abreu Marques: your media department was really busy in the last week, so you also had another announcement that made. You launched S2, so your next generation hands-free charging system for heavy duty ⁓ electric with the unit already delivered to large-scale port customers. This also, I succeeding your prior model, that you have in the field for years. seven days, you've shipped the M1 and the S2. So could you also walk us through how much ⁓ core is shared under the hood? So are these generally different products? Crijn Bouman: so of the S2 was a bit longer in the making, so this was more an availability announcement. ⁓ indeed, so a lot of shared technology. So ⁓ see product actually as the Rocsys platform, which has the robotics and automation components, ⁓ also ⁓ the software intelligence, proactive services, cetera, That's sort of the value to the customer. S2 and M1 are both components in the Rocsys platform. So they are the automation hardware components in the platform. And there's a lot a lot shared. the software stack is mostly shared, very similar. They're running around cloud system ⁓ with APIs to the fleet management. The computer stack, the robot motion control stack are same on the software level, but also on the hardware level. Critical sensors are similar or the same. As an example, ⁓ have developed a camera module that exactly built for this type of application. So from really low temperatures to blistering heat outside, rain, snow, sun, icing, dirty conditions, ports, for example, can be really ⁓ abrasive in terms the salty air and like the dust particles in the air. So built, for example, that camera module. completely based on our experience and that's in all the products. they're under the hood, they're very similar. The difference is that ⁓ the S2 is a single bay, that's where the S stands for and M1 is a multi-bay system, world's first system for robotaxi fleets basically. Daniel Abreu Marques: Yeah, and I found interesting is also like you mentioned it now several times, but also the language around this product is now ⁓ not like a charging robot. You talk about it like it's the Rocsys platform. ⁓ the robots plus the portal APIs, proactive care, and so on. So this sounds like a software defined infrastructure play more or less. So is this ⁓ a positioning shift here? And also do you see? that the robots are becoming a node in a larger fleet operation system and the long-term value is then more in the platform layer, not the hardware itself. Crijn Bouman: Yeah, so indeed the long term value capture is more in the platform layer. So if you look from the customer perspective, we are providing a service with a ton of functionality in the platform. However, hardware is still very important, right? But also see that ⁓ the landscape in robotics is maturing and there are many more interesting components available which can leverage. So for every new task at the depot, we basically look, you know, what is available in terms of components, what is part of our platform, what is the best configuration and how should we integrate this to create the value for the customer and the value for the customers in the end, the task performed. So it's more in the application layer. So we've over time evolved indeed to see more the platform layer as the value capture and ⁓ with the robotics landscape maturing. We are leveraging every, all the developments of others into the hardware side, still delivering the same value to the customer. Make sure the fleet runs Daniel Abreu Marques: Yeah, makes sense. And one thing that I'm curious about, I don't know if you can answer it or not. how much costs such a robot ⁓ with multi-bay ⁓ setup? Crijn Bouman: Yeah, can describe it in broad strokes. So we are charging a fee per bay per year. So are serving, we are serving the parking bay actually. And the fee per bay per year ⁓ depends the parameters the depot and the fleet operation. ⁓ I think what I can say. Daniel Abreu Marques: OK, so you will be still the owner of the robot. Crijn Bouman: it's, we, we, we provide the service to the customer and we commit to that. That's also in our core values. We own it. Basically we commit to delivering the value because for our customer, they are not interested in the hardware. ⁓ The customer is interested make the fleet run. That's how they make their money. you know, transporting people and goods, that's how they make their money and all the rest is, supporting that mission. So, we provide the service to make that happen. Daniel Abreu Marques: And how long are the typical contractual lengths? Crijn Bouman: We see that the surface length typically is combined with the length of ⁓ site a certain parameter in the customer operation. So they'd like to have predictable costs for a certain period of time. And yeah, in the end, we don't provide the product, we provide the service So ⁓ we make that they have the right configuration and can support their fleet. Daniel Abreu Marques: So now we have a picture of what Rocsys does and also why it matters and where you are today. And now I want to get into some of the strategic thinking that I think it really sets you apart. from most people in this space, you have written a piece comparing the robotaxi industry to aviation. So airlines became commoditized bus companies in the air, while airports and ground handlers captured the real long-term value. you've positioned Rocsys as the the menzies of the Robotaxi era. Could you walk us through that thinking and where did this analogy came from? Crijn Bouman: Yeah, I mean, the analogy with aviation, are just many similarities. Both are about transporting people, but there's also a very clear business to consumer element from the booking experience to the travel experience. But if you look at the backend, it's actually a hardcore industrial operation, know, massive and optimization efforts, huge CapEx. So it's very similar. Then if you look at the task level, there are also a lot of similarities. fleet ⁓ an airport, airplanes refueled, they are inspected and the interior is cleaned to turn around and put them on the next mission, right? It's very similar to what's required in a robotaxi depot. So those are a of similarities. then, but the real aha moment came more from the ⁓ location scarcity. an airport course, need to close to a city. you can't build 20 airports in the city. That makes no sense. So similarly, it's very similar to a robotaxi hub. So basically there can only be a certain amount of robotaxi hubs in the city. but they are crucial ⁓ to provide a service. So, our thinking was the created in these hubs will be massive. This will be super strategic and super important for the fleet owner. So. a lot of value created and that's where we want to play, where the most value is created. And that's also where we can add the most value by our customers to run those hubs in the most efficient ways, basically. So yeah, there's so many parallels that we thought this makes actually, to look at industry with that lens ⁓ and learn from Daniel Abreu Marques: Yeah, absolutely makes sense. And if the robotaxi operators are heading down the airline path, more or less ⁓ became commoditized like bus companies in the years. Do you think any of the robotaxi ⁓ providers actually understand today and who's making also the smartest moves to avoid becoming a commodity Crijn Bouman: Yeah. So I think honestly, many players understand this. would not dare to say all of them, but many, of them understand this. ⁓ There a couple of things I can't mention, which are not in the public domain, but many players understand that you have to own a larger part of the value chain than just, you know, the to consumer interface. And ⁓ I think a good which is in the public domain is Uber. You know, Uber has publicly shown that, you know, are of course a very popular B2C ride hail app, but they now announced that they are heavily investing into the backend. So they are basically running the operations, financing fleet, to like running depots, et cetera, ⁓ coordinating all those So you could see that a sign. And of course I can't think for them, but those things have been publicly announced. I know this thinking is ⁓ in the AV in-crowd is really a big realization for many players. We need to own a larger part of the value chain to position us for the long-term basically. Daniel Abreu Marques: So maybe also what I'm really curious about is what happens inside a depot. So because I think most people have no idea how maybe also messy depot operations actually are. does the design of the depot infrastructure, particularly also with automated charging, impact the fleet uptime, also scalability, and the overall unit economics of such a robotaxi service? Crijn Bouman: Yeah, so, the, so, ⁓ in modeling, we thought about like all the tasks that are required to put the fleet. And of course we speak with almost every robotaxi company in the world. So, there's of course charging, is inspecting the vehicle, there's cleaning the vehicle on the exterior and interior. There's also repairs. So a actually, so we looked a lot into, for example, rental car business. So there's actually a lot of similarities with rental car business. All the tasks that are required to give a vehicle to the next owner are very close to what is required for a robotaxi. So there's a lot of jobs to be done and you need to make that as efficient as possible. the focus for the M1 product is to, so in general, we focus on automating the more repetitive tasks or the tasks that in the end maybe contribute to health and safety risk like very ⁓ repetitive tasks with heavy high power cables, et cetera, these kinds of things. we try to support our customers with running the depot as efficiently as possible because that's, yeah, crucial for the unit economics basically, and also for the cost in the end for the customer experience. Daniel Abreu Marques: And maybe can you give us like number? So what does a manual depot actually look like in terms of staffing? you, for example, let's say you run a 500 robotaxi fleet today, how many do I need at a depot to keep those vehicles charged and running? Crijn Bouman: Yeah, ⁓ quite a few actually. ⁓ so I walk you through our of thought there. So basically, when look at taxis or, or ride hail, you see, you know, how many miles they do per day. You look at the vehicle battery capacity, which is there. And ⁓ what that would mean if you run that 24 seven, basically. So. First conclusion, for example, is that charging needs to happen three to four times per day. It's just an analogy from what is happening today in taxi services or ride-hail services charging is very frequent, it's three to four times per day. And then the vehicles also need to be ⁓ cleaned on the inside, and now and then inspected. But the most frequent task is charging. So charging sets the tempo, it defines the rhythm of the depot. And if you add up all those tasks, like all tasks plus all the walk time on the depot, in end, we for like, you mentioned 500, I think, right? So the ratios, we come up our modeling and again, the answer is always nuanced. It depends on the operation, the city, ⁓ all kinds of but the ratio is about to eight to one to 12, roughly. So... one to eight to one to 12 operators per vehicle, vehicles per operator, sorry. for 500, you would need somewhere between 40 and 60 operator head count roughly to the show 24 seven basically. It's ⁓ quite. Daniel Abreu Marques: And how does this ratio like 1 to 12 change when you implement automated charging? Crijn Bouman: Yeah, so a 75 % improvement. So an operator could serve, for example, seven bays instead of four. That is a 75 % improvement. So also quite significant 75 % improvement. Daniel Abreu Marques: And also described earlier charging is or less the highest priority task in a depot. So every other job like the cleaning, inspection, tire checks, gets interrupted when vehicle needs to be plugged in. So automated charging doesn't just time on charging itself. So it also frees up the depot crew to actually do their work without constant interruption. Can you walk us through that dynamic actually plays out on the ground? Crijn Bouman: Yeah, so that is one thing we started to understand when we were modeling these operations is you have to decide what is the priority and somebody cannot do two things at the same time, right? So if there's two tasks, what do you prioritize? then if you see that ⁓ sets the tempo for the depot, it defines the rhythm and throughput is the most important parameter. So how many vehicles can you put on the road per hour? That's the most important for customer experience. You know, how long is your wait time for a robotaxi, but also the revenue. So the thing that has the most impact there needs to prioritized. in the end, that's ⁓ charging. And therefore try to squeeze the other tasks ⁓ the charging cycle. But also means if operator is another job, And a new vehicle comes in, they need to stop doing what they're doing, walk over to the other ⁓ spot, ⁓ plug in the vehicle and walk back. So you introduce a lot of tasks switching there, which a lot of walk time. And in the end, that creates a high degree of inefficiency. ⁓ It's also a fun job for the operator. we try, when you get rid of the charging task, basically everybody gets more productive, but also more happy ⁓ at the depot, pretty sure. Daniel Abreu Marques: Yeah. you also mentioned that the depot real estate is really like an emerging constraint. So the sites near the cities with enough ⁓ capacity are scarce and expensive. So if you would be advising a robotaxi operator entering a new European city today, how would you tell them to think about their depot strategy as a competitive moat rather than just like a cost center? Crijn Bouman: Yeah, I mean, super strategic. I it comes from the scarcity equation as well. Like you cannot build a couple of depots in a city. Just, you know, if you go to San Francisco or London or whatever, that land is just not available. So we've seen in and in the U.S. there's sort of a race to the right real estate. So it's going to a defining factor in how quickly you can scale service and how... what service level you can actually achieve for your customer. So very, very strategic. ⁓ would recommend first of all, you know, partner up with players that have a huge portfolio of real estate or have a huge funnel of real estate that know what they are doing. Because ⁓ estate time, know, ⁓ from, you acquiring the site to permitting to construction. getting the power to site all these things take a lot of time. I say I would recommend to partner up with players that have that real estate or have that expertise and a big funnel, a of real estate. And on other hand, also, you know, build a pilot as soon as possible because you would need to, you need to of understand your and the kind of real estate and the layout you're looking for. But you will only figure this out once you start doing it. And that's what we've seen a lot. know, once you start doing it, you start to realize the real question. ⁓ how to integrate automation in those sites. If you have all of that upfront, your site building process can be much more efficient. So that would be my recommendation. Daniel Abreu Marques: And Rocsys also then offer some kind of consulting? So can customers come to you and say, hey, I want to ⁓ have an autonomous or automated depot. you me with configuring it? Crijn Bouman: the automation piece, yes. So basically we can help customers to think through all the assumptions, et cetera, and all the ⁓ which are required. Yes, absolutely. On power side, that would be more on the partnership with the charger or power players, basically. Daniel Abreu Marques: OK, makes sense. And maybe let's look at the 10,000 vehicle robotaxi city. And you mentioned in such fleets like three to four charges per day, that would mean like 30,000 to 40,000 plug-in events daily. if you then assume a 2 % error rate or something like this, that means like hundreds of failed charges per day. So can you also explain differences in charging failures between like automated charging and manual charging or are there significant differences? Crijn Bouman: Yeah. So, I mean, that's also something we have learned over time is the amount of actions required. So if you bring in a vehicle three times per day and you have 10,000 vehicles is 30,000 instances, but you have plug in and plug out. that's 60,000 manual actions. So it's really, really a lot. And I think you mentioned like a 2 % error rate. That's absolutely not acceptable because That would be a huge amount of misses. So we are offering 99.9, which is 0.1 % error rate. And that is absolutely the minimum to be viable, basically. So that is one thing you need to offer like a super high reliability. It's the hardest part of our job, getting the reliability right. And secondly, you need a really good fallback process for if things do not work out well. Because in the end for the customer earns money with the fleet running. So have built over the years also a very layered approach to this because in the port terminal as well, right? If it doesn't work, the whole terminal will stop in the end. Right? So it's really important. So I walk you through quickly how that sort of works. So basically we navigate on computer vision. ⁓ That's first principle. Then we use feedback to ⁓ create a higher reliability on the plugin task. So basically same as humans would insert a plug, you with your vision, you more or less aim for the hole. Then you feel if it's right. And when you feel it's right, you push it in. That's also how our system works. So we use first the computer vision and then we use physical feedback of the real world into the adaptation of the robot behavior to get to a higher success rate. Then if that fails, we have also intelligent self recovery. So Basically, the robot will try multiple times, but it will not try the exact same move, but make new recordings and try a new attempt in a different manner than the first one. So that is the third layer, ⁓ And if that doesn't work, there is a remote fix procedure. So the robotic system will call a remote operator and the remote operator will have access to the sensors and the camera, for example, and they can remotely support the robot to complete the action. And if that... even fails, then there is an onsite recovery process, which is very simple for the onsite operators to So if there's people cleaning, they can also do some very simple actions to support the robot. Yeah. Then if that doesn't work, that's the last resort is that the robot, can the connector from the robot and switch to manual operation while in parallel, a mechanic is sent to site to fix the system. So we have built like a six layer approach to create like the best possible uptime for the customer. And that took some years to get that right, but it's so important. The fleet must run, show must go on. Daniel Abreu Marques: Yeah, I can imagine and you mentioned like this 99.9 % failure rate. How does it look like for human? Are human also in that direction or are they worse than that? Crijn Bouman: Humans in the end get it right, it may take a long time. And there are also other errors like people dropping a cable on the ground and ⁓ it breaks ⁓ those kinds failures. So it's a bit of a different category of failures, but we think we can be more consistent. I must say, honestly, we haven't benchmarked that exactly, but we... There's a lot of know, people are also, it's not the nicest job to have. are outside in the rain, sun, the day, eight hours. So ⁓ as bit of a graphic illustration we did, I told you about our proprietary modeling. We do a lot of tests here in-house. We recently, did like a plugin test with people from our own team, just continuously plugging. And everybody ⁓ after a dozen times were asking like, sorry, when is this test over? And it's really, you don't realize it's sitting behind your desk, but it's really not the nicest job to have. So, also a lot of failures coming from that and understandable, ⁓ understandably. I would say. Daniel Abreu Marques: Yeah, absolutely. Another thing is that the industry currently is also buzzing around ⁓ physical AI, embodied AI, most on the type is also like centered on humanoid robots. So where does a purpose-built robotics charging arm sit on the spectrum between a humanoid robot and a smart industrial tool? And does it matter also how your investors categorize you? Crijn Bouman: So basically we are a physical AI company, we are completely agnostic to what type of ⁓ is used in application. So consistent value for us lies in the application layer. ⁓ We what kind of robotics, what kind of sensors, what kind of software is required to perform certain task at hand. And ⁓ also what application makes the most sense. Is it an arm or a rail or a mobile robot on wheels or Is it humanoid, or is it maybe a mix of these? So we are very open to any type of robotic implementation. look from the task perspective, what needs to get done for the customer and what is required from the platform and what is the best ⁓ ⁓ implementation there. Now, if look at the M1 design, we've spent quite a bit of time on this, as I explained earlier. So... For example, you know, we've been considering like mobile robots, AMRs driving around vehicles. We've been considering humanoids, we've been considering the rail implementation. for example, if you compare humanoids or AMRs, the mobile robots on So ⁓ one of key things in the application is cycle time. So how fast can you cycle? And a robot a rail can travel to the next bay in a few seconds. where if you are like a mobile robot on wheels, you have to drive around vehicles, you have to navigate across oncoming traffic, you have to navigate around charging cables lying on the ground. So it will completely destroy your cycle time. So with a rail design, basically, can get sort of superhuman abilities. You can go the next vehicle in a matter of seconds. So that's how we look at it. And we are like every new task, we look with a very open mind, like ⁓ what makes sense, ⁓ what is today. What is reliable enough is reliability is key. then we apply our application layer on top of it and offer it as a complete service basically. Daniel Abreu Marques: Yeah. And another podcast, you've been quite I would say, on the challenges of scaling deep tech in Europe versus the US. you said ⁓ that the US is a much better place to scale deep tech than Europe. ⁓ So what specifically about the ecosystem makes it in your opinion? Crijn Bouman: Yeah, to be fair, I'm actually pretty optimistic what is happening in Europe now. the for the optimism is that people in Europe now deeply start to realize that we have been too slow and lagging. So I think that realization was less there a few years ago, but it's starting to come and that is great. So I'm very optimistic about Europe. We see in Europe breaking records on a global scale, which is super exciting to see. We can get And yeah, as a European venture, we have been very active in the United States because that's where the action is. But, and that's also where, you know, ⁓ is why I'm optimistic. We to realize every day that China and at this point in time are just the better place. ⁓ As an I was in Las Vegas last week. I took a couple of robotaxi rides. There are multiple different providers now offering robot taxis in Las Vegas. And, know, when you tell that ⁓ Europe, like, yeah, you mean pilots. No, no, no. I just download the app and book it. Right. So, we to ⁓ keep realism here. And I'm we have fantastic universities, super high skilled ⁓ people here, ⁓ conditions, backbone, cetera. I'm optimistic. But the optimism comes from the realization the realism that today. We are not there and I feel that people feel that now as well and that we need to step up our game, right? So. Daniel Abreu Marques: And maybe let's stay on Europe as you mentioned, our market reality is somehow shifting. ⁓ a year European Robotaxis felt really theoretical. ⁓ now London, for example, is turning into a four-way battleground with Waymo launching with its Jaguar and Moove and targeting Q3 2026. Wayve just raised also over a billion launching with Uber in London. And then we also have the Chinese. ⁓ players like Baidu and with their Apollo Go service coming through both Uber and Lyft and that's just one city. And we have others like Munich with Momenta, we have Pony.ai, partnered with Bolt for a broader European rollout and as someone who builds the charging infrastructure these fleets will need. Did the speed of this conversion surprise you? Crijn Bouman: Yeah, of course, we are deeply into the ecosystem and we have a lot of knowledge that is not in the public market yet. So we're not really surprised, but it's nice to see. I applaud also the UK for bringing this momentum now as a first with the legislation framework, which is there. But yeah, so we've always been thinking this is a global game, running a robotaxi fleets or autonomous mobility is really a global play. you need all continents to perform. So we always had the that this will come in Europe and ⁓ Asia everywhere. The players, the amount of money going into that. ⁓ We recently of course, the 16 billion of Waymo. You know, this a global game. This amount of money destined to play a global game. So it's logical that it expands everywhere. Daniel Abreu Marques: And when you look at Europe over the next, let's say one, two years, do you see real commercial demand for the autonomous charging building up? Or do you say it's still like two, three years away from actual like fleet scale deployments that would need your robots? Crijn Bouman: Yeah, so we see it happening now in a small like trial scale because everybody wants to understand sort of the equation. But real scaling game will start in two to three years ⁓ Europe. However, you there's so many learnings everywhere. Starting early a benefit. But it's indeed two to three years away. We will see regional but yeah, two to three years away, I would say. Daniel Abreu Marques: And how is Rocsys then positioning for this European Because you also to make sure that you're not too early, but you also don't want to be too late. Crijn Bouman: Well, we are for sure not too late. We are the only company in the world that is ready actually to scale with this. have six years of deployment ⁓ ⁓ markets. So are definitely ready. ⁓ Scaling is happening in China and in the US as we speak. And ⁓ Europe, know, the benefit is there could be a bit of second mover advantage that a lot of the lessons in China and US are earlier translated to Europe. So think, yeah, are well on track. We are not too early, not too late. ⁓ We right in time actually to our advice or customers to do the right things from the day one and to be ready for scale in probably like 2027 Daniel Abreu Marques: ⁓ also ⁓ huge part of your offering is also like hardware and hardware companies also ⁓ always have like the struggles said much, but you have to jump from like the first hand building early units to mass manufacturing. ⁓ So is there a specific manufacturing scaling challenge for Rocsys or is everything already like in place to really scale? Crijn Bouman: Yeah. I mean, you were careful. ⁓ Hardware is freaking hard. really not simple. So, and at scale is really, really freaking hard. So, absolutely agree that. know that in and out. ⁓ for my previous company, have ton of experience on the manufacturing side, but especially on how to outsource manufacturing efficiently. ⁓ Also on our management team or chief operations, Albert, comes from a hardware background, a supply chain background. So we have quite a bit of expertise in the team, but especially our focus is outsourcing. So we are laser focused to be like the most capital efficient version of a physical AI company you can think of. do not invest substantially into manufacturing footprint. Everything is done with the right partner in the right geography that really understands how you do this and with the right quality control in place. And as on manufacturing side, it's also supporting your equipment in the field. So ⁓ we've built the ground up a system to coordinate ⁓ third party service ⁓ for the scaling So basically also on the surface side, are extremely capital efficient. We will only be part of the coordination, but ⁓ ⁓ a very efficient training program for third parties to scale. We like to be as much as a software company as a physical AI company can be basically. ⁓ know, the core to manufacturing challenges, understand that it's freaking hard, never underestimated. ⁓ sure that you put your people on where you add most value and leave the things that are not true added value, leave it to others. Daniel Abreu Marques: So maybe let's zoom out one more time. And I to talk about the energy and grid side of this, because I think there's also some more to So you have a large robotaxi depot with a lot of these fast charger drawings simultaneously, that creates enormous grid load. And as you automate when and how the vehicles connect, you're also controlling the megawatts that are flowing. So have you thought about Rocsys as a kind of Crijn Bouman: Okay. Daniel Abreu Marques: could that ability to precisely time those charging sessions become a real revenue stream in itself? Crijn Bouman: Yeah, so actually with our products we are a bit more focused on the automation layer, the automation flow, we typically partner with more the power ⁓ players. Yeah, we are more in the like the logistics side than we are in the power ⁓ side, physical jobs to be done at the depot. But you this is like a whole new layer of optimization on top of the power layer basically. So we will. ⁓ We want to enable our customers to basically build also that level of optimization on top of the grid layer actually. it brings like the value, the best possible value, the combination the automation of the jobs to be done and the power optimization. So, and that's going be a very exciting play when this really comes to scale, I think. Daniel Abreu Marques: So we are close to the end and just one last question. So for our listeners, what do you want to look them after in terms of Rocsys in the next coming months or weeks? Is there anything coming up that we should be curious about? Crijn Bouman: of course you should always be curious for things coming off from Rocsys side. There will be many things, of course, which I can't speak to now, but sure. you know, AV market be very exciting. In the 2016, 17, was like the over promise, I would say. And then there was a bit of a hiccup, but now it's real. It's going to ⁓ scale and it's going to in and US. also in Europe, I'm very optimistic ⁓ the market. ⁓ may be some differences per country, but the next couple of years it's going to happen. And we come with products that support that rollout in the regions where it makes sense. So I'm very for the future watch the news. ⁓ We'll probably have coming up in the coming year. Daniel Abreu Marques: Perfect. Crijn, I think we've covered a lot of Crijn today. So thank you for your time and thank you for coming on Autonomy Insiders and for this really insightful conversation about Rocsys. Crijn Bouman: Okay. Thank you, Daniel. Thanks so much. Thanks for the great in-depth questions. Daniel Abreu Marques: See you soon. Crijn Bouman: Okay, see you soon. Bye bye.