Latest / Autonomy Insiders / The Robotaxi Industry's Real Bottleneck Is Not AI. It's the Curb

Transcript

The Robotaxi Industry's Real Bottleneck Is Not AI. It's the Curb · Autonomy Insiders

• Daniel Abreu Marques: Welcome to Autonomy Insiders, the show where global industry leaders in autonomous driving unpack their real-world insights. Bern Grush, Pudocity: pickup and drop off it's not an engineering problem. It's a spatial management problem, and robo taxis are competing for a fixed resource at the curb. each additional robo taxi is competing with every other vehicle, for that same curb space. And that competition is unstructured. That's a really critical problem. so basically, safety will improve in a somewhat straight line, but curb conflicts and congestion issues that come with those curb conflicts that will compound with scale. So basically, as the industry scales, it will get safer. And it will get more congested at the same time. So in the end, as things improve for the technology, it's really congestion that's going to inhibit scale. Daniel Abreu Marques: Welcome to Autonomy Insiders. I'm your host Daniel and everyone in this industry debates sensors, foundation models, or remote assistance ratios, and almost nobody debates the last ten meters. Where does a robo taxi actually stop? Who decides? And what happens when thousands of them want the same curb at 8 AM? So my guest today argues that the robotaxi industry's Achilles heal is not the AI, it's the pickup and drop-off and Byrne Crush is the CEO of Puder City, a Toronto-based startup building what he calls Robo Taxi Orchestration. He's also executive director of the Urban Robotics Foundation, lead drafter of the ISO standards governing how automated vehicles and robots use curbs and sidewalks, and a member of the World Economic Forum Global Future Council on Autonomous Systems. So Bern, I'm so happy to have you on the show today. Bern Grush, Pudocity: thank you. I'm thrilled for this conversation as well. and thank you very much for inviting me to this. Daniel Abreu Marques: So maybe let's start directly with your call claim. So you've said that the robot taxi industry's killer seal is not the eye or the sensors, but the passenger experience when stopping. So that's a proactive statement in an industry obsessed with autonomy metrics. So make the case for us. Bern Grush, Pudocity: Well, the industry and the and the media about the industry are all currently focused on safety, and that's critical. I'm not taking anything away from safety issues. But safety is an engineering problem, and usually engineering problems get solved eventually. So we may be impatient with it, but I think it will be solved. But really The current trajectory of robotaxi mishaps is that they on a per vehicle basis they keep shrinking. They get more and more in the news because the fleets are growing, but the on a per vehicle basis they shrink. Even if they'll never get to zero. The issue is that pickup and drop off it's not an engineering problem. It's a spatial management problem, and robo taxis are competing for a fixed resource at the curb. This is really about logistics and operations. It's not about mechatronic safety. And each additional robo taxi is competing with every other vehicle, I mean not just other robo taxis for that same curb space. And that competition is unstructured. That's a really critical problem. so basically, safety will improve in a somewhat straight line, but curb conflicts and and congestion issues that come with those curb conflicts that that will compound with scale. So basically, as the industry scales, it will get safer. And it will get more congested at the same time. So in the end, as things improve for the technology, it's really congestion that's going to inhibit scale. And that's what I mean by the Achilles heel, because the industry wants to scale, just from the industry perspective itself, and it can't do that unless it gets this problem of the what we call Puto orchestration or pick up and drop off orchestration. Daniel Abreu Marques: And the skeptic would maybe say Waymo does hundreds of thousands of rides per week and people get picked up just fine. So why isn't this a solve problem or at least a problem to solve itself with scale? Bern Grush, Pudocity: Well, it's not a solved problem, not only just for the for the sort of logistics answer that I just gave, but it's not a solved problem because in the states, the US states regulations prohibit cities from interfering with the operations, the pilots and the trials and the early instances of robotaxi fleets. And so the cities can't manage their curb as they normally would. And so that's you know, that's in the way. and what I want you to consider is that if you have just say a thousand robotaxis, which is what we're at in San Francisco and where we're going to be going in many, many cities, a thousand robotaxis generates about twenty-five million pick up and drop off and staging, because sometimes they have to wait between rides, about twenty-five million events. Pick up and drop of events per year. That's a lot of opportunities for blocking, first respondents, for blocking bicycle lanes, all those sorts of things that you see in the news. Even if only a small percent, even if only five percent of those cause a problem, congestion problems and so on, that's well over a million annual opportunities for problems That will appear in the news or that will appear on social media and so on. So it's a lot of problems. fundamentally as you double the fleet, you more than double those conflict opportunities. And it gets worse with scale. It doesn't solve itself with scale. Daniel Abreu Marques: And human drivers also make opportunistic stopping decisions. So they double park, they wave, they make eye contact, they they break small rules gracefully. So what exactly is lost when you remove the human and why can't AI simply replicate that improvisation? Bern Grush, Pudocity: Well, there's a huge difference between how we operate socially with, you know, all of our eye contacts and hand waving and that sort of thing there's an expectation with half of us for backing down when we have conflicts because it's it's another human, there's a social conflict, but we don't want to back down from a machine. We feel that we're first and the machines are so far programmed to be, you know, subservient or to back off or to be extra cautious. And we know that. We human drivers know that. So it's not only the machine to machine competition, it's the human to machine competition as well. So we're not gonna solve that with just AI. I'm not saying we can't, make I AI better, but we're not going to solve it with AI because of the human machine issue, which is not going to be AI'd away. That's a permanent social factor. Daniel Abreu Marques: is Pudo underrated because it's genuinely hard or because it's unclamorous? Because no one raises a series B on curb management? So how much of this blind spot is actually also maybe a capital allocation? Bern Grush, Pudocity: Well, it's both hard and unglamorous. I mean basically you're dealing with multiple operators in public spaces, you're dealing with multiple city departments, you have to deal with parking authority, with the public works, with the emergency services, with the events management, and so it's unglamorous, Capital is really enthusiastic about autonomy. I think that's pretty clear throughout. And it's somewhat blind to operations because we haven't spoken about operations. We expected operations to be quite easy, but these operational snarls, as we mentioned earlier, is really what's suppressing scale. When capital comes to understand that, I think that in the investment will come. But right now, very few investors I think are understanding this, but I think that's starting to change. Hope. I want to point out that at one point in time, go back to 100 years ago or 90 years ago, parking meters were completely unglamorous as well, but eventually it became a core municipal infrastructure. So you know, Pluto orchestration is going to end up you know following that same arc. It just takes a while. so fundamentally Pluto orchestration is a governance problem. And who's going to decide whether and where that happens and what the usage fees should be, what the schedule should be, how the locations should be marked, all those sorts of things. These are all municipal infrastructure issues and they're not really AI issues. they're not the kinds of issues that are interesting for investors, for example. these are all really municipal infrastructure issues and you know, for example, when I drive my car, I don't decide where the lines are painted or where the stop sign is placed. Puto infrastructure is traffic infrastructure. So I'm talking about the same level of technology as our our current traffic systems and traffic marking and traffic lights and so forth. Very unglamorous for sure. Daniel Abreu Marques: Yeah, so I think we we have established the claims. So now let's get concrete and I want to understand what actually goes wrong at a curb mechanically when fleet scales. So walk me through the failure modes. When a robotaxi poodle goes wrong today, let's say in Phoenix or in San Francisco, what does that concretely look like for the passengers, for the traffic, for the city? Bern Grush, Pudocity: Well, for the passenger you will have things such as the vehicle stops far away or in an unexpected place or maybe on the wrong side of the street. I've had people tell me that they ended up chasing after a robotaxi that was creeping ahead trying to find a place to pull over for over two blocks. I've heard that from more than one person just because there's no place. I mean, that's just an able-bodied passenger. I want you to consider a passenger who's blind or in a wheelchair, that sort of thing, or so there becomes not just that it's hard to find or it's inconvenient. It's it becomes a safety hazard in that case. From a traffic perspective, you have vehicles double parked in lanes, they're blocking the bike lanes, blocking the bus lanes, and so forth. And a lot of the comments that we hear about blocking first responsors, some portion of that, and I it's hard to tell, is not because the machine has had a failure or the robot's had a failure. It's because the robot's waiting for a passenger and it has no other place to wait. And it doesn't have enough smarts to know to get out of the way if there was a bus or a siren or traffic and so forth. It's also the case from the city perspective. They're getting unnecessary obstruction calls, like a 911 call here in the US, but you'll get you know traffic complaints and so forth. there's also from a city perspective a loss of parking revenue, which is another complex problem as well. It's not visible to all these social media talking about these things, but it is already a measured issue. So that's pretty important. Daniel Abreu Marques: And today's incidents are anecdotes or like blocked bike lanes, confused clusters of vehicles. at what fleet density does this flip from an annoyance to let's say a systemic problem? So do you have a number or threshold in mind? Bern Grush, Pudocity: Well, it's not a threshold of like a fleet count. It's not like how many robots it's really it's about the density within a particular time and place. So this can become a problem with five vehicles if all five of them are coming to the same place during a peak hour. So it's really about that relationship between density and the amount of the space. So basically the flip between an annoyance to a actual systems problem, it happens when the the arrival rate, so if you think it's about this is about queuing theory. So the arrival rate of the robotax exceeds the space at the curb. So that's a little bit of a technical answer, but you have to look at each local circumstance and that's why you need a system to manage that. And when I spoke earlier about this being unstructured, we need to structure that so that we don't run into these local problems. It's not congestion everywhere, it's just hundreds and hundreds of small local congestions which then spreads everywhere. Daniel Abreu Marques: If we look at the curb, robo taxis compete for the curb with delivery vans, with buses, bikes, maybe also outdoor dining and parked cars. So cities already fail at managing this without autonomy. So are robotaxis then creating a new problem in your opinion, or is it just exposing an old? Bern Grush, Pudocity: Well again it's both. I mean they're exposing an old one because the curb has generally, not everywhere, has generally been underpriced and undermanaged for up something of a century. here in the US, a professor named Donald Schupe, has developed a lot of theories about how this can and should be monetized. And his theories need to be reviewed and reused for this stopping problem. It's not the same as because it's not the same as storing a vehicle for two hours. It's only stopping for a few minutes, but it's a similar problem. So the the other issue is that Robotex is actually at a new dimension because machines need a sort of a machine readable allocation. This issue of how us humans work it out, I mean not only do we collaborate with you know with the eye and hands indications as we work human to human or driver to driver. we also circle and circle and circle until we finally find a place and we cannot have robotaxis circling and circling to find a place to pull over. And they don't. They just stop in the lane. So instead of circling they now block the lane. so that makes it a different problem because robotaxis double parking quite frequently whereas humans double park They do double park but they they double park less often. So that is an important difference. Daniel Abreu Marques: And you also touched already a bit on the accessibility dimension that most people skip. So for example, wheelchair users you mentioned, or elderly persons, someone with luggage in the rain maybe as well. So how does unmanaged Pluto fail these passengers specifically and why should operators also care commercially, not just morally? Bern Grush, Pudocity: Well, it's unreasonable to expect someone who's blind or in a wheelchair to enter a robo taxi that's waiting like double parked that's waiting in the travel lane. That's just I mean, no one would expect that. so from an accessibility perspective, not having proper pickup and drop off places excludes those kinds of users. Consider an elderly rider that might be relatively able bodied, except that they're older. you can't expect that person to chase after a robo taxi for a block or two because it can't find a place to stop, for example. But there's also an important issue, and this is from the robo taxi perspective. These failures, they generate dissatisfied riders. sometimes that might it may cause, I don't know if this has happened yet, but it may cause a refund, or that will cause lost customers who don't want to deal with some of these issues I've just described. and also the accessibility issues and so forth. So all of this is going to have a negative impact commercially as well. Daniel Abreu Marques: And we we also talked about that unlike the human drivers, this autonomous vehicles need a reliable or reserved space. but reserving curb space is also probably politically explosive, let's say. So you're effectively taking a public space and allocating it maybe to to private fleets. So in your opinion, isn't that a fight that the industry loses? Bern Grush, Pudocity: Well, So first of all, importantly, reservation is not ownership. I mean reservation it's timed, it's it should be paid. I mean, the system that we work with permits monetization, but it's really about shared access. It's the same as an airport landing slot or a parking meter session. We think this is politically easy to push. through because when the city sets rules and the prices and any operator can book the same space. That might mean not just, you know, not just robotaxis. It may be that these spaces need to be shared with other loading docks, for example. That's a consideration. So this is like metering public space. It's really not, you know, privatizing that space. You know, another way to think about this is consider the way we manage space for bus stops. that's monetized because we pay for the bus, all of these instances, airport bus stops and what we're talking about for Robotype, it's all it's all governing public space and that's critical. You're not again back to the scale issue, you're not gonna scale without that governance. Daniel Abreu Marques: And multiple operators will also run fleets in the same city, probably. So Waymo doesn't talk to Zeux, Zux doesn't talk to Tesla. So what happens, for example, at a stadium or an airport when three competing fleets search to the same curb with no coordination layer? Bern Grush, Pudocity: that's a perfect example. the vehicles coming from each fleet or each fleet operator, they're gonna wanna optimize their own results. They wanna get as close as possible or get to their fair or they're rider as quickly as possible and so forth. So they all, because they're all looking after their own efficiency, that is one of the sources of creating congestion. And of course what you're going to get is you know gridlock from that obviously. And the potential in a large circumstances, if you have as an example, an event at a stadium let's just say there's 20,000 people at that event And let's just say 200 of them order a robotic taxi, and they need to come in a sort of a fifteen or twenty minute period. That's unworkable. I think you can, you know, certainly easy to see that how that would create problem just for the users to find their own taxi, for example. But I'm just, you know, more concerned about the congestion issues at this point. And it's you know, it's the operators are not going to talk to each other. I think we all know that. I mean, that's implicit in your question. So, you know. The way it has to be is that each one needs to speak to the curb in the same sense the same analogy as each airline, in a sense communicates with the runway and not with the other airline. So it's again, the analogy still holds. Daniel Abreu Marques: And you didn't just write papers about this, you also started a company. So let's talk about Pluto City and also what orchestration actually means. so give me the mechanics. So a passenger, books ride, vehicle approaches and where does Pluto City sit in that flow and also what does it actually decide? Bern Grush, Pudocity: Well yeah, thanks for that question. So Puto City, the company has a product. The product is called a Curb Reserver. and that is a software orchestration as a service. It sits in the cloud. And what it does it sits in between the fleet's dispatch system. It's not between a vehicle and a passenger, it's between the dispatch system and the city's rules. So the city, sets up its rules the rules are schedules and places and the the vehicle operator requests a space for either a pickup or drop off for example or staging and the system provides that space from its map as nearly as possible to what the vehicle operator has asked for. So basically the orchestration system is saying here's a start time and an end time and a space for you to do whatever it is that you need to do with that space. We don't ask what you're doing. We don't care if you're doing pickup or drop off, or you're dropping off a package or picking up a package. What we care about is the dimensions of your vehicle and the space that you need and the time you need it for. And that's basically all that's happening. It's kind of it's just a reservation system, like a dating system between a between a car and a space. Daniel Abreu Marques: And you also already call it orchestration as a service and that also implies that it's a neutral layer between operators and cities as as you just talked about. So why does it need to be an independent company rather than a feature of an Waymo, an Uber or someone else in house? Bern Grush, Pudocity: Well, first of all, it has to operate for all at least all autonomous vehicles. In other words, we can't have a separate system for FedEx and a separate system for Waymo, not only because they don't talk to each other, but because we want to make sure that when they're sharing that space. When I say sharing that space, they're sharing it consecutively, obviously. we wanna make sure that neither arrives early and stands and awaits in the in the through lane. So we basically have to have a system that's independent of any of those operators. Again, back to the airlines. The airlines don't run air traffic control. that's run by government agencies that operate airports. the same thing here. It's a government agency in this case a municipality would operate that because it's really the city that's orchestrating it. It's not it's not the fleets that are orchestrating it. The fleets orchestrate the fleet, but they don't orchestrate where to stop. That's that's a very important distinction. we're not orchestrating the route. We're not telling the vehicle where to go. We're saying here's the space for you at this time. How you get there and what you're doing there is not what we're orchestrating. Very important distinction. Daniel Abreu Marques: And I think another hard question that every startup faces is also who pays you? So is it then the operator, the city, the passengers in your example? Bern Grush, Pudocity: we're a municipal traffic management system. So the software is licensed to the city and therefore the data that the software uses belongs to the city. I wanna point it out, that's pretty important as well. So the license fee is paid by the property operator. So the property operator's case of city, who owns or manages the curb. Is the municipality. So that's who is paying for that system. Now, of course, you would expect and hope that the system would be monetized because it needs money to support itself. And it's also the case that monetization has a very important value for demand management. And so that monetization would come from the Fleet operator, but of course the fleet operator is getting it from the passenger. So the the short answer to their question is the passengers are paying for it. But we don't require monetization, only the city does that. So we do the orchestration, and if the city monetizes that, then eventually the passenger pays for it. Daniel Abreu Marques: And you already touched also a bit on that the orchestration can then also extend beyond road taxis, so it could be shuttles, transit, delivery, even other forms of autonomous vehicles. is that a vision or a roadmap and also maybe where do you actually start and what do you deliberately do not in year one, for example? Bern Grush, Pudocity: Well, we have started in year one with pickup and drop-off, and it doesn't matter whether it is for goods or passengers. It turns out that already the system can even handle a fleet of one. I'm gonna point this out because what's not obvious is that at some point in three or five or nine years, when some families own a personally level four vehicle. That vehicle also needs to be if it's going to go downtown and it's gonna let its passenger off and it's gonna go somewhere to park, it also needs to be orchestrated. So we've been using the word robo taxi, but it's really an automated vehicle. and robotaxi sounds like a service for passengers, but it's an automated vehicle for goods or passengers, whether it's a service vehicle or a privately owned vehicle, it handles all of those. What's being added later? that's not done already is for charging, waiting, staging, and cleaning. It's the same problem. this is really critical. Every robotic vehicle has tasks. And a task is drop this person there, pick that person there, drop this package here, go get a charge, go get clean. So when you think about robots, instead of as a passenger there, think about a robot has to know where to go where is it that it should stop to do its task. So therefore the task of getting its battery charged or the task of dropping you off is the same thing from a robot's perspective. It's go to some place and stop. And people don't see that, but the system do that's why the system does all those things. It's just that we haven't added charging and we are in the process of adding charging and and waiting, but we Yeah, we just done pick up and drop off. So goods are goods are passengers, right? Daniel Abreu Marques: Super interesting and cities are also notoriously slow customers and A V operators are notoriously closed ecosystems. So which sales motion is harder and what has surprised you in early conversations? Bern Grush, Pudocity: Well, yes, it will be a long process. that's certain. And right now, while they're getting started, the operators they don't want to be constrained. I I wouldn't if I were one of these companies. I'd say like we're busy enough with the problems that we have. don't add to my list of problems. But it's become evident and it's it's even become evident in the news now. it's been evident in the academic literature for well over five years, but it's become evident more and more that these need to have a place to pull over. So I wanna tell you what has surprised me first or most is that the states in the US this is the US issue and I hope it's different in the EU but in the US the states have ensured that the cities Have no say. It's the same thing in my country in Canada. Ontario, the province, says what these fleets can or cannot do, and will not let the city have any say in how they're managed or where they can go and so forth. That's been the biggest surprise, and I think that is because the state legislatures didn't even consider this as an issue. They're focused. appropriately on safety because the promise that's made to to government is that these vehicles will reduce annual fatality rates. Probably they will. It seems that's the way it's going. And that's a very good thing to be worried about. And so they're have not been concerned about the issue of curb management. It's just not been on the radar. That's an oversight and that's led us to the point where we are now and that's why it's so very, very difficult. I hope, because EU seems to be regulatory forward, I hope that the EU won't have that same short sighted view that has happened in the US. And that's been that's been the biggest surprise. Daniel Abreu Marques: And what do you think is Pluto City's mode? So if this works, what stops a Google or a curb data incumbent like a parking payments company from copying it? Bern Grush, Pudocity: Well, there's patents, and there's standards compliance. Of course, standards compliance is not a steep moat, but the patents are important. part of this is that overall as we ramp up the the moat grows. I mean part of it is the the the manner in which we protect privacy because of the Again, that comes from what the standard says you need to use. if you do more than that, which some companies may be inclined to do, that would be a problem. we won't do that because we are municipal thinking and not commerce thinking. So we're concerned about traffic management more than anything else. It's also that some of our standards work have a machine learning Capability that will be critical. And what that does is that makes our system more and more valuable as we do more and more cities. And that lead creates a moat. that's probably as important a moat as the as the patents themselves. It's also the network effect as you add more and more cities. whoever's out there adding more and more cities, It's hard not to want to be part of that. sort of data lake that's making it possible for your city to be optimized sooner. so it's an optimization moat, a patent moat and and the network effect mode, those are the biggest Daniel Abreu Marques: and Poodle City is built on standards, as you mentioned, that you personally helped draft. So that's maybe also an unusual position. So writing rules and also building the product. So maybe let's go there as well. So you are the global project lead drafting the ISO forty-four forty eight for public area mobile robots, and Pudu City builds on the Poedu communication standard ISO twenty-five six hundred fourteen. for listeners who glaze over at ISO numbers. So what do these standards actually govern maybe in one or two sentence each and why should an operator also care about Bern Grush, Pudocity: Well, in a sense it's all the same. So but there is a distinction as which side of the curb we're on. So if you look at the curb on the sidewalk side, it started out to be what are called public area mobile robots. So those would be delivery robots, cleaning robots, you know, sweeping robots, security robots that operate on a sidewalk. That is now extended into public facilities such as the airport and so forth, which is odd because that's not the roadway, that's in a in a private building. But the same kinds of problems because they're moving amongst humans. So what we're concerned about is the behavior of the robots. We're not concerned about the mechatronics or the batteries or the size of the wheels. We're concerned about their behavior and what we call the legibility. Do the humans that are operating nearby understand what these devices are doing. So that's the 4448 side. The 25614 says is this is the road side of the curb that's where our robo taxis are that we've been talking about until now. That's a much smaller standard than the 4448. it's a much narrower issue. It's really concerned about what is the message language and the data flow operations for the vehicle to negotiate with the curb. you know, requesting, reserving, arriving, dwelling and so forth. why I care about all of that is if we don't get these things right, then none of these technologies can scale. they'll be socially disruptive and I'm gonna use a strange term. I'm concerned about urban livability and if done right it could improve urban living. If they're not, it will make it worse. So I'm you know certainly concerned about that. That's why it's really all about the the call it human factors, the relationship between us humans and the machines. So the interface between human and machines is an area of study called human factors engineering, and that is very important for understanding how fleets of all of these robots can operate in cities. That's where all my work comes from. or all my all my interest comes from that perspective. And I think you can see that in the in the standards Daniel Abreu Marques: And standard work is also probably slow, consensus driven, unpaid, and you spent years on it. So what convinced you that standards are the leverage point rather than say lobbying or just shipping a product? Bern Grush, Pudocity: Well, I think that without standards you don't have an industry. Well, you have an industry, but you without standards you can't scale. You you can't interoperate. And if you can't have those standards, then what will you build your regulations on? So in a sense, and kind of my mantra is no standards means no industry and no regulations means no deployment. We don't act as a regulator, we act as a standards drafter. But it's it's again without the standards, on what basis to which you regulate. And that's by the way, we have 20-some different standards from 20-some different states in the US for what they consider delivery robots. It's one tiny segment of all this. And there are differences from state to state. They're not huge differences, but they're shortcomings and differences between them. And that's all because there's no standards in place at the time that they wrote those standards. So that's why I think they're important. Daniel Abreu Marques: And some might play devil's advocate and say standards can also freeze immature technology too early. So how do you know that the Pluto problem is understood well enough to standardize it now? Bern Grush, Pudocity: Well, we're only standardizing the interface. we have zero impact on how a robotaxi is designed, how it moves, how it navigates. None of that at all. Because that is immature and that will change a lot. So you're completely right on that. We're only determining the the reservation message. And how those messages are passed back and forth. So that's fully understood. we understand how to describe a vehicle, its size, is width, its weight, is height, and that sort of thing. We understand how to describe a space, same sets of issues. We know how a map works. So there's nothing that we're doing that's not already understood, and there's nothing that we're doing that changes anything about how robotaxis would be designed or managed as a fleet. So we're not managing a fleet. We're managing the reservations for the pickup and drop off. It's a very, very tiny slice of the whole system. And that's the that's a critical slice to make it work in cities. And not only is it not too soon, it's already late because you see the kinds of problems that we have in any city where we have more than fifty or seventy robo taxis, these problems that we're looking to solve already. Okay, so we're late, not early in that regard. Daniel Abreu Marques: And you also sit on the World Economic Forum Global Future Council on Autonomous Systems. So can you tell a bit about what is that body actually debating right now about autonomy in cities and also where do you find yourself maybe in in the minority? Bern Grush, Pudocity: So that's a small group. There's there's multiple councils. I don't know the total number of councils. It it would be a few tens, I think. but our council on autonomous systems has about fifteen or sixteen people in it. and I'm the only one that's concerned about what Call public area mobile robots. So the discussion is really wide-ranging. So we talk about the role of AI, we talk about LLMs, we talk about simulation, we talk about physical AI. So we all have different expertise. Here's what's common amongst all of us. We're all concerned about the widening regulatory gap, that we all see that as a problem. I personally see the technology moving faster and the regulatory landscape moving slower. I think all of us on that group agree, and I'm assuming that our listeners would agree with that as well. And that gap is widening in that sense. each of us were selected for you know a different purpose, so that I would naturally have a minority position because we all do. but mine is that I focus specifically on the interface between robotic systems and what we what we call as non-involved pedestrians. These are pedestrians that are not trained and are not protected and are not knowledgeable, they're not working there. They're just happen to be in the same space. You're out walking your dog and there's a delivery robot that comes by. And you had nothing to do with that robot. So I'm concerned about that interface between you and that robot. This your safety and your understanding of what's going on. And all it's not just pedestrians, but remember these, delivery robot crosses the street. Now, what about cars? are they visible to cars, for example? one of the issues that I work on is how does a robot cross an intersection? Or how's the robot cross the road? Sounds like a chicken joke, but it's not. it's not evident exactly how that procedure should should roll out in all of its exceptions. It's about the exceptions, of course. So that puts me in a absolute minority. I'm the only person asking how's the robot cross the road? Daniel Abreu Marques: But these are important questions. Bern Grush, Pudocity: Yeah, I think so. Daniel Abreu Marques: And most of the deployment stories today are more American or Chinese, this show here is European focused as well. European cities are denser, more transit oriented, probably also more regulation friendly than US cities. So is the poodle problem in your opinion harder in let's say Barcelona or Munich than in Phoenix or is Europe actually better positioned to get this right from day one? Bern Grush, Pudocity: Because, in my view, the EU is regulatory forward or thinks about regulations before innovation or before deployment, I think you have a much better chance. It's also the case that the problem is actually worse in the EU because of density. The average urban density in the EU is much higher, about something like factor three or four, but much higher than the average urban density in the United States. Yes, we have more cars, but not so many more as people say. So you're going to have a more difficult if if you don't manage pickup and drop-off, if you ignored it as the US states have forced their cities to ignore it. If you do that, then your problem will be even worse. read all the problems in San Francisco. It's gonna be worse in Berlin, if you treat them the same way. So I'm holding out hope that EU cities will set up, pick up and drop off orchestration systems. before they get past a small fleet? But we'll see. I hope so. I think you hope so too. Daniel Abreu Marques: And yeah, absolutely. Absolutely, And here in Europe, eighteen EU transport ministers signed joint declaration on cross border A V testing and also you know, pilots are kicking off like Bolton Pony AI in Luxembourg are launching. So if a European transport minister gave you ten minutes, what one PUDA related requirement would you tell them to write into deployment permits now before the fleets really scale? Bern Grush, Pudocity: Well at as a kind of a rough quote, whatever the city name is, our city, we reserve the right to set up reservable locations for pickup and drop off. and staging and waiting. And any time that you stop on our streets, we reserve the right to tell you where it is that you're going to stop or and to manage it. We also reserve the right to monetize it and and to schedule it and so forth. But basically you wanna leave space for the city to orchestrate when the time comes even if you don't do it from day one. But you know it doesn't cost anything to tell the fleet that you're going to do that, but if you tell them later after they're already operating. It's the same sort of thing here. if you go to San Francisco now with they have a thousand robotaxis working there and two or three different companies, and you tell the major company there, Waymo, that we're gonna orchestrate you, they're gonna say no. I mean I would if I were them. But the city of New York is already hesitating to let them in. and if New York said, We'll let you in if you agree to be orchestrated, Waymo would say yes. So you need to ask first. You need to set that out first. That's the that's the most important thing that these ministers can do. Daniel Abreu Marques: sense and I think one part that you all also already touched on earlier is that US CDs are often preempted by state law from regulating the AVs so basically who should own curb allocation decisions the city the operator the market we are pricing the neutral orchestrator so I think you you already gave me your opinion a bit earlier but but yeah just maybe a honest ranking from your side. Bern Grush, Pudocity: Well, look, the municipality should set the rules about its curve. Where where, when, and whether or not it's it's monetized. That should be up to the city. that's the first so you need to have a neutral orchestrator. I think we've made that clearer earlier in this conversation. It can't be done by one of the companies or it has to be done by the city. And the assist that we wrote is Operated by the city. In other words, it's the maps and the rules come from the city. So basically it's a city operation. We're just the software that operates what the city says. And it's also the case that you know market pricing is going to be a very, very important city instrument throughout this. And again, that market pricing is up to the city. It shouldn't you know, a state or a federal government shouldn't, I mean, they could minimize. Or they could set a they could set a floor or a ceiling, that's not unreasonable, but cities have a chance to set dynamic pricing. They should be allowed to do that. I make it sound really simple. of course there has to be some set of rules that the companies aren't harmed or that the cities aren't harmed. But in the end, the cities need to have as much control as is necessary for them to manage their traffic and their safety and their loss of parking revenue, which is, important for many cities as well. Daniel Abreu Marques: And all these curves also probably generate a lot of data, so demand patterns, twelve times, movement of people. So who should in your opinion own this poodle data? Bern Grush, Pudocity: hundred percent the city's own it. remember that the city is in our scheme of this, in our way of thinking about this, the city licenses the platform that they put their map, they put their rule, the city does, and the taxi companies or the goods movement companies or the autonomous vehicle companies use the city curb. So all of that data of when and what was used, not what it was for. you can't even know whether there was a passenger or not, or whether it was goods or not. You the that information isn't used, so it's not available from our system. But the size of the vehicle and the time it was used and the place it was used, that's available. And that's belongs to the city, all of it. Daniel Abreu Marques: let's close by looking forward and maybe also get some of your least consensus takes. So you co-wrote the book The End of Driving, and I think I also can spot it right behind you in your in your bookshelf. So looking at the next 10 years of the city streets, so what's one widely held belief in the AV industry apart? Urban deployment that you think is simply wrong? Bern Grush, Pudocity: Well, there has been, and it's it's fading slowly, but there's been a long held and fully unsubstantiated belief that autonomy at scale will reduce congestion and parking demand. the assumption was that everyone would use a robotaxi and no one would own a car. That's not true, but that's the assumption from a decade ago. We debunked that in the book very largely. So but really it matters what is the mixture of Robotic taxis, robotic personal cars, household robocars, and household non-automated vehicles. So what's that mixture like? That is going to take, in my opinion, some people think it's 10 years away. I think it's 50 years away before it's all robotic. But that doesn't matter. That's unsubstantiated as well. But that belief that congestion will be wiped out and parking will be wiped out is Completely unsubstantiated. And so we in the book spend a lot of time talking about why that won't happen. And you know, basically one of the elements in the book is Putha, one part of one chapter is about what we've been talking about here. And ungoverned, this is just going to create more congestion, not less. You can already see it. You already see it. Daniel Abreu Marques: And maybe let's close with a concrete calls to actions for three listener groups. So an A V operator, executive, a city transport official and maybe also an investor. So one thing each of them should do about Pluto, maybe in the next, I don't know, twelve months. Bern Grush, Pudocity: Well the robo fleet operators and by the way, goods if someone like FedEx or UPS are involved with robotic vehicles as well, they will at some point. They should Lobby for orchestration, not resist it. Basically, they should say give us space to operate instead of making us stop on the roadway. It's the opposite of what it appears or what we think is happening, they should be lobbying for orchestration just to optimize. they want to be safer, they want to shorten their pickup and drop off. one of the points that we make is that if any robotic vehicle can go directly to a spot, do what whatever its task is, it's pickup or drop-off, or whatever its task is, go right there. execute his task and go out, these vehicles will do in a city would average a cup at at least two additional trips per day. So that creates more business if they're orchestrated. basically the pickup and drop off will optimize and it actually increases their profit if you want to think about it that way. The transportation official should and we talked about it earlier about you know what what should be written into licensing clauses or permits and so on. They should write we reserve the right for or orchestrating pickup and drop-offs that must be in those permission documents. To the investor who's not who's been completely enamored with autonomy, which is a great thing to be enamored with. Orchestration is the only way these fleets can scale. And it's a little bit like, the old American story, the people that made the most money during the gold rush were the people that made that sold the picks and shovels. This is the same story. So everyone's investing in the automated vehicle and no one is considered no one is considered, you know, the operations layer that's actually going to make a difference. Daniel Abreu Marques: That are really, really concrete calls to actions and I think it's a good point to end. So thank you very much, Byrne, for taking the time. It was really insightful. I think we we learned a lot today. Bern Grush, Pudocity: you're very welcome. Daniel Abreu Marques: so we learned that not only one of the biggest problems for autonomous vehicles is to drive, but also how to stop and where to stop. so thank you very much for yeah shedding some light on on the whole Puru theme so happy to had you on today. Bern Grush, Pudocity: Very welcome. Had a good time. Thank you.