The automotive industry is closed and proprietary.
But Shinpei Kato, founder and CTO of Tier IV, thinks they are going to be forced to change. Teir IV has brought together a global community of programmers and corporate partners to create the Autoware project.
Tier IV’s goal to develop a completely open-source software platform to drive autonomous vehicles is ambitious, and they have already completed some of the most advanced road-tests of driverless cars in Japan.
Today we explore the business bottlenecks in rolling out autonomous vehicles, why open-source makes the automotive industry nervous, and why the first successful driverless car won’t be what you think it will.
It’s a great conversation, and I think you’ll enjoy it.
- Introducing Autoware and Tier IV
- What keeps the auto industry from adopting open source
- The only way a college professor can actually run a startup
- The challenges in building an industrial open source community
- How to road test driverless cars in Japan
- Japan’s first fully-autonomous taxi service
- When we will see driverless taxies as part of our everyday life
- The bottleneck that keeps robot-taxis from going mainstream
- Which autonomous vehicles we are going to see first.
- Tier IV’s business model
- How open-source might be Japan’s secret weapon in global AI
Links from the Founder
- Everything you ever wanted to know about Tier IV
- Check out Shinpei’s personal home page
- Learn about the Autoware Foundation
- The Tier IV safety report
- Some other media coverage of Tier IV
Welcome to Disrupting Japan, straight talk from Japan’s most successful entrepreneurs.
I’m Tim Romero and thanks for joining me.
Open-source software has completely changed how we think about operating systems, networking, and databases. The whole Internet basically runs on open-source software, but can a 100% open-source software power an autonomous car?
Well, one Japanese founder not only thinks it can, but he’s betting his company on it, and that startup has already conducted some of the most advanced road tests in Japan.
Today, we sit down with Shinpei Kato, founder of autonomous driving startup Tier IV, and Shinpei is also the chairman of the Autoware Foundation, Autoware, being the open-source project to develop software for fully autonomous vehicles.
With so much driverless car news coming out of the US, you might not know about what’s happening in Japan, but it’s pretty amazing. We talk about what’s involved in road testing driverless cars in Japan some frightening things people are doing to their cars, the challenges of building an open-source platform in an industry that has historically been fiercely secretive and proprietary and why Japan’s first driverless cars are not going to look anything like what you think they will.
But you know, Shinpei tells that story much better than I can, so let’s get right to the interview.
So I’m sitting here with Shinpei Kato of Tier IV who is developing an autonomous driving software, so thanks so much for sitting down with us.
Shinpei: Thank you very much for inviting me to this fantastic show.
Tim: Oh, it’s our pleasure. Listen, before we get into the details, can you explain the relationship between Tier IV and Autoware, because the two different entities are really closely connected and like, together, they form Tier IV’s business strategy.
Shinpei: So I used to be at Nagoya University and I had led a project of autonomous driving where we started developing software for autonomous vehicles, so I had a lot of attention from industries that made me decide I should do startup rather than the university research. Tier IV was founded to facilitate RND of this open-source software called Autoware. Two years later, it became much more which made me decide founding Autoware community rather than using Tier IV as just one Japanese startup. The Autoware community can be composed of worldwide industry and academia, Tier IV handed over Autoware to the community.
Tim: So Tier IV is started before Autoware?
Shinpei: When you say Autoware, it means one is Autoware as a software project which started before Tier IV, but when you say Autoware, it can also mean Autoware Foundation as a community. This Autoware Foundation was founded after Tier IV was founded.
Tim: So my understanding is that Autoware is a complete open-source platform for autonomous vehicles, but how do I put it? That’s an incredibly complex set of technologies, so is Autoware really, is it a complete platform now or how much does it actually do?
Shinpei: When you consider the building blocks of autonomous vehicles, it’s huge, yes, as you mentioned, it’s complex. Autoware stands for open-source software. I would say it’s a small piece but still, it’s just a piece of building blocks, so what I wanted to do was to publish Autoware as a basic platform on top of which other commercial building blocks can build. You need hardware, you need cloud, there are many, many other pieces in the building blocks, but Autoware can underlie these building blocks.
Tim: Wow, the automotive industry from the very beginning has been this extremely proprietary and closed set of technologies. I mean, it’s always run that way, so what has been the reaction of the auto makers to the idea of building on top of an open-source platform?
Shinpei: I think that there are several points of view but in general, open-source software must be useful for the automotive industry whether they use it or not for the commercial products because when you have open-source software, you would have your people educated using autonomous software or your broad-type RND systems can also use this open-source software. Now, it’s up to you if you want to keep using this open-source software for your commercial products, or of course, you can switch.
Tim: Okay, that makes sense. Listen, before we dive into the business side of things, I want to back up and talk for a minute about you. So you mentioned before, this started originally when you were an associate professor at Nagoya University, and so far, your career has been entirely in academia, at Nagoya University, University of Tokyo, so are you still involved with academia? Do you kind of balance between running a startup and being an assistant professor now?
Shinpei: I’m still deeply involved in academia but I want to make a kind of new career. For me, I don’t really see in the past that university professors actually really do the business, I don’t know why, but to me, why not? So I could be kind of a pioneer to show professors can actually do the business.
Tim: How do you balance your time? Because like, being a startup founder, it’s more than a full-time job.
Shinpei: Yeah, that’s a good question, so Tier IV is deep tech. If Tier IV is not deep tech, I think it was very difficult for me to balance between startup and university, but given that Tier IV is deep tech, what I do in university can actually underlie what we do in a startup.
Tim: So your academic research is also really tied into Autoware and Tier IV?
Shinpei: Exactly, that’s why I love open-source. If you don’t use open-source, then you may see some legal problems if you just use your university outcome to your private company, but here, we search outcome from the university can go to open-source, then my startup uses this open-source of their business, that’s the strategy.
Tim: Yeah, I love the concept of that. I mean, I’m a fan of almost anything that is open-source. The more open and sharing, and collaboration we have in the world, the better. Well, actually, tell me a bit about the Autoware community? Who’s supporting and who’s building the platform?
Shinpei: So Autoware Foundation is three years old. Including Tier IV, more than half of the community, they actually come from startups, but now, we’re getting more large companies in the community, but still the core actually startup.
Tim: Yeah, I mean, I think, yeah, I mean, large company participation is always a great sign for open-source projects of any kind, but most, like, really successful open-source projects have, oh, you know, this core group of individual-like people who are contributing and the companies kind of form around them, and so are there also, like, individuals that are also contributing to this open-source platform?
Shinpei: You’re right, the core group of Autoware Foundation, the startup too, wants to jump into the automatic driving market but they believe that they cannot do it by themselves, right? You want to make an alliance with the partner companies in order to compete with or cooperate with automotive makers or big tech giants. They want to build a system not by themselves but by using this alliance, that’s the core group of the foundation.
Tim: Yeah, well, I mean, it makes sense. It kind of levels that playing field so that small startups can tackle a small piece of the program and together build something great.
Shinpei: It’s something very large scale.
Tim: Yeah, but, so this would be, like, a simple basic question but for open-source software, if you’re looking at something like Linux or mySQL or something, developers can install it on their own laptops and play with it, and experiment and see how they like it, but how do you do that with self-driving car technology? How do, like, individual startups and contributors booted up to try it out?
Shinpei: In practice, they actually play it on the cars, seriously. They download Autoware and they have their car, make it modified in order to accommodate software stack like bi-wire modification, so if you have a car with bi-wire modification, you can play around the cars, but yes, you are right, this is not scalable, so today, it is interesting to see there are more working simulators, so if you have simulators, can drastic improvement in quality of graphics, so you can use actually game engines, for example, as a simulator and emulate sensors information, then run Autoware on top that you can test: perception, localization, planning, control.
Tim: Okay, that makes sense, so it’s step by step, running in simulation and then for some of them like you say, if they’re really dedicated, they can modify their vehicles. In one of your articles, you’ve mentioned that you’ve run tests in over 60 location in this Japan for this technology, so what are these tests like? What are you testing? What are you learning from them?
Shinpei: There are two types of testing, technology testing and user experience testing, so technology testing, you want to see if Autoware can drive in, for example, hotel, but for user experience testing on the other hand, it’s also interesting to see how people who ride in kind of robot taxi react but you may also want to see how people in local areas who really look for this technology for mobility basis, it’s valuable for us to see if people in suburbs or local places would actually feel beneficial on Autoware-based robot taxis and buses.
Tim: Okay, and how do you get permission to run the road test? I understand regulations are really strict in Japan.
Shinpei: So I would say the regulation in Japan is very clear rather than strict, so in theory, you don’t need any permission to run a road test in Japan if you have a safety driver in the driver seat but in practice, because you don’t want to make people surprised by operating robot taxi even though taxi driver is inside, so as a protocol, yeah, we talk to a place in the local authority then get not permission, get agreement with local police, then you’re ready to go, but if you want to operate driverless cars, it’s a different story because the cars must be protected by regulation so you have to submit kind of regulation waiver for the police, also road authority, then you will be tested on site and if you pass the test, you’re ready to go.
Tim: Okay, so far, all of your tests have been these kind of emergency drivers in the vehicle?
Shinpei: I would say most of the few tests that we do, we have safety drivers inside but Tier IV is getting more recognized in Japan because we are the first to showcase real driverless car on the public road, so we went through the whole processes with the police and road operating, so we are now qualified to operate driverless cars that is a car without any human driver on the public road.
Tim: Which is an amazing accomplishment, I mean, actually, let’s talk about that. So on your website, you have, like, several different types of autonomous vehicles that you’re working. What is the type of car where you have authorization to run driverless on public roads.
Shinpei: It’s a robot taxi based on the Japan taxi model which is the most popular vehicle type in Japan. They actually use that car for real taxi services, so we modified it to accommodate Autoware, then submit emission to the police. Now, we are qualified to operate driverless robot taxi even in Tokyo.
Tim: So a really successful pilot, you got a lot of great press attention for that. What is the timeframe before we’ll be able to see driverless taxis as part of our everyday life?
Shinpei: If you talk about the deployment, I would say five years to ten years, but if you say the first real commercial robot taxi, then I think it will appear in a few years.
Tim: And what’s the bottleneck there, is it more on the technology side? Is it more on the automotive production side? Why is it a few years instead of next year?
Shinpei: Technology side, right? So if you have more vehicles, then you have more risk, so the technology today enables probably a few vehicles at the same time to be operated without any human driver, but if it scales out to 100 or 1,000, or 10,000, then today’s technology may not be sufficient to operate all of them at once.
Tim: Around the world, these robot taxis are getting all the attention cause I think that’s one of the most scalable use cases for it, but you also have an autonomous vehicle called the Miliee which I think is a fascinating use case, and can you tell us, like, a little bit about what that is?
Shinpei: Sure, so the design concept of Miliee is not for high-speed public roads. It focused on low-velocity areas because if you lower the speed, you have a lower risk. If you lower the scope of the area, you also have a lower risk. The Miliee is because of real-world application of autonomous driving.
Tim: Yeah, and I mean, it looks like a golf cart with a bunch of sensors attached to it but there’s one thing that I’m curious about with the technology that supports autonomous vehicles, so for example, a low-speed autonomous vehicle that has to run around crowded backstreets with lots of pedestrians but moving in a very slow pace is going to be very different than, say, autonomous vehicles that are doing like long-haul trucking, which would be very different from taxis or buses, so in all of these different use cases, how similar is the AI technology that is used in each of them? Are they very similar technology, they’re just tweaks or are they almost different from the ground up in how you approach solving these problems?
Shinpei: In the testing phase, low-speed mobility do not require the same level of AI as robot taxi but if you consider real-world applications, then you shouldn’t have any edge cases, you have to ensure that the whole possible use cases must be covered even though it’s a low-speed vehicle, so ultimately speaking Miliee and robot taxis, they would require the same level of technology. The concept here is, given that we use the same level of technology, area, very limited such as a park and unlimited area such as a city, through the same technology or must come first to the real-world application, then once you have the success, then you can go to a robot taxi application.
Tim: So on a practical level, what sort of autonomous vehicles do you think we’ll be seeing first on the road?
Shinpei: What do you mean by on the road?
Tim: Okay, yeah, that’s a tricky thing, so what sort of autonomous vehicles do you think we’ll be seeing deployed first?
Shinpei: For sure, geo-fenced road sticks application, so if you move things instead of people, you get low risk, right? But if you move people, then you also have to consider the people inside the car.
Tim: So these would be things like factory automation or…
Shinpei: Yes, exactly, cargo deliveries, like experience should come first to the real application.
Tim: And since it’s a really well-defined area, it’s easy to set up beacons and other things to kind of help the AVs?
Tim: What do you think is the first use case of AVs carrying people?
Shinpei: That’s, I believe, case of Miliee, low-speed limited area, limited condition. Not only Miliee, but also for example, robot shuttle in a private area, they basically have the same risk level as Miliee, so such categories come to the realization in a few years.
Tim: Makes sense. Let me zoom out a bit and talk a bit about Tier IV, so what is the fundamental business model of Tier IV? Is it joint research projects? Is it software development? Is it integration?
Shinpei: That’s actually a very interesting question for me. My true answer is I don’t know because market is not there, right? How to business model if the market is not there, but my assumption is that since we are based on open-source software and open-source software attracts more people, more companies, they want to use open-source software as a basis of their business programs, but they don’t know how to integrate open-source software for their production quality system. Once you go into production quality, you need somebody who can help. Imagine that Linux, students can use Linux, playing around. Once you go to the business, you want to to use Ubuntu. Red Hat because business model is basically an interpolation, optimization, customization over Linux, but Tier IV is following that style. We are the main distributor of Autoware.
Tim: That makes a lot of sense, so I guess the big question here seems to be like how open these automakers will be to open-source. It’s one of those questions, like, if you look at the market as a whole, everyone benefits from having open-source because it allows people to specialize and learn from each other, and it keeps costs down, but any one actor, whether it’s Toyota or GM, or whoever, doesn’t look at the whole market; they’re just looking at themselves, and looking at the members of Autoware, there’s a lot of corporate members, there’s governments, there’s university support, but so far, the corporate members are mostly like software or AI, or systems integrators, and there aren’t any automakers yet. Why is that?
Shinpei: I like this conversation, so consider two aspects, from our side, if you invite one automotive maker, then other automotive makers will not join, that’s the automotive industry. So we want to be a neutral place. From automotive maker’s point of view, today’s automotive industry for autonomous driving is something like before 1990s in the PC market, so they don’t need to use open-source today but as autonomous driving emerges and more connected applications emerge, I believe they will not be able to develop the technology by themselves. Now, start relying on third-party or partnership to build product. If we go to that future, then open-source is very strong. If you look at the history, if you need your partners to build your system, then if you have a partnership with closed community, you can’t get market.
Tim: That’s interesting, yeah, I could see that, so it would be no direct involvement until everyone comes in at once?
Tim: That makes a lot of sense. If you’re thinking the first vehicles that are going to get rolled out are going to be things like the Miliee or cargo, physical goods transport, then it wouldn’t be the automakers in the early stages, it would be, well, I mean, maybe it is, but some small division within those automakers, yeah.
Shinpei: Yeah, you can showcase using a limited application, then once automotive makers start relying on software companies, now, they may want to use open-source.
Tim: So if I’m understanding you right, 10 years from now, you see Tier IV as being the main trusted integrator for all of the different automotive and automotive-related companies that are using this platform in their vehicles?
Shinpei: I hope so. In addition to that, so in 10 years, we definitely want to be a main integrator of autonomous driving technology following, say, Ubuntu, Red Hat style business, but when you consider open-source business, there is another class of business model which is kind of Android. Android is a completely different business model. It’s not an integration business; it’s a platform business. When your partners or end users, they use Android, you have a business opportunity to make them install another application or service, so if Tier IV has commercial service using Autoware, more users use Autoware, most business opportunities for Tier IV. Yes, we need to be an integrator first because we have to deploy Autoware first, but once the Autoware is deployed, Tier IV can be another phase as a service provider using Autoware platform.
Tim: Yeah, that makes a lot of sense. Let me ask you about Japan in general. So there’s so much amazing fundamental research going on in Japan with what you guys are doing and Preferred Networks and Toyota have really done some interesting research as well, but how does Japan’s autonomous driving technology as an industry stack up to what’s being done in the United States and China?
Shinpei: That’s as a whole exactly the reason why I started Autoware, Japan is a small country, Japan alone never reaches the same place as United States and China because they are super countries but Japan is an island country, but making partnership in global places, so Autoware wants to be such a hub coming across global companies alliance, so Japan alone probably can never reach the same place as United States and China but Japan and global countries alliance may be able to be at the same place as United States and China.
Tim: Yeah, that makes sense, and I mean, that’s why the open-source component is so important. It allows everyone in any country to contribute to that.
Shinpei: Yeah, that’s a beautiful story, right? Because a small product, small university team in a small country such as Japan can lead a very large scale alliance using open-source community, it becomes the same level of products or technology as what the super countries are doing.
Tim: So is Autoware similar to the Apollo project in China? Cause that’s also open-source, isn’t it?
Shinpei: Yes, Apollo, in some sense, is kind of great presence for Autoware because if there is only one open-source project Autoware, then we didn’t make such a success, but there was another open-source project, Apollo, so Autoware and Apollo made this open-source community today.
Tim: Okay, it’s like every founder knows you never want to be the only one in your market.
Shinpei: I know.
Tim: Cause then nobody wants, they don’t feel safe, it doesn’t feel like a real thing until there’s some competition in the market.
Shinpei: Yeah, so in that sense, yeah, Apollo is significant. The main difference between Apollo and Autoware is that Apollo belongs to Baidu. Autoware doesn’t belong to any particular company.
Tim: So it’s open-source but still controlled by Baidu?
Shinpei: Yes, so imagine that Android belongs to Google. Linux doesn’t belong to anybody. I don’t know, there is no which is better, the style of open-source community. I didn’t want to have Autoware stick to a digital company; I wanted to have Autoware as a truly neutral open-source project, so I decided to hand over all the rights of Autoware from Tier IV to Autoware Foundation which is a nonprofit organization.
Tim: That’s awesome. Well, listen, Shinpei, before we wrap up, I want to ask you what I call my “Magic Wand” question.
Tim: And that is, if I gave you a magic wand and I told you that you could change one thing about Japan, anything at all – the education system, the way people think about risks, the way people think about startups – anything at all to make it better for startups and innovation in Japan, what would you change?
Shinpei: Speak English?
Tim: English speaking? Well, why is that? No, a number of people say that, so why is that important? Why would that make Japan more innovative?
Shinpei: So for me, education level of Japan think it’s better than any other countries but Japan is not good at being a leader in the global community. We can be part of the global community, but sometimes, we are not really leading.
Tim: And do you think that’s mostly because of, like, language ability?
Shinpei: Communication of the culture. We don’t have to change education level, we don’t have to change our skill or the system. I said Japan regulation is not strict at all; it’s very clear, so I love it. But the communication, it’s not a problem but if you want to lead the community, then I think we need another communication.
Tim: That’s interesting and I mean, your English is quite good but as a team that’s trying to build this global open-source community, have you found that challenging?
Shinpei: Yes, it was challenging, but since I’m Japanese, I think or kind of I was fortunate to be in such a place, people trust me, I’m in a good position to coordinate people, I have learned a lot how to communicate with people when I studied in the US. If I was not in the US, then I think even if people trust me, I may not have been able to coordinate people because this is not something Japanese people are good at.
Tim: Do you think that is just English ability or is there also a kind of this cultural…
Shinpei: Many things.
Shinpei: Yeah. I was kidding about English speaking. It’s communication.
Tim: Now, but I think there’s something there too.
Shinpei: Yes, so Japanese people or culture, we are very flexible but may not be so reasonable and logical, so the communication problem may actually come from the lack of chronological and reasonable thinking, but since we are flexible, probably, we have another good aspect, but if you want to communicate or coordinate the community, then you need a logical thinking or you need reasonable decision which probably Japanese people are not very good at.
Tim: Well, but being flexible means you can learn.
Shinpei: Sure, that’s maybe how I am
Tim: Okay, that’s great. Hey, Shinpei, I want to thank you so much for sitting down with me today. I really appreciate it.
Shinpei: Yeah, thank you very much, I enjoyed a lot.
And we’re back.
I love that people are hacking their cars so that they can load an experiment with self-driving software. I mean, I don’t want to ride in those cars with them but I think it’s pretty cool that they’re doing it.
Moving forward, it’s going to be interesting to see if Shinpei and the Tier IV team can develop a sustainable open-source community around Autoware. Many potential contributors might be hesitant because of the lack of participation from the auto industry.
But in the end, having the big automakers on board might not be as important as it seems.
When we talk about autonomous vehicle technology, if we only look at how the technology is being deployed in cars, in passenger vehicles, it does look like the US and China is way ahead of Japan, and passenger vehicles are the most difficult and potentially the biggest market.
But the man who coined the term ‘disruptive innovation’ Clayton Christensen, his theory of disruptive innovation teaches us that disruptive technology introduced by the major players rarely succeeds. Successful disruptive technologies are almost always initially deployed in unsexy low-margin, small niche markets. The innovators then transform these niche markets and use a new expertise in product improvements they make there to move upstream and eventually displace the large players.
Perhaps we’re about to see the same dynamic play out in driverless cars. It might be that the winners will not be the Teslas and Volvos and GM trying to develop technology that is good enough for use on the road, but rather the companies who start out by building outstanding autonomous vehicles for on-site transport or revolutionary low-speed vehicles like the Miliee, and then move up the value chain into the bigger and more profitable markets.
Of course, at this point, there’s no way to know how all this will play out. Today, we’re told that driverless cars are about five years away, just like they have been for the last 15 years, and maybe we really will see them on the road in five years, but in this technology race, I would bet on the companies who are actually shipping amazing and fully functional autonomous vehicles, even if they’re only selling in small unsexy niche markets. Actually, especially if they’re only selling in small unsexy niche markets.
Because those are the companies that end up changing the world.
If you want to talk more about driverless cars, Shinpei and I would love to hear from you, so come by DisruptingJapan.com/show174, and let’s talk about it. If you leave a comment, I guarantee, Shinpei or I or maybe both will respond.
And hey, if you get the chance, check us out on LinkedIn or Facebook, but even better, if you like the show, tell people about it. Disrupting Japan is my labor of love. It’s free forever and we have no advertising budget. People hear about the podcast because listeners like you enjoy it and they tell their friends about it.
But most of all, thanks for listening, and thank you for letting people interested in Japanese startups and innovation know about the show.
I’m Tim Romero and thanks for listening to Disrupting Japan.