Oren Buskila: 0:03

Actually, once you have a lidar, it's also possible. Now it becomes really possible to have an over the air update that makes the car from one that has very good ADAS to one that's actually autonomous.

Tom Raftery: 0:20

Good morning, good afternoon or good evening wherever you are in the world. This is the digital supply chain podcast, the number one podcast focusing on the digitization of supply chain. And I'm your host, global Vice President at SAP. Tom Raftery. Hi, everyone. Welcome to the digital supply chain podcast. My name is Tom Raftery, with SAP and with me on the podcast today I've got my special guest Oren, or would you like to introduce yourself?

Oren Buskila: 0:46

Yeah, hi, Tom. I'm Oren Buskila. I'm the Chief r&d officer and one of the co founders in a company called Innoviz technologies. We make LIDAR and perception for self driving cars.

Tom Raftery: 1:01

Okay, that's got to be a small market right now I got to think, but presumably you're betting that that's a market that's going to explode sometime in the next. I mean, depending on who you listen to. If you're listening to Elon Musk, it's already here, you just have to press that fsd button. If you listen to some of the other people who were a bit more cautious about it, it's 10-15 years away to even infinity away.

Oren Buskila: 1:30

Yeah, this is a today obviously, we don't see too many autonomous cars on the road, right? Actually, none. No company has reached that. That's the point yet. That's a fact. But this is definitely coming. And, and once this, this becomes a reality. Yeah, that this, this will be a huge market, because humans are terrible drivers. And, and we, you know, we we show that we demonstrate that every day on the roads. And this is no, it's tragic. But this has caused this is causing $1.3 million million deaths, excuse me a year, and around 50 times as many injuries. And this is a problem that that we we need to solve and, and almost all of those accidents are caused by human error, something like 93% of that. So the way the way to solve this, this terrible problem globally is to give the give the wheel to the to the car to drive itself because you know, a car or computer doesn't doesn't read WhatsApp messages during during the drive and it doesn't fall asleep. But it's still very it's, of course, it's very hard to develop it to this stage. But but this is our mission. And one of the reasons we don't see self driving cars today at all, is because the technology, the technology that we're building at innovyz isn't yet ready for the road. It's been it's going there, it's almost there. But LIDAR is such as redo are just not there yet.

Tom Raftery: 3:18

Okay. So, if the Lidars such as the ones that you create are not there yet. What are you making today.

Oren Buskila: 3:29

So we are working on on those items. And we actually we going to, to deliver the ladder to the first autonomous car, this is going to to be shipped in volume to the to the public roads, this is going to be a BMW car later next year, and innovyz was was nominated as the as a supplier of the LIDAR to that to that self driving BMW. So we're working now on the final stages of, of of validating the product that we're providing to BMW installed the innovation one, it's a automotive grade, high performance, low cost lighter. So this is what what we'll focus in most of our efforts now. And in parallel, we're already working on the next generation of sliders called the university. This is the name of our product, which will be something like 30 times more powerful than the universe one and yet three times less costly. So while the universe one is a good solution for premium cars, like the BMW then of his two will be a perfect, perfect solution for for the mass market for the you know the Honda's Toyota's the Chevrolets of the market. But this will be in 2024.

Tom Raftery: 4:54

Okay, okay, good. Just a timeframe. Famously, Elon Musk He has said that you should only do autonomous cars with cameras, and you shouldn't use LIDAR. So what do you know that he doesn't know? or What does he know? That you don't know? Or how does that work out?

Oren Buskila: 5:16

Well, it's really it's really not not Oren B skila against Elon Musk, it's El n Musk against the world, as, as you probably know, is he's ba ically maybe the last almost th last person standing with wi h this opinion and is is en itled to it. But I think re lity has proven so far, that it s somewhere between very, ve y, very difficult to im ossible to, to build a car, a ca, self driving car, without a la der. So maybe we should ex lain what that is to the li teners, because we've been ta king about for for almost 10 mi utes without explaining it. So LIDAR is basically a sensor, it s a laser radar of a light ra ar, this is why it's called a LI AR. And it's sort of like a ca era, only that it sees in 3d, ou camera only sees in 2d. A lo of can see in 3d, it can it ca sense the, the range of any pi el in in the in the picture th t that it creates. It's also an active sensor, in the sense th t it doesn't need any ambient li ht to see a camera needs, ne ds, the sun, the sunlight, or li ht from road illumination to se what's ahead of it. But Al ah is active, its uses its ow laser source to scan the sc ne and incense, the signing on the car. So a letter is it so t of fills in all of the ca es in which a camera fails. An the places where the camera fa ls are some of the reasons wh why we've seen Tesla's you kn w, driving into parked trucks wa too often, or even or even, yo know, over an overturned tr ck on the roads, which Tesla mi sed in the past causing ac ually the death of the driver in that specific case. So a ca era doesn't see well in the da k. It especially doesn't see we l, when does Quick Change of sh des, shadows on the road en ering a tunnel or going out of it or, say shadows of trees. On the road, a camel really has a ard time with such co ditions. Also, sometimes you ju t get sort of those like op ical illusions that even even a erson could fall for, let al ne a camera. But a LIDAR gi es the car the absolute sense of the range to each object. So th re is no missing say, a tire on the road, a camera is very, ve y hard for it to see a dark ti e on a dark road, especially du ing night. A lot of can de ect it. So a camera with a LI AR together creates a much mo e complete picture of the sc ne of the road ahead of the ca. And actually there's also a th rd sensor in the game, which is a radar. Radar uses el ctromagnetic waves in a much lo ger wavelength than a li hter. It's it isn't as good in terms of the resolution as a ca era or LIDAR. But it's very it s very robust to to en ironmental conditions like ra n and fog. This is the main re son why they are traitors. So ba ically every this was sort of a ong intro into into first wh t the latter is. Now. The fa t is that almost all car co panies today understand and ag ee that you need all of the th ee sensors in order to ac ieve self driving. And and El n Musk is still the one that in ists that you don't need it. Bu the actual results do not ag ee. The facts with Elon Mu k's opinion so far. And it's al o I'd say I think Elon Musk is taking the maybe maybe the pragmatic approach to selling cars now. Right? People want to want to believe that they buy a Tesla now and it's it's just a matter of an over the update and it will be autonomous next year. It's always next year. So if Elon Musk's If he said that you need a lighter, you needed a lighter than, you know, he couldn't sell for self driving cars. But I think I think that as soon as ladders do meet the cost target that they need to the maturity, the volume. I don't think there's any reason why why why Tesla won't use that as well.

Tom Raftery: 10:23

Okay. Yep. Makes sense. Just for people who may have seen pictures of self driving cars, those test ones that are going around the roads in like Phoenix, Arizona or the Bay Area. Typically they have a dome on the roof with a spinning mirrors, which is what spins out the the the LIDAR waves, yours is different, right? Yours is a solid state. So no, no big dome and spinning mirrors is am I correct? In that

Oren Buskila: 11:00

correct. So what the dome that you mentioned is, is a spinning LIDAR or no, that we use, we call them spinners in the in the industry. And this is maybe a very straightforward to design a LIDAR because you just you just use a lot, lots of lasers, lots of detectors, you use a big mechanical motor to spin all of it. So it's like the brute force approach to design a ladder. So it you know, it achieved the objective of, you know, making ladders available for, say, research and development projects 10 years ago, but it doesn't scale. And the reason is that it's it's very big. This is why you can see those from afar. Those big domes, it's very big, it's very expensive. And I'm talking in the high 10s of 1000s of dollars per unit. And it also tends to break it's not very reliable, because of all of this hardware and all this mechanics, nothing reliable. So obviously this This can't go to Mass production, or you know, if it does, it will, you know, quadruple the price of the car, which is unacceptable. Of course. So it's enough is we developed a LIDAR that is has high performance, but is is many times smaller, cheaper and and designed for automotive reliability, which is a challenge in itself, you know, 15 years on the road without a single failure. This is very hard. So this is what what we attend have is designed. This is this is why BMW chose this, this product over the the alternatives. And we we just believe that if we don't do it, this this is on us the fact that you know, Thomas cars keep on being pushed, pushed further away. We it's our responsibility to, you know, to make them a reality. And we started the company, something like six years ago. This has been our mission since then we'll now over 300 engineers here working on that towards that target we recently IPO on NASDAQ by the way innovators not publicly listed, this is all part of the of this bigger mission to to allow our customers to develop and deliver the autonomous cars. Very good

Tom Raftery: 13:40

congrats on the recent listening, by the way, thank you. How do lighters cope with inclement weather conditions, rain, snow, fog, things like that.

Oren Buskila: 13:51

So they're usually much less susceptible to those conditions and cameras. But also usually not completely immune to them. So you do it's somewhere in between between cameras and, and radars. And this is why usually need the three. But something that's interesting about it is that also part of what makes car lighter special is that because it's a self driving car, and the driver is supposed to be able to be maybe asleep or certainly not not attentive to the road. The car needs to know if if the self driving mode is not performing well. So let's say that there's there's rain or fog that affects the the cameras range and the LIDAR range. The current needs to know that so the LIDAR needs to tell the car Hey there, I think I think there's some fog on the road and the range is probably 30% lower now or something like that depends on the fog. So maybe we should reduce the speed a bit. So this is something that you need to build into the like intelligence to be built into the LIDAR to be able to be used in this in the context of self driving car.

Tom Raftery: 15:08

Okay, we've been talking a lot about cars so far under the episode but for for people who are listening to this podcast because it's a supply chain podcast, there may be more concerned with trucks or they may be more concerned with delivery vehicles or service vehicles are those smaller autonomous delivery trucks as well, little delivery vehicles that go on footpaths, you know, the kind of new road vehicles for example, are you developing for all of those or just cars are what what kind of range of vehicles are you looking at.

Oren Buskila: 15:42

So basically, every thing that is robotic, will will need a LIDAR. So everything that's that you mentioned, like long haul trucks, right? This this is using those are using cameras and radars and also ladders. Today, these are just the r&d fleets. But obviously they will need all of those also in production. And and say sidewalk delivery robots and even fulfillment center robots, all of those need need data because there's there's no sense of like it in, you know, producing this, this 3d image of the world. So all of those are important, all of those will be huge, huge markets. But we believe that enough is that the the, the most important one, and also probably the one that will mature the fastest is the is the passenger car market. The reason why it's important is because, you know, this is I think in the in the foreseeable future, say the next five to 10 years, this is where the biggest opportunities are in terms of both saving lives, changing people's lives, saving people's time. And also, commercially wise, this is where the big numbers are. And something like there's there's something as 80 million cars produced globally. And all of those or this will gradually shift into into a state where, where all of the cars will have LIDAR, some will have more. So we see this as the biggest opportunity in the in the next few years. Also the opportunity to make the biggest impact. And the thing is the car doesn't have to be fully autonomous, to actually need the LIDAR cameras, cars are becoming safer and safer. Because of better as a this is his advanced driver assistance systems and these are getting better and better. And the the next phase the next generation of those systems needs a LIDAR and actually once you have a ladder it's also possible now it becomes really possible to have an over day update that makes the car from one that has very good data to one that's actually autonomous autonomous say in in certain conditions. It's called level three it's highway highway autonomy. Sure, but once you have you have the hardware that you need if you have a camera and a radar as well and the right computer, but the ladder is like the missing puzzle piece now. So this is why we're targeting the OEM market the passenger car market first but it's important to note we're also working with many many customers in the trucking business in the robotics business we're just seeing that this this is picking up more slowly but going forward those will also be very important markets to to analyze okay and

Tom Raftery: 19:01

what kind of size device are we talking about roughly? Is it like the size of a mouse you know and I don't mean the animal computer mouse are our you know a wallet are a book are you know what kind of size because they I'm assuming they need to be kind of Roof Mounted so they get the best possible view is that is that safe a safe assumption.

Oren Buskila: 19:26

So obviously it has to be in the front of the car. Well at least it has to be at least one in the front of the car. This is for highway autonomy, and are several assembly or mounting locations. Each each car maker has its own preferred design, or sometimes even several designs. So we've seen customers targeting the grill, some VMU the headlights and some indeed the roof roof mount location, which is actually the one of the of the best in terms of possible performance, the size. So this this is why the logo actually has has to be has to be small because if it will, you know, keep being that big as the ones you see on the on those way more cars, then there will be no place to put it back on the roof, which is just not very appealing. Terrible, I mean that the car designers may be the most powerful people in the car company for good reason because you know, we will still buy we buy cars with our no gut in the heart and not not only with with our mind. So it has to look to look good. And this is the lesson is at home can see what I'm holding now that what that you can see, but I'm holding something that's roughly about the size of a computer mouse, maybe a computer mouse and a half. This is the this is the hang of is one. very slim, this is four, four and a half centimeters high. And this basically fits in could fit in the grille of a car. On the roof. We've even seen the designs where it fits in the side mirrors, especially for trucks, okay, so this is very compact, and you can feel it not only in the front, but some customers are also using several of those around the car to have a full 360 degree coverage. This is much more important for level four and level five autonomy, which which means self driving, not only on the highway, but basically anywhere.

Tom Raftery: 21:38

Okay, interesting. There's a LIDAR apparently built into the back of my phone, which is obviously considerably smaller than the one you held up there for the car. What's the difference between what's the difference between the one in my phone and the one that you have for the for the cars,

Oren Buskila: 21:55

the biggest difference, most notable, most important is the range. So the one in your in your phone in the in the iPhone, it has a range of several meters. So the laser can be very, can be weak, it doesn't have to be very powerful because it only needs to reach a few few meters, the field of view is very limited, you don't need to see the like like our innovis one we see 120 degrees because we need to see several lanes in each side of the of the car. So it's mostly mostly the the range vendors some some other performance aspects. We actually considered in when when we started the company, we thought, well, we should take the technology we want to show you at about like, it should be the car market. But because we understood we have we have this technology that can reach hundreds of meters of range with very high resolution. So we said this is not something that could fit in the form. But it's actually this is not the right for us, this wasn't the right application for it. hundreds, hundreds of pieces of range. What needs this kind of ranges is an object that travels at you know, 150 kilometers per hour, this is a car so that our technology is really adapted optimized for for cars and in other vehicles.

Tom Raftery: 23:20

Okay, super. What about the the software that runs it.

Oren Buskila: 23:57

So I say there are probably several, like two layers, two layers of software, there's just the the processing software that just you know, make sure processing and control make sure that the ladder will work will function. This is an embedded software that runs on this on the slider hardware. And then there's another layer for the perception, right? If the if the LIDAR produces just the point cloud, the point is the 3d equivalent of a picture. The perception understands what's in this point cloud. So whether there's a truck and a car and a pedestrian in a cyclist. So this is like like the brain behind the eyes behind the LIDAR. And, and this, this is something another part of what we do within a business. This is like the, the the second leg does the How to like then there's the perception of software like and this is this is a huge, huge, huge challenge in itself because this is just as important and as complex as the ladder we have dozens of engineers working solely on that. And the car, the car makers, they they try to have this output from each of the sensors so the camera tells them I see a pedestrian here and a cyclist here and the cam and the radar surf tries to tell them the same and then the lighter also produces this object list. And to put it simply they they inhale fusion layer, they try to make sense of all of those inputs together, see that they match. And if they match that, then the calc can be sure that it's actually getting a good good understanding that it can rely on of the environment. So in the camera, you probably know the name of the of the of the most well known company that that develops and provides this system, the perception based on camera it this is Mobileye, another Israeli company, by the way, and acquired by Intel a couple of years ago. And we need to do basically the same that Mobileye does, just on a 3d representation of the world rather than 2d. So so this is also very important, and very challenging. But the good thing is, once we solve it, on our hardware, this we make it easier for all of the, our customers to just use this already processed point cloud with the object interface, instead of having to develop it themselves, say like, like, way more? Did we developed all of this perception themselves, but we deliver it as a solution to our customers,

Tom Raftery: 26:50

okay? And if I'm an OEM if I'm a car manufacturer, how do I take the output from your LIDAR, and then mix it up with output from cameras and output from radars and decide which one is right. If there's a, I don't want to say a conflict, but if there's, you know, two sensors, one saying there's a car there another saying, No, there's no car there? How does? How does the car? Say, yes, there is something in front of me or not?

Oren Buskila: 27:20

Good question. And this is actually one of the most important reasons why you need to have more than one, or even two sensors. And if they need to be different sensors, because if they're both the same, then they can, they can fail. In the same exact conditions, if you have three cameras looking at the same, the same thing, they might all be wrong. But if you have different, it's called modalities, you'll get one camera one light or one radar, then each one sees the world differently, then you can actually count on on the aggregate on the integrated input of all of those. So you could say that, like the maybe they're the various the over simplistic way to say is two out of three, two out of three agrees agree that then there's there's probably something that about, it's obviously more complex than that, and it depends on the situation, you can you can also assume that different OEMs have their own proprietary systems are basically basically the trying to match the, the outputs of all of those sensors and see that it makes sense. And if it doesn't, by the way, then you know, you you have you have some sort of, of a risk, one of the sensors might be damaged, or maybe there's some unusual conditions on the road, which causes some of the sensors to, to get confused. So maybe the right thing to do is to is to make a safe stop. And, you know, call call the, the the tow service, for example. It can it can be you know, any of several possible actions in this case, but but at least the car knows if, if if there's a mismatch, then it needs to be very careful now.

Tom Raftery: 29:16

Because cars are mostly computers these days, the best course of action is probably just to just turn it off and turn it on again.

Oren Buskila: 29:23

Just don't do this while driving on the highway on the Autobahn at 20 miles per hour. Right?

Tom Raftery: 29:31

You said your your next device coming out in 2024, I think you call it the innovyz. Two is going to be substantially better and substantially cheaper than the individual one. And so it will be more for the mass market by being substantially cheaper is that just because the technology is getting better? Or is it because you've cut out some feature from it or how have you managed to cut the costs.

Oren Buskila: 29:57

So it's it's both one of the of the reasons that we're actually relying on advances in some of the of the core components, core technologies that we use in the LIDAR. But I think the bigger part of this, of this change and this cost saving is just much, much better engineering. And we do this based on everything that we learned in the in the with the universe one, both in the design and the manufacturing and the reliability Testing everything that we that we learned in the universe one we implemented our implementing in the universe to, so we saved something like three to four times of the optics, we we just we eliminated all of the rest, we we just kept something like 25 to 30% of the optics. So we saved all of those, all of this cost, the optics that we did save, we designed to be much simpler to make. So it costs less per part, we reduce the number of electronic parts, we simplified the manufacturing process. So basically everything is is now optimized. And we we reached an almost three times cost saving compared to the to the universe one, while actually boosting the performance by a considerable amount. And we, we really believe that this is the the product that will enable self driving in for the masses, right? I think this this is this is the first slide that most people will actually see on the road. And we're seeing really amazing feedback from, from the first customers who have their look into this product.

Tom Raftery: 31:56

Okay. All right, or in we're coming towards the end of the podcast. Now, is there any question I haven't asked that you wish I had, or any aspect of this that we've not touched on that you think it's important for people to be aware of?

Oren Buskila: 32:10

Well, you've asked a lot of important questions and interesting questions, and we could talk about it, you know, for for hours and days and weeks. And that not really finished. But I think many, many people are wondering what happened to all of those forecasts, or expectations, like two or three years ago that that self driving is is around the corner on it's like, you know, companies said that it will be in 2020 and 2021. And I think some people may may be disappointed with what like with a delay that entire industry is experiencing. And what I want to say to those to all those people, your listeners is first I'm I'm with you understand it. And I think we all would wish it to be to be here earlier because it's really, really improved our lives. But this is very similar to all many other complex technologies, it does take more, sometimes longer than then originally expected. But but this is I can guarantee this is on the way this is happening. This is going to be the reality we are seeing the reality happening will help will helping to build it. And in in a matter of a few years, you will experience it yourself. So you should you should keep that anticipation This is coming.

Tom Raftery: 33:41

Fantastic. Fantastic. Oren, if people want to know more about yourself, or about your interviews, one or two, or universe technologies, or any of the aspects that we talked about on the podcast today, where would you have me direct them.

Oren Buskila: 33:57

So there's the innovyz Technologies website, where we link it to many other four sites, we give some final resources. There's YouTube innovyz Technologies on YouTube, where you can see lots of really, really cool point clouds, like recorded data from all over the world there. And as there's our web pages on on LinkedIn and and Facebook will, we and I and my partner and and co founder Omar Caleb the CEO, we post regularly interesting updates on the company and also also more point clouds of the ladder. So you're all welcome to give. give this a try. And I think you'll see it's interesting.

Tom Raftery: 34:46

Super, super Oren. That's been great. Thanks a million for coming on the podcast today.

Oren Buskila: 34:51

Thank you, Tom. This has been a pleasure.

Tom Raftery: 34:53

Okay, we've come to the end of the show. Thanks everyone for listening. If you'd like to know more about digital supply chains, head on over to sa p.com slash digital supply chain or, or simply drop me an email to Tom raftery@sap.com. If you'd like to show please don't forget to subscribe to it and your podcast application of choice to get new episodes as soon as they're published. Also, please don't forget to rate and review the podcast. It really does help new people to find the show. Thanks. Catch you all next time.