This post was originally published on Autocar
Innoviz lidar is used to enable level 3 autonomous driving in the BMW 7 Series
Lidar unlocks autonomy, and to unlock lidar, you need to speak to the plain-speaking boss of Innoviz
Lidar is the technology that’s going to unlock eyes-off, hands-free driving.
In Europe, the company at the forefront of development is Israel’s Innoviz, which supplies BMW to enable level-three self-driving in the 7 Series and will also drive similar level-three set-ups in Porsche and Audi.
‘Lidar’ is an acronym for ‘light detection and ranging’. Essentially a 3D camera, it fires lasers ahead of it to create a ‘point cloud’ that helps the car understand the relationship between objects around it.
European cars have so far used the technology sparingly, if at all, but Chinese car makers love it, even though they’ve not gone beyond level-two (eyes on) assisted driving. That’s despite the fact the technology is also expensive and tricky to master.
Innoviz co-founder and CEO Omer Keilaf is very confident indeed that his technology is going to drive the autonomous revolution. He’s also very open about the competition, why level three is actually harder than level four (driverless) urban autonomy and how China both represents the biggest threat and actually doesn’t worry him at all.
We spoke to him at the recent CES tech event in Las Vegas…
Omer Keilaf on why L3 is tougher than L4
“People expect me to say that the level -four lidar should have better resolution and range because there’s no driver, but that’s actually not the case. The requirements are higher for level three, because of the speed on the highway. You need to see that small object 200 meters away, because you’re driving so fast and you need to react.
“Level four is for urban situations and is much slower. The real difference between levels three and four is the lack of driver. You need the lidar to be super resilient to anything that might happen. And if the car goes into a puddle, then you need to stop to apply the cleaning system. This is the worst situation: where the car needs to stop and everyone in the car has a problem.”
On unscrupulous lidar companies
“When companies design a lidar and don’t take into account real life, you won’t see the problem in a five-minute demo, because lidar companies know where to drive to avoid seeing this problem. There are dirty secrets – like really dirty.”
On level-three programmes that failed
“Several level-three programmes failed because the OEM didn’t know really enough about how complicated lidar is.
“One OEM after two years discovered that the lidar they were using was so strong that it actually burned cameras in the streets; take a picture of the car and the image sensor on your iPhone will be burned.”
On the cost issue
“How much was Tesla’s Full Self Driving [system] priced at? $15,000? $100 a month? What would be a relevant price point for a level-three system for the customer? I would argue that the $500 lidar is not the problem.
“Let’s go to the end game, level five: you no longer have to pay insurance, because you will pay the car company, and you get a chauffeur.
“I agree that to get to 30% and 50% of the market, it needs to get cheaper. Will it get there? Absolutely, no doubt.”
On the relative simplicity of lidar and how to cut costs
“If you open the box, you will be surprised how simple it is.
“We have our own ASIC chip, which is about $40, but that’s because that’s in low volume, right? There’s nothing in there that is fundamentally expensive. In the future, there will be more circuit-integrated components. Today you see many, many discrete parts. That’s because that OEM wants this field of view, that OEM wants this resolution, that OEM wants this size and connected with this internet.
“There’s no standard right now. I need to keep the system very agile, and flexibility is cost, right? Once I start to print this 10 million times and the cost of saving $2 is an investment of $2,000,000, I will do that, right? That’s industrialisation.”
On using too much computing power and energy
“It’s easier to process lidar data than camera. If you look at Tesla, the first translation of the camera is to make it 3D, because machine vision requires 3D understanding. I can use a very small part of the picture. The number of pixels I need to process to translate is significantly lower, because it’s not a big image in high resolution. I can use significantly less processing power to determine whether it’s a truck, car or human being.”
On why Chinese cars have so many lidars
“In China, you will find many cars driving with lidars at level two, and I would say ‘why do you need it?’. You don’t need redundancy; you have the driver observing and ready to react if something happens.
“I think that the Chinese have used lidars to leapfrog some of the hurdles they had with the software. I actually expect that they will use less lidars over time, because their software will mature.
“Once you go to level three, you can’t be in a situation like in level two, when the lidar can be reset every five minutes if it has a bug. At level two, who cares? It’s not good, but you can actually go to the market. The driver is responsible and you can have any problem you might think of. “
On how China is helping Innoviz
“China is creating a crystal ball that allows the West to understand where it needs to go. They are fitting their cars with higher grades, so then the West understand that they need to get there faster; they know that they can’t stand behind too long. For me, China is just a good way to create sense of urgency.”
On why Chinese lidars ultimately can’t win in the West
“They are only domestic. Up until now, they haven’t managed to succeed outside of China, because of the regulation. Lidar is very sensitive in terms of creation of maps, and so there’s a very high tariff to buy Chinese lidars in the US.
“No OEMs in the West are going use Chinese lidars for their vehicles. Israel is the Switzerland of lidars: we can sell to anyone. We don’t meet the requirements for military uses, we don’t have a problem selling in China and obviously we don’t have a problem selling in the US.”
On the lack of competition outside China
“Out of China, I don’t have real competitors. In China, I will have a challenge to win in the domestic market against competitors like Hesai. But that’s waste of time: domestic OEMs are not willing to pay money; they would go with domestic lidars.”
On the likelihood that Chinese lidar makers are getting state support
“I open the box and their product is more expensive than mine and they’re selling it at a cheaper price. How do they do it? I have no idea. If you talk to a car company, they will say the same thing about the cars they see from China. We don’t understand how they sell it in the prices they sell it. So for sure they have been funded. I don’t know how. But anyway, it doesn’t matter. We are lucky, in a way, that they can’t sell outside of China. For me, it works well.”
On when Innoviz will be profitable
“We make money not only through the sales of lidars. if I only sell lidars, then I need to sell around 200,000, and obviously we’re not there yet. But as a tier-one supplier in the space, I get a lot of money for my development. So we just announced two weeks ago that we closed an NRE [non-recurring engineering] payment plan for around $80 million from an OEM, although the real number is actually closer to $100m; $100m of NRE is more than a year of funding for the company.”
On programs Innoviz is involved in
“We’re currently involved in around three programs. We have the BMW 7 Series. We have two level-three programes with the Volkswagen Group, for Audi and Porsche, and one level four. We’re partnered with [Israeli-American ADAS and autonomous specialist] Mobileye.”
On why he doesn’t worry about Mobileye’s claim that it wants to dump Innoviz lidar for its own imaging radar
“And last year they said they wanted to replace my lidar with their own lidar. You know what happened? They stopped their lidar programme. Of course they will say that; they would prefer to use their own technology. But the requirements that we’re getting for lidars can’t match the power they’re developing. If I were in their shoes, I would say the same.”
On why autonomous driving is just the start for lidar
“Essentially a lidar is a 3D camera. I will give you an application that will convince you that this is going to be the biggest thing in the world. We humans are consuming data from displays all the time, but the first time that we were able to consume 3D data was only through these 3D goggles – augmented reality, virtual reality. But that’s all synthetic right now. Let’s say that now I have such a 3D camera that is recording a football game and you wear your goggles and you can see the game anywhere you want, standing next to the player, in the goal. Or standing next to the singer in the concert. Now you’re convinced it will happen, right? It’s a consumer experience that will happen. 3D cameras will be everywhere.”