Arbe Robotics raises $9M to build high-resolution radars for autonomous cars

As carmakers and tech companies continue to improve the mechanics and reliability of self-driving cars, Arbe Robotics, a Tel Aviv startup developing a high-resolution radar system to help those vehicles detect and identify objects, has raised $9 million to help see itself to its first in-car trials in the next year or two. There are five of these trials in the works right now, founder and CEO Kobi Marenko tells us, all in the US with large partners.

(I asked multiple times, and Marenko would not give me any more clues as to who they are beyond what I have detailed here.)

O.G. Tech Ventures, a new VC backed by Eyal Ofer, the Israeli shipping and real estate magnate; and OurCrowd, the equity crowdfunding platform led the round, with previous investors Canaan Partners, iAngels, and Taya Ventures also participating. This brings the total raised by Arbe to $12 million.

Considering how strongly linked venture funding is to developing engineering-intensive AI and sensor startups, this funding puts Arbe somewhere squarely in the middle of the field of others working in the same space as it is.

Echodyne — which today mainly seems to focus on drones but is also eyeing up road vehicles — is now at $44 million raised. Autoliv is publicly traded with a market cap of over $10 billion and has also recently moved into the market. YC-incubated Zendar, on the other hand, has only disclosed around $4.3 million in funding to date. And Uhnder describes itself as “fully funded” but has been very under the radar (heh) when it comes to any more details about what that entails.

Marenko describes Arbe as capital efficient and said that the company had actually wanted to raise about $2 million less than it had. “There is a push right now to take as much money as you can, but we try to take a little as we need to make it to the next station,” he said. “With this $9 million, we are able to finish the product, to do more tests, to build a support center in the U.S. and to get to the next station.” He’s also prepared to raise more when the time comes, and if that coincides with actual commercial deals beyond trials, it will be at a much better valuation for the company.

If Arbe sounds familiar to you from another context, regular TechCrunch readers might recall that Arbe won our pitch-off event in Tel Aviv in 2016. At the time, the company was focused more on drones than self-driving cars. In the period between then and now, it has pivoted its focus somewhat in response to what it can see as a more immediate market opportunity in road vehicles. “The main focus right now is automotive,” Markenko said. “It’s been taking us on a long ride, and we can’t fly in between the two.” He is also a fan of puns and metaphors, it seems.

Up to now, there has been a lot of focus on the role that LIDAR can play in sensors in autonomous or partially-autonomous cars. This laser-light-based system can created highly detailed pictures of objects or landscapes based on the variations in return time that you get for a beamed light signal to return back to a sensor, but the catch is that you need the right light and weather conditions for this to work at its best.

As Marenko describes it, Arbe’s system can sit alongside that to provide a complementary level of information using high-resolution radar.

The key with Arbe’s system is that the company has developed intelligent software and a radar that work regardless of exterior conditions, to provide the most accurate results not just in terms of seeing objects but being able to identify what they actually are. The two main components, both developed by Arbe itself, are an ultra high resolution radar; and Simultaneous Localization and Mapping (SLAM), algorithms that provide improved target separation and enhanced resolution, and remove false alarms that confuse systems and lead to malfunctions and accidents, with work in a 4D imaging system.

Marenko claims that the product is able to distinguish objects with a one-degree resolution between them (for an easy-ish explanation of what that means, see here: the basic idea is that if there are two objects close together, Arbe’s system can detect that there are two objects, and what they are).

The company’s product, demonstrated in action in September to its customers, “shows the actual resolution … and our ability to distinguish between people, or a tree, or a motorcycle,” Marenko said.

One big challenge for Arbe and others in its field will be to see how well they can scale what they are building: the name of the game will be unit costs and the ability to mass produce their technology.

“The main question that we’re getting is the ability to bring this into mass production,” he said. Arbe is now in the “final stages” of securing a manufacturing partner, not as an investor but as someone to build its chips.

Here’s a video we made with Arbe earlier this year as it was gearing up to show off its tech to customers in its first live deployments: