With $8M in New Capital, Vayavision Refines Raw Data Fusion for AVs
The race to get driverless cars on the road is a global endeavor, and one nation putting its pedal to the metal is Israel. A host of Israeli companies are working to develop the technologies that help these vehicles function properly, and one of those companies is called Vayavision.
Based in Tel Aviv, Vayavision is focused on raw data fusion and perception systems. This week, it raised $8 million in a funding round led by Viola Ventures, Mizmaa Ventures, and OurCrowd, with participation from Mitsubishi UFJ Capital and LG Electronics.
The operating systems of autonomous vehicles are essentially designed to function like human brains, and to maneuver, they must be able to “see” their surroundings. Visual outputs come from lidar (light detection and ranging), radar, and camera data generated by the vehicle’s sensors. Vayavision says its patented software fuses the raw data from these perception sensors to create a precise 3D picture of the car’s environment, which allows vehicles to detect lanes, classify and track objects, and recognize traffic and road signs.
“The problem is that no sensor is good enough on its own,” explains Ronny Cohen, Vayavision co-founder and CEO. And if the sensor data is unreliable or inaccurate, it could have deadly results, he adds.
“We gather all the raw data and fuse it together at the pixel level so the car can make optimal decisions,” he says. “We get good results with higher efficiency and less power and bandwidth consumption.”
Plenty of companies are involved in perception software for self-driving vehicles; Cohen cites fellow Israeli company Mobileye as an example of one such company which concentrates on interpreting single-lensed camera data. Cohen says most sensing solutions for AVs are based on “object fusion,” in which each sensor registers an independent object and then must reconcile the data to determine what’s correct. Raw data fusion is needed to reach “more advanced perception paradigms” and achieve an adequate level of safety, he maintains.
Vayavision was launched in 2016 after the company’s co-founder and chief technical officer, Youval Nehmadi, flew from Israel to California for DARPA’s 2007 urban challenge—the same competition in which Waymo got its start, Cohen says. Nehmadi was so inspired by what he saw in California that when he returned home, he started a PhD in sensor fusion for vehicles the following year. In the process of pursuing his PhD, he came up with the intellectual property that underpins Vayavision’s software.
The company is currently running a number of pilots with automakers and Tier 1 suppliers, although Cohen declined to name names. So far, he says, the pilots have returned “strong results,” so Vayavision plans to double the number of tests it does next year. The 25-person company is also contemplating opening offices in Germany and the U.S. to be closer to its potential customers.
Cohen says Vayavision will use its new investment capital to create a “full viable product and integration” as well as bolster business relationships with its pilot partners. He says his company is ahead of the safety curve and its software can be used before full, level 5 autonomy is achieved.
“Realistically, automakers aren’t ready yet” for full autonomy, Cohen says.
“Even Waymo, thought to be the leader, has problems,” he adds. “The industry was facing a lot of excitement initially, but it’s not as simple as it seems—it’s not just about quality, but cost and scalability. Making a system which is cost-effective is crucial, and our ability to work with less processing and power is critical to our overall use case.”