Myriad sensors go into making autonomous cars tick. Drive.ai‘s fleet of Nissan NV200 vans pack lidar (laser-based sensors that measure the distance between themselves and objects) in addition to radar, GPS, and inertial measurement units. Google spinoff Waymo’s Chrysler Pacifica Hybrid Vans are similarly equipped with lidar, radar, and cameras.
But Soroush Salehian and Mina Rezk, the cofounders of Silicon Valley startup Aeva, contend that those and other arrays skip over a critical vector: high-resolution, real-time velocity. That’s why 18 months ago they set out to design a device that, like lidar, could gauge the distance between objects, but one that could also accurately record reflectivity and the distance traveled by those objects per unit time.
“[We set out] to create next-gen sensing systems for autonomous vehicles,” Salehian told VentureBeat. “There has been a lot of focus on making systems with [better] performance and range, but we want to bridge the gap across different types of technologies … [Our sensor] combines the key advantages of lidar, radar, machine vision, and high-accuracy motion sensing.”
Aeva today announced a $45 million funding round led by Lux Capital and Canaan. Notably, the latter was an early investor in PrimeSense, the Tel Aviv-based 3D sensing company that pioneered the technology behind Microsoft’s Kinect sensor and the Face ID sensors in the latest iPhone models. (Apple acquired the company in 2013.)
Rezk, who previously designed optical hardware for Nikon and worked alongside Salehian at Apple’s Special Projects Group (which is rumored to be deeply involved with its ongoing autonomous car project), laid out in a phone conversation the advantages of Aeva’s single-chip sensor design. Conventional lidar, he explained, can only provide indirect velocity data; it rapidly beams individual pulses of laser light, producing frames milliseconds apart that occasionally omit nearby objects. By contrast, Aeva’s compact cuboid of a sensor sends out a continuous wave embedded with a unique signature.
That fundamental difference in technique confers not only a range, power consumption, and resolution advantage — 200 meters, under 100 watts, and down to the centimeters per second — but improved performance in inclement weather and with reflective objects like metal railings. By achieving photon detection of the signal as it bounces off objects, the sensor’s able to perform multiple measurements in the air without risking interference from other laser sensors, and without compromising range.
Perhaps best of all? It drives down the per-sensor price tag to just “a few hundred bucks,” Salehian said. That’s about the going rate San Francisco startup Luminar says it’s targeting with its mechanical lidar system, which has a range of 250 meters. And it’s highly competitive with offerings from lidar incumbent Velodyne, which sells its top shelf, 300-meter range unit for $75,000 and its most popular SKU — the 100-meter VLP-16 — for $4,000.
“[Ours is] very different from a typical lidar,” Salehian said. “It has the core technological advantages of a lidar sensor but also possess the ability to measure velocity instantaneously very accurately … Only Aeva’s sensing system knows what an object is from over 200 meters away and can simultaneously identify if it’s a car or a person, recognize how fast it’s going … and predict what it’s about to do in the following seconds.”
Aeva’s cofounders are nothing if not ambitious, but they have competition in Blackmore, a company prototyping frequency-modulated continuous wave (FCMW) lidar sensors that fire continuous beams of light. Two other startups — Quanergy and Oryx Vision — are exploring similar technology, as is Velodyne.
But Rezk and Salehian are confident their platform won’t be easily replicated.
“One thing that’s very unique about us is that we’ve found a way to work with components and processes and develop tech that we’re able to produce at scale,” Rezk said. “We’re going through a rigorous process … We’re on a daily basis working to improve the product — to shrink the size down and work toward the final design.”
Aeva, which has about 50 employees, has already begun selling its sensors to OEMs and manufacturers. Salehian and Rezk declined to name names, but Lux Capital partner Shahin Farshchi said that early partners include “key” automotive and ride-sharing companies that represent “many millions” of units a year.
“Soroush and his cofounder, Mina Rezk, proposed a new type of sensor that immerses robots in a way that is not possible with conventional sensing technology,” he said. “Aeva has delivered a single sensor that empowers a machine to perceive its surroundings, like humans do, but with far greater fidelity.”