A California-based start-up claims to have created a 3D mapping technology that fuses a LIDAR instrument, a camera, and a motion sensor in one to not only identify objects in a scene and their respective velocities, but also their future motion.
Aeva Inc.'s co-founders, Soroush Salehian and Mina Rezk -- who both hail from the Special Projects Group of Apple -- have developed a tiny box that they claim is the future of autonomous mobility according to an insider from The Verge who visited their warehouse last week.
This device can measure the distance and velocity of objects in a scene, virtually in real-time, relative to one another and the device itself, then predict the direction and speed that they will go via artificial intelligence. According to the engineers, however, this is not the innovation.
The key difference of Aeva's invention, Rezk explained to The Verge, is that this 4D LIDAR (Light Detection and Ranging) is continuous rather than based on an algorithm that sends out intervals of light pulses like most other systems. Aeva uses a laser without an algorithm that measures field depth, scene reflectivity, and object velocity simultanously and instantaneously, which is then fed to an AI system to understand what all this data means.
It is important to note that this small box is not the brain of an autonomous vehicle, but rather a better set of eyes -- a device to improve existing LIDARs. It was not created to be a complete self-driving system; those who use this 4D LIDAR will still be required to develop the AI software needed for a fully-autonomous vehicle. The team aims to create improved sensors to provide systems with data.
Aeva claims that the sensors used in this box are better, cheaper, and more efficient than those currently on the market that typically are a patchwork of various three-party elements. Even better, this easily manufactured box can work with virtually any autonomous driving software regardless of the brand.
The pair's goals are tall, but they are accordingly hoping to evolve their tech quickly.