Mobileye's contribution to the self-driving-future:
Mobileye is a company that develops advanced driver assistance systems based on vision. With this technology Mobileye provides cars with warnings for collision prevention. The company was founded in Jerusalem by Amnon Shashua. Today Mobileye's purpose is translated into developing the perfect 'eye' for a self-driving car. Not for a Mobileye self-driving car. Shahsua does not want to create its own vehicle brand. Instead he opts to sell his knowledge. Over the years Mobileye's driver assistance systems has been bought by Tesla, GM, Audi, BMW, Volvo, Volkswagen, Nissan, Renault, etc.
Except for the anti-collision-software some car producers (for example General Motors and Volkswagen) are also acquiring Mobileye's road data and digital maps that are (and have been) gathered in road testing. Together they call it the Road Experience Management mapping system. Also in this map-selling business Mobileye is leading the pack, but foresees competition from a variety of sectors that have access to big amounts of geographical data (for example IBM, GE, Nvidia or Here, the former Nokia mapping division, which is owned by Audi, BMW and Daimler).
'Everywhere' vs 'Somewhere'
This title refers to the difference in research and development techniques between Mobileye and Google. Mobileye, and most of the auto industry, are in the 'everywhere'-camp. This camp wants to create a driverless future enabling semi-automatic features tot a man driven car, testing, learning and mapping 'everywhere' to improve the software's artificial intelligence. Google on the other hand does its testing with fully autonomous vehicles in specific locations (= somewhere; for example Mountain View, California), driving the same roads over and over ,while perfecting their maps and the AI.
Mobileye's approach, according to CEO Shashua, will mature much faster. "Our Mobileye self-driving car tests are held everywhere, which makes the step from semi-autonomous driving without maps, to fully driverless with real-time data, crowdsources maps and a strong artificial intelligence, a gap we can bridge in 3 to 5 years." Shashua also sees big challenges in Google's approach of scaling up very detailed geographical data into maps that could hold up all over the globe. Google's comeback was that its IP portfolio is much broader than Mobileye's and that is not interested in semi-autonomous driving tech.
What does the future hold for Mobileye?
The next comparison Shashua makes with Google is the price tag. The Lidar-technology used by Google is very expensive. One of Google's prototype main sensor costs around $80,000. In comparison, Shashua predicts that automakers, using Mobileye's technology, will be able to produce a driverless car as soon as 2021, with techs and sensors costing $1,000 or less. So Mobileye predicts that their Advanced Driver System will beat Lidar and Mobileye is by far the biggest ADAS software supplier on the market. On top of that Mobileye has become partner of the U.S. Department of Transportation in improving bus safety through project Shield+ (see video below). Not a bad starting point.
But there is still work. Among the challenges is the ability for cameras to see in low-light conditions and the big competition from other mapping providers. But Mobileye's crowd-sourced mapping project, together with GM, would be able to bypass that. With this new technique they would for example be able to update their central database on highway markings. While driving, autonomous cars could cross-check what their cameras are seeing against what other cars have already seen before. According to Mobileye, this Road Experience Management system can process road details at 10-kilobytes per kilometer. This will drastically reduce the load on the vehicle’s data, and will provide specific geographical data at an accuracy of about 10 cm, far more precise than today’s GPS.
Mobileye is the top supplier of camera-based sensor systems used in driver assistance systems. Together with STMicroelectronics, Mobileye will be releasing in 5th series of their EyeQ sensor fusion technology. Release is expected somewhere in 2018. The system on a chip (SoC) will collect and fuse the broad range of data being generated by up to 20 different sensors on a vehicle such as cameras, speed and motion sensors and radar. And making sure all possible outcomes are covered, Mobileye announced to even include connections with Lidar. There also will be an SDK so automakers will be able to differentiate their offerings.
The EyeQ5 will be designed to serve as the central processor for future fully-autonomous driving and goes into direct competition with Nvidia's Drive PX or NXP's BlueBox. In recent months, BMW, Volkswagen, General Motors and Nissan agreed to use EyeQ5 systems in future car fleets.
Last updated: 02/08/16