Mobileye Taps AWS for the Lifeblood of AVs: Cloud Computing

A self-driving car is a computing system as much as it is a vehicle, so Mobileye’s selection of Amazon Web Services as its preferred cloud computing partner for autonomous vehicles is an important choice.

Intel’s autonomous vehicle subsidiary announced the decision on November 26 as the AWS re:Invent conference opened in Las Vegas. The company announced that AWS will provide computer storage, database, analytics, machine learning and edge computing resources that will help Mobileye develop self-driving technology more quickly. The companies didn’t attach a dollar figure to the deal, but it could prove quite big within the next several years.

Mobileye notes it’s running core workloads on AWS and will build a data lake on Amazon S3 (Simple Storage Service) to store and analyze hundreds of petabytes of data. This will allow it to run through new iterations of its AV technology in shorter cycles, Mobileye said.

Mobileye’s systems combine cameras and other sensors around vehicles for advanced driving assistance in current cars and full automation in pilot projects. Last month, the Tel Aviv-based company announced a deal with Volkswagen and the Israeli government to put an autonomous ride-hailing service on the road by 2022. Other customers include BMW, General Motors, Honda and Nissan.

A rival to GPU pioneer Nvidia, which provides its Drive platform and powerful onboard computers to many AV companies, Mobileye says its technology allows autonomous cars to drive safely while blending in with human drivers. Its Responsibility-Sensitive Safety (RSS) framework uses a formal definition of safety to keep vehicles in a “safe state” in which they are incapable of causing accidents.

Collecting and analyzing huge amounts of data from sensors on test vehicles is a key part of developing self-driving systems. Data about a car’s environment, driving decisions and outcomes is used to train the artificial intelligence that controls autonomous vehicles. Access to vast cloud resources and analytical tools may provide a big advantage to cloud giants like Alphabet’s Waymo and China-based Internet company Baidu.

AWS offers this type of scalability and agility to companies that don’t have it in house. It already claims AV customers that include Toyota Research Institute and several startups, including Drive.ai, nuTonomy and TuSimple.

In addition to cloud capacity, AWS provides access to deep learning frameworks such as Apache MXNet, TensorFlow and PyTorch and onboard computing and storage with AWS Snowball Edge. Its Greengrass software is designed to let IoT systems in the field, including cars, carry out functions including computing, machine learning, data caching and syncing even if they’re only intermittently connected to the Internet.

TuSimple, a trucking startup developing Level 4 conditional autonomy, noted in an AWS video that its trucks each generate up to 5 terabytes of data per day. Before it started using Snowball Edge, that data used to be collected on onboard hard drives, with a 15-hour daily upload process. In addition, TuSimple noted that AWS computing resources reduced its AI training time from days to hours, allowing it to run 10,000 event simulations in 30 minutes.

Stephen Lawson is a freelance writer based in San Francisco. Follow him on Twitter @sdlawsonmedia.


Leave a comment

Your email address will not be published. Required fields are marked *