[ad_1]
Developing autonomous vehicles requires a great deal of computing power. To train driving systems, companies simulate different driving scenarios in virtual worlds over and over again: In 2020, Cruise said its systems ran for 200,000 hours a day in simulation on its Hydra platform.
“Simulation is an incredibly important function at Cruise as it accelerates the development, testing and deployment of our autonomous vehicles through specialized tools and frameworks,” states a job listing.
“The simulation and test infrastructure team owns the platforms that facilitate the programming, execution and data flow of the simulation and testing of autonomous vehicles at an extremely high scale.”
Waymo, an autonomous competitor and supposed market leader, also uses large server farms to process simulated trips. For every 22 years of the driver’s experience with the Cruise systems watch, Waymo makes 100 (of course, with different simulations that each will argue is better). As an Alphabet company, its ‘Carcraft’ platform is hosted on Google’s servers, although as we saw with the Verily jump from Alphabet to Azure, that is not guaranteed.
Once autonomous vehicles hit the road, they will need to carry computing power with them. Every vehicle will need powerful sensors and even more powerful computers to quickly process information, understand its surroundings, and avoid crashes.
Cars will also need to communicate with each other, ideally to help optimize traffic and warn of future problems. It is not clear how exactly this will work. Early advocates of autonomous vehicles insisted that everything an autonomous system will need will be in the vehicle itself, with driving data occasionally uploaded for training at a later date.
But last August, Alphabet admitted that building vehicles for today’s world and streets is “more difficult than we think.” They created another company, Cavnue, which seeks to build a highway in Michigan to test another idea: make the highway smarter.
Cavnue hopes to incorporate Edge’s computing and sensors onto the road, offloading some of the work from the vehicles themselves. Among the members of the company’s advisory committee is GM.
