ATLANTA — When you want to test a pilot without having to fly, you put them in a flight simulator. When you want to test an autonomous vehicle without having to drive a mile of asphalt, you give it a digital world to navigate through.
Unlike a vision-based flight simulator, AV simulation means the autonomous driving platform is subject to a steady stream of digital data that mimics, exactly, the data stream it would receive operating in the real world.
Simulation is essential to building enough autonomous driving data to fully prepare autonomous vehicles to operate efficiently and safely, said Norman marks, global director, automotive industry, at Nvidia.
He noted that billions of miles need to be driven by test vehicles in the real world to create an accurate model. “Hence, there is a critical need for computer simulation to ensure deep neural networks are applied and tested against all global driving conditions,” he said
Speaking at the automotiveIT International conference here, Marks offered insights into the powerful simulation technologies Nvidia is developing specifically for the automotive industry.
In the future, Marks said, Nvidia technology could also be used to system check new AVs when they come off the assembly line, making sure the cars’ autonomous driving platforms are roadworthy before being shipped to customers.
-By Jason Booth
(Editor’s note: A longer version of this story will appear in the summer issue of automotiveIT International magazine. For a free subscription, please go to www.automotiveIT.com/subscribe)