The future of autonomous driving leaped forward when NVIDIA introduced DRIVE™ PX2, a deep-learning computer platform for in-vehicle artificial intelligence that solves the daunting challenges of self-driving cars. The DRIVE PX2 incorporates two Tegra® processors and two discrete Pascal graphics processing units (GPUs) to provide 8 teraflops of computing power – the equivalent of 150 MacBook Pros. NVIDIA claims that it can achieve 24 trillion operations a second. Although it’s about the size of a lunchbox, with that much processing power, the NVIDIA DRIVE PX2 must be liquid cooled to prevent overheating.
Jen-Hsun Huang, NVIDIA’s CEO, says it’s the first supercomputer made for cars. Huang says that much computing power is necessary for building capable self-driving cars capable of coping with sudden, unexpected changes in the driving environment.
NVIDIA claims the PX2 can “process the inputs of 12 video cameras, plus lidar, radar and ultrasonic sensors” that allow the system to quickly adapt to vehicles, pedestrians, road debris and signs and to anticipate potential threats for safer navigation.
Deep Learning Requires An NVIDIA Supercomputer
NVIDIA also announced DIGITS, a deep learning platform that it’s already testing with its own self-driving cars. Deep learning will allow self-driving cars to train themselves over time for all sorts of unexpected scenarios. DIGITS helped NVIDIA produce its own deep neural network, dubbed Drivenet. It features nine “inception layers,” — nine separate neural networks embedded with in each other. Running information through the network just once takes 40 billion operations. Drivenet is able to identify five different classes of objects, including pedestrians and motorcyclists. Drivenet also helps PX2 overcome natural obstacles such as poor weather (rain, snow, fog) and extreme lighting conditions.
Despite the fact NVIDIA has its own neural network, carmakers will still build their own deep neural networks, Huang predicts. Audi is using Drivenet to analyze visual data taken from a snowstorm. After just one night they were able to detect data that the human eye can’t see.
Volvo is NVIDIA’S first car company customer. It will purchase several hundred Drive PX 2 supercomputers as part of an effort to develop self-driving cars as part of a company-wide initiative, known as DriveMe, which sets new safety benchmarks. Volvo’s vision is to eliminate death or serious injury in new Volvos by the year 2020.
NVIDIA hopes the Drive PX2 supercomputer will lead to safer driving, new mobility services, and the redesign of urban architecture. “Our vision is to create the computing platform by which we can achieve these goals,” Huang said. “The car is paying attention all of the time. It is a virtual copilot that will keep you out of harm’s way.”