Nvidia Continues 'Driving' Autonomous Vechicles
March 21, 2019

Nvidia Continues 'Driving' Autonomous Vechicles

Nvidia’s technology drives many industries, from M&E to research and science. One sector that is reaping large benefits is the automotive arena.

At GTC 2019, Nvidia made a handful of announcements as it pertains to the automotive industry, including the availability of the Nvidia DRIVE Constellation autonomous vehicle simulation platform.

The cloud-based, open, scalable simulation platform enables large virtual fleets of self-driving cars to log millions of miles across virtual worlds that incorporate a wide range of driving scenarios, from the routine to the downright dangerous.

Constellation consists of two side-by-side servers. One, the DRIVE Constellation Simulator, uses Nvidia GPUs running DRIVE Sim software to generate the sensor output from the virtual car driving in a virtual world. The other server, the DRIVE Constellation Vehicle, contains the DRIVE AGX Pegasus AI car computer, which processes the simulated sensor data. The driving decisions from the Constellation Vehicle are fed back into the Constellation Simulator, enabling “bit-accurate and timing-accurate hardware-in-the-loop testing.”

More and more manufacturers are exploring the use of autonomous vehicles, increasing the need for testing and safety standards. Using simulation to road test these vehicles is certainly safer, more efficient, and more cost-effective than doing so in the real world. For instance, scenarios calling for rainy conditions would have to wait for Mother Nature to cooperate in the real world, whereas they can be set up instantly in a simulation.

In other related news, Nvidia introduced DRIVE AV Safety Force Field (SFF), a computational defensive driving software suite that shields autonomous vehicles from collisions. The decision algorithms protect against the unpredictability of real-world traffic for a safe and comfortable driving experience. 

The SFF analyzes and predicts the dynamics of the surrounding environment by collecting sensor data and determining a set of actions to protect the vehicle and other road users. It runs on the Nvidia DRIVE platform and is an open platform that can be used with any driving software.

In related news, Nvidia announced that it has partnered with Toyota Research Institute-Advanced Development (TRI-AD) to develop, train, and validate self-driving vehicles. The partnership expands on the companies’ ongoing relationship to use the Nvidia DRIVE AGX Xavier AV computer, and extends to AI computing infrastructure using Nvidia GPUs, simulation using the DRIVE Constellation platform, and in-car AV computers based on the DRIVE AGX Xavier or DRIVE AGX Pegasus. According to Nvidia, the agreement in part includes development of an architecture that can be used across many vehicle models and types.