Skip to main content

Nvidia’s new simulator brings virtual learning to autonomous vehicle developers

Nvidia’s Drive Constellation autonomous vehicle simulation system is ready for manufacturers to use to rack up millions of virtual miles testing and proving their technologies.

Nvidia announced the availability of its data center simulator for testing self-driving automotive technologies at the company’s annual GPU Technology Conference in San Jose, California. Designed to speed up autonomous driving development, Drive Constellation is an open, cloud-based platform technology vendors can use to validate systems efficiently, safely, and much faster than with vehicles on real roads.

Drive Constellation is an answer to the conundrum that has plagued autonomous vehicle technology: How can you test vehicles in real-world driving situations to prove they’re safe before they’ve been validated and certified as safe?

Self-driving car technology companies have logged millions of miles on public roads under restricted conditions. However, the public, automotive manufacturers, and government regulatory agencies are all leery of the dangers of testing unfinished, unproven technology where people can be injured or killed. Higher order statistics consistently show human error is responsible for more than 90 percent of fatal accidents, but allowing un-piloted or partially piloted vehicles on public roadways for testing is still fraught with its own difficulties.

Public road testing is also limited and slow, compared to the possibilities of simulated testing. For example, if you wanted to know how a car’s autonomous system would work if a mattress flew off the back of a pickup truck the car was behind on the highway traveling 70 miles per hour, setting up a real-world test is out of the question. The same test on a simulator would be relatively easy, and no lives or property would be at risk.

Nvidia introduced Drive Constellation at the GPU Tech conference last year as the solution to the difficult problem of testing autonomous vehicle technologies.

The Nvidia platform uses two servers that work together: Drive Constellation Simulator and Drive Constellation Vehicle. The Simulator server utilizes Nvidia graphics processing units (GPUs) and Drive Sim software to generate output from a virtual car driving on virtual roads. The Vehicle server processes the data from the Simulator using Nvidia’s DRIVE AGX Pegasus A.I. car computer. When the Vehicle makes driving decisions during simulation, the decision data is fed back to the Simulator immediately for continuous processing.

The Drive Constellation platform is ready for manufacturers who want to work with Nvidia to validate autonomous vehicle technologies.

Nvidia announced Toyota Research Institute-Advanced Development (TRI-AD) is the first Drive Constellation customer. “We believe large-scale simulation tools for software validation and testing are critical for automated driving systems,” said TRI-AD CEO James Kuffner.

Editors' Recommendations