Autonomous vehicle development is a time and resource-intensive business, requiring dozens of test vehicles, thousands of hours of data collection and millions of miles of driving to hone the artificial brains of the cars of tomorrow. What if you could do most of that in the cloud? That's the question Nvidia hopes to answer with the release of its Nvidia Drive Constellation testing platform for self-driving cars. The announcement came during the keynote address at Nvidia's 2019 GPU Technology Conference in San Jose Monday.
Drive Constellation is, basically, a simulation and validation platform that allows automakers and developers to test their autonomous vehicles and technologies in a virtual environment that lives in a specially-designed cloud server. Constellation works by first generating an extremely realistic environmental simulation using Nvidia's RTX-powered servers, then feeding that simulation into the autonomous vehicle software or platform requiring evaluation. The vehicle can then be tested in the virtual road rather than on a public road. It's like plugging your self-driving car's brain into The Matrix, where it can be tested, trained and developed over and over again.
The benefit of simulated testing is that it is customizable, repeatable and controllable. Devs can test exact lighting, weather conditions, traffic and pedestrian behaviors over and over, or they can test dozens of iterations of those conditions at the same time. Rather than having a fleet of dozens or hundreds of autonomous cars testing on the road, Drive Constellation will allow a developer to test thousands in the cloud before hitting the road IRL.
Constellation is able to render fully customizable environments for virtual autonomous vehicle testing with amazing fidelity.
On stage at the 2019 Nvidia GPU Technology Conference keynote, Nvidia CEO Jensen Huang presented a strikingly realistic demonstration of what Constellation might look like to an autonomous vehicle developer, changing the weather, time of day and traffic conditions for the software simulation in real-time. The demonstration was capped off by showcasing more than a dozen random simulations at the same time.
According to Nvidia, Constellation is also unique in that it is open-source, which allows developers to bring their own bespoke sensor data, vehicle specifications and testing conditions to the cloud simulation, saving development time by testing both the hardware and software simultaneously and increasing the accuracy of the testing since almost no variables change once the test is completed and the car hits the real roads.
Huang announced that Nvidia Drive platform and Drive Constellation tech has been fully integrated into the development workflow for its first automotive partner, Toyota -- specifically the Toyota Research Institute Advanced Development (TRI-AD) autonomous vehicle team.
"Toyota, the world's largest car company, is partnering with us from end-to-end, from deep learning systems to simulation systems to in-car computers to collaboration for AI," stated Huang at the keynote address. "Toyota and the TRI-Advanced Development team are partnering with Nvidia to create the future of autonomous vehicles."
Nvidia also announced that it is opening the Nvidia Drive Constellation platforms to more partners. Among the first in line to use the technology for testing and validation are autonomous platform developers Cognata and IPG and hardware providers ON Semiconductor and Velodyne.
Nvidia Drive Constellation technologies will also factor in the newly-announced partnership with German certification body TUV SUD, serving as the basis for developing autonomous vehicle testing and licensing standards in the future.