NVIDIA VR + AI = Billions Of Miles Of Virtual Driving

NVIDIA announced today that it has combined its historical strength in high-performance 3D graphics with its Artificial Intelligence (AI) technologies to create a cloud-based platform for simulating and testing the driving operations of autonomous vehicles. By enabling developers to virtualize the closed-loop system of sensors, image processing, routing, and driving controls, the company’s new DRIVE Constellation is helping its customers get safer vehicles to market, faster and cheaper.

What did NVIDIA announce?

The NVIDIA DRIVE Constellation is a cloud-computing platform that uses two different GPU-equipped servers and software to simulate the operations of an autonomous vehicle. As the recent fatal accident with an Uber smart vehicle has demonstrated, autonomous driving is still far from ready for mass production and adoption. The variability of millions of roads, driving conditions, weather, traffic, and unexpected events lead to many corner cases that can create hazards for people and liabilities for auto manufacturers. Many auto companies including Toyota, Mercedes-Benz , Uber, Audi , VW, and Tesla are using NVIDIA’s DRIVE computational platform to turn images from LIDAR, RADAR, and video into navigation, driving decisions, and control. It makes sense that NVIDIA would extend its graphics and AI technologies to further accelerate and improve the testing process.

One of the two servers that make up the platform runs the NVIDIA DRIVE Sim software, using NVIDIA GPUs to simulate the output of the image sensors and provide photo-realistic visualization of virtual vehicle travel. In fact, it can drive down billions of miles of roads in a variety of weather, road, and lighting conditions in a fraction of the time and cost a physical testing system would require—all in the safety of a datacenter. Other potentially dangerous conditions can be scripted to test the virtual vehicle’s ability to respond to an event, such as a pedestrian or animal suddenly crossing the road. The simulated sensor output is constantly fed to the second server, which contains a DRIVE Pegasus autonomic driving platform that uses AI to “act” on the sensor input. The resulting driving decisions are fed back to the DRIVE Sim server to complete the closed-loop system.

Conclusions

NVIDIA is now delivering both the on-board supercomputers required for real-time autonomous vehicle control as well as a virtual testing platform to reduce costs and improve safety. While its competitors invest hundreds of millions into basic AI silicon trying to catch up, NVIDIA keeps extending its lead. Furthermore, the newly announced DRIVE Constellation demonstrates that the company is delivering complete platforms, not just silicon devices, and its close collaboration with leading automakers and universities virtually assures that the company will remain a step ahead in the exciting race to full level 5 autonomous driving.

Disclosure: Moor Insights & Strategy, like all research and analyst firms, provides or has provided research, analysis, advising and/or consulting to many high-tech companies in the industry, including NVIDIA and many others. The author does not have any investment positions in any of the companies named in this article.