SAN JOSE, Calif. — NVIDIA today announced a number of new releases around the NVIDIA DRIVE autonomous vehicle platform and software suite, as well as a collaboration with Toyota to develop and train self-driving vehicles.
An update to the NVIDIA DRIVE platform gives greater safety to autonomous vehicles (AVs), teaching autonomous vehicles the concept of defensive driving, and giving greater collision avoidance capabilities to AVs.
Off the roads, a new simulation platform for AVs can simulate millions of miles of driving in virtual reality, giving car manufacturers the ability to test vehicle performance without needing to operate massive test fleets for data collection.
Eliminating human error from vehicle operation
The new NVIDIA DRIVE AV Safety Force Field (SFF) adds an increased level of safety to the DRIVE platform by employing defensive driving algorithms to give greater safety to passengers and other vehicles in real-world driving conditions.
The framework is programmed to act on sensor data from the AV to ensure that the actions of the vehicle will never create, escalate, or contribute to an unsafe situation, and can also take action to mitigate potential danger.
“By removing human error from the driving equation, we can prevent the vast majority of collisions and minimize the impact of those that do occur,” said David Nister, vice president of Autonomous Driving Software at NVIDIA. “SFF is mathematically designed such that autonomous vehicles equipped with SFF will, like magnets that repel each other, keep themselves out of harm’s way and not contribute to unsafe situations.”
The National Highway Traffic Safety Administration estimates that approximately 94% of crashes are caused by human error. The SFF, which was tested in city and highway driving conditions using real-world data, looks to greatly reduce the rate of crashes by removing that human element.
SFF is unique is its ability to take into account both braking and steering constraints, and is programmed to follow one core principle of collision avoidance, as opposed to a large set of rules and expectations. The platform can be combined with any self-driving vehicle software to add an increased level of safety and to prevent an AV from taking unsafe actions on the road.
Virtual AV testing now available through NVIDIA DRIVE Constellation
NVIDIA also announced today that its DRIVE Constellation AV simulation platform is now available. The cloud-based platform, introduced at last year’s GPU Technology Conference, can simulate millions of miles of driving by self-driving vehicles across a range of scenarios, from everyday drives to unique and dangerous driving conditions.
The DRIVE Constellation data center is comprised of two side-by-side servers, with the first, the DRIVE Constellation Simulator — using NVIDIA GPUs running DRIVE Sim software to generate the sensor output from the virtual car driving in a virtual world. The second server, the DRIVE Constellation Vehicle, contains the DRIVE AGX Pegasus AI car computer, which processes the simulated sensor data.
The driving decisions from DRIVE Constellation Vehicle are fed back into DRIVE Constellation Simulator, enabling bit-accurate, timing-accurate hardware-in-the-loop testing.
The platform will also help craft self-driving car regulatory standards, given its ability to run far more test cases than could be accomplished with a traditional fleet of self-driving cars. German safety agency TÜV SÜD is already using Constellation to create their self-driving validation and testing standards.
“Together with the support of our ecosystem partners, we’re making available large-scale, cloud-based, open simulation that thoroughly and safely validates self-driving cars under endless challenging situations,” said Zvi Greenstein, general manager at NVIDIA.
Safer autonomous transportation at core of Toyota partnership
Another user of the Constellation simulation platform is Toyota Research Institute Advanced Development (TRI-AD). A new collaboration between NVIDIA and TRI-AD announced earlier today will see Toyota using the DRIVE Constellation platform for components of their simulation workflow, and helping to bring autonomous vehicles to the market.
Previous agreements between the two companies have seen Toyota utilize the NVIDIA DRIVE AGX Xavier AV computer in its development and testing of self-driving cars.
“Our vision is to enable self-driving vehicles with the ultimate goal of reducing fatalities to zero, enabling smoother transportation, and providing mobility for all,” said Dr. James Kuffner, CEO of TRI-AD. “We believe large-scale simulation tools for software validation and testing are critical for automated driving systems.”
The new agreement also includes the development of an architecture for AVs that can be used across a variety of vehicle models, helping to accelerate the process of bringing self-driving vehicles to market, while also simulating billions of miles of driving in a wide variety of scenarios and conditions. In addition, Toyota will use in-car AV computers based on DRIVE AGX Xavier or DRIVE AGX Pegasus hardware.