The EX90’s core NVIDIA’s DRIVE Orin system-on-a-chip (SoC) is the fundamental computing solution that orchestrates everything in the car: from driving deep learning for Volvo’s AI-based active safety and driver assistance systems to generally helping to introduce safe autonomous driving in the future.
But the use of NVIDIA technology does not stop there. In the next development stage, Volvo will intergrate NVIDIA Blackwell GPU architecture. This upcoming step – planned to take place within the next few years – and the integration of this architecture means that Volvo Cars will introduce vehicles built on NVIDIA DRIVE Thor, which can handle up to 1,000 TOP. This is four times as many operations per second as a DRIVE Orin SoC, while offering seven times higher energy efficiency.
This Thor integration, the company says in a press release, “will help to further future-proof our next generation of cars.”
One of the points of DRIVE Thor, and the integration of the NVIDIA Blackwell GPU architecture, will enable Volvo Cars to deploy even more advanced driver assistance and safety features, develop autonomous driving and introduce generative AI-based features and experiences in the car.
Light-Weight Files and Real Time Rendering
That said, it can be noted that a major advantage of the Gaussian splatting technique is that it creates lightweight files and enables real-time rendering, which provides greater flexibility for interactive applications. It has been argued that photogrammetry alternatives requires longer time frames and produces larger file sizes, unlike the faster and more efficient Gaussian splatting. The latter allows for direct rendering of volume data without converting them to surface or line primitives.
In the case of CAD systems, geographic information systems (GIS) and vector computer graphics, a geometric primitive is the simplest geometric shape that the system can handle. The simplest primitives are points and straight line segments. However, with the splatting technique, which was originally introduced by Lee Westover in the early 1990s, this computational technique has been pushed much further and can now be of great use in the development of driver assistance solutions.
What makes the Blackwell architecture so powerful?
But as indicated above, the Blackwell GPU architecture opens up new chapters in a number of areas that require speed and the ability to handle large volumes of data. It includes several revolutionary accelerated computing technologies, catalyzing advances in computing in more industries than just automotive.
Jensen Huang, CEO of NVIDIA, calls it a monumental breakthrough:
“Absolutely, for the last thirty years we have been dedicated to developing accelerated computing to drive monumental breakthroughs like deep learning and AI.”
He highlights three key advances in particular:
- Generative AI: It facilitates real-time applications that leverage massive language models with trillions of parameters.
- Exascale Computing: Equips users to tackle some of the most complex challenges that require exascale computing power, defined as achieving at least one quintillion calculations per second.
- Security: Introduces robust security features designed to protect confidential AI training and inference processes.
Virtual and real-world testing
In general, Volvo Cars uses virtual environments along with real-world testing for software training, development, and validation because they are secure, scalable, and cost-effective. The virtual environments are developed in-house in collaboration with Zenseact, an AI and software company founded by Volvo Cars.
The AI project is part of a doctoral program for leading Swedish universities to explore whether neural rendering techniques will be integrated into future security initiatives. The study is sponsored by Wallenberg AI, Autonomous Systems, and Software (WASP).
Smart use of data
Volvo Cars has a long history of using data and advanced technology to improve safety. Data collected by Volvo Cars’ safety research team has played a crucial role in the development and testing of some of the world’s most prominent safety features.
In the 1970s, Volvo Cars began using data to improve safety through its safety research team. Initially, the team arrived at the scene of an accident with tape measures, skid marks and other crash indicators. The data and knowledge gathered from accidents has inspired many life-saving innovations, such as the Whiplash Injury Protection System and the Side Impact Protection System.
New advanced technology is now enabling the company to be even smarter about how it uses data to prevent risky situations.
