In an era where technological advancements are reshaping every aspect of our lives, the automotive industry is embracing cutting-edge solutions to enhance safety on the roads. Volvo Cars, in collaboration with Nvidia, is at the forefront of this revolution, leveraging AI-generated, life-like 3D environments to accelerate the development of its safety software. This breakthrough is allowing the company to identify potential hazards and refine its vehicle safety systems at an unprecedented speed and scale.
Traditionally, vehicle safety testing relied heavily on real-world scenarios and physical testing. However, with the advent of virtual environments and advanced computing, the development of safety features has entered a new dimension. Volvo’s use of AI-generated virtual worlds is a key example of how this technology can revolutionise the way we approach safety, making the testing process more efficient and far-reaching than ever before.
At the heart of this innovation lies a technique known as Gaussian splatting. This advanced computational method enables the creation of realistic 3D scenes using tiny data points, or “splats,” to represent objects in a scene. Unlike traditional mesh-based 3D models, Gaussian splatting fills a virtual space with these points, which each carry properties such as position, size, orientation, and transparency. When combined with real-world incident data collected by Volvo’s advanced sensors, this technique allows the creation of virtual traffic scenarios that are not only realistic but can be manipulated to test how vehicles react in a wide variety of situations.
Alwin Bakkenes, Head of Global Software Engineering at Volvo Cars, explains: “We already have millions of data points of moments that never happened that we use to develop our software. Thanks to Gaussian splatting, we can select one of the rare corner cases and explode it into thousands of new variations of the scenario to train and validate our models against. This has the potential to unlock a scale that we’ve never had before and even to catch edge cases before they happen in the real world.”
These virtual environments enable Volvo to expose its safety software to a vast array of traffic situations, including edge cases that might be rare but are potentially dangerous, such as animals crossing the road, objects falling off vehicles, or drivers engaging in erratic behaviour. By recreating these complex scenarios in a virtual setting, Volvo can simulate and test how its vehicles would react without the need for extensive physical testing, which can be costly, time-consuming, and impractical.
The ability to simulate such a wide range of potential scenarios in real-time offers a level of flexibility that traditional testing methods cannot match. These scenarios can be adjusted on the fly, allowing engineers to test the software under different conditions—whether that means adding or removing road users, altering traffic behaviour, or introducing new obstacles. In doing so, Volvo can ensure that its advanced driver assistance systems (ADAS) are capable of responding to the unpredictable nature of real-world driving.
Moreover, the integration of Nvidia’s AI supercomputing platform enhances the process even further. Powered by Nvidia’s DGX systems, this platform provides the computational power needed to process vast amounts of data and generate insights that drive the development of safer vehicles. This collaboration enables Volvo to gather and contextualise data from its cars’ onboard sensors, which include emergency braking, sharp steering, and manual interventions, to create a more complete picture of what happens during potential accidents.
As a result, Volvo can develop and refine its software with far greater efficiency. What once took months to simulate and test can now be accomplished in a matter of days. The company’s safety software is tested against a wide array of scenarios, ensuring that it performs well not only under normal conditions but also in rare, high-risk situations.
Zenseact, Volvo’s in-house AI and software company, plays a pivotal role in this initiative. Founded in 2017 as a joint venture between Volvo and Autoliv, Zenseact specialises in developing AI-driven technologies for autonomous driving and advanced safety systems. Through Zenseact, Volvo has been able to push the boundaries of AI and software development, creating safety innovations that will reduce the risk of accidents and ultimately save lives.
The scale of this project is massive. Volvo’s commitment to using virtual worlds for safety testing is part of a broader effort to revolutionise how cars are built and tested. This innovative approach is just one aspect of Volvo’s long-term investment in autonomous driving and AI-powered vehicle safety. The company’s new electric vehicles, including the EX90, are built on Nvidia’s DRIVE Thor system, a centralised supercomputer that orchestrates every aspect of the car, from driving assistance to autonomous driving.
“We are investing in a future where cars are not just smarter, but also safer,” says Jensen Huang, CEO of Nvidia. “With our DRIVE Thor system, Volvo is able to create software-defined cars that deliver a continuous stream of safety updates and improvements, ensuring that vehicles are always getting safer over time.”
Through this partnership, Volvo and Nvidia are not just improving safety today—they are paving the way for the next generation of vehicles, which will be able to anticipate and respond to hazards in real time, creating a future where the roads are safer for everyone.
