Automotive AI

Volvo uses Generative AI to elevate vehicle safety

News, 17 August 2025

Volvo Cars has taken a leap forward by employing generative AI to craft virtual worlds—advancing the development and testing of its safety software, such as ADAS (Advanced Driver Assistance Systems). By synthesizing real-world sensor data into thousands of high-fidelity scenarios, the automaker can now expose its safety systems to edge cases at unprecedented speed and scale, transforming months of testing into mere days. 

Volvo has long been synonymous with safety, and it’s pushing the envelope further by integrating AI-generated virtual environments into its safety software workflow. Using a technique called Gaussian splatting—which converts real-world imagery into highly realistic 3D scenes—Volvo can generate and manipulate thousands of unique scenarios. These include rare but hazardous situations like erratic traffic, pedestrian anomalies, or unexpected obstacles. 

This virtual capability lets Volvo expose its safety systems, including ADAS, to uncommon “edge cases” wherein software typically fails. With this, the development cycle accelerates dramatically—from months to mere days. These virtual environments are developed in-house in collaboration with Zenseact, Volvo’s AI-focused software division, and are part of a PhD-led research initiative supported by the Wallenberg AI, Autonomous Systems and Software Program (WASP)

Moreover, Volvo’s approach is supercharged with the support of NVIDIA’s AI supercomputing platform. Sensor-rich data from vehicles are processed on powerful systems like NVIDIA DGX, enabling training and validation at scale. This infrastructure enhances Volvo’s ability to build, test, and evolve active safety systems more efficiently. 

SEO Keywords

Volvo generative AI safety, Gaussian splatting Volvo, virtual testing ADAS, Volvo Zenseact AI, automotive safety innovation, AI in vehicle testing

References

Back to top button