🎙️ Voice is AI-generated. Inconsistencies may occur.

2 min read

Volvo Using AI, Your SUV's Data to Create a Virtual World to Test Safety

A combination approach to studying safety is part of Volvo's overall initiative to have its vehicles involved in zero auto accidents.

  • Español
  • 中国人
  • Français
  • Deutsch
  • Portuguese
  • हिन्दी

An artificial intelligence-generated virtual world is serving as a stand-in for the real world as Volvo works to make its cars as safe as possible.

The Swedish automaker is testing advanced driver assistance systems (ADAS) as part of its quest to eliminate all accidents. Volvo's long-held zero-accident goal aims to prevent as many of the World Health Organization-estimated 1.19 million annual traffic incident-related deaths globally as possible.

The virtual-world project is a result of partnerships between Volvo Cars; Zenseact, an artificial intelligence and software company founded by Volvo Cars; and the Wallenberg AI, Autonomous Systems and Software Program (WASP), which is the largest individual research program in Sweden.

Using sensors on Volvo vehicles, the automaker can collect and analyze incident data, a process enabled by connected car technology. Among the behaviors that send information to Volvo are emergency braking incidents, sharp steering and manual intervention when a vehicle's ADAS system is working.

In a press release, Volvo said that the company uses the data to reconstruct and explore the incidents to "better understand how incidents can be avoided" using Gaussian splatting, an advanced data-rendering technique that creates high-fidelity three-dimensional scenes and objects using real-world visuals for reference.

Gaussian splatting Volvo
An example of how Gaussian splatting depicts a person in a roadway. Volvo Cars

Technicians can manipulate the virtual environment. They are able to change variables in a scenario by adding/deleting roads, changing traffic behaviors and removing/inserting obstacles—all of which can generate different outcomes—for example.

The manipulation can also create extreme hazards to explore the behavior of vehicles in complex accidents that are infrequent in the real world, yet dangerous and part of what Volvo must plan for if it is to eliminate all accidents its vehicles are involved in.

Gaussian splatting allows these manipulations to occur faster than previous versions of the research technology allowed, moving the timeline from months to days. It's possible because of a recently expanded partnership between Volvo Cars and NVIDIA, which enabled the new EX90 battery-electric SUV and forthcoming ES90 battery-electric sedan to be software-defined vehicles that have software and hardware connected to the internet, updatable on the fly via over-the-air technology.

"Data holds the key to improving automotive safety. Thanks to Gaussian splatting, we can now quickly multiply the millions of datapoints we have, turning real-world sensor sequences into thousands of variations of edge cases. We believe this could be the mileage multiplier that we've been looking for to accelerate how quickly we reach our ambition for a zero-collision future," Alwin Bakkenes, head of global software engineering at Volvo Cars, told Newsweek.

Gaussian splatting Volvo
Gaussian splatting combined with other safety-analyzing technology is shown identifying roadside hazards. Volvo Cars

The vehicles' new supercomputing platform, powered by NVIDIA DGX systems, is able to contextualize data that Volvo says will lead to improving and accelerating the development of artificial intelligence.

It is possible because of the establishment of one of the largest data centers in the Nordics, which was jointly set up by Zenseact and Volvo Cars.

The virtual environment is only one piece of the puzzle. Volvo is combining its digital efforts with real-world vehicle software training, development and validation.

The Latest in AI

EXCLUSIVE
05.29.25
HEALTH
05.01.25
MOBILITY
04.29.25
MOBILITY
04.22.25
HEALTH
04.23.25
See all Stories