Five AI, the Bosch Mobility subsidiary, is developing new simulation technology intended to speed up and reduce the cost of autonomous vehicle development, supported by UK government funding under the CAM Pathfinder program. The approach combines Gaussian scene reconstruction with generative AI to build photorealistic 3D digital environments from real-world imagery, offering a level of scalable realism that has been difficult to achieve with conventional simulation methods.
Diffusion-based AI models will be used to alter these reconstructed environments, generating variations in lighting, weather and other conditions. This is intended to let engineers test AV systems against a far broader set of scenarios than could practically be captured through real-world data collection alone.
The tool is being built to support development of Level 2++ and Level 3 automated driving functions, and will operate alongside other Bosch Mobility simulation components as part of a wider platform. That platform is intended to reconstruct photorealistic digital twins for scenario creation and evaluation of vehicle responses, with the stated ambition of eventually supporting training and validation of Level 4 systems.
Dave Buffey, head of Bosch Mobility UK, said the software represents a significant innovation effort by Five AI's team, with Bosch working closely with the company to address the technical challenges automakers face in integrating autonomous driving technology into production vehicles. Five AI has supplied simulation software to automotive manufacturers since 2015.
Source: Automotive Testing Technology International — read the original article