A research team from the Friedrich-Alexander University Erlangen-Nuremberg (FAU) is working on significantly improving the perception capabilities of autonomous vehicles. In the "BAVAR-RADAR" project, researchers rely on Artificial Intelligence (AI) and novel datasets to deliberately overcome the previously limited resolution of radar sensors.

Radar sensors are considered robust and operate reliably even in rain, fog, or darkness. However, they often only provide rough outlines of the surroundings. For automated driving, this is a significant disadvantage, as it is difficult to distinguish objects and movements cannot be captured precisely—particularly in critical traffic situations.

More precise environmental perception and increased traffic safety thanks to digital twin

The team led by Prof. Dr. Vasileios Belagiannis addresses this challenge with an innovative approach: combining real measurement data with synthetic data from so-called digital twins. These virtual representations of real traffic scenes allow targeted expansion of training data and realistic simulation.

On this basis, researchers train neural networks that can reconstruct detailed three-dimensional point clouds from the comparatively sparse radar signals. Generative AI models help to consolidate the data and balance discrepancies between simulated and real measurements.

The project's goal is to make existing radar systems significantly more powerful—without additional hardware. As a result, driver assistance systems and autonomous vehicles could recognize their surroundings more accurately and react faster, for example, when children suddenly appear from behind parked cars.

By 2028, the technology is set to be demonstrated in a test vehicle. In addition to FAU, the University of Applied Sciences Hof and industry partner Valeo, which provides the test vehicle and real sensor data, are also involved.

The project demonstrates how AI-based signal processing can enhance road traffic safety and underscores the potential of data-driven innovations for the mobility of the future.