Focus area I: Perception

Reliable perception of the vehicle environment is essential for cooperative automated driving, as it underpins all downstream functions such as prediction and decision-making. Failures in perception can lead to severe safety risks, as demonstrated by early fatal accidents in automated driving. Despite progress, perception remains a major challenge.

The goal of the focus area Perception is to derive an accurate representation of the current state of the environment surrounding the automated vehicle under consideration (EGO vehicle) from sensor data, thereby reducing resulting uncertainties and quantifying them for further processing in the focus area Prediction. In doing so, Beyond Validation AI focuses on two areas that offer great innovation potential for reliable perception: (1) online monitoring and reduction of uncertainties in AI-based raw sensor data processing; (2) consideration of V2X communication as an additional source of information to reduce uncertainties, which in turn, however, causes other uncertainties. In addition, research in the focus area Perception also addresses traffic situations in which parts of the scene are only partially visible or completely occluded. V2X communication and intelligent infrastructure systems extend perception beyond what the EGO vehicle can directly detect, thereby facilitating the identification of remaining uncertainties as well as potentially occluded road users.