Sensor Fusion as an Enabling Technology for Safety-critical Driver Assistance Systems

Abstract

Driver assistance systems are incorporating more and more advanced safety functions. As these functions have to react quickly and reliably in emergency situations with a false alarm rate close to zero a high integrity of the environmental perception is required. This elevated level of signal integrity can be achieved by data fusion, where the information of several, in general heterogeneous sensors is combined to obtain a better model of the environment in terms of accuracy, object integrity, object identity, etc. As an example, we demonstrate the power of sensor fusion by an automatic emergency brake (AEB) system whose environmental perception is based upon a video camera and a radar sensor. In particular we discuss the improvement of kinematic attributes such as object lateral distance as well as the object’s confidence or probability of existence.