Early Sensor Data Fusion Techniques for Collision Mitigation Purposes

Nowadays there is a growing interest towards active safety systems in the automotive industry and various systems are developed. The key feature of all this novel driver assistance and safety systems is the accurate, robust and reliable perception of the vehicle's environment. This knowledge is essential for applications like the envisaged collision mitigation. However, current off-the-shelf single sensor approaches cannot always fulfil these challenging demands. Therefore, in our work we develop a multi sensor approach to examine the capabilities of early sensor data fusion. The term "early fusion" derives from a model-based fusion concept, which relies on only slightly pre-processed data and thus provides the possibility of consolidated and integrative sensor data interpretation. Contrary to track-based fusion the sensor data are merged in an early stage of the processing chain to preserve most of the environmental information content and therefore to tap almost the full sensor potential. This paper gives an overview on this novel approach to fusion