Performance evaluation of automotive sensor data fusion

Abstract

Modern passenger and commercial vehicles are equipped with an increasing number of sensors that explore the driving surroundings. Sensor data fusion algorithms accumulate all available sensor information and provide models describing the surroundings. In this paper, an approach to evaluate the performance of such automotive sensor data fusion algorithms is presented. The method calculates performance indices on several evaluation levels. Since the number of performance indices decreases from low to high evaluation levels, the indices can be arranged in a pyramidal structure. While a large number of indices on lower levels describe many aspects and details of the evaluation process, few indices on higher levels give a more overall answer to the performance issue.