Lifetime-aware information aggregation under uncertainty for advanced driver assistance systems

Abstract

Advanced Driver Assistance Systems require a huge amount of sensor information. Information sensed by the vehicle’s on-board sensors is often not sufficient due to physical limitations, thus, vehicles exchange information with each other to enhance their knowledge. However, both local and remote measurements are only a noisy view on the environment. If a vehicle receives faulty information about its environment, its decision-making may perform wrong actions. In this paper, we focus on the accuracy of information w. r. t. the correctness of the provided information. To increase the accuracy, we use a data representation that depicts every measurement as a probability vector. This representation contains all available information and tracks accuracy while aggregating and fusing information. To incorporate measurements sensed at different times and locations, we model the environmental changes. As information expires after a certain time, we combine a Markov model with an aging model to keep the aggregated information stable even after message expiration. We achieve this by reducing the impact of information with increasing age. Our extensive evaluation shows that our novel approach outperforms state-of-the-art approaches significantly.