Advances in Multi-Sensor Data Fusion for Ubiquitous Positioning: Novel Approaches for Robust Localization and Mapping

In this paper we argue why robust positioning in transportation applications is best achieved by multi-sensor fusion. Furthermore, we suggest that sensor fusion processing be performed in a probabilistic fashion and that in the majority of relevant practical applications one should draw on utility theory in order to make decisions that will be of the highest expected benefit given the current circumstances. Simply stated, it is a fact that all sensors are prone to errors or failure. Only if we model these errors correctly, and account for all possible failure modes, are we able to implement systems that reap the benefits of multi-sensor fusion: increased reliability and a valuable indication of the currently achieved accuracy. We provide examples for cooperative automotive applications that apply utility-based information dissemination in a vehicle-to-vehicle communications setting as well as an outlook to collaborative Simultaneous Localization and Mapping (SLAM).