Uncertainty Quantification of 5G Positioning as a Location Data Analytics Function

Mobile positioning is a fundamental service of 5G as it enables a number of applications that rely on location information and location-based analytics. In many applications, it is important to quantity the uncertainty associated with position estimation, for example, for confidence assessment on the location data and anomaly detection, as well as for location data fusion from heterogenous technologies. In this paper, we propose uncertainty quantification as a location data analytics function. First, we introduce an indicator of positioning uncertainty based on the residual measurement error, which does not require the ground truth knowledge. Then, we train and update an uncertainty map of a monitored environment by leveraging the position estimates and location-based measurements collected by multiple users. Such uncertainty map can be used to predict the positioning uncertainty level in any point of a monitored environment. Finally, we propose an implementation of such functionality as an analytics function within the 5G architecture. The functionality is then deployed in a virtualized environment and, using system-level simulations under different propagation conditions, we show how the uncertainty predicted through the proposed method is highly correlated with the true positioning error.