Probabilistic Non-Line-of-Sight Detection in Reliable Urban GNSS Vehicle Localization based on an Empirical Sensor Model

Satellite based vehicle localization is an important requirement for a variety of innovative automotive applications. When putting such applications to dense urban areas, so called non-line-of-sight satellite observations-also known as multipath-need to be handled carefully. In this paper, this problem is addressed by proposing a real-time probabilistic multipath mitigation algorithm for robust and reliable vehicle localization with low-cost GNSS sensors for urban environments. Another main contribution of this paper is the derivation of an empirical signal-to-noise distribution from a long-term measurement campaign. It will be demonstrated that by using this additional information throughout the vehicle localization algorithm, the position accuracy can be increased by 10% with an enhanced integrity compared to previous work. The proposed algorithms are carefully evaluated with real-world data and compared to a high-reliable ground truth reference sensor.