Non-line-of-sight mitigation for reliable urban GNSS vehicle localization using a particle filter

GNSS based localization in the context of Advanced Driver Assistance Systems and autonomous driving raises its attention regarding positioning performance not only related to accuracy but integrity as well. Especially, for safety relevant applications the proper computation of confidence levels under degraded environmental conditions is of major importance. Low cost solutions that integrate GNSS and additional in-vehicle sensor information are able to bridge short periods of time with limited GNSS accessibility and can therefore improve availability and accuracy. Additionally, non-line-of-sight (NLOS) and multipath effects in urban areas need special attention as these error influences violate the estimated confidence and introduce unobservable offsets to the position solution. The mitigation of local influences in urban areas increases the demands for the integration of proper error models for NLOS and multipath errors. The algorithmic detection of these effects and the proper propagation of all uncertainties within a Bayes framework is one of the key technologies towards the adoption of GNSS for safety critical applications. This paper proposes a probabilistic NLOS detection algorithm that is able to improve both - accuracy and integrity of the position estimate in urban areas. As an extension of a previous implementation by the authors based on an unscented Kalman Filter the proposed system is implemented as a particle filter in order to meet automotive requirements in terms of real time and scalability. Both approaches are compared by an evaluation of a data set from an urban test drive in terms of accuracy and integrity.