Probabilistic GNSS signal tracking for safety relevant automotive applications

Signal processing based on software defined radio (SDR) carries out innovative potential for advanced data fusion algorithms. In the field of satellite navigation, SDR is a key enabling technology for new integration approaches based on GNSS base band processing. As multipath and non-line-of-sight (NLOS) errors are the main challenge for GNSS in automotive applications, much research for the mitigation of those influences is still ongoing. Existing localization approaches are suitable to a limited extent for meeting the requirements of services and applications, especially in the context of driver assistance and autonomous driving where a proper computation of confidence levels is of mayor importance. If unhandled, multipath and NLOS errors introduce a bias to the GNSS observations resulting in an overestimation of the position confidence which violates the integrity of subsequent applications. This paper proposes an integrated probabilistic approach for a continuous implementation of the Bayesian framework within the signal processing of GNSS receivers. Special attention will be on modeling ambiguities on signal level that in traditional GNSS receivers ends up in multipath or NLOS errors. The idea is based on the theory of multiple object tracking to resolve ambiguities within the measurement space. The outcome of this paper is a model which condenses local error sources as generic system properties and proposes a Bayesian filter which improves the integrity and accuracy of the position estimate in urban areas without additional environmental knowledge. This approach is a step towards the applicability of low cost GNSS for safety relevant applications in the automotive area.