Tightly Coupled Multi-GNSS Receiver Fusion for Robust Position Estimation in Urban Environments

R. Streiter, S. Bauer, G. Wanielik

Abstract: Satellite based localization is still the key technology for vehicle positioning in automotive applications. Especially Low-Cost GNSS receivers that are employed with vehicles are not suitable for precise localization issues in Advanced Drivers Assistance systems (ADASs) and Intelligent Transportation Systems (ITS). To enable applications like tolling, which charges the user on road or lane usage, or green driving assistants for efficient path planning a further improvement of positioning accuracy is required. In order to mitigate the low positioning performance of these receivers a well-known approach is the fusion of the GNSS receivers’ measurements with other sensors available onboard. Those are for example odometry sensors like yaw rate or acceleration sensors. This paper proposes to apply several GNSS receivers in a single vehicle in combination with a tightly coupled integration of the receivers’ raw measurements with the vehicles odometry information. These data is typically fused by a Bayes Filter implementation that provides stable position estimation. Such filters use process models that can describe the physical constraints of the vehicle movement. For example, it can be assumed that a vehicle moves with a roughly constant turn rate and velocity, approximately. Considering the current measurements and the constraints of such an underling motion model the vehicle state (e.g. the velocity and position) can be estimated including the uncertainties of the estimation itself. Hereby the most trivial approach is to fuse high-level sensor output like the GNSS positioning solution with additional sensor measurements. Unfortunately, when calculating the position solution at the receiver, some information is lost by the receiver’s data reduction. To avoid this, more complex filter models can be used to perform the vehicle state estimation directly based upon low-level pseudorange measurements. This approach is usually called Tightly Coupling. Latest research results prove that the concept of tightly coupling of odometry and GNSS raw data can improve the accuracy of position solutions dramatically - especially in challenging environments where pseudorange measurements are strongly multipath and non-line-of-sight affected. Within the authors’ previous work, it was shown that providing more redundant positioning measurements from other GNSS receivers can further improve the accuracy in a loosely coupled scenario. Within this paper the advantages of tightly coupled approaches are introduced into the multi receiver architecture. Besides the redundancy of the pseudorange measurements---which mitigates the influence of measurement noise and single receiver errors---an implicit heading sensor is derived from the antenna configuration geometry to improve the filters estimations. The designed sensor and process model of the filter implementation will be shown in the final paper. However, the sole knowledge of an estimated position is often not sufficient for the implementation of safety relevant ADAS applications. Next to accuracy, integrity and confidence considerations---as known from aviation applications---are becoming more and more important. Especially in the European research project GAIN---where positioning is safety relevant---the integrity concept is a major cornerstone of the ongoing development. While uncertainties are handled within the Bayes filter implementation – the straight forward concept contains no correlation description between pseudorange measurements. That leads to misleading results in the integrity estimation which is often underestimated due to the missing correlation model in the filter implementation. In order to address the problem, the antenna configurations’ unknown correlations were determined using an optimization process. In order to evaluate the performance of the proposed positioning algorithm, real sensor data was recorded using the measurement vehicle “Carai” from the chair of communications engineering of the Chemnitz University of Technology. It was equipped with four uBlox Lea 6 – out-of-the-shelf mass market GNSS receivers. The antenna configuration is roof, hood and two rear receivers left and right. For the ground truth generation a NovAtel Span GNSS system with real time RTK and inertial measurement unit was used. Two different scenarios were considered. First a test drive under open sky conditions on a highway was run to assess the systems general viability. Furthermore, an urban test drive was run to evaluate the performance under strongly multipath affected conditions. Integrity and accuracy are evaluated for both scenarios and compared to the authors previous loosely coupled multi antenna implementation. Based on previous experiences and preliminary observations a significant improvement of the accuracy is expected and a more robust confidence interval will be estimated. A detailed description of the system architecture and the evaluation process will be given at the final paper. The advantage of Multi GNSS receiver fusion for positioning accuracy and the probabilistic tightly coupled GNSS/Odometry integration were demonstrated in previous positioning research activities. The proposed system integrates both approaches into a common Multi Antenna Tightly Coupled positioning algorithm that is supposed to increase the position accuracy and integrity of low cost GNSS receivers. The robust integrity estimation can decrease the false alarm rate of depending ADAS applications. An outlook that can be predicted so far is the need for a generic representation of the correlation of close-by GNSS receiver measurements.
Published in: Proceedings of the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2013)
September 16 - 20, 2013
Nashville Convention Center, Nashville, Tennessee
Nashville, TN
Pages: 1285 - 1289
Cite this article: Streiter, R., Bauer, S., Wanielik, G., "Tightly Coupled Multi-GNSS Receiver Fusion for Robust Position Estimation in Urban Environments," Proceedings of the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2013), Nashville, TN, September 2013, pp. 1285-1289.
Full Paper: ION Members/Non-Members: 1 Download Credit
Sign In