Evaluation and Comparison of GNSS Navigation Algorithms including FDE for Urban Transport Applications

Abstract

Integrity is one of the most important parameters characterizing the Global Navigation Satellite Systems (GNSS) performance for safety-related applications. It is defined as a measure of trust to be placed in the correctness of the information supplied by the total system. This concept is more and more important, especially for applications in urban environment with standalone GNSS system, where GNSS performance is strongly degraded by the signal propagation conditions. Non line of Sight (NLOS) signals, caused by the surrounding obstacles, are of particular interest, as the error they induced remains a large cause of inaccuracy. In order to get rid of these constraints and to guarantee a better performance of positioning accuracy as well as integrity, several approaches can be used, such as positioning weighting models, 3D city models, Fault Detection and Exclusion (FDE) techniques etc. Since the concepts of integrity and accuracy are strongly related to each other, the main objectives of this paper are to reduce and to bound position errors in order to pave the way for further work of integrity. In the work presented here, we will introduce 2 classes of methods which contribute to the NLOS signal mitigation as well as integrity performance. The first class concerns the use of different weighting models in order to reduce the influence of the NLOS on the overall measurements. The second class consists of Fault Detection and Exclusion (FDE) algorithms which allow users to identify and remove the outliers. Using the real GPS data collected from an experiment in Nantes, firstly, we compare the Ordinary Least Square (OLS) and the Weighted Least Square (WLS) solutions with four different weighting models whose variances depend respectively on: 1) Dirichlet Process Mixture (DPM) [1], 2) the Carrier-to-Noise Ratio C/N0 [2], 3) the satellite elevation [3], 4) the hybridization of the C/N0, the satellite elevation and an indicator of LOS/NLOS signal [4], which comes from an “urban trench” model [5]. The simulation results prove that the hybridized weighting model can achieve a significant improvement in terms of accuracy compared to the other three models: the 95% horizontal position accuracy can be improved by about 78.39%. Then, with the same data and the DPM weighting model as the basic position estimation algorithm, four FDE algorithms are respectively implemented: 1) Subset testing 2) Global and sequential local test 3) Forward Back testing [6] 4) Danish method [6][7]. The simulation results show that the position accuracy can also be well improved with FDE algorithms since unreliable positioning caused by large errors can be excluded. Thus, integrity performance can be better guaranteed but system availability is inevitably reduced.