Robust Vehicle Positioning in Multipath Environments Based on Graph Optimization

Abstract

This paper proposes a high-precision positioning method using graph-based optimization with only a global navigation satellite system (GNSS) in an urban environment with GNSS multipath signals. GNSS has become an important core sensor for automatic driving and pedestrian navigation, and it is important to improve the positioning accuracy in urban environments where multipath occurs. We propose a novel graph-optimization-based GNSS positioning method. The feature of the proposed method is that it fully utilizes pseudorange and Doppler observations of a multi-GNSS constellation and constructs a graph to estimate position, velocity, and receiver clock biases. We also add constraints on past and current nodes of the graph using time-difference observations of the GNSS carrier phase. The carrier phase observations of GNSS have rarely been used for graph optimization. By adding a constraint with a precise carrier phase, we can improve the accuracy of position estimation. In addition, a robust optimization method is used to exclude outliers due to GNSS multipath signals and optimize the entire graph to achieve highly accurate positioning in multipath environments. We evaluated the accuracy of the proposed method by collecting GNSS data in a real urban environment using a vehicle. As a result, the horizontal positioning error of the proposed method is reduced to 1.37 m, while the conventional GNSS-only positioning method has a horizontal positioning error of 5 to 10 m, indicating that the proposed method can achieve high-precision positioning in a multipath environment.