Factor Graph Optimization-Based RTK/INS Integration With Raw Observations for Robust Positioning in Urban Canyons

The integration of Global Navigation Satellite Systems (GNSS) and Inertial Navigation Systems (INS) is subject to severe performance degradation in urban canyons due to signal blockages. We propose a novel tightly coupled RTK/INS robust positioning framework based on factor graph optimization (FGO), which directly fuses GNSS pseudorange and carrier-phase observations with a low-cost inertial measurement unit (IMU). The framework enhances positioning accuracy through full exploitation of integer ambiguity characteristics via persistent constraints maintained by ambiguity propagation with marginalization, while systematic sliding window size optimization under the ambiguity-constant model replaces empirical heuristics by balancing accuracy against computational efficiency. Furthermore, multiple robust estimation methods, including Gaussian mixture models (GMM) and their enhanced variants, were integrated into the framework to mitigate outliers. Vehicle-based experiments conducted in urban environments using low-cost IMU demonstrate that the proposed method achieves horizontal positioning accuracy of approximately 10 cm. Although FGO-based RTK/INS demonstrating more modest improvements than standalone RTK, ambiguity propagation achieves a practically meaningful 11.2% 3D accuracy enhancement. Considering both accuracy and computational efficiency, a sliding window size of 4-s is recommended for both FGO-based RTK and tightly coupled RTK/INS. While GMM-based methods demonstrate superior accuracy in GNSS-challenged environments despite increased computational demands, M-estimation and switch constraints provides balanced accuracy and efficiency.