Two-Step Graph Optimization for GNSS/INS Integration in Arctic Shipborne Navigation

Abstract

Reliable positioning in the Arctic remains a critical challenge due to low satellite elevation angles, frequent signal blockages by ice ridges, and strong ionospheric disturbances, all of which degrade the performance of global navigation satellite systems (GNSSs). Most GNSS/inertial navigation system (INS) integration techniques, ranging from extended Kalman filter (EKF) to conventional factor-graph-based frameworks, have seen limited validation in the Arctic, where sparse satellite geometry and prolonged signal outages pose conditions far more severe than typical mid-latitude environments. To bridge this gap and improve robustness in such conditions, we propose a two-step graph optimization (TSGO) framework for loosely coupled (LC) GNSS/INS integration tailored to Arctic shipborne applications. In the first stage, TSGO performs local graph-based optimization on raw GNSS pseudorange and Doppler observations to mitigate outliers and improve positioning accuracy. The second stage incorporates the optimized GNSS results into a global factor graph alongside inertial measurements to refine the trajectory through globally consistent smoothing. Field experiments conducted aboard an Arctic research vessel demonstrate that TSGO significantly outperforms conventional EKF and optimization-based GNSS/INS (OB-GINS) methods. Compared to EKF, TSGO reduces the maximum horizontal error by over 70% and the root mean square error (RMSE) by 24%. It also achieves 33% and 10% improvements in these metrics over OB-GINS. These results highlight the effectiveness and robustness of TSGO in high-latitude GNSS-challenged environments.