GNSS/INS integrated system based on weighted factor graph optimization with dynamic performance evaluation

Abstract Accurate and robust positioning systems are crucial for vehicular applications. In order to improve positioning performance in complex urban environments, such as urban canyons and dense forests, we propose a global navigation satellite system (GNSS) and inertial navigation system (INS) integrated approach based on weighted factor graph optimization (FGO) with dynamic performance evaluation (DPE). This approach is designed to enhance the estimation of vehicle states, including position, velocity, and attitude. First, we introduce a DPE framework to assess the observation quality of GNSS pseudorange measurements. Specifically, we use Fisher information and Kullback–Leibler divergence to comprehensively evaluate the observation performance of pseudorange measurements and the actual information gain that pseudorange measurements contribute to constraining the uncertainty of the system state. Finally, considering the different observational qualities of all pseudorange information, we use information weight assignment to handle the measurement accuracy of all pseudorange information, and then use weighted FGO with a sliding window for fusion estimation of GNSS and INS to achieve robust vehicle positioning. Simulation experiment and real scenario experiment demonstrate that our proposed method can provide more reliable positioning in complex urban environment. Compared with the existing adaptive weighted FGO throughout the experiment, the overall positioning accuracy of the proposed method in the three-axis direction is improved by 13.6%. This shows that the proposed method can be widely applied to robust vehicle positioning.