M-Estimation Based Robust Factor Graph Fusion Method for Integrated Navigation under Non-Gaussian Noise
Jingxin Zhao, Rong Wang, Zhi Xiong, Jianye Liu · 2023
A robust factor graph fusion method based on Mestimation for the inertial navigation system and global navigation satellite system (INS/GNSS) is proposed in this paper. The factor nodes of the designed factor graph are composed of prior factor nodes and GNSS factor nodes with unary factor properties, and INS factor nodes with binary factor properties, in which the variable nodes are represented by extended navigation states. The traditional factor graph fusion algorithm always considers the measurement noise as a Gaussian distribution, which cannot adapt to the denial environment and carrier maneuvering, resulting in greatly reduced performance. In this paper, measurement noise with non-Gaussian characteristics is modeled as Gaussian mixture noise, which is then processed by M-estimation-based method. The incremental smoothing algorithm is used to enhance the effectiveness of the factor graph fusion method. The navigation accuracy of the proposed robust method for non-Gaussian measurement information is more than 45% higher than that of the fusion algorithm based on the existing factor graph, which has been verified by the experimental simulations.