复杂城市环境下GNSS/INS组合导航可观测度分析及鲁棒滤波方法

In complex urban environment, urban canyon, building occlusion, multipath, and non-line-of-sight effects have serious influence on the quality of GNSS signal. By introducing the observability analysis method in dynamic time-varying systems, the observability of each system state-variable can be quantitatively evaluated online. And the mechanism of the influence of abnormal GNSS signals on the integrated navigation performance can be analyzed. Aiming at the continuous and reliable navigation where the GNSS signals are changed or denied for a short time, an adaptive factor is proposed based on the relationship between recursively calculated values and theoretically defined values in the estimation error covariance matrix. It can reflect the abnormity of GNSS/INS fusion and filtering process, and provide reference for designing an adaptive robust Kalman filtering algorithm(ARKF). Experimental simulation results show that the maximum positioning error can be limited to less than 1 m, which can be utilized as an important indicator of vehicle's safety. Therefore, the robust filtering method enhances the robustness and environmental adaptability of GNSS/INS integrated navigation systems, and guarantees high-precision navigation and fusion positioning of lane-level for self-driving cars in complex urban environments.