基于自适应评估的鲁棒因子图导航算法研究  

Research on Robust Factor Graph Navigation Algorithm Based on Adaptive Evaluation

作　　者：赵静欣 王融[1] 胡博 魏帅迎 包文龙 王聪聪 ZHAO Jingxin;WANG Rong;HU Bo;WEI Shuaiying;BAO Wenlong;WANG Congcong(Navigation Research Center,College of Automation Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 211106;Aerospace Times Feihong Technology Company Limited,Beijing 100094)

机构地区：[1]南京航空航天大学自动化学院导航研究中心,南京211106 [2]航天时代飞鸿技术有限公司,北京100094

出　　处：《导航与控制》2025年第1期61-70,共10页Navigation and Control

基　　金：国家自然科学基金(编号:62073163,61703208,61873125);江苏省“青蓝工程”;中央高校基础科研经费(编号:NT2022009,NZ2020004,NZ2019007)。

摘　　要：在实际导航过程中,多源融合导航系统的传感器会随应用场景的变化出现量测精度变化的情况。针对传统因子图算法优化过程中无法处理传感器量测精度动态变化的问题,提出了一种基于自适应评估的鲁棒因子图算法。通过引入量测信息评估指标和自适应权重函数,动态调整因子图融合优化时的实时计算惯性预积分预测值与辅助传感器量测值的残差,动态调整对应因子节点的融合信息权重。与传统因子图算法相比,该算法在辅助传感器量测信息异常时能提升因子图算法的优化精度与鲁棒性。仿真实验结果表明,在辅助传感器出现量测误差时,和传统因子图算法相比,所提出的基于自适应评估的鲁棒因子图导航算法具有更高的鲁棒性和精度。In the actual navigation process,the sensors of the multi-source fusion navigation system will change the measurement accuracy with the change of application scenarios.In this paper,a robust factor graph algorithm based on adaptive evaluation is proposed to address the issue of traditional factor graph algorithms being unable to handle dynamic changes in sensor measurement accuracy during the optimization process.By introducing measurement information evaluation indicators and adaptive weight functions,the real-time calculation of the residual between the inertial pre-integration prediction value and the auxiliary sensor measurement value during factor graph fusion optimization is dynamically adjusted,and the fusion information weight of the corresponding factor nodes is dynamically adjusted.Compared with traditional factor graph algorithms,this algorithm can improve the optimization accuracy and robustness of the factor graph algorithm when the measurement information of various auxiliary sensors is abnormal.The simulation experiment results show that the proposed robust factor graph navigation algorithm based on adaptive evaluation has higher robustness and accuracy compared to traditional factor graph algorithms when measurement errors occur in auxiliary sensors.

关 键 词：因子图 多源融合导航 鲁棒自适应优化 传感器故障 预积分 

分 类 号：U666.1[交通运输工程—船舶及航道工程]