基于变分贝叶斯算法的多模型车载组合导航算法  

Multi-model Vehicle Integrated Navigation Algorithm Based on Variational Bayesian Method

作　　者：王红茹 朱东琴 国强[1] 戚连刚 WANG Hongru;ZHU Dongqin;GUO Qiang;QI Liangang(College of Information and Communication Engineering,Harbin Engineering University,Harbin 150001,China)

机构地区：[1]哈尔滨工程大学信息与通信工程学院,哈尔滨150001

出　　处：《实验室研究与探索》2024年第2期98-103,109,共7页Research and Exploration In Laboratory

基　　金：哈尔滨工程大学本科教育教学改革项目(JG2022B0810)。

摘　　要：在行车过程中强机动引起全球卫星导航系统(GNSS)量测噪声产生野值,表现出厚尾特性导致常规状态估计精度下降的问题,为此提出一种基于变分贝叶斯的SINS/GNSS组合导航信息融合方法。构建车载组合导航系统模型,采用Student’s t分布对量测异常情况下噪声建模,并用变分贝叶斯的方法对系统状态和隐变量进行求解,实现对模型参数的后验估计。针对城市行车存在GNSS测量失效的问题,利用交互式多模型算法实现了GNSS量测中断情况下的SINS/GNSS和SINS/OD子系统的动态交互融合。通过跑车实验进行验证,实验结果表明,所提算法可有效抑制GNSS量测野值噪声对SINS/GNSS/OD组合导航系统的影响,与传统交互式多模型算法相比,具有较高的精度和鲁棒性。In order to solve the problem of the accuracy declination of general state estimation in the existence of outliers in GNSS measurement noise which has the thick tail characteristics caused by strong maneuvering driving,SINS/GNSS integrated navigation information fusion method is proposed based on variational Bayesian method.Firstly,a model of vehicle integrated navigation system is established.Abnormal measurement noise is modeled as the Student’s t distribution.The system state and hidden variables are given by variational Bayesian method.Therefore,posterior estimation value of the model parameters can be obtained.Then,the dynamic interactive fusion of SINS/GNSS and SINS/OD subsystems is implemented by using the interactive multi-model to handle the problem of invalid measurement of GNSS in urban driving.Finally,the offline car experimental results show that the proposed method can effectively reduce GNSS measurement outlier noise sharply exerted on SINS/GNSS/OD integrated navigation system.It has much higher accuracy and robustness compared with the traditional interactive multi-model.

关 键 词：车载组合导航 变分贝叶斯 Student’s t分布 交互式多模型 野值噪声 

分 类 号：TN967[电子电信—信号与信息处理]