一种多传感器组合导航系统的自适应分布式滤波算法  

An Adaptive Distributed Filtering Algorithm for Multi-Sensor Integrated Navigation System

作　　者：王玮[1] 孙炜玮 潘新龙 乔玉新[1] 韩真真 WANG Wei;SUN Weiwei;PAN Xinlong;QIAO Yuxin;HAN Zhenzhen(College of Intelligent Science and Engineering,Yantai Nanshan University,12 Daxue Road,Longkou 265713,China;Naval Aeronautical University,188 Erma Road,Yantai 264001,China;Shandong Donghai Thermal Power Co Ltd,Donghai Industrial Park,Longkou 265713,China)

机构地区：[1]烟台南山学院智能科学与工程学院,山东省龙口市大学路12号,265713 [2]海军航空大学,山东省烟台市二马路188号,264001 [3]山东东海热电有限公司,山东省龙口市东海工业园区,265713

出　　处：《大地测量与地球动力学》2023年第12期1275-1280,共6页Journal of Geodesy and Geodynamics

基　　金：国家自然科学基金(62076249);山东省自然科学基金(ZR2020MF154)。

摘　　要：针对联邦滤波器和集中式卡尔曼滤波器在实际应用中存在的问题及组合导航中量测噪声统计特性不准确引起滤波精度下降的问题,提出利用变分贝叶斯估计和分布式估计的多传感器组合导航系统自适应分布式滤波方法。首先,基于集中式卡尔曼滤波技术,推导分布式滤波的一种表达方法,该方法具有估计最优性、故障自动隔离且对导航系统无污染的特点;然后,结合变分贝叶斯估计技术,提出自适应分布式滤波方法,对系统状态和时变的量测噪声方差进行同步实时估计;最后,采用SINS/GNSS/CNS/ADS组合导航系统进行仿真验证。仿真结果表明,本文算法能实时跟踪突变或缓变的量测噪声方差,较联邦滤波算法可有效降低量测噪声统计特性不准确给系统带来的不利影响,进而提高组合导航系统的滤波精度。Aiming at the problems existing in the practical application of federated filter and centralized Kalman filter as well as the problems of decreased filtering accuracy due to inaccurate statistical characteristics of measurement noise in integrated navigation,we propose an adaptive distributed filtering method of multi-sensor integrated navigation system using variational Bayesian estimation and distributed estimation.Firstly,based on the centralized Kalman filter,we derive an expression distributed filtering method which has the characteristics of optimal estimation,automatic fault isolation and no pollution to the overall navigation system.Then we propose an adaptive distributed filtering method based on variational Bayesian estimation technique,which we use to estimate the system state and time-varying measurement noise variances synchronously and in real time.Finally,we use the SINS/GNSS/CNS/ADS integrated navigation system for simulation verification.The simulation results show that this algorithm can track the abruptly or slowly varying variance of the measurement noise in real time,can effectively improve the filtering accuracy of the overall integrated navigation system and reduce the adverse effects of inaccurate statistical characteristics of the measurement noise compared with the federated filter algorithm.

关 键 词：变分贝叶斯 自适应滤波 分布式估计 组合导航 联邦滤波 

分 类 号：P228[天文地球—大地测量学与测量工程]