基于变分贝叶斯与子空间辨识的融合定位算法  

Fusion Localization Algorithm Based on Variational Bayes and Subspace Identification

作　　者：夏金凤 刘延旭 XIA Jinfeng;LIU Yanxu(School of Computer and Information Engineering,Dezhou Univer sity,Dezhou Shandong 253023,China)

机构地区：[1]德州学院计算机与信息学院,山东德州253023

出　　处：《德州学院学报》2023年第6期44-49,55,共7页Journal of Dezhou University

基　　金：德州学院校级科研项目资助(2022xjrc111)。

摘　　要：面向全球导航卫星拒止环境下无人机室内高精度高可靠定位导航需求,室内电磁环境复杂导致的传感器噪声时变和无人机高灵动性导致的无人机运动过程噪声时变,降低了定位状态估计值的精度和稳定性的问题,而传统变分贝叶斯滤波方法仅能估计过程或观测噪声,为此,提出了基于变分贝叶斯与子空间辨识的无人机融合定位算法,在变分贝叶斯框架下基于子空间辨识理论,实现了时变过程噪声的动态估计,解决了现有变分贝叶斯框架无法同时估计时变观测和过程噪声的问题,最后通过实测数据集的UWB/IMU/光流融合定位测试验证结果表明,该方法提升了无人机室内定位的精度和鲁棒性。In the context of global navigation satellite system(GNSS)denied environments for unmanned aerial vehicle(UAV),there is a need for high-precision and high-reliability indoor positioning.The complex electromagnetic environment indoor leads to time-varying sensor noise,and the high agility of UAV results in time-varying motion noise,which reduces the accuracy and stability of position estimation.Traditional variational Bayesian filtering methods can only estimate either process or observation noise.Therefore,we proposed a UAV fusion positioning algorithm based on variational Bayesian filtering and subspace identification.Within the variational Bayesian framework and utilizing subspace identification theory,the proposed algorithm dynamically estimates time-varying process noise,solved the problem of variational Bayesian frameworks being unable to simultaneously estimate time-varying observation and process noise.Finally,experimental results from a real-world dataset of UWB/IMU/optical flow fusion positioning that the proposed algorithm enhances the accuracy and robustness of indoor UAV po sitioning.

关 键 词：变分贝叶斯 子空间辨识 融合定位 时变噪声方差估计 

分 类 号：TN92[电子电信—通信与信息系统]