文献类型：期刊文献

中文题名：基于扩散卡尔曼滤波算法的目标跟踪估计

英文题名：TARGET TRACKING ESTIMATION BASED ON DIFFUSION KALMAN FILTERING ALGORITHM

作者：程华[1];张雪婷[1];房一泉[1]

机构：[1]华东理工大学信息科学与工程学院,上海200237

年份：2021

卷号：38

期号：2

起止页码：191

中文期刊名：计算机应用与软件

外文期刊名：Computer Applications and Software

收录：CSTPCD;;北大核心:【北大核心2020】；

基金：中国教育和科研计算机网网络中心创新项目(NGII20160606)。

语种：中文

中文关键词：目标跟踪;分布式估计;卡尔曼滤波算法;扩散矩阵;均方偏差;平均能耗

外文关键词：Target tracking;Distributed estimation;Kalman filtering algorithm;Diffusion matrix;Mean-square deviation;Average energy consumption

摘要：针对节点网络上的目标跟踪,提出一种基于扩散Kalman滤波算法的分布式跟踪估计。假设该节点网络系统按照线性状态空间模型演进,网络中的每个节点获取与未观察到的状态线性相关的测量值;对于每个测量值和每个节点,采用来自邻近区域的数据计算出一个局部状态估计值;采用一个基于扩散矩阵和连接矩阵的扩散步骤,将前面计算得到的邻域估计值在整个网络上扩散,从而使得每个节点获得关于系统状态的最佳跟踪估计。仿真实验结果表明,该算法不仅在平均均方偏差性能方面可与集中式Kalman滤波跟踪算法相比拟,而且在平均能耗方面优于对比算法。
In view of the target tracking problem over a network of nodes,a distributed tracking estimation based on the diffusion Kalman filtering algorithm is proposed.It was assumed that the node network system evolved according to the linear state-space model,and each node in the network took measurements linearly related to the unobserved state;for every measurement and for every node,a local state estimate was computed using the data from the neighborhood;a diffusion step based on diffusion matrix and link matrix was adopted to diffuse the estimates of the neighborhood computed previously across the network,so that each node could get the best tracking estimates of the system state.The simulation results show that the proposed tracking algorithm not only can be comparable with the centralized Kalman filtering tracking algorithm but also superior to other distributed Kalman filtering algorithms in the term of mean-square deviation performance,and is better than several compared algorithms in the average energy consumption.