基于贝叶斯估计的加权最小二乘分布式融合  

Weighted Least Squares Distributed Fusion Based on Bayesian Estimation

作　　者：徐苏[1] 杨红[2] 

机构地区：[1]淮海工学院工程训练中心,江苏连云港222000 [2]广州大学物理与电子工程学院,广东广州510006

出　　处：《探测与控制学报》2011年第6期46-51,共6页Journal of Detection & Control

基　　金：国家自然科学基金项目资助(60774032);教育部高等学校博士学科点专项科研基金项目资助(新教师基金课题)(20070561006);广东省自然科学博士启动项目资助(9451064101002853)

摘　　要：针对不同观测矩阵加权最小二乘算法没有考虑到模型参数本身信息的共同缺点,提出基于贝叶斯估计的带不同观测矩阵的加权最小二乘分布式(Bayes Estimation Weighted Least Square,BEWLS)融合Kalman滤波算法。该方法首先采用推广的离散卡尔曼滤波对非线性系统线性化,然后在考虑模型参数本身的先验信息条件下,采用基于贝叶斯(Bayes)估计方法对Kalman滤波算法的观测方程进行加权最小二乘融合。BEWLS融合算法减少了计算负担,提高了融合精度,便于实时应用。理论和仿真证明:BEWLS融合具有优越性。For the common shortcomings that weighted least squares algorithm for different observation matrix did not take the model parameters information into account,Bayesian estimation weighted least square（BEWLS） fusion Kalman filter algorithm with different observation matrix was presented based on Bayes estimation.The method first applied extended discrete Kalman filter to linearize nonlinear systems,then Bayes estimation was used to undertake weighted least square fusion for the observation equations of algorithm in considering the model parameters of their own prior information.BEWLS algorithms could reduce the computational burden and improve the fusion accuracy and was suitable for real time application.The BEWLS fusion advantages were proved theoretically and testified by simulation.

关 键 词：非线性离散系统 卡尔曼滤波 多传感器信息融合 分布式观测融合 BAYES估计 

分 类 号：O211.64[理学—概率论与数理统计]