传感器网络中鲁棒状态信息融合抗差卡尔曼滤波器

研究了无线传感器网络中的分布式鲁棒状态信息融合问题. 在局部状态估计层, 基于鲁棒统计学理论提出了适用于噪声相关情况的抗差(扩展)卡尔曼滤波器. 在融合中心层, 针对局部估计相关未知性和不完整性, 给出了不依赖于互协方差阵的稳健航迹融合方法—–内椭球逼近法. 仿真结果证实了算法的有效性: 所提出的抗差卡尔曼滤波器在野值存在情况下, 性能退化远低于传统卡尔曼滤波器(28.6%比428.6%); 所提出的内椭球逼近法获得比协方并交叉法更好的融合估计性能, 且不需要局部估计相关性的先验知识.

Anti-outlier Kalman filter-based robust estimation fusion in wireless sensor networks

The problem of distributed robust estimation fusion is considered for a hierarchical wireless sensor network(WSN).Based on the theory of robust statistics(RS),a novel anti-outlier(extended) Kalman filter(KF) is presented for local state estimation in a clustered WSN with correlated measuring noises.In the fusion center(FC),a cross-covarianceindependent track fusion approach –-internal ellipsoidal approximation fusion(IEAF) is developed to fuse the local estimates,among which the correlations are usually unknown or incomplete.Simulation results illustrate the significance of the proposed approaches: the presented anti-outlier KF deteriorates in performances much less than the traditional KF(28.6% VS.428.6%) in the presence of outlier;the proposed IEAF has higher fusion accuracy than the fusion estimator of covariance intersection(CI),and doesn’t need any prior knowledge.