一种适用于稀疏无线传感器网络的改进分布式UIF算法

An Improved Distributed Unscented Information Filter Algorithm for Sparse Wireless Sensor Networks

分布式无迹信息滤波(Distributed unscented information filter,DUIF)算法是一种有效的非线性分布式状态估计多源信息融合方法,然而当将该算法应用于稀疏无线传感器网络(Wireless sensor networks,WSN)时,稀疏WSN中存在的无效节点会引起使滤波趋于发散的平均一致误差.针对该问题,本文提出一种改进DUIF算法.该算法不改变DUIF算法的级联结构,而是将其底层和上层滤波器分别改进为局部无迹信息滤波器(Local unscented information filter,LUIF)和加权平均一致性滤波器.LUIF对每个节点的局部多源观测信息进行局部融合,得到局部的后验估计信息向量和矩阵,进而将它们作为加权平均一致性滤波器的输入,最终得到不包含平均一致误差的分布式后验估计结果.其中,加权平均一致性滤波器是通过对由LUIF输出的局部后验估计信息向量和矩阵分别进行平均一致性滤波而得以在改进DUIF算法框架下实现的.同时,在此过程中,相邻节点之间的状态估计互相关信息也被引入改进DUIF算法的输出结果中,进一步增强了滤波的可靠性.仿真实验结果表明,改进DUIF算法能够在稀疏WSN中对机动目标进行有效跟踪,在估计精度和抑制滤波发散方面明显优于标准DUIF算法.

The distributed unscented information filter algorithm （DUIF） is an e？cient non-linear distributed multi-source information fusion approach. However, when applying the DUIF algorithm to the sparse wireless sensor network （WSN）, the invalidating nodes existing in the sparse WSN will induce the average-consensus error which may make the DUIF algorithm divergent. To solve the problem, an improved algorithm is proposed in this paper, which does not change the cascade structure of DUIF algorithm. The improved DUIF algorithm introduces a local unscented information filter （LUIF） and a weighted average consensus filter as the bottom and top filters, respectively. The LUIF fuses the local multi-source information of each node, and outputs the local posterior estimating information vector and matrix. Then it makes these local vectors and matrixes as the input to the weighted average consensus filter, and gets the distributed posterior estimating results which do not contain the the average-consensus error. By means of making the output of LUIF as the input of an average consensus filter, the weighted average consensus filter is realized under the framework of the improved DUIF algorithm. Meanwhile, the cross correlation information between the neighbouring nodes is also introduced into the output of the improved DUIF algorithm, which improves the reliability of the fiter. The simulation results show that the proposed improved DUIF algorithm can e？ciently track the target in sparse WSN, and is obviously better at estimating accuracy and inhibitting filter divergence than the standard DUIF algorithm.