无线传感器网络中丢包扩散卡尔曼算法的改进

Improved diffusion Kalman algorithm with packet-dropping in wireless sensor networks

针对无线传感器网络数据融合时的丢包问题,对传统扩散卡尔曼滤波算法进行改进,并探讨丢包问题的改善方案。丢包情况下,改进算法在传统算法增量更新时,从邻居节点集合中剔除有丢包的节点,在融合更新时,节点重新调整融合权值,以减小丢包对估计值的影响。仿真结果表明,在一定条件下,改进算法的平均误差偏置比传统算法小2～3dB。由于"领导"节点较普通节点对系统的估计值影响更大,所以,可以通过选择能量大的节点作为"领导"节点,或者加强对"领导"节点的维护来改善丢包问题。

In this paper,an improved diffusion Kalman filter algorithm in connection of the ＂packet-dropping＂ problem for data fusion in wireless sensor networks is proposed and an improved scheme is achieved.When packet-dropping exists,the nodes with packet-dropping from their neighbor nodes sets are excluded by the improved algorithm in order to reduce the bad effect of packet-dropping on estimated values.In profile of the data fusion,nodes are designed to readjust the fusion weights and to reduce the bad impacts of the packet-dropping on the estimated values.Simulation results show that the improved algorithm has a better performance than the traditional one.This mainly reflects on the lower average error bias under several certain circumstances.Besides,＂leader＂ nodes turn to have greater effects on the estimated value of the system than general nodes.Therefore,the ＂packet-dropping＂ problem can be improved by choosing the nodes with enough energy as ＂leadership＂ or strengthening the maintenance of those ＂leader＂ nodes.