无线传感器网络的分布式目标跟踪研究  

Study on distributed target tracking in wireless sensor networks

作　　者：周伟[1,2] 石为人[2] 张洪德[2] 王小刚[2] 易军[1] 

机构地区：[1]重庆科技学院电气与信息工程学院,重庆401331 [2]重庆大学自动化学院,重庆400044

出　　处：《仪器仪表学报》2013年第7期1485-1491,共7页Chinese Journal of Scientific Instrument

基　　金：国家科技重大专项(2011BAJ03B13);十二五国家科技支撑计划项目(2011BAK07B03);重庆市教委科学技术研究项目(KJ121410)资助

摘　　要：针对无线传感器网络节点计算能力和能量受限问题,提出一种分布式并行扩展卡尔曼粒子滤波算法。在网络动态分簇模型上,簇头将粒子集划分为多个子集,并分配到簇内各个传感器节点中并行运行,最后在簇头进行信息融合,得到目标状态估计。算法提高了粒子滤波效率,避免单个节点能量过度消耗,均衡了网络能耗。同时,算法利用扩展卡尔曼滤波器来产生粒子滤波的重要性密度函数,使得重要性密度函数抽样样本更加接近后验概率密度产生的样本。仿真结果表明,算法对运动目标能实现较好的预测和跟踪,跟踪精度高,并能有效平衡网络能耗。实验结果说明了提出算法的有效性和可行性。Aiming at the problem of limited computing capacity and energy in wireless sensor networks,in this paper a distributed parallel extended Kalman particle filtering algorithm is proposed. Based on the network dynamic cluster model, the cluster head （CH） divides the particle set into multiple subsets,which are assigned to the sensor nodes and run in parallel. The computed results of the sensor nodes are fused on the CH, and the optimal state estimation of the target is ob- tained. The proposed algorithm improves the particle filtering efficiency and avoids excessive energy consumption in each sensor node,and balances the network energy consumption. The algorithm uses extended Kalman filter algorithm to gener- ate the importance density function of particle filtering ,which makes the sample of the important density function closer to the sample generated by the posterior probability density. The simulation results show that, the proposed algorithm can predict and track the moving target better, has high tracking accuracy and can effectively balance the network energy consumption. The experimental results verify the effectiveness and feasibility of the proposed algorithm.

关 键 词：无线传感器网络 目标跟踪 粒子滤波 分布式 

分 类 号：TP393.0[自动化与计算机技术—计算机应用技术]