无线传感网络中分布式粒子滤波的目标追踪算法

Target tracking based on distributed particle filter in wireless sensor network

在给出无线传感网络的传感器配置模型的基础上,提出了一种分布式粒子滤波DPF(d istributedparticle filter)算法,并实现了对网络中的一个运动目标的追踪。利用传感器模型,可将网络划分为一系列不相联系的传感器组,并在每个传感器组上运行一个局部粒子滤波,通过中心节点将估计状态传递给下一个传感器组的中心节点,依次实施对目标定位、跟踪。为减少网络间的通讯负荷和节约传感器节点能量,使用了高斯混合器模型(GMM),对局部粒子滤波的粒子和相应的权值进行近似;并提出了根据估计误差自适应激活传感器的优化算法。仿真结果表明,基于GMM近似的DPF在保持较高的估计精度的同时,能够大幅度减少网络间的通讯负荷。

A distributed particle filter algorithm was proposed to track a moving target in a wireless sensor network based on the sensor model. With the sensor model, the wireless sensor network could be separated into a of series disjointed sensor cliques, and the local particle filter was running on each sensor clique. The estimated state distribution was propagated by the wireless communication channel from the central node of a sensor clique to the next central node. To reduce the communication burden, the Gaussian mixture mode（GMM） was proposed to approximate the belief; and an adaptive heuristic sensor configuration algorithm was also proposed. The tracking results of the simulation show that the proposed DPF based on GMM approximation has a high performance of tracking, and can reduce the communication burden significantly.