适于无线传感网络目标追踪的一种改进无迹粒子滤波时延差估计算法

在无线传感网络（WSN)中运用基于粒子滤波的时延差估计方法进行目标追踪，其性能的关键是设计精确的粒子滤波器建议分布。为了解决追踪过程中粒子贫化问题，提出了一种基于改进无迹粒子滤波器的时延差估计算法。利用最小二乘法估计目标初始时刻位置，在卡尔曼滤波框架下运用高斯-牛顿迭代法则融合最新观测信息，并引入尺度调节衰减因子不断修正重要性密度函数，从而使建议分布更加逼近真实。将其与时延差定位方法结合，并在WSN环境下进行仿真实验。结果显示，改进的算法在整体粒子数有限的情况下追踪精度更高，收敛性较好，尤其适合环境噪声非高斯的复杂WSN目标追踪应用。

An Time Delay Difference Estimation Algorithm Based on Improved Unscented Particle Filter Suitable for Target Tracking in Wireless Sensor Network

For the time delay difference tracking estimation methods based on particle filter in wireless sensor network(WSN) , the key issue is to generate an accurate proposal distribution for particle filter. An time delay difference estimation algorthm based on improved unscented particle filter (IUPF) is proposed to overcome the degeneracy phenomenon of particles. The least square method is used to achive the initial target position, and then the unscented particle filter (UPF) and Gauss-Newton rule are used to incorporate the most current observations and provide more accurate importance density function for the particle filter by introducing a scaled correction factor．Finally, IUPF is applied to the time delay difference localization estimation methods in WSN. The simulation results show that, when the particle number is limited, the proposed algorithm can improves the target tracking accuracy and achieve faster convergence speed under non-Gauss noise environment in WSN.