基于PCRLB的目标跟踪节点选择算法  

Sensor Selection Algorithm for Target Tracking with PCRLB

作　　者：庞小双 王邢波 

机构地区：[1]南京邮电大学自动化学院,江苏南京210023

出　　处：《计算机技术与发展》2017年第10期54-59,共6页Computer Technology and Development

基　　金：国家自然科学基金杰出青年基金(60825304);国家"973"重点基础研究发展计划项目(2009cb320600)

摘　　要：针对能量、带宽、存储等资源限制的无线传感器网络下的目标跟踪问题,提出了基于扩展H∞滤波的后验-克拉美罗下界(PCRLB)传感器节点的选择算法。该算法可随时间动态选择一个最优传感器集合并将均方根误差(RMSE)作为优化目标跟踪的性能。无线传感器网络中对于非线性、非高斯的动态系统,采用蒙特卡罗方法计算基于状态估计误差的一步向前Cramer-Rao下界,利用扩展H∞滤波器对目标状态和PCRLB进行逼近估计,并以此作为传感器节点选择标准以实现传感器的在线选择。基于Matlab工具箱的计算机仿真结果表明,相对于随机传感器节点选择算法和基于最近邻的传感器节点选择算法,基于后验-克拉美罗下界的目标跟踪传感器观测节点选择算法具有更好的有效性和优越性。For target tracking problems limited by resources like energy,bandwidth and storage under wireless sensor networks, a kind of sensor selection algorithm of Posterior Cramer-Rao Lower Bound （PCRLB） based on the extended H∞ filter is proposed, which can se- lect an optimal sensor set dynamically with time and optimize tracking performance in terms of Root Mean Square Error （RMSE）. Monte Carlo method is adopted to compute one-step look-ahead CRLB on the state estimation error in a nonlinear,possibly non-Gaussian and dynamic system with extended H∞ filter for approximate estimation of target state and PCRLB which are presented as the sensor selection criterion to realize the sensor options online. Simulation results with Matiab toolbox show that it has owned better effectiveness and superi- ority than stochastic sensor node selection algorithm and sensor node selection algorithm based on nearest neighbor.

关 键 词：目标跟踪 节点选择 均方根误差 后验-克拉美罗下界 扩展H∞滤波 无线传感器网络 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]