分布式Unscented粒子滤波跟踪

Target tracking using distributed Unscented particle filter in sensor network

提出了一种新的分布式粒子跟踪算法,该算法主要考虑传感网络能量受限、通信受限等特性,改善了通常的分布式粒子滤波粒子数目大、节点间信息交换多的弊端,能够用较少的节点计算得到对机动目标更好的跟踪结果,实现了改进的分布式粒子滤波(DUPF)。DUPF算法的主要思想是利用UnscentedKalman滤波改进分布式粒子滤波算法形成一个建议分布,用来生成粒子分布,在这个基础上,通过分布式粒子滤波实现目标的在线跟踪。仿真实验表明,和分布式粒子滤波相比,DUPF只需要其25%的粒子数目就能达到同样的跟踪精度,即可用较少的节点和通信消耗,实现高精度的目标跟踪。

A new distributed target tracking algorithm,Distributed Unscented Particle Filter（DUPF）, is proposed to improve the usual distributed particle filter methods with more particles and more information communication between the two nodes. In consideration of the enengy-limited sensor network and the imperfect communication,a few node calculations are used in this thesis to get a better tracking results for manoeuvering targets. The Unscented Kalman Filter（UKF） in new DUPF is used to improve the particle filter to generate the proposed particle distribution, so the on line tracking for a target can be realized by the DUPF. A simulation experiment indicates that the number of particles needed by DUPF is only 25 % that of the common distributed particle filter,which shows that DUPF has gotten more accurate tracking results with less communication nodes and energy cosumption.