基于广义标签多伯努利滤波的可分辨群目标跟踪算法  

Multiple Resolvable Groups Tracking Using the GLMB Filter

作　　者：朱书军 刘伟峰 崔海龙[1] ZHU Shu-Jun;LIU Wei-Feng;CUI Hai-Long(Institute of Systems Science and Control Engineering, School of Automation, Hangzhou Dianzi University, Hangzhou 310018;Science and Technology on Information System Engineering Laboratory, Nanjing 210007)

机构地区：[1]杭州电子科技大学自动化学院系统科学与控制工程研究所,杭州310018 [2]信息系统工程重点实验室,南京210007

出　　处：《自动化学报》2017年第12期2178-2189,共12页Acta Automatica Sinica

基　　金：国家自然科学基金(61333011;61271144;61273170;61402140);浙江省自然科学基金(LY15F030020);杭州电子科技大学控制科学与工程重点学科资助~~

摘　　要：针对杂波条件下可分辨群目标的状态估计、目标个数与子群个数估计问题,提出了一种基于标签随机有限集(Label random finite set,L-RFS)框架下的可分辨群目标跟踪算法,该算法主要包括两个方面:可分辨多群目标动态建模和多群目标的跟踪估计.本文工作主要包括:1)结合图论中的邻接矩阵对可分辨群目标运动进行动态建模.2)利用基于L-RFS的广义标签多伯努利滤波(Generalizes label multi-Bernoulli,GLMB)算法对目标的状态和个数进行估计,并且通过估计邻接矩阵得到群的结构和个数估计.3)通过个数不同、结构不同的三个子群目标在二维平面分别做线性和非线性运动进行算法验证.仿真分析表明本文算法能够准确估计出群目标中各目标的状态、个数以及子群的个数,并且能获得目标的航迹估计.Aiming at the estimation of states, the number of targets and subgroups, a resolvable group target traking algorithm is proposed based on the framework of label random finite set （L-RFS）. The proposed algorithm focus on two aspects： dynamic modeling and tracking estimation for multiple resolvable group targets. Specifically, in the first step, the adjacent matrix is fused in the dynamic models. In the second step, the estimated state sets of the targets and the number of targets are estimated by using the generalized labeled multi-Bernoulli （GLMB） filter in the L-RFS framework. Finally, from the estimated adjacent matrix, the structures and number of subgroups are shown. Two experiments of a linear system and a nonlinear system, which involve three groups of targets with different shapes and structure, are given to show that the given algorithm is effective in estimating the resolvable group tareets.

关 键 词：可分辨群目标跟踪 广义标签多伯努利滤波 邻接矩阵 随机有限集 图论 

分 类 号：O157.5[理学—数学] TN713[理学—基础数学]