无线传感网跟踪任务中的目标运动模型估计与节点调度  

Target Motion Model Estimation and Sensor Scheduling for Tracking Tasks in Wireless Sensor Networks

作　　者：王祺尧 冯辉[1,2] 胡波 罗灵兵[1,2] WANG Qiyao;FENG Hui;HU Bo;LUO Lingbing(Department of Electronic Engineering, School of Information Science and Technology, Fudan University, Shanghai 200433, China;Research Center of Smart Networks and Systems, School of Information Science and Technology, Fudan University, Shanghai 200433, China)

机构地区：[1]复旦大学信息科学与工程学院电子工程系,上海200433 [2]复旦大学信息科学与工程学院智慧网络与系统研究中心,上海200433

出　　处：《复旦学报（自然科学版）》2019年第2期221-230,共10页Journal of Fudan University：Natural Science

基　　金：国家自然科学基金(61501124);国家重点研发计划(213)

摘　　要：在无线传感器网络进行目标跟踪的过程中,合理的节点调度算法可以兼顾跟踪精度和能量消耗,延长网络的使用寿命.然而,当目标运动模型未知时,难以实现高效的节点调度.为解决目标运动模型未知场景下的跟踪问题,本文将监控区域中的目标移动和传感器观测建模为隐马尔可夫模型(HMM),并提出了HMMQMDP算法,把问题分解为运动模型估计和节点调度两个阶段:运动模型估计阶段是根据传感器采集的观测序列估计目标状态转移概率;节点调度阶段则被建模为部分可观测马尔可夫决策过程(POMDP),综合考虑决策的短期和长期损失,应用QMDP算法近似求解优化策略.仿真结果表明:该算法可以根据观测样本有效地学习和估计目标运动模型,提升节点调度算法的效果.In the process of target tracking of wireless sensor networks, a well-designed sensor scheduling algorithm can consider the tracking accuracy and energy consumption and extend the service life of the network. However, it is difficult to achieve efficient sensor scheduling when the target motion model is unknown. In order to solve the tracking problem when the target motion model is unknown, in this paper, the target movement and sensor observations in the surveillance area are modeled as hidden Markov model (HMM). HMM-QMDP algorithm is proposed to decouple the problem into two stages: motion model estimation and sensor scheduling. At the motion model estimation stage the target state transition probabilities are estimated based on the observed sequences acquired by sensors. The sensor scheduling stage is modeled as partially observable Markov decision process (POMDP), considering both the short- and long-term losses of decision making, and QMDP algorithm is applied to approximate the optimization strategy. The simulation results show that this algorithm can effectively learn and estimate the target motion model based on the observed samples and enhance the effect of sensor scheduling algorithm.

关 键 词：无线传感器网络 运动模型估计 节点调度 隐马尔可夫模型 部分可观测马尔科夫决策过程 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]