【作者】 孟伟；

【导师】 吴成东； 肖文栋；

【作者基本信息】 东北大学 ， 模式识别与智能系统， 2008， 硕士

【摘要】 无线传感器网络被认为是21世纪十大新兴技术之一,引起社会各界广泛关注。源定位是无线传感器网络重要应用之一,其中包括声源定位、气体污染源定位、灾难援救如火灾、地震灾害等。目前国内外的研究主要集中于单源定位,而源目标多以多源形式出现,因此展开多源定位研究具有重要实际意义。多源定位是一个难点问题,本文主要对基于能量的多声源、火灾定位算法以及定位问题中传感器节点部署和节点选择策略展开研究。本文主要研究工作与成果有：研究了基于能量的多声源定位算法。提出了一种多声源定位算法,算法思想是将节点从多个声源接收到的信号进行分解,将多源定位转化为单源定位问题,这样可以在保证定位精度的基础上减少算法计算复杂度,从而有效减少能量消耗。实验证明,该算法定位精度满足需求,算法能耗得到有效控制。研究了分布式多声源定位问题。分析了节点环的形成,该节点环使算法在各个节点内循环独立运行。实验证明,与集中式定位方法比较,分布式定位算法使网络通信消耗大幅度降低,且定位精度满足需求。研究了火灾模型及其着火点定位问题。提出了一种高斯多峰火点模型,该模型较好地描述了火灾初始阶段场景；研究了利用无线传感器网络节点自适应采样数据,基于EM-MR算法进行火点定位以及参数估计的方法,仿真实验证明该算法具有定位精度高、计算复杂度低等优点。研究了定位问题中节点部署和节点选择策略。根据克拉美-罗下限(CRLB)原理分析了节点部署策略；从信息论的角度分析了节点选择问题,提出了基于Mahalanobis距离方法和扩展卡尔曼滤波(EKF)算法的节点选择算法。该算法综合考虑了节点信息效用度和节点间通信消耗等因素,并应用扩展卡尔曼滤波算法进行状态概率更新。仿真实验结果证明了方法的有效性。更多还原

【Abstract】 Recently, wireless sensor network (WSN) which is considered as one of the ten most important technologies in 21st century has received a great deal of attention in the society. Source localization is one of its applications, including acoustic source localization, plume source localization and disaster rescuing such as fire and quake rescuing. So far, researches in literature mainly focus on single-source localization, however, targets always emerge in form of multiple sources. So research on multi-source localization has a significant practical meaning. In this thesis, energy-based multiple acoustic source and fire localization problems also sensor placement and sensor selection methods are addressed.The main research work and achievements of this thesis are as follows:The problem of energy-based multiple acoustic source localization is investigated. A multiple acoustic source localization algorithm is proposed. The basic idea of the method is to decompose the observed sensor data and consequently convert problem of multi-source localization into single source localization. The experimental results show that the algorithm can achieve satisfied localization accuracy with low computational complexity.Furthermore, the problem of distributed multi-source localization is researched. The forming of sensor node cycle is analyzed. Algorithm is run by each node and circles through the node cycle. The simulation results proved that compared with centralized way distributed localization algorithm has lower communication cost and satisfied localization accuracy.Moreover, the research of the problems of fire modeling and localization is carried out. A Gaussian multi-peak model is proposed. This model can well depict the real situation of fire. The fire locations and other parameters are estimated through adaptive sampling with sensor nodes using EM-MR algorithm. According to simulation results, the algorithm has high estimation accuracy and low computational complexity.In addition, sensor placement and selection strategies for localization are studied. Firstly, according to CRLB analysis, sensor placement strategy is proposed. Then based on the information theory, a heuristic method for sensor selection using Mahalanobis distance measurement and EKF algorithm is suggested. This method takes both information utility and communication cost into consideration and use EKF incrementally update the belief state. The method’s efficiency can be proved by the experimental results.更多还原