【作者】 石琴琴；

【导师】 李德仁；

【作者基本信息】 上海交通大学 ， 模式识别与智能系统， 2009， 博士

【摘要】 无线传感器网络是由大量随机分布的,集成有传感器、数据处理单元和通信模块的微小节点通过自组织的方式构成的网络。近年来,无线传感器网络技术取得了飞速发展,在工农业、军事国防、环境监测等许多领域都有着广阔的应用前景。无线传感器网络作为一个全新的研究领域,向科技工作者提出了大量具有挑战性的课题,节点自定位问题就是其中之一。实现节点的自身定位是无线传感器网络进行目标识别、监控、跟踪等众多应用的前提,节点自定位问题的研究已经成为无线传感器网络研究领域非常重要的一个研究方向。无线传感器网络中的传感器节点往往布设规模大且随机布放,又因其能量有限,可靠性差,无线模块的通信距离有限等诸多限制条件,对用于实现其自定位的相关技术和算法提出了很高的要求,使用常规的GPS方法定位或现场工程测量方法定位都是不恰当的。应用于无线传感器网络节点的自定位系统须要适合其自身的特点,通常要求具备自组织性、健壮性、能量高效性以及分布式计算等特性。本文的研究工作围绕无线传感器网络节点的自定位这一课题展开,在对近年来该领域所取得的研究成果进行深入学习与理解的基础上,将研究目标确定为探索在现有的节点硬件条件下,从算法研究的角度来解决节点的自定位问题。本文的主要贡献包括以下内容:1.在许多无线传感器网络的应用中,为了便于计算及描述,往往将节点的布设区域抽象简化为二维空间,也因此将节点自定位问题简化在二维空间层面进行解决。本文提出一个分布式的节点自定位系统,适用于可将定位环境近似为二维平面的应用场景,该系统包含通常的三个定位步骤:①未知节点与信标节点之间的距离估计,②未知节点初步自定位计算,③未知节点位置求精计算。与以往的分布式节点自定位系统相比,该系统在节点初步自定位计算阶段及位置求精计算阶段所使用的算法上面进行了创新性的研究工作:(1)在节点初步自定位计算阶段提出使用算法组合Min-max+LI来确定节点的位置。Min-max算法是既有的用于无线传感器网络节点自定位计算的经典算法,LI算法(平面距离交会算法)是原应用于工程测量领域的平面控制点加密算法,两种算法的组合使用可互补其在定位计算中的优缺点。(2)在节点位置求精计算阶段提出使用SD算法(最速下降算法),将节点位置求精过程转化为求非线性方程组最优解的过程来实现优化。以理论分析结合仿真实验的方式论证所提出的系统的可行性与有效性。实验结果表明,综合考量定位精度、定位覆盖率、计算量、通信负荷等方面,本文提出的系统优于一些前人研究中提出的典型的分布式节点自定位系统。2.针对无线传感器网络节点的自定位问题,尽管已经提出了多种解决方案,但这些方案多数是将传感器节点的布设区域假设或简化为二维平面来进行设计和评估的。本文分析了在无线传感器网络的一些实际应用中实现节点在三维空间自定位的必要性,并就定位系统的模型设计及算法使用方面进行创新性研究:(1)文中论述了使用移动信标节点布设机制的优越性,提出一个采用空载的移动信标节点实现全网信号覆盖的节点三维自定位系统。该定位系统的运作模式为:布设于应用区域内的未知节点被动地接收由空载的移动信标节点发射的UWB信号来获取定位所需信息,并藉此采用TOA测距技术测量其与信标节点之间的空间距离,而后使用基于距离测量的三维位置计算算法实现分布式的自定位。(2)提出采用原应用于工程测量领域的空间控制点加密算法――SDI算法(空间距离交会算法)作为系统中使用的三维位置计算算法,来实现分布式的节点三维自定位计算。论述了所提出的定位系统模型的一些特点及优点,并以理论分析结合仿真实验的方式论证使用SDI算法实现无线传感器网络节点三维自定位的可行性与有效性。依据仿真实验结果,主要可得出两方面的结论:一方面是SDI算法对测距误差敏感,在测距精确的情况下是一种表现良好的三维位置计算算法,因为UWB TOA测距技术可提供精确的空间测距,故本文提出的定位系统整体可获得良好的定位表现;另一方面是当测距误差较大时,SDI算法在高程定位精度方面优于其他两种典型的位置计算算法Lateration与Min-max的表现,但Min-max算法在平面定位精度方面具有最稳定的表现,因此,当在应用环境或其他制约条件的限制下,只能采用较粗略的测距技术(如RSSI)时,SDI算法与Min-max算法可组合使用,分别用于获得节点三维位置的高程坐标分量与平面坐标分量。更多还原

【Abstract】 Wireless sensor networks (WSNs) are self-organized networks composed of a great deal of small nodes randomly distributed in the sensing area. A certain kind of sensor, a data processing unit and a communication module are the common components of a node in the WSNs. In the recent few years, WSNs have accessed rapid development and reveal vast application prospects in many fields, such as military affairs, industry, agriculture, environment, medical treatment, etc. As a brand-new research field, WSNs pose many challenging topics to the research workers. The problem of node localization, that is, determining where a given node is physically or relatively located in a network, is one of the challenging tasks and yet extremely crucial for many applications. The self-localization of nodes in WSNs has been the basis for most of the applications, such as taget recognization, monitoring, and tracking. The node localization has been an important and critical research direction in the study of WSNs.Because of the large and random deployment of the sensor nodes in WSNs and some restrictive conditions of the sonsor nodes themselves, such as short battery life, low reliability, and limited wireless communication distance, WSNs pose high demands on the techniques and algorithms which are used for the node localization. Some routine localization methods, such as GPS and surveying method, are not suited for the node localization in WSNs. The node localization systems for WSNs are commonly requied self-orgornized, robust, energy-efficient, and computation-distributed.Our study is focused on the node localization in WSNs, and the main object is to solove the node localization problem from the aspect of algorithm design, under the current hardware conditions of sensor nodes. The main contribution of this dissertation includes: 1. In many WSNs’applications, for computation simplicity and ease of presentation, the sensing areas are commonly assumed flat and the node localization problem is solved in two-dimensional (2D) space only. In this paper, we present a distributed node localization system for WSNs, which fits for the application scenarios where the localization environment can be simplified as a 2D space. The system includes three common steps:①determine node-beacon distances,②compute node positions, and③refine the positions. Compared with other current distributed node localization systems, our system innovates in the second step and the third step.(1) An algorithm combination Min-max+LI is proposed to be the position derivation algorithm. Min-max is a representative node position derivation algorithm for WSNs, and LI is a control point densification algorithm used in engineering survey field originally. The combination usage of the two algorithms is an optimisation to let them make up each other in terms of their respective advantages and disadvantages in the localization computation.(2) SD method is presented for the refinement, and thus, the refinement procedure is transformed into a solution procedure for a nonlinear equation system. The feasibility and effectivity of our sytem are demonstrated through analysis in theory and simulation. Results show that our proposed sytem can perform better than some representative distributed node localization schemes presented in previous researches in terms of the trade-off among accuracy, coverage, computation cost, and communication overhead.2. Most previous approaches on node localization are designed and evaluated considering only 2D applications, but sometimes it is unreasonable to just simplify the node localization problem to 2D level due to the complexity of the actual application scenarios. The necessity of solving the node localization problem considering three-dimensional (3D) environments is discussed, and some innovative work are done on the system model design and algorithm application of 3D node localization in WSNs.(1) The advantages of using the mobile beacon mechanism are demonstrated, and a 3D node localization system using an aircraft-carried mobile beacon for the localization signal coverage is proposed. The working modes of the system is that the unknown nodes in the sensing area receive the UWB signals from the mobile beacon passively and measure the distance to the mobile beacon using TOA technique, and then, the nodes localize themselves locally.(2) SDI is proposed as the 3D node position derivation algorithm, and it is a spatial control point densification algorithm used in engineering survey field originally.Some features of our proposed system are discussed, and the feasibility and effectivity of using SDI as the position derivation algorithm are demonstrated through analysis in theory and simulation. Based on the results, mainly two conclusions are obtained. One is that SDI is sensitive to the node-beacon distance error, and hence, it is a suitable 3D node position derivation algorithm when the distance measurement is precise. Since UWB TOA technique can provide precise distance measurements, our system can perform well in general. The other is that SDI performances much better than the other two representative algorithms, Min-max and Lateration, in terms of the vertical positioning error under all the range situations while Min-max has the best performance on horizontal positioning when the range precision is poor. Therefore, in the cases where only corse ranging results can be obtained (e.g., using RSSI for ranging), both SDI and Min-max should be considered for 3D node position derivation, in which, Min-max is used to derive the horizontal position of a node while SDI is used to drive the vertical position.更多还原