【作者】 任维政；

【导师】 邓中亮；

【作者基本信息】 北京邮电大学 ， 微电子与固体电子学， 2011， 博士

【摘要】 无线传感器网络(Wireless Sensor Networks, WSN)节点定位技术是无线传感器网络的重要支撑技术之一,传感器节点的位置信息在无线传感器网络的诸多应用领域中扮演着重要的角色。本论文面向大型建筑室内外定位、导航服务的实际应用需求,针对无线传感器网络节点的自身定位问题进行了研究。在调研和分析已有的各种定位算法的基础上,从非测距定位方法、测距定位方法和移动节点的跟踪算法三个方面分别提出了基于自适应粒子群优化(Adaptive Particle Swarm Optimization, APSO)的WSN定位算法、基于差分似然估计(Difference Maximum Likelihood Estimation, DMLE)的WSN节点定位算法和基于粗糙神经网络的自适应多模交互式(Rough Set and Neural Network Adaptive Interacting Multiple Model, RSAIMM)的WSN跟踪算法。为了验证本论文提出算法的有效性,开发了大型建筑室内及周边定位原型系统,对本论文提出的算法进行了验证。(1)基于APSO的WSN定位算法针对无线传感器网络要求低成本、低功耗的要求,为了克服现有非测距WSN节点定位方法计算量大、定位精度受节点密度影响较大的缺点,本论文将粒子群理论引入非测距定位算法中,提出了一种基于自适应粒子群优化的WSN定位算法。仿真结果表明,该算法在不需要增加任何额外硬件设备和通信负荷的情况下,比DV-Hop算法定位精度提高了20%以上,定位精度受节点密度的影响与DV-Hop算法相比明显减小。(2)基于DMLE的WSN节点定位算法为了克服接收信号强度测量误差对无线传感器网络节点自身定位精度的影响,在对极大似然估计定位算法和接收信号强度指数(Received Signal Strength Indication, RSSI)模型分析的基础上,定义了个体差异差分系数、距离差分系数和距离差分定位方程,将测距差分修正和极大似然估计相结合提出了一种RSSI测距差分修正极大似然估计定位算法。仿真结果表明,该算法有效抑制了由于环境变化所引起的RSSI测量误差,定位精度可达2.5m以下。(3)基于RSAIMM的WSN跟踪算法为了克服复杂环境和运动模式对节点跟踪带来的影响,进一步提高跟踪精度,采用基于粗糙神经网络的双滤波器并行结构,提出了基于粗糙神经网络的自适应跟踪算法。当目标在机动和非机动之间变化时,粗糙神经网络在线自动输出匹配特征值,以足够准确的系统方差适应目标的运动变化并保持对目标状态的高精度跟踪。仿真结果表明,该算法与传统的多模交互式跟踪算法相比,跟踪精度提高了23.15%。(4)大型建筑室内及周边定位原型系统为了验证提出的定位、跟踪算法的有效性,给出了系统结构设计、功能设计和软件架构的整体方案,在此基础上,开发了面向定位跟踪应用的WSN验证平台,设计了客户服务终端、无线传感器网络节点以及系统服务器软件。原型系统定位、跟踪性能达到了预期的实验效果,证明了本论文提出定位算法的有效性,为下一步的研究奠定了基础。更多还原

【Abstract】 Wireless sensor networks (WSN) nodes localization technology is one of the important supporting technology of WSN, and location information of sensor nodes plays an important role in many WSN application areas. Intended for location and navigation application on requirements of indoor and peripheral areas of large buildings, the issues of WSN node’s own localization are presented in this paper. After research and analysis of a variety of existing localization algorithms, a WSN localization algorithm based on adaptive particle swarm optimization (APSO), WSN node localization algorithm based on difference maximum likelihood estimation(DMLE), and WSN tracking algorithm based on rough set and neural network adaptive interacting multiple model(RSAIMM) are put forward respectively, according to free ranging localization method, RSSI ranging localization method and tracking algorithm of moving node. Localization demonstration systems of indoor and peripheral areas of large building are developed to verify the validity of algorithms in this paper.(l)WSN location algorithm based on APSOIn response to the requirement of WSN for low cost and low power consumption, and in order to overcome existing large scales of computing and location accuracy easily affected by the node density of existing non-ranging WSN node positioning method, the algorithm of particle swarm theory into the non-locating position is induced coming up with WSN localization algorithm based on APSO. Without additional hardware device and communication load, positioning accuracy of the new algorithm is improved more than20%than DV-Hop algorithm, and positioning accuracy affected by the node density is significantly reduced compared with the DV-Hop algorithm.(2) Differential likelihood estimation localization algorithm based on DMLETo overcome the effect of received signal strength measurement errors on WSN node itself positioning precision, and based on the foundation of maximum likelihood estimation positioning algorithm and RSSI model analysis, a RSSI ranging difference amendment Maximum likelihood estimation positioning algorithm is presented, by defined individual difference coefficient, distance difference coefficient and distance difference positioning equation, the combination difference amendment, and maximum likelihood estimation. The RSSI measurement error due to environmental changes and positioning accuracy, effectively inhibited in the algorithm, can reach within2.5m.(3)WSN tracking algorithm based on RSAIMMIn order to overcome the effect of the complex environment and movement patterns on node tracking and further to increase tracking accuracy, adaptive tracking algorithm based on rough neural network introduced by adapting dual-filter parallel structures based on rough neural network. When the goals change between the motor and non-motor, rough neural network on-line can automatically output matching eigenvalues with the change of sufficient and accurate system variance matching target motion and high accuracy tracking of the target state maintained. The tracking accuracy of the algorithm can be improved by23.15%, compared with the traditional multi-mode interactive tracking algorithms.(4) Indoor and peripheral areas of large building location prototype systemIn order to verify the effectiveness of the proposed localization and tracking algorithm, the overall program of the system structure design, functional design and software architecture are developed. On this basis, a WSN verification platform for location and tracking applications is developed, by designing customer service terminals, nodes in wireless sensor networks, and server software. Prototype system’s location and tracking performance has achieved the desired effect, proved the validity of the thesis proposed location algorithm, and laid the foundations for future study.更多还原