基于压缩感知和BP神经网络的无线传感器网络定位算法研究

发布时间：2018-03-04 06:37

本文选题：无线传感器网络　切入点：BP神经网络　出处：《南京邮电大学》2016年硕士论文　论文类型：学位论文

【摘要】：无线传感器网络(WSN)可以被广泛应用于农业、环境生态监测、智能家居、目标跟踪等领域中,有效而准确地定位传感器节点对于帮助它们完成各项工作是至关重要的。本文重点研究非视距(NLOS)环境下基于BP神经网络和压缩感知(CS)的WSN节点定位:(1)介绍了WSN节点定位的基础知识、节点位置基本的计算方法和定位性能的评价指标,研究了几个比较经典的基于测距的和基于非测距的定位算法,重点阐述了本文应用的基于测距的RSSI算法和TDOA算法的原理。(2)针对WSN节点定位中存在NLOS的问题,结合BP神经网络具有较强的抗噪声能力、良好的数据逼近能力以及能并行处理数据的特点,提出了利用BP神经网络优化NLOS环境下WSN节点定位的方法。首先分析了误差的来源,然后合理地选择训练算法和神经网络结构,提高了BP神经网络的收敛速度,最后用可靠节点对神经网络进行训练,并且利用训练好的神经网络实现对未知节点的定位。仿真结果表明,该方法能够有效地抑制NLOS误差,定位精度优于传统的Taylor算法和Chan算法。(3)针对WSN节点定位中存在的功耗大、定位算法复杂的问题,提出了CS和BP神经网络相结合进行WSN节点定位的方法。首先利用压缩感知进行初始定位,确定网格中是否存在节点,如果存在,则应用BP神经网络修正测量误差,然后使用三边测量法计算出未知节点的具体位置。本算法可以同时进行多目标定位,克服了传统的压缩感知方法只能在网格中心进行定位的缺点,降低了RSSI定位带来的较大的误差,同时大大减小了需要上传的数据量,在降低功耗的同时提高了定位精度,可以适用于条件恶劣的室外环境。
[Abstract]:WSNs can be widely used in agriculture, ecological monitoring, smart home, target tracking and so on. It is very important to locate sensor nodes effectively and accurately in order to help them accomplish all kinds of work. In this paper, we focus on the study of WSN node positioning: 1) based on BP neural network and compression sensing (CSS) in the environment of non-line-of-sight (NLOS). The introduction of WSN section is given. The basic knowledge of Point-Positioning, The basic calculation method of node location and the evaluation index of location performance are studied, and several classical localization algorithms based on ranging and non-ranging are studied. The principle of RSSI algorithm based on ranging and the principle of TDOA algorithm applied in this paper is emphasized. Aiming at the problem of NLOS in WSN node location, the BP neural network has strong anti-noise ability. The method of optimizing the location of WSN nodes in NLOS environment using BP neural network is put forward. The source of errors is analyzed firstly, and then the training algorithm and neural network structure are reasonably selected. The convergence speed of BP neural network is improved. Finally, the reliable nodes are used to train the neural network, and the trained neural network is used to locate the unknown nodes. The simulation results show that the proposed method can effectively suppress the NLOS error. The localization accuracy is better than the traditional Taylor algorithm and Chan algorithm. This paper presents a method of WSN node localization based on CS and BP neural network. Firstly, the initial localization is made by compression perception, and if there are nodes in the grid, the measurement error is corrected by BP neural network. Then the location of unknown nodes can be calculated by means of trilateral measurement. This algorithm can locate multiple targets simultaneously, which overcomes the disadvantage that traditional compression sensing can only locate in the center of the grid. The large error caused by RSSI positioning is reduced, and the amount of data needed to be uploaded is greatly reduced, and the positioning accuracy is improved while the power consumption is reduced, which can be applied to the outdoor environment with poor conditions.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP212.9;TN929.5

【相似文献】