受限空间基于RSS指纹的定位算法研究

发布时间：2018-02-01 13:51

本文关键词： 无线定位 RSS指纹超分辨率受限空间 Kriging算法　出处：《中国矿业大学》2016年博士论文　论文类型：学位论文

【摘要】：随着物联网的广泛应用,越来越多的无线传感器被部署,众所周知,传感器除感知被测参数外,还需明确其位置信息,没有位置信息的感知信息是没有意义的,因此无线传感器的精确定位已成为物联网技术发展中的共性关键技术问题。在众多无线定位方法中,基于接收信号强度RSS(Received Signal Strength)指纹模型的定位方法因其部署简单、硬件成本低、适用范围广等特点受到研究者的关注。但该方法存在定位误差分布不均匀、RSS指纹建模工作量大和精度低、定位精度不高的问题,本论文围绕解决以上问题展开研究,研究内容主要包括:针对定位误差分布不均匀问题。研究RSS指纹模型定位方法的定位误差分布模型,得到误差分布规律。利用该误差分布规律得到不同精度区域的划分,为基于超分辨率原理的低分辨率区域划分奠定基础,同时利用该规律亦可进一步提高定位方法的定位精度。针对传统RSS指纹建模方法工作量大、精度低的问题。提出一种基于Kriging的RSS指纹生成算法,该算法首先对定位区域的信号强度空间场进行结构分析,在充分了解该场的性质前提下,选择理论变差函数模型;然后,在无偏估计和最小估计方差的准则下,利用观测值求解理论变差函数,得到相应的权值系数;最后对待估点上的RSS指纹利用Kriging估计器进行计算。相比传统逐点采集RSS指纹的方法,只需在定位区域内采集少量点的RSS指纹,就能准确估计出其它待估点的RSS指纹,解决传统RSS指纹建模工作量大、精度低的问题。针对RSS指纹定位算法定位精度不高的问题。提出了一种基于模糊核聚类FKC(fuzzy kernel clustering)支持向量机SVM(Support Vector Machine)的定位算法,在不提高定位算法计算复杂度的前提下,采用支持SVM升维技术,提高指纹分辨率,从而提升算法的定位精度,采用模糊技术加快支持向量机的训练过程。为进一步提高定位算法的定位精度,在上述定位算法的基础上,根据定位误差模型将定位区域划分成若干个低分辨率LR(Low-Resolution)区域方案,提出了利用超分辨原理的新的定位算法,使定位算法的定位精度最高可提升50%。以上的研究成果解决了传统基于RSS指纹模型定位方法存在的问题,不仅为物联网中无线传感器的精确定位提供技术支撑,还为无线精确定位方法的研究提供了新思路。
[Abstract]:With the wide application of the Internet of things, more and more wireless sensors are deployed, as everyone knows, in addition to the perception of sensor parameters to be measured, is required to clear the location information without location information awareness information is of no significance, therefore the accurate positioning of the wireless sensor has become a common problem in the development of networking technology in many. Wireless positioning method, the received signal strength based on RSS (Received Signal Strength) positioning method of fingerprint model because of its simple deployment, low cost, wide application range by researchers attention. But this method has the positioning error distribution is not uniform, RSS fingerprint modeling workload and low precision, the positioning accuracy is not high the problem, this paper focuses on solving above problems are studied, the research content mainly includes: Aiming at the problem of the uneven distribution of the positioning error method to study the localization of RSS fingerprint model. The positioning error distribution model, get the error distribution. Using the error distribution are divided into different precision area, low resolution region super resolution based on the principle of laying the foundation, positioning at the same time using the law can enhance the precision of positioning method. According to the workload of the traditional fingerprint RSS modeling method, the problem of low accuracy. This paper puts forward a RSS fingerprint generation algorithm based on Kriging algorithm, the first signal intensity space of the positioning area field structure analysis, to fully understand the nature of the field under the premise, the variogram model selection theory; then, the unbiased estimation and minimum variance criterion, using the observed value theory for solving the poor function, get the corresponding weight coefficient; finally to estimate RSS fingerprint point was calculated by using the Kriging estimator. Compared with the traditional methods of sampling point by point RSS fingerprint, only To collect RSS fingerprint of a few points in the positioning area, can accurately estimate the other estimated RSS fingerprint point, solve the traditional RSS fingerprint modeling workload, the problem of low accuracy. The RSS fingerprint positioning accuracy is not high. This paper proposes a fuzzy kernel clustering algorithm based on FKC (fuzzy kernel clustering) support vector machine SVM (Support Vector Machine) algorithm, without increasing the computational complexity of the localization algorithm under the premise, to support the SVM dimension raising technology, improve the fingerprint resolution, so as to enhance the positioning accuracy of the algorithm, the support vector machine to speed up the training process using fuzzy technology. In order to further improve the accuracy of the algorithm, in the based on the above algorithm, according to the positioning error model positioning area is divided into several low resolution LR (Low-Resolution) regional scheme is proposed by using the principle of the new super resolution An algorithm, the positioning accuracy of the positioning algorithm can improve the highest research results more than 50%. to solve the traditional RSS fingerprint model positioning method based on the problems, not only provide technical support for the accurate positioning of wireless sensor networking, provides a new idea for wireless location method research.

【学位授予单位】：中国矿业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP212.9;TP391.44;TN929.5

【相似文献】