【作者】 张丽颖；

【导师】 吴帆；

【作者基本信息】 北京邮电大学 ， 电子科学与技术， 2019， 硕士

【摘要】 日益增长的位置服务(LBS,Location Based Serve)商业需求推动了近年来诸多位置估计技术的发展。LBS广泛应用于大型商场、医院、仓库、停车场等场合并在人流监控、医疗救助、货物调配等方面有着广阔的应用前景。一方面,复杂多变的外界环境及多径效应使得位置估计精度受到严重影响,另一方面,针对商场等的三维空间位置估计需求使得位置估计系统的部署难度和密度增加。因此,适应于三维空间环境的位置估计技术和高效的系统部署成为研究重点。本文以三维空间位置估计问题为背景,研究了影响位置估计精度的关键因素:定位系统的天线结构和位置估计算法。本文提出GSLS(Gradient-Based Simplified Least Squares)方法用于角度估计,并将该方法应用到球形天线位置估计系统中实现三维(3D,Three Dimension)空间位置估计。本文的主要研究内容包括以下几个方面:第一,对现有的TSLS(Three-Stage Simplified Least Squares)方法进行改进,提出了GSLS方法实现角度估计。TSLS方法基于扇形天线实现角度估计,但是其AOA(Angle Of Arrival)估计过程需要不断根据环境信噪比变化调整角度融合(DFU,DoA Fusion)权重以达到最小AOA估计误差,这在实际中是繁琐且难以实现的。本文提出的GSLS方法通过构建接收信号强度(RSS)与AOA间的联合概率密度函数并求解,取代了TSLS方法中的DoA(Direction OfArrival)估计及融合步骤,减小了DFU过程中权重对AOA估计精度的影响,降低了AOA估计误差;同时GSLS采用迭代法求解最优估计,与TSLS中的非迭代法相比,更易收敛到最优估计结果,AOA估计精度更高。第二,将GSLS方法引入球形天线的空间定位系统中实现了基于球形天线的AOA估计过程;为了验证系统的有效性,本文通过理论分析和仿真实验对自由空间和存在多径的室内空间的球形天线的AOA估计及位置估计性能进行了分析。结果表明:部署相对稀疏条件下(K≤S 5),球形天线位置估计系统在100m*100m*20m的室内空间位置估计误差均值为3.29m,水平位置估计误差均值为2.69m,AOA估计误差均值为6.76deg,其中方向角误差均值为1.16%,方位角误差均值为2.85%,AOA估计及位置估计误差较低,能够满足空旷室内空间如体育馆、大型仓库的定位需求。第三,针对室内定位需求,提出将基于球形天线的AOA估计结果作为指纹进行位置估计。文章通过典型环境下的位置估计实验,分析得到AOA指纹定位对设备异质性不敏感但对环境扰动敏感。由此,本文提出应用Kalman滤波器进行RSS值预处理,仿真表明,经过Kalman滤波器的处理,基于AOA的指纹位置估计方法对环境扰动和设备异质性不敏感,位置估计精度比基于RSS的指纹位置估计方法提高24.7%。同时由于AOA指纹的向量维度更小,位置估计中的数据库指纹匹配过程耗时降低25%,位置估计实时性更好。更多还原

【Abstract】 The growing business demand of Location Based Serve(LBS)has promoted the development of many location estimation technologies in recent years.LBS is widely used in shopping malls,hospitals,warehouses,parking lots and other occasions,and has broad application prospects in flow monitoring,medical assistance,cargo allocation and so on.On the one hand,the location estimation accuracy is seriously affected by the complex and changeable environment and multipath effect.On the other hand,the deployment difficulty and density of the location estimation system are increased because of the demand for three-dimensional spatial location estimation such as shopping malls.Therefore,location estimation technology and efficient system deployment adapted to three-dimensional space environment become the focus of research.In this thesis,the antenna structure and position estimation algorithm of position estimation system,which are the key factors affecting the accuracy of position estimation,are studied.In this thesis,a GSLS(Gradient-Based Simplified Least Squares)method is proposed for angle estimation,and the method is applied to spherical antenna position estimation system to realize three-dimensional(3D,Three Dimension)spatial position estimation.The main contents of this thesis include the following aspects:Firstly,the existing TSLS(Three-Stage Simplified Least Squares)method is improved,and the GSLS method is proposed to realize angle estimation.TSLS method realizes angle estimation based on sector antenna,but its AOA(Angle of Arrival)estimation process needs to constantly adjust the weight of angle fusion(DFU,DoA Fusion)according to the change of environmental signal-to-noise ratio to achieve the minimum AOA estimation error,which is cumbersome and difficult to achieve in practice.The proposed GSLS method replaces the Direction of Arrival estimation and fusion steps in TSLS method by constructing a joint probability density function between received signal strength(RSS)and AOA,reduces the influence of weight on the accuracy of AOA estimation in DFU process,and reduces the error of AOA estimation.At the same time,GSLS uses iterative method to solve the optimal estimation,and non-iterative method in TSLS.It is easier to converge to the optimal estimation result and has higher accuracy than AOA estimation.Secondly,the GSLS method is introduced into the spherical antenna space positioning system to realize the AOA estimation process based on the spherical antenna.In order to verify the effectiveness of the system,the AOA estimation and position estimation performance of spherical antenna in free space and multi-path indoor space are analyzed through theoretical analysis and simulation experiments.The results show that under the condition of relative sparse deployment(K<5),the mean error of indoor space position estimation of spherical antenna position estimation system is 3.29M at 100m*100m*20m,the mean error of horizontal position estimation is 2.69m,and the mean error of AOA estimation is 6.76deg.The mean error of direction angle,azimuth and AOA are 1.16%,2.85%,respectively.The error of A estimation and location estimation is low,which can meet the positioning requirements of open indoor space such as gymnasium and large warehouse.Thirdly,aiming at indoor positioning requirements,AOA estimation results based on spherical antenna are proposed as fingerprints for position estimation.Based on the position estimation experiments in typical environments,it is concluded that AOA fingerprint localization is insensitive to device heterogeneity but sensitive to environmental disturbances.The simulation results show that the fingerprint location estimation method based on OA is insensitive to environmental disturbance and equipment heterogeneity,and the accuracy of location estimation is 24.7%higher than that of fingerprint location estimation method based on RSS.At the same time,because the vector dimension of AOA fingerprint is smaller,the time-consuming of database fingerprint matching in location estimation is reduced by 25%,and the real-time performance of location estimation is better.更多还原