基于CSI测距的轻量级指纹室内定位技术研究

发布时间：2018-08-13 10:52

【摘要】：随着Wi-Fi和移动设备的普及,基于Wi-Fi的室内定位受到了越来越多研究者的关注。接收信号强度指示(RSSI)作为一种主流的方案常用于基于测距的定位系统和指纹定位系统。近年来,有设备(如Intel 5300无线网卡)支持获取物理层的信道状态信息(CSI)。CSI是比RSSI能以更细粒度表征信号特性的一种指标,为基于Wi-Fi的室内定位技术开辟了新的空间,因而被广大研究者所关注。本文基于CSI进行室内定位研究,主要贡献如下:1.提出了一种基于CSI测距的轻量级指纹定位方案,RanLi Fi(Range-based Lightweight Fingerprint)方案,包括预处理阶段、离线阶段和在线阶段。预处理阶段在离线和在线阶段之前都会进行,用于处理得到稳定的数据,离线阶段构建室内指纹地图,在线阶段完成测距及指纹匹配并得到目标的位置。通过采用一种新型的环状采样点策略,可以将基于测距的定位方法和基于指纹的方案巧妙结合,通过指纹方案弥补测距方案精度不高的劣势,并通过测距方案降低指纹方案中指纹匹配的开销。具体而言,首先计算得到目标与无线接入点(AP)之间的距离d,然后将指纹匹配范围缩小到以无线接入点为圆心,半径最接近d值的环路上进行,最后将匹配程度最高的指纹对应的位置作为目标的位置。2.基于巴氏系数法,提出了一种B-AoE(Bhattacharyya based Average of Energy of Interest)指纹。通过对6种候选指纹方案从内在相似性和外在差异性两方面进行评价,发现B-AoE指纹很好地满足了相同位置足够相似,不同位置容易区分的特性要求。在RanLiFi的离线阶段和在线阶段,将获取到的CSI数据生成B-AoE指纹。3.为了准确测距,提出一种基于感兴趣能量的自适应传播模型,称为S-EoI(Self-adaptive and Energy of Interest based)模型。通过分析发现,在20MHz频宽的条件下,并不能准确分离视距路径和非视距路径信号,此外,即使在同一室内环境中,每次信号传输的衰减程度也是有差异性的。基于此,本文定义了一种稳定性好且能反映环境变化的感兴趣能量值,并给出了一种自适应的方法计算每次信号传输的路径损耗系数,基于二者进行距离计算。S-EoI模型最大的优点在于无需对不同室内环境进行参数训练,自适应的路径损耗系数可准确反映当次信号的衰减情况。4.实现了基于CSI测距的轻量级指纹定位系统,Ran LiFi系统。系统包括数据获取模块、预处理模块、指纹地图构建模块和定位导航模块,并支持容错机制和反馈机制,用于保证系统的实用和稳定性。数据获取模块采用Intel 5300无线网卡获得原始的CSI数据,预处理模块、指纹地图构建模块和定位导航模块分别对应实现了Ran Li Fi方案中预处理阶段、离线阶段和在线阶段中的方法。在指纹地图构建模块,采用滑动窗口机制控制数据量并生成指纹,而在定位导航模块,为了满足实时性,以1s为时间片采集数据生成指纹用于进行匹配。5.从多个角度对RanLiFi定位系统进行了性能评价。实验表明,本文所提S-Eo I测距方案平均误差为1.7米,B-AoE指纹方案在98%的情况下可正确进行位置区分;在容错机制支持下,若以1.5米为间隔进行采样点设置,RanLi Fi系统可取得平均0.8米的定位精度,时间性能达到了秒级。
[Abstract]:With the popularity of Wi-Fi and mobile devices, indoor positioning based on Wi-Fi has attracted more and more researchers'attention. Received Signal Intensity Indicator (RSSI), as a mainstream scheme, is often used in ranging-based positioning systems and fingerprint positioning systems. In recent years, devices (such as Intel 5300 wireless adapters) have supported the acquisition of channel states in the physical layer. Information (CSI). CSI is an indicator that can characterize signal characteristics with finer granularity than RSSI. It opens up a new space for Wi-Fi-based indoor positioning technology and attracts the attention of many researchers. This paper studies indoor positioning based on CSI. The main contributions are as follows: 1. A lightweight fingerprint positioning scheme based on CSI ranging, RanLi Fi (RanLi Fi) Range-based Lightweight Fingerprint) scheme, including pre-processing phase, offline phase and online phase. The pre-processing phase will be carried out before offline and online phase, used to process stable data, offline phase to build indoor fingerprint map, online phase to complete ranging and fingerprint matching and get the location of the target. A novel ring sampling point strategy is proposed, which combines the ranging-based localization method with the fingerprint-based scheme. The fingerprint scheme compensates for the low precision of the ranging scheme and reduces the cost of fingerprint matching in the fingerprint scheme. Then the fingerprint matching range is reduced to a loop with the wireless access point as the center and the radius closest to the D value. Finally, the position of the fingerprint with the highest matching degree is taken as the target position. 2. Based on Papanico coefficient method, a B-AoE (Bhattacharyya based Average of Energy of Interest) fingerprint is proposed. The B-AoE fingerprint can satisfy the requirement that the same location is sufficiently similar and different locations are easy to distinguish. In the off-line and on-line phases of RanLiFi, the obtained CSI data are generated into B-AoE fingerprint.3. In order to accurately measure the distance, a base is proposed. The adaptive propagation model of energy of interest is called S-EoI (Self-adaptive and Energy of Interest based) model. It is found that the line-of-sight path and non-line-of-sight path signals can not be separated accurately under the condition of 20 MHz bandwidth. Moreover, the attenuation degree of each signal transmission is different even in the same indoor environment. Based on this, this paper defines an energy of interest which is stable and can reflect the changes of the environment, and gives an adaptive method to calculate the path loss coefficient of each signal transmission. The biggest advantage of S-EoI model is that it does not need to train the parameters of different indoor environments and adaptively uses the path loss coefficient. The system includes data acquisition module, pre-processing module, fingerprint map construction module and positioning navigation module, and supports fault-tolerant mechanism and feedback mechanism to ensure the practicability and stability of the system. The data acquisition module uses Intel 5300 wireless network card to acquire the original CSI data, preprocessing module, fingerprint map construction module and positioning and navigation module respectively implement the methods in the pre-processing stage, off-line stage and online stage of Ran Li Fi scheme. The performance of RanLiFi positioning system is evaluated from various angles. The experimental results show that the proposed S-Eo I ranging scheme has an average error of 1.7 meters, and the B-AoE fingerprint scheme can be correctly positioned in 98% of the cases. With the support of fault-tolerant mechanism, RanLi-Fi system can achieve an average positioning accuracy of 0.8 meters, and its time performance reaches second level.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN92

【相似文献】