基于移动设备的室内定位与导航

发布时间：2018-08-13 11:09

【摘要】：随着无线网络技术的发展与现代化城市建设的快速发展,基于位置感知的服务(Location Based Services,LBS)在个人位置服务、医疗领域、电子商务、紧急救援、智能家居等多方面显示出巨大的活力,是近年来备受关注的一大研究热点。而具有较高精度的室内定位与导航技术是实现LBS的基础与关键。传统的全球定位系统GPS(GlobalPositioningSystem)和蜂窝移动通信技术在室外时拥有较高的定位精度,但是在室内环境中GPS信号会受到遮挡,导致定位精度大大下降。相比于已有的室内定位技术,从部署成本、定位精度、后期维护、传输速度、可移植性等方面综合考虑,基于WiFi(WirelessFidelity)接收信号强度(Received Signal Strength,RSS)的室内定位技术不需要布设其他硬件设备,通过充分利用已有的WiFi设施,即可在任何具有WiFi模块的智能移动设备上实现定位,在众多定位算法中具有一定的优势。然而,RSS容易受到外界环境干扰,严重影响室内定位系统的稳定性与准确性,单纯的基于WiFi信号的定位无法满足人们对室内定位服务的精度要求。针对上述问题,本文通过对WiFi信号接收强度的特征分析,提出基于信息融合的室内定位算法,将基于RSS的WiFi位置指纹定位算法与行人航位推算算法(Pedestrian Dead Reckoning,PDR)通过Kalman滤波器进行数据融合实现定位,并在智能移动终端上开发实现了集室内定位、导航、追踪为一体的应用系统。本文的主要内容和创新点包括:(1)三维室内空间模型的构建。结构清晰、具有良好的表达能力和视觉效果的室内空间模型是实现室内LBS的基础。相比于室外环境,室内空间结构的复杂性对室内建模提出很大的挑战。本文根据现有的室内数据文件,设计并构建了基于"结点-弧段"结构的三维室内空间网络模型,用以表达室内空间要素的空间属性与拓扑结构,作为地图可视化与室内导航的基础。并进一步具体论述了基于Voronoi图的室内走廊中轴线提取原理,实现了建筑单层路径的自动提取,提高了建模的效率。(2)改进的基于RSS的位置指纹定位方法。本文通过对RSS的深入研究,从室内定位的角度分析了不同因素对RSS的影响。并针对RSS的复杂性与多变性特点,提出基于空间收敛的WKNN(Weighted K-Nearest Neighbor)室内定位算法,实现了较为精确地室内定位。同时,通过采用不同的接入点(AccessPoint,AP)选择和匹配机制,去除冗余的AP数据并优化AP定位子集合,提高定位算法的效率与精度。通过与同机制下的算法进行比较,本文提出的算法在实时性和定位精度方面均有提高。在实验环境下,以1.5米的采样间隔创建位置指纹数据库,在使用6个AP进行定位的情况下,获得的平均定位误差为1.68 m。(3)基于Kalman滤波的多数据融合室内实时追踪与导航。在实时室内导航过程中,基于RSS的位置指纹定位算法易受室内环境变化的影响,存在定位不稳定且精度不高的现象,对于运动物体的位置描述也存在不规则跳跃现象;而PDR算法可直接利用移动设备自带的传感器,通过对行人运动状态的估计进行相对位置预测,但是定位存在不可消除的累计误差。本文建立Kalman滤波器对两者定位信息进行数据融合和轨迹平滑,以实现在室内导航过程中获得较高精度的室内实时动态定位精度。并通过行人轨迹方向检测,在转弯处对Kalman滤波进行重置,来降低线性运动模型在转弯处的累积定位误差。同时,文中利用设备气压计数据来识别导航过程中用户的上下楼行为,实现了适用于智能移动设备的多楼层的定位与导航系统。通过实验验证,在室内动态追踪与导航过程中系统的平均定位误差为1.2m。将算法结果与PDR、WiFi定位进行对比,本算法在随着时间上的累积误差上表现最为平稳。
[Abstract]:With the development of wireless network technology and the rapid development of modern city construction, Location Based Services (LBS) has shown tremendous vitality in many aspects, such as personal location service, medical field, electronic commerce, emergency rescue, smart home and so on. It has become a hot research topic in recent years. High-precision indoor positioning and navigation technology is the foundation and key to realize LBS. Traditional GPS (Global Positioning System) and cellular mobile communication technology have higher positioning accuracy outdoors, but GPS signals in indoor environment will be blocked, resulting in a significant reduction in positioning accuracy. Bit technology, from the deployment cost, positioning accuracy, post-maintenance, transmission speed, portability and other aspects of a comprehensive consideration, based on WiFi (Wireless Fidelity) Received Signal Strength (RSS) indoor positioning technology does not require the deployment of other hardware devices, by making full use of existing WiFi facilities, you can have any WiFi module. However, RSS is susceptible to external environment interference, which seriously affects the stability and accuracy of indoor positioning system. Simple WiFi-based positioning can not meet the accuracy requirements of indoor positioning services. Based on the analysis of the characteristics of the reception intensity of WiFi signals, an indoor localization algorithm based on information fusion is proposed. The location algorithm of WiFi position fingerprint based on RSS and Pedestrian Dead Reckoning (PDR) are fused by Kalman filter to realize the localization, and the indoor localization is realized on the intelligent mobile terminal. The main contents and innovations of this paper include: (1) the construction of three-dimensional indoor space model. The indoor space model with clear structure, good expressive ability and visual effect is the foundation of indoor LBS. Compared with outdoor environment, the complexity of indoor space structure has a great impact on indoor modeling. According to the existing indoor data files, this paper designs and constructs a three-dimensional indoor space network model based on "node-arc" structure to express the spatial attributes and topological structure of indoor space elements, which is the basis of map visualization and indoor navigation. The principle of axes extraction realizes the automatic extraction of building single-layer path and improves the efficiency of modeling. (2) An improved location fingerprint method based on RSS is proposed. WKNN (Weighted K-Nearest Neighbor) indoor localization algorithm based on WKNN (Weighted K-Nearest Neighbor) achieves more accurate indoor localization. At the same time, by using different access point (AP) selection and matching mechanism, redundant AP data is removed and AP localization subset is optimized to improve the efficiency and accuracy of localization algorithm. The algorithm presented in this paper improves the real-time performance and positioning accuracy. In the experimental environment, a location fingerprint database is created with a sampling interval of 1.5 meters, and the average positioning error is 1.68 m when six APs are used for positioning. (3) Multi-data fusion indoor real-time tracking and navigation based on Kalman filtering. In the navigation process, the location fingerprint localization algorithm based on RSS is vulnerable to the influence of indoor environment changes, there is instability and low precision in the location, and there is also irregular jumping phenomenon in the position description of moving objects; PDR algorithm can directly use the sensors of mobile devices to estimate the state of pedestrian movement. In this paper, Kalman filter is established to fuse the positioning information and smooth the trajectory, so as to achieve high precision indoor real-time dynamic positioning accuracy in the process of indoor navigation. In order to reduce the cumulative positioning error of the linear motion model at the turning point, the device barometer data is used to identify the user's upstairs and downstairs behavior in the navigation process, and the multi-floor positioning and navigation system suitable for intelligent mobile devices is realized. The average positioning error is 1.2m. Compared with PDR and WiFi, the algorithm is the most stable in cumulative error over time.
【学位授予单位】：华东师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN92

【相似文献】