本文为加拿大渥太华大学(作者:Hind Mukhtar)的硕士论文,共103页。
由于基于位置的服务和高带宽应用的增加,需要对用户进行精确定位以改进资源管理和波束形成,因此对定位的需求一直在增长。室外定位传统上是通过全球定位系统(GPS)来实现的,但是由于障碍物和多径效应,在城市环境中的定位性能下降,因此需要更好的定位技术。本文提出了一种利用激光雷达(LIDAR)或卫星图像和基站信道状态信息(CSI)数据进行图像分类和深度学习的级联方法来预测户外移动车辆和用户的位置。使用3种不同的数据集对算法的性能进行了评估,前两个使用毫米波波段的模拟数据和从弗吉尼亚州阿灵顿的罗斯林附近采集的激光雷达图像。结果表明,分层结构提高了定位精度,平均绝对误差(MAE)为6.55m,而单卷积神经网络(CNN)的平均绝对误差(MAE)为9.82m。第三个数据集来自LTE移动通信系统以及丹麦大学的卫星图像,结果表明,分层方法的MAE为9.45m,而单前馈神经网络(FFNN)的MAE为15.74m。
Demand for localization has been growingdue to the increase in location-based servicesand high bandwidth applicationsrequiring precise localization of users to improve resourcemanagement and beamforming. Outdoor localization has beentraditionally done through Global Positioning System (GPS), however it’s performance degrades in urban settingsdue to obstruction and multi-path effects, creating the need forbetter localization tech-niques. Thisthesis proposes a technique using a cascaded approach composed ofimageclassification and deep learning using LIDAR or satellite images and Channel State In-formation (CSI) data from base stations to predict the location of moving vehicles andusers outdoors. The algorithm’s performance is assessed using3 different datasets. Thefirst two use simulated data in theMilli-meter Wave (mmWave) band and lidar imagesthat are collected from the neighbourhood of Rosslyn inArlington, Virginia. The resultsshow animprovement in localization accuracy as a result of the hierarchicalarchitecture,with a Mean Absolute Error (MAE) of 6.55m for the proposedtechnique in comparisonto a MAE of 9.82m using one Convolutional Neural Network(CNN). The third datasetuses measurements from an LTE mobile communicationsystem along with satellite imagesthat take place at the University ofDenmark. The results achieve a MAE of9.45 m forthe heirchichal approach in comparison to a MAE of 15.74 m for oneFeed-Forward NeuralNetwork (FFNN).
-
引言
-
背景理论
-
文献回顾
-
研究方法
-
研究结果
下载地址:
https://url92.ctfile.com/f/1850492-519471753-bb18d8
(访问密码:3660)
更多精彩文章请关注公众号: