物联网学报 ›› 2021, Vol. 5 ›› Issue (2): 107-115.doi: 10.11959/j.issn.2096-3750.2021.00208

• 理论与技术 • 上一篇    下一篇

基于生成对抗网络的高精度室内无线定位方法

王福展, 朱晓荣, 陈美娟, 朱洪波   

  1. 南京邮电大学通信与信息工程学院，江苏 南京 210003
  • 修回日期:2020-05-29 出版日期:2021-06-30 发布日期:2021-06-01
  • 作者简介:王福展（1995- ），男，南京邮电大学硕士生，主要研究方向为无线室内定位
    朱晓荣（1977- ），女，博士，南京邮电大学教授、博士生导师，主要研究方向为5G网络、物联网、无线定位算法等
    陈美娟（1971- ），女，博士，南京邮电大学副教授、硕士生导师，主要研究方向为5G网络、物联网
    朱洪波（1956- ），男，博士，南京邮电大学教授、博士生导师，主要研究方向为泛在无线通信与物联网、宽带移动通信、下一代网络、无线通信与电磁兼容等

High-precision indoor wireless positioning method based on generative adversarial network

Fuzhan WANG, Xiaorong ZHU, Meijuan CHEN, Hongbo ZHU   

  1. College of Telecommunications ＆Information Engineering, Nanjing University of Posts and Telecommunications, Nanjing 210003, China
  • Revised:2020-05-29 Online:2021-06-30 Published:2021-06-01

在线阅读

37

PDF下载

487

摘要：

无线信号在传播过程中容易受到干扰，这限制了传统室内定位方法在实际生活中的应用。而基于位置的指纹定位技术具有普适性强的优点，是当前的研究热点。指纹数据的数量是影响指纹定位精度的重要因素，但是采集大量指纹数据的成本较大。因此，如何使用少量指纹数据实现较高定位精度成为指纹定位技术的难点。针对此问题，提出了一种基于生成对抗网络（GAN, generative adversarial network）的高精度室内无线定位方法。首先，在室内等间隔密集地采集指纹数据，构造初始指纹数据集，选取初始指纹数据集中部分指纹数据，使用 GAN 利用部分指纹数据得到大量指纹数据；然后，基于这些生成数据，使用k最近邻（KNN, k-nearest neighbor）分类算法模型和随机森林模型进行定位预测。实验结果表明，该方法能够基于少量指纹数据实现较高的无线定位精度，定位精度达15.4 cm。

关键词: 指纹定位, 生成对抗网络, 室内, KNN, 随机森林

Abstract:

Because wireless signals are susceptible to interference during the propagation process, the application of traditional indoor positioning methods in real life is limited.Because location-based fingerprint positioning technology has the advantage of strong universality, it has become a current research hotspot.The number of fingerprint data is an important factor affecting the accuracy of fingerprint positioning, but the cost of collecting a large amount of fingerprint data is large.Therefore, how to use a small amount of fingerprint data to achieve higher positioning accuracy is a difficult point of fingerprint positioning technology.Aiming at this problem, a high-precision indoor wireless positioning method based on generative adversarial network was proposed.Firstly, fingerprint data was collected densely at equal intervals indoors, and the initial fingerprint data set was constructed, the part of the fingerprint data was selected in the initial fingerprint data set, and the generative adversarial network was used to obtain a large amount of fingerprint data from part of the fingerprint data.Then, based on these generated data, a KNN (k-nearest neighbor) model and a random forest model were used for location prediction.Experimental results show that this method can achieve high wireless positioning accuracy based on a small amount of fingerprint data, and the positioning accuracy can reach 15.4 cm.

Key words: fingerprint localization, generative adversarial network, indoor, KNN, random forest

中图分类号: 

No Suggested Reading articles found!