Research on indoor single-station precise passive positioning technology based on neural network
-
摘要: 为了提高室内环境下对目标的定位精度，提出一种室内单站精确定位技术. 该技术利用室内电波传播多径效应构成的复杂信道信息，基于机器学习，构建卷积神经网络架构，通过卷积提取不同位置目标到达接收传感器的多径时延特征信息；然后通过多层全连接层深度神经网络的模型训练，将基于复杂信道的定位问题转化为回归模型的问题，建立信道指纹与位置之间的非线性关系来完成被动定位. 训练和仿真结果表明，在室内复杂电波传播环境下，基于神经网络的室内单站精确定位技术能够实现单接收站情况下对目标的精确定位. 本文主要对3
$ \times $ 3网格大小的金属散射体进行定位，接收站位于室内时，平均定位误差为0.621个网格（12.42 cm）；接收站位于室外时，能够分别实现信噪比20 dB、30 dB、40 dB情况下44.09 cm、21.42 cm、20.96 cm的平均定位误差. 本文方法为室内复杂环境下的目标定位提供了一种新的定位方法.Abstract: In order to improve the positioning accuracy of targets in the indoor environment, a single-station precise passive positioning technology is proposed in this paper. Based on the complex channel information formed by the multipath effect of indoor radio wave propagation, a convolution neural network architecture is constructed, which can extract the multipath delay characteristic information from different location targets to the receiving sensor. Then, a deep network with several dense layers is designed to train the convoluted data, in which way, we transform the location problem based on complex channels into the problem of regression model and establish the nonlinear relationship between signal fingerprint and position, and finally the passive location of the target can be realized. The training and simulation results show that in the indoor environment of complex radio wave propagation, the technology we proposed can accurately locate the target in the case of a single receiving station. In this paper, we mainly locate the metal scatterer with the size of 3$ \times $ 3 grids. When the receiving sensor is located indoors, the average positioning error is 0.621 grid(12.42 cm). When outdoors, the average positioning error is 44.09 cm, 21.42 cm, 20.96 cm under the SNR of 20, 30, 40 dB, respectively, which provides a new method for indoor target positioning of complex environments.-
Key words:
- indoor positioning /
- neural network /
- multipath effect /
- machine learning /
- Green function
-
图 5 接收信号与不同卷积核的卷积结果.
Fig. 5 Convolution results of received signals with different convolution kernels
图 8 训练和测试均方误差曲线（室内单接收站）
Fig. 8 Training loss curve and test loss curve (single receiving station inside)
图 9 训练和测试均方误差曲线（室外单接收站）.
Fig. 9 Training loss curve and test loss curve (single receiving station outside).
表 1 两种神经网络的测试结果
Tab. 1 Results of the two neural networks
神经网络 K 训练
均方
误差测试
均方
误差平均测试距离误差 误差/网格 误差/cm CNN 203 0.046 0.359 0.621 12.42 Dense 203 0.901 12.283 3.873 77.46 1637 0.008 0.070 0.285 5.70 
下载: 导出CSV
表 2 不同信噪比下的测试结果
Tab. 2 Results of different SNRs
信噪比/dB 训练均方误差 测试均方误差 平均测试距离误差 误差/网格 误差/cm 20 1.22 4.994 2.204 44.09 30 0.327 1.142 1.071 21.42 40 0.275 0.999 1.048 20.96
下载: 导出CSV
-
[1] SEGURA M, CHEEIN F A, TOIBERO J, et al. Ultra Wide-band localization and SLAM: a comparative study for mobile robot navigation[J]. Sensors,2011,11(2):2035-2055. [2] CHE F, AHMED A, AHMED Q Z, et al. Machine learning based approach for indoor localization using ultra-wide bandwidth (UWB) system for industrial internet of things (IIoT). [C]// The 5th International Conference on the UK-China emerging technologies (UCET), 2020. DOI: 10.1109/UCET51115.2020.9205352 [3] HOEFER W J R. Computational time reversal: a frontier in electromagnetic structure synthesis and design[J]. IEEE transactions on microwave theory techniques,2015,63(1):3-10. [4] HOEFER W J R. Superresolution imaging by computational time reversal in scattering media[C]// IEEE International Symposium on Antennas and Propagation & USNC/URSI National Radio Science Meeting. Vancouver: IEEE, 19-24 July, 2015. [5] 刘召伟, 徐凤燕, 王宗欣. 基于参数拟合的室内多用户定位算法[J]. 电波科学学报,2008,23(6):1090-1105.LIU Z W, XU F Y, WANG Z X. A new multi-user indoor location algorithm based on difference distance-loss model with parameter fitting[J]. Chinese journal of radio science,2008,23(6):1090-1105. (in Chinese) [6] 李若南, 李金宝. 一种无源被动室内区域定位方法的研究[J]. 计算机研究与发展,2020,57(7):1381-1392.LI R N, LI J B. Research on a device-free passive indoor regional localization method[J]. Journal of computer research and development,2020,57(7):1381-1392. (in Chinese) [7] WU K, XIAO J, YI Y, et al. CSI-based indoor localization[J]. IEEE transactions on parallel and distributed systems,2013,24(7):1300-1309. [8] XIAO J, WU K, YI Y, et al. FIFS: fine-grained indoor fingerprinting system[C]// The 21st International Conference on Computer Communications and Networks (ICCCN). Munich: IEEE, 30 July-2 August, 2012. [9] WU K, XIAO J, YI Y, et al. FILA: Fine-grained indoor localization[C]// Proceedings IEEE INFOCOM. Orlando, March 25-30, 2012. [10] WANG X, GAO L, MAO S. CSI Phase fingerprinting for indoor localization with a deep learning approach[J]. IEEE internet of things journal,2016,3(6):1113-1123. [11] WANG X, GAO L, MAO S, et al. CSI-based fingerprinting for indoor localization: a deep learning approach[J]. IEEE transactions on vehicular technology,2017,66(5):763-776. [12] ZHANG H Y, TONG G X, XIONG N X. Fine-grained CSI fingerprinting for indoor localisation using convolutional neural network[J]. IET communications,2020,14(18):3266-3275. [13] 葛德彪, 闫玉波. 电磁波时域有限差分方法[M]. 西安: 西安电子科技大学出版社, 2011: 10. -
点击查看大图
图(11) / 表(2)
计量
- 文章访问数: 219
- HTML全文浏览量: 46
- PDF下载量: 15
- 被引次数: 0
