电磁频谱地图构建：稀疏高斯过程回归方法

Electromagnetic Spectrum Map Construction ：a Sparse Gaussian Process Regression Approach

摘要

针对大面积场景电磁频谱地图构建面临的大数据量、大计算开销的挑战，提出一种基于变分推断的稀疏高斯过程回归方法，实现电磁频谱地图构建。首先，将电磁频谱地图构建问题建模为一个回归问题，利用已采集的数据拟合地理位置与对应接收信号强度之间的映射关系。其次，将该映射关系建模为一个高斯随机过程，利用高斯过程回归的非参数化特性构建预测模型。最后，仅选取部分已采集数据作为模型的数据输入，通过最大化变分分布与后验分布之问的相似性训练得到一个稀疏的高斯过程回归预测模型。仿真实验和基于实采数据的实验表明，该方法能够在显著降低计算复杂度、加快运算速度的基础上，构建得到高精度的电磁频谱地图。

Abstract

To address the challenges of massive data volume and substantial computational overhead in large area electromagnetic spec- trum maps construction，a sparse Gaussian process regression method based on variational inference is proposed.First，the electromag- netic spectrum map construction problem is modeled as a regression problem，and mapping relationship between geographic location and corresponding received signal strength is fitted using collected data.Second，the mapping relationship is modeled as a Gaussian sto- chastic process，and the non-parametric property of Gaussian process regression is used to construct a prediction model.Finally，only part of the collected data is selected as input of the model，and a sparse Gaussian process regression prediction model is obtained by maximizing the similarity between variational distribution and posterior distribution.Simulation experiments and real data experiments show that this method can significantly reduce computational complexity and speed up operation process while obtaining high-precision electromagnetic spectrum maps.

关键词

稀疏高斯过程回归 / 频谱感知 / 电磁频谱地图构建

Key words

sparse Gaussian process regression / spectrum awareness / electromagnetic spectrum map construction

参考文献

[1] Romero D, Kim S. Radio map estimation:a data-driven approach to spectrum cartography[J]. IEEE Signal Processing Magazine, 2022, 39(6):53-72.

[2] Wang J, Zhu Q M, Lin Z P, et al. Sparse Bayesian learning-based 3d radio environment map construction sampling optimization,sce- nario-dependent dictionary construction,and sparse recovery[J]. IEEE Transactions on Cognitive Communications and Networking, 2024, 10(1):80-93.

[3] Jiang H, Mukherjee M, Zhou J, et al. Channel modeling and charac- teristics for 6G wireless communications[J]. IEEE Network, 2021, 35 (1):296-303.

[4] Shi W Q, Jiang H, Xiong B P, et al. RIS-Empowered V2V commu- nications:three-dimensional beam domain channel modeling and a- nalysis[J]. IEEE Transactions on Wireless Communications, 2024, 23 (11):15844-15857.

[5] TU C, ZHU W L, ZHANG M, et al. A broadband intelligent direc- tion finding method based on CNN features[J]. Information Coun- termeasure Technology, 2022, 1(2):75-84.

[6] Hata M. Empirical formula for propagation loss in land mobile ra- dio service[J]. IEEE Transactions on Vehicular Technology, 1980, 29 (3):317-325.

[7] Straka T, Vojtech L, Neruda M. Graphical heatmap-based approach to indoor radio signal propagation:adapting advanced ray tracing and global illumination algorithms[J]. IEFE Transactions on An- tennas and Propagation, 2024, 72(7):6045-6059.

[8] Xia H Y, Zha S, Huang J J, et al. Radio environment map construc- tion by residual kriging based on Bayesian hierarchical model[C]// Proceedings of International Symposium on Electromagnetic Compatibility(ISEMC), 2023:1-5.

[9] TAO S F, WU Y J, LUO J, et al. Radio environment map construc- tion method for complex scenes based on inverse obstacle distance weighted[J]. Journal of Electronics ＆ Information Technology, 2024, 46(8):3210-3218.

[10] ZHAO W L, TIAN X Y, CHEN C, et al. Wi-Fi Fingerprint locali- zation uniting spline interpolation[J]. Journal of Electronics ＆ In- formation Technology, 2024, 46(9):3563-3570.

[11] Mario R C, Carla E G, Taewoong H, et al. A REM update method- ology based on clustering and random forest[J]. Applied Sciences, 2023, 13(9):5362,doi: 10.3390/app13095362.

[12] Zhen P, Zhang B N, Xu Y Q, et al. Radio environment map con- struction based on gaussian process with positional uncertainty[J]. IEEE Wireless Communications Letters, 2022, 11(8):2162-2337.

[13] Konstantinos D P, Alireza S, Wei Y, et al. Bayesian active learning for sample efficient 5G radio map reconstruction[J]. IEEE Trans- actions on Wireless Communication, 2024, 30(12):19382-19396.

[14] Xu Y Q, Zhang B N, Zhang X K, et al. Radio environment map construction with Gaussian process and kernel transformation[C]// Proceedings of International Conference on Communication Image,and Signal Processing( CCISP),2021:350-355.

[15] Wang X Y, Wang X Y, Mao S W, et al. Indoor radio map construc- tion and localization with deep gaussian processes[J]. IEEE Inter- net of Things Journal, 2020, 7(11):11238-11249.

[16] Nicolo D F, Michele R, Gianluigi P, et al. Model-free radio map es- timation in massive MIMO systems via semi-parametric gaussian regression[J]. IEEE Wireless Communications Letters, 2022, 11 (3):473-477.

[17] Zhang Y L, Ma L. Radio map crowd sourcing update method using sparse representation and low rank matrix recovery for WLAN in-door positioning system[J]. IEEE Wireless Communications Let- ters, 2021, 10(6):1188-1191.

[18] CHEN Z B, HU J M, ZHANG B N, et al. Spectrum map construc- tion algorithm based on tensor tucker decomposition[J]. Journal of Electronics ＆ Information Technology, 2023, 45(11):4161-4169.

[19] Suto K, Rannai S, Sato K, et al. Image-driven spatial interpolation with deep learning for radio map construction[J]. IEEE Wireless Communications Letters, 2021, 10(6):1222-1226.

[20] Zhang S Y, Wijesinghe C, Ding Z. RME-GAN:a learning frame- work for radio map estimation based on conditional generative ad- versarial network[J]. IEEE Internet of Things Journal, 2023, 10 (20):18016-18027.

[21] Wang J, Zhu Q M, Lin Z P, et al. Sparse Bayesian learning-based hierarchical construction for 3D radio environment maps incorpora- ting channel shadowing[J]. IEEE Transactions on Wireless Com- munication, 2024, 23 (10):14560-14574.

[22] Wang J, Zhu Q M, Lin Z P, et al. Sparse Bayesian learning-based hierarchical construction for 3D radio environment maps incorpora- ting channel shadowing[J]. IEEE Transactions on Wireless Com- munications, 2024, 23(10):14560-14574.

[23] Lyu C, Liu X, Mihaylova L. Review of recent advances in Gaussian process regression methods[C]// Proceedings of the UK Workshop on Computational Intelligence, 2022:226-237.

[24] David R Burt, Edward R C, Wilk V D. Convergence of sparse vari- ational inference in Gaussian processes regression[J]. Journal of Machine Learning Research, 2020, 21(119):1-63.

[25] Michalis C S. Variational learning of introducing variables in sparse gaussian process[C]// Proceedings of International Conference on Artificial Intelligence and Statistics(AISTATS), 2009:567-574.

[26] 屠铖, 朱文丽, 张旻, 等. 基于 CNN 特征的宽频段智能测向方法[J]. 信息对抗技术, 2022, 1(2):75-84.

[27] 陶诗飞, 只昱江, 罗佳, 等. 基于反障碍距离加权的复杂场景电磁频谱地图构建方法[J]. 电子与信息学报, 2024, 46(8): 3210-3218.

[28] 赵万龙, 田新元, 陈超, 等. 联合 Spline 插值的 Wi-Fi 指纹匹配定位算法[J]. 电子与信息学报, 2024, 46(9):3563-3570.

[29] 陈智博, 胡景明, 张邦宁, 等. 基于张量 Tucker 分解的频谱地图构建算法[J]. 电子与信息学报, 2023, 45(11):4161-4169.

基金资助

国家白然科学基全项目(61971473)

国家白然科学基全项目(62372456)