基于机器学习和卫星图像的路径损耗预测

基于反向传播神经网络（back propagation neural network,BPNN）构建了一种路径损耗预测模型. 通过卫星图像的红、绿、蓝（red, green and blue,RGB）通道的颜色信息来表征无线通信电波传播路径的环境特征,结合路测点与基站的距离特征构建数据集,迭代训练网络参数,以预测传播路径损耗. 结果表明,对跨基站路测点的预测结果与实测数据之间的相关系数达到0.83,绝对平均误差控制在0.66 dB,标准差控制在6.65 dB,说明在缺乏某一场景的详细模型和材质参数时,本文模型也能可靠预测无线通信电波的传播路径损耗. 此外,本文信道模型与传统信道建模方法多方面的对比与分析表明,本文模型在相同计算资源下可以提供和传统信道建模方法相差很小的预测结果,同时大大缩短预测所需的时间,说明本文模型对传播路径损耗做出快速预测的能力可以用于无线通信网络系统的优化.     

一种林区毫米波通信的路径损耗预测模型

由于林区环境复杂,实地测量路损存在困难,因此基于自由空间邻近模型(Close-In,CI)引入了路径损耗指数矫正因子,并考虑视距和非视距传输概率以及雨衰提出了适用于林区的毫米波路径损耗预测模型.以60 GHz毫米波为例,考虑不同的树木高度、树冠半径、天线高度、树木密度进行模拟仿真并与传统的自由空间路损模型进行比较.仿真...     

基于改进射线跟踪法和BP神经网络算法的室外微蜂窝毫米波信道特性研究

在28 GHz与39 GHz毫米波频段室外微蜂窝场景下,基于改进射线跟踪法和反向传播(back propagation, BP)神经网络算法对毫米波单发单收信道及单发多收信道进行建模仿真研究. 在得到的无线信道仿真数据基础上,研究分析了毫米波信道的路径损耗、均方根(root-mean-square, RMS)时延扩展(delay spread, DS)、接收功率等传播特性. 通过与现有文献的测量结果对比分析验证了改进射线跟踪法的正确性与有效性. 通过BP神经网络方法拟合得到的路径损耗模型参数结果与改进射线跟踪法仿真得到的路径损耗参数结果对比发现,两者吻合程度很高,验证了BP神经网络算法能很好地对室外微蜂窝毫米波信道大尺度参数进行预测. 同时,文中给出了一种普遍适用的用来表征室外微蜂窝视距(line-of-sight, LoS)与非视距(non-line-of-sight, NLoS)场景下28 GHz与39 GHz毫米波信道的路径损耗模型. 结果表明:LoS场景下的RMS DS和接收功率都小于NLoS场景下得到的结果;LoS场景与NLoS场景下RMS DS、水平方向到达角、多径簇的个数累积分布函数均服从高斯分布;RMS DS在毫米波频段微蜂窝场景下,随着频率的升高而增大,到达接收端的多径成簇呈现稀疏性.     

基于3D电子地图和信道实测数据的市区路径损耗机器学习模型研究

随着5G移动通信系统的发展部署以及网络性能的优化,高精度和低复杂度的路径损耗预测模型尤为重要。该文针对大型城市场景,使用目前5G热点频段700 MHz, 2.4 GHz, 3.5 GHz的实测数据,将收发端位置、3维距离、相对余隙、建筑物密度、平均高度等作为环境特征,建立了基于3D电子地图的机器学习路径损耗预测模型,结果表明在复杂城市环境下,该文方法因其预测精度高而优于传统的基于收发端距离的路径损耗模型。另外,该文提出了基于频率迁移学习的路径损耗预测模型,并用均方误差、平均绝对百分比误差、均方根误差、决定系数等指标对其性能进行评估。该文方法可以解决建筑物遮挡严重的复杂城市环境以及在无大量测试数据的路径损耗预测问题,精确地预测城市环境中视距非视距混合信道的路径损耗值。     

神经网络传播模型的设计与应用

传播模型是网络规划的基础之一,其对路径损耗的准确预测能力从一定程度上决定了网络规划的质量。在工程中得到广泛应用的经典传播模型在预测本地无线环境引起的路径损耗方面存在不够准确的缺点。提出了一种以人工神经网络为核心的自适应传播模型,通过利用现网路测数据对神经网络进行训练,充分利用神经网络所具备的高度非线性映射能力和自适应性,使其具备根据本地无线环境特征准确预测传播损耗的能力。实验表明,基于神经网络的传播模型对本地无线环境引起的路径损耗预测的准确性比经典模型更好,并且其适用性也较强。     

28 GHz室内毫米波信道路径损耗模型研究

毫米波信道建模是第五代（the 5th Generation,5G）移动通信系统的关键技术,而路径损耗是表征毫米波信道传播大尺度衰落影响的重要参数.为了更好地理解毫米波信道的传播特性,应进行广泛的信道测量与建模.因此,对28 GHz室内环境进行了信道测量,并给出了相应的毫米波信道路径损耗模型,同时基于入射及反弹射线法/镜像法仿真分析了路径损耗传播特性.研究结果表明:实测结果与仿真结果一致性吻合良好,从而验证了入射及反弹射线法/镜像法的正确性;自由空间邻近（Close-In,CI）参考距离路径损耗模型表达式更简洁,鲁棒性更强.最后,本文给出了一种普遍适用的用来表征室内视距（Line-of-Sight,LOS）与非视距（Non-Line-of-Sight,NLOS）环境28 GHz与60 GHz毫米波信道的路径损耗模型.     

基于BP神经网络模型的RSSI测距方法研究

针对传统信号传播路径损耗模型接收的信号强度指示(received signal strength indication, RSSI)测距误差较大, 提出了基于反向传播(back propagation, BP)神经网络模型的RSSI测距方法.首先, 研究分析传统信号传播路径损耗模型及测距误差; 其次, 利用BP神经网络构建新的路径损耗模型, 并将该模型应用到RSSI测距中, 对基于BP神经网络模型的RSSI测距方法进行研究; 最后, 通过实验和MATLAB仿真对测距方法进行验证.仿真结果表明:BP神经网络模型的RSSI测距误差比传统信号传播路径损耗模型的RSSI测距误差要小.     

密集多径环境下UWB测距的NLOS误差减小方法

为减小密集多径环境下超宽带(UWB)测距结果中因为障碍物引起的非视距(NLOS)误差,提出了一种有效的NLOS误差减小方法.此方法考虑了NLOS误差的产生原理及特点,以信号传播的路径损耗模型为基础,通过对接收信号中不同时间到达单径的能量比较,实现了对NLOS误差的粗略估计,进而以此估计值对测距结果进行校正.在方法的具体实现上,给出了一种计算量较小、复杂度较低的单径检测算法.对实测数据的处理验证了方法的正确性,结果表明本文提出的NLOS误差减小方法使UWB测距精度有了较大提升.     

非视距紫外光通信系统中霾衰减特性研究

为了研究霾环境下应用紫外光通信的系统特性,研究了霾粒子的物理特性及谱分布特性,利用散射理论分析了霾粒子在日盲紫外光波段的散射特性;并利用经典Luettgen单散射信道模型,研究了霾环境下非视距日盲紫外光传输的路径损耗特性。通过分析路径损耗与通信距离、能见度以及系统角度之间的关系仿真结果,得到了非视距紫外光传输系统中霾衰减的理论特性:在较短通信距离时,系统的路径损耗受天气状况（能见度）影响较大;能见度较好时,通信距离对路径损耗的影响将会突出,实际中应尽量选取能见度大于2 km的天气条件。文中的工作对设计霾天环境下紫外光通信系统及优化系统性能提供了一定的理论参考,同时对系统工程化实现也具有一定的指导意义。     

基于射线跟踪的空地毫米波传播损耗预测

传播毫米波传播损耗模型假设收发端天线高度比较低,不直接适用于空地传输场景,本文基于射线跟踪原理提出了一种空地场景下毫米波通信的传播损耗统计模型。该模型推导分析了视距、反射、绕射和无信号四种不同传播情况的路径概率,并综合考虑了环境、频率、距离和天线高度等因素对路径损耗的影响。仿真分析结果表明,本文模型对不同场景和传播高度都具有良好的适用性,对于非视距情况的细化分类使其计算结果,比传统模型更为精确且与射线跟踪仿真结果更为吻合,可用于空地毫米波通信系统设计和算法优化等领域。      

Review of Wireless Personal Communications Radio Propagation Models in High Altitude Mountainous Areas at 2.6 GHz

This paper presents five commonly used radio propagation models (RPMs) which are suitable for the prediction of path loss in macrocell environments of LTE wireless communication systems. These RPMs’ application in high altitude mountainous areas networks (HAMANETs) environment requires further validation and studies. Through using the measured path loss in the HAMANETs at 2.6 GHz to calculate the predicted value of the five RPMs and the measured value’s mean error (ME), root mean square error, and error standard deviation (ESTD), we verified the predicted value of the SPM model that is closer to the actual measurement. On this basis, the empirical propagation model in HAMANETs environment is corrected. When correcting, a method to calculate base station’s effective antenna height and propagation distance is provided by using the altitude above sea level data. This method can reduce the error that the mountainous areas are simplified to the flat-terrain in the existed propagation models. A linear least square method is used to calculate the optimal propagation model. Finally, the ME is the smallest, and the ESTD is less than 8 dB, which indicate that the corrected propagation model is more suitable for the actual environmental path loss’s prediction. The results show that the path loss factor of the test area is about 65 dB, including the influence of the high altitude, mountains, vegetation, and air humidity in HAMANETs environment. The study results can provide useful advice to the evaluation and verification of personal wireless communications in the HAMANETs. Furthermore, using the correction method proposed in this paper can correct propagation models suitable for the different propagation environments to improve the accuracy and efficiency of wireless network optimization.     

Neural network-based path loss model for cellular mobile networks at 800 and 1800 MHz bands

Cellular radio communication systems have become essential for data/voice/video/multimedia applications. The performance of cellular communication radio systems is assessed by considering the design specifications of frequency planning, channel assignment and interference mitigation strategies among others. Frequency planning is the most important component to improve capacity or quality of cellular radio systems. Large-scale path loss values between the base station and mobile stations are the key regulating factors that limit the performance of cellular systems, especially in urban/vegetation region. There is a necessity to develop a suitable path loss prediction model for predicting path loss values based on received signal strength measurements. In this paper, an ANN-based path loss model was used for macro cell measurement data obtained in the Vijayawada urban region, India. The Multi-Layer Perceptron (MLP) neural network model was considered. The prediction results indicate that the ANN model outperformed the Auto Regressive Moving Average (ARMA) and COST-231-WImodels. The outcome of this research work will be immensely useful for improving coverage and ensuring better frequency planning of cellular radio systems.     

Modeling of Soldering Quality by Using Artificial Neural Networks

Multilayer perceptrons (MLPs ) are well-known artificial neural networks (ANNs) that are used in many different applications. In this paper, MLP neural networks were used to predict product quality in a wave soldering research case. The aims were to construct process models and to determine whether the formation of soldering defects could be predicted reliably by using the method. In addition, the scope of the research included demonstrating the prediction performance of the created models. A MLP-based variable selection procedure with a back-propagation algorithm was used to create defect formation models and to find the most important factors affecting the number of detected defects. The process parameters were used as inputs for the MLP network and each defect type in turn as a model output. In conclusion, the results were promising, and the method used showed potential considering the wider use of the data processing procedure in the electronics or any other industry.     

Cloud security situation prediction method based on grey wolfoptimization and BP neural network

Aiming at the accuracy and error correction of cloud security situation prediction, a cloud security situation prediction method based on grey wolf optimization (GWO) and back propagation (BP) neural network is proposed.Firstly, the adaptive disturbance convergence factor is used to improve the GWO algorithm, so as to improve theconvergence speed and accuracy of the algorithm. The Chebyshev chaotic mapping is introduced into the positionupdate formula of GWO algorithm, which is used to select the features of the cloud security situation prediction dataand optimize the parameters of the BP neural network prediction model to minimize the prediction output error.Then, the initial weights and thresholds of BP neural network are modified by the improved GWO algorithm toincrease the learning efficiency and accuracy of BP neural network. Finally, the real data sets of Tencent cloudplatform are predicted. The simulation results show that the proposed method has lower mean square error (MSE)and mean absolute error (MAE) compared with BP neural network, BP neural network based on genetic algorithm(GA-BP), BP neural network based on particle swarm optimization (PSO-BP) and BP neural network based onGWO algorithm (GWO-BP). The proposed method has better stability, robustness and prediction accuracy.     

一种深度残差网络构建的5G无线智能传播模型

为建立更为准确的全覆盖、全应用、全频谱的5G无线信道模型,提出通过认知无线电与深度神经网络相结合的方法研究无线电波传播特性。根据传统无线传播模型并考虑到不同传播环境,根据信道大尺度衰落特性（包括路径损耗、阴影衰落和小尺度衰落特性）的统计结果,通过BP算法提取特征,并应用FeatureTools进行深度特征综合建立特征方程,计算特征变量与传播损耗的相关系数,进行相关系数的置信区间及变量独立性检验,最终筛选出22个特征并排序。基于深度残差网络建立传播路径损耗的回归模型,结合批正则化过拟合测算平均接收功率,为建立更精确的无线信道模型提供了量化依据,并最终在测试数据集上取得均方根误差8.36（本地）和10.03（云端）的成绩,对工程实践具有较强的参考价值。     

基于优化神经网络算法的电离层foF2预测

为了提高基于反向传输（back propagation,BP）神经网络的电离层f_oF₂预测的精度,采用了一种改进粒子群优化神经网络的方法,对BP网络的初始权值进行优化,防止出现神经网络训练中的局部最优.通过比较基于粒子群优化的神经网络预测结果与遗传算法优化的神经网络预测结果,我们发现对于BP神经网络,两种方法都有很好的性能.此外,和电离层经验模型国际参考电离层模型（international reference ionosphere 2016,IRI2016）结果进行对比,结果表明,本文提出的自适应变异粒子群（adaptive mutation particle swarm optimization,AMPSO）优化神经网络能有效提高f_oF₂的预测精度,并在低纬地区有更好的预测效果.     

Macrocell path loss prediction using artificial intelligence techniques

The prediction of propagation loss is a practical non-linear function approximation problem which linear regression or auto-regression models are limited in their ability to handle. However, some computational Intelligence techniques such as artificial neural networks (ANNs) and adaptive neuro-fuzzy inference systems (ANFISs) have been shown to have great ability to handle non-linear function approximation and prediction problems. In this study, the multiple layer perceptron neural network (MLP-NN), radial basis function neural network (RBF-NN) and an ANFIS network were trained using actual signal strength measurement taken at certain suburban areas of Bauchi metropolis, Nigeria. The trained networks were then used to predict propagation losses at the stated areas under differing conditions. The predictions were compared with the prediction accuracy of the popular Hata model. It was observed that ANFIS model gave a better fit in all cases having higher R² values in each case and on average is more robust than MLP and RBF models as it generalises better to a different data.