CN112367129B - 5G reference signal received power prediction method based on geographic information - Google Patents
5G reference signal received power prediction method based on geographic information Download PDFInfo
- Publication number
- CN112367129B CN112367129B CN202010729363.4A CN202010729363A CN112367129B CN 112367129 B CN112367129 B CN 112367129B CN 202010729363 A CN202010729363 A CN 202010729363A CN 112367129 B CN112367129 B CN 112367129B
- Authority
- CN
- China
- Prior art keywords
- model
- cell
- height
- neural network
- signal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 16
- 238000012937 correction Methods 0.000 claims abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 13
- 238000013473 artificial intelligence Methods 0.000 claims abstract description 8
- 210000004027 cell Anatomy 0.000 claims description 80
- 238000003062 neural network model Methods 0.000 claims description 54
- 230000006870 function Effects 0.000 claims description 24
- 238000011176 pooling Methods 0.000 claims description 18
- 238000004364 calculation method Methods 0.000 claims description 13
- 238000010586 diagram Methods 0.000 claims description 12
- 239000013598 vector Substances 0.000 claims description 12
- 230000004913 activation Effects 0.000 claims description 10
- 210000002569 neuron Anatomy 0.000 claims description 9
- 238000010276 construction Methods 0.000 claims description 8
- 238000013527 convolutional neural network Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims 2
- 230000001537 neural effect Effects 0.000 claims 1
- 238000010248 power generation Methods 0.000 claims 1
- 238000012876 topography Methods 0.000 claims 1
- 238000013461 design Methods 0.000 description 4
- 230000008054 signal transmission Effects 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- CCEKAJIANROZEO-UHFFFAOYSA-N sulfluramid Chemical group CCNS(=O)(=O)C(F)(F)C(F)(F)C(F)(F)C(F)(F)C(F)(F)C(F)(F)C(F)(F)C(F)(F)F CCEKAJIANROZEO-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/309—Measuring or estimating channel quality parameters
- H04B17/318—Received signal strength
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/373—Predicting channel quality or other radio frequency [RF] parameters
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/391—Modelling the propagation channel
- H04B17/3913—Predictive models, e.g. based on neural network models
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Electromagnetism (AREA)
- Computer Networks & Wireless Communication (AREA)
- Data Mining & Analysis (AREA)
- Signal Processing (AREA)
- Mathematical Physics (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
本发明公开了一种基于地理信息的5G参考信号接收功率预测方法。属于人工智能技术领域,具体步骤包括:1、根据地图信息构建特征地图;2、使用人工智能构建误差迭代修正模型,将构建的特征地图与实际的信号接受功率共同训练构建的误差迭代修正模型;其中,所述特征地图以小区的建筑位置信息、发射机相对地面的高度、机器下倾角、垂直电下倾角、栅格与发射机的距离、栅格与信号线的相对高度、传播路径的损耗、载波频率、用户天线高度纠正项、小区发射机相对地面的高度、小区发射机发射功率、信号接收海拔高度作为构建特征地图的输入,最终以信号接收功率为输出。
The invention discloses a method for predicting the received power of a 5G reference signal based on geographic information. It belongs to the technical field of artificial intelligence, and the specific steps include: 1. Constructing a feature map according to map information; 2. Using artificial intelligence to construct an error iterative correction model, and jointly training the constructed feature map and the actual signal receiving power to construct an error iterative correction model; The feature map is based on the building location information of the cell, the height of the transmitter relative to the ground, the machine downtilt angle, the vertical electrical downtilt angle, the distance between the grid and the transmitter, the relative height between the grid and the signal line, and the loss of the propagation path. , carrier frequency, user antenna height correction item, the height of the cell transmitter relative to the ground, the cell transmitter transmit power, and the signal receiving altitude are used as the input to construct the feature map, and finally the signal received power is the output.
Description
技术领域technical field
本发明属于人工智能技术领域,具体涉及一种基于地理信息的5G参考信号接收功率预测方法。The invention belongs to the technical field of artificial intelligence, and in particular relates to a method for predicting the received power of a 5G reference signal based on geographic information.
背景技术Background technique
随着5G NR技术的发展,5G在全球范围内的应用也在不断地扩大;运营商在部署5G网络的过程中,需要合理地选择覆盖区域内的基站站址,进而通过部署基站来满足用户的通信需求;在整个无线网络规划流程中,高效的网络估算对于精确的5G网络部署有着非常重要的意义;无线传播模型正是通过对目标通信覆盖区域内的无线电波传播特性进行预测,使得小区覆盖范围、小区间网络干扰以及通信速率等指标的估算成为可能;由于无线电波传播环境复杂,会受到传播路径上各种因素的影响,如平原、山体、建筑物、湖泊、海洋、森林、大气、地球自身曲率等,使电磁波不再以单一的方式和路径传播而产生复杂的透射、绕射、散射、反射、折射等,所以建立一个准确的模型是一项非常艰巨的任务。With the development of 5G NR technology, the application of 5G worldwide is also expanding; in the process of deploying 5G networks, operators need to reasonably select the base station sites in the coverage area, and then deploy base stations to meet the needs of users. In the entire wireless network planning process, efficient network estimation is very important for accurate 5G network deployment; the wireless propagation model is precisely by predicting the radio wave propagation characteristics in the target communication coverage area, so that the cell It is possible to estimate indicators such as coverage, inter-cell network interference and communication rate; due to the complex propagation environment of radio waves, it will be affected by various factors on the propagation path, such as plains, mountains, buildings, lakes, oceans, forests, atmosphere , the curvature of the earth itself, etc., so that electromagnetic waves no longer propagate in a single way and path, resulting in complex transmission, diffraction, scattering, reflection, refraction, etc., so it is a very difficult task to establish an accurate model.
现有的无线传播模型可以按照研究方法进行区分,一般分为:经验模型、理论模型和改进型经验模型;经验模型的获得是从经验数据中获取固定的拟合公式,典型的模型有Cost 231-Hata、Okumura等;理论模型是根据电磁波传播理论,考虑电磁波在空间中的反射、绕射、折射等来进行损耗计算,比较有代表性的是Volcano模型;改进型经验模型是通过在拟合公式中引入更多的参数从而可以为更细的分类场景提供计算模型,典型的有Standard Propagation Model(SPM)。Existing wireless propagation models can be distinguished according to research methods, generally divided into: empirical model, theoretical model and improved empirical model; the acquisition of empirical model is to obtain a fixed fitting formula from empirical data, and the typical model is Cost 231 -Hata, Okumura, etc.; the theoretical model is based on the electromagnetic wave propagation theory, considering the reflection, diffraction, refraction, etc. of electromagnetic waves in space to calculate the loss, the more representative is the Volcano model; the improved empirical model is based on fitting More parameters are introduced into the formula to provide calculation models for more detailed classification scenarios, typically Standard Propagation Model (SPM).
在实际传播模型建模中,为了获得符合目标地区实际环境的传播模型,需要收集大量额外的实测数据、工程参数以及电子地图用来对传播模型进行校正;此外无线LTE网络已在全球普及,全球几十亿用户,每时每刻都会产生大量数据;如何合理地运用这些数据来辅助无线网络建设就成为了一个重要的课题。In the actual propagation model modeling, in order to obtain a propagation model that conforms to the actual environment of the target area, a large amount of additional measured data, engineering parameters and electronic maps need to be collected to correct the propagation model; Billions of users generate a large amount of data every moment; how to use these data reasonably to assist the construction of wireless networks has become an important issue.
近年来,大数据驱动的AI机器学习技术获得了长足的进步,并且在语言、图像处理领域获得了非常成功的运用;伴随着并行计算架构的发展,机器学习技术也具备了在线运算的能力,其高实时性以及低复杂度使得其与无线通信的紧密结合成为了可能。In recent years, AI machine learning technology driven by big data has made great progress, and has been used very successfully in the fields of language and image processing; with the development of parallel computing architecture, machine learning technology also has the ability of online computing, Its high real-time performance and low complexity make it possible to integrate it closely with wireless communication.
发明内容SUMMARY OF THE INVENTION
针对上述问题,本发明提供了一种基于地理信息的5G参考信号接收功率预测方法;运用机器学习来建立无线传播模型,并利用模型准确预测在新环境下无线信号覆盖强度,从而大大减少网络建设成本,提高网络建设效率。In view of the above problems, the present invention provides a 5G reference signal received power prediction method based on geographic information; machine learning is used to establish a wireless propagation model, and the model is used to accurately predict the wireless signal coverage strength in a new environment, thereby greatly reducing network construction. cost and improve the efficiency of network construction.
本发明的技术方案是:本发明所述的一种基于地理信息的5G参考信号接收功率预测方法,其特征在于:包括以下步骤:The technical solution of the present invention is: a method for predicting the received power of a 5G reference signal based on geographic information according to the present invention, which is characterized by comprising the following steps:
步骤(1.1)、根据地图信息构建特征地图;Step (1.1), construct a feature map according to the map information;
步骤(1.2)、使用人工智能构建误差迭代修正模型,将构建的特征地图与实际的信号接受功率共同训练构建的误差迭代修正模型;Step (1.2), use artificial intelligence to construct an error iterative correction model, and jointly train the constructed error iterative correction model with the actual signal receiving power;
在步骤(1.1)中,所述特征地图具体构建过程的步骤如下:In step (1.1), the steps of the specific construction process of the feature map are as follows:
(1.1.1)、首先,构建高度特征图:其包括构建发射机相对地面的高度特征,构建栅格与信号线的相对高度特征,构建小区发射机相对地面的高度特征,构建小区站点所在栅格(Cell X,Cell Y)的建筑物高度特征,构建小区站点所在栅格(Cell X,Cell Y)的海拔高度特征及信号接收端所在的栅格(X,Y)上的海拔高度特征;(1.1.1) First, construct the height feature map: it includes constructing the height feature of the transmitter relative to the ground, constructing the relative height feature of the grid and the signal line, constructing the height feature of the cell transmitter relative to the ground, constructing the grid where the cell site is located. The building height feature of the grid (Cell X, Cell Y), and the altitude feature of the grid (Cell X, Cell Y) where the cell site is located and the altitude feature of the grid (X, Y) where the signal receiving end is located;
(1.1.2)、其次,构建场景特征图:对发射机与信号接收端的位置进行定位,具体操作是:5G信号由发射机发出,最终由信号接收端接收;(1.1.2) Second, build a scene feature map: locate the position of the transmitter and the signal receiving end, the specific operation is: the 5G signal is sent by the transmitter and finally received by the signal receiving end;
从发射机发出的5G信号在传输到信号接收端中,通过海拔、建筑及场景不同的20种地形的统计构建场景特征图;The 5G signal sent from the transmitter is transmitted to the signal receiving end, and the scene feature map is constructed through the statistics of 20 kinds of terrains with different altitudes, buildings and scenes;
(1.1.3)、最后,构建信号特征图:通过Cost 231-Hata特征获取发射机与信号接收端的位置,计算由发射机到信号接收端的距离,获取发射机的5G信号发射功率;将Cost231-Hata特征、发射机到信号接收端的距离、发射机的5G信号发射功率构建成信号特征图;(1.1.3) Finally, construct the signal feature map: obtain the position of the transmitter and the signal receiving end through the Cost 231-Hata feature, calculate the distance from the transmitter to the signal receiving end, and obtain the 5G signal transmission power of the transmitter; The Hata feature, the distance from the transmitter to the signal receiving end, and the 5G signal transmission power of the transmitter are constructed into a signal feature map;
在所述步骤(1.1.2)中,所述的地形包括:海洋,内陆湖泊,湿地,城郊开阔区域,市区开阔区域,道路开阔区域,植被区,灌木植被,森林植被,城区超高层建筑(>60m),城区高层建筑(40m-60m),城区中高层建筑(20m-40m),城区<20m高密度建筑群,城区<20m多层建筑,低密度工业建筑区域,高密度工业建筑区域,城郊,发达城郊区域,农村,CBD商务圈地形的栅格数目;In the step (1.1.2), the terrain includes: ocean, inland lake, wetland, suburban open area, urban open area, road open area, vegetation area, shrub vegetation, forest vegetation, urban super high-rise Buildings (>60m), high-rise buildings in urban areas (40m-60m), high-rise buildings in urban areas (20m-40m), high-density buildings in urban areas <20m, multi-storey buildings in urban areas <20m, low-density industrial building areas, high-density industrial buildings The number of grids of the terrain of the area, suburban, developed suburban area, rural area, CBD business circle;
在步骤(1.1.3)中,所述Cost 231-Hata特征的计算方法如下:In step (1.1.3), the calculation method of the Cost 231-Hata feature is as follows:
其中,d表示发射机到信号接收端的距离,(Cell X,Cell Y)与(X,Y)分别表示小区站点所在栅格与信号接收端所在的栅格;Among them, d represents the distance from the transmitter to the signal receiving end, (Cell X, Cell Y) and (X, Y) respectively represent the grid where the cell site is located and the grid where the signal receiving end is located;
PL=46.3+33.9 log10 f-13.82 log10 hb-α+(44.9-6.55 log10 hue)log10 d+Cm (2)PL=46.3+33.9 log 10 f-13.82 log 10 h b -α+(44.9-6.55 log 10 h ue )log 10 d+C m (2)
其中,PL表示传播路径损耗,f表示载波频率,hb表示基站天线有效高度,α表示用户天线高度纠正项,hue表示用户天线有效高度,Cm表示场景纠正常数;Among them, PL represents the propagation path loss, f represents the carrier frequency, h b represents the base station antenna effective height, α represents the user antenna height correction term, h ue represents the user antenna effective height, and C m represents the scene correction constant;
其中,RSRP与PL的关系:RSRP=Pt-PL (3)Among them, the relationship between RSRP and PL: RSRP=P t -PL (3)
RSRP表示Cost 231-Hata特征计算的信号接收功率,Pt表示小区发射机发射功率;RSRP represents the signal received power calculated by the Cost 231-Hata feature, and P t represents the transmit power of the cell transmitter;
所述的Cm由场景特征图经过卷积神经网络计算所得;hb与hue由栅格与信号线的相对高度、小区发射机相对地面的高度、小区站点所在栅格(Cell X,Cell Y)的海拔高度、信号接收端所在的栅格(X,Y)上的海拔高度及小区站点所在栅格(Cell X,Cell Y)的建筑物高度经过BP神经网络计算所得;The C m is calculated by the scene feature map through the convolutional neural network; h b and h ue are determined by the relative height of the grid and the signal line, the height of the cell transmitter relative to the ground, the cell site where the cell is located (Cell X, Cell The altitude of Y), the altitude on the grid (X, Y) where the signal receiving end is located, and the building height of the grid (Cell X, Cell Y) where the cell site is located are calculated by the BP neural network;
其中,所述栅格与信号线的相对高度的计算方法如下:Wherein, the calculation method of the relative height of the grid and the signal line is as follows:
Δhv=Height+Cell Altitude-tan(Electrial Downtilt+MechanicalDowntilt) (4)Δh v =Height+Cell Altitude-tan(Electrial Downtilt+MechanicalDowntilt) (4)
式(4)中,Δhv=Height+Cell Altitude-tan(Electrial Downtilt+MechanicalDowntilt)中的Height表示小区发射机相对地面的高度,In formula (4), Δh v =Height+Cell Altitude-tan(Electrial Downtilt+MechanicalDowntilt), Height in the cell transmitter relative to the ground,
Δhv=Height+Cell Altitude-tan(ElectrialDowntilt+Mechanical Downtilt)分别表示小区站点所在栅格(Cell X,Cell Y)的海拔高度、信号接收端所在的栅格(X,Y)上的海拔高度、小区发射机垂直电下倾角及小区发射机垂直机械下倾角;其中每个栅格为5米乘5米的正方形;Δh v =Height+Cell Altitude-tan(ElectrialDowntilt+Mechanical Downtilt) represents the altitude of the grid (Cell X, Cell Y) where the cell site is located, the altitude on the grid (X, Y) where the signal receiving end is located, The vertical electrical downtilt angle of the cell transmitter and the vertical mechanical downtilt angle of the cell transmitter; each grid is a square of 5 meters by 5 meters;
在步骤(1.2)中,所述构建误差迭代修正模型的具体步骤如下:In step (1.2), the specific steps of constructing the error iterative correction model are as follows:
(1.2.1)、设计一个神经网络模型Model1(x):pre=Model1(x) (5)(1.2.1), design a neural network model Model 1 (x): pre=Model 1 (x) (5)
其中,所述x表示神经网络模型Model1(x)输入值,所述的输入值包括:场景特征图、高度特征图与信号特征图;pre表示神经网络模型Model1(x)的输出;Wherein, the x represents the input value of the neural network model Model 1 (x), and the input value includes: the scene feature map, the height feature map and the signal feature map; pre represents the output of the neural network model Model 1 (x);
其中,所述神经网络模型包含误差反向传播算法、卷积神经网络、激活层、池化层、全连接层;使用均方误差作为损失函数,通过亚当优化器来训练;Wherein, the neural network model includes an error back-propagation algorithm, a convolutional neural network, an activation layer, a pooling layer, and a fully connected layer; the mean square error is used as the loss function, and the Adam optimizer is used for training;
针对场景特征图:使用不同大小的卷积核进行卷积运算,首先在输入层中的边界进行补0操作,在卷积层中每个卷积核与补0后的输入序列从序列首端做内积运算一直到序列末端,得到输出层的值,形成新的特征图;再通过全连接层得到场景纠正常数;For the scene feature map: use convolution kernels of different sizes to perform convolution operations. First, perform zero-filling operations on the boundaries in the input layer. In the convolutional layer, each convolution kernel and the input sequence after zero-filling are performed from the beginning of the sequence. Do the inner product operation until the end of the sequence, get the value of the output layer, and form a new feature map; then get the scene correction constant through the fully connected layer;
针对高度特征图:使用误差反向传播算法对高度特征进行计算;得到基站天线有效高度及用户天线有效高度;根据式(1)-(3)得到Cost 231-Hata特征计算的信号接收功率;For the height feature map: use the error back propagation algorithm to calculate the height feature; obtain the effective height of the base station antenna and the effective height of the user antenna; obtain the signal received power calculated by the Cost 231-Hata feature according to equations (1)-(3);
针对信号特征图:将RSRP、d、Pt与场景特征构成特征向量,使用不同大小的卷积核进行卷积运算,在卷积层中每个卷积核与输入层中边界补0的输入序列从序列首端做内积运算一直到序列末端,得到输出层的值,形成新的特征图f1;For the signal feature map: RSRP, d, P t and scene features are formed into feature vectors, and convolution kernels of different sizes are used to perform convolution operations. In the convolution layer, each convolution kernel and the input layer in the input layer are filled with 0 input. The sequence performs the inner product operation from the beginning of the sequence to the end of the sequence, obtains the value of the output layer, and forms a new feature map f 1 ;
将f1通过Relu激活函数激活,所述Relu激活函数如式(6)所示:Activate f 1 through the Relu activation function, and the Relu activation function is shown in formula (6):
Ac=max(0,f1) (6)Ac=max(0, f 1 ) (6)
其中,Ac表示激活层输出的矩阵,对Ac进行最大池化操作,ma表示最大池化的长度,Poi表示最大池化的结果:Among them, Ac represents the matrix output by the activation layer, and the maximum pooling operation is performed on Ac, ma represents the length of the maximum pooling, and Po i represents the result of the maximum pooling:
Poi=max({Aci,Aci+1…Aci+ma-2,Aci+ma-1}) (7)Po i =max({Ac i , Ac i+1 ...Ac i+ma-2 , Ac i+ma-1 }) (7)
其中,Aci为矩阵中第i个元素,将池化结果Poi输入到全连接层中进行分类,把分布式特征映射到样本标记空间;所述全连接层由每个池化层连成一个一维向量,先经过隐含层神经元的计算,最后再连接一个神经元输出构成;每个神经元与的计算方法如式(8)所示:Among them, A i is the ith element in the matrix, the pooling result Po i is input into the fully connected layer for classification, and the distributed features are mapped to the sample label space; the fully connected layer is connected by each pooling layer. A one-dimensional vector is first calculated by neurons in the hidden layer, and finally connected to a neuron output; the calculation method of each neuron is as shown in formula (8):
pre=∑i Poi·W (8)pre=∑ i Po i ·W (8)
最终;pre为Model1预测的结果,W表示网络中的参数;Finally; pre is the result predicted by Model 1 , and W represents the parameters in the network;
(1.2.2)、训练神经网络模型Model1(W,b)期望得到最小化的损失函数:(1.2.2), training the neural network model Model 1 (W, b) expects to get the minimized loss function:
Loss(Model1(W,b),Label)=(pre-Label)2 (9)Loss(Model 1 (W, b), Label) = (pre-Label) 2 (9)
其中,Label表示每个数据的标签,W与b分别表示神经网络模型Model1中的参数与偏置;Among them, Label represents the label of each data, and W and b represent the parameters and biases in the neural network model Model 1 respectively;
(1.2.3)、计算误差函数:(1.2.3), calculate the error function:
Error(Model1(x),Label)=pre-Label (10)Error(Model 1 (x), Label) = pre-Label (10)
其中,Error表示每个数据x经过训练后神经网络模型Model1(x)映射之后得到的pre与预期的Label产生的误差;Among them, Error represents the error between the pre and the expected Label obtained after each data x is mapped by the neural network model Model 1 (x) after training;
(1.2.4)、设计一个与神经网络模型Model1(x)结构相同的神经网络模型Model2(x),用来修正神经网络模型Model1(x)在预测参考信号功率时产生的误差:(1.2.4), design a neural network model Model 2 (x) with the same structure as the neural network model Model 1 (x) to correct the error generated by the neural network model Model 1 (x) when predicting the reference signal power:
pre2=Model2(x) (11)pre 2 = Model 2 (x) (11)
其中,pre2表示神经网络模型Model1(x)预测信号功率产生的误差;Among them, pre 2 represents the error generated by the neural network model Model 1 (x) predicting the signal power;
(1.2.5)、令Label2=Error,训练神经网络模型Model2(x)期望得到最小化的损失函数:(1.2.5), let Label 2 =Error, train the neural network model Model 2 (x) and expect to get the minimized loss function:
Loss(Model2(x),Label)=(pre2-Label2)2 (12)Loss(Model 2 (x), Label)=(pre 2 -Label 2 ) 2 (12)
Label2表示神经网络模型Model2(x)预测的值,Error表示神经网络模型Model1(x)在预测信号功率时产生的误差;Label 2 represents the value predicted by the neural network model Model 2 (x), and Error represents the error generated by the neural network model Model 1 (x) when predicting the signal power;
(1.2.6)、使用神经网络模型Model2(x)的结果修正神经网络模型Model1(x)产生的误差,其计算方法如下:(1.2.6), use the results of the neural network model Model 2 (x) to correct the error generated by the neural network model Model 1 (x), and the calculation method is as follows:
Pre3=Model1(x)+Model2(x) (13);Pre3 = Model1( x )+Model2(x)(13);
在步骤(1.2.2)中,所述训练神经网络模型Model1(W,b)期望得到最小化的损失函数的具体步骤如下:对于给定的迭代次数,首先基于在整个数据集上求出的损失函数loss(W)对输入的参数向量W计算梯度向量;然后对参数W进行更新:对参数W减去梯度值乘学习率的值,也就是在反梯度方向,更新参数;In step (1.2.2), the specific steps of training the neural network model Model 1 (W, b) to obtain the minimized loss function are as follows: For a given number of iterations, first obtain the The loss function loss(W) calculates the gradient vector for the input parameter vector W; then updates the parameter W: subtract the value of the gradient value multiplied by the learning rate from the parameter W, that is, in the reverse gradient direction, update the parameter;
其中,loss(W)表示参数梯度下降方向,即loss(W)的偏导数,η为学习率;Label表示样本的真实值;当完成迭代时,实现W的更新与模型的建立:in, loss(W) represents the direction of parameter gradient descent, that is, the partial derivative of loss(W), η is the learning rate; Label represents the true value of the sample; when the iteration is completed, the update of W and the establishment of the model are realized:
Loss(W)=(Label-(Model1(W)+Model2(W)))2 (14)Loss(W)=(Label-(Model 1 (W)+Model 2 (W))) 2 (14)
本发明的有益效果是:本发明充分考虑了信号传播经过的所有地理信息,并通过地理信息构建的地图特征进行分析,最后使用人工智能模型对信号接收功率进行预测,最终均方误差降低到了40-50dBm之间,平均每个样本的误差在6-7dBm之间;而使用Cost 231-Hata方差将达到346.79dBm,平均每个样本的误差在18-19dBm之间。The beneficial effects of the present invention are as follows: the present invention fully considers all geographic information through which the signal propagates, analyzes the map features constructed by the geographic information, and finally uses an artificial intelligence model to predict the signal received power, and finally the mean square error is reduced to 40% Between -50dBm, the average error per sample is between 6-7dBm; while using Cost 231-Hata, the variance will reach 346.79dBm, and the average error per sample is between 18-19dBm.
附图说明Description of drawings
图1(a)是本发明中建筑物高度热力示意图;1(b)是本发明中海拔高度热力图;1(c)本发明中地物类型索引图;Figure 1 (a) is a schematic diagram of the thermal power of building height in the present invention; 1 (b) is a thermal diagram of altitude in the present invention; 1 (c) is an index diagram of feature types in the present invention;
图2是本发明中发射机参数示意图;Fig. 2 is a schematic diagram of transmitter parameters in the present invention;
图3是本发明中地形特征结构示意图;Fig. 3 is the topographic feature structure schematic diagram in the present invention;
图4是本发明中Model1(x)神经网络模型示意图;Fig. 4 is Model 1 (x) neural network model schematic diagram in the present invention;
图5是本发明中均方差损失结构图;Fig. 5 is the structure diagram of mean square error loss in the present invention;
图6是本发明中弱覆盖识别率曲线图。FIG. 6 is a graph showing the recognition rate of weak coverage in the present invention.
具体实施方式Detailed ways
为了更清楚地说明本发明的技术方案,下面结合附图对本发明的技术方案做进一步的详细说明:In order to illustrate the technical solution of the present invention more clearly, the technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings:
本实施例的主要思想为:首先将构建高度特征图,场景特征图与信号特征图;接着根据这些特征图设计神经网络模型Model1(x),并使用卫星真实采集的数据集来对神经网络模型Model1(x)中的权重进行优化;最终在模型一的基础上使用结构相同的模型二对模型一产生的误差进行误差修正;The main idea of this embodiment is as follows: first, a height feature map, a scene feature map and a signal feature map will be constructed; then a neural network model Model 1 (x) will be designed according to these feature maps, and the neural network will be analyzed by using the data set actually collected by the satellite. The weights in Model 1 (x) are optimized; finally, on the basis of
具体的,如图所述;一种基于地理信息的5G参考信号接收功率预测方法,包括以下步骤:Specifically, as shown in the figure; a method for predicting the received power of a 5G reference signal based on geographic information, comprising the following steps:
步骤(1.1)、根据卫星地图采集真实的地图信息,通过这些地图信息来构建特征地图;Step (1.1), collect the real map information according to the satellite map, and construct the feature map through these map information;
步骤(1.2)、使用人工智能构建误差迭代修正模型,将构建的特征地图与实际的信号接受功率共同训练构建的误差迭代修正模型;其作用是:将数据训练好的误差迭代修正模型来预测的信号接收功率;所述误差迭代修正模型以每个信号源与接受信号的地理信息提取的特征作为输入,以接收方接收到信号源发出的信号功率为输出;Step (1.2), use artificial intelligence to build an error iterative correction model, and jointly train the constructed feature map and the actual signal receiving power to build an error iterative correction model; Signal received power; the error iterative correction model takes the features extracted from each signal source and the geographic information of the received signal as input, and takes the signal power received by the receiver from the signal source as the output;
其中的,实际的信号接收功率是指:在实际场景中,通过发射机发射的5G信号在信号接收端接收到的实际功率。Among them, the actual signal received power refers to: in the actual scenario, the actual power received at the signal receiving end of the 5G signal transmitted by the transmitter.
进一步的,在步骤(1.1)中,Further, in step (1.1),
真实采集的地图信息,共有8个字段(如表一),各字段对应含义如表一所示;考虑地图类型的多样性和复杂性,城区、农村、湖泊等实际地物被抽象为数字,这些数字称为地物类型名称编号(Clutter Index),在表二中可以看到地物类型名称编号所对应的实际地物类型;The real collected map information has a total of 8 fields (as shown in Table 1), and the corresponding meanings of each field are shown in Table 1; These numbers are called Clutter Index, and in Table 2 you can see the actual type of objects corresponding to the Clutter Index;
表一,地图数据的字段含义:Table 1, field meanings of map data:
表二,地物类型名称的编号含义:Table 2, the number meaning of the name of the feature type:
与工程参数数据一样,地图数据也进行了栅格化处理,每个栅格代表了5m×5m的区域,其中(X,Y)记录了地图所在栅格的左上角坐标;Like the engineering parameter data, the map data is also rasterized, each grid represents an area of 5m × 5m, where (X, Y) records the coordinates of the upper left corner of the grid where the map is located;
所述特征地图的构建过程,具体步骤如下:The specific steps of the construction process of the feature map are as follows:
(1.1.1)、首先,构建高度特征图:其包括构建发射机相对地面的高度特征,构建栅格与信号线的相对高度特征,构建小区发射机相对地面的高度特征,构建小区站点所在栅格(Cell X,Cell Y)的建筑物高度特征,构建小区站点所在栅格(Cell X,Cell Y)的海拔高度特征及信号接收端所在的栅格(X,Y)上上的海拔高度特征;关于建筑物高度、海拔高度与地物类型索引分别为图1(a)、(b)、(c)所示;关于发射机的相关参数如图2所示;(1.1.1) First, construct the height feature map: it includes constructing the height feature of the transmitter relative to the ground, constructing the relative height feature of the grid and the signal line, constructing the height feature of the cell transmitter relative to the ground, constructing the grid where the cell site is located. The building height feature of the grid (Cell X, Cell Y), the altitude feature of the grid (Cell X, Cell Y) where the cell site is located, and the altitude feature on the grid (X, Y) where the signal receiving end is located ; The building height, altitude and ground object type index are shown in Figure 1(a), (b), (c) respectively; the relevant parameters of the transmitter are shown in Figure 2;
(1.1.2)、其次,构建场景特征图:对发射机与信号接收端(如手机、电脑、智能终端等具有5G信号接收功能的设备)的位置进行定位,具体操作是:5G信号由发射机发出,通过直线发送,最终由信号接收端接收;(1.1.2) Second, build a scene feature map: locate the position of the transmitter and the signal receiving end (such as mobile phones, computers, smart terminals and other devices with 5G signal receiving function), the specific operation is: 5G signals are transmitted by It is sent by the machine, sent through a straight line, and finally received by the signal receiving end;
从发射机发出的5G信号在传输到信号接收端中,通过海拔、建筑及场景不同的20种地形的统计构建场景特征图;所述的地形包括:海洋,内陆湖泊,湿地,城郊开阔区域,市区开阔区域,道路开阔区域,植被区,灌木植被,森林植被,城区超高层建筑(>60m),城区高层建筑(40m-60m),城区中高层建筑(20m-40m),城区<20m高密度建筑群,城区<20m多层建筑,低密度工业建筑区域,高密度工业建筑区域,城郊,发达城郊区域,农村,CBD商务圈地形的栅格数目;具体的场景特征图如图3所示;The 5G signal sent from the transmitter is transmitted to the signal receiving end, and the scene feature map is constructed through the statistics of 20 types of terrain with different altitudes, buildings and scenes; the terrain includes: oceans, inland lakes, wetlands, suburban open areas , urban open area, open road area, vegetation area, shrub vegetation, forest vegetation, urban super high-rise buildings (>60m), urban high-rise buildings (40m-60m), urban medium and high-rise buildings (20m-40m), urban areas <20m The number of grids of high-density buildings, multi-storey buildings of less than 20m in urban areas, low-density industrial building areas, high-density industrial building areas, suburban areas, developed suburban areas, rural areas, and CBD business circles; the specific scene feature map is shown in Figure 3 Show;
通过不等式可以计算出信号由发射机到信号接收端所在直线上每个点半径为五个栅格的的地形特征坐标;其中(CellX,CellY),(X,Y)与(fx,fy)分别为发射机坐标、接收方坐标和途径的地形坐标;统计出所有(fx,fy)的地形栅格个数就是所构建的地形的特征图;through the inequality It can calculate the terrain feature coordinates of each point on the line from the transmitter to the signal receiving end with a radius of five grids; where (CellX, CellY), (X, Y) and (fx, fy) are the transmission The coordinates of the machine, the coordinates of the receiver and the terrain coordinates of the route; the number of all (fx, fy) terrain grids is counted as the feature map of the constructed terrain;
(1.1.3)、最后,构建信号特征图:通过Cost 231-Hata特征获取发射机与信号接收端的位置,计算由发射机到信号接收端的距离,获取发射机的5G信号发射功率;将Cost231-Hata特征、发射机到信号接收端的距离、发射机的5G信号发射功率构建成信号特征图;(1.1.3) Finally, construct the signal feature map: obtain the position of the transmitter and the signal receiving end through the Cost 231-Hata feature, calculate the distance from the transmitter to the signal receiving end, and obtain the 5G signal transmission power of the transmitter; The Hata feature, the distance from the transmitter to the signal receiving end, and the 5G signal transmission power of the transmitter are constructed into a signal feature map;
所述Cost 231-Hata特征的计算方法,其具体步骤操作如下:The calculation method of the described Cost 231-Hata feature, its concrete steps are as follows:
其中,d表示发射机到信号接收端的距离,(CellX,CellY)与(X,Y)分别表示小区站点所在栅格与信号接收端所在的栅格;Among them, d represents the distance from the transmitter to the signal receiving end, (CellX, CellY) and (X, Y) respectively represent the grid where the cell site is located and the grid where the signal receiving end is located;
PL=46.3+33.9 log10 f-13.82 log10 hb-α+(44.9-6.55 log10 hue)log10 d+Cm (2)PL=46.3+33.9 log 10 f-13.82 log 10 h b -α+(44.9-6.55 log 10 h ue )log 10 d+C m (2)
其中;PL表示传播路径损耗(dB),f表示载波频率(MHz),hb表示基站天线有效高度(m),hue表示用户天线有效高度(m),α表示用户天线高度纠正项(dB),d表示发射机到信号接收端的距离(km),Cm表示场景纠正常数(dB);Among them; PL represents the propagation path loss (dB), f represents the carrier frequency (MHz), h b represents the effective height of the base station antenna (m), h ue represents the effective height of the user antenna (m), α represents the user antenna height correction term (dB ), d represents the distance from the transmitter to the signal receiving end (km), and C m represents the scene correction constant (dB);
其中,RSRP与PL的关系为:RSRP=Pt-PL (3)Among them, the relationship between RSRP and PL is: RSRP=P t -PL (3)
RSRP表示Cost 231-Hata特征计算的信号接收功率,Pt表示小区发射机发射功率(dBm);RSRP represents the signal received power calculated by the Cost 231-Hata feature, and P t represents the cell transmitter transmit power (dBm);
所述的Cm由场景特征图经过卷积神经网络计算所得;hb与hue由栅格与信号线的相对高度、小区发射机相对地面的高度、小区站点所在栅格(Cell X,Cell Y)的海拔高度、信号接收端所在的栅格(X,Y)上的海拔高度及小区站点所在栅格(Cell X,Cell Y)的建筑物高度经过BP神经网络计算所得;The C m is calculated by the scene feature map through the convolutional neural network; h b and h ue are determined by the relative height of the grid and the signal line, the height of the cell transmitter relative to the ground, the cell site where the cell is located (Cell X, Cell The altitude of Y), the altitude on the grid (X, Y) where the signal receiving end is located, and the building height of the grid (Cell X, Cell Y) where the cell site is located are calculated by the BP neural network;
其中,所述栅格与信号线的相对高度的计算方法如下:Wherein, the calculation method of the relative height of the grid and the signal line is as follows:
Δhv=Height+Cell Altitude-tan(Electrial Downtilt+MechanicalDowntilt) (4)Δh v =Height+Cell Altitude-tan(Electrial Downtilt+MechanicalDowntilt) (4)
式(4)中,Δhv=Height+Cell Altitude-tan(Electrial Downtilt+MechanicalDowntilt)中的Height表示小区发射机相对地面的高度,In formula (4), Δh v =Height+Cell Altitude-tan(Electrial Downtilt+MechanicalDowntilt), Height in the cell transmitter relative to the ground,
Δhv=Height+Cell Altitude-tan(Electrial Downtilt+MechanicalDowntilt)分别表示小区站点所在栅格(Cell X,Cell Y)的海拔高度、信号接收端所在的栅格(X,Y)上的海拔高度、小区发射机垂直电下倾角及小区发射机垂直机械下倾角;其中每个栅格为5米乘5米的正方形;Δh v =Height+Cell Altitude-tan(Electrial Downtilt+MechanicalDowntilt) represents the altitude of the grid (Cell X, Cell Y) where the cell site is located, the altitude on the grid (X, Y) where the signal receiving end is located, The vertical electrical downtilt angle of the cell transmitter and the vertical mechanical downtilt angle of the cell transmitter; each grid is a square of 5 meters by 5 meters;
在步骤(1.2)中,所述构建误差迭代修正模型的具体步骤如下:In step (1.2), the specific steps of constructing the error iterative correction model are as follows:
(1.2.1)、设计一个神经网络模型Model1(x):pre=Model1(x) (5)(1.2.1), design a neural network model Model 1 (x): pre=Model 1 (x) (5)
所述x表示神经网络模型Model1(x)的输入值,所述的输入值包括:场景特征图、高度特征图与信号特征图;pre表示神经网络模型Model1(x)的输出;The x represents the input value of the neural network model Model 1 (x), and the input value includes: a scene feature map, a height feature map and a signal feature map; pre represents the output of the neural network model Model 1 (x);
其中,所述神经网络模型包含误差反向传播算法、卷积神经网络、激活层、池化层、全连接层;使用均方误差作为损失函数,通过亚当优化器来训练;Wherein, the neural network model includes an error back-propagation algorithm, a convolutional neural network, an activation layer, a pooling layer, and a fully connected layer; the mean square error is used as the loss function, and the Adam optimizer is used for training;
针对场景特征图:使用不同大小的卷积核进行卷积运算,首先在输入层中的边界进行补0操作,在卷积层中每个卷积核与补0后的输入序列从序列首端做内积运算一直到序列末端,得到输出层的值,形成新的特征图;再通过全连接层得到场景纠正常数;For the scene feature map: use convolution kernels of different sizes to perform convolution operations. First, perform zero-filling operations on the boundaries in the input layer. In the convolutional layer, each convolution kernel and the input sequence after zero-filling are performed from the beginning of the sequence. Do the inner product operation until the end of the sequence, get the value of the output layer, and form a new feature map; then get the scene correction constant through the fully connected layer;
针对高度特征图:使用误差反向传播算法对高度特征进行计算;得到基站天线有效高度及用户天线有效高度;根据式(1)-(3)得到Cost 231-Hata特征计算的信号接收功率;For the height feature map: use the error back propagation algorithm to calculate the height feature; obtain the effective height of the base station antenna and the effective height of the user antenna; obtain the signal received power calculated by the Cost 231-Hata feature according to equations (1)-(3);
针对信号特征图:将RSRP、d、Pt与场景特征构成特征向量,使用不同大小的卷积核进行卷积运算,在卷积层中每个卷积核与输入层中边界补0的输入序列从序列首端做内积运算一直到序列末端,得到输出层的值,形成新的特征图f1;For the signal feature map: RSRP, d, P t and scene features are formed into feature vectors, and convolution kernels of different sizes are used to perform convolution operations. In the convolution layer, each convolution kernel and the input layer in the input layer are filled with 0 input. The sequence performs the inner product operation from the beginning of the sequence to the end of the sequence, obtains the value of the output layer, and forms a new feature map f 1 ;
将f1通过Relu激活函数激活,所述Relu激活函数如式(6)所示:Activate f 1 through the Relu activation function, and the Relu activation function is shown in formula (6):
Ac=max(0,f1) (6)Ac=max(0, f 1 ) (6)
Ac表示激活层输出的矩阵,对Ac进行最大池化操作,其中,ma表示最大池化的长度,Poi表示最大池化的结果:Ac represents the matrix output by the activation layer, and performs the maximum pooling operation on Ac, where ma represents the length of the maximum pooling, and Po i represents the result of the maximum pooling:
Poi=max({Aci,Aci+1...Aci+ma-2,Aci+ma-1}) (7)Po i =max({Ac i , Ac i+1 ...Ac i+ma-2 , Ac i+ma-1 }) (7)
其中,Aci为矩阵中第i个元素,将池化结果Poi输入到全连接层中进行分类,把分布式特征映射到样本标记空间;所述全连接层由每个池化层连成一个一维向量,先经过隐含层神经元的计算,最后再连接一个神经元输出构成;每个神经元与的计算方法如式(8)所示:Among them, A i is the ith element in the matrix, the pooling result Po i is input into the fully connected layer for classification, and the distributed features are mapped to the sample label space; the fully connected layer is connected by each pooling layer. A one-dimensional vector is first calculated by neurons in the hidden layer, and finally connected to a neuron output; the calculation method of each neuron is as shown in formula (8):
pre=∑i Poi·W (8)pre=∑ i Po i ·W (8)
最终;pre为Model1预测的结果,W表示网络中的参数;Finally; pre is the result predicted by Model 1 , and W represents the parameters in the network;
(1.2.2)、训练神经网络模型Model1(W,b)期望得到最小化的损失函数:(1.2.2), training the neural network model Model 1 (W, b) expects to get the minimized loss function:
Loss(Model1(W,b),Label)=(pre-Label)2 (9)Loss(Model 1 (W, b), Label) = (pre-Label) 2 (9)
其中,Label表示每个数据的标签,W与b分别表示神经网络模型Model1中的参数与偏置;Among them, Label represents the label of each data, and W and b represent the parameters and biases in the neural network model Model 1 respectively;
(1.2.3)、计算误差函数:(1.2.3), calculate the error function:
Error(Model1(x),Label)=pre-Label (10)Error(Model 1 (x), Label) = pre-Label (10)
其中,Error表示每个数据x经过训练后神经网络模型Model1(x)映射之后得到的pre与预期的Label产生的误差;Among them, Error represents the error between the pre and the expected Label obtained after each data x is mapped by the neural network model Model 1 (x) after training;
(1.2.4)、设计一个与神经网络模型Model1(x)结构相同的神经网络模型Model2(x),用来修正神经网络模型Model1(x)在预测参考信号功率时产生的误差:(1.2.4), design a neural network model Model 2 (x) with the same structure as the neural network model Model 1 (x) to correct the error generated by the neural network model Model 1 (x) when predicting the reference signal power:
pre2=Model2(x) (11)pre 2 = Model 2 (x) (11)
其中,pre2表示神经网络模型Model1(x)预测信号功率产生的误差;Among them, pre 2 represents the error generated by the neural network model Model 1 (x) predicting the signal power;
(1.2.5)、令Label2=Error,训练神经网络模型Model2(x)期望得到最小化的损失函数:(1.2.5), let Label 2 =Error, train the neural network model Model 2 (x) and expect to get the minimized loss function:
Loss(Model2(x),Label)=(pre2-Label2)2 (12)Loss(Model 2 (x), Label)=(pre 2 -Label 2 ) 2 (12)
Label2表示神经网络模型Model2(x)预测的值,Error表示神经网络模型Model1(x)在预测信号功率时产生的误差;Label 2 represents the value predicted by the neural network model Model 2 (x), and Error represents the error generated by the neural network model Model 1 (x) when predicting the signal power;
(1.2.6)、使用神经网络模型Model2(x)的结果修正神经网络模型Model1(x)产生的误差,其计算方法如下:(1.2.6), use the results of the neural network model Model 2 (x) to correct the error generated by the neural network model Model 1 (x), and the calculation method is as follows:
Pre3=Model1(x)+Model2(x) (13);Pre3 = Model1( x )+Model2(x)(13);
其中,训练神经网络模型Model1(W,b)期望得到最小化的损失函数的具体步骤如下:对于给定的迭代次数,首先基于在整个数据集上求出的损失函数loss(W)对输入的参数向量W计算梯度向量;然后对参数W进行更新:对参数W减去梯度值乘学习率的值,也就是在反梯度方向,更新参数;Among them, the specific steps of training the neural network model Model 1 (W, b) to obtain the minimized loss function are as follows: For a given number of iterations, first, based on the loss function loss(W) calculated on the entire data set, the input The parameter vector W calculates the gradient vector; then update the parameter W: subtract the value of the gradient value multiplied by the learning rate from the parameter W, that is, in the reverse gradient direction, update the parameter;
其中,loss(W)表示参数梯度下降方向,即loss(W)的偏导数,η为学习率;Label表示样本的真实值;当完成迭代时,实现W的更新与模型的建立in, loss(W) represents the direction of parameter gradient descent, that is, the partial derivative of loss(W), η is the learning rate; Label represents the true value of the sample; when the iteration is completed, the update of W and the establishment of the model are realized
Loss(W)=(Label-(Model1(W)+Model2(W)))2 (14)Loss(W)=(Label-(Model 1 (W)+Model 2 (W))) 2 (14)
实施例:Example:
本发明已经应用在基于地理信息的5G信号接收功率分析系统当中,使用者需要选择出5G基站的位置,并选择出信号接收端的位置。此时,根据地图的建筑与海拔系统可以构建出高度特征;接着,该系统可以根据5G基站的位置、接收端的位置与卫星地图提供的信息进行计算,得出信号由发射机到信号接收端所需要经过的地形特征;然后,根据已知的基站参数与地图信息构建出信号特征;最终,将这些特征输入到已经训练完成的Model1(x)与Model2(x)中,根据公式(13)计算出Pre3;此时,Pre3就是该系统最终的输出结果,即,信号接收端对于选定基站的位置最终能接收到的5G信号功率;该系统对于5G基站的建设具有指导性作用,对于需要接收5G信号的位置,模拟计算出不同位置的5G基站最终能接收到的信号功率,减少了基站建设的试错成本。The present invention has been applied to the 5G signal received power analysis system based on geographic information, and the user needs to select the location of the 5G base station and the location of the signal receiving end. At this time, the height feature can be constructed according to the building and altitude system of the map; then, the system can calculate based on the location of the 5G base station, the location of the receiving end and the information provided by the satellite map, and obtain that the signal is transmitted from the transmitter to the signal receiving end. The terrain features that need to be passed; then, the signal features are constructed according to the known base station parameters and map information; finally, these features are input into the already trained Model 1 (x) and Model 2 (x), according to formula (13 ) to calculate Pre3; at this time, Pre3 is the final output result of the system, that is, the 5G signal power that the signal receiving end can finally receive for the location of the selected base station; this system has a guiding role for the construction of 5G base stations, and The location where the 5G signal needs to be received is simulated to calculate the final signal power that the 5G base station in different locations can receive, which reduces the trial and error cost of base station construction.
本系统对真实情况进行了测试,并用两个性能参数来对系统的性能进行了衡量,用来表示神经网络模型Model1(x)与Model2(x)的性能参数说明;This system has tested the real situation, and measured the performance of the system with two performance parameters, which are used to represent the performance parameter description of the neural network models Model 1 (x) and Model 2 (x);
均方差损失:LOSS=(label-Y)2/n (16)Mean squared loss: LOSS=(label-Y) 2 /n (16)
其中label表示训练数据集中已知的信号接收功率,Y表示模型最终输出的结果;n表示样本数量;最终上述模型的损失如图5所示;LOSS值越小表示预测得越准确;本发明充分考虑了信号传播经过的所有地理信息,并通过地理信息构建的地图特征进行分析,最后使用人工智能模型对真实采集数据集进行验证,对信号接收功率进行预测,最终均方误差降低到了40-50dBm之间,平均每个样本的误差在6-7dBm之间;而使用Cost 231-Hata均方误差将达到346.79dBm,平均每个样本的误差在18-19dBm之间。where label represents the known received signal power in the training data set, Y represents the final output result of the model; n represents the number of samples; the final loss of the above model is shown in Figure 5; the smaller the LOSS value, the more accurate the prediction; the present invention is sufficient All the geographic information that the signal propagates through is considered, and the map features constructed by the geographic information are used for analysis. Finally, the artificial intelligence model is used to verify the real collected data set, and the received signal power is predicted, and the final mean square error is reduced to 40-50dBm The average error per sample is between 6-7dBm; while using Cost 231-Hata, the mean square error will reach 346.79dBm, and the average error per sample is between 18-19dBm.
弱覆盖识别率:Weak coverage recognition rate:
表三TP、FP、FN和TN的定义:Table 3 Definitions of TP, FP, FN and TN:
PCRR综合考虑Precision(准确率)和Recall(召回率)的目标,其计算公式如下:PCRR comprehensively considers the goals of Precision (accuracy rate) and Recall (recall rate), and its calculation formula is as follows:
其中Precision可以理解为预测结果为弱覆盖的栅格实际也是弱覆盖的概率,其定义如下:Among them, Precision can be understood as the probability that the grid whose prediction result is weak coverage is actually also weakly covered, which is defined as follows:
Recall可以理解为真实结果为弱覆盖的栅格有多少被预测成了弱覆盖的概率,其定义如下:Recall can be understood as the probability of how many grids with weak coverage are predicted to be weak coverage, which is defined as follows:
最终神经网络模型的弱覆盖识别率如图6所示;在通常情况下,弱覆盖识别率至少需要达到20%才可以使用,本发明的弱覆盖识别率在本案例中达到了35%。The weak coverage recognition rate of the final neural network model is shown in Figure 6; under normal circumstances, the weak coverage recognition rate needs to reach at least 20% before it can be used, and the weak coverage recognition rate of the present invention reaches 35% in this case.
最后,应当理解的是,本发明中所述实施例仅用以说明本发明实施例的原则;其他的变形也可能属于本发明的范围;因此,作为示例而非限制,本发明实施例的替代配置可视为与本发明的教导一致;相应地,本发明的实施例不限于本发明明确介绍和描述的实施例。Finally, it should be understood that the embodiments described in the present invention are only used to illustrate the principles of the embodiments of the present invention; other modifications may also belong to the scope of the present invention; therefore, by way of example and not limitation, the embodiments of the present invention are alternatives to Configurations may be considered consistent with the teachings of the present invention; accordingly, embodiments of the present invention are not limited to those expressly introduced and described herein.
Claims (1)
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010729363.4A CN112367129B (en) | 2020-07-27 | 2020-07-27 | 5G reference signal received power prediction method based on geographic information |
PCT/CN2020/119194 WO2022021602A1 (en) | 2020-07-27 | 2020-09-30 | 5g reference signal received power prediction method based on geographic information |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010729363.4A CN112367129B (en) | 2020-07-27 | 2020-07-27 | 5G reference signal received power prediction method based on geographic information |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112367129A CN112367129A (en) | 2021-02-12 |
CN112367129B true CN112367129B (en) | 2022-02-11 |
Family
ID=74516890
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010729363.4A Active CN112367129B (en) | 2020-07-27 | 2020-07-27 | 5G reference signal received power prediction method based on geographic information |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN112367129B (en) |
WO (1) | WO2022021602A1 (en) |
Families Citing this family (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113644996B (en) * | 2021-10-13 | 2021-12-31 | 华中师范大学 | Cell-level RSRP estimation method based on deep learning |
CN114611388B (en) * | 2022-03-02 | 2022-10-25 | 中国人民解放军战略支援部队航天工程大学 | Wireless channel characteristic screening method based on artificial intelligence |
CN114826459B (en) * | 2022-03-08 | 2023-02-03 | 南京航空航天大学 | Spectrum map accurate construction method based on cross-domain reasoning |
CN115022889B (en) * | 2022-05-30 | 2023-07-11 | 中国人民解放军空军工程大学 | A 5G network RSRP prediction model and its establishment method and application |
CN115334561A (en) * | 2022-08-30 | 2022-11-11 | 中国电信股份有限公司 | Signal quality prediction method, signal quality prediction device, electronic device, and non-volatile storage medium |
CN115296761B (en) * | 2022-10-10 | 2022-12-02 | 香港中文大学(深圳) | A Channel Prediction Method Based on Electromagnetic Propagation Model |
CN116248207A (en) * | 2023-02-13 | 2023-06-09 | 北京吉亚斯科技有限公司 | Gias Microwave Interference Prediction Model |
CN117395700B (en) * | 2023-12-06 | 2024-10-11 | 南京大学 | A wireless signal coverage prediction method in hilly and mountainous areas considering terrain characteristics |
CN118300720B (en) * | 2024-04-29 | 2024-12-20 | 南京大学 | A method for predicting diffraction loss of wireless signals |
CN118678377B (en) * | 2024-08-16 | 2024-11-12 | 深圳市大数据研究院 | Spectrum coverage map construction method and device based on heterogeneous graph neural network |
CN120018067B (en) * | 2025-04-17 | 2025-07-08 | 北京久佳信通科技有限公司 | A method and system for processing short messages based on scene adaptation |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111405585A (en) * | 2020-03-19 | 2020-07-10 | 北京联合大学 | A Neighborhood Relation Prediction Method Based on Convolutional Neural Networks |
Family Cites Families (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR100439674B1 (en) * | 2001-09-27 | 2004-07-12 | (주) 엘지텔레콤 | Apparatus and method for failure prediction of radio port using neural network |
CN108683463A (en) * | 2017-12-15 | 2018-10-19 | 南京邮电大学 | A kind of propagation properties analysis method and analysis system |
CN108513306B (en) * | 2018-06-12 | 2021-12-14 | 北京中网华通设计咨询有限公司 | Network coverage prediction method based on test big data |
CN109151866B (en) * | 2018-11-27 | 2019-03-05 | 南京华苏科技有限公司 | A method for antenna azimuth correction based on user data |
WO2020154432A1 (en) * | 2019-01-22 | 2020-07-30 | Trellisware Technologies, Inc. | Controlling data communication quality in software-defined heterogenous multi-hop ad hoc networks |
CN109842887B (en) * | 2019-03-06 | 2021-10-29 | 南京邮电大学 | A Resource Mapping Method Considering User Mobility for Ultra-Dense Scenarios |
CN110912627B (en) * | 2019-12-24 | 2021-05-14 | 电子科技大学 | A data-driven cell received power prediction method |
CN111405586A (en) * | 2020-03-19 | 2020-07-10 | 长沙航空职业技术学院 | Wireless intelligent propagation method based on Cost231-Hata model |
-
2020
- 2020-07-27 CN CN202010729363.4A patent/CN112367129B/en active Active
- 2020-09-30 WO PCT/CN2020/119194 patent/WO2022021602A1/en active Application Filing
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111405585A (en) * | 2020-03-19 | 2020-07-10 | 北京联合大学 | A Neighborhood Relation Prediction Method Based on Convolutional Neural Networks |
Also Published As
Publication number | Publication date |
---|---|
WO2022021602A1 (en) | 2022-02-03 |
CN112367129A (en) | 2021-02-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112367129B (en) | 5G reference signal received power prediction method based on geographic information | |
Zhang et al. | Cellular network radio propagation modeling with deep convolutional neural networks | |
EP2659697B1 (en) | Method for the prediction of coverage areas of a cellular network | |
CN110418354A (en) | A Propagation-Free Model-Based Wireless Network Planning Method Based on Machine Learning | |
CN113329437B (en) | Wireless network signal propagation path loss prediction method and electronic equipment | |
JP2004524751A (en) | Prediction method based on measurement of radiation path loss | |
Morocho-Cayamcela et al. | Breaking wireless propagation environmental uncertainty with deep learning | |
CN114501530B (en) | Method and device for determining antenna parameters based on deep reinforcement learning | |
CN115314133B (en) | Path loss data enhancement method and system based on matrix completion | |
CN113644996B (en) | Cell-level RSRP estimation method based on deep learning | |
JP6696859B2 (en) | Quality estimation device and quality estimation method | |
Nguyen et al. | Deep learning for path loss prediction in the 3.5 GHz CBRS spectrum band | |
CN114611388A (en) | Wireless channel characteristic screening method based on artificial intelligence | |
CN119644337B (en) | High-space-time resolution weather radar beam propagation path inversion method | |
US20250080996A1 (en) | Novel scatterer density-based predictive channel modeling method | |
Nagao et al. | Geographical clustering of path loss modeling for wireless emulation in various environments | |
Nagao et al. | A study on urban structure map extraction for radio propagation prediction using XGBoost | |
CN115022889B (en) | A 5G network RSRP prediction model and its establishment method and application | |
CN117639981A (en) | Environment reconstruction and deterministic channel modeling method based on point cloud | |
CN115460537B (en) | Base station positioning method, device and network equipment | |
CN110995365B (en) | Radio frequency gene library establishing method and illegal radio wave detection method | |
Zhang et al. | FedRME: Federated Learning for Enhanced Distributed Radiomap Estimation | |
CN117135639B (en) | Wireless communication coverage prediction method, system and equipment based on neural network | |
Ebhota et al. | Signal power loss prediction based on artificial neural networks in microcell environment | |
Jassim et al. | A novel approach for outdoor signal strength prediction using machine learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |