WO2020107712A1 - 基于用户数据实现天线方位角纠偏的方法 - Google Patents

基于用户数据实现天线方位角纠偏的方法 Download PDF

Info

Publication number
WO2020107712A1
WO2020107712A1 PCT/CN2019/075043 CN2019075043W WO2020107712A1 WO 2020107712 A1 WO2020107712 A1 WO 2020107712A1 CN 2019075043 W CN2019075043 W CN 2019075043W WO 2020107712 A1 WO2020107712 A1 WO 2020107712A1
Authority
WO
WIPO (PCT)
Prior art keywords
cell
prediction
azimuth
user data
data
Prior art date
Application number
PCT/CN2019/075043
Other languages
English (en)
French (fr)
Inventor
王计斌
朱格苗
闫兴秀
Original Assignee
南京华苏科技有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 南京华苏科技有限公司 filed Critical 南京华苏科技有限公司
Priority to US16/960,686 priority Critical patent/US11206555B2/en
Publication of WO2020107712A1 publication Critical patent/WO2020107712A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
    • H04W16/22—Traffic simulation tools or models
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/142—Network analysis or design using statistical or mathematical methods
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/147—Network analysis or design for predicting network behaviour
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/16—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using machine learning or artificial intelligence
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
    • H04W16/18—Network planning tools
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/02—Arrangements for optimising operational condition
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/06—Testing, supervising or monitoring using simulated traffic
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/08—Testing, supervising or monitoring using real traffic
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/10—Scheduling measurement reports ; Arrangements for measurement reports
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
    • H04W16/24—Cell structures
    • H04W16/28—Cell structures using beam steering
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W88/00—Devices specially adapted for wireless communication networks, e.g. terminals, base stations or access point devices
    • H04W88/18—Service support devices; Network management devices

Definitions

  • This application relates to the field of network technology, and in particular to a method for realizing antenna azimuth correction based on user data.
  • the quality of the antenna feed system often lacks effective means of monitoring and problem detection. Therefore, the accuracy and rationality of working parameters have become the most prominent problem of optimization.
  • the reasons for the antenna working parameter errors are: the optimization staff did not submit the updated working parameters according to the regulations after the station adjustment, or the task was tight and the measurement was not strictly updated accurately; The area of the cell is small; there are many antennas in the sky and the antenna feeders are complex. The optimizer cannot find some antennas or find the wrong cell; the widely used measurement equipment is the compass, which is caused by the complex electromagnetic environment of the sky and a certain distance compared to the antenna. Measurement error.
  • the present invention provides an application of big data processing technology based on the collection, storage and analysis capabilities of mobile communication network data, using machine learning algorithms to comprehensively analyze multiple data sources to form a comprehensive evaluation of the network, thereby Clean the antenna working parameters. It can quickly find the problem of abnormal antenna coverage, improve the quality of the community's sky, provide strong support for community engineering construction, network maintenance, and wireless optimization, realize the improvement of network quality, and provide a powerful tool for the network intelligent operation and maintenance platform to directly update the working parameters. Technical support.
  • the method for implementing antenna azimuth correction based on user data includes the following steps:
  • Step 1 Collect data, collect user data and worker parameter data samples
  • Step 2 For data processing, first delete the cells whose cell longitude and latitude are empty in the work parameter table, select only the cell with the cell coverage type as outdoor, and check the user data and work parameter data at a grid-level multi-day time granularity Match by the cell identification code, and then perform deduplication processing on the matched data group, count the number of user data under each cell, only keep the cells with the number of user data greater than a predetermined value, and calculate users based on the cell latitude and longitude and the user data latitude and longitude The distance from the data to the cell, and the abnormal value detection is performed to delete the user data with a long distance, and then the angle of the user data to the cell is calculated from the north direction;
  • Step 3 The sampling point-based intensity prediction method, the sampling point density-based prediction method, the sampling point intensity and density-based prediction method, the sampling point intensity-based stratified statistical prediction method, and the sampling point intensity-based prediction method are adopted respectively.
  • Sector prediction method to calculate the predicted azimuth of each cell;
  • Step 4 Use the user data and the surveyed station parameter data as the training set, use the five prediction methods of step 3 to predict the training set, and then use the Monte Carlo method to train the weights of the five prediction methods according to the prediction results;
  • Step 5 According to the prediction effect obtained by the weights of step 4 on the training set, select the optimal weight and determine the prediction model configured with the optimal weight, the prediction model is used to predict the azimuth based on user data and output the azimuth The predicted value. The predicted azimuth value is used to correct the azimuth of the antenna.
  • the present invention uses five basic prediction algorithms and weighting algorithms to form a joint algorithm to predict the data, which can ensure the accuracy of the predicted data Sex.
  • the data set includes: time, international mobile user identification code, cell identification code used by the user, longitude of the user, latitude of the user, reference signal received power, cell identification code, cell longitude, cell latitude, azimuth, cell Name, cell coverage type.
  • the method for predicting the strength based on sampling points is specifically: finding the average value of the angles of the first n sampling points with the largest reference signal received power (Reference Signal Received Power, RSRP) value under each cell identification code as the azimuth prediction value, n being a natural number .
  • RSRP Reference Signal Received Power
  • the hierarchical statistical prediction algorithm based on the strength of the sampling points is: under each cell identification code, the distance is deduplicated and then n-1 different percentiles are calculated, arranged from small to large, according to user data to the cell
  • the prediction algorithm based on the combination of the strength and density of the sampling point is: under each cell identification code, the sampling point is divided into a sector every N degrees according to the angle to the cell, divided into 360/N shares, N ⁇ [1,360],360/ N takes an integer; count the total number of sampling points under each cell identification code, take out the sectors where the number of sampling points in the 360/N sector area is greater than d% of the total sampling points, d ⁇ [1,99], calculate these sectors
  • the sector-based prediction algorithm based on the strength of the sampling point is: under each cell identification code, the sampling point is divided into a sector every N degrees according to the angle of the cell into 360/N shares, and the RSRP value under each sector is obtained The largest RSRP average of the first n sampling points, and finally the sector angle with the largest RSRP average is the predicted azimuth angle.
  • the sector-based prediction algorithm based on the density of the sampling point is: under each cell identification code, the sampling point is divided into a sector every N degrees according to the angle of the cell into 360/N parts, and the midpoint of each sector is counted For the number, the sector angle with the largest number of points is used as the predicted azimuth angle.
  • the Monte Carlo method is: randomly extracting a part of data from the training set as training samples, randomly selecting P times to obtain P training samples; randomly generating Q weight combinations for each training sample, where Q is a natural number; For each combination of weights, the ratio of the number of deviations between the predicted azimuth of the training model and the azimuth of the survey station within R° is counted as the confidence of the combination of weights, and the value range of R is [0,360 ); For each training sample, determine the weight combination with the greatest confidence.
  • the present invention combines the five prediction algorithms together to avoid some of the shortcomings of the individual algorithm, and uses the method of giving multiple prediction model weights to organically combine multiple algorithms to give the algorithm with better prediction effect high weight.
  • the relatively poor algorithm is given a low weight, which ensures the accuracy of data prediction and the stability of data prediction for different regions.
  • the RSRP heat map is obtained by rasterizing the user data, and the predicted value of the azimuth angle predicted by the prediction model based on the user data is drawn on the heat map to facilitate maintenance personnel to search based on the heat map
  • the parameters such as wrong latitude and longitude, wrong azimuth, and reverse cell connection in the worker parameter.
  • 1 is a flow chart of antenna parameter cleaning in a method for realizing antenna azimuth correction based on user data in the present invention
  • FIG. 2 is a RSRP heat map of a 10-meter grid-level OTT sampling point in a cell of a certain place using the embodiment of FIG. 1;
  • FIG. 3 is a comparative analysis diagram of the results of the forecasting project participating in a survey station in a province using the embodiment of FIG. 1;
  • FIG. 4 is a comparative analysis diagram of the results of applying the embodiment of FIG. 1 in an important scene analysis/prediction project participation in a survey station in a residential area;
  • FIG. 5 is a practical effect diagram of finding errors in latitude and longitude by using the embodiment of FIG. 1;
  • FIG. 6 is a practical effect diagram of finding an azimuth angle error using the embodiment of FIG. 1;
  • FIG. 7 is a practical effect diagram of finding a reverse connection cell by using the embodiment of FIG. 1.
  • the data to be predicted is accurate and stable, but it is often not good in practical applications, because the general algorithm has certain limitations and applicability, which leads to poor prediction data.
  • the present invention combines several algorithms together to avoid the shortcomings of a single algorithm.
  • the method of assigning weights is used to organically combine multiple algorithms to give a good algorithm a higher weight and a relatively poor algorithm. Low weight, this can ensure the accuracy of the prediction effect.
  • the selected regions are different, the stability of the forecast can also be guaranteed.
  • the joint algorithm is applied in practice, and the expected effect is achieved, and excellent results are achieved in terms of both stability and accuracy.
  • the antenna parameter cleaning model is composed of prediction model algorithm, weight algorithm, and optimal weight algorithm.
  • the method for implementing antenna azimuth correction based on user data includes the following steps:
  • (1) Data collection collect 30-day user data and industrial parameter data.
  • the main indicator variables used are: time (rpt_time), international mobile user identification code (imsi), user-used cell identification code (s_ci), user's location Longitude, latitude of the user, reference signal received power (s_rsrp), cell identification code (ci), cell longitude (latitude), cell latitude (latitude), azimuth (azimuth), cell name (cell_name) ), cell coverage type (cover_type).
  • Algorithm one (based on the sampling point strength prediction method): Find the average of the angles of the first ten sampling points with the largest RSRP value under each ci as the predicted azimuth angle based on the algorithm one.
  • Delete the data of the first and fifth rings only retain the data of the middle 2, 3, and 4 rings, and then find the average of the angles of the first five sampling points with the largest RSRP value in each ring of the 2, 3, and 4 rings;
  • the mean value of the mean value of the sampling points corresponding to the 4th ring is used as the predicted azimuth value based on the second algorithm.
  • Algorithm three (prediction method based on the combination of sampling point strength and density): under each ci, the sampling points are divided into 72 parts according to the angle to the cell (that is, a sector of 5 degrees), and the total number of sampling points under each ci is counted. Take out the sectors of the 72 fan-shaped areas where the number of sampling points is greater than 4% of the total sampling points, calculate the RSRP average of the first five sample points with the largest RSRP value in these sectors, take out the 2 fan-shaped areas with the largest RSRP average, calculate these two The average value of each sector angle is used as the predicted value of the azimuth angle predicted based on the third algorithm.
  • Algorithm four prediction algorithm based on sector strength of sampling points: under each ci, the sampling points are divided into 72 parts according to the angle to the cell (that is, a sector of 5 degrees), and the RSRP value of each sector is the largest The average RSRP of the first 10 sampling points, and finally the sector angle with the largest RSRP average is taken as the predicted azimuth angle based on Algorithm 4.
  • Algorithm five prediction algorithm based on density of sampling points in sectors: under each ci, the sampling points are divided into 72 parts according to the angle to the cell (that is, a sector at 5 degrees), and the number of points in each sector is counted The sector angle with the largest number of points is used as the predicted value of the azimuth based on the algorithm five.
  • the method of using Monte Carlo training weights is to randomly extract half of the data from the training set for training.
  • a total of 50 random selections that is, 50 training samples.
  • Each sample randomly generates 10,000 weight combinations (in order to test whether the number of iterations has an impact on the prediction results, here randomly generate 1000, 3000, 5000, 7000, 9000, 10000, 11000, 13000, 20000 weight combinations, selected 10,000 weight combinations with the best results), the criterion for evaluating each combination is the ratio of the number of predicted azimuths and the azimuth of the survey station within 20 degrees, and this number ratio is used as the confidence of the weight combination, For each training sample, the final weight combination with the largest output ratio is the weight combination with the largest output confidence.
  • Z in formula (1) is the actual prediction error
  • angel predict is the predicted azimuth
  • angle true is the azimuth of the survey station
  • cost represents the error coefficient corresponding to the weight combination
  • M is the total survey in the training sample.
  • K is the set angle error threshold.
  • the confidence of the weight combination is 1-cost. For example, if the cost is 20%, the confidence of the weight combination is 1-20%, that is, 80%.
  • the experiment is divided into two parts.
  • the first part is to put the training data into five prediction model algorithms to train, predict and get error data, and then put the training data into the joint algorithm for training, prediction and get error data.
  • the second part is to compare the errors of the training and verification sets of the joint algorithm with the five prediction algorithms to evaluate the effect of the joint algorithm.
  • the first is data collection and processing.
  • a total of 30 days, OTT (Over The Top) data of 691 cells were collected.
  • the so-called OTT data refers to the data collected through the OTT business of the communication industry.
  • Each cell has thousands or even tens of thousands of data and surveys the 691 cells to obtain real information of the cell.
  • the data In order to ensure the integrity of the data, the data needs to be processed with outliers, the data lacking the cell identification code and the latitude and longitude information should be eliminated, and the data with abnormal distances and repeated data of the sampling points should be deleted.
  • the prediction effects of the joint algorithm and the five algorithms are compared, and the errors of the joint algorithm and the five algorithms on the training set and the verification set are calculated respectively.
  • the calculation error is less than 20 degrees (including 20 degrees).
  • FIG. 2 is an RSRP thermal map of a 10-meter grid-level OTT sampling point of a community in Quanzhou in an embodiment. It can be seen from the figure that the identification code, name, 30-day sampling point, azimuth and predicted azimuth angle of the cell, and the strength of the RSRP value of each sampling point.
  • Figure 3 is a comparative analysis of the results of a city's forecasting project participating in a survey station in the embodiment.
  • FIG. 5 is an actual effect diagram of the latitude and longitude error in the embodiment.
  • the surveyor surveyed the site’s working parameters on site and found the site at the original latitude and longitude position, which was consistent with the original system’s conclusion that the latitude and longitude error was determined.
  • the survey staff analyzed the data based on the OTT sampling point and found the site in another place, verifying the system's ability to identify and judge latitude and longitude.
  • 6 is an actual effect diagram of the azimuth error in the embodiment, the azimuth angle predicted by the OTT sampling point in this cell is 75 degrees, and the azimuth angle actually surveyed is 60 degrees, but the azimuth angle in the worker parameter table is 20 degrees. The error between the predicted azimuth angle and the actual survey azimuth angle is small.
  • FIG. 7 is an actual effect diagram of the reverse connection cell in the embodiment.
  • the fan-shaped area marked by the black bold line and the fan-shaped area marked by the white bold line shown in (a) and (b) in FIG. 7 represent two neighboring cells, and the periphery of the cell
  • Each small circle of black, gray, and white represents the sampling point; see (a) in FIG. 7. It can be seen that, using the above prediction model provided by the present invention, based on the data of the sampling point, the first cell (the sector marked by the black bold line) is predicted.
  • the predicted value of the azimuth of the area) is 300 degrees, while the azimuth of the first cell actually surveyed is 187 degrees; referring to (b) in FIG. 7, it can be seen that the second cell (white) is predicted based on the data of the sampling points
  • the fan-shaped area marked by the bold line) has a predicted azimuth of 120 degrees, and the actual azimuth of the second cell surveyed is 293 degrees. It can be seen that the prediction error of these two adjacent cells is very large, so it provides a reference for maintenance personnel. Data, maintenance personnel on-site inspection found that the antennas of these two cells are suspected to be reversed, and the platform software has the function of quickly identifying the reversed cells.
  • the steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly by hardware, a software module executed by a processor, or a combination of both.
  • the software module may be placed in random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, or any other form of storage medium known in the art.

Abstract

本发明公开一种基于用户数据实现天线方位角纠偏的方法，将五种预测算法联合在一起，规避了单独算法的一些缺点，利用赋予多种预测模型权重的方法，将多种算法有机地组合在一起，将预测效果较好的算法赋予高权重，而将相对不好的算法赋予低的权重，既保证了数据预测的准确性，也保证了对于不同地区的数据预测的稳定性。本发明通过对分析的数据画栅格级用户数据RSRP热力图，获得小区周围的采样点信息以及预测的方位角，通过更深层次的分析，可以找出工参中经纬度错误、方位角错误、接反小区等问题。

Description

基于用户数据实现天线方位角纠偏的方法
本申请要求于2018年11月27日提交中国专利局、申请号为201811420967.X、申请名称为“基于用户数据实现方位角纠偏的方法”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及网络技术领域，尤其涉及基于用户数据实现天线方位角纠偏的方法。
背景技术
在大数据时代，全面详实的移动通信数据获取需求取代了以往的路测等传统数据的随机粗放获取模式，因为获取的数据更加准确，分析的结果准确性有了保证。
移动通信中每天都会产生大量的数据，面对如此大量而复杂的数据，要发现数据之间的关系以预测其余相关的变量，需要一个稳定而准确的模型，这对算法提出了更高的要求。为了获得这样准确的模型，分析常规的建模过程是有必要的。首先要借助统计和可视化的方法分析数据的特征，然后再对数据进行建模分析。但是如果只考虑单个模型可能会模型选择错误，在实际运用当中，由于各个区域的不同，产生的数据也会有所差异，如果数据不够多或者数据没有累计到一定的程度，那么在选用模型时可能会出现错误。而且一旦选定了某种模型，很有可能没有机会再去选择其余的模型。当数据发生改变时，预测效果可能会变差。因此我们需要考虑全面的信息，建立一种行之有效的算法对数据进行预测。
在移动通信中，天馈系统运行质量常常缺乏有效的监控和发现问题的手段。因此工参准确性和合理性成为优化最突出的问题。产生天线工参错误的原因有：优化人员调站后未按照规定提交更新工参，或者任务紧张没有严格测量准确更新；例行拉网天馈工参勘查任务未严格执行，实际准确测量勘查的小区占比少；天面多运营天馈设备林立馈线复杂，优化人员找不到某些天线或者找错小区；广泛使用的测量设备为罗盘仪，而天面复杂电磁环境和一定距离比照天线造成 测量误差。
运营商长期以来依赖勘站的方式获取小区方位角等数据，但该方式不仅耗时、耗人和耗力而且运维的成本较高。
发明内容
为了解决上述问题，本发明基于移动通信网络数据的采集、存储及分析能力，运用机器学习算法，提供一种应用大数据处理技术，将多个数据源综合分析，形成对网络全方位评价，从而对天线工参进行清洗。可以快速的发现天线覆盖异常的问题，提高小区的天面质量，为小区工程建设、网络维护、无线优化提供有力支撑，实现提升网络质量，为网络智能化运维平台直接更新工参提供强有力的技术支撑。
基于用户数据实现天线方位角纠偏的方法，包括以下步骤：
步骤1、收集数据，收集用户数据和工参数据样本；
步骤2、数据处理，首先将工参表中小区经度和纬度为空的小区删除，只选取小区覆盖类型为室外的小区，在栅格级多天时间粒度上对所述用户数据和工参数据通过小区识别码进行匹配，然后对匹配后的数据组进行去重处理，统计每个小区下的用户数据数量，仅保留用户数据数量大于预定值的小区，根据小区经纬度和用户数据经纬度，计算用户数据到小区的距离，并且进行异常值检测删除距离较远的用户数据，再从正北方向计算用户数据到小区的角度；
步骤3、分别采用基于采样点强度预测方法、基于采样点密度分扇区的预测方法、基于采样点强度和密度结合的预测方法、基于采样点强度的分层统计预测方法和基于采样点强度分扇区的预测方法，计算每个小区的预测方位角；
步骤4、使用用户数据和实测的勘站工参数据作为训练集，用步骤3的五种预测方法对训练集进行预测，再根据预测结果使用蒙特卡罗方法训练出五种预测方法的权重；
步骤5、根据步骤4的权重在训练集上得到的预测效果，选出最优权重，确定配置有所述最优权重的预测模型，所述预测模型用于基于用户数据预测方位角输出方位角预测值，所述方位角预测值用于对天线方位角进行纠偏。
大数据量时，预测值与真实值常常会有较大的误差，为了避免误差过大， 本发明采用五种基本预测算法和权重算法结合形成联合算法对数据进行预测，可以保证预测数据的准确性。
所述数据组包括：时间、国际移动用户识别码、用户使用的小区识别码、用户所在的经度、用户所在的纬度、参考信号接收功率、小区识别码、小区经度、小区纬度、方位角、小区名称、小区覆盖类型。
所述基于采样点强度预测方法具体为：求每个小区识别码下参考信号接收功率(Reference Signal Received Power，RSRP)值最大的前n个采样点角度的均值作为方位角预测值，n为自然数。
所述基于采样点强度的分层统计预测算法为：每个小区识别码下，先将距离去重再计算n-1个不同的百分位数，按照从小到大排列，根据用户数据到小区的距离划分m环，m≥3,第一环：距离<＝第一个百分位数；第二环：第一个百分位数<距离<＝第二个百分位数；第三环：第二个百分位数<距离<＝第三个百分位数；第四环：第三个百分位数<距离<＝第四个百分位数；第n环：第n-1个百分位数<距离，删除最小环和最大环的数据，保留中间环数据；确定位于所述中间环中每环参考信号接收功率(RSRP)值最大的前n个采样点角度均值，确定所述前n个采样点角度均值的均值作为方位角预测值，所述n为自然数。
所述基于采样点强度和密度结合的预测算法为：每个小区识别码下，将采样点按照到小区的角度每N度划分一个扇区分成360/N份，N∈[1,360],360/N取整数；统计每个小区识别码下的采样点总数，取出360/N个扇形区域中采样点个数大于总采样点d％的扇区，d∈[1,99],计算这些扇区中RSRP值最大的前n个采样点RSRP均值，n为自然数，取出RSRP均值最大的扇形区域t个，t的范围为[1,360/N]之间的整数，计算这t个扇区角度的均值作为方位角预测值。
所述基于采样点强度分扇区的预测算法为：每个小区识别码下，将采样点按照到小区的角度每N度划分一个扇区分成360/N份，求每个扇区下RSRP值最大的前n个采样点的RSRP均值，最后取RSRP均值最大的扇区角度为方位角预测值。
所述基于采样点密度分扇区的预测算法为：每个小区识别码下，将采样点按照到小区的角度每N度划分一个扇区分成360/N份，统计每个扇区中点的 个数，取点的个数最多的扇区角度作为方位角预测值。
所述蒙特卡罗方法为：随机从所述训练集中抽取一部分数据作为训练样本，共随机选取P次得到P份训练样本；针对每份训练样本随机生成Q个权重组合，所述Q为自然数；针对每个权重组合，统计所述训练模型输出的预测方位角与勘站方位角的偏差在R°以内的个数占比作为该权重组合的置信度，所述R的取值范围为[0,360)；针对每份训练样本，确定置信度最大的权重组合。
本发明将五种预测算法联合在一起，规避了单独算法的一些缺点，利用赋予多种预测模型权重的方法，将多种算法有机地组合在一起，将预测效果较好的算法赋予高权重，而将相对不好的算法赋予低的权重，这样保证了数据预测的准确性，也保证了对于不同地区的数据预测的稳定性。再者，通过对用户数据进行栅格化处理得到RSRP热力图，并且在该热力图中绘制出通过预测模型基于该用户数据所预测出的方位角预测值，以方便维护人员基于该热力图查找工参中经纬度错误、方位角错误、接反小区等问题。
附图说明
图1为本发明基于用户数据实现天线方位角纠偏的方法进行天线工参清洗的流程图；
图2为应用图1实施例在某地某小区10米栅格级OTT采样点RSRP热力图；
图3为应用图1实施例在某省预测工参与勘站结果对比分析图；
图4为应用图1实施例在重要场景分析/居民区预测工参与勘站结果对比分析图；
图5为应用图1实施例发现经纬度错误实际效果图；
图6为应用图1实施例发现方位角错误实际效果图；
图7为应用图1实施例发现接反小区实际效果图。
具体实施方式
为了使本发明的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本发明进行进一步详细说明。应当理解，此处所描述的具体实施例仅 仅用以解释本发明，并不用于限定本发明。
在预测小区方位角时，需要预测的数据准确而稳定，但是在实际运用中往往不好，因为一般的算法有一定的局限性和适用性，导致预测的数据不好。但是本发明将几种算法联合在一起，可以避免单个算法的缺点，利用赋予权重的方法将多种算法有机的组合在一起，赋予好的算法较高的权重，而赋予相对较不好的算法低权重，这样可以保证预测效果的准确性。当所选的地区不同时，也可以保证预测的稳定性。本发明将联合算法运用于实际中，取得了预期的效果，在稳定性和准确性方面都取得了俱佳的效果。
如图1，天线工参清洗模型是由预测模型算法、权重算法、最优权重算法组成。
基于用户数据实现天线方位角纠偏的方法，包括以下步骤：
(1)数据收集：收集30天用户数据和工参数据,主要使用的指标变量有：时间(rpt_time)、国际移动用户识别码(imsi)、用户使用的小区识别码(s_ci)、用户所在的经度(longitude)、用户所在的纬度(latitude)、参考信号接收功率(s_rsrp)、小区识别码(ci)、小区经度(longitude)、小区纬度(latitude)、方位角(azimuth)、小区名称(cell_name)、小区覆盖类型(cover_type)。
(2)数据处理：首先将工参表中小区经度和纬度为空的小区删除并且只选取小区覆盖类型为室外的小区，在栅格级30天时间粒度上对上述步骤(1)中的各数据源通过小区识别码进行匹配，然后对整合后的数据进行去重处理，统计每个小区下的用户数据数量，仅保留用户数据数量大于300的小区，根据小区经纬度和用户数据经纬度，计算用户数据到小区的距离，并且进行异常值检测删除距离较远的用户数据。再从正北方向统计计算用户数据到小区的角度。
(3)算法分析：五种预测算法采用如下算法。
算法一(基于采样点强度预测方法)：求每个ci下RSRP值最大的前十个采样点角度的均值，作为基于该算法一预测出的方位角预测值。
算法二(基于采样点强度的分层统计预测算法)：每个ci下，先将距离去重再计算20％、40％、60％、80％的分位数，根据OTT采样点到小区的距离划分五环，第一环：距离<＝20％的分位数。第二环：20％<距离<＝40％的分位数。第三环：40％<距离<＝60％的分位数。第四环：60％<距离<＝80％的分位数。第 五环：80％<距离。删除第1、5环的数据，只保留中间的2、3、4环数据，再求2、3、4环中每环RSRP值最大的前五个采样点角度均值；最后求2、3、4环对应的所述采样点角度均值的均值，作为基于该算法二预测出的方位角预测值。
算法三(基于采样点强度和密度结合的预测方法)：每个ci下，将采样点按照到小区的角度分成72份(即5度一个扇区)，统计每个ci下的采样点总数，取出72个扇形区域中采样点个数大于4％总采样点的扇区，计算这些扇区中RSRP值最大的前五个采样点RSRP均值，取出RSRP均值最大的扇形区域2个，计算这两个扇区角度的均值作为基于该算法三预测出的方位角预测值。
算法四(基于采样点强度分扇区的预测算法)：每个ci下，将采样点按照到小区的角度分成72份(即5度一个扇区)，求每个扇区下RSRP值最大的前10个采样点的RSRP均值，最后取RSRP均值最大的扇区角度作为基于算法四预测出的方位角预测值。
算法五(基于采样点密度分扇区的预测算法)：每个ci下，将采样点按照到小区的角度分成72份(即5度一个扇区)，统计每个扇区中点的个数，取点的个数最多的扇区角度作为基于该算法五预测出的方位角预测值。
权重算法的具体实现步骤如下。
(1)使用采样点数据和真实的勘站工参数据作为训练集，用上面的五种预测模型算法对训练集进行预测，再根据五种算法的预测结果使用蒙特卡罗方法训练出五种算法的权重。
(2)使用蒙特卡罗训练权重的方法是随机从训练集中抽取一半的数据进行训练，共随机选取50次，即有50份训练样本。每份样本都随机生成一万个权重组合(为了检验迭代次数是否对预测结果有影响，此处随机生成1000、3000、5000、7000、9000、10000、11000、13000、20000个权重组合，选取了结果最好的一万个权重组合)，评估每个组合的标准是预测方位角与勘站方位角的偏差在20度以内的个数占比，该个数占比作为权重组合的置信度，对于每份训练样本，最终输出占比最大的权重组合，即输出置信度最大的权重组合。
在具体实现时，按照如下公式(1)和(2)，针对每份训练样本确定置信度最大的权重组合。
Z＝∠(angel predict,angel true)  公式(1)
Figure PCTCN2019075043-appb-000001
其中，公式(1)中的Z为实际预测误差，angel predict为预测方位角，angle true为勘站方位角；公式(2)中cost表示权重组合对应的误差系数，M为训练样本中总勘站小区数量，K为设定的角度误差阈值。
基于上述公式可知，权重组合的置信度为1-cost,例如，cost为20％，则权重组合的置信度就为1-20％，即为80％。
最优权重算法甄选具体实现步骤如下：
(1)上述得到50份最佳的权重组合，取最终50份权重的均值或者中位数作为最终的权重，分别运用50份权重的均值和中位数对验证集进行预测，选取预测结果最好的权重组合，各方法的权重组合如下：W_方法1＝0.02383308，W_方法2＝0.48847700，W_方法3＝0.08321591，W_方法4＝0.08895580，W_方法5＝0.27992355。
(2)最终通过加权平均计算上面五种算法的预测结果作为最终的方位角预测值。
为了评估联合算法的效果，选取了福建省691个小区的OTT数据和勘站数据进行实验，以得到基于五种预测算法预测结果的准确性、稳定性。
实验步骤如下：
首先进行数据收集，数据的处理，分别运用五种预测算法和联合算法进行数据预测，得到数据预测结果，然后整理五种预测算法和联合算法的结果以及真实勘站数据，比较联合算法和五种预测算法的准确性，以此综合评价联合算法模型的效果。
实验分为两部分，第一部分是将训练数据放到五种预测模型算法中训练，预测，得到误差数据，接着将训练数据放到联合算法中进行训练，预测，得到误差数据。
第二部分是将联合算法与五种预测算法的训练集、验证集的误差进行对比，以评估联合算法的效果。
实验数据
首先是数据的收集和处理，共收集了30天，691个小区的OTT(Over The Top) 数据，所谓OTT数据是指通过通信行业的OTT业务所收集的数据。每个小区有上千甚至上万条数据并且对这691个小区进行勘站获取小区真实信息数据。
为了确保数据的完整性，需要对数据进行异常值处理，对于缺少小区识别码和经纬度信息的数据予以剔除，删除采样点中距离异常的数据和重复的采样点数据。
实验方法
首先将训练数据放到模型中训练，预测，将每个算法得到的预测数据和误差数据保存，接着将训练数据放到联合算法中训练，预测，将得到的预测数据和误差数据保存，最后是比对联合算法和五种算法的预测效果，分别计算联合算法和五种算法在训练集、验证集上的误差，计算误差小于20度(包含20度)的个数占比。
实验结果
在运用联合算法和五种算法对比后，得到联合算法的预测准确性更高，使用联合算法进行预测，得到如下所示的一些相关图。图2是实施例中泉州市某个小区10米栅格级OTT采样点RSRP热力图。从图中可以看出该小区的识别码、名称、30天的采样点以及工参方位角和预测的方位角，以及每个采样点的RSRP值的强弱。图3是实施例中某市预测工参与勘站结果对比分析图，从图中可以看出预测偏差小于45度的小区占比达到96.88％，占比都较大，说明预测模型有较好的准确性。图4是居民小区中预测工参与勘站结果对比分析图，其准确率达到90％以上。分别计算了预测误差小于等于20度、35度、45度的小区个数占比，图中的占比都较大，说明预测模型有较好的准确性。在道路、城中村等其他场景中也可达到相同的效果。图5是实施例中经纬度错误实际效果图，勘测人员现场勘察站点工参，在原工参经纬度位置上未找到该站点，与原系统判断经纬度错误的结论一致。勘察人员根据OTT采样点对数据进行分析，在另一地方找到该站点，验证了系统对于经纬度识别判断功能。图6是实施例中方位角错误实际效果图，该小区中根据OTT采样点预测的方位角为75度，而实际勘测的方位角为60度，但是工参表中的方位角为20度，预测的方位角与实际勘测的方位角误差较小，经验证发现，该小区方位角疑为人为录入错误，平台软件具备快速甄别小区方位角异常功能。图7是实施例中接反小区实际效果图,图7中 (a)和(b)示出的黑色加粗线条标注的扇形区域和白色加粗线条标注的扇形区域表示两个相邻小区，小区周边的黑灰白色的各小圆圈表示采样点；参见图7中的(a)图可知，利用本发明提供的上述预测模型，基于采样点的数据，预测出第一个小区(黑色加粗线条标注的扇形区域)的方位角预测值为300度，而实际勘测第一个小区的方位角为187度；参见图7中的(b)图可知，基于采样点的数据，预测出第二个小区(白色加粗线条标注的扇形区域)的预测方位角为120度，而实际勘测第二个小区的方位角为293度，可见，这两个相邻小区的预测误差很大，如此，给维护人员提供参考数据，维护人员现场核查发现这两个小区的天线疑为人为接反，平台软件具备快速甄别接反小区的功能。
本发明方案所公开的技术手段不仅限于上述实施方式所公开的技术手段，还包括由以上技术特征任意组合所组成的技术方案。
最后，还需要说明的是，在本文中，诸如第一和第二等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来，而不一定要求或者暗示这些实体或操作之间存在任何这种实际的关系或者顺序。而且，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者智能设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者智能设备所固有的要素。在没有更多限制的情况下，由语句“包括一个......”限定的要素，并不排除在包括所述要素的过程、方法、物品或者智能设备中还存在另外的相同要素。
本说明书中各个实施例采用递进的方式描述，每个实施例重点说明的都是与其他实施例的不同之处，各个实施例之间相同相似部分互相参见即可。对于实施例公开的装置、系统、智能设备和存储介质而言，由于其与实施例公开的方法相对应，所以描述的比较简单，相关之处参见方法部分说明即可。
专业人员还可以进一步意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、计算机软件或者二者的结合来实现，为了清楚地说明硬件和软件的可互换性，在上述说明中已经按照功能一般性地描述了各示例的组成及步骤。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来 使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
结合本文中所公开的实施例描述的方法或算法的步骤可以直接用硬件、处理器执行的软件模块，或者二者的结合来实施。软件模块可以置于随机存储器(RAM)、内存、只读存储器(ROM)、电可编程ROM、电可擦除可编程ROM、寄存器或技术领域内所公知的任意其它形式的存储介质中。
对所公开的实施例的上述说明，使本领域专业技术人员能够实现或使用本申请。对这些实施例的多种修改对本领域的专业技术人员来说将是显而易见的，本文中所定义的一般原理可以在不脱离本申请的精神或范围的情况下，在其它实施例中实现。因此，本申请将不会被限制于本文所示的这些实施例，而是要符合与本文所公开的原理和新颖特点相一致的最宽的范围。

Claims (8)

  1. 基于用户数据实现天线方位角纠偏的方法，其特征在于，包括以下步骤：
    步骤1、收集用户数据和工参表，所述工参表中包括小区工参数据；步骤2、删除所述工参表中小区经纬度为空的小区参数，选取所述工参表中小区覆盖类型为室外的小区工参数据，对所述用户数据进行栅格化处理筛选用户数据，通过小区识别码对所选取的小区工参数据和所筛选的用户数据进行匹配得到匹配成功的数据组，所述数据组包括小区工参数据及与其匹配的用户数据，针对所述数据组，对用户数据进行去重处理，并针对每个小区统计归属于该小区的用户数据数量，选择用户数据数量大于第一预定值的小区作为目标小区，根据目标小区的小区经纬度和与其匹配的用户数据中的用户所在的经纬度，计算用户到小区的距离，并选择距离小于第二预定值的用户数据作为目标用户数据，根据所述目标用户数据，基于正北方向计算用户到小区的角度；
    步骤3、根据所述距离和所述角度，分别采用基于采样点强度的预测算法、基于采样点密度分扇区的预测算法、基于采样点强度和密度结合的预测算法、基于采样点强度的分层统计预测算法和基于采样点强度分扇区的预测算法，计算每个小区的预测方位角；
    步骤4、采用所述目标用户数据和勘站方位角作为训练集，采用蒙特卡罗算法对包含所述步骤3中的五种预测算法的预测模型进行训练，以训练所述五种预测算法的权重；
    步骤5、根据步骤4的权重在训练集上得到的预测效果，选出最优权重,确定配置有所述最优权重的预测模型，所述预测模型用于基于用户数据预测方位角输出方位角预测值，所述方位角预测值用于对天线方位角进行纠偏。
  2. 根据权利要求1所述的基于用户数据实现天线方位角纠偏的方法，其特征在于，所述数据组包括：时间、国际移动用户识别码、用户使用的小区识别码、用户所在的经度、用户所在的纬度、参考信号接收功率、小区识别码、小区经度、小区纬度、方位角、小区名称、小区覆盖类型。
  3. 根据权利要求1所述的基于用户数据实现天线方位角纠偏的方法，其特征在于，所述基于采样点强度的预测算法用于：确定每个小区识别码对应的 参考信号接收功率(RSRP)值最大的前n个采样点角度的均值作为方位角预测值,所述n为自然数。
  4. 根据权利要求1所述的基于用户数据实现天线方位角纠偏的方法，其特征在于，所述基于采样点强度的分层统计预测算法用于：针对每个小区识别码对应的采样点，先将距离去重再计算n-1个不同的百分位数，按照从小到大排列，根据用户到小区的距离划分m环，m≥3,其中，第一环：距离<＝第一个百分位数；第二环：第一个百分位数<距离<＝第二个百分位数；第m环：第m-1个百分位数<距离，删除位于最小环和最大环的数据，保留位于中间环的数据，确定位于所述中间环中每环参考信号接收功率(RSRP)值最大的前n个采样点角度均值，确定所述前n个采样点角度均值的均值作为方位角预测值，所述n为自然数。
  5. 根据权利要求1所述的基于用户数据实现天线方位角纠偏的方法，其特征在于，所述基于采样点强度和密度结合的预测算法用于：针对每个小区识别码对应的采样点，将采样点按照到小区的角度每N度划分一个扇区分成360/N个扇区，N∈[1,360],360/N取整数；统计每个小区识别码下的采样点总数，选择所述360/N个扇区中采样点个数大于总采样点d％的扇区，所述d∈[1,99],针对所选的扇区，计算该扇区中参考信号接收功率(RSRP)降序排序靠前的前n采样点的RSRP均值，所述n为自然数，选取所述RSRP均值降序排序靠前的前t个扇区，所述t取值范围为[1,360/N]之间的整数，计算所述t个扇区角度的均值作为方位角预测值。
  6. 根据权利要求1所述的基于用户数据实现天线方位角纠偏的方法，其特征在于，所述基于采样点强度分扇区的预测算法用于：针对每个小区识别码对应的采样点，将采样点按照到小区的角度每N度划分一个扇区分成360/N个扇区，确定每个扇区下参考信号接收功率(RSRP)值降序排序靠前的前n个采样点的RSRP均值，选择所述RSRP均值最大的扇区角度作为方位角预测值。
  7. 根据权利要求1所述的基于用户数据实现天线方位角纠偏的方法，其特征在于，所述基于采样点密度分扇区的预测算法用于：针对每个小区识别码对应的采样点，将采样点按照到小区的角度每N度划分一个扇区分成360/N个扇区，统计每个扇区中点的采样点个数，选择采样点的个数最多的扇区角度 作为方位角预测值。
  8. 根据权利要求1所述的基于用户数据实现天线方位角纠偏的方法，其特征在于，所述采用蒙特卡罗算法具体用于：
    随机从所述训练集中抽取一部分数据作为训练样本，共随机选取P次得到P份训练样本；
    针对每份训练样本随机生成Q个权重组合，所述Q为自然数；
    针对每个权重组合，统计所述训练模型输出的预测方位角与勘站方位角的偏差在R°以内的个数占比作为该权重组合的置信度，所述R的取值范围为[0,360)；
    针对每份训练样本，确定置信度最大的权重组合。
PCT/CN2019/075043 2018-11-27 2019-02-14 基于用户数据实现天线方位角纠偏的方法 WO2020107712A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US16/960,686 US11206555B2 (en) 2018-11-27 2019-02-14 Method for implementing antenna azimuth correction based on user data

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811420967.XA CN109151866B (zh) 2018-11-27 2018-11-27 基于用户数据实现天线方位角纠偏的方法
CN201811420967.X 2018-11-27

Publications (1)

Publication Number Publication Date
WO2020107712A1 true WO2020107712A1 (zh) 2020-06-04

Family

ID=64806232

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/075043 WO2020107712A1 (zh) 2018-11-27 2019-02-14 基于用户数据实现天线方位角纠偏的方法

Country Status (3)

Country Link
US (1) US11206555B2 (zh)
CN (1) CN109151866B (zh)
WO (1) WO2020107712A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2022021602A1 (zh) * 2020-07-27 2022-02-03 南京邮电大学 一种基于地理信息的5g参考信号接收功率预测方法

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109151866B (zh) 2018-11-27 2019-03-05 南京华苏科技有限公司 基于用户数据实现天线方位角纠偏的方法
WO2021005454A1 (en) * 2019-07-05 2021-01-14 Telefonaktiebolaget Lm Ericsson (Publ) A method of measuring aas emf
CN110536310B (zh) * 2019-08-29 2022-03-15 南京华苏科技有限公司 基于用户数据识别天线接反的方法
CN110839251B (zh) * 2019-11-27 2022-09-13 南京华苏科技有限公司 基于用户数据识别天线前后抑制比异常的方法
CN111741493B (zh) * 2020-08-19 2020-11-24 南京华苏科技有限公司 一种基于aoa与mdt的方位角的纠偏方法及装置
CN112469060B (zh) * 2020-12-08 2023-05-16 中国联合网络通信集团有限公司 一种天线参数确定方法及装置
CN112702194B (zh) * 2020-12-16 2023-04-07 中国联合网络通信集团有限公司 室分小区故障定位方法、装置及电子设备
WO2024032872A1 (en) * 2022-08-08 2024-02-15 Telefonaktiebolaget Lm Ericsson (Publ) Methods and nodes for predicting azimuth values of cells in communications networks

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106844757A (zh) * 2017-02-17 2017-06-13 山东浪潮商用系统有限公司 一种天线方位角异常数据挖掘方法
CN106921989A (zh) * 2015-12-25 2017-07-04 中国移动通信集团北京有限公司 一种通信网络场强分布确定方法及装置
CN108124268A (zh) * 2016-11-30 2018-06-05 中国移动通信有限公司研究院 一种参数准确性识别方法及网络设备
CN108306699A (zh) * 2018-02-08 2018-07-20 南京华苏科技有限公司 一种基于增益预估的天馈优化方法
CN108375363A (zh) * 2017-12-05 2018-08-07 中国移动通信集团福建有限公司 天线方位角偏转核查方法、装置、设备及介质
CN109151866A (zh) * 2018-11-27 2019-01-04 南京华苏科技有限公司 基于用户数据实现天线方位角纠偏的方法

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103188693B (zh) * 2011-12-30 2016-08-10 中国移动通信集团江苏有限公司 基于地理信息系统的天线下倾角确定方法及装置
EP2901737B1 (en) * 2012-09-27 2019-02-20 Telefonaktiebolaget LM Ericsson (publ) Method and communication node for determining positioning measurement uncertainty and position determination.

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106921989A (zh) * 2015-12-25 2017-07-04 中国移动通信集团北京有限公司 一种通信网络场强分布确定方法及装置
CN108124268A (zh) * 2016-11-30 2018-06-05 中国移动通信有限公司研究院 一种参数准确性识别方法及网络设备
CN106844757A (zh) * 2017-02-17 2017-06-13 山东浪潮商用系统有限公司 一种天线方位角异常数据挖掘方法
CN108375363A (zh) * 2017-12-05 2018-08-07 中国移动通信集团福建有限公司 天线方位角偏转核查方法、装置、设备及介质
CN108306699A (zh) * 2018-02-08 2018-07-20 南京华苏科技有限公司 一种基于增益预估的天馈优化方法
CN109151866A (zh) * 2018-11-27 2019-01-04 南京华苏科技有限公司 基于用户数据实现天线方位角纠偏的方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
ANONYMOUS: "3rd Generation Partnership Project; Technical Specification Group Radio Access Network; Evolved Universal Terrestrial Radio Access (E-UTRA); Further advancements for E-UTRA physical layer aspects (Release 9)", 3GPP STANDARD ; TECHNICAL REPORT ; 3GPP TR 36.814, no. V9.2.0, 25 March 2017 (2017-03-25), pages 1 - 105, XP051297630 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2022021602A1 (zh) * 2020-07-27 2022-02-03 南京邮电大学 一种基于地理信息的5g参考信号接收功率预测方法

Also Published As

Publication number Publication date
CN109151866B (zh) 2019-03-05
CN109151866A (zh) 2019-01-04
US11206555B2 (en) 2021-12-21
US20200351678A1 (en) 2020-11-05

Similar Documents

Publication Publication Date Title
WO2020107712A1 (zh) 基于用户数据实现天线方位角纠偏的方法
JP6770125B2 (ja) 建築物改装に伴う損害防止の推定
KR101982159B1 (ko) 카테고리별로 구분된 유동인구 정보를 이용한 유동인구 측정 방법
Long et al. Mapping block-level urban areas for all Chinese cities
Bégin et al. Assessing Volunteered Geographic Information (vgi) Quality Based on CONTRIBUTORS'Mapping Behaviours
CN110536310B (zh) 基于用户数据识别天线接反的方法
CN109949063B (zh) 一种地址确定方法、装置、电子设备及可读存储介质
CN114997534B (zh) 基于视觉特征的相似降雨预报方法和设备
CN116595121B (zh) 一种基于遥感技术数据显示监测系统
Chatzipoulka et al. Urban geometry, SVF and insolation of open spaces: London and Paris
CN116437291A (zh) 一种基于手机信令的文化圈规划方法和系统
CN111475746A (zh) 兴趣点位置挖掘方法、装置、计算机设备和存储介质
CN114662774A (zh) 一种城市街区活力预测方法、存储介质和终端
Rodrigues et al. Extracting 3D maps from crowdsourced GNSS skyview data
Wang et al. Evaluation of information transfer and data transfer models of rain-gauge network design based on information entropy
CN110321528B (zh) 一种基于半监督地理空间回归分析的高光谱影像土壤重金属浓度评估方法
Zhao et al. Urban spatial structure analysis: quantitative identification of urban social functions using building footprints
CN115879594A (zh) 一种基于地理探测器的城市定居人口分布趋势预测方法
CN114710742A (zh) 一种基于多链插值构建指纹地图的室内定位方法
CN110347938B (zh) 地理信息处理方法、装置、电子设备及介质
CN109874170B (zh) 地理坐标系盲检测方法、装置、设备及介质
Jardón et al. Spatial Markov chains implemented in GIS
Primawan Optimizing WLAN access point placement using geospatial technique
Le et al. Research and application of remote sensing and GIS technologies in determining and forecasting land use changes by Markov chain in y Yen District-Nam Dinh Province
CN117715086A (zh) 一种基于mdt数据的网络覆盖分析方法、装置、设备及介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19890086

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19890086

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 19890086

Country of ref document: EP

Kind code of ref document: A1