WO2020125349A1 - 一种场强测试方法 - Google Patents

一种场强测试方法 Download PDF

Info

Publication number
WO2020125349A1
WO2020125349A1 PCT/CN2019/121051 CN2019121051W WO2020125349A1 WO 2020125349 A1 WO2020125349 A1 WO 2020125349A1 CN 2019121051 W CN2019121051 W CN 2019121051W WO 2020125349 A1 WO2020125349 A1 WO 2020125349A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
field strength
target area
strength prediction
data combination
Prior art date
Application number
PCT/CN2019/121051
Other languages
English (en)
French (fr)
Inventor
武继龙
Original Assignee
中兴通讯股份有限公司
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 中兴通讯股份有限公司 filed Critical 中兴通讯股份有限公司
Priority to EP19897593.0A priority Critical patent/EP3902313A4/en
Publication of WO2020125349A1 publication Critical patent/WO2020125349A1/zh

Links

Images

Classifications

    • 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
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R29/00—Arrangements for measuring or indicating electric quantities not covered by groups G01R19/00 - G01R27/00
    • G01R29/12—Measuring electrostatic fields or voltage-potential
    • 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
    • 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
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/04—Arrangements for maintaining 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

Definitions

  • the present invention relates to the field of communication technology, and in particular, to a field strength prediction method for a signal coverage area.
  • environmental factors are an important factor in determining the accuracy of the model.
  • the first method can make full use of environmental information, and can achieve high accuracy in the modeling of field strength prediction.
  • the measurement of environmental information requires a lot of manpower and material resources, and a very high time cost.
  • the second method can meet the needs of rapid modeling, since all known data sets are used as training samples for modeling, the resulting model has low accuracy and large errors, and it is not available in solving field strength coverage problems. high.
  • the present invention provides a method for predicting field strength, comprising: selecting a first data combination from a data set in a non-target area, the signal corresponding to the first data combination and the second data combination in the target area The consistency degree of the fluctuation rule of the corresponding signal conforms to a preset rule; the field strength prediction model of the target area is trained, and the training samples of the field strength prediction model include the first data set.
  • an embodiment of the present invention also provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores An instruction executed by a processor, the instruction is executed by the at least one processor, so that the at least one processor performs the method described in the above aspects.
  • an embodiment of the present invention also provides a non-transitory computer-readable storage medium, the non-transitory computer-readable storage medium storing computer-executable instructions, the computer-executable instructions are used to The method described in the above aspects.
  • an embodiment of the present invention also provides a computer program product, the computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program includes program instructions, when the When the program instructions are executed by a computer, the computer is caused to perform the methods described in the above aspects.
  • FIG. 1 is a field strength prediction process of Embodiment 1 of the present invention.
  • Embodiment 1 of the present invention is a specific operation flow of Embodiment 1 of the present invention.
  • FIG. 3 is a comparison diagram of the MAE of the field strength prediction method of Embodiment 2 of the present invention and the prior art field strength prediction MAE;
  • FIG. 4 is a comparison diagram of the MAE of the field strength prediction method of Embodiment 3 of the present invention and the prior art field strength prediction MAE;
  • FIG. 5 is a schematic diagram of a hardware structure of an electronic device that performs a field strength prediction method according to an embodiment of the present invention.
  • FIG. 1 is a field strength prediction process in Embodiment 1 of the present invention.
  • the field strength prediction method includes: S13: selecting a first data combination from a data set in a non-target area, and the degree of consistency of the fluctuation rule of the signal corresponding to the signal corresponding to the second data combination of the target area in the signal corresponding to the first data combination Meet the preset rules; S14: Train the field strength prediction model of the target area, and the training samples of the field strength prediction model include the first data set.
  • the field strength prediction method provided by the embodiment of the present invention as compared with the prior art where all data sets of the non-target area are used for target area modeling, the first data combination selected from all data sets of the non-target area in the present invention is used as Training samples of the field strength prediction module, and the degree of consistency of the fluctuation law of the signal corresponding to the first data combination and the signal corresponding to the second set of data conforms to the preset rules, that is, the first data combination and the second selected by the present invention
  • the data combination has a certain degree of correlation, that is, the data in the non-target area data set that differs greatly from the fluctuation of the second data combination is removed. Therefore, the first data combination is used as the training sample for the field strength prediction model , Can make the field strength prediction model have higher accuracy, and the prediction method does not need to carry out on-site measurement of various environmental factors.
  • the non-target area may be divided into multiple sub-areas, and the target area may be further divided into multiple sub-areas, but usually the target area may be one sub-area.
  • the target area may be a sub-area in the non-target area. For example, if the non-target area is N cells, and each cell is a sub-area, then the k-th cell of the N cells can be used as the target area, N and k are both positive integers, and the value of k is greater than 0 and less than or equal to N .
  • the elements in the first data combination and the second data combination may include antenna parameters of the base station, which are simply referred to as parameters.
  • the antenna parameters of the base station can be adjusted to increase the signal coverage strength of the base station for the weak coverage area of the cell.
  • Antenna parameters usually have multiple dimensions: transmit power, antenna direction angle, antenna downtilt angle, antenna height and so on. Due to the many dimensions of the parameter, the adjustment direction and the size range are large. In reality, the maximum transmission power is used, and adjusting the antenna height will cause the signal coverage strength to change sharply. Therefore, the signal coverage strength is usually adjusted by adjusting the antenna direction angle.
  • the elements of the first data combination may also include other signal data, such as the distance, relative position, signal strength, etc. of the sub-area and the base station in the non-target area.
  • the elements in the second data combination may also include the distance, relative position, etc. of the non-target area and the base station.
  • S11 and S12 are also included before S13.
  • S11 Divide the data set into multiple sets of work parameter data according to the change of the work parameter
  • S12 Divide the second data combination of the target area into multiple sets of second data according to the change of the work parameter.
  • the parameters in S11 and S12 may be the antenna direction angle. That is, the data set in S11 is divided into multiple sets of industrial parameter data according to different angles with respect to the antenna direction angle of the base station. The second data combination in S12 is also divided into multiple sets of second data according to different angles with respect to the antenna direction angle of the base station.
  • the working parameters in S11 and S12 may also include, but are not limited to, transmission power, antenna downtilt angle, antenna height, and so on.
  • the working parameter in S11 and S12 may be a single working parameter, and the working parameter data is a single working parameter data.
  • the single working parameter data refers to only one parameter or dimension of the working parameter, for example, only the working direction of the antenna Or only the antenna height.
  • the working parameters in S11 and S12 may also include at least two parameters, for example, including two parameters, the antenna direction angle and the antenna height, and the data with the same antenna direction angle and antenna height in the data set may be divided into a group of working parameters In the second data combination, the data in which the antenna direction angle and the antenna height are the same are divided into a group of second data.
  • the elements of the first data combination may also include the distance between the sub-area and the base station in the non-target area.
  • the elements of the data set of the non-target area may also include other signal data, for example, all sub-areas and The distance of the base station.
  • S11 may further include: the signal data corresponding to each set of work parameter data is sorted according to distance to form ordered signal data, and the work parameter data is a single work parameter data.
  • multiple sets of work parameter data obtained from S11 select target work parameter data, and all target work parameter data constitute a first data combination, and the correlation degree of the target work parameter data with respect to multiple sets of second data complies with preset rules In other words, the degree of consistency of the fluctuation rule of the signal corresponding to the target industrial parameter data with respect to the signal of at least one set of second data conforms to the preset rule.
  • the degree of consistency of the fluctuation law of the signal corresponding to the first data combination relative to the signal corresponding to the second data combination may be calculated based on the correlation coefficient, such as the Pearson correlation coefficient. Specifically, it may be based on the Pearson correlation coefficient, Calculate the degree of correlation between the target worker parameter data and the second data.
  • S13 specifically includes: measuring the degree of consistency of the fluctuation law of the signal corresponding to all the data combinations in the data set of the non-target area relative to the signal corresponding to the second data combination, and sorting according to the degree of consistency.
  • the data combination corresponding to the top several of the consistency degree is used as the first data combination or the data combination corresponding to the consistency degree within a preset range is used as the first data combination.
  • the data combination corresponding to the top 10 or 20 of the consistency degree is used as the first data combination, that is, a data combination with a higher degree of consistency with the second data combination is mined from all data sets as The first data combination excludes the data combination with a lower degree of agreement; for example, the data combination corresponding to the degree of agreement between 0.8 and 1 is used as the first data combination.
  • the value of the degree of agreement does not exceed 1, when the degree of agreement is 1.
  • the time means that the fluctuation law of the two sets of data is exactly the same.
  • the training samples of the field strength prediction model may further include second data, that is, the training samples of the field strength prediction model include the first data combination and the second data combination.
  • the target area may also include a third data combination.
  • the third data combination may be used as a test sample of the field strength prediction model to test the effect of the field strength prediction model for field strength prediction.
  • all data combinations in the target area can be divided into multiple sets of data combinations according to the changes in the value of the work parameter, and 1/3 of the multiple sets of data combinations are the second data combination (the change of the work parameter value is known), in addition
  • the 2/3 group is the third data combination (it can be considered that the change of the working parameter value is unknown), the data amount of the third data data combination is greater than the data amount of the second data combination, so that the working parameter value of the test sample changes more than the known
  • the change of the working parameter value can more effectively verify the prediction effect of the field strength prediction model.
  • the working parameter values in all data combinations in the target area include 6 types (taking the antenna direction angle as an example, 10 degrees, 20 degrees, 30 degrees, 40 degrees, 50 degrees, and 60 degrees); then the antenna direction angle can be selected as The data combination corresponding to 10 degrees and 20 degrees is used as the second data combination, and the remaining data combination is the third data combination.
  • the data volume of the third data combination and the data volume of the second data combination can also be other ratios, for example, 1/4 of the multiple data combinations are the second data combination (the parameter value change is known), in addition Group 3/4 is the third data combination (it can be considered that the change of the working parameter value is unknown).
  • the data amount of the third data combination may be less than or equal to the data amount of the second data combination.
  • the training samples in the field strength prediction model include the first data combination and the second data combination
  • the test samples include the third data combination.
  • the ratio of the data volume of the training samples to the data volume of the test samples in the field strength prediction model is between 5 and 15, such as 8, 10, or 12, etc.
  • the ratio of the sum of the data amount of the first data combination and the second data combination to the data amount of the third data combination is between 5 and 15.
  • the field strength prediction model of the embodiment of the present invention may be a model in the field of deep learning, such as a fully connected neural network model, or a model in the field of machine learning, such as the Xgboost model.
  • S14 it may further include S16: performing field strength prediction on the target area based on the field strength prediction model.
  • the relevant data of the target area is input into the field strength prediction model, and the predicted field strength value of the target area is obtained after calculation of the field strength prediction model.
  • S15 may also be included: verify the validity of the field strength prediction model, and when the field strength prediction model is valid, enter S16, otherwise suspend or terminate.
  • MAE Mean Absolute Error
  • A Mean Absolute Error
  • FIG. 2 is a specific operation flow of the field strength prediction method in Embodiment 1 of the present invention.
  • Each group of data corresponds to the fluctuation law, that is, the fluctuation law of each group of data is obtained; from the processed data, the data corresponding to the target area is selected as the standard training data, and from all the processed data, the fluctuation law that is more consistent with the standard training data is selected.
  • Data combination, the standard training data and the data combination are fused into training samples, so that the model training samples have a high correlation, which is beneficial to training a model with higher accuracy; select the model, use the training samples for training; select evaluation Methods to evaluate the test results of the model.
  • Embodiment 2 and Embodiment 3 are specific descriptions of the solution provided in Embodiment 1, and should not be considered as limiting the application scenarios of the solution.
  • the embodiment of the present invention uses a full-link neural network as a basic model for field strength prediction in 67 cell data that does not contain any terrain information and 67 cell data in a full data set.
  • the verification scheme method improves the performance of the model.
  • S21 Divide the data of each cell in 67 cells into multiple groups of single parameter data according to the change of the parameter value contained in the cell data.
  • the working parameters may include transmission power, antenna direction angle, antenna downtilt angle, antenna height, etc.
  • the change of the working parameter data may specifically be the change of the antenna direction angle angle.
  • S22 Sort the signal data in each set of single parameter data in S21 according to the order from the smallest distance to the base station to form an ordered set of signal data. Assuming that M sets of single parameter data are obtained in S21, M sets of ordered signal data are formed.
  • the target cell may include 1/3X of the change in the work parameter value as the known cell.
  • Parameter value changes the corresponding data is used as standard training data (equivalent to the second data combination in Example 1), other 2/3X parameters change as unknown parameters, and the corresponding data as test data (Equivalent to the third data combination in Embodiment 1).
  • the selection of 1/3X as the change of the known working parameter value is to make the working parameter value of the test data change more than that of the known data, which can better verify the effectiveness of this embodiment.
  • S24 For the standard training data in S23, according to each set of single parameter data included, calculate each set of ordered signal data formed by S22 separately to obtain the Pearson coefficient between the ordered signal data of single parameter data And generate a Pearson coefficient value table. Specifically, for example, if M is equal to 100 and X is equal to 60, then each group of ordered data signals in the 100 groups of ordered data signals will be successively subjected to the Pearson coefficient with respect to the 20 groups of standard training data. After 2000 calculations Get the Pearson coefficient value table.
  • S25 According to the Pearson value table generated in S24, select the standard training data, each group of single parameter data, the top ten non-target cell single parameter data with the most consistent fluctuation pattern.
  • the first 10 non-target cell single parameter data with the most consistent fluctuation law are fused with the standard training data and used as the model training set to train a fully linked neural network model.
  • the most consistent size of the fluctuation law is to make the ratio of the data volume of the training set to the data volume of the test set within the interval [5,15], so as to avoid the reduction of prediction accuracy caused by redundant or inefficient data.
  • S26 Use the neural network model trained by S25 to predict the test data of S23, and use the mean absolute error (Mean Absolute Error, MAE) as the evaluation standard of the model prediction performance.
  • the MAE of the cell number 4 obtained in the embodiment of the present invention is 4.53 dB. If the prior art is adopted, that is, the remaining 66 cells are used as the training data to train the model, and the MAE is predicted to be 5.96 dB under the same experimental conditions. It can be seen that the predicted MAE of the embodiment of the present invention can be reduced by 1.43 dB.
  • S27 Select different cells among 67 cells as the test set, and repeat S23 to S26.
  • the field strength prediction method of the embodiment of the present invention ignores the environment in the average MAE of the predicted field strength and the true field strength of 13 test cells compared with the prior art.
  • the method of using full data for information is reduced by 1.77dB.
  • FIG. 3 is a comparison between the MAE of the predicted field strength and the true field strength of the 13 test cells in the embodiment of the present invention and the MAE of the method of ignoring environmental information and using full data in the prior art.
  • the white box in FIG. 3 is the MAE of the predicted field strength and true field strength of the 13 test cells in the embodiment of the present invention, and the black box is the MAE of the method using full data in the prior art.
  • Table 2 is the MAE values of the predicted field strength and the true field strength of the 13 test cells corresponding to FIG. 3 and the MAE values of the prior art using the full data method. Among them, the field strength prediction method of the embodiment of the present invention is tested in 13 The average MAE of the predicted field strength and the real field strength of the cell is reduced by 1.77dB compared with the method of ignoring environmental information and using full data in the prior art.
  • the difference between this embodiment and Embodiment 2 is that the model is converted from a fully-linked neural network to an Xgboost model as a basic model for field strength prediction.
  • the steps from S31 to S34 in the specific steps of this example are completely the same as the steps from S21 to S24 in Embodiment 2, and the difference between S35 and S37 is the change of the basic model.
  • the work parameter may include the transmission power, the antenna direction angle, the antenna downtilt angle, the antenna height, etc.
  • the change of the work parameter data may be specifically the change of the antenna direction angle angle.
  • S32 Sort the signal data in each set of single parameter data in S31 according to the order from the smallest distance to the base station to form an ordered set of signal data. Assuming that M groups of single working parameter data are obtained in S21, M groups of ordered signal data are formed.
  • S33 Select the cell with the number 4 from the 67 cells as the target cell. Assuming that the target cell includes the change in the X parameter value, 1/3X of the change in the work parameter value included in the target cell can be used as the known cell. Parameter value changes, the corresponding data is used as standard training data (equivalent to the second data combination in Example 1), other 2/3X parameters change as unknown parameters, and the corresponding data as test data (Equivalent to the third data combination in Embodiment 1). The selection of 1/3X as the change of the known working parameter value is to make the working parameter value of the test data change more than that of the known data, which can better verify the effectiveness of this embodiment.
  • S34 For the standard training data in S33, according to each set of single industrial parameter data included, calculate each set of ordered signal data formed by S32 separately to obtain the Pearson coefficient between the ordered signal data of single industrial parameter data And generate a Pearson coefficient value table. Specifically, for example, if M is equal to 100 and X is equal to 60, then each group of ordered data signals in the 100 groups of ordered data signals will be successively subjected to the Pearson coefficient with respect to the 20 groups of standard training data. After 2000 calculations Get the Pearson coefficient value table.
  • S35 According to the Pearson value table generated in S24, select the standard training data, each group of single parameter data, the top ten non-target cell single parameter data with the most consistent fluctuation pattern.
  • the first 10 non-target cell single parameter data with the most consistent fluctuation law are fused with the standard training data and used as the model training set to train an Xgboost model.
  • the most consistent size of the fluctuation law is to make the ratio of the data volume of the training set to the data volume of the test set within the interval [5,15], so as to avoid the reduction of prediction accuracy caused by redundant or inefficient data.
  • S36 Use the Xgboost model trained in S35 to predict the test set in S33, and use MAE as the evaluation standard for model prediction performance.
  • the MAE of the cell number 4 obtained in the embodiment of the present invention is 2.37 dB. If the prior art is adopted, that is, the remaining 66 cells are used as the training data to train the model, and the MAE is predicted to be 5.54 dB under the same experimental conditions. It can be seen that the predicted MAE of the embodiment of the present invention can be reduced by 2.17 dB.
  • S27 Select different cells from the 67 cells as the test set, and repeat S23 to S26.
  • the field strength prediction method of the embodiment of the present invention reduces the average MAE between the predicted field strength and the true field strength in 13 test cells compared with the method using full data. 2.41dB.
  • FIG. 4 is a comparison between the MAE of the predicted field strength and the true field strength of the 13 test cells in the embodiment of the present invention and the MAE of the method of ignoring environmental information and using full data in the prior art.
  • the white box in FIG. 4 is the MAE of the predicted field strength and the true field strength of the 13 test cells in the embodiment of the present invention, and the black box is the MAE of the method using full data in the prior art.
  • Table 3 is the MAE values of the predicted field strength and the true field strength of the 13 test cells corresponding to FIG. 4 and the MAE values of the prior art using the full data method. Among them, the field strength prediction method of the embodiment of the present invention is tested in 13 The average MAE of the predicted field strength and the true field strength of the cell is 2.41dB lower than that of the existing method that ignores environmental information and uses full data.
  • Embodiments of the present invention provide a non-transitory (non-volatile) computer storage medium that stores computer-executable instructions, and the computer-executable instructions can execute the method in any of the foregoing method embodiments.
  • An embodiment of the present invention provides a computer program product.
  • the computer program product includes a computer program stored on a non-transitory computer-readable storage medium.
  • the computer program includes program instructions. When the program instructions are executed by a computer To make the computer execute the method in any of the above method embodiments.
  • FIG. 5 is a schematic diagram of a hardware structure of an electronic device for performing a field strength prediction method provided by an embodiment of the present invention.
  • the device includes one or more processors 610 and a memory 620. Take a processor 610 as an example.
  • the device may further include: an input device 630 and an output device 640.
  • the processor 610, the memory 620, the input device 630, and the output device 640 may be connected through a bus or in other ways. In FIG. 5, connection through a bus is used as an example.
  • the memory 620 can be used to store non-transitory software programs, non-transitory computer executable programs, and modules.
  • the processor 610 executes non-transitory software programs, instructions, and modules stored in the memory 620 to execute various functional applications and data processing of the electronic device, that is, to implement the processing methods of the foregoing method embodiments.
  • the memory 620 may include a storage program area and a storage data area, where the storage program area may store an operating system and application programs required for at least one function; the storage data area may store data, and the like.
  • the memory 620 may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices.
  • the memory 620 may optionally include memories remotely disposed with respect to the processor 610, and these remote memories may be connected to the processing device through a network. Examples of the aforementioned network include, but are not limited to, the Internet, intranet, local area network, mobile communication network, and combinations thereof.
  • the input device 630 can receive input digital or character information, and generate signal input.
  • the output device 640 may include a display device such as a display screen.
  • the one or more modules are stored in the memory 620, and when executed by the one or more processors 610, execute the method in any of the above method embodiments.
  • the above-mentioned products can execute the method provided by the embodiment of the present invention, and have function modules and beneficial effects corresponding to the execution method.
  • function modules and beneficial effects corresponding to the execution method For technical details that are not described in detail in this embodiment, refer to the method provided in this embodiment of the present invention.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Electromagnetism (AREA)
  • General Physics & Mathematics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本发明公开了一种场强预测方法，包括：从非目标区域的数据集合中选择第一数据组合，第一数据组合所对应的信号与目标区域的第二数据组合所对应的信号的波动规律一致程度符合预设规则；进行目标区域的场强预测模型的训练，场强预测模型的训练样本包括第一数据集合。

Description

一种场强测试方法
交叉引用
本发明要求在2018年12月20日提交中国专利局、申请号为201811563761.2、发明名称为“一种场强测试方法”的中国专利申请的优先权，该申请的全部内容通过引用结合在本发明中。
技术领域
本发明涉及通信技术领域，尤其涉及一种信号覆盖区的场强预测方法。
背景技术
在场强预测等相关技术领域内，环境因素是决定模型精确度的一个重要因素。目前，对于环境信息的使用主要有两种方式：第一，对环境各种因素进行实地的精细的人工测量；第二，忽略环境信息，使用非目标区域内所有的数据集合(全数据)作为训练样本进行建模。第一种方式可以充分利用环境信息，在场强预测的建模中能够达到较高的精度，但是，环境信息的测量需要大量的人力物力，以及非常高的时间成本，在快速建模需求中并不适用。第二种方式虽然能够满足快速建模需求，但是由于采用所有的已知数据集合作为训练样本进行建模，其所产生的模型精度较低，误差较大，在解决场强覆盖问题中可用性不高。
为此，急需一种不需要耗费大量的人力物力，又具有较高精度的场强预测方法。
发明内容
为了解决上述问题，本发明提供一种场强预测方法，包括：从非目标区域的数据集合中选择第一数据组合，所述第一数据组合所对应的信号与目标区域的第二数据组合所对应的信号的波动规律一致程度符合预设规则；进行 所述目标区域的场强预测模型的训练，所述场强预测模型的训练样本包括所述第一数据集合。
为实现上述发明目的，本发明实施例还提供了一种电子设备，包括：至少一个处理器；以及与所述至少一个处理器通信连接的存储器；其中，所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器执行以上各个方面所述的方法。
为实现上述发明目的，本发明实施例还提供了一种非暂态计算机可读存储介质，所述非暂态计算机可读存储介质存储有计算机可执行指令，所述计算机可执行指令用于执行以上各个方面所述的方法。
为实现上述发明目的，本发明实施例还提供了一种计算机程序产品，所述计算机程序产品包括存储在非暂态计算机可读存储介质上的计算机程序，所述计算机程序包括程序指令，当所述程序指令被计算机执行时，使所述计算机执行以上各个方面所述的方法。
附图说明
此处所说明的附图用来提供对本发明的进一步理解，构成本发明的一部分，本发明的示意性实施例及其说明用于解释本发明，并不构成对本发明的不当限定。在附图中：
图1为本发明实施例1的场强预测过程；
图2为本发明实施例1的具体操作流程；
图3为本发明实施例2的场强预测方法的MAE与现有技术场强预测MAE的对比图；
图4为本发明实施例3的场强预测方法的MAE与现有技术场强预测MAE的对比图；
图5为本发明实施例提供的执行场强预测方法的电子设备的硬件结构示意图。
具体实施方式
为使本发明的目的、技术方案和优点更加清楚，下面将结合本发明具体实施例及相应的附图对本发明技术方案进行清楚、完整地描述。显然，所描述的实施例仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
实施例1
图1为本发明实施例1中的场强预测过程。该场强预测方法包括：S13:从非目标区域的数据集合中选择第一数据组合，所述第一数据组合所对应的信号相对目标区域的第二数据组合所对应的信号的波动规律一致程度符合预设规则；S14：进行所述目标区域的场强预测模型的训练，所述场强预测模型的训练样本包括所述第一数据集合。
本发明实施例提供的场强预测方法，相对于现有技术中将非目标区域的所有数据集合用于目标区域建模，本发明从非目标区域的所有数据集合中选择的第一数据组合作为场强预测模块的训练样本，且该第一数据组合所对应的信号与第二组数据所对应的信号的波动规律一致程度符合预设规则，即本发明所选择的第一数据组合与第二数据组合具有一定程度的相关程度，也就是去除了非目标区域的数据集合中与第二数据组合的波动规律相差较多的数据，因此，以该第一数据组合作为场强预测模型的训练样本，可以使得场强预测模型具有较高的精度，且该预测方法无需对环境的各种因素进行实地测量。
具体的，非目标区域可以划分为多个子区域，目标区域也可以再分为多个子区域，但是通常目标区域可以是一个子区域。该目标区域可以是非目标区域中的一个子区域。例如，非目标区域为N个小区，每个小区为一个子区域，那么N个小区中的第k个小区可以作为目标区域，N和k均为正整数，且k的值大于0小于等于N。
其中，第一数据组合和第二数据组合中的元素均可以包括基站的天线工参，简称为工参。实际应用中，对于出现无线信号弱覆盖问题的小区，可以调整基站的天线工参以提升基站对本小区弱覆盖区域的信号覆盖强度。天线工参通常有多个维度：发射功率、天线方向角、天线下倾角、天线高度等等。由于工参维度多，调整方向和大小范围大，现实中的发射功率多采用最大功率，而调整天线高度会导致信号覆盖强度急剧变化，因此，通常通过调整天线方向角来调整信号覆盖强度。
第一数据组合的元素还可以包括其他的信号数据，例如非目标区域内子区域与基站距离、相对位置、信号强度等等。同理，第二数据组合中的元素也可以包括非目标区域与基站距离、相对位置等等。
在S13之前还包括S11和S12。
S11：将数据集合按所述工参的数值变化划分成多组工参数据；S12:将目标区域的第二数据组合按所述工参的数值变化划分成多组第二数据。
如上分析，由于实际应用中通常通过调整天线方向角来调整信号覆盖强度，因此，S11和S12中的工参可以为天线方向角。即S11中数据集合按相对基站的天线方向角的角度不同而划分成多组工参数据。S12中的第二数据组合也按相对基站的天线方向角的角度不同而划分成多组第二数据。
当然，S11和S12中的工参也可以包括但不限于发射功率、天线下倾角、天线高度等。具体的，S11和S12中工参可以为单一工参,工参数据单一的工参数据，单一工参数据是指其中的工参的参数或维度只有一个，例如只有天线方向角这一工参或只有天线高度这一工参。此外，S11和S12中的工参也可以包括至少两个参数，例如包括天线方向角和天线高度两个参数，可以将数据集合中天线方向角和天线高度均相同的数据分为一组工参数据，第二数据组合中天线方向角和天线高度均相同的数据分为一组第二数据。
如上所述，第一数据组合的元素还可以包括非目标区域内子区域与基站距离，进一步的，非目标区域的数据集合的元素也可以包括其他的信号数据， 例如非目标区域内所有子区域与基站的距离。S11中还可以包括：每一组工参数据对应的信号数据依距离进行排序，以形成有序的信号数据，该工参数据为单一工参数据。
S13中，从S11中获得的多组工参数据选择目标工参数据，所有的目标工参数据即组成第一数据组合，该目标工参数据相对多组第二数据的相关程度符合预设规则，换言之，该目标工参数据对应的信号相对至少一组第二数据的信号的波动规律一致程度符合预设规则。
在S13中，可以基于相关系数，例如皮尔逊相关系数，计算第一数据组合所对应的信号相对第二数据组合所对应的信号的波动规律一致程度，具体的，可以是基于皮尔逊相关系数，计算目标工参数据相对第二数据的相关程度。
S13中，具体包括：衡量所述非目标区域的数据集合中所有的数据组合所对应的信号相对所述第二数据组合所对应的信号的波动规律一致程度，并按一致程度高低排序，将所述一致程度前若干名所对应的数据组合作为所述第一数据组合或者将所述一致程度在预设范围内所对应的数据组合作为所述第一数据组合。例如，将所述一致程度前10、或20名等所对应的数据组合作为所述第一数据组合，即从所有的数据集合中挖掘出与第二数据组合的一致程度较高的数据组合作为第一数据组合，排除一致程度较低的数据组合；又例如是将一致程度在0.8至1之间所对应的数据组合作为第一数据组合，一致程度的数值不超过1，当一致程度为1时表示两组数据组合的波动规律完全相同。
S14中，场强预测模型的训练样本还可以包括第二数据，即场强预测模型的训练样本包括第一数据组合和第二数据组合。除第二数据组合外，目标区域还可以包括第三数据组合，第三数据组合可以作为场强预测模型的测试样本，用于检验场强预测模型用于场强预测的效果。具体的，目标区域中的所有数据组合可以依据工参的数值变化划分成多组数据组合，多组数据组合 中的1/3组别为第二数据组合(工参数值变化已知)，另外2/3组别为第三数据组合(可认为工参数值变化未知)，第三数据数据组合的数据量大于第二数据组合的数据量，使得测试样本的工参数值变化多于已知的工参数值变化，可以更加有效地验证场强预测模型的预测效果。例如，目标区域的所有数据组合内的工参数值包括6种(以天线方向角为例，10度，20度，30度，40度，50度，60度)；则可以选择天线方向角为10度和20度所对应的数据组合作为第二数据组合，其余的数据组合为第三数据组合。
当然，第三数据组合的数据量与第二数据组合的数据量也可以是其他比值，如多组数据组合中的1/4组别为第二数据组合(工参数值变化已知)，另外3/4组别为第三数据组合(可认为工参数值变化未知)。此外，多组数据组合中也可以是第三数据组合的数据量小于或等于第二数据组合的数据量。
可见，场强预测模型中训练样本包括第一数据组合和第二数据组合，测试样本包括第三数据组合。为了避免冗余和预测精度低的问题，场强预测模型中训练样本的数据量与测试样本的数据量的比值在5至15之间，例如8、10或12等等。具体的，第一数据组合和第二数据组合的数据量之和与第三数据组合数据量的比值在5至15之间。
本发明实施例的场强预测模型可以是深度学习领域的模型，如全连接神经网络模型，也可以是机器学习领域的模型，如Xgboost模型。
S14之后，还可以包括S16：基于所述场强预测模型，对所述目标区域进行场强预测。目标区域的相关数据输入场强预测模型，该场强预测模型计算后得出目标区域的预测场强值。
在S14和S16之间，还可以包括S15：验证所述场强预测模型的有效性，当所述场强预测模型有效时，进入S16,否则中止或终止。可以利用上述测试样本输入场强预测模块模型得到目标区域的场强预测值，并将该场强预测值与场强真实值通过平均绝对误差(Mean Absolute Error，MAE)的方法比对，得到场强预测值与场强真实值的差值为A。现有技术中使用非目标区域内所 有的数据集合作为训练样本进行建模，其得到的场强预测值与场强真实值的差值为B。通过比较A和B,可以评估本发明实施例的场强预测模型的有效性。
图2为本发明实施例1中的场强预测方法的具体操作流程。首先获得非目标区域的全集数据，即所有数据集合；将全集数据预处理为所需的数据形式，具体的，可以依据工参数值变化划分成需要的数据形式；从处理后的数据中找出每组数据对应波动规律，即获取每组数据的波动规律；从处理后的数据选择与目标区域对应的数据为标准训练数据，从所有处理后的数据中选择与标准训练数据波动规律较为一致的数据组合，将标准训练数据与该数据组合融合成为训练样本，进而使得模型训练的样本具有较高的相关性，以利于训练出精度较高的模型；选择模型，使用训练样本进行训练；选择评估方法，对模型测试结果进行评估。
下文将以小区作为目标为例，通过对实施例二和实施例三的介绍，说明本申请实施例提供的方案在实际中的具体应用过程。可以理解，实施例二和实施例三是对实施例一提供的方案的具体说明，而不应视为对该方案应用场景等进行限制。
实施例2
本发明实施例在67个不包含任何地形地貌信息的小区数据，67个小区数据组成的全数据集中，使用全链接神经网络，作为场强预测的基础模型。验证方案方法对模型性能提升大小。
具体实施步骤如下。
S21：将67个小区中每个小区数据，根据本小区数据中所包含的工参数值变化，划分成多组单一工参数据。其中，工参可以包括发射功率、天线方向角、天线下倾角、天线高度等等，S21中，工参数据变化可以具体为天线方向角角度变化。
S22：将S21中的每一组单一工参数据中信号数据，根据与基站的距离从小到大的顺序进行排序，形成一组有序的信号数据。假设S21中得到M组 单一工参数据，则形成M组有序的信号数据。
S23：从67个小区中选择编号为4的小区作为目标小区，假设目标小区包括的工参数值变化为X个，可以将目标小区包含的工参数值变化中的1/3X，作为已知工参数值变化，其所对应的数据做为标准训练数据(相当于实施例1中的第二数据组合)，其他2/3X工参数值变化作为未知工参变化，其所对应数据做为测试数据(相当于实施例1中的第三数据组合)。选择1/3X作为已知工参数值变化，是为了使测试数据的工参数值变化多于已知数据工参数值变化，可以较好地验证本实施例的有效性。
S24：对S23中的标准训练数据，按所包含的每一组单一工参数据，分别计算S22形成的每一组有序信号数据，得到单一工参数据的有序信号数据间的皮尔逊系数，并生成一个皮尔逊系数值表。具体的，例如M等于100，X等于60，则将100组有序的数据信号中的每一组有序的数据信号分别相对20组标准训练数据依次进行皮尔逊系数，经过2000次的计算后得到皮尔逊系数值表。
参表1，为部分小区中不同工参数据，两两之间皮尔逊系数值表，其中a_b表示编号为a的小区内的第b个公参变化)。
表1
Figure PCTCN2019121051-appb-000001
S25：根据S24生成的皮尔逊值表，选择标准训练数据中，每组单一工参数据，波动规律最一致的前10个非目标小区单一工参数据。波动规律最一致的前10个非目标小区单一工参数据与标准训练数据融合后一起作为模型 的训练集合,训练一个全链接的神经网络模型。其中，波动规律最一致的数量大小，是为了使训练集合的数据量与测试集合的数据量比值在[5,15]这个区间内，以避免冗余或非有效数据造成预测精度降低。
S26：使用S25训练的神经网络模型，对S23的测试数据进行预测，以平均绝对误差(Mean Absolute Error，MAE)作为模型预测性能的评估标准。本发明实施例得到的编号为4的小区的MAE为4.53dB，如果采用现有技术中，即采用其余66个小区作为训练数据训练模型，同样的实验条件下预测MAE为5.96dB。可见，本发明实施例的预测MAE相比可降低1.43dB。
S27：选择67个小区中不同的小区作为测试集，重复S23到S26，本发明实施例的场强预测方法在13个测试小区预测场强与真实场强的平均MAE比现有技术中忽略环境信息使用全数据的方法降低了1.77dB。
图3为本发明实施例中13个测试小区预测场强与真实场强的MAE与现有技术中忽略环境信息使用全数据的方法的MAE的对照。图3中白色框为本发明实施例13个测试小区预测场强与真实场强的MAE，黑色框为现有技术中使用全数据的方法的MAE。
表2为与图3对应的13个测试小区预测场强与真实场强的MAE数值与现有技术中使用全数据方法的MAE数值，其中，本发明实施例的场强预测方法在13个测试小区预测场强与真实场强的平均MAE比现有技术中忽略环境信息使用全数据的方法降低了1.77dB。
表2
Figure PCTCN2019121051-appb-000002
Figure PCTCN2019121051-appb-000003
实施例3
本实施例与实施例2的区别在于将模型从全链接神经网络转化为Xgboost模型，作为场强预测的基础模型。本实例具体步骤中S31到S34的步骤与实施例2中S21到S24的步骤完全一致，S35到S37的区别在于基础模型的改变。
具体实施步骤如下。
S31：将67个小区中每个小区数据，根据本小区数据中所包含的工参数值变化，划分成多组单一工参数据。其中，工参可以包括发射功率、天线方向角、天线下倾角、天线高度等等，S31中，工参数据变化可以具体为天线方向角角度变化。
S32：将S31中的每一组单一工参数据中信号数据，根据与基站的距离从小到大的顺序进行排序，形成一组有序的信号数据。假设S21中得到M组单一工参数据，则形成M组有序的信号数据。
S33：从67个小区中选择编号为4的小区作为目标小区，假设目标小区包括的工参数值变化为X个，可以将目标小区包含的工参数值变化中的1/3X，作为已知工参数值变化，其所对应的数据做为标准训练数据(相当于实施例1中的第二数据组合)，其他2/3X工参数值变化作为未知工参变化，其所对应数据做为测试数据(相当于实施例1中的第三数据组合)。选择1/3X作为已知工参数值变化，是为了使测试数据的工参数值变化多于已知数据工参数值变化，可以较好地验证本实施例的有效性。
S34：对S33中的标准训练数据，按所包含的每一组单一工参数据，分别计算S32形成的每一组有序信号数据，得到单一工参数据的有序信号数据 间的皮尔逊系数，并生成一个皮尔逊系数值表。具体的，例如M等于100，X等于60，则将100组有序的数据信号中的每一组有序的数据信号分别相对20组标准训练数据依次进行皮尔逊系数，经过2000次的计算后得到皮尔逊系数值表。
参上述表1，为部分小区中不同工参数据，两两之间皮尔逊系数值表，其中a_b表示编号为a的小区内的第b个公参变化。
S35：根据S24生成的皮尔逊值表，选择标准训练数据中，每组单一工参数据，波动规律最一致的前10个非目标小区单一工参数据。波动规律最一致的前10个非目标小区单一工参数据与标准训练数据融合后一起作为模型的训练集合,训练一个Xgboost模型。其中，波动规律最一致的数量大小，是为了使训练集合的数据量与测试集合的数据量比值在[5,15]这个区间内，以避免冗余或非有效数据造成预测精度降低。
S36：使用S35训练的Xgboost模型，对S33中的测试集合进行预测，以MAE作为模型预测性能的评估标准。本发明实施例得到的编号为4的小区的MAE为2.37dB，如果采用现有技术中，即采用其余66个小区作为训练数据训练模型，同样的实验条件下预测MAE为5.54dB。可见，本发明实施例的的预测MAE相比可降低2.17dB。
S27：选择67个小区中不同的小区作为测试集，重复S23到S26，本发明实施例的场强预测方法在13个测试小区预测场强与真实场强的平均MAE比使用全数据的方法降低了2.41dB。
图4为本发明实施例中13个测试小区预测场强与真实场强的MAE与现有技术中忽略环境信息使用全数据的方法的MAE的对照。图4中白色框为本发明实施例13个测试小区预测场强与真实场强的MAE，黑色框为现有技术中使用全数据的方法的MAE。
表3
Figure PCTCN2019121051-appb-000004
Figure PCTCN2019121051-appb-000005
表3为与图4对应的13个测试小区预测场强与真实场强的MAE数值与现有技术中使用全数据方法的MAE数值，其中，本发明实施例的场强预测方法在13个测试小区预测场强与真实场强的平均MAE比现有技术中忽略环境信息使用全数据的方法降低了2.41dB。
本发明实施例提供了一种非暂态(非易失性)计算机存储介质，所述计算机存储介质存储有计算机可执行指令，该计算机可执行指令可执行上述任意方法实施例中的方法。
本发明实施例提供了一种计算机程序产品，所述计算机程序产品包括存储在非暂态计算机可读存储介质上的计算机程序，所述计算机程序包括程序指令，当所述程序指令被计算机执行时，使所述计算机执行上述任意方法实施例中的方法。
图5是本发明实施例提供的执行场强预测方法的电子设备的硬件结构示意图，如图所示，该设备包括一个或多个处理器610以及存储器620。以一个处理器610为例。该设备还可以包括：输入装置630和输出装置640。
处理器610、存储器620、输入装置630和输出装置640可以通过总线或者其他方式连接，图5中以通过总线连接为例。
存储器620作为一种非暂态计算机可读存储介质，可用于存储非暂态软 件程序、非暂态计算机可执行程序以及模块。处理器610通过运行存储在存储器620中的非暂态软件程序、指令以及模块，从而执行电子设备的各种功能应用以及数据处理，即实现上述方法实施例的处理方法。
存储器620可以包括存储程序区和存储数据区，其中，存储程序区可存储操作系统、至少一个功能所需要的应用程序；存储数据区可存储数据等。此外，存储器620可以包括高速随机存取存储器，还可以包括非暂态存储器，例如至少一个磁盘存储器件、闪存器件、或其他非暂态固态存储器件。在一些实施例中，存储器620可选包括相对于处理器610远程设置的存储器，这些远程存储器可以通过网络连接至处理装置。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。
输入装置630可接收输入的数字或字符信息，以及产生信号输入。输出装置640可包括显示屏等显示设备。
所述一个或者多个模块存储在所述存储器620中，当被所述一个或者多个处理器610执行时，执行上述任意方法实施例中的方法。
上述产品可执行本发明实施例所提供的方法，具备执行方法相应的功能模块和有益效果。未在本实施例中详尽描述的技术细节，可参见本发明实施例所提供的方法。
以上的具体实例，对本发明的目的、技术方案和有益效果进行了进一步详细说明，所应理解的是，以上所述仅为本发明的具体实施例而已，并不用于限制本发明，凡在本发明的精神和原则之内，所做的任何修改、等同替换、改进等，均应包含在本发明的保护范围之内。

Claims (13)

  1. 一种场强预测方法，其中，包括：
    从非目标区域的数据集合中选择第一数据组合，所述第一数据组合所对应的信号相对目标区域的第二数据组合所对应的信号的波动规律一致程度符合预设规则；
    进行所述目标区域的场强预测模型的训练，所述场强预测模型的训练样本包括所述第一数据集合。
  2. 根据权利要求1的场强预测方法，其中，所述第一数据组合中和所述第二数据组合中的元素均包括工参；
    在所述非目标区域的数据集合中选择第一数据组合之前，还包括：
    将数据集合按所述工参的数值变化划分成多组工参数据；
    将目标区域的第二数据组合按所述工参的数值变化划分成多组第二数据；
    在所述从非目标区域的数据集合中选择第一数据组合，所述第一数据组合所对应的信号相对目标区域的第二数据组合所对应的信号的的波动规律一致程度符合预设规则中，包括：
    从所述多组工参数据中选择至少一组目标工参数据，所有所述目标工参数据组成所述第一数据组合，所述目标工参数据相对至少一组所述第二数据的相关程度符合预设规则。
  3. 根据权利要求2所述的场强预测方法，其中，所述工参包括基站的天线方向角。
  4. 根据权利要求2所述的场强预测方法，其中，基于皮尔逊相关系数，计算所述目标工参数据相对所述第二数据的相关程度，或计算所述第一数据组合所对应的信号相对所述第二数据组合所对应的信号的波动规律一致程度。
  5. 根据权利要求1至4任一项所述的场强预测方法，其中，在所述从非目标区域的数据集合中选择第一数据组合，所述第一数据组合所对应的信号相对目标区域的第二数据组合所对应的信号的波动规律一致程度符合预设规 则，包括：
    衡量所述非目标区域的数据集合中所有的数据组合所对应的信号相对所述第二数据组合所对应的信号的波动规律一致程度，并按一致程度高低排序，将所述一致程度前若干名所对应的数据组合作为所述第一数据组合或者将所述一致程度在预设范围内所对应的数据组合作为所述第一数据组合。
  6. 根据权利要求2所述的场强预测方法，其中，所述场强预测模型的训练样本还包括所述第二数据组合，所述目标区域还包括第二数据组合之外的第三数据组合，所述第三数据组合为所述场强预测模型的测试样本。
  7. 根据权利要求6所述的场强预测方法，其中，所述场强预测模型的训练样本数据量与测试样本的数据量比值在5至15之间。
  8. 根据权利要求1所述的场强预测方法，其中，所述目标区域场强预测模型为全连接神经网络或Xgboost模型。
  9. 根据权利要求2所述的场强预测方法，其中，所述数据集合的元素还包括非目标区域的子区域至基站的距离，在所述将数据集合按所述工参的数值变化划分成多组工参数据中，还包括将每一组所述工参数据的信号数据依所述距离进行排序，所述工参数据为单一工参数据。
  10. 根据权利要求1所述的场强预测方法，其中，所述目标区域与所述非目标区域相互独立；或者，所述目标区域为所述非目标区域的子区域。
  11. 根据权利要求1所述的场强预测方法，其中，在所述进行所述目标区域的场强预测模型的训练之后，还包括：基于所述场强预测模型，对所述目标区域进行场强预测。
  12. 一种电子设备，其中，包括：
    至少一个处理器；以及，
    与所述至少一个处理器通信连接的存储器；其中，
    所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器执行权利要求1-11中任一 项所述的方法。
  13. 一种非暂态计算机可读存储介质，其中，所述非暂态计算机可读存储介质存储有计算机可执行指令，所述计算机可执行指令用于执行权利要求1-11中任一项所述的方法。
PCT/CN2019/121051 2018-12-20 2019-11-26 一种场强测试方法 WO2020125349A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
EP19897593.0A EP3902313A4 (en) 2018-12-20 2019-11-26 FIELD STRENGTH TEST PROCEDURE

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811563761.2A CN111355543B (zh) 2018-12-20 2018-12-20 一种场强预测方法
CN201811563761.2 2018-12-20

Publications (1)

Publication Number Publication Date
WO2020125349A1 true WO2020125349A1 (zh) 2020-06-25

Family

ID=71100164

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/121051 WO2020125349A1 (zh) 2018-12-20 2019-11-26 一种场强测试方法

Country Status (3)

Country Link
EP (1) EP3902313A4 (zh)
CN (1) CN111355543B (zh)
WO (1) WO2020125349A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112183576A (zh) * 2020-08-25 2021-01-05 北京邮电大学 一种基于不均衡数据集的Time-LSTM分类方法
CN116806026A (zh) * 2023-08-25 2023-09-26 山东高速信息集团有限公司 一种基于高速公路建设的5g网络同步铺设方法及设备

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114002574A (zh) * 2021-10-29 2022-02-01 上海华力微电子有限公司 半导体结构的测试方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110070841A1 (en) * 2009-09-22 2011-03-24 Jesse Caulfield Method, system, and computer-readable medium for improved prediction of spectrum occupancy and estimation of radio signal field strength
CN106231621A (zh) * 2016-07-29 2016-12-14 武汉大学 一种用于fdd‑lte系统中传播模型的多场景自适应优化方法
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
CN108271165A (zh) * 2016-12-30 2018-07-10 中国移动通信集团浙江有限公司 一种基站网络信号覆盖状态的预测方法及系统

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7099860B1 (en) * 2000-10-30 2006-08-29 Microsoft Corporation Image retrieval systems and methods with semantic and feature based relevance feedback
CN105843818A (zh) * 2015-01-15 2016-08-10 富士通株式会社 训练设备和训练方法、判断设备、以及推荐设备
CN106384023A (zh) * 2016-12-02 2017-02-08 天津大学 基于主路径的混合场强预测方法
CN108183754B (zh) * 2018-01-05 2021-06-22 湘潭大学 一种gsm基站电磁辐射预测方法
CN108229588B (zh) * 2018-02-08 2020-04-07 重庆师范大学 一种基于深度学习的机器学习识别方法
CN111327377B (zh) * 2018-12-14 2022-08-19 中兴通讯股份有限公司 场强预测方法、装置、设备及存储介质

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110070841A1 (en) * 2009-09-22 2011-03-24 Jesse Caulfield Method, system, and computer-readable medium for improved prediction of spectrum occupancy and estimation of radio signal field strength
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
CN106231621A (zh) * 2016-07-29 2016-12-14 武汉大学 一种用于fdd‑lte系统中传播模型的多场景自适应优化方法
CN108271165A (zh) * 2016-12-30 2018-07-10 中国移动通信集团浙江有限公司 一种基站网络信号覆盖状态的预测方法及系统

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3902313A4 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112183576A (zh) * 2020-08-25 2021-01-05 北京邮电大学 一种基于不均衡数据集的Time-LSTM分类方法
CN112183576B (zh) * 2020-08-25 2022-12-27 北京邮电大学 一种基于不均衡数据集的Time-LSTM分类方法
CN116806026A (zh) * 2023-08-25 2023-09-26 山东高速信息集团有限公司 一种基于高速公路建设的5g网络同步铺设方法及设备
CN116806026B (zh) * 2023-08-25 2023-11-03 山东高速信息集团有限公司 一种基于高速公路建设的5g网络同步铺设方法及设备

Also Published As

Publication number Publication date
CN111355543A (zh) 2020-06-30
EP3902313A4 (en) 2022-09-28
CN111355543B (zh) 2023-01-17
EP3902313A1 (en) 2021-10-27

Similar Documents

Publication Publication Date Title
WO2020125349A1 (zh) 一种场强测试方法
WO2019062734A1 (zh) 基于Wi-Fi热点的室内定位方法及装置
CN104613930B (zh) 一种测距的方法、装置及移动终端
CN105911539B (zh) 一种到达时间差的误差补偿值确定方法及装置
CN108375363B (zh) 天线方位角偏转核查方法、装置、设备及介质
CN109671174A (zh) 一种电塔巡检方法及装置
CN107992898A (zh) 一种电子元件焊点缺陷检测方法
CN103096363A (zh) 模拟弱无线网络的方法和装置
CN107809766A (zh) 一种用于网络优化的机器学习样本的生成方法及装置
Dong et al. Chaotic species based particle swarm optimization algorithms and its application in PCB components detection
CN116563841B (zh) 配电网设备标识牌的检测方法、检测装置和电子设备
CN104502967B (zh) 快速获取地震勘探观测系统面元信息的方法及装置
CN114461618A (zh) 一种计量缺失数据补全方法、装置及终端设备
CN116522565B (zh) 一种基于bim的电力工程设计配电网规划方法及计算机设备
CN109934151B (zh) 一种基于movidius计算芯片和Yolo face的人脸检测方法
CN105813103B (zh) 基站俯仰角的测量方法和装置
CN113852914B (zh) 待检验点的位置定位方法及定位装置
CN113822892B (zh) 仿真雷达的评测方法、装置、设备及计算机存储介质
CN113989632A (zh) 一种遥感图像桥梁检测方法、装置、电子设备及存储介质
CN108595764A (zh) 电路模块行为级建模方法及系统
CN109362027B (zh) 定位方法、装置、设备及存储介质
Kincses et al. Modelling the spatial structure of Europe
CN110045362B (zh) 基于rnn的雷达目标航迹识别方法和装置
CN107145938A (zh) 基于测井信息的储层岩石中值半径预测方法
Lu et al. Extreme learning machine with dead zone and its application to WiFi based indoor positioning

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: 19897593

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2019897593

Country of ref document: EP

Effective date: 20210720