Network coverage quality evaluation method, device and equipment

Abstract

This invention discloses a method, apparatus, and device for network coverage quality assessment. The method includes: dividing the area to be assessed into multiple grids to obtain grid division results; wherein, the grid division results include primary serving cell information and neighboring cell information corresponding to each sampling point within the multiple grids; checking whether a target network frequency band exists in the primary serving cell information set and neighboring cell information set corresponding to each grid in the grid division results, and performing corresponding data processing on the corresponding grids according to different check results to obtain coverage intensity data of the target network frequency band at the corresponding sampling points within each grid; based on the coverage intensity data, statistically analyzing the coverage quality index of each grid under the target network frequency band, and performing cluster analysis on all grids to obtain the coverage quality partitions of the target network frequency band in the area to be assessed. This invention can effectively reduce the cost of coverage assessment, improve the accuracy and comprehensiveness of assessment results, and accurately locate areas with weak coverage.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a method, apparatus, and electronic device for assessing network coverage quality. Background Technology

[0002] With the development of next-generation networks, wireless network frequency bands are gradually being refarmed and reused for evolution, while previous-generation network systems only guarantee continuous coverage of the basic coverage layer. Currently, coverage quality assessment for target networks mainly employs simulation software prediction methods and frequency sweepers or drive test terminals for traversal testing.

[0003] The simulation software prediction method uses simulation software to input factors affecting signal propagation, such as building height, antenna height, power, and frequency, to establish a propagation model for coverage prediction. It relies on theoretical calculations to predict coverage quality. However, its accuracy is limited by the precision of the propagation model and the completeness of electronic map data. It cannot fully match complex real-world environments such as urban buildings and vegetation, and it is also difficult to reflect real-time dynamic changes such as user movement and network load fluctuations, resulting in a large deviation between the prediction results and the actual network coverage.

[0004] The frequency sweeper or drive test terminal traversal testing method requires a systematic and traversal scan of the target area to collect signal strength, interference, and other key signal data within the frequency range. However, it suffers from long testing cycles, high costs, and low efficiency. First, it requires specialized equipment and personnel, and may need to be conducted multiple times at different times and locations, increasing human and material costs. Second, drive tests can only provide limited coverage data, making it difficult to cover the entire network area, especially in large-scale cities or complex terrains, where comprehensive coverage assessment is challenging. Finally, the periodicity and limitations of drive tests may prevent the timely detection and resolution of network problems, affecting the real-time nature of network maintenance and optimization.

[0005] Therefore, there is an urgent need for a low-cost, high-precision and comprehensive method to assess the coverage quality of the target network, in order to support the evolution of the existing network towards the network and identify holes and weak field areas in the frequency band coverage of the target network. Summary of the Invention

[0006] The purpose of this invention is to provide a network coverage quality assessment method, device, and electronic device that can achieve refined acquisition of the coverage strength of the target network frequency band at relevant sampling points within the grid. It can complete the full-domain coverage quality partitioning through quantitative indicators and cluster analysis, without relying on complex simulation modeling or large-scale traversal testing, thereby effectively reducing the cost of coverage assessment, improving the accuracy and comprehensiveness of assessment results, and accurately locating areas with weak coverage.

[0007] A first aspect of the present invention provides a method for assessing network coverage quality, comprising: The area to be evaluated under the coverage of a multi-layer network is divided into multiple grids to obtain grid division results; wherein, the grid division results include: the primary serving cell information and neighboring cell information corresponding to each sampling point within the multiple grids; In the grid division results, it is checked whether the target network frequency band exists in the primary serving cell information set and neighbor cell information set corresponding to each grid, and the corresponding grid is processed accordingly based on different check results to obtain the coverage intensity data of the target network frequency band at the corresponding sampling point in each grid. Based on the corresponding coverage intensity data, the coverage quality index of each grid cell under the target network frequency band is statistically analyzed, and cluster analysis is performed on all grid cells to obtain the coverage quality partition of the target network frequency band in the area to be evaluated.

[0008] Optionally, the step of performing corresponding data processing on the corresponding grids based on different inspection results to obtain coverage intensity data of the target network frequency band at the corresponding sampling point in each grid includes: If the detection result shows that the target network frequency band does not exist in either the primary cell information set or the neighbor cell information set, then based on the preset wireless network coverage quality prediction model, the signal strength of the target network frequency band at the corresponding sampling point in the corresponding grid is obtained; wherein, the wireless network coverage quality prediction model is trained through the network measurement dataset of co-coverage cell pairs; the co-coverage cell pairs include the cell where the target network frequency band is located and the cell where the known network frequency band is located with the shared antenna feed configuration.

[0009] Optionally, the step of performing corresponding data processing on the corresponding grids based on different inspection results to obtain the coverage intensity data of the target network frequency band at the corresponding sampling point within each grid further includes: If the check result indicates that the target network frequency band exists in the primary serving cell information set of the grid, then the signal strength of the target network frequency band at the corresponding sampling point is directly extracted from the corresponding primary serving cell information set. If the inspection result indicates that the target network frequency band exists only in the neighbor cell information set of the grid, then from the corresponding neighbor cell information set, determine the signal strength of the target network frequency band in multiple neighbor cells of the corresponding sampling point, and select the maximum value as the signal strength of the target network frequency band at the corresponding sampling point; The signal intensity at the corresponding sampling point is used as the coverage intensity data of the corresponding grid.

[0010] Optionally, the wireless network coverage quality prediction model takes into account the received signal strength of a known network frequency band at the corresponding sampling point and the distance between the grid where the corresponding sampling point is located and the base station cell to which it belongs, and outputs the signal strength of the target network frequency band at the corresponding sampling point.

[0011] Optionally, the model type of the wireless network coverage quality prediction model is a linear regression model.

[0012] Optionally, the step of dividing the evaluation area under multi-layer network coverage into multiple grids to obtain grid division results includes: The sampling point data reported by users in the area to be evaluated is obtained and cleaned to obtain the cleaned sampling point data. The area to be evaluated is rasterized, and the cleaned sampling point data is mapped to the corresponding raster according to geographical location to obtain the raster division result.

[0013] Optionally, the coverage quality indicators include at least one of the following: average received level of the grid, number of target network sampling points, number of target network weak coverage sampling points, and coverage rate.

[0014] Optionally, the sampling point data includes at least one of application software crowdsourcing data and minimized road test data.

[0015] A second aspect of the present invention provides a network coverage quality assessment device, comprising: The grid processing module is used to divide the area to be evaluated under the coverage of a multi-layer network into multiple grids to obtain grid division results; wherein, the grid division results include: the primary serving cell information and neighboring cell information corresponding to each sampling point within the multiple grids; The coverage strength acquisition module is used to check whether the target network frequency band exists in the primary serving cell information set and neighbor cell information set corresponding to each grid in the grid division result, and to perform corresponding data processing on the corresponding grid according to different check results, so as to obtain the coverage strength data of the target network frequency band at the corresponding sampling point in each grid. The quality assessment module is used to calculate the coverage quality index of each grid cell under the target network frequency band based on the corresponding coverage strength data, and to perform cluster analysis on all grid cells to obtain the coverage quality partition of the target network frequency band in the area to be assessed.

[0016] A third aspect of the present invention provides an electronic device including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the network coverage quality assessment method according to any embodiment of the first aspect.

[0017] Compared with existing technologies, embodiments of the present invention provide a network coverage quality assessment method, apparatus, and electronic device. The embodiments of the present invention first divide the area to be assessed under multi-layer network coverage into multiple grids to obtain grid division results. The grid division results include: primary serving cell information and neighboring cell information corresponding to each sampling point within the multiple grids. Next, in the grid division results, it is checked whether the target network frequency band exists in the primary serving cell information set and neighboring cell information set corresponding to each grid. Based on different check results, corresponding data processing is performed on the corresponding grids to obtain the coverage intensity data of the target network frequency band at the corresponding sampling points within each grid. Finally, based on the corresponding coverage intensity data, the coverage quality index of each grid under the target network frequency band is statistically analyzed, and cluster analysis is performed on all grids to obtain the coverage quality partitions of the target network frequency band in the area to be assessed. This invention enables refined acquisition of coverage strength at relevant sampling points within a grid for a target network frequency band. It completes full-area coverage quality zoning through quantitative indicators and cluster analysis, without relying on complex simulation modeling or large-scale traversal testing. This effectively reduces coverage assessment costs, improves the accuracy and comprehensiveness of assessment results, and accurately locates areas with weak coverage. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating an embodiment of the network coverage quality assessment method provided by the present invention; Figure 2 This is a flowchart illustrating an embodiment of the rasterization process provided by the present invention; Figure 3 This is a flowchart illustrating an embodiment of the wireless network coverage quality prediction model provided by the present invention. Figure 4 This is an example diagram of an embodiment of the training feature data provided by the present invention; Figure 5 This is a schematic diagram of an embodiment of the propagation curve of a pair of cells with the same coverage provided by the present invention; Figure 6 This is a schematic diagram of an embodiment of the network coverage quality assessment device provided by the present invention; Figure 7 This is a schematic diagram of the structure of an embodiment of the electronic device provided by the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] See Figure 1 This is a flowchart illustrating an embodiment of the network coverage quality assessment method provided by the present invention.

[0021] A first aspect of the present invention provides a method for evaluating network coverage quality, including steps S1 to S3, as detailed below: Step S1: Divide the area to be evaluated under the coverage of the multi-layer network into multiple grids to obtain grid division results; wherein, the grid division results include: the primary serving cell information and neighboring cell information corresponding to each sampling point within the multiple grids; Step S2: In the grid division result, check whether the target network frequency band exists in the primary serving cell information set and neighbor cell information set corresponding to each grid, and perform corresponding data processing on the corresponding grid according to different check results to obtain the coverage intensity data of the target network frequency band at the corresponding sampling point in each grid; Step S3: Based on the corresponding coverage intensity data, calculate the coverage quality index of each grid cell under the target network frequency band, and perform cluster analysis on all grid cells to obtain the coverage quality partition of the target network frequency band in the area to be evaluated.

[0022] It should be noted that in step S1, sampling point data reported by the user in the area to be evaluated is obtained based on application software (APP) crowdsourcing data and Minimization Drive Test (MDT) data. Each sampling point data includes at least primary serving cell information and neighboring cell information; the primary serving cell information includes: network type, base station and cell number, sampling latitude and longitude, primary serving cell frequency and received signal level (i.e., signal strength); the neighboring cell information includes the Physical Cell Identifier (PCI), center frequency, and received signal level of m neighboring cells respectively. When a user is located at a certain location, the base station cell that the terminal is currently connecting to and providing service to is the primary serving cell, and the terminal will continuously scan the surrounding neighboring cell signals to prepare for cell handover. In the existing APP crowdsourcing and MDT data, each sampling point data records the primary serving cell currently connected to the terminal, and records a maximum of 6 neighboring cells with the strongest signals, i.e., the maximum value of m is 6. The area to be evaluated is rasterized to obtain multiple grids, providing a data foundation for subsequent refined analysis. The primary serving cell information corresponding to all sampling points within a grid constitutes the primary serving cell information set of that grid, and correspondingly, the neighboring cell information corresponding to all sampling points within that grid constitutes the neighboring cell information set.

[0023] In step S2, the embodiments of the present invention fully consider the impact of new technologies such as Massive MIMO (Massive Multiple Input Multiple Output) and beamforming on the coverage of the primary serving cell, and perform differentiated processing according to the measurement of the target network frequency band in each grid: (1) When the target network frequency band exists in the primary serving cell information set, the corresponding sampling points are directly cleaned, and the cleaned sampling points are used as the effective sampling points of the target network in this grid; (2) When the target network frequency band exists only in the neighboring cell information set, the strongest frequency point selection method is adopted, and the target network coverage strength of the corresponding sampling point is indirectly evaluated by using the neighboring cell signal; (3) When the target network frequency band does not exist in either the primary serving cell information set or the neighboring cell information set, the coverage strength of the target network in the relevant sampling points in the grid is predicted by machine learning algorithm.

[0024] In step S3, based on the coverage strength data obtained in step S2, the coverage quality index of each grid cell under the target network frequency band is calculated. Then, based on the geographic proximity of the grid cells and the similarity constructed from the coverage quality index, grid cells of the same category are integrated geographically to form coverage quality partitions for the target network frequency band in the area to be evaluated. Each coverage quality partition corresponds to a clear coverage quality level and geographic range, accurately identifying areas with weak coverage, no coverage, etc., providing a clear basis for subsequent network planning resource allocation and network optimization measures.

[0025] The embodiments of the present invention can achieve refined acquisition of the coverage strength of the target network frequency band at relevant sampling points within the grid, and complete the full-domain coverage quality partitioning through quantitative indicators and cluster analysis, without relying on complex simulation modeling or large-scale traversal testing, thereby effectively reducing the coverage assessment cost, improving the accuracy and comprehensiveness of the assessment results, and effectively adapting to the target network coverage quality assessment needs under multi-layer network evolution.

[0026] In an optional embodiment, dividing the area to be evaluated under multi-layer network coverage into multiple grids to obtain grid division results includes: The sampling point data reported by users in the area to be evaluated is obtained and cleaned to obtain the cleaned sampling point data. The area to be evaluated is rasterized, and the cleaned sampling point data is mapped to the corresponding raster according to geographical location to obtain the raster division result.

[0027] Specifically, such as Figure 2The diagram shown is a flowchart of an embodiment of the rasterization processing provided by the present invention. The sampling point data of the present invention includes at least one of APP crowdsourcing data and MDT data. After obtaining the initial sampling point data, it is necessary to clean it, that is, remove invalid data such as empty location information, empty primary server cell network measurement information, and invalid data where the distance between the primary server cell location and the sampling point location exceeds a reasonable range, to obtain cleaned sampling point data (e.g., ...). Figure 2 (The "cleaned data" in the document).

[0028] Furthermore, in this embodiment of the invention, a Geographic Information System (GIS) is used to rasterize / gridize the area to be evaluated, resulting in multiple grids. Based on the location information (latitude and longitude) of the sampling points in the cleaned data, each sampling point is assigned to its corresponding grid, thus obtaining the raster division result of the area to be evaluated (i.e., a structured representation of all cleaned sampling point data, which can be a table or a table format). Figure 2 (The "grid matrix" in the text).

[0029] exist Figure 2 In GIS, Tab and Shp are commonly used vector map formats. Each row in the raster matrix represents a sample point data point. n Indicates grid number, Simple n The sampling point number is represented by the primary serving RSRP (RSRP_s), which is the primary serving cell signal strength / received level of the sampling point; the primary serving frequency (EARFCN_s), which is the center frequency of the primary serving cell of the sampling point; the neighboring cell nRSRP (RSRP_n), which is the signal strength of the nth neighboring cell of the sampling point; and the neighboring cell n frequency (EARFCN_n), which is the center frequency of the nth neighboring cell of the sampling point. In an optional embodiment, the step of performing differential processing on the corresponding grid according to different inspection results to obtain the coverage strength data of the target network frequency band at the corresponding sampling point in each grid includes: If the detection result shows that the target network frequency band does not exist in either the primary cell information set or the neighbor cell information set, then based on the preset wireless network coverage quality prediction model, the signal strength of the target network frequency band at the corresponding sampling point in the corresponding grid is obtained; wherein, the wireless network coverage quality prediction model is trained through the network measurement dataset of co-coverage cell pairs; the co-coverage cell pairs include the cell where the target network frequency band is located and the cell where the known network frequency band is located with the shared antenna feed configuration.

[0030] Furthermore, the step of performing corresponding data processing on the corresponding grids based on different inspection results to obtain the coverage intensity data of the target network frequency band at the corresponding sampling point within each grid also includes: If the check result indicates that the target network frequency band exists in the primary serving cell information set of the grid, then the signal strength of the target network frequency band at the corresponding sampling point is directly extracted from the corresponding primary serving cell information set. If the inspection result indicates that the target network frequency band exists only in the neighbor cell information set of the grid, then from the corresponding neighbor cell information set, determine the signal strength of the target network frequency band in multiple neighbor cells of the corresponding sampling point, and select the maximum value as the signal strength of the target network frequency band at the corresponding sampling point; The signal intensity at the corresponding sampling point is used as the coverage intensity data of the corresponding grid.

[0031] It should be noted that, in this embodiment of the invention, the measurement distribution of the target network frequency band in its sampling point data differs within different grids, and different processing operations are employed based on different distribution results / inspection results. The specific distribution categories of the target network frequency band in the grid are as follows: (1) Type 1 (the result of the inspection is that the target network frequency band exists in the main serving cell information set of the grid): In the grid corresponding to Type 1, the target network is the main serving cell of some sampling points, and there is direct signal data. At this time, the receiving level of the main serving cell in the grid is directly extracted as the target network, and it is used as the effective sampling data in the grid of this type.

[0032] In other words, if there are some sampling points in a certain grid, and the primary serving cell itself is the target network, the signal strength of the primary serving cell frequency point measured at that sampling point is the actual coverage strength of the target network frequency band at that location. No additional calculation is needed; it can be used directly.

[0033] (2) Type 2 (the result of the inspection is that the target network frequency band exists in the neighbor cell information set of the grid only): In the grid corresponding to Type 1, the primary serving cell of all sampling points is not the target network, but the target network frequency band exists in the neighbor cells that the terminal can scan at some sampling points. At this time, the "strongest frequency point selection method" is used to take the strongest signal of the target network in the neighbor cell as the valid data of the sampling point.

[0034] In other words, the target network signal has already covered this grid, but the signal strength is insufficient and it has not been selected as the primary serving cell by the terminals in the grid. However, some terminals can still detect the existence of the target network.

[0035] For grids containing target network measurements within neighboring cells, which may contain measurement information from multiple target networks, the strongest frequency point selection method is used to extract the strongest level value corresponding to the target network frequency point in the neighboring cell as the effective data within this type of grid, and then re-integrate the effective sampling points of the target frequency points within the grid.

[0036] Specifically, sampling data from the primary serving cell in the raster matrix that does not contain the target network frequency point is extracted, and it is determined whether the target network frequency point (i.e., the target network frequency band) exists in the neighboring cells. If it exists, the RSRP values ​​of all the corresponding neighboring cells with the target frequency point are used as the input set. ={ 1, 2,..., },in Indicates the first The received signal level of each target neighboring cell; for the function ( ), calculate set The maximum value in ( ) = max(X); function ( Returns the calculated maximum level of the target network neighboring cell for this sample, and the average of the maximum RSRP of the target frequency point among all sampling points in this grid, as a characterization of the coverage capability within this grid.

[0037] (3) Type 3 (Detection results show that the target network frequency band is not present in either the primary server cell information set or the neighboring cell information set): In the grid corresponding to Type 3, all sampling points in the primary server and neighboring cells only have signal records of the known network frequency band (such as frequency band A), and no signal records of the target network frequency band (such as frequency band B). This is because of the current network layering strategy. When the terminal turns off inter-frequency measurement, it cannot measure the coverage quality of the target network, resulting in some grids having only single-frequency network sampling points.

[0038] For this type of grid, a machine learning algorithm is used to train a wireless network coverage quality prediction model using network measurement datasets of co-coverage cell pairs, predicting the coverage quality of the target frequency band. Co-coverage cell pairs include the cell containing the target network frequency band and the cell containing a known network frequency band with a shared antenna feed configuration. A shared antenna feed configuration means that the cell containing the target network frequency band and the cell containing the known network frequency band share the same physical antenna, having the same antenna height, azimuth angle, and downtilt angle.

[0039] like Figure 3 The diagram shown is a flowchart illustrating an embodiment of the wireless network coverage quality prediction model provided by this invention. Figure 3In this model, a single base station simultaneously transmits signals from frequency band A (the primary serving cell, not the target network) and frequency band B (the target network). The area to be evaluated is divided into two types of grids: single-frequency grids (orange) and multi-frequency grids (green). Single-frequency grids only contain sampling data for frequency band A, not frequency band B; multi-frequency grids (green) contain sampling data for both frequency band A and frequency band B (sample data used to train the model). Based on the multi-frequency grids, the signal strength (RSRP_A) of frequency band A at relevant sampling points, the signal strength (RSRP_B) of frequency band B at relevant sampling points, and network measurement data such as the distance d1 from the grid containing frequency band A / frequency band B to the cell are extracted. These data are organized into a structured feature matrix and used to train the wireless network coverage quality prediction model.

[0040] In an optional embodiment, the wireless network coverage quality prediction model takes into account the received signal strength of a known network frequency band at a corresponding sampling point and the distance between the grid where the corresponding sampling point is located and the base station cell to which it belongs, and outputs the signal strength of the target network frequency band at the corresponding sampling point.

[0041] like Figure 3 As shown, for a sampling point in an orange single-frequency grid, the RSRP_A of frequency band A and the distance d2 from the grid where the sampling point is located to the base station cell are extracted from the signal information of the primary serving cell (such as frequency band A) recorded by the APP crowdsourcing / MDT data. These are then input into the wireless network coverage quality prediction model. The model outputs the predicted RSRP_B of the sampling point in the target network frequency band (such as frequency band B) to complete the coverage quality prediction of the target network frequency band in the single-frequency grid.

[0042] Obviously, during model training, it is necessary to collect network measurement datasets (from multi-frequency grids) of the same coverage cell pairs, and use the signal strength of the known network frequency band at the relevant sampling points, as well as the distance from the grid where the relevant sampling point is located to the base station cell, as the feature data for training. The signal strength of the target network frequency band at the relevant sampling points is used as the label data for training. The preset initial model (such as a linear regression model) is then trained to obtain the wireless network coverage quality prediction model.

[0043] The training process of the model will be described in detail below: Step 1: Data preparation.

[0044] By associating with existing network operating parameters, information on co-coverage cell pairs corresponding to the target network frequency band (such as frequency band B) can be obtained from multi-frequency grid data, and key features such as the distance between the grid and the cell, and the difference in received signal level can be extracted. Figure 4 The image shown is an example diagram of an embodiment of the training feature data provided by the present invention. Figure 4In this process, the rasterized data is correlated with the existing network operating parameters to obtain complete data for each sampling point: "Sampling Point Identifier" and "Raster Identifier," used to locate the specific sampling point and its corresponding raster; "Primary Serving Cell Identifier" and "Co-Coverage Cell Pair Identifier," used to determine the primary serving cell of the sampling point, as well as other cells co-covering the primary serving cell, such as... Figure 4 In the text, "cell pair 1" indicates that cell A and cell B form a cell pair with the same coverage; "grid and cell distance" is the distance from the center point of the grid to the corresponding base station cell (such as d1, d2, ..., dn); "center frequency_S" and "received level_S" are used to record the center frequency and signal strength of the primary serving cell corresponding to the sampling point.

[0045] For the signal strength of two frequency bands within the same grid and covering cell pair (e.g., RSRP_A and RSRP_B), the received level difference RSRP_Δ = RSRP_A - RSRP_B is calculated. Finally, the grid identifier, covering cell pair, grid-cell distance, and RSRP difference are organized into structured feature data to provide data support for subsequent model analysis and training.

[0046] Step 2: Model training and model building.

[0047] Based on the general Cost231Hata propagation model of wireless signals, the received signal level at any location within a wireless network cell can be predicted. The calculation formula is as follows: ① Path loss: PL A =46.3 + 33.9 Log(F A -13.82 Log(H)+(44.9-6.55 Log(H)) Log(D) + C; ② Calculation of received signal level: RSRP A =PT–PL A ; Among them: PL A F represents the path loss in frequency band A, in dB; A The frequency of band A is represented in MHz; D represents the distance between the transmitting and receiving antennas in km; H represents the effective height of the base station antenna in m; C represents the environmental correction factor; RSRP A PT represents the received signal level of frequency band A measured by the terminal, in dBm; PT represents the transmit power of the cell antenna, in dBm; the typical values ​​for C in this embodiment of the invention are (dense urban areas: -2dB, urban areas: -5dB, dense suburbs: -8dB, suburbs: -10dB, rural areas: -26dB).

[0048] For two frequency band cells (A and B) in a dual-layer network with a shared antenna feed configuration, the effective antenna height (H), transmit / receive distance (D), and environmental correction factor (C) of the base stations at the same location are identical. Based on the above propagation model, the received signal level difference (RSRP_Δ) between frequency bands A and B at the same location is derived as follows: ③ Definition of level difference: RSRP Δ =RSRP A -RSRP B ; ④ Simplified derivation of the difference: RSRP Δ =(PT A -33.9 LogF A )-(PT) B -33.9 LogF B ); Among them, PT A PT B F A F B These are the transmit power and frequency of cells in frequency band A and frequency band B, respectively, and are constants in the parameter configuration. Therefore, in a shared antenna feeder coverage scenario, RSRP Δ For a fixed value, RSRP A With RSRP B There is a certain linear relationship between them.

[0049] like Figure 5 The diagram shown is a schematic representation of an embodiment of the propagation curve of a pair of cells with the same coverage provided by the present invention. Figure 5 In the meantime, the actual propagation curves of frequency band A and frequency band B fluctuate around the ideal propagation curve. At the same distance, the difference between the two remains stable and can be described by a linear relationship.

[0050] This invention employs a linear regression model from machine learning to establish a wireless network coverage quality prediction model. Known RSRP values ​​of band A and band B in a shared coverage area of ​​bands A and B are used as training data. To improve the accuracy of the model's predictions, "grid-to-site distance" is introduced as a correction factor for the differences in antenna height and diffraction attenuation between multi-layer networks. This determines the optimal linear equation describing the relationship between the signal strength of band B and related characteristics (such as the signal strength of band A and the grid-to-site distance), thereby achieving coverage quality prediction for band B within the main coverage area of ​​a single band A.

[0051] The basic form of a linear regression model is: RSRP B =ω0+ω1 RSRP A +ω2 distance; Where ω0, ω1, and ω2 are the coefficients of the model, i.e., ω0 is the intercept term, used to correct for the basis error; ω1 is the RSRP. A The coefficient ω1 is used to describe the linear relationship between the signal strength of frequency band A and frequency band B; ω2 is the distance coefficient used to describe the influence of the transmission and reception distance on the signal strength of frequency band B.

[0052] Step 3: Use machine learning to train the correlation between the primary network cell and the target network cell.

[0053] First, initialize the linear regression model by setting the initial values ​​of the model coefficients ω0, ω1, and ω2 to 0. Then, input the feature data and label data into the model, and iteratively update the model coefficients by minimizing the loss function, allowing the model to gradually fit the training data.

[0054] The loss function is constructed using the least squares method, with the objective of minimizing the sum of squared errors between the actual observed values ​​and model predictions of the RSRP in the target network band (B band).

[0055] in, It is the first The model prediction value for each sample. It is the first The actual observations of each sample are used. Through iterative iteration, a set of coefficients ω0, ω1, and ω2 is found that minimizes the loss function.

[0056] In the least squares method, the minimum value of the loss function is found by taking the partial derivative of the loss function and setting it to zero. For linear regression models, the coefficients can be solved using the following formula:

[0057]

[0058] in, and These represent the RSRP of the B band. B RSRP of Band A A The mean.

[0059] In practice, optimization algorithms such as gradient descent can be used to solve for the coefficients that minimize the loss function, and finally obtain the optimal ω0, ω1 and ω2, so that the model can accurately fit the training data, thereby achieving the prediction of the RSRP value of the B band within the grid of the A band signal measured only.

[0060] It is worth noting that the embodiments of the present invention do not impose specific limitations on the model type of the wireless network coverage quality prediction model, as long as it is trained using network measurement datasets of the same coverage cell pairs and can accurately predict the RSRP value of the target network frequency band. In particular, the preset linear regression model is the preferred model type for the wireless network coverage quality prediction model.

[0061] In an optional embodiment, the coverage quality metrics include at least one of the following: average received level of the grid, number of target network sampling points, number of target network weak coverage sampling points, and coverage rate.

[0062] It should be noted that, in this embodiment of the invention, the grid division results of the entire sampling point are combined, and each grid is used as a statistical unit to quantify and statistically analyze key KPI indicators (i.e., coverage quality indicators) such as the average RSRP, the number of target network sampling points, the number of target network weak coverage sampling points, and coverage rate within each grid. Based on the above multi-dimensional quantitative indicators, the "DBSCAN clustering algorithm" is used to complete the association division and classification analysis of the entire grid, identify and output the location information of the target network's poor coverage weak areas, thereby guiding the precise deployment of network resources and improving resource utilization efficiency.

[0063] To further demonstrate the technical effects achieved by the network coverage quality assessment method provided by this invention, the invention will be further described below with reference to application examples from the inventors' research and development process: This invention uses the 4G FDD1800M frequency band in urban areas as the target network frequency band. Based on the actual test data from the APP crowdsourcing test, it constructs a target network-specific coverage evaluation model, completes the model's effectiveness verification, accurately outputs the overall target network coverage quality evaluation results, and quickly identifies weak areas such as poor coverage quality and weak field holes. This is used to guide network construction, optimize the precise deployment of resources, complete the pre-planning for the evolution of 4G network frequency bands, realize the precise implementation of optimized governance, and improve the overall operational efficiency of multi-layer collaborative networks.

[0064] (a) Rasterization of raw sampled data (1) Determine the source data format for the APP crowdsourcing test. The dataset should include at least: network type, cell ECI, longitude, latitude, main cell received level RSRP_s, main cell carrier frequency number EARFCN_s, 6 neighboring cell received levels RSRP_n1~RSRP_n6, and 6 neighboring cell carrier frequencies EARFCN_n1~EARFCN_n6.

[0065] (2) By associating the urban GIS raster data, the APP crowdsourced testing source data is rasterized to determine the raster number to which the original sample belongs, and the raster division results are obtained (as shown in Table 1), which provides a data basis for subsequent differential processing.

[0066] Table 1. Example of raster division results

[0067] (b) Classify and process different types of grids: Based on the distribution of target network frequency bands of the primary serving cell and neighboring cells within each grid, divide the grids into three categories and adopt differentiated processing strategies: (1) Type 1 - Target Network Sampling within the Grid: For grids where the primary serving cell has an FDD1800M target network signal (as shown in Table 2), the sampling data within the grid is cleaned and filtered, taking into full account the impact of Massive MIMO, beamforming, and other technologies on the actual coverage of the primary serving cell: primary serving sampling data in non-target network frequency bands are removed, and only valid sampling data of the primary serving cell in the FDD1800M frequency band (as shown in the sampling data surrounded by the black dashed line in Table 2) are retained as the coverage strength data for this type of grid. In Table 2, "Band_s" represents the network frequency band name corresponding to the primary serving cell.

[0068] Table 2. Example of sampling point data for raster type 1

[0069] (2) Type 2 - Only neighboring cell information within the grid has target network sampling: For grids where the main serving cell has no FDD1800M signal but the neighboring cell list contains target network frequency band signals (as shown in Table 3), extract the level data of all FDD1800M neighboring cells under each sampling point, and use the strongest frequency point selection method to select the maximum received level of the target network neighboring cell within a single sampling point as the equivalent coverage strength data of the target network at that location.

[0070] Table 3. Example table of sampling point data for raster type 2

[0071] In Table 3, the data enclosed by the black dashed box is the sampling data of the FDD1800 frequency band, and the data corresponding to the thick black solid box is the sampling data of the non-FDD1800 frequency band.

[0072] (3) Type 3 - Target network measurement data is not available in the information of the main service cell and neighboring cells within the grid: For grids where no FDD1800M signal is collected in the information of the main service cell and neighboring cells, based on the multi-layer network coverage feature data (as shown in Table 4), and using Python tools to introduce a linear regression model, a wireless network coverage quality prediction model is trained to predict the coverage strength data of the target frequency band.

[0073] Table 4. Example Table of Multi-Layer Network Co-Coverage Feature Data

[0074] After model training and parameter solving, the standard linear regression model (i.e., the wireless network coverage quality prediction model) for this co-coverage cell pair is: RSRP_fdd=ω0+ω1 RSRP_d+ω2 distance; where ω0=0.778, ω1=-0.014, ω2=-13.095; RSRP_fdd is the signal strength / received level of FDD1800M in the same coverage cell pair; RSRP_d is the signal strength / received level of mid-band D in the same coverage cell pair.

[0075] The minimum root mean square error (RMSE) of the wireless network coverage quality prediction model is 5.26, indicating good predictive performance with a small deviation between the predicted results and actual observations. Using this prediction model, given a known received signal level (RSRP_d) of -80dBm in the co-covered D-band, the predicted RSRP value (i.e., RSRP_fdd_pred) for the FDD1800 band at the same location is -77.17dBm. Using the same principle, the signal level of the target network FDD1800 under this type of grid across the entire network is predicted (as shown in Table 5) to improve the sampling point data of the target network and enhance the accuracy of target network coverage quality assessment.

[0076] Table 5. Example of target network level prediction results for type 3 raster

[0077] (c) Evaluate the coverage quality of the target network based on the output results. The network coverage quality assessment method provided by this invention processes the rasterized data from the APP crowdsourcing test and calculates in batches various core coverage indicators of the FDD1800M target network, including average received level, total effective sampling, number of weak coverage samples, and regional coverage rate, to form the overall regional coverage quality statistical results, as shown in Table 6.

[0078] Table 6. Coverage Quality of FDD1800M Target Network

[0079] This invention utilizes the "DBSCAN clustering algorithm" to aggregate and divide regions based on grid geospatial proximity and coverage quality index similarity, accurately locating weak coverage areas. This is used to guide the precise deployment of network resources in advance for network evolution, filling in blind spots and weak areas, and improving resource efficiency. Figure 6 The diagram shown is a schematic representation of an embodiment of the target network weak coverage area distribution provided by the present invention.

[0080] (d) Conduct on-site measurements to address the issue of weak coverage areas. To verify the accuracy of the model's output regarding weak coverage areas in the FDD1800 band, a field traversal test was conducted (results are shown in Table 7). The verification revealed that the FDD1800 band exhibited varying degrees of weak coverage in all four identified problem areas. Therefore, timely coverage improvements and gap filling are necessary to enhance the coverage quality of the FDD1800 band in line with network evolution.

[0081] Table 7. Example of Field Test Verification in Areas with Weak Target Network Coverage

[0082] In summary, the network coverage quality assessment method provided by this invention has the following beneficial effects: (1) This embodiment of the invention is based on APP crowdsourced testing data / MDT data, and uses the actual network measurement data reported by the terminal. It ignores the process impact of network parameter configurations such as frequency band capability differences and antenna height on network coverage assessment, and directly takes the actual measurement results as the guide. At the same time, it fully considers the personalized network characteristics of the primary serving cell, neighboring cells and cells with the same coverage for the measurement, and creates a targeted target network coverage quality assessment model for network evolution. This embodiment of the invention is more accurate and scientific, and the calculation results are clear and intuitive. Compared with the method of predicting the link quality of cells in the same coverage frequency band based on measurement reports, it has higher accuracy and more comprehensive assessment capabilities.

[0083] (2) In this embodiment of the invention, even ignoring the differences in network resource configuration such as frequency band capability, antenna height, and antenna gain, personalized cell-level network quality prediction can be achieved by using big data machine learning algorithms guided by the user's actual network quality gridded data. This not only allows for accurate assessment of the coverage quality of the target network, but also, by combining the location information in the crowdsourced testing data, it can quickly locate weak coverage issues, intuitively present the location of areas with weak coverage, and accurately guide the orderly allocation of network resources.

[0084] (3) The embodiments of the present invention are fast and efficient, and can quickly realize the comprehensive analysis and evaluation of the target network coverage quality in various scenarios. At the same time, it supports network measurement data carrying location information such as MDT. It has good applicability and promotion, can effectively save manpower and material resources, realize the precise allocation of resources, and better guide the precise implementation of network planning and optimization work.

[0085] See Figure 7 This is a schematic diagram of an embodiment of the network coverage quality assessment device provided by the present invention.

[0086] A second aspect of the present invention provides a network coverage quality assessment device, comprising: The grid processing module 11 is used to divide the area to be evaluated under the coverage of a multi-layer network into multiple grids to obtain grid division results; wherein, the grid division results include: the primary serving cell information and neighboring cell information corresponding to each sampling point within the multiple grids; The coverage strength acquisition module 12 is used to check whether the target network frequency band exists in the primary serving cell information set and neighbor cell information set corresponding to each grid in the grid division result, and to perform corresponding data processing on the corresponding grid according to different check results, so as to obtain the coverage strength data of the target network frequency band at the corresponding sampling point in each grid. The quality assessment module 13 is used to calculate the coverage quality index of each grid cell under the target network frequency band based on the corresponding coverage intensity data, and to perform cluster analysis on all grid cells to obtain the coverage quality partition of the target network frequency band in the area to be assessed.

[0087] It should be noted that the network coverage quality assessment device provided in the second aspect embodiment of the present invention can realize all the processes of the network coverage quality assessment method described in any of the first aspect embodiments. The functions and technical effects of each module and unit in the device are the same as the functions and technical effects of the network coverage quality assessment method described in any of the first aspect embodiments, and will not be repeated here.

[0088] See Figure 7 This is a schematic diagram of the structure of an embodiment of the electronic device provided by the present invention.

[0089] A third aspect of the present invention provides an electronic device including a processor 21, a memory 22, and a computer program stored in the memory 22 and configured to be executed by the processor 21, wherein the processor 21 implements the network coverage quality assessment method described in any of the first aspects of the present invention when executing the computer program.

[0090] Preferably, the computer program can be divided into one or more modules / units (such as computer program one, computer program two, ...), and the one or more modules / units are stored in the memory 22 and executed by the processor 21 to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device.

[0091] The processor 21 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor, or the processor 21 can be any conventional processor. The processor 21 is the control center of the electronic device, connecting various parts of the electronic device through various interfaces and lines.

[0092] The memory 22 mainly includes a program storage area and a data storage area. The program storage area can store the operating system, applications required for at least one function, etc., and the data storage area can store related data, etc. In addition, the memory 22 can be a high-speed random access memory, or a non-volatile memory, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, and a flash card, etc., or the memory 22 can also be other volatile solid-state storage devices.

[0093] It should be noted that the aforementioned electronic devices may include, but are not limited to, processors and memory, as will be understood by those skilled in the art. Figure 7 The structural block diagram shown is merely an example of the structure of the aforementioned electronic device and does not constitute a limitation on the structure of the electronic device. The aforementioned electronic device may include more or fewer components than shown, or combine certain components, or use different components. The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for assessing network coverage quality, characterized in that, include: The area to be evaluated under the coverage of a multi-layer network is divided into multiple grids to obtain grid division results; wherein, the grid division results include: the primary serving cell information and neighboring cell information corresponding to each sampling point within the multiple grids; In the grid division results, it is checked whether the target network frequency band exists in the primary serving cell information set and neighbor cell information set corresponding to each grid, and the corresponding grid is processed accordingly based on different check results to obtain the coverage intensity data of the target network frequency band at the corresponding sampling point in each grid. Based on the corresponding coverage intensity data, the coverage quality index of each grid cell under the target network frequency band is statistically analyzed, and cluster analysis is performed on all grid cells to obtain the coverage quality partition of the target network frequency band in the area to be evaluated.

2. The network coverage quality assessment method as described in claim 1, characterized in that, The step of performing corresponding data processing on the corresponding grids based on different inspection results to obtain coverage intensity data of the target network frequency band at the corresponding sampling points within each grid includes: If the detection result shows that the target network frequency band does not exist in either the primary cell information set or the neighbor cell information set, then based on the preset wireless network coverage quality prediction model, the signal strength of the target network frequency band at the corresponding sampling point in the corresponding grid is obtained; wherein, the wireless network coverage quality prediction model is trained through the network measurement dataset of co-coverage cell pairs; the co-coverage cell pairs include the cell where the target network frequency band is located and the cell where the known network frequency band is located with the shared antenna feed configuration.

3. The network coverage quality assessment method as described in claim 2, characterized in that, The step of performing corresponding data processing on the corresponding grids based on different inspection results to obtain the coverage intensity data of the target network frequency band at the corresponding sampling point within each grid also includes: If the check result indicates that the target network frequency band exists in the primary serving cell information set of the grid, then the signal strength of the target network frequency band at the corresponding sampling point is directly extracted from the corresponding primary serving cell information set. If the inspection result indicates that the target network frequency band exists only in the neighbor cell information set of the grid, then from the corresponding neighbor cell information set, determine the signal strength of the target network frequency band in multiple neighbor cells of the corresponding sampling point, and select the maximum value as the signal strength of the target network frequency band at the corresponding sampling point; The signal intensity at the corresponding sampling point is used as the coverage intensity data of the corresponding grid.

4. The network coverage quality assessment method as described in claim 2, characterized in that, The wireless network coverage quality prediction model takes the received signal strength of a known network frequency band at a corresponding sampling point as input, and the distance between the grid where the corresponding sampling point is located and the base station cell to which it belongs, and outputs the signal strength of the target network frequency band at the corresponding sampling point.

5. The network coverage quality assessment method as described in claim 2, characterized in that, The model type of the wireless network coverage quality prediction model is a linear regression model.

6. The network coverage quality assessment method as described in claim 1, characterized in that, The process of dividing the evaluation area under multi-layer network coverage into multiple grids to obtain grid division results includes: The sampling point data reported by users in the area to be evaluated is obtained and cleaned to obtain the cleaned sampling point data. The area to be evaluated is rasterized, and the cleaned sampling point data is mapped to the corresponding raster according to geographical location to obtain the raster division result.

7. The network coverage quality assessment method as described in claim 1, characterized in that, The coverage quality indicators include at least one of the following: average received level of the grid, number of target network sampling points, number of target network weak coverage sampling points, and coverage rate.

8. The network coverage quality assessment method as described in claim 6, characterized in that, The sampling point data includes at least one of application software crowdsourcing data and minimal road test data.

9. A network coverage quality assessment device, characterized in that, include: The grid processing module is used to divide the area to be evaluated under the coverage of a multi-layer network into multiple grids to obtain grid division results; wherein, the grid division results include: the primary serving cell information and neighboring cell information corresponding to each sampling point within the multiple grids; The coverage strength acquisition module is used to check whether the target network frequency band exists in the primary serving cell information set and neighbor cell information set corresponding to each grid in the grid division result, and to perform corresponding data processing on the corresponding grid according to different check results, so as to obtain the coverage strength data of the target network frequency band at the corresponding sampling point in each grid. The quality assessment module is used to calculate the coverage quality index of each grid cell under the target network frequency band based on the corresponding coverage strength data, and to perform cluster analysis on all grid cells to obtain the coverage quality partition of the target network frequency band in the area to be assessed.

10. An electronic device, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the network coverage quality assessment method as described in any one of claims 1 to 8.

Images