Network edge coverage area discovery method and apparatus

Abstract

The present application relates to the technical field of mobile network, and provide a kind of network edge coverage area discovery method and device, the method comprises: based on the current index item of each to be identified cell in current period, in three dimensions from each to be identified cell in current period the first edge coverage cell set of identification;The current index item is input edge coverage cell prediction model, obtain the future index item in future period that the edge coverage cell prediction model output;Based on the future index item, in the three dimensions from each to be identified cell in the future period the second edge coverage cell set of identification;Take the intersection of the first edge coverage cell set and the second edge coverage cell set, based on the edge coverage cell in the intersection, determine edge coverage area.The method of the present application can discover edge coverage area in network in time.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mobile networks, in particular to a network edge coverage area discovery method and device. BACKGROUND

[0002] The network coverage edge area scenario has the characteristics of poor coverage, high capacity demand, and poor network perception. In recent years, due to the large-scale development of cities, the user structure has changed, and suburban expansion areas, newly built buildings, construction sites, etc. are typical edge coverage scenarios. The number of complaints in this area is increasing, and the matching degree of network structure and user structure is insufficient, and the user's wireless communication perception is poor. It is necessary to identify the edge scenario, and the existing technology mainly discovers through complaint site investigation and manual analysis. The existing identification method of edge user perception demand mainly includes the following two kinds:

[0003] (1) Through complaint identification: the existing method is to collect complaints, confirm new buildings, construction sites and suburban expansion areas, etc. Edge coverage scenarios, and use on-site investigation to confirm user demand, but due to the lag of problem discovery, incomplete evaluation methods, and insensitive user demand around complaints, it is not possible to timely and comprehensively master the total edge coverage area.

[0004] (2) Through manual analysis: by extracting MR coverage rate (MR coverage rate is an important indicator to measure the performance of a mobile network, which refers to the proportion of sampling points that meet certain signal strength requirements in the total sampling points in a mobile network) and capacity indicators, combined with satellite map to check if there is a problem in the area, which is mainly used for site planning dimension, but due to the large amount of manual checking work, and the certain lag of satellite map, some edge areas cannot be discovered in time. SUMMARY

[0005] The present application provides a network edge coverage area discovery method and device to solve the problem that the existing technology cannot timely discover the network coverage edge area through manual complaint and manual analysis identification method.

[0006] The present application provides a network edge coverage area discovery method, comprising:

[0007] Based on the current index items of each to-be-identified cell in the current period, a first edge coverage cell set in the current period is identified from each to-be-identified cell in three dimensions, the three dimensions being: coverage dimension, capacity dimension, and user perception dimension;

[0008] The current index items are input into an edge coverage cell prediction model to obtain future index items in a future period output by the edge coverage cell prediction model;

[0009] identify a second edge coverage cell set in the future period from each to-be-identified cell in the three dimensions based on the future indicator item;

[0010] Take the intersection of the first edge coverage cell set and the second edge coverage cell set, and determine an edge coverage area based on the edge coverage cells in the intersection;

[0011] The edge coverage cell prediction model is trained based on a first indicator item in a first time period in a historical period as a sample and a second indicator item in a second time period in the historical period as a label, and the first time period is earlier than the second time period.

[0012] According to the network edge coverage area discovery method provided by the application, a first edge coverage cell set in a current period is identified from each to-be-identified cell in three dimensions based on a current indicator item of each to-be-identified cell in the current period, and the method comprises the following steps:

[0013] In the coverage dimension, based on the current indicator items related to MR, TA and network disconnection, a coverage degradation cell in each to-be-identified cell is determined, the coverage degradation cell is added to the first edge coverage cell set, and a third priority is given;

[0014] In the capacity dimension, based on the current indicator items related to traffic, user number and PRB utilization rate, a value edge cell in the coverage degradation cell is determined, and a second priority is given to the value edge cell, the value edge cell being a cell in the coverage degradation cell that has a growth trend in at least one of traffic, user number and PRB utilization rate;

[0015] In the user perception dimension, for a value edge cell that meets the complaint, capacity perception inflection point or uplink and downlink edge quality difference rule, a first priority is given.

[0016] According to the network edge coverage area discovery method provided by the application, in the coverage dimension, based on the current indicator items related to MR, TA and network disconnection, a coverage degradation cell in each to-be-identified cell is determined, and the method comprises the following steps:

[0017] MR: if MR daily sampling points are met and any one of MR indicator thresholds of MR indicator items is met, it is determined to be yes;

[0018] TA: if any one of TA indicator thresholds of TA indicator items is met, it is determined to be yes;

[0019] Network disconnection: if the number of attachments and the number of detachments are met at the same time, it is determined to be yes;

[0020] For any to-be-identified cell, if any one of the corresponding MR, TA and network disconnection sub-dimensions is determined to be yes, the any to-be-identified cell is determined to be a coverage degradation cell.

[0021] The network edge coverage area discovery method provided by the application determines the value edge cell in the coverage deterioration cell based on the current index related to the traffic, the number of users and the PRB utilization rate in the capacity dimension, and comprises the following steps:

[0022] Traffic: if the index threshold of the daily average traffic and the daily average traffic increase is met at the same time, it is determined to be yes;

[0023] Number of users: if the index threshold of the maximum number of RRC users and the daily average user increase is met at the same time, it is determined to be yes;

[0024] PRB utilization rate: if the index of the busy time PRB utilization rate and the daily average utilization rate increase is met at the same time, it is determined to be yes;

[0025] For any coverage deterioration cell, if any of the corresponding traffic, the number of users and the PRB utilization rate is determined to be yes, the any coverage deterioration cell is determined to be a value edge cell.

[0026] The network edge coverage area discovery method provided by the application, the edge coverage cell prediction model is a long-term and short-term time sequence network model.

[0027] The network edge coverage area discovery method provided by the application determines the edge coverage area based on the edge coverage cells in the intersection, and comprises the following steps:

[0028] Based on the edge coverage cells in the intersection, the three-line positioning method is used to determine the edge coverage area, and the TA average distance index of the coverage dimension is used as the vector line segment length variable of the three-line positioning method.

[0029] The network edge coverage area discovery method provided by the application determines the edge coverage area based on the edge coverage cells in the intersection, and comprises the following steps:

[0030] The associated sectors of the base stations of the edge coverage cells in the intersection are determined, the starting longitude and latitude, the azimuth angle and the length of each associated sector are obtained, and the length is the TA average distance.

[0031] The starting longitude and latitude, the azimuth angle and the TA average distance are used to generate a vector line segment with a starting point, a direction and a length.

[0032] If there are three or more associated sectors, then a plurality of sectors are drawn by the vector line segment corresponding to each associated sector with a radius, and a grid area with a preset length associated with the overlapping area of the plurality of sectors is determined as the edge coverage area.

[0033] In the case of two associated sectors, two sectors are drawn by a vector line segment corresponding to each associated sector as a radius, and a preset length of a grid area in an overlapping area of the two sectors is determined as the edge coverage area.

[0034] In the case of one associated sector, a terminal position of a vector line segment corresponding to the associated sector is determined, and a grid area in which the terminal position is located is determined as the edge coverage area.

[0035] The application further provides a network edge coverage area discovery device, comprising:

[0036] The first identification module is configured to identify, from each of the to-be-identified cells, a first edge coverage cell set in a current period in three dimensions based on current index items of the to-be-identified cells in the current period, the three dimensions being coverage dimension, capacity dimension and user perception dimension.

[0037] The model prediction module is configured to input the current index items into an edge coverage cell prediction model to obtain future index items in a future period output by the edge coverage cell prediction model.

[0038] The second identification module is configured to identify, from each of the to-be-identified cells, a second edge coverage cell set in the future period in the three dimensions based on the future index items.

[0039] The edge coverage area determination module is configured to take an intersection of the first edge coverage cell set and the second edge coverage cell set, and determine an edge coverage area based on edge coverage cells in the intersection.

[0040] The edge coverage cell prediction model is trained by taking first index items in a first time period in a historical period as samples and second index items in a second time period in the historical period as labels, the first time period being earlier than the second time period.

[0041] The application further provides an electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the network edge coverage area discovery method according to any of the above when executing the computer program.

[0042] The application further provides a non-transitory computer readable storage medium having a computer program stored thereon, the computer program being executable by a processor to implement the network edge coverage area discovery method according to any of the above.

[0043] The application further provides a computer program product comprising a computer program, the computer program being executable by a processor to implement the network edge coverage area discovery method according to any of the above.

[0044] The network edge coverage area discovery method and device provided by the application can identify the first edge coverage cell set in the current period from each to-be-identified cell in three dimensions of coverage dimension, capacity dimension and user perception dimension, input the current index item into an edge coverage cell prediction model to obtain a future index item in a future period output by the edge coverage cell prediction model, identify the second edge coverage cell set in the future period from each to-be-identified cell in the three dimensions based on the future index item, take the intersection of the first edge coverage cell set and the second edge coverage cell set, and determine the edge coverage area based on the edge coverage cells in the intersection. Since the edge coverage cell is identified in three dimensions, the first edge coverage cell set in the current period can be more accurately identified, the future index item in the future period is predicted based on the current index item corresponding to the current period, the edge coverage cell is identified in three dimensions based on the future index item, the second edge coverage cell set in the future period is identified, and the edge coverage area is determined based on the edge coverage cells in the two sets, so that the edge coverage area in the network can be discovered in time. BRIEF DESCRIPTION OF DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0046] Figure 1 FIG. 1 is a flowchart of the network edge coverage area discovery method provided by the application;

[0047] Figure 2 FIG. 2 is a structural diagram of the network edge coverage area discovery device provided by the application;

[0048] Figure 3 FIG. 3 is a schematic diagram of the three-line positioning method in the network edge coverage area discovery method provided by the application;

[0049] Figure 4 FIG. 4 is a structural diagram of the electronic device provided by the application. DETAILED DESCRIPTION

[0050] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be clearly and completely described below with reference to the drawings in the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0051] The network edge coverage area discovery method of the embodiment of the present application, as shown in the figure, comprises: Figure 1

[0052] Step S110: Based on the current index item of each to-be-identified cell in the current period, a first edge coverage cell set in the current period is identified from each to-be-identified cell in three dimensions, i.e. coverage dimension, capacity dimension and user perception dimension. The current period is defined as a time period of a preset length (e.g. one week) before the current time. In this embodiment, the index item can be divided into index items of coverage dimension, capacity dimension and user perception dimension, and the corresponding index item and index threshold or judgment rule can be set according to the actual network situation. When the index item corresponding to the to-be-identified cell meets the index threshold or judgment rule, it is considered that the to-be-identified cell is the first edge coverage cell.

[0053] Step S120: The current index item is input into the edge coverage cell prediction model to obtain a future index item in a future period output by the edge coverage cell prediction model. The future period is defined as a time period of the preset length (e.g. one week) after the current time. The edge coverage cell prediction model is trained with a first index item in a first time period in a historical period as a sample and a second index item in a second time period in the historical period as a label, and the first time period is earlier than the second time period. For example, taking a one-month historical period as an example, each week is a time period. Based on the index items in this one month, the edge coverage cell prediction model is trained with the index item of the previous week as a sample and the index item of the next week as a label, so as to obtain the trained edge coverage cell prediction model.

[0054] Step S130: Based on the future index item, a second edge coverage cell set in the future period is identified from each to-be-identified cell in the three dimensions. In this step, the identification manner is basically the same as that of step S110.

[0055] Step S140: The intersection of the first edge coverage cell set and the second edge coverage cell set is taken, and based on the edge coverage cells in the intersection, an edge coverage area is determined. Specifically, the edge coverage area can be determined according to the three-line positioning method.

[0056] ​In the network edge coverage area discovery method of the embodiment, the first edge coverage cell set in the current period can be more accurately identified due to the edge coverage cell identification in three dimensions, and the future index item in the future period is predicted by the model according to the current index item corresponding to the current period, the edge coverage cell in the future period is identified based on the future index item in three dimensions, and the second edge coverage cell set in the future period is identified, and the edge coverage area is determined through the common edge coverage cell in the two sets, so that the edge coverage area in the network can be discovered in time. Moreover, the intersection of the first edge coverage cell set and the second edge coverage cell set is taken, so as to avoid that a cell is mistakenly considered as an edge coverage cell due to a sudden situation in the current period, and only the cell that is identified as an edge coverage cell in the current period and the future period is considered as an edge coverage cell, thereby improving the accuracy of identification.

[0057] In some embodiments, step S110 comprises:

[0058] In the coverage dimension, the coverage degradation cell in each to-be-identified cell is determined based on the current index item related to MR, TA (time advanced) and off-network, the coverage degradation cell is added to the first edge coverage cell set, and the third priority is given;

[0059] In the capacity dimension, the value edge cell in the coverage degradation cell is determined based on the current index item related to traffic, user number and PRB (Physical Resource Block, physical resource block) utilization rate, the value edge cell is given the second priority, and the value edge cell is a cell in the coverage degradation cell that has a growth trend in at least one of traffic, user number and PRB utilization rate;

[0060] In the user perception dimension, the value edge cell that meets the complaint, capacity perception inflection point or uplink and downlink edge quality difference rule is given the first priority.

[0061] With the change of network environment, the third priority (i.e. low priority) area is gradually transferred to medium and high priority after triggering the conditions of capacity dimension and user perception dimension subsequently, and through the division of priority, it is helpful to better allocate resources to the edge coverage area of the network and preferentially process the edge coverage area with high priority.

[0062] Specifically, the index items and standard thresholds or judgment rules of the three dimensions are shown in Tables 1-4.

[0063] Table 1: Index items and standard thresholds related to coverage dimension

[0064]

[0065]

[0066] Based on the current index items of MR, TA and off-network, determine the coverage degradation cell in each to-be-identified cell, specifically including:

[0067] MR: If the MR daily sampling points are met, and any one of the MR index thresholds of the MR index items is met, it is judged to be yes.

[0068] TA: If any one of the TA index thresholds of the TA index items is met, it is judged to be yes.

[0069] Off-network: If the number of attachments and the number of detachments are met at the same time, it is judged to be yes.

[0070] For any to-be-identified cell, if any one of the corresponding MR, TA and off-network is judged to be yes, it is determined that the any to-be-identified cell is a coverage degradation cell.

[0071] Table 2: Index items and standard thresholds related to capacity dimension

[0072]

[0073] Under the capacity dimension, based on the current index items related to traffic, number of users and PRB utilization rate, determine the value edge cell in the coverage degradation cell, including:

[0074] Traffic: If the index thresholds of daily traffic and daily traffic increase are met at the same time, it is judged to be yes.

[0075] Number of users: If the index thresholds of RRC maximum number of users and daily number of users increase are met at the same time, it is judged to be yes.

[0076] PRB utilization rate: If the indexes of busy time PRB utilization rate and daily utilization rate increase are met at the same time, it is judged to be yes.

[0077] For any coverage degradation cell, if any one of the corresponding traffic, number of users and PRB utilization rate is judged to be yes, it is determined that the any coverage degradation cell is a value edge cell.

[0078] Table 3: Index items and determination rules of capacity perception inflection point in user perception dimension

[0079]

[0080] Capacity perception inflection point: If the conditions of number of users, uplink / downlink utilization rate and busy time uplink / downlink traffic (GB) are met at the same time, it is judged to be yes.

[0081] Table 4: Edge quality perception determination rule in user perception dimension

[0082]

[0083] Edge quality difference: any one of the LTE / NR, uplink / downlink edge quality difference is judged to be yes.

[0084] In some embodiments, the edge coverage cell prediction model is a long-and short-term time-series network model (LSTNET). In this embodiment, due to the involvement of more variables (i.e., index items) and the complexity of the time series relationship, the training effect of the traditional machine learning model is limited, and the training of CNN and RNN has advantages and disadvantages. CNN has an advantage in short-term prediction, and RNN has an advantage in long-term prediction. The use of LSTNET model can effectively solve the problem of multiple input variables and complex time series relationship. The input and output of the LSTNET model are shown in Table 5 below:

[0085] Table 5 Input and output of LSTNET model

[0086]

[0087] It should be noted that the index items of the input and output of the edge coverage cell prediction model are slightly different from those in Tables 1-4. The index items in Tables 1-4 that are not directly collected or can be calculated by directly collecting the index items are not used as input and output of the model, for example: amplitude data, proportion data. MR weak coverage sampling points are not used as output index items because the number of MR weak coverage sampling points will not change in a short period of time. Moreover, in the above step S130, for the index items not directly output by the model, the second edge coverage cell set in the future period can be identified according to the specific identification steps in step S110 after calculating the related index items.

[0088] The AI prediction algorithm based on the LSTNET model mainly includes the following steps:

[0089] Step 1: Initialize the training set

[0090] Select the edge perception model key indicators as the input end, and the variables include the index items MR coverage rate, MR weak coverage sampling point, TA average distance, attach / detach times, SA time length residence ratio, daily traffic, uplink / downlink busy time PRB utilization rate, RRC maximum user number, and uplink / downlink low rate ratio. The time series is selected as the day busy time, the MR coverage rate and the TA average distance are 7 days polling, and the other index items are day granularity.

[0091] Step 2: Establish the LSTNET model

[0092] Long-Short Term Network for Time Series (LSTNET) is a deep learning framework designed for multivariate time series forecasting, which can effectively solve the mixed problem of long and short term patterns. LSTNET uses convolutional neural network (CNN) and recurrent neural network (RNN) to extract short-term local dependence patterns between variables and discover long-term patterns of time series trends, while using traditional autoregressive models to solve the scale insensitivity problem of neural network models. The input is the training set data, and the output is the coverage and capacity-aware data.

[0093] The LSTNET model designs four components: convolution component, recurrent component, recurrent-jump component, and time attention natural regression component. Through layer-by-layer recursion, the complete prediction function is finally obtained.

[0094] Step 3: Data training

[0095] Set the convolution component (first layer): this component is used to extract short-term features in the time dimension (the input part of the prediction model, i.e., MR coverage, MR weak coverage sampling points, TA average distance, attach / detach times, SA time length residence ratio, daily traffic, uplink and downlink busy time PRB utilization rate, RRC maximum user number, uplink and downlink low rate ratio), and extract local dependence between variables. The convolution layer contains multiple filters, the filter width is w, the height is n (the height is set to be the same as the number of variables), k represents the kth filter, and X represents the input feature value matrix, such as the dependence of busy time uplink and downlink PRB utilization rate and uplink and downlink low rate ratio. k represents the convolution operation value of the processed feature, which is calculated from the scanned input matrix X. Here * represents the convolution operation, and the activation function is RELU(x) = max(0, x). The output of the convolution layer k will be transmitted to the recurrent component and the recurrent-jump component at the same time.

[0096] h k = RELU(W k *X + b k )

[0097] where W k represents the feature value matrix of the kth filter, and b k represents the hyperparameter.

[0098] Set the recurrent component (second layer): the recurrent component is a recurrent layer with gated recurrent unit GRU, using RELU function as the hidden layer update activation function. Θ is an element-wise (bit multiplication) operation, and σ is a sigmoid activation function. x t is the input of this layer at time t, and w represents the weight.

[0099] w of any index xr in the weight feature matrix at time point t (e.g. MR coverage) xr Very low, the value after sigmoid is 0, reset gate γ t This feature at time t will be discarded. Conversely, if the weight w of any index xu in the weight feature matrix (e.g. PRB utilization rate when uplink and downlink are busy) xu Very large, the value after sigmoid is 1, then it becomes update gate to keep this feature at time t. Finally, by reset gate γ t , update gate u t , temporary hidden layer state C t Common output final hidden layer state h t (neural network feature operation value of a certain hidden layer).

[0100] Reset gate: γ t = σ(x t w xr +h t-1 w hr +b r )

[0101] Update gate: u t = σ(x t w xu +h t-1 w hu +b u )

[0102] Temporary hidden layer state: C t = RELU(x t w xc + γ t Θ(h t-1 w hc )+b c )

[0103] Final hidden layer state: h t = (1-u t )Θh t-1 +u t ΘC t

[0104] Where b r , b u and b c represent hyperparameters, xc represents any index in the weight feature matrix, w xc represents the weight of xc, w hr , w hr and w hc represent the final hidden layer state h t-1weights. Due to the limited time span of t and t-1, the recurrent component can only handle short-term dependencies. If the time span is large, such as the feature correlation at time t-p, the recursive component cannot effectively train. In order to extract relatively long-term dependencies, a recursive structure with time jump connection is needed, that is, the recursive-jump component.

[0105] Set the recursive-jump component (the third layer): replace t-1 with t-p for capturing long-term dependencies as the input and output of the recurrent component, p is the time length of interval sampling, and the final output is the final hidden layer state h t . For example, in the feature matrix, xu represents the MR coverage rate, and the change period is long. By associating t and t-p, it is retained to the update gate for output h t (neural network feature operation value of a hidden layer).

[0106] Reset gate: γ t = σ(x t w xr +h t-P w hr +b r )

[0107] Update gate: u t = σ(x t w xu +h t-p w hu +b u )

[0108] Temporary hidden layer state: C t = RELU(x t w xc + γ t Θ(h t-P w hc )+b c )

[0109] Final hidden layer state: h t =(1-u t )Θh t-p +u t ΘC t

[0110] The main function of the recurrent component and the recursive-jump component is to process the nonlinear part of the LSTNET. The final nonlinear part target function is represented as In the following formula, "+" on the left side is the recurrent component part, and the right side is the recursive-jump component part. W R is the weight of the recurrent part, is the calculation result value of the recurrent part; Wi s is the weight of the recursive jump part, is the calculated result value of the recurrent skip-connection part, p is the interval sample length, t represents the time point of the output prediction result, i represents the most relevant time point, b is a hyperparameter, and superscripts R and S are used to distinguish different components.

[0111]

[0112] Set time attention layer: used to solve the weight W in the above step R and Wi s , which learns the weighted combination of the hidden representations of each window position of the input matrix. Specifically, the attention weight w t (the value dimension is q*1) at the current timestamp t can be calculated as follows:

[0113]

[0114] Autoregressive component (fourth layer): the linear prediction part of the LSTNET model, in the LSTNET architecture, a classic autoregressive (AR) model is used as the linear component. The prediction result of the AR component is denoted as (the predicted value of feature variable i at time t); the weight coefficient of the AR model is denoted as q ar , which is the size of the input window on the input matrix, y t-k,i , which is the value of the i-th feature variable at time t-k, b ar is a hyperparameter. The essence of autoregression is to predict y at time t by accumulating y at the previous q ar -1 time points through the weight coefficient , that is

[0115]

[0116] Finally, the target prediction function is obtained as follows: is the nonlinear part of LSTnet, which is obtained by the recurrent component and the recurrent-skip component, is the linear part of LSTnet, which is obtained by the autoregressive component. The final target prediction function is:

[0117] Step 4: Hyperparameter optimization

[0118] Although LSTNET can effectively solve the coverage characteristics of long-term and short-term time series, due to the difference in multi-dimensional feature statistics in the communication network, the dynamic change and trend characteristics of the indicators are quite different, and the dynamic period length coverage characteristics are difficult to extract through the traditional LSTNET. There are many key parameters in the LSTNET model, such as neurons, convolution layers, and kernel parameters, which can only be set reasonably to ensure the fitting effect of the model with sufficient precision. The prediction results of the parameters selected by manual time delay or expert experience are difficult to guarantee the stability of the precision, especially in the scene with large periodic dynamic characteristics of multi-dimensional features. Therefore, in the process of constructing LSTNET, this paper introduces the artificial intelligence method to select the model parameters, and realizes the selection of optimal parameters through the particle swarm algorithm. In the combination of multiple parameters, the parameter combination suitable for the model is selected for iterative training.

[0119] Considering that the multi-element coverage sequence variables have strong coupling, and each feature has an accurate periodicity of different time granularity and different time length, the prediction result ytrain corresponding to the parameters and the true value ytrue in the training set can be obtained through the default parameter configuration of the model. At this time, the error of the model on the training set should be E=ytrue-ytrain.

[0120] The error sequence and feature variable data (ytrue) are superimposed on the results of different hyperparameter configurations of the particle swarm to perform training. The particle swarm algorithm is an optimization algorithm that searches for the optimal solution in the particle search space. A set of random particles is randomly selected in the solution space, and each particle represents a solution. The pros and cons of each solution can be calculated by the fitness function. Each initial particle has a random initial velocity vector. In the iteration process, the position of the particle is constantly updated with the velocity vector, and the velocity vector is constantly updated in the direction of the current optimal particle. Finally, the parameter configuration that makes the error E optimal (the loss function square absolute error MAE is close to 0 for optimal, and MAE is less than 5 for small error) is determined.

[0121] Step 5: Data prediction

[0122] Based on the big data AI learning algorithm, the LSTNET neural network machine learning algorithm is used to continuously improve the accuracy of the model, solve the problem of hysteresis, and automatically identify the perception needs and value of edge users in combination with the edge perception model.

[0123] In some embodiments, in step S140, the edge coverage area is determined based on the edge coverage cells in the intersection, including: based on the edge coverage cells in the intersection, the edge coverage area is determined by using a three-line positioning method, and the TA average distance index of the coverage dimension is used as the vector line segment length variable of the three-line positioning method. In this embodiment, the TA average distance is introduced as the vector line segment length in the three-line positioning method, which is equivalent to considering the actual coverage distance of the site of the edge coverage cell, which is more close to the user position of the edge coverage area. In the three-line positioning method, the central latitude and longitude is not calculated by taking the average, but the latitude and longitude is taken as the starting point, the azimuth angle is taken as the direction, and the TA average distance is taken as the length, to form a vector line segment. The central latitude and longitude is obtained by taking the central latitude and longitude of a plurality of vector line segments, so that the accuracy of the position aggregation of the edge coverage area is obviously improved. In practical application, a Python function is encapsulated into a runtime library to avoid tedious code and greatly improve the running efficiency. In actual use, only the calculate_center function (parameters: longitude, latitude, azimuth angle, TA average distance) needs to be directly called and the calculation result can be batch generated.

[0124] As shown in Figure 3 TA "three-line positioning method" determines the area position, according to the latitude and longitude (as the starting point) of the associated sector, the azimuth angle (as the vector azimuth angle), and the TA average distance (as the length) to form three straight line segments, and use the calculate_center function to determine the central latitude and longitude of the three line segments; according to the central latitude and longitude, match the grid layer, and output the grid number corresponding to the region.

[0125] Azimuth angle: the current base station generally uses three-face antenna (sector) coverage, and the coverage angle of the antenna is called azimuth angle, and the angle value is the included angle degree number from the north to the coverage direction. TA average distance: refers to the average distance from the user sampling point to the base station, which is calculated by the time advance, and the unit of measurement is meter;

[0126] In this embodiment, the three-line positioning method step includes:

[0127] The associated sectors of the base stations of the edge coverage cells in the intersection are determined, and the starting point latitude and longitude, azimuth angle and length of each associated sector are obtained, and the length is the TA average distance. Specifically, the number of edge coverage cells in the intersection may be five, of which the sectors of three edge coverage cell base stations are associated sectors, and the sectors of the other two edge coverage cell base stations are associated sectors. At this time, two edge coverage areas can be finally determined. Whether it is an associated sector is determined according to the current azimuth angle of the edge coverage cell base station antenna. If the difference between the azimuth angles of the sectors of the two base stations is greater than 90 degrees and less than 270 degrees, it can be considered as an associated sector.

[0128] A vector line segment with a start point, a direction and a length is generated from the start point latitude and longitude, the azimuth angle and the TA average distance.

[0129] In the case of three or more associated sectors, a plurality of sectors are drawn by the vector line segment corresponding to each associated sector as a radius, and a preset length grid area in an overlapping area of the plurality of sectors is determined as the edge coverage area. The grid is a preset grid in urban planning, and the length of a single grid is usually 50m*50m, as shown in Figure 3 The grid area of all grids occupied by the overlapping area of the three sectors (as shown by the shaded area in Figure 3 ) is determined as the edge coverage area. Figure 3

[0130] In the case of two associated sectors, two sectors are drawn by the vector line segment corresponding to each associated sector as a radius, and a preset length grid area in an overlapping area of the two sectors is determined as the edge coverage area.

[0131] In the case of one associated sector, the end point position of the vector line segment corresponding to the associated sector is determined, and a grid area in which the end point position is located is determined as the edge coverage area. The size of the grid area is usually the size of a single grid.

[0132] The network edge coverage area discovery device provided by the application is described below. The network edge coverage area discovery device described below can be referred to in correspondence with the network edge coverage area discovery method described above.

[0133] As shown in Figure 2 , the network edge coverage area discovery device of the application comprises:

[0134] The first identification module 210 is configured to identify a first edge coverage cell set in a current period from each to-be-identified cell in three dimensions based on a current index item of each to-be-identified cell in the current period, the three dimensions being coverage dimension, capacity dimension and user perception dimension.

[0135] The model prediction module 220 is configured to input the current index item into an edge coverage cell prediction model to obtain a future index item in a future period output by the edge coverage cell prediction model.

[0136] The second identification module 230 is configured to identify a second edge coverage cell set in the future period from each to-be-identified cell in the three dimensions based on the future index item.

[0137] ​The edge coverage area determination module 240 is configured to determine the intersection of the first edge coverage cell set and the second edge coverage cell set, and determine an edge coverage area based on edge coverage cells in the intersection.

[0138] The edge coverage cell prediction model is trained by taking a first index item in a first time period in a historical time period as a sample and a second index item in a second time period in the historical time period as a label.

[0139] In the network edge coverage area discovery device, the first edge coverage cell set in the current time period can be accurately identified in three dimensions, the future index item in the future time period is predicted by using the model according to the current index item corresponding to the current time period, the second edge coverage cell set in the future time period is identified in three dimensions based on the future index item, and the edge coverage area is determined by the edge coverage cells in the two sets, so that the edge coverage area in the network can be discovered in time.

[0140] Optionally, the first identification module 210 comprises:

[0141] The coverage deterioration identification module is configured to determine, in the coverage dimension, a coverage deterioration cell in each to-be-identified cell based on the current index item related to MR, TA and network disconnection, add the coverage deterioration cell to the first edge coverage cell set, and assign a third priority to the coverage deterioration cell.

[0142] The capacity demand identification module is configured to determine, in the capacity dimension, a value edge cell in the coverage deterioration cell based on the current index item related to traffic, the number of users and PRB utilization, and assign a second priority to the value edge cell, the value edge cell being a cell in the coverage deterioration cell that has a growth trend in at least one of traffic, the number of users and PRB utilization.

[0143] The user perception identification module is configured to, in the user perception dimension, assign a first priority to the value edge cell that meets the complaint, the capacity perception inflection point or the uplink and downlink edge quality difference rule.

[0144] Optionally, the coverage deterioration identification module is specifically configured to:

[0145] MR: if a daily sampling point of MR is met and any one of MR index thresholds of the MR index item is met, it is determined to be yes.

[0146] TA: if any one of TA index thresholds of the TA index item is met, it is determined to be yes.

[0147] Network disconnection: if the number of attachments and the number of detachments are met at the same time, it is determined to be yes.

[0148] For any to-be-identified cell, if the corresponding MR, TA and off-network any one sub-dimension is judged as yes, then the any to-be-identified cell is determined as a coverage degradation cell.

[0149] Optionally, the capacity demand identification module is specifically configured to:

[0150] Traffic: the index thresholds of daily average traffic and daily average traffic increase are met at the same time, and then judged as yes;

[0151] User number: the index thresholds of RRC maximum user number and daily average user number increase are met at the same time, and then judged as yes;

[0152] Utilization: the indexes of busy time PRB utilization and daily average utilization increase are met at the same time, and then judged as yes;

[0153] For any coverage degradation cell, if the corresponding traffic, user number and PRB utilization any one sub-dimension is judged as yes, then the any coverage degradation cell is determined as a value edge cell.

[0154] Optionally, the edge coverage cell prediction model is a long-term and short-term time series network model.

[0155] Optionally, the edge coverage area determination module is specifically configured to determine the edge coverage area based on the edge coverage cells in the intersection, and use the three-line positioning method, and the TA average distance index of the coverage dimension as the vector line segment length variable of the three-line positioning method.

[0156] Optionally, the edge coverage area determination module is specifically configured to: determine the associated sectors of the base stations of the edge coverage cells in the intersection, obtain the start longitude and latitude, azimuth angle and length of each associated sector, the length being the TA average distance; generate a vector line segment with a start point, a direction and a length from the start longitude and latitude, the azimuth angle and the TA average distance; in the case of three or more associated sectors, then a plurality of sectors are drawn by taking each associated sector corresponding vector line segment as a radius, and a preset length of the grid area associated with the overlapping area of the plurality of sectors is determined as the edge coverage area; in the case of two associated sectors, then two sectors are drawn by taking each associated sector corresponding vector line segment as a radius, and a preset length of the grid area associated with the overlapping area of the two sectors is determined as the edge coverage area; in the case of one associated sector, the end point position of the vector line segment corresponding to the one associated sector is determined, and the grid area where the end point position is located is determined as the edge coverage area.

[0157] Figure 4 An example of an entity structure schematic diagram of an electronic device is shown as Figure 4As shown, the electronic device can include a processor 410, a communications interface 420, a memory 430, and a communications bus 440, wherein the processor 410, the communications interface 420, and the memory 430 complete mutual communication through the communications bus 440. The processor 410 can invoke a logical instruction in the memory 430 to execute a network edge coverage area discovery method, which includes:

[0158] Based on the current indicator item of each to-be-identified cell in the current period, a first edge coverage cell set in the current period is identified from each to-be-identified cell in three dimensions, which are coverage dimension, capacity dimension, and user perception dimension.

[0159] The current indicator item is input into an edge coverage cell prediction model to obtain a future indicator item in a future period output by the edge coverage cell prediction model.

[0160] Based on the future indicator item, a second edge coverage cell set in the future period is identified from each to-be-identified cell in the three dimensions.

[0161] An intersection of the first edge coverage cell set and the second edge coverage cell set is taken, and based on the edge coverage cells in the intersection, an edge coverage area is determined.

[0162] The edge coverage cell prediction model takes a first indicator item in a first time period in a historical period as a sample and a second indicator item in a second time period in the historical period as a label to be trained, and the first time period is earlier than the second time period.

[0163] In addition, the logical instruction in the memory 430 described above can be implemented in the form of a software function unit and sold or used as an independent product, which can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in various embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0164] In another aspect, the present application also provides a computer program product, which comprises a computer program, the computer program being stored in a non-transitory computer-readable storage medium, and the computer program being executable by a processor to enable the computer to perform the network edge coverage area discovery method provided by the above-mentioned method, and the method comprises the following steps of:

[0165] Based on the current indicator item of each to-be-identified cell in the current period, a first edge coverage cell set in the current period is identified from each to-be-identified cell in three dimensions, the three dimensions being: coverage dimension, capacity dimension and user perception dimension.

[0166] The current indicator item is input into an edge coverage cell prediction model to obtain a future indicator item in a future period output by the edge coverage cell prediction model.

[0167] Based on the future indicator item, a second edge coverage cell set in the future period is identified from each to-be-identified cell in the three dimensions.

[0168] An intersection of the first edge coverage cell set and the second edge coverage cell set is taken, and an edge coverage area is determined based on the edge coverage cells in the intersection.

[0169] The edge coverage cell prediction model is trained by taking a first indicator item in a first time period in a historical period as a sample and a second indicator item in a second time period in the historical period as a label, the first time period being earlier than the second time period.

[0170] In another aspect, the present application also provides a non-transitory computer-readable storage medium, which stores a computer program, and the computer program is executable by a processor to implement the network edge coverage area discovery method provided by the above-mentioned method, and the method comprises the following steps of:

[0171] Based on the current indicator item of each to-be-identified cell in the current period, a first edge coverage cell set in the current period is identified from each to-be-identified cell in three dimensions, the three dimensions being: coverage dimension, capacity dimension and user perception dimension.

[0172] The current indicator item is input into an edge coverage cell prediction model to obtain a future indicator item in a future period output by the edge coverage cell prediction model.

[0173] Based on the future indicator item, a second edge coverage cell set in the future period is identified from each to-be-identified cell in the three dimensions.

[0174] An intersection of the first edge coverage cell set and the second edge coverage cell set is taken, and an edge coverage area is determined based on the edge coverage cells in the intersection.

[0175] The edge coverage cell prediction model is trained by taking a first index item in a first time period in a historical time period as a sample and a second index item in a second time period in the historical time period as a label.

[0176] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0177] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software plus necessary universal hardware platforms, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of software products, which can be stored in a computer readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and include a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0178] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A network edge coverage area discovery method, characterized by, The method comprises the following steps: based on the current indicator of each to-be-identified cell in the current period, a first edge coverage cell set in the current period is identified from each to-be-identified cell in three dimensions, the three dimensions being: coverage dimension, capacity dimension and user perception dimension; the current indicator is input into an edge coverage cell prediction model to obtain a future indicator in a future period output by the edge coverage cell prediction model; based on the future indicator, a second edge coverage cell set in the future period is identified from each to-be-identified cell in the three dimensions; the intersection of the first edge coverage cell set and the second edge coverage cell set is taken, and an edge coverage area is determined based on the edge coverage cells in the intersection; wherein the edge coverage cell prediction model is trained by taking a first indicator in a first time period in a historical period as a sample and a second indicator in a second time period in the historical period as a label, the first time period being earlier than the second time period; based on the edge coverage cells in the intersection, the edge coverage area is determined, comprising: based on the edge coverage cells in the intersection, the edge coverage area is determined by using a three-line positioning method, and the TA average distance indicator of the coverage dimension is taken as a vector line segment length variable of the three-line positioning method, the vector line segment being a vector line segment with a starting point, a direction and a length generated by the starting longitude and latitude, the azimuth angle and the TA average distance of the associated sector of each edge coverage cell base station in the intersection.

2. The network edge coverage area discovery method of claim 1, wherein, based on the current indicator of each to-be-identified cell in the current period, a first edge coverage cell set in the current period is identified from each to-be-identified cell in three dimensions, comprising: in the coverage dimension, based on the current indicators related to MR, TA and network disconnection, the coverage degradation cells in each to-be-identified cell are determined, the coverage degradation cells are added to the first edge coverage cell set, and the third priority is given; in the capacity dimension, based on the current indicators related to traffic, user number and PRB utilization rate, the value edge cells in the coverage degradation cells are determined, and the value edge cells are given the second priority, the value edge cells being the cells in the coverage degradation cells that have a growth trend in at least one of traffic, user number and PRB utilization rate; in the user perception dimension, for the value edge cells that meet the complaint, capacity perception inflection point or uplink and downlink edge quality difference rules, the first priority is given.

3. The network edge coverage area discovery method of claim 2, wherein, in the coverage dimension, based on the current indicators related to MR, TA and network disconnection, the coverage degradation cells in each to-be-identified cell are determined, comprising: MR: if MR daily sampling points are met and any one of MR indicator thresholds of MR indicators is met, it is judged to be yes; TA: if any one of TA indicator thresholds of TA indicators is met, it is judged to be yes; network disconnection: if the number of attachments and the number of detachments are met at the same time, it is judged to be yes; for any to-be-identified cell, if any one of the corresponding MR, TA and network disconnection sub-dimensions is judged to be yes, the any to-be-identified cell is determined to be a coverage degradation cell.

4. The network edge coverage area discovery method of claim 2, wherein, In the capacity dimension, based on the current index items related to traffic, number of users and PRB utilization, the value edge cell in the coverage degradation cell is determined, including: Traffic: if the index thresholds of daily average traffic and daily average traffic increase are met, it is determined to be yes; Number of users: if the index thresholds of RRC maximum number of users and daily average number of users increase are met, it is determined to be yes; PRB utilization: if the indexes of busy time PRB utilization and daily average utilization increase are met, it is determined to be yes; For any coverage degradation cell, if any of the sub-dimensions of traffic, number of users and PRB utilization is determined to be yes, the any coverage degradation cell is determined to be a value edge cell.

5. The network edge coverage area discovery method of claim 1, wherein, The edge coverage cell prediction model is a long-term and short-term time series network model.

6. The network edge coverage area discovery method of claim 1, wherein, Based on the edge coverage cells in the intersection, the edge coverage area is determined by using the three-line positioning method, including: Determine the associated sectors of the base stations of each edge coverage cell in the intersection, obtain the starting longitude and latitude, azimuth and length of each associated sector, and the length is the TA average distance; Generate a vector line segment with a starting point, a direction and a length from the starting longitude and latitude, the azimuth and the TA average distance; In the case of three or more associated sectors, a plurality of sectors are drawn by taking each associated sector as a radius, and a grid area with a preset length associated with the overlapping area of the plurality of sectors is determined as the edge coverage area; In the case of two associated sectors, two sectors are drawn by taking each associated sector as a radius, and a grid area with a preset length associated with the overlapping area of the two sectors is determined as the edge coverage area; In the case of one associated sector, the endpoint position of the vector line segment corresponding to the one associated sector is determined, and the grid area where the endpoint position is located is determined as the edge coverage area.

7. A network edge coverage area discovery apparatus, characterized by Including: A first identification module is configured to identify a first edge coverage cell set in a current period from each to-be-identified cell in three dimensions based on current index items of each to-be-identified cell in the current period, the three dimensions being coverage dimension, capacity dimension and user perception dimension; A model prediction module is configured to input the current index items into an edge coverage cell prediction model to obtain future index items in a future period output by the edge coverage cell prediction model; A second identification module is configured to identify a second edge coverage cell set in the future period from each to-be-identified cell in the three dimensions based on the future index items; An edge coverage area determination module is configured to take an intersection of the first edge coverage cell set and the second edge coverage cell set, and determine an edge coverage area based on edge coverage cells in the intersection; The edge coverage cell prediction model is trained by taking first index items in a first time period in a historical period as samples and second index items in a second time period in the historical period as labels, and the first time period is earlier than the second time period. The edge coverage area determination module is specifically configured to determine the edge coverage area by using a three-line positioning method based on the edge coverage cells in the intersection, and a TA average distance index item of the coverage dimension is used as a vector line segment length variable of the three-line positioning method, and the vector line segment is a vector line segment with a starting point, a direction and a length, which is generated by the starting point longitude and latitude, the azimuth angle and the TA average distance of the associated sectors of the base stations of the edge coverage cells in the intersection.

8. An electronic device comprising: A memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the network edge coverage area discovery method according to any one of claims 1 to 6 when executing the computer program. 9.A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the network edge coverage area discovery method according to any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the network edge coverage area discovery method according to any one of claims 1 to 6. The computer program is executed by the processor to implement the network edge coverage area discovery method according to any one of claims 1 to 6.

Images