Cell function deployment method, device, medium, and program product

Abstract

Provided in the present application are a cell function deployment method, a device, a medium, and a program product. The method comprises: acquiring first feature data of a target cell, and a function policy model corresponding to a scenario of the target cell (S101); inputting the first feature data into the function policy model, so as to obtain predicted performance values corresponding to a plurality of function policies (S102); and on the basis of the predicted performance values, determining, from among the plurality of function policies, a target function policy to be deployed to the target cell (S103).

Description

Cell function deployment method, device, medium and program product

[0001] Cross-reference to related applications

[0002] The present application is based on the Chinese patent application No. 2024107373923, filed on June 6, 2024, and claims priority to the Chinese patent application No. 2024107373923, the entire contents of which are incorporated herein by reference. TECHNICAL FIELD

[0003] The present application relates to the field of communication technology, and in particular to a cell function deployment method, device, medium and program product. BACKGROUND

[0004] In a wireless communication system, the function deployment of a physical cell is often determined after the station is opened or after the version is upgraded. Except for a small part of functions with adaptive adjustment, most functions in different scenarios adopt the same configuration strategy. However, due to the influence of factors such as user tidal effect, idle / busy period conversion, and change of user channel statistical characteristics in the cell, the cell scene may change significantly, and the focus of cell performance observation in different scenes will also change. If a fixed function configuration strategy is adopted, the cell performance will be limited, resulting in a decrease in the spectral efficiency of the cell and a deterioration in user experience. SUMMARY

[0005] Embodiments of the present application provide a cell function deployment method, an electronic device, a computer readable storage medium and a computer program product.

[0006] In a first aspect, embodiments of the present application provide a method for determining a cell function strategy, comprising: obtaining first feature data of a target cell and a function strategy model corresponding to a scene of the target cell; inputting the first feature data into the function strategy model to obtain performance prediction values corresponding to a plurality of function strategies; and determining a target function strategy deployed to the target cell from the plurality of function strategies according to the performance prediction values.

[0007] In a second aspect, embodiments of the present application provide an electronic device, comprising: one or more processors; a memory having one or more programs stored thereon, when the one or more programs are executed by the one or more processors, the one or more processors implement the method for determining a cell function strategy as described in the first aspect of the present application.

[0008] In a third aspect, an embodiment of the present application provides a computer readable storage medium, characterized in that the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the method for determining a cell function policy according to the first aspect of the present application.

[0009] In a fourth aspect, an embodiment of the present application provides a computer program product, characterized in that the computer program product comprises a computer program, and the computer program is executed by a processor to implement the method for determining a cell function policy according to the first aspect of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0010] FIG. 1 is a flow diagram of a method for determining a cell function policy according to an embodiment of the present application;

[0011] FIG. 2 is a flow diagram of a method for determining a scenario of a target cell according to an embodiment of the present application;

[0012] FIG. 3 is a flow diagram of a method for training a scenario classification model according to an embodiment of the present application;

[0013] FIG. 4 is a flow diagram of a method for obtaining a second training set according to an embodiment of the present application;

[0014] FIG. 5 is a flow diagram of a method for training a function policy model according to an embodiment of the present application;

[0015] FIG. 6 is a flow diagram of a method for determining a cell function policy according to an embodiment of the present application;

[0016] FIG. 7 is a structural diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0017] In order for those skilled in the art to better understand the technical solutions of the present application, the technical solutions provided by the present application will be described in detail below with reference to the drawings.

[0018] In the following, example embodiments will be described more fully with reference to the accompanying drawings, in which example embodiments can be embodied in different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application will be thorough and complete, and fully convey the scope of the application to those skilled in the art.

[0019] As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0020] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0021] In the following description, reference is made to the accompanying drawings which form a part hereof, and in which are shown by way of illustration various embodiments for practicing the present application. It is to be understood that other embodiments can be utilized and structural or logical changes can be made without departing from the scope of the present application. The following detailed description, therefore, is not to be taken in a limiting sense, as the scope of the present application is defined by the appended claims.

[0022] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and the present application, and will not be interpreted in an overly literal or overly formal sense unless expressly so defined herein.

[0023] For the convenience of better understanding the solutions of the embodiments of the present application, the related art is introduced first as follows.

[0024] The function deployment of the physical cell of the wireless communication system mainly focuses on the following aspects:

[0025] (1) Physical cell identifier automatic configuration:

[0026] The physical cell identifier (PCI) is a basic parameter of the wireless network configuration, which is used to distinguish the wireless signals of different cells.

[0027] Due to the limited number of available PCIs, the PCI reuse will inevitably occur.

[0028] In the process of deploying a new base station, the physical cell identifier automatic configuration capability is needed to allocate a PCI for each cell controlled by the base station.

[0029] It is ensured that there is no same PCI in the coverage range of the related cells, and the PCIs adopted by the neighbor cells are different.

[0030] (2) Automatic neighbor relation configuration:

[0031] When the wireless network topology changes (such as a new base station is added / removed or a base station fails to work, etc.), the base station needs to be able to automatically maintain the neighbor relation with the surrounding base stations.

[0032] This is of great significance for inter-cell handover, load balancing, interference coordination, etc.

[0033] (3) Inter-cell interference coordination:

[0034] Due to the limited nature of wireless resources, especially frequency resources, inter-cell reuse of wireless resources is necessary.

[0035] When the wireless resources used by neighboring cells are the same, inter-cell interference will occur.

[0036] Inter-cell interference coordination is to coordinate the use of wireless resources by neighboring cells, and to avoid using the same frequency resources in the same time period as much as possible, or to allow neighboring cells to use the same frequency resources under acceptable interference conditions.

[0037] (4) Wireless resource management:

[0038] A physical cell is a unit object for wireless resource management, meeting two conditions:

[0039] The cell as a whole unit allocates channels to users, and users can only see one logical node providing services for them.

[0040] Regardless of where the user is in the area, the probability of each channel providing service to the user is equal among all available channels in the area.

[0041] (5) Service combination and function split:

[0042] When deploying a cell, the impact of service combination on function split and deployment also needs to be considered.

[0043] Different split methods can provide more optimization opportunities, such as split method B allows different services to use different coding techniques and allows joint decoding algorithms to effectively overcome interference.

[0044] (6) Deployment method and coverage:

[0045] Depending on different application scenarios (such as hotspots, stadiums, large shopping malls, and airports, etc.), the deployment method and coverage of the physical cell will be different.

[0046] For example, wide-area coverage using fiber deployment can enable all wireless access network functions to be deployed centrally, enabling co-located deployment with core network functions, and maximizing the cooperative diversity gain.

[0047] In summary, the functional deployment of a physical cell in a wireless communication system is a comprehensive process that needs to consider multiple aspects, including automatic configuration of physical cell identification, automatic neighbor relationship configuration, inter-cell interference coordination, radio resource management, service combination and functional split, and deployment mode and coverage, etc.

[0048] An Artificial Intelligence (AI) model refers to a model constructed using artificial intelligence technology. It can improve the accuracy of its predictions and decisions through learning and training. Specifically, an AI model is a mathematical model based on artificial intelligence technology that can learn and train on a large amount of data to make predictions and decisions on new data. This model can process various types of data, including text, images, and sound, and extract useful information from it to make corresponding predictions and decisions.

[0049] In a wireless communication system, the functional deployment of a physical cell is often determined after the station is opened or after the version is upgraded. Except for a small part of the functions with adaptive adjustment, most functions use the same configuration strategy in different scenarios. However, due to factors such as user tidal effect, idle / busy period transition, and changes in user channel statistical characteristics, the cell scene may change significantly. The focus of cell performance observation in different scenarios will also change. If the same configuration strategy is continued, the cell performance will be limited, resulting in a decrease in spectral efficiency and a deterioration in user experience.

[0050] To solve the technical problem of uniform fixed configuration strategy limiting cell performance, the present application provides a cell function deployment method, device, medium and program product, which intelligently deploys cell-level functions based on scene recognition, so that the comprehensive performance of different cells in different scenarios can reach the optimal condition, thereby improving cell throughput and improving user experience.

[0051] It should be noted that the preset AI model provided in the embodiments of the present application can include but is not limited to mainstream machine learning algorithm models, deep learning algorithm models, artificial intelligence large models, etc., such as decision tree models, multi-layer perception, support vector machines, radial basis function networks, recurrent neural networks, convolutional neural networks (CNN), and training methods including but not limited to supervised learning, unsupervised learning, reinforcement learning, etc.

[0052] The method provided by the embodiments of the present application can be applied to a radio access network device, including but not limited to a base station (BTS) in a Global System of Mobile communication (GSM) or a Code Division Multiple Access (CDMA), a base station (NodeB, NB) in a Wideband Code Division Multiple Access (WCDMA), an evolved NodeB (eNB), an access point (AP) or a relay station in an LTE network, or a base station (gNB) in an NR network, and the like, without limitation.

[0053] Please refer to FIG. 1, which is a flowchart of a method for determining a cell function policy provided by the embodiments of the present application. As shown in FIG. 1, the method in the embodiments of the present application can include but is not limited to the following steps S101-S103.

[0054] In step S101, first feature data of a target cell and a function policy model corresponding to a scene of the target cell are obtained.

[0055] In some examples, the first feature data can be determined according to associated feature information of the cell. The associated feature information of the cell can include but is not limited to a certain amount of user channel measurement information in the cell, a reference signal receiving power (RSRP), user distribution information such as user direction of arrival (DOA) distribution information, user energy projection distribution information, and the like, a number of access users of the cell, a physical resource block (PRB) utilization rate of the cell, neighboring cell interference information, function policy information, user historical scheduling information, and the like. In a specific implementation, the first feature data can be obtained by performing joint statistical characteristic analysis on the associated feature information of the cell to obtain a joint statistical characteristic analysis result, and then constructing the first feature data according to the joint statistical characteristic analysis result. The first feature data can be in a vector form. The joint statistical characteristic analysis includes but is not limited to central tendency analysis, dispersion tendency analysis, correlation analysis, normal distribution test, regression analysis, variance analysis, clustering analysis, time series analysis, and the like.

[0056] In a possible implementation of the present application, the first feature data of the target cell is obtained by the following method: acquiring at least one of the following information of the target cell in a period of time (for example, 15 min granularity): Radio Resource Control (RRC) access user number, Physical Resource Block (PRB) utilization rate, average Channel Quality Indicator (CQI), cell spectral efficiency (SE), user perceived rate Rate, or Neighbor Interference (NI) information; and constructing the first feature data based on the above information, for example, data1=[RRC_std, PRB_Ratio_std, CQI_std, NI_std].

[0057] In another possible implementation of the present application, the first feature data of the target cell is obtained by the following method: acquiring an energy projection vector representing the degree of user line-of-sight (Los) / non-line-of-sight (NLOS) in the target cell, for example, data1=[energy projection vector of 64 beams], and constructing the first feature data based on the energy projection vector of the target cell.

[0058] It should be noted that the information used to construct the first feature data can be flexibly selected and adjusted according to actual conditions during specific implementation.

[0059] In the embodiments of the present application, the performance of the target cell after deploying the preset function strategy is predicted by using the function strategy model corresponding to the scene of the target cell, wherein the function strategy and the function strategy model are corresponding.

[0060] In the embodiments of the present application, the scene of the target cell can be set artificially, can be determined by a preset algorithm, or can be obtained by predicting based on the feature data of the cell by using an AI model.

[0061] It can be understood that after the first feature data and the function strategy model corresponding to the scene of the target cell are obtained, the performance prediction can be started.

[0062] In step S102, the first feature data is input into the function strategy model to obtain performance prediction values corresponding to a plurality of function strategies.

[0063] In step S103, the target function strategy deployed to the target cell is determined from the plurality of function strategies according to the performance prediction values.

[0064] It can be understood that by inputting the first feature data into the function strategy model, the function strategy model can simulate and predict the performance prediction value according to the first feature data, and the performance prediction value is used to represent the numerical value of the cell performance when the function strategy is run. The larger the performance prediction value is, the better the cell performance under the function strategy is. The function strategy most suitable for the target cell under the current scenario can be selected through the performance prediction value.

[0065] In the embodiment of the application, the first feature data of the target cell and the function strategy model are obtained, the first feature data is input into the obtained function strategy model to obtain the corresponding performance prediction value, and then the target function strategy deployed to the target cell is determined from the plurality of function strategies according to the performance prediction value. Through the method provided in the embodiment of the application, the cell-level function can be intelligently deployed based on the scene, so that the comprehensive performance of different cells in different scenes can be optimized to improve the cell throughput and improve the user experience.

[0066] It can be understood that in the embodiment of the application, the function strategy needs to be determined according to the scene where the target cell is located, so that the comprehensive performance of the cell can be optimized, and therefore the scene of the target cell needs to be determined before the function strategy is determined. Please refer to FIG. 2, which is a flowchart of a scene determination method of a target cell provided in the embodiment of the application. As shown in FIG. 2, the method for determining the target cell in the embodiment of the application can include but is not limited to steps S201 to S202.

[0067] Step S201, obtaining second feature data of a target cell.

[0068] Step S202, inputting the second feature data into a scene classification model to obtain a scene corresponding to the target cell.

[0069] Through the scene determination method of the target cell as shown in FIG. 2, the scene where the target cell is located can be determined, and then the target function strategy is determined according to the scene where the target cell is located.

[0070] In some examples, the second feature data of the target cell can be obtained by: acquiring radio resource control (RRC) access user number, physical resource block (PRB) utilization rate, average channel quality indicator (CQI), cell spectral efficiency (SE), user perceived rate (Rate), neighbor cell interference (NI), function policy label, and the like of the target cell in a period of time (e.g., 15 min granularity), and all or part of the information is combined into a scene classification input feature vector, and the input features are standardized, and the second feature data is determined based on the standardized scene classification input feature vector. For example, the radio resource control (RRC) access user number, PRB utilization rate, average CQI information, and cell SE information of the target cell are acquired, and are denoted as data 2 = [RRC_std, PRB_Ratio_std, CQI_std], and thus the second feature data is obtained.

[0071] In some examples, the second feature data of the target cell can also be obtained by: acquiring an energy projection vector representing the degree of user line-of-sight (Los) / non-line-of-sight (NLOS) in the target cell, for example, data2 = [64 beam energy projection vectors], and constructing the second feature data based on the energy projection vector of the target cell.

[0072] In some examples, the scene classification model in the embodiment of the present application is obtained by pre-training. Please refer to FIG. 3, which is a flow diagram of a scene classification model training provided by an embodiment of the present application. As shown in FIG. 3, the training steps of the scene classification model in the embodiment of the present application can include but are not limited to steps S301 to S302.

[0073] Step S301, acquiring a first training set.

[0074] Step S302, acquiring an initial scene classification model, performing first training on the initial scene classification model based on the first training set, and obtaining the scene classification model.

[0075] In some examples, the first training set includes a plurality of first training sample data, and the first training sample data includes but is not limited to at least one of the following data: radio resource control (RRC) access user number, physical resource block (PRB) utilization rate, average channel quality indicator (CQI), cell spectral efficiency (SE), user perceived rate (Rate), neighbor cell interference (NI), function policy label, and energy projection vector.

[0076] In an embodiment of the present application, the training method of the scene classification model is as follows:

[0077] (1) Obtain a certain amount of user channel measurement information, RSRP, user DOA distribution information, user energy projection distribution information, cell access user quantity, cell PRB utilization rate, adjacent cell interference information, function policy information, user historical scheduling information, and other associated feature data in a cell. Perform joint statistical characteristic analysis on the obtained information, and screen and construct a first training set for training a scene classification model. The joint statistical characteristic analysis can include but is not limited to: central tendency analysis, deviation tendency analysis, correlation analysis, normal distribution test, regression analysis, variance analysis, cluster analysis, time series analysis, and the like. The selection of the joint statistical characteristic analysis method can be selected according to actual needs, and the selection of the joint statistical characteristic analysis method is not excessively limited in this application.

[0078] (2) Based on the constructed first training set, the initial scene classification model is trained to obtain a trained scene classification model.

[0079] It should be noted that the scene classification model applied in the embodiments of the present application can be an unsupervised clustering (such as Kmeans clustering) model, a supervised scene recognition (requiring data with partial scene labels) model, or a scene classification model introducing expert experience.

[0080] It should be noted that the cell scene classification can be a network-level mass cell scene centralized classification, or a single cell-level scene classification.

[0081] In a possible implementation of the present application, the training of the scene classification model can be performed according to the following method: obtaining radio resource control (RRC) access user number, PRB utilization rate, average channel quality (CQI), cell spectral efficiency (SE), user perceived rate (Rate), adjacent cell interference (NI), and function policy ID information in a period of time, and combining all or part of the information into a first training sample data of scene classification, and performing standardization processing on the input features, for example, the standardized scene classification first training sample data is recorded as: data1 = [RRC_std, PRB_Ratio_std, CQI_std] (the scene classification input feature can also be an energy projection vector for representing the Los / NLos degree of users in a cell, for example, data1 = [64 beam energy projection vectors], which can be flexibly selected and adjusted according to actual conditions); collect a certain amount of data to construct a data set to obtain a first training set containing scene classification input features, for example, 90,000 data sets are collected to form a scene classification data set; and the 90,000 data sets are classified by a clustering algorithm (such as a Kmeans algorithm) to obtain a qualified scene classification model. Through the above method, the scene classification model required by the embodiments of the present application can be trained.

[0082] After obtaining the trained scene classification model, the scene classification model and the first training set can also be used to build a second training set for training the function strategy model. Referring to FIG. 4, a flowchart for obtaining the second training set according to an embodiment of the present application is shown. As shown in FIG. 4, the obtaining of the second training set can include, but is not limited to, the following steps S401 to S403.

[0083] In step S401, the scene classification model is used to determine the scene label corresponding to each first training sample data, wherein the scene label is used to represent the type of the scene.

[0084] In step S402, the performance label corresponding to the first training sample data is determined according to the scene label corresponding to the first training sample data.

[0085] In step S403, the second training set is built based on all the first training sample data and the performance label corresponding to the first training sample data, wherein the second training set is used to train the function strategy model.

[0086] In some examples, the 90,000 data sets are obtained as described above, and then input into the scene classification model for classification. It is assumed that the total number of classified scene labels is 3, which are scene 1, scene 2, and scene 3. The obtained scene labels are used to determine the performance label corresponding to the first training sample data in the subsequent step S402.

[0087] In some examples, in step S402, the performance label corresponding to the first training sample data is determined according to the scene label corresponding to the first training sample data, which can include: obtaining the class center point vector corresponding to the scene label according to the scene label corresponding to the first training sample data; and determining the performance label corresponding to the first training sample data according to the class center point vector.

[0088] In an embodiment of the present application, the 90,000 data (i.e., the first training sample data) are divided into three categories of data, i.e., scene 1, scene 2, and scene 3 according to the scene. It is assumed that each scene corresponds to 30,000 data. The classified class center point vectors are as follows: scene 1: K1 = [0.25, 0.15, 0.35]; scene 2: K2 = [0.5, 0.4, 0.55]; and scene 3: K3 = [0.8, 0.9, 0.75]. The cell spectrum efficiency SE and the user perception rate Rate of each data are weighted according to the class center point vector of the current scene to obtain the performance label New_SE corresponding to the first training sample data. For example, the weighting method is as follows:

[0089] Taking scenario 1 as an example, the class center point vector is K1 = [w1, w2, w3] = [0.25, 0.15, 0.35], and the New_SE is New_SE = w1*SE + w2*SE + (1-w1)*Rate + (1-w2)*Rate = 0.25*SE + 0.15*SE + 0.75*Rate + 0.85*Rate = 0.4*SE + 1.6*Rate.

[0090] In some examples, constructing the second training set based on all the first training sample data and the performance labels corresponding to the first training sample data can include: determining feature sample data according to the first training sample data, and forming second training sample data according to the feature sample data and the performance labels corresponding to the first training sample data; determining scene labels and function policy labels corresponding to the second training sample data according to the first training sample data corresponding to the second training sample data; and classifying all the second training sample data according to the scene labels and the function policy labels to obtain a plurality of second training sets, wherein the second training sample data in a second training set correspond to a same scene label and a same function policy label.

[0091] In an embodiment of the present application, all the first training sample data is first classified according to scenarios to obtain first training sample data of a plurality of scenarios. For the first training sample data of each scenario, a function policy label corresponding to each first training sample data is obtained, for example, the function policy label corresponding to the first training sample data is 0, indicating that the first training sample data is obtained when the cell MU function is closed in the corresponding scenario; the function policy label corresponding to the first training sample data is 1, indicating that the first training sample data is obtained when the cell MU function is opened in the corresponding scenario; and all the first training sample data in each scenario is classified according to the function policy label, so that second training sample data classified by scenario and function policy is obtained. Second training sample data of the same scenario and the same function policy constitutes a second training set.

[0092] In some examples, the first training sample data corresponding to scenario 1 is represented as [RRC_std, PRB_Ratio_std, CQI_std, NI_std, function policy label], the feature sample data [RRC_std, PRB_Ratio_std, CQI_std, NI_std] is extracted from the first training sample data, the feature sample data and the performance label New_SE of the first training sample data are combined to form a second training sample data, which can be denoted as data2 = [RRC_std, PRB_Ratio_std, CQI_std, NI_std, New_SE]. Wherein, when the function policy label is 0, it indicates that the cell MU function is closed; when the function policy label is 1, it indicates that the cell MU function is opened; for example, data2 = [RRC_std = 20%, PRB_Ratio_std = 10%, CQI_std = 80%, NI_std = 13%, 0, 1.8], which indicates that in the preset scenario, the cell MU function is closed, and the actual performance of the cell is 1.8; data2 = [RRC_std = 19%, PRB_Ratio_std = 9%, CQI_std = 80%, NI_std = 13%, 1, 1.6], which indicates that in the preset scenario, the cell MU function is opened, and the actual performance of the cell is 1.6.

[0093] Suppose there are three scenarios and two different function policies in total, and the function policies are function policy 1 and function policy 2, the second training sample data in the second training set corresponds to the same scenario label and the same function policy label, then:

[0094] The training sample data of function policy 1 under scenario 1 is denoted as: data_S1_F1 = [RRC_std, PRB_Ratio_std, CQI_std, NI_std, New_SE];

[0095] The training sample data of function policy 2 under scenario 1 is denoted as: data_S1_F2 = [RRC_std, PRB_Ratio_std, CQI_std, NI_std, New_SE];

[0096] The training sample data of function policy 1 under scenario 2 is denoted as: data_S2_F1 = [RRC_std, PRB_Ratio_std, CQI_std, NI_std, New_SE];

[0097] The training sample data of function policy 2 under scenario 2 is denoted as: data_S2_F2 = [RRC_std, PRB_Ratio_std, CQI_std, NI_std, New_SE];

[0098] The training sample data of the function strategy 1 under the scenario 3 is denoted as: data_S3_F1 = [RRC_std, PRB_Ratio_std, CQI_std, NI_std, New_SE];

[0099] The training sample data of the function strategy 2 under the scenario 3 is denoted as: data_S3_F2 = [RRC_std, PRB_Ratio_std, CQI_std, NI_std, New_SE].

[0100] The second training sample data of the same scenario and the same function strategy form a second training set. For example, a plurality of training sample data data_S1_F1 form a second training set data_S1_F1 Set, which can be used to train a function strategy model corresponding to the scenario 1 and the function strategy 1.

[0101] In some examples, each scenario has a plurality of corresponding function strategy models, and each function strategy model has a corresponding function strategy label. For example, the scenario 1 includes a function strategy model 1A and a function strategy model 1B, the function strategy model 1A corresponds to a function strategy label 0, indicating a cell MU function off strategy; and the function strategy model 1B corresponds to a function strategy label 1, indicating a cell MU function on strategy.

[0102] In some examples, the training method of the function strategy model of the embodiment of the present application is as follows: obtaining an initial function strategy model and a second training set according to a scenario label and a function strategy label; performing second training on the initial function strategy model based on the second training set to obtain a function strategy model, which is used to predict the performance of a target cell under a corresponding function strategy, and obtain a performance prediction value. The second training set includes a plurality of second training sample data, and the second training sample data includes feature sample data and a performance label.

[0103] The scenario label corresponding to the function strategy model 1A is "scenario 1", and the function strategy label corresponding to the function strategy model 1A is "0", and then the corresponding second training set data_S1_F1 Set is obtained.

[0104] Please refer to FIG. 5, which is a flowchart of a function strategy model training method provided by the embodiment of the present application. As shown in FIG. 5, in the embodiment of the present application, the second training on the initial function strategy model based on the second training set to obtain the function strategy model can include but is not limited to steps S501 to S503.

[0105] In step S501, the feature sample data is input into the initial function strategy model to obtain a performance prediction value corresponding to the feature sample data.

[0106] In step S502, a prediction error value between the performance prediction value corresponding to the feature sample data and the performance label is determined based on a preset target function.

[0107] In step S503, the initial function strategy model is trained according to the prediction error value corresponding to each feature sample data, to obtain a function strategy model.

[0108] In an embodiment of the present application, an initial function strategy model corresponding to a scene is first selected, and a preset target function for the initial function strategy model is constructed. A neural network model can be selected as the initial function strategy model of each function strategy in each scene, and the target function can be constructed by the following formula:

[0109] In the formula, G represents the prediction error value, New_SEi represents the real performance value, P_New_SE_i represents the performance prediction value, N represents the number of feature sample data, and W is a preset parameter.

[0110] After the initial function strategy model is constructed, the prediction error value between the performance prediction value corresponding to the feature sample data and the performance label is determined based on the preset target function.

[0111] Subsequently, the initial function strategy model is trained according to the prediction error value corresponding to each feature sample data, to obtain a function strategy model. For example, in the case of the three scenes and two different function strategies described above, six different second training sets are classified, and six different function strategy models are obtained after training, which are denoted as AI_S1_F1, AI_S1_F2, AI_S2_F1, AI_S2_F2, AI_S3_F1, and AI_S3_F2. AI_S3_F1 is taken as an example, which represents the function strategy model of function strategy 1 in scene 3.

[0112] In an embodiment of the present application, each scene has a plurality of corresponding function strategy models, and each function strategy model corresponds to a function strategy.

[0113] In an embodiment of the present application, the first feature data is input into the function strategy model to obtain performance prediction values corresponding to a plurality of function strategies, including inputting the first feature data into each function strategy model to obtain performance prediction values output by each function strategy model.

[0114] It can be understood that in each different scene, a plurality of different function strategy models are used to adapt to function strategy deployment in different time periods or different situations, and the performance prediction value is used to determine which function strategy is most suitable for the target cell in the current situation to obtain the optimal performance.

[0115] In the embodiment of the present application, the target function strategy deployed to the target cell is determined from the performance prediction values, which can include: determining the maximum performance prediction value from the performance prediction values output by each function strategy model; and determining the target function strategy deployed to the target cell according to the function strategy model corresponding to the maximum performance prediction value and the function strategy corresponding to the function strategy model.

[0116] In some examples, after the model training is completed, the model is deployed and enters the model inference stage. It is assumed that at this time, three function strategy models under different scenarios are deployed, and each scenario includes two function strategy models. At this time, in an application cycle, the second feature data input of the current application cycle is obtained in 15 min granularity, the cell scenario of the current application cycle is inferred by the scene classification model, for example, the scene classification model judges that the current scenario is scenario 2, and then the two function strategy models AI_S2_F1 and AI_S2_F2 under scenario 2 are applied. The performance labels after strategy deployment and the performance prediction values P_New_SE1 and P_New_SE2 are obtained after the inference of the two strategy sub-models, and the larger performance prediction value is selected as the optimal function strategy. For example, P_New_SE2>P_New_SE1, so in the current application cycle, function strategy 2 should be applied under scenario 2.

[0117] The scheme of the embodiment of the present application will be described below through a specific application example.

[0118] Please refer to FIG. 6, which is a flowchart of a method for determining a cell function strategy provided by the embodiment of the present application. As shown in FIG. 6, the specific process of the method for determining a cell function strategy in the embodiment of the present application includes a model training part and a model inference part, which can include but are not limited to the following steps S601-S610.

[0119] Step S601: In a preset time period, obtain the feature data of a cell as first training sample data at a preset time granularity (such as 15 min), and construct a first training set based on the obtained multiple first training sample data.

[0120] The first training sample data includes at least one of the following data: radio resource control (RRC) access user number, physical resource block (PRB) utilization rate, average channel quality (CQI), cell spectral efficiency (SE), user perceived rate (Rate), neighbor cell interference (NI), function strategy label, and energy projection vector.

[0121] In some examples, the first training sample data is denoted as: data1 = [RRC_std, PRB_Ratio_std, CQI_std, NI_std, function policy label]. Wherein, RRC_std represents the number of radio resource control (RRC) access users, PRB_Ratio_std represents the PRB utilization ratio, CQI_std represents the average channel quality, and NI_std represents the adjacent interference degree.

[0122] In step S602, the initial scene classification model is trained for cell scene classification using the first training set. The training manner can be unsupervised training. After the scene classification training is completed, a scene classification model is obtained, which is used for scene classification model inference in model inference.

[0123] In this example, it is assumed that the total number of classified scene categories is 3, which are scene 1, scene 2, and scene 3.

[0124] In step S603, the data in the first training set is divided by scene. It is assumed that the first training set has a total of 90,000 data, and each scene is divided into 30,000 data. Then, based on the divided data in each scene, the class center point vector corresponding to the data of each scene is determined. It is assumed that the class center point vectors of various scenes are as follows:

[0125] Scene 1: K1 = [0.25, 0.15, 0.35];

[0126] Scene 2: K2 = [0.5, 0.4, 0.55];

[0127] Scene 3: K3 = [0.8, 0.9, 0.75];

[0128] For each first training sample data, the class center point vector corresponding to the first training sample data is weighted and operated to determine the performance label corresponding to the first training sample data.

[0129] In step S604, a second training set for training a function policy model is constructed based on the first training set.

[0130] The first training sample data is obtained by scene. The first training sample data data1 is converted into data2 = [RRC_std, PRB_Ratio_std, CQI_std, NI_std, function policy label, New_SE] based on the performance label New_SE corresponding to the first training sample data.

[0131] When the function policy label is 0, it means that the cell MU function is closed; when the function policy label is 1, it means that the cell MU function is opened; for example, when data2 = [RRC_std = 20%, PRB_Ratio_std = 10%, CQI_std = 80%, NI_std = 13%, 0, 1.8], it means that in such a scenario, the cell MU function is closed, and the actual performance of the cell is 1.8; when the input vector data2 = [RRC_std = 19%, PRB_Ratio_std = 9%, CQI_std = 80%, NI_std = 13%, 1, 1.6], it means that in such a scenario, the cell MU function is opened, and the actual performance of the cell is 1.6.

[0132] Then, all data2 are classified according to the scene label and the function policy label to obtain a plurality of second training sets.

[0133] Suppose there are three scenes and two different function policies in total, and the function policies are function policy 1 and function policy 2, respectively, and the second training sample data in the second training set correspond to the same scene label and the same function policy label, for example:

[0134] The second training sample data of function policy 1 under scene 1 is recorded as data_S1_F1 = [RRC_std, PRB_Ratio_std, CQI_std, NI_std, New_SE];

[0135] The second training sample data of function policy 2 under scene 1 is recorded as data_S1_F2 = [RRC_std, PRB_Ratio_std, CQI_std, NI_std, New_SE];

[0136] The second training sample data of function policy 1 under scene 2 is recorded as data_S2_F1 = [RRC_std, PRB_Ratio_std, CQI_std, NI_std, New_SE];

[0137] The second training sample data of function policy 2 under scene 2 is recorded as data_S2_F2 = [RRC_std, PRB_Ratio_std, CQI_std, NI_std, New_SE];

[0138] The second training sample data of function policy 1 under scene 3 is recorded as data_S3_F1 = [RRC_std, PRB_Ratio_std, CQI_std, NI_std, New_SE];

[0139] The second training sample data of the function strategy 2 under the scenario 3 is denoted as: data_S3_F2 = [RRC_std, PRB_Ratio_std, CQI_std, NI_std, New_SE].

[0140] In step S605, an initial function strategy model corresponding to the scenario is selected, and a preset objective function for the initial function strategy model is constructed. For example, a neural network model can be selected as the initial function strategy model of each function strategy under each scenario.

[0141] In step S606, the initial function strategy model is trained according to the prediction error value corresponding to each feature sample data, to obtain a function strategy model. For example, in the case of the three scenarios and two different function strategies described above, six different second training sets are classified, and six different function strategy models are obtained after training, which are denoted as: AI_S1_F1, AI_S1_F2, AI_S2_F1, AI_S2_F2, AI_S3_F1, and AI_S3_F2. AI_S3_F1 is taken as an example, which represents the function strategy model of the function strategy 1 under the scenario 3.

[0142] In step 607, after the model training is completed, the model deployment is performed, the trained scenario classification model and the function strategy model are combined, and after the model deployment, the model inference is performed to predict the strategy and intelligently deploy.

[0143] The model combination mode can include but is not limited to model cascading, model parallel, etc. For example, the AI submodel 1 (scenario classification model) provides a label ID of scenario division for the current cell, the label ID is obtained, the AI submodel 2 (function strategy model) under the corresponding scenario is activated through the scenario label ID for model inference, and then the optimal strategy of the current cell is obtained. For example, the input vectors are constructed respectively, the AI submodel 1 and the AI submodel 2 are independently inferred and calculated, and finally the inference results of the two models are integrated to make a decision on the optimal function strategy.

[0144] In step S608, the feature data of the current application cycle with a granularity of 15 min is obtained within an application cycle, the scenario classification model is inferred, and the cell scenario of the current application cycle is obtained through the scenario classification model inference. The feature data includes but is not limited to: the number of radio resource control (RRC) access users, the physical resource block (PRB) utilization rate, the average channel quality (CQI), the cell spectral efficiency (SE), the user perceived rate (Rate), the neighbor cell interference (NI), the function strategy label, and the energy projection vector.

[0145] Step S609, selecting a function strategy model according to the scene type determined by the scene classification, to determine the optimal function strategy of the cell according to the performance prediction value output by the function strategy model.

[0146] For example, if the current scene is determined to be scene 2, two function strategy models AI_S2_F1 and AI_S2_F2 are selected under scene 2. The two function strategy models obtain performance prediction values P_New_SE1 and P_New_SE2 corresponding to the performance labels after strategy deployment respectively, and the larger prediction value is the optimal function strategy. For example, P_New_SE2>P_New_SE1, so function strategy 2 should be applied under scene 2 in the current application cycle. The two function strategy models AI_S2_F1 and AI_S2_F2 obtain performance prediction values P_New_SE1 and P_New_SE2 corresponding to the performance labels after strategy deployment respectively, and the larger prediction value is the optimal function strategy. For example, P_New_SE2>P_New_SE1, so function strategy 2 should be applied under scene 2 in the current application cycle.

[0147] Step S610, deploying the optimal function strategy for the cell.

[0148] Continuing with the previous example, function strategy 2 is deployed for the cell in the current cycle.

[0149] The method for determining a function strategy of a cell provided by the embodiments of the present application can determine a target function strategy that is suitable for the scene in which the cell is located, so that the comprehensive performance of different cells in different scenes can be optimal, thereby improving the throughput of the cell and improving the user experience.

[0150] The embodiments of the present application also provide an electronic device, as shown in FIG. 7, the electronic device 1400 includes:

[0151] one or more processors 1410;

[0152] a memory 1420 having one or more programs stored thereon, when the one or more programs are executed by the one or more processors 1410, the one or more processors 1410 implement the method for determining a function strategy of a cell provided by any of the embodiments of the present application.

[0153] The memory 1420, as a non-transitory network system, can be used to store non-transitory software programs and non-transitory computer executable programs. In addition, the memory 1420 can include a high-speed random access memory, and can also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some embodiments, the memory 1420 can optionally include a memory 1420 disposed remotely with respect to the processor 1410, which can be connected to the processor 1410 through a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0154] The memory 1420 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1420 can store an operating system and other application programs. When the technical solutions provided by the embodiments of the present specification are implemented by software or firmware, the related program codes are stored in the memory 1420 and are called and executed by the processor 1410 to implement the method of the embodiments of the present application.

[0155] The processor 1410 can be implemented in the form of a general-purpose CPU (central processing unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute related programs to implement the technical solutions provided by the embodiments of the present application.

[0156] In some embodiments, the electronic device further comprises:

[0157] - an input / output interface for realizing information input and output;

[0158] - a communication interface for realizing communication interaction between the device and other devices, which can realize communication through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.);

[0159] - a bus for transmitting information between various components (such as the processor 1410, the memory 1420, the input / output interface, and the communication interface) of the device;

[0160] - wherein the processor 1410, the memory 1420, the input / output interface, and the communication interface can realize communication connection between each other inside the device through the bus.

[0161] An embodiment of the present application further provides a computer readable storage medium, which stores computer executable instructions for implementing the method for determining a cell function strategy provided by any of the embodiments of the present application.

[0162] An embodiment of the present application further provides a computer program product, which comprises a computer program or computer instructions stored in a computer readable storage medium, and a processor of a computer device reads the computer program or computer instructions from the computer readable storage medium, and the processor executes the computer program or computer instructions, so that the computer device implements the method for determining a cell function strategy provided by any of the embodiments of the present application.

[0163] In the embodiments of the present application, the first feature data and the function strategy model of the target cell are acquired, the first feature data is input into the acquired function strategy model to obtain a corresponding performance prediction value, and then the performance prediction value is used to determine a target function strategy deployed to the target cell from multiple function strategies. Through the method provided by the embodiments of the present application, the target function strategy that is adapted to the scene where the cell is located can be determined, so that the comprehensive performance of different cells in different scenes can be optimal, thereby improving the cell throughput and improving the user experience.

[0164] The system architecture and application scenarios described in the embodiments of the present application are used to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. It can be known by those skilled in the art that, as the system architecture evolves and new application scenarios appear, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0165] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware, and the computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, storage, databases, or other media in this application is intended to include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0166] Those skilled in the art can understand that all or some steps of the above-mentioned methods and systems can be implemented as software, firmware, hardware and appropriate combinations thereof. Some or all physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software can be distributed on a computer readable medium, which can include computer storage media (or non-transitory media) and communication media (or transitory media). As known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information such as computer readable instructions, data structures, program modules or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. In addition, those skilled in the art know that communication media generally includes computer readable instructions, data structures, program modules or other data in modulated data signals such as carrier waves or other transmission mechanisms, and can include any information delivery medium.

[0167] The above describes some embodiments of the present application with reference to the accompanying drawings, and is not limited to the scope of the present application. Any modification, equivalent replacement and improvement made by those skilled in the art without departing from the scope and essence of the present application shall be within the scope of the present application.

Claims

1. A method for determining a cell function strategy, comprising: obtaining first feature data of a target cell and a function strategy model corresponding to a scenario of the target cell; inputting the first feature data into the function strategy model to obtain performance prediction values corresponding to multiple function strategies; determining a target function strategy deployed to the target cell from the multiple function strategies according to the performance prediction values.

2. The method for determining a cell function policy according to claim 1, wherein, The scenario of the target cell is determined by the following steps: obtaining second feature data of a target cell; inputting the second feature data into a scenario classification model to obtain a scenario corresponding to the target cell.

3. The method for determining a cell function policy according to claim 2, wherein, The scenario classification model is obtained by the following steps: obtaining a first training set, wherein the first training set comprises multiple first training sample data; obtaining an initial scenario classification model, and performing first training on the initial scenario classification model based on the first training set to obtain the scenario classification model.

4. The method for determining a cell function policy according to claim 3, wherein, The first training sample data comprises at least one of the following data: a radio resource control (RRC) access user number, a physical resource block (PRB) utilization rate, an average channel quality indicator (CQI), a cell spectral efficiency (SE), a user perceived rate (Rate), a neighbor cell interference (NI), a function strategy label, and an energy projection vector.

5. The method for determining a cell function policy according to claim 3, wherein, After obtaining the scenario classification model, the method comprises: determining a scenario label corresponding to each first training sample data based on the scenario classification model, wherein the scenario label is used to represent a type of the scenario; determining a performance label corresponding to the first training sample data according to the scenario label corresponding to the first training sample data; constructing a second training set based on all the first training sample data and the performance label corresponding to the first training sample data, wherein the second training set is used to train the function strategy model.

6. The method for determining a cell function policy according to claim 5, wherein, The construction of the second training set based on all the first training sample data and the performance label corresponding to the first training sample data comprises: determining feature sample data according to the first training sample data, and forming second training sample data according to the feature sample data corresponding to the first training sample data and the performance label; determining the scenario label and a function strategy label corresponding to the second training sample data according to the first training sample data corresponding to the second training sample data; classifying all the second training sample data according to the scenario label and the function strategy label to obtain multiple second training sets, wherein the second training sample data in the second training set corresponds to the same scenario label and the same function strategy label.

7. The method for determining a cell function policy according to claim 5, wherein, The determination of the performance label corresponding to the first training sample data according to the scenario label corresponding to the first training sample data comprises: obtaining a class center point vector corresponding to the scenario label according to the scenario label corresponding to the first training sample data; determining the performance label corresponding to the first training sample data according to the class center point vector.

8. The method for determining a cell function policy according to claim 1, wherein, Each of the scenarios has a plurality of corresponding function strategy models, each of the function strategy models has a corresponding function strategy label, and the function strategy models are trained by the following steps: An initial function strategy model and a second training set are obtained according to the scenario label and the function strategy label; The initial function strategy model is second trained based on the second training set to obtain the function strategy model.

9. The method for determining a cell function policy according to claim 8, wherein, The second training set includes a plurality of second training sample data, the second training sample data includes feature sample data and a performance label, and the initial function strategy model is second trained based on the second training set to obtain the function strategy model, including: The feature sample data is input into the initial function strategy model to obtain a performance prediction value corresponding to the feature sample data; A prediction error value between the performance prediction value corresponding to the feature sample data and the performance label is determined based on a preset target function; The initial function strategy model is trained according to the prediction error value corresponding to each of the feature sample data to obtain the function strategy model.

10. The method for determining a cell function policy according to claim 1, wherein, Each of the scenarios has a plurality of corresponding function strategy models, each of the function strategy models corresponds to a function strategy, and the first feature data is input into the function strategy model to obtain a performance prediction value corresponding to a plurality of function strategies, including: The first feature data is input into each of the function strategy models to obtain the performance prediction value output by each of the function strategy models.

11. The method for determining a cell function policy according to claim 10, wherein, The target function strategy deployed to the target cell is determined from a plurality of the function strategies according to the performance prediction value, including: A maximum performance prediction value is determined from the performance prediction value output by each of the function strategy models; The target function strategy deployed to the target cell is determined according to the function strategy model corresponding to the maximum performance prediction value and the function strategy corresponding to the function strategy model.

12. An electronic device, comprising: one or more processors; memory storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method for determining a cell function strategy according to any one of claims 1-11.

13. A computer readable storage medium storing a computer program, wherein, The computer program is executed by the processor to implement the method for determining a cell function strategy according to any one of claims 1-11.

14. A computer program product comprising a computer program, wherein, The computer program is executed by the processor to implement the method for determining a cell function strategy according to any one of claims 1-11.

Images