Network performance prediction method and device, equipment, storage medium and program product

Abstract

The invention provides a network performance prediction method and device, equipment, a storage medium and a program product, and the method comprises the steps: collecting wireless communication data corresponding to each grid in a target region, the wireless communication data comprising antenna information of a base station, distribution information of a terminal, and a performance index of a reference signal sent by the base station to the terminal; for each grid, estimating a channel angle power spectrum according to the sparse features of the wireless communication data and the reference signal; and according to the channel angle power spectrum and the beam parameter of the base station, predicting a communication performance index and a sensing performance index corresponding to each grid under the beam parameter. According to the embodiment of the invention, collaborative optimization of the communication perception integrated network is realized, and the network performance prediction accuracy is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wireless communication, and particularly relates to a network performance prediction method, device, equipment, storage medium and program product. BACKGROUND

[0002] In the field of integrated sensing and communication (ISAC), the existing technology mainly focuses on multi-antenna channel modeling and beamforming design to integrate communication and sensing functions. Like the air-ground collaborative communication and sensing integration optimization method, although it solves some problems, it does not consider the integrated sensing and communication modeling; the multi-antenna dual-function radar-communication base station design focuses on resource allocation, has poor adaptability in complex dynamic environments and high computational complexity; the target-level-based statistical channel modeling framework is insufficient for communication performance optimization and has limited scalability; the localized channel modeling technology ignores high-order channel characteristics and has poor dynamic adaptability; the spectrum efficiency prediction method relies on a large amount of road test data, has poor interpretability, and does not comprehensively consider the dual-function indicators of communication and sensing. Overall, the existing technology does not fully consider the collaborative design of communication and sensing tasks when predicting network performance, thereby having problems such as poor dynamic environment adaptability and low network performance prediction accuracy. SUMMARY

[0003] The purpose of the present application is to provide a network performance prediction method, device, equipment, storage medium and program product, which is used to solve the problem that the existing technology does not fully consider the collaborative design of communication and sensing tasks when predicting network performance, thereby having problems such as poor dynamic environment adaptability and low network performance prediction accuracy.

[0004] In order to solve the above technical problems, the embodiment of the present application provides a network performance prediction method, comprising:

[0005] Collecting wireless communication data corresponding to each grid in a target area, wherein the wireless communication data includes antenna information of a base station, distribution information of a terminal, and performance indicators of reference signals sent by the base station to the terminal;

[0006] For each grid, estimating a channel angle power spectrum according to the wireless communication data and sparse characteristics of the reference signals;

[0007] According to the channel angle power spectrum and pre-configured beam parameters of the base station, predicting the corresponding communication performance indicators and sensing performance indicators of each grid under the beam parameters.

[0008] Optionally, the method further comprises:

[0009] input the predicted communication performance indicators and the perception performance indicators into a pre-trained neural network model to obtain predicted spectrum efficiencies of each of the grids corresponding to the beam parameters respectively, wherein the neural network model is used for spectrum efficiency prediction according to the communication performance indicators and the perception performance indicators, and the predicted spectrum efficiencies are generated.

[0010] Optionally, the wireless communication data corresponding to each of the grids in the target area respectively includes:

[0011] collecting raw data in the target area;

[0012] performing data cleaning and normalization on the raw data to obtain wireless communication data;

[0013] dividing the target area into a plurality of grids to determine the wireless communication data corresponding to each of the grids respectively.

[0014] Optionally, the communication performance indicators include communication received power and communication signal-to-interference-and-noise ratio of the reference signal, and the perception performance indicators include perception signal-to-interference-and-noise ratio.

[0015] The method further includes:

[0016] predicting, according to the channel angle power spectrum and the pre-configured beam parameters of the base station, the communication received power of the reference signal corresponding to each of the grids respectively under the beam parameters.

[0017] calculating, according to the predicted communication received power, the communication signal-to-interference-and-noise ratio corresponding to each of the grids respectively under the beam parameters.

[0018] predicting, according to the channel angle power spectrum, the beam parameters of the base station, and the line-of-sight transmission paths between the base station and each of the grids respectively, the perception signal-to-interference-and-noise ratio corresponding to each of the grids respectively under the beam parameters.

[0019] Optionally, the method further includes:

[0020] adjusting the beam parameters of the base station by adjusting a precoding matrix and a steering matrix of the base station.

[0021] Optionally, the predicting, according to the channel angle power spectrum, the beam parameters of the base station, and the line-of-sight transmission paths between the base station and each of the grids respectively, the perception signal-to-interference-and-noise ratio corresponding to each of the grids respectively under the beam parameters includes:

[0022] For each of the grids, spatial information corresponding to a first radar signal transmitted by the base station to the grid is obtained according to the channel angle power spectrum, wherein the spatial information comprises angle information and scattering cross-section information;

[0023] According to the spatial information corresponding to each of the grids, the beam parameters of the base station, and the line-of-sight transmission paths between the base station and each of the grids, the perceived signal-to-interference-and-noise ratio of each of the grids corresponding to the beam parameters is predicted.

[0024] Optionally, the method further comprises:

[0025] Obtaining historical communication data corresponding to a plurality of historical grids in the target area in a historical period, wherein the historical communication data comprises at least one of base station configuration information, performance indicators and transmission indicators of the reference signal;

[0026] According to the historical communication data, calculating a spectral efficiency label of each of the historical grids, wherein the spectral efficiency label comprises average spectral efficiency and / or instantaneous spectral efficiency;

[0027] Constructing a training data set, wherein the training data set comprises a plurality of historical grids, historical communication data and spectral efficiency labels corresponding to each of the historical grids;

[0028] Training a deep neural network according to the training data set to generate the neural network model.

[0029] Optionally, the feature types of the historical communication data comprise category features, numerical features and active antenna type features;

[0030] The training of the deep neural network according to the training data set to generate the neural network model comprises:

[0031] According to the feature types, the historical communication data in the training data set is divided into category feature data, numerical feature data and active antenna type data;

[0032] The category feature data is processed by one-hot encoding to obtain first training data;

[0033] The active antenna type data is processed by embedding to obtain second training data;

[0034] According to the first training data, the second training data and the numerical feature data, the historical communication data in the training data set is updated;

[0035] The deep neural network is trained based on the updated training dataset to generate the neural network model.

[0036] This invention also provides a network performance prediction device, comprising:

[0037] The first acquisition module is used to acquire wireless communication data corresponding to each grid in the target area. The wireless communication data includes antenna information of the base station, distribution information of the terminal, and performance indicators of the reference signal sent by the base station to the terminal.

[0038] The first estimation module is used to estimate the channel angle power spectrum for each of the grids based on the sparse characteristics of the wireless communication data and the reference signal.

[0039] The first prediction module is used to predict the communication performance index and sensing performance index of each grid under the beam parameters, based on the channel angle power spectrum and the pre-configured beam parameters of the base station.

[0040] This invention also provides a network device, including: a processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the network performance prediction method as described in any of the preceding embodiments.

[0041] This invention also provides a readable storage medium, comprising: a program stored on the readable storage medium, wherein when the program is executed by a processor, it implements the steps of the network performance prediction method as described in any of the preceding claims.

[0042] This invention also provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the network performance prediction method as described in any of the preceding embodiments.

[0043] At least one of the above technical solutions of the present invention has the following beneficial effects:

[0044] In the scheme, a network performance prediction method is provided, which specifically comprises the following steps: first, collecting wireless communication data corresponding to each grid in a target area, wherein the wireless communication data comprises antenna information of a base station, distribution information of a terminal, and a performance index of a reference signal transmitted by the base station to the terminal; in a prediction stage, first, for each grid, a channel angle power spectrum is estimated according to the wireless communication data and a sparse feature of the reference signal; then, according to the channel angle power spectrum and a pre-configured beam parameter of the base station, a communication performance index and a perception performance index corresponding to each grid under the beam parameter are predicted. The present application can realize the cooperative optimization of a communication-perception integrated network according to real collected road test data, improve the overall efficiency of the system, improve the network performance prediction accuracy, realize the effective prediction of the key performance indicators (communication performance index and perception performance index) of the communication-perception integrated network when the network parameters change, and is suitable for network optimization work and has strong dynamic environment adaptability. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 A flowchart of a network performance prediction method of an embodiment of the present application is shown in the figure.

[0046] Figure 2 An input data processing flowchart of a neural network model of an embodiment of the present application is shown in the figure.

[0047] Figure 3 A structure diagram of a network performance prediction system of an embodiment of the present application is shown in the figure.

[0048] Figure 4 A structure diagram of a network performance prediction device of an embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, and not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0050] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a particular order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally a class, not limited to the number of objects, for example, the first object can be one or more. In addition, "and / or" in the specification and claims indicates at least one of the connected objects, and the character " / ", generally indicates that the objects before and after are in a "or" relationship.

[0051] As Figure 1 shown, the embodiment of the present application provides a network performance prediction method, comprising:

[0052] Step S101, collecting wireless communication data corresponding to each grid in the target area respectively, wherein the wireless communication data includes antenna information of the base station, distribution information of the terminal and performance indicators of reference signals sent by the base station to the terminal;

[0053] In step S101, in the commonly used communication integrated networking, different types of antenna devices are used at the base station end, different devices have different numbers of sending antennas N, the base station sends M reference signals for detecting users near the base station. In order to reduce measurement error, users located in the same area (for example, an area with a size of 100m x 100m) are divided into the same grid. For communication users and sensing targets, single-antenna receiving devices are used, and the base station sends downlink data, and each user or sensing target can simultaneously receive signals from multiple base stations.

[0054] In the network performance prediction test, first, wireless communication data is collected by actual engineering measurement, which includes but is not limited to antenna information of the base station, distribution information of the terminal and performance indicators of reference signals sent by the base station to the terminal, wherein the antenna information of the base station includes but is not limited to the number of antennas N, the number of beams M, the antenna spacing and , representing the antenna spacing of the antenna panel X dimension, ​The antenna spacing in the Y dimension of the antenna panel is represented; the distribution information of the terminal includes, but is not limited to, the distribution information of the user corresponding to the terminal and the grid position of the terminal; the performance indicators of the reference signal include, but are not limited to, the reference signal received power (RSRP) and the signal to interference plus noise ratio (SINR). Furthermore, RSRP includes the RSRP of multiple beams of the primary serving cell base station and the RSRP of multiple beams of the neighboring cell base station for grid reception, and SINR includes the SINR of multiple beams of the primary serving cell base station and the SINR of multiple beams of the neighboring cell base station.

[0055] Step S102: For each of the grids, estimate the channel angular power spectrum based on the sparse characteristics of the wireless communication data and the reference signal;

[0056] In step S102, when estimating the channel angular power spectrum (APS), since there are only a few major multipaths in the channel, the channel APS can usually be considered sparse. Therefore, based on the sparse characteristics of the reference signal, the APS estimation problem is modeled as a sparse recovery problem for solution. The method for solving the channel APS can be the Orthogonal Matching Pursuit (OMP) method, which can effectively reduce computational complexity, but this invention does not limit it.

[0057] Step S103: Based on the channel angle power spectrum and the pre-configured beam parameters of the base station, predict the communication performance index and sensing performance index corresponding to each grid under the beam parameters.

[0058] In step S103, communication performance indicators include, but are not limited to, communication RSRP and SINR; sensing performance indicators include, but are not limited to, sensing SINR; and beam parameters include, but are not limited to, the base station's precoding matrix and steering matrix. By adjusting the base station's precoding matrix and steering matrix, at least one of the following can be affected: beam direction angle, beam tilt angle, beamwidth, beam power factor, and beamforming gain. Based on step S103, the communication performance indicators and sensing performance indicators corresponding to each grid cell under different beam parameters can be predicted.

[0059] In the embodiment of the present application, a network performance prediction method is provided, which specifically comprises: first, collecting wireless communication data corresponding to each grid in a target area, wherein the wireless communication data includes antenna information of a base station, distribution information of a terminal, and a performance index of a reference signal transmitted by the base station to the terminal; in the prediction stage, first, for each grid, estimating a channel angle power spectrum according to the wireless communication data and the sparse characteristics of the reference signal; and then, predicting a communication performance index and a perception performance index corresponding to each grid under the beam parameters according to the channel angle power spectrum and the pre-configured beam parameters of the base station. The present application can realize the cooperative optimization of the communication-perception integrated network according to the real collected road test data, improve the overall system efficiency, improve the network performance prediction accuracy, and realize that the key performance indicators (communication performance index and perception performance index) of the communication-perception integrated network can be effectively predicted when the network parameters change, which is suitable for network optimization work and has strong dynamic environment adaptability.

[0060] Optionally, the method further comprises:

[0061] inputting the predicted communication performance index and the predicted perception performance index into a pre-trained neural network model to obtain a predicted spectrum efficiency of each grid under the beam parameters output by the neural network model, wherein the neural network model is used for spectrum efficiency prediction according to the communication performance index and the perception performance index to generate the predicted spectrum efficiency.

[0062] In the embodiment of the present application, according to the communication performance index and the perception performance index predicted in step S103, the neural network model is used for spectrum efficiency prediction, which can predict the predicted spectrum efficiency of each grid corresponding to different beam parameters. The above process comprehensively considers the low-order and high-order network performance indicators, and the dual performance indicators of communication and perception, realizes the cooperative optimization of the communication-perception integrated network, improves the overall system efficiency, adopts the neural network model to improve the inference speed, meets the rapidly changing network demand, and performs spectrum efficiency prediction according to the simulation data (communication performance index and perception performance index) generated in step S103, instead of relying only on road test data for prediction, which greatly reduces the data dependency. The key performance indicators (communication performance index, perception performance index and predicted spectrum efficiency) of the communication-perception integrated network can all be effectively predicted when the network parameters change, which is suitable for network optimization work and has strong dynamic environment adaptability.

[0063] In one embodiment, the collected wireless communication data corresponding to each grid in the target area comprises:

[0064] collecting original data in the target area;

[0065] Data cleaning and normalization are performed on the original data to obtain wireless communication data;

[0066] The target area is divided into multiple grids, and the wireless communication data corresponding to each grid is determined.

[0067] In the embodiment of the application, when collecting wireless communication data, first, RSRP and SINR data of multiple cells and multiple grids are collected through base stations and user terminals to ensure data coverage of different environmental conditions and network scenarios. Then, the collected wireless communication data is cleaned to remove abnormal values and noise data, and RSRP and SINR are normalized to eliminate the influence of different orders of magnitude data on subsequent modeling. Finally, the target area is divided into multiple grids, and users in each grid use single-antenna receiving equipment to ensure the consistency of user data in the same grid, and wireless communication data corresponding to each grid is generated.

[0068] In one embodiment, optionally, an embodiment for estimating channel angle power spectrum according to the wireless communication data and the sparse characteristics of the reference signal is provided, which is as follows:

[0069] The statistical relationship between the beam-level RSRP measurement value of the first grid and the expected channel APS is as follows:

[0070]

[0071] wherein, represents the beam-level RSRP measurement value of the first grid, represents the expected calculation, represents a predefined precoding matrix of M beams, and the item of the matrix can be represented as , m represents the mth beam, and n represents the nth antenna, represents a uniform planar array (UPA) antenna steering vector matrix, and the item of the matrix can be represented as , wherein, represents an antenna angle component, represents the number of antennas in the X dimension of the antenna panel, represents the number of antennas in the Y dimension of the antenna panel, represents the angle of departure in the vertical direction, represents the angle of departure in the horizontal direction, is the M beam-level RSRP measurement value, represents the APS expectation in the continuous time slot.

[0072] In estimating the APS, first by estimating , since there are only a few main multipaths in the channel, the channel APS can be generally considered to be sparse, therefore, based on the sparsity of the reference signal, the APS estimation problem is modeled as a sparse recovery problem, the channel APS is solved using the OMP method, and is expressed as . The specific OMP algorithm is as follows:

[0073] ‌Requirements: K, W, S, rsrp;

[0074] 1: initialization k=0, S=∅, =0, ;

[0075] ‌2: repeat;

[0076] 3: ;

[0077] 4: ;

[0078] 5: ;

[0079] 6: ;

[0080] 7: ;

[0081] 8: ;

[0082] 9: until or ;

[0083] ‌Ensure: .

[0084] wherein K represents an iteration counter, S represents a support set, represents a precoding matrix of the base station, represents a steering matrix of the base station, represents an initial residual value, represents an updated residual value after the k+1th update. The above OMP algorithm belongs to the prior art and will not be described in detail here.

[0085] In one embodiment, the communication performance indicator includes a communication received power and a communication signal-to-noise ratio of the reference signal, and the sensing performance indicator includes a sensing signal-to-noise ratio;

[0086] The communication performance index and the perception performance index corresponding to each of the grids under the beam parameters are predicted according to the channel angle power spectrum and the beam parameters of the base station.

[0087] The communication received power of the reference signal corresponding to each of the grids under the beam parameters is predicted according to the channel angle power spectrum and the beam parameters of the base station.

[0088] The communication signal-to-noise ratio corresponding to each of the grids under the beam parameters is calculated according to the predicted communication received power.

[0089] The perception signal-to-noise ratio corresponding to each of the grids under the beam parameters is predicted according to the channel angle power spectrum, the beam parameters of the base station and the line-of-sight transmission path between the base station and each of the grids.

[0090] In step S103, the beam parameters of the base station can be adjusted manually by adjusting the precoding matrix and the steering matrix of the base station.

[0091] In the prediction, for the communication RSRP, the prediction method includes: first, adjusting to the required beam parameters, then, assuming that the channel APS is unchanged in the continuous time slot T, calculating the RSRP corresponding to each of the grids under the beam parameters according to the channel APS in step S102 and the adjusted beam parameters, and the formula is as follows:

[0092]

[0093] wherein, represents the predicted beam-level RSRP prediction value of the i-th grid, represents the adjusted precoding matrix, represents the adjusted steering matrix, represents the adjusted steering matrix, represents the estimated channel APS in step S102.

[0094] In addition, the embodiment of the present application also provides a verification method for RSRP prediction accuracy, which specifically includes: measuring the RSRP prediction accuracy by comparing the mean absolute error (MAE) of the predicted reference signal received power under the change of the beam parameters and the real reference signal received power after the change of the beam parameters, and the formula is as follows:

[0095]

[0096] wherein, denotes the L1 norm calculation, denotes the total number of samples, denotes the real measurement value of the reference signal received power of the grid after the beam parameter adjustment, denotes the predicted measurement value of the reference signal received power of the grid after the beam parameter adjustment.

[0097] For the communication SINR, the RSRP of the base station k to the grid can be expressed as: max is used to denote the maximum beam is the representative of the base station k to the grid . According to the predicted communication RSRP, the communication SINR corresponding to each grid under the beam parameter is calculated, and the formula is as follows:

[0098]

[0099] For the perception SINR, assuming that the base station is used as a single station co-located radar, on the basis of the channel APS, according to the beam parameter of the base station and the line-of-sight transmission path, the perception SINR corresponding to each grid under the beam parameter is predicted.

[0100] In an embodiment, the method further comprises:

[0101] adjusting the beam parameter of the base station by adjusting the precoding matrix and the steering matrix of the base station.

[0102] In the embodiment of the application, an antenna rotation physical model is constructed in advance according to the antenna data of the base station, the antenna rotation physical model can adjust the precoding matrix and the steering matrix of the base station, and further adjust the beam parameter of the base station, and the specific method comprises the following steps: first, adjusting the precoding matrix of the base station through the antenna rotation physical model, then determining the adjustment of the steering matrix according to the mechanical rotation angle and the electronic rotation angle, and further adjusting the beam parameter, so that when the communication performance index and the perception performance index are obtained, after the channel APS is obtained, the precoding matrix and the steering matrix of the base station are adjusted through the antenna rotation physical model, and the communication performance index and the perception performance index of the grid under different beam parameters can be obtained.

[0103] In an embodiment, the method further comprises:

[0104] For each of the grids, according to the channel angle power spectrum, spatial information corresponding to a first radar signal transmitted by the base station to the grid is obtained, wherein the spatial information comprises angle information and scattering cross section information.

[0105] According to the spatial information corresponding to each of the grids, beam parameters of the base station, and line-of-sight transmission paths between the base station and each of the grids, respectively, a perceived signal-to-interference-and-noise ratio (SINR) corresponding to each of the grids under the beam parameters is predicted.

[0106] In the embodiment of the present application, it is assumed that the base station is used as a single station co-located radar, and a perception target is in a grid, one perception target corresponds to one grid, therefore, the angle of arrival (AoA) and the angle of departure (AoD) are assumed to be equal, and therefore, the target response matrix of the perception target can be represented as:

[0107]

[0108] wherein, denotes radar cross section (RCS), denotes estimated echo path loss, it is assumed that a line-of-sight (LOS) path exists between the base station and the perception target, and the path direction with the maximum gain is the LOS direction. Therefore, the echo signal received by the base station can be represented as:

[0109]

[0110] wherein, is additive white gaussian noise (AWGN) on the base station side.

[0111] The receiving beamforming vector is used to accept the desired echo signal. Therefore, in order to measure how much environmental information can be obtained from the reflection signal of the first target, the perception SINR of the perception target at the base station (i.e., the perception SINR of the grid corresponding to the perception target at the base station) can be represented as:

[0112]

[0113] In an embodiment, optionally, the method further comprises:

[0114] acquire historical communication data corresponding to each of the historical grids in the target area in a historical period, the historical communication data including at least one of base station configuration information, performance indicators of the reference signals, and transmission indicators;

[0115] calculate a spectral efficiency label of each of the historical grids according to the historical communication data, wherein the spectral efficiency label includes average spectral efficiency and / or instantaneous spectral efficiency;

[0116] construct a training data set, wherein the training data set includes the historical communication data and the spectral efficiency label corresponding to each of the historical grids;

[0117] train a deep neural network according to the training data set to generate the neural network model.

[0118] In the embodiments of the present application, the historical communication data is collected through actual engineering measurement, and the base station configuration information includes but is not limited to New Radio Serving Physical Cell Identity (NRServing PCI), Next Generation NodeB Identifier (gNodeB ID), Coverage Scenario, Active Antenna Unit Type (AAUType). The performance indicators of the reference signals include but are not limited to RSRP of Synchronization Signal Block (SSB) beam 0 and RSRP of SSB beam 7, SINR of SSB beam 0 and SINR of SSB beam 7. The transmission indicators of the reference signals include but are not limited to Average Modulation and Coding Scheme (AvgMCS), Average Rank (Avg Rank).

[0119] The spectral efficiency label includes average spectral efficiency and / or instantaneous spectral efficiency, wherein the average spectral efficiency is calculated in the following manner:

[0120] Let denote the downlink spectral efficiency of M sampling points to be predicted, wherein denote the average spectral efficiency of the jth sampling point, and in the embodiments of the present application, statistical data obtained by base station side measurement is used to represent , and the calculation formula is as follows:

[0121]

[0122] wherein, denotes the number of service initiations required by the jth sampling point, denotes the cumulative spectrum efficiency within the jth service number, in order to obtain the conversion factor 180 also needs to be considered.

[0123] The calculation method of the instantaneous spectrum efficiency is as follows:

[0124] Taking the qth base station as an example, the downlink channel transmitted by the qth base station to the users in a certain grid is represented as wherein, Q represents the total number of base stations. For the instantaneous downlink spectrum efficiency of the grid , it can be represented as:

[0125]

[0126] wherein, denotes the transmission rank of the grid , and denotes the downlink SINR received by the user at the kth flow.

[0127] Then, according to the historical communication data and the spectrum efficiency label, a training data set is constructed;

[0128] Finally, based on the loss function, the deep neural network is trained using the training data set to generate the neural network model.

[0129] Taking the spectrum efficiency label as the instantaneous downlink spectrum efficiency as an example: the prediction target of the neural network model is to predict the instantaneous downlink spectrum efficiency corresponding to the user in each grid, and the formula is as follows:

[0130]

[0131] wherein, the parameter mapping represents the prediction process, and is the trainable parameter.

[0132] The prediction target of the neural network model is to minimize the average percentage error (Mean Absolute Percentage Error, MAPE) between the predicted spectrum efficiency and the real spectrum efficiency through the loss function, and the formula of the loss function can be represented as:

[0133]

[0134] ​In addition, it should be noted that during model training, each piece of test data in the training data set is regarded as an input sequence. In order to avoid overfitting, an early stopping mechanism is adopted during model training, and in order to enhance the generalization ability of the model to unknown active antenna unit types, part of the samples in each training batch are randomly reclassified as the "unknown active antenna unit type" type.

[0135] In one embodiment, the feature types of the historical communication data include category features, numerical features and active antenna type features.

[0136] The method comprises the following steps:

[0137] The historical communication data in the training data set is divided into category feature data, numerical feature data and active antenna type data according to feature types.

[0138] The category feature data is processed by one-hot encoding to obtain first training data.

[0139] The active antenna type data is processed by embedding to obtain second training data.

[0140] The historical communication data in the training data set is updated according to the first training data, the second training data and the numerical feature data.

[0141] The neural network model is generated by training a deep neural network according to the updated training data set.

[0142] As shown in FIG. 1, Figure 2 The main body of the neural network model adopted by the embodiment of the application adopts a deep neural network, contains nine hidden layers, and the number of neurons in the composition is: [64, 128, 256, 1024, 512, 128, 32, 16, 1]. Based on this, a data processing method before inputting the neural network model is provided. First, the historical communication data in the training data set is divided into category feature data, numerical feature data and active antenna type data, for example: the category feature data includes NR Serving PCI, gNodeB ID and coverage scenario; the numerical feature data includes AvgMCS, Avg Rank, SINR of SSB and RSRP of SSB; and the active antenna type data includes AAU Type. Then, the one-hot encoding is applied to the category feature data to retain the category information, and the numerical feature is directly input into the model to utilize the magnitude change. Secondly, for the active antenna type data, the embedding technology is introduced to convert it into a compact numerical vector, so as to balance the complexity and information retention ability of the input feature.

[0143] As Figure 3 shown, based on the network performance prediction method provided by the embodiment of the application, the embodiment of the application further provides a network performance prediction system, comprising: a data acquisition module, a data-driven channel modeling module, an antenna rotation physical model and a neural network model, wherein the data acquisition module is used to acquire wireless communication data, that is, to implement the method corresponding to step S101, the data-driven channel modeling module is used to estimate the channel angle power spectrum of each grid, that is, to implement the method corresponding to step S102, the antenna rotation physical model is used to adjust the beam parameter of the base station, and then to predict the communication performance index and the perception performance index corresponding to each grid under different beam parameters, and the neural network model is used to predict the downlink spectrum efficiency, that is, to implement the method corresponding to step S104.

[0144] Integrating the above-mentioned modules or models into a unified software platform can realize the automation process of data acquisition, channel modeling and spectrum efficiency prediction. In addition, based on the above-mentioned system, a dynamic adjustment and optimization strategy can also be realized, and the specific method is as follows:

[0145] Through the data acquisition module, the network performance index is continuously monitored, and the latest data is input into the data-driven channel modeling module, the antenna rotation physical model and the neural network model, and the dynamic adjustment and optimization strategy is adjusted.

[0146] A user interface is provided to allow network administrators to monitor the optimization process, view real-time performance indicators, and manually adjust optimization parameters or start a new optimization cycle as needed.

[0147] In summary, by introducing the network performance prediction method, the embodiment of the application successfully solves the defects of the existing communication and perception integrated network performance modeling technology in dynamic environment adaptability, computational complexity, comprehensive performance optimization, data dependency and model interpretability, and achieves significant technical, economic and social effects. In actual application, the RSRP estimation error is reduced by 8.3%, and the spectrum efficiency prediction error is reduced by 55%. In addition, by considering the key performance indicators of the dual functions of communication and perception, the spectrum utilization is improved, and a large amount of data acquisition cost is saved, significantly reducing the operation and maintenance cost. On the social level, the application promotes the development of low-altitude economy, intelligent transportation and other fields, and is expected to improve the safety and efficiency of unmanned aerial vehicle flight. Technically, the interpretability of the model is enhanced, making the prediction results more transparent and reliable, and improving the credibility and operability in practical application. Overall, through efficient, flexible and accurate communication and perception integrated network performance modeling, the application not only improves the system performance and economic benefits, but also promotes the process of social intelligent development.

[0148] As Figure 4As shown, the embodiment of the present application further provides a network performance prediction device, comprising:

[0149] The first acquisition module 401 is configured to acquire wireless communication data corresponding to each grid in the target area, wherein the wireless communication data comprises antenna information of a base station, distribution information of a terminal, and a performance index of a reference signal transmitted by the base station to the terminal.

[0150] The first estimation module 402 is configured to estimate a channel angle power spectrum for each grid according to the wireless communication data and a sparse feature of the reference signal.

[0151] The first prediction module 403 is configured to predict a communication performance index and a perception performance index corresponding to each grid under a preconfigured beam parameter of the base station according to the channel angle power spectrum and the beam parameter.

[0152] Optionally, the device further comprises:

[0153] The second prediction module is configured to input the predicted communication performance index and the predicted perception performance index into a pre-trained neural network model to obtain a predicted spectrum efficiency of each grid under the beam parameter output by the neural network model, wherein the neural network model is used to perform spectrum efficiency prediction according to the communication performance index and the perception performance index to generate the predicted spectrum efficiency.

[0154] Optionally, the first acquisition module 401 comprises:

[0155] The first acquisition sub-module is configured to acquire original data in the target area.

[0156] The first processing sub-module is configured to perform data cleaning and normalization processing on the original data to obtain wireless communication data.

[0157] The first division sub-module is configured to divide the target area into a plurality of grids to determine the wireless communication data corresponding to each grid.

[0158] Optionally, the communication performance index in the first prediction module 403 comprises a communication received power and a communication signal-to-interference-and-noise ratio of the reference signal, and the perception performance index comprises a perception signal-to-interference-and-noise ratio.

[0159] The first prediction module 403 comprises:

[0160] The first prediction sub-module is configured to predict a communication received power of the reference signal corresponding to each grid under the preconfigured beam parameter of the base station according to the channel angle power spectrum and the beam parameter.

[0161] a first calculating sub-module, configured to calculate, according to the predicted communication received power, a communication signal-to-interference-and-noise ratio corresponding to each of the grids under the beam parameter of the base station;

[0162] a second predicting sub-module, configured to predict, according to the channel angle power spectrum, the beam parameter of the base station, and a line-of-sight transmission path between the base station and each of the grids, a perception signal-to-interference-and-noise ratio corresponding to each of the grids under the beam parameter of the base station.

[0163] Optionally, the apparatus further comprises:

[0164] a first adjusting device, configured to adjust the beam parameter of the base station by adjusting a precoding matrix and a steering matrix of the base station.

[0165] Optionally, the second predicting sub-module comprises:

[0166] a first obtaining unit, configured to, for each of the grids, obtain, according to the channel angle power spectrum, spatial information corresponding to a first radar signal transmitted by the base station to the grid, wherein the spatial information comprises angle information and scattering cross-section information;

[0167] a first predicting unit, configured to predict, according to the spatial information corresponding to each of the grids, the beam parameter of the base station, and the line-of-sight transmission path between the base station and each of the grids, the perception signal-to-interference-and-noise ratio corresponding to each of the grids under the beam parameter of the base station.

[0168] Optionally, the apparatus further comprises:

[0169] a first obtaining module, configured to obtain historical communication data corresponding to a plurality of historical grids in the target region in a historical period, wherein the historical communication data comprises at least one of base station configuration information, performance indicators of the reference signal, and transmission indicators;

[0170] a first calculating module, configured to calculate, according to the historical communication data, a spectral efficiency label of each of the historical grids, wherein the spectral efficiency label comprises average spectral efficiency and / or instantaneous spectral efficiency;

[0171] a first constructing module, configured to construct a training data set, wherein the training data set comprises historical communication data and spectral efficiency labels corresponding to a plurality of historical grids, respectively;

[0172] a first training module, configured to train a deep neural network according to the training data set, to generate the neural network model.

[0173] Optionally, the feature types of the historical communication data in the first obtaining module include category features, numerical value features and active antenna type features.

[0174] The first training module comprises:

[0175] The second dividing sub-module is configured to divide the historical communication data in the training data set according to feature types to obtain category feature data, numerical value feature data and active antenna type data.

[0176] The second processing sub-module is configured to process the category feature data through one-hot encoding to obtain first training data.

[0177] The third processing sub-module is configured to process the active antenna type data through embedding to obtain second training data.

[0178] The first updating sub-module is configured to update the historical communication data in the training data set according to the first training data, the second training data and the numerical value feature data.

[0179] The first training sub-module is configured to train a deep neural network according to the updated training data set to generate the neural network model.

[0180] It should be noted that the embodiments of the device correspond to the embodiments of the above method, and all implementation manners in the embodiments of the above method are applicable to the embodiments of the device, and the same technical effects can be achieved.

[0181] The embodiments of the present application also provide a network device, which comprises a processor, a memory and a program stored in the memory and executable on the processor, and the program, when executed by the processor, implements the network performance prediction method according to any one of the above embodiments and can achieve the same technical effects. To avoid repetition, details are not described here.

[0182] The embodiments of the present application also provide a readable storage medium, which comprises a program stored in the readable storage medium, and the program, when executed by a processor, implements the steps of the network performance prediction method according to any one of the above embodiments and can achieve the same technical effects. To avoid repetition, details are not described here. The computer readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0183] The embodiment of the present application further provides a computer program product comprising computer instructions, which, when executed by a processor, implement the steps of the network performance prediction method according to any one of the above and achieve the same technical effects. To avoid repetition, details are not described herein.

[0184] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another, without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0185] Obviously, various modifications and changes can be made to the present application by those skilled in the art without departing from the spirit and scope of the present application. Thus, it is intended that the present application encompass such modifications and changes as fall within the scope of the appended claims and their equivalents.

Claims

1. A method for predicting network performance, characterized in that, include: The wireless communication data corresponding to each grid cell within the target area is collected. The wireless communication data includes the antenna information of the base station, the distribution information of the terminals, and the performance indicators of the reference signal sent by the base station to the terminal. For each of the grids, the channel angular power spectrum is estimated based on the sparse characteristics of the wireless communication data and the reference signal; Based on the channel angle power spectrum and the pre-configured beam parameters of the base station, predict the communication performance index and sensing performance index corresponding to each grid under the beam parameters.

2. The network performance prediction method according to claim 1, characterized in that, The method further includes: The predicted communication performance index and the sensing performance index are input into a pre-trained neural network model to obtain the predicted spectral efficiency of each grid corresponding to the beam parameters, as output by the neural network model. The neural network model is used to predict the spectral efficiency based on the communication performance index and the sensing performance index to generate the predicted spectral efficiency.

3. The network performance prediction method according to claim 1, characterized in that, The wireless communication data corresponding to each grid cell within the target area is collected, including: Collect raw data within the target area; The raw data is cleaned and normalized to obtain the wireless communication data; The target area is divided into multiple grids, and the wireless communication data corresponding to each grid is determined.

4. The network performance prediction method according to claim 1, characterized in that, The communication performance indicators include the communication received power and the communication signal-to-interference-plus-noise ratio of the reference signal, and the sensing performance indicators include the sensing signal-to-interference-plus-noise ratio; The step of predicting the communication performance index and sensing performance index of each grid under the beam parameters based on the channel angle power spectrum and the pre-configured beam parameters of the base station includes: Based on the channel angle power spectrum and the pre-configured beam parameters of the base station, predict the communication reception power of the reference signal corresponding to each grid under the beam parameters; Based on the predicted communication received power, the communication signal-to-interference-plus-noise ratio (SIR) of each grid under the beam parameters is calculated. Based on the channel angle power spectrum, the beam parameters of the base station, and the line-of-sight transmission path between the base station and each of the grids, the perceived signal-to-interference-plus-noise ratio (SINNR) of each grid under the beam parameters is predicted.

5. The network performance prediction method according to any one of claims 1 or 4, characterized in that, The method further includes: The beam parameters of the base station are adjusted by adjusting the precoding matrix and steering matrix of the base station.

6. The network performance prediction method according to claim 4, characterized in that, The step of predicting the perceived signal-to-interference-plus-noise ratio (SINNR) for each grid under the specified beam parameters based on the channel angular power spectrum, the base station's beam parameters, and the line-of-sight transmission path between the base station and each grid includes: For each of the grid cells, spatial information corresponding to the first radar signal sent by the base station to the grid cell is obtained based on the channel angular power spectrum, wherein the spatial information includes angular information and scattering cross section information; Based on the spatial information corresponding to each grid, the beam parameters of the base station, and the line-of-sight transmission path between the base station and each grid, the perceived signal-to-interference-plus-noise ratio (SINNR) corresponding to each grid under the beam parameters is predicted.

7. The network performance prediction method according to claim 2, characterized in that, The method further includes: The historical communication data corresponding to multiple historical grids within the target area during a historical period is obtained. The historical communication data includes at least one of base station configuration information, performance indicators of the reference signal, and transmission indicators. Based on the historical communication data, a spectral efficiency tag is calculated for each of the historical grids, wherein the spectral efficiency tag includes average spectral efficiency and / or instantaneous spectral efficiency; Construct a training dataset, wherein the training dataset includes historical communication data and spectral efficiency labels corresponding to a plurality of historical grids, each of the historical grids; A deep neural network is trained based on the training dataset to generate the neural network model.

8. The network performance prediction method according to claim 7, characterized in that, The feature types of the historical communication data include category features, numerical features, and active antenna type features; Wherein, training a deep neural network based on the training dataset to generate the neural network model includes: The historical communication data in the training dataset are segmented according to feature type to obtain categorical feature data, numerical feature data, and active antenna type data; The category feature data is processed by one-hot encoding to obtain the first training data; The second training data is obtained by processing the active antenna type data through embedding. Update the historical communication data in the training dataset based on the first training data, the second training data, and the numerical feature data; The deep neural network is trained based on the updated training dataset to generate the neural network model.

9. A network performance prediction device, characterized in that, include: The first acquisition module is used to acquire wireless communication data corresponding to each grid in the target area. The wireless communication data includes antenna information of the base station, distribution information of the terminal, and performance indicators of the reference signal sent by the base station to the terminal. The first estimation module is used to estimate the channel angle power spectrum for each of the grids based on the sparse characteristics of the wireless communication data and the reference signal. The first prediction module is used to predict the communication performance index and sensing performance index of each grid under the beam parameters, based on the channel angle power spectrum and the pre-configured beam parameters of the base station.

10. A network device, characterized in that, include: A processor, a memory, and a program stored in the memory and executable on the processor, the program implementing the network performance prediction method as described in any one of claims 1 to 8 when executed by the processor.

11. A readable storage medium, characterized in that, include: The readable storage medium stores a program that, when executed by a processor, implements the steps of the network performance prediction method as described in any one of claims 1 to 8.

12. A computer program product, characterized in that, It includes computer instructions that, when executed by a processor, implement the steps of the network performance prediction method as described in any one of claims 1 to 8.

Images