Channel prediction model training method and device, electronic equipment and readable storage medium

Abstract

This application discloses a channel prediction model training method, apparatus, electronic device, and readable storage medium, applied in the field of communication technology. The method includes: acquiring historical channel state information, historical channel data, and prior information of at least one preset base station over a historical time period; determining the usage environment scenario of each preset base station; generating multiple training samples based on the historical channel state information, historical channel data, prior information of each base station, and usage environment scenario; and iteratively optimizing the channel prediction model to be trained based on the multiple training samples to obtain the channel prediction model. This application solves the technical problem of low generalization of channel prediction models.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a channel prediction model training method, apparatus, electronic device, and readable storage medium. Background Technology

[0002] With the rapid development of science and technology, communication technology has become increasingly mature. In order to improve the performance of communication systems, the demand for the accuracy of channel state information estimation of base stations is also increasing.

[0003] Currently, channel prediction models trained using historical channel state information from various base stations often use this information as the basis for channel state prediction. However, this approach can lead to lower prediction accuracy for base stations in other scenarios, resulting in lower generalization ability of the channel prediction model. Summary of the Invention

[0004] The main objective of this application is to provide a channel prediction model training method, apparatus, electronic device, and readable storage medium, aiming to solve the technical problem of low generalization of channel prediction models in the prior art.

[0005] To achieve the above objectives, this application provides a channel prediction model training method, which includes:

[0006] The system acquires historical channel status information, historical channel data, and historical base station prior information for at least one preset base station within a historical time period, and determines the usage environment scenario of each preset base station.

[0007] Multiple training samples are generated based on the historical channel state information, historical channel data, base station prior information, and usage environment scenarios.

[0008] Based on the multiple training samples, the channel prediction model to be trained is iteratively optimized to obtain the channel prediction model.

[0009] To achieve the above objectives, this application also provides a channel prediction model training apparatus, the channel prediction model training apparatus comprising:

[0010] The acquisition module is used to acquire historical channel status information, historical channel data and historical base station prior information of at least one preset base station in a historical time period, and to determine the usage environment scenario of each preset base station.

[0011] The generation module is used to generate multiple training samples based on the historical channel state information, historical channel data, base station prior information, and usage environment scenarios.

[0012] The iteration module is used to iteratively optimize the channel prediction model to be trained based on the multiple training samples to obtain the channel prediction model.

[0013] This application also provides an electronic device, the electronic device comprising: a memory, a processor, and a program of the channel prediction model training method stored in the memory and executable on the processor, wherein when the program of the channel prediction model training method is executed by the processor, the steps of the channel prediction model training method as described above can be implemented.

[0014] This application also provides a computer-readable storage medium storing a program for implementing a channel prediction model training method, wherein when the program for implementing the channel prediction model training method is executed by a processor, it implements the steps of the channel prediction model training method as described above.

[0015] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the channel prediction model training method described above.

[0016] This application provides a channel prediction model training method, apparatus, electronic device, and readable storage medium. The method involves acquiring historical channel state information of at least one preset base station over a historical time period, prior information of each preset base station, and environmental information of each preset base station. Multiple training samples are generated based on these historical channel state information, prior information of each base station, and environmental information. The channel prediction model to be trained is iteratively optimized based on these training samples to obtain the channel prediction model. Since the training samples are generated from the historical channel state information, prior information of each base station, and environmental information, and are used to iteratively optimize the channel prediction model, the optimized channel prediction model uses historical channel state information, prior information of each base station, and environmental information as decision-making criteria for channel state information prediction. Therefore, the generalization ability and prediction accuracy of the channel prediction model are improved. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1This is a flowchart illustrating the first embodiment of the channel prediction model training method of this application;

[0020] Figure 2 This is a flowchart illustrating the training process of a channel prediction model to be trained, as described in the channel prediction model training method of this application.

[0021] Figure 3 This is a flowchart illustrating the training process of another channel prediction model to be trained, which is involved in the channel prediction model training method in this application embodiment.

[0022] Figure 4 This is a flowchart illustrating the overall process of training a channel prediction model to be trained, as described in the channel prediction model training method of this application.

[0023] Figure 5 This is a flowchart illustrating the overall process involved in the channel prediction model training method in the embodiments of this application.

[0024] Figure 6 This is a schematic diagram of the hardware operating environment involved in the channel prediction model training method in the embodiments of this application.

[0025] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0026] To make the above-mentioned objectives, features, and advantages of this application more apparent and understandable, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] Example 1

[0028] This application provides a channel prediction model training method. In the first embodiment of the channel prediction model training method of this application, refer to... Figure 1 The channel prediction model training method includes:

[0029] Step S10: Obtain historical channel status information, historical channel data and historical base station prior information of at least one preset base station in a historical time period, and determine the usage environment scenario of each preset base station.

[0030] In this embodiment, it should be noted that the preset base station is a base station that has been constructed and is in stable use. The base station prior information includes at least one of channel data time-frequency correlation information, channel data frequency domain information, and channel data noise amplitude information. The usage environment scenario includes at least one of transportation hubs, indoor areas, outdoor areas, special environments, and environments prone to crowding. Among them, transportation hubs include at least one of subway stations, train stations, and bus stations; indoor areas include at least one of shopping malls, office buildings, and residential areas; outdoor areas include at least one of roads, railways, and country roads; special environments include at least one of seaside, mountainous areas, and enclosed basements; and environments prone to crowding include at least one of airports, hospitals, and hotels. The channel data is the data transmitted in the signal path of the base station, and the type of channel data includes at least one of pilot signal type data and time-frequency full channel type data.

[0031] In one feasible embodiment, historical channel state information, historical channel data, and historical base station prior information of at least one preset base station at any historical time period are obtained.

[0032] It is understandable that, since the channel data of base stations changes in real time, if historical channel data of a preset base station is collected over a long period of time, the channel prediction model trained based on the collected historical channel data may have low generalization due to the fact that the channel data of the preset base station has changed.

[0033] In another feasible embodiment, in order to overcome the above-mentioned defects, historical channel state information, historical channel data and historical base station prior information of at least one preset base station in the most recent historical time period are obtained.

[0034] By collecting historical channel data from preset base stations in the most recent historical time period, the generalization ability of the channel prediction model trained based on the collected historical channel data is avoided to a certain extent due to the real-time changes in the channel data of the base stations. Therefore, the prediction accuracy of the channel prediction model is improved.

[0035] In another feasible embodiment, each of the preset base stations is equipped with a Flume program. Each preset base station calls its own deployed Flume program to collect its corresponding channel state information, channel data, and base station prior information. The historical channel state information, historical channel data, and historical base station prior information of each preset base station in the historical time period are stored in the Kafka cluster in the offline server.

[0036] In one feasible embodiment, the usage environment scenario of each of the preset base stations is obtained by manual identification.

[0037] In another feasible embodiment, any one of the preset base stations is taken as the target base station. Environmental information of the environment where the target base station is located and environmental information of the environment where at least one classified base station is located are collected. A first similarity between the environmental information of the environment where the target base station is located and the environmental information of the environment where each of the classified base stations is located is calculated. A target classified base station with the largest corresponding first similarity is selected from among the classified base stations. The classified environment scene corresponding to the target classified base station is taken as the usage environment scene where the target base station is located. Alternatively, channel data of each of the classified base stations are collected. A second similarity between the historical channel data of the target base station and the channel data of each of the classified base stations is calculated. A target classified base station with the largest corresponding second similarity is selected from among the classified base stations. The classified environment scene corresponding to the target classified base station is taken as the usage environment scene where the target base station is located.

[0038] In another feasible embodiment, a preset configuration file is obtained, wherein the preset configuration file includes the correspondence between the base station's environmental information and the base station's channel data and the base station's usage environment scenario. Any one of the preset base stations is taken as the target base station, and the environmental information of the environment in which the target base station is located is collected. Based on the environmental information of the environment in which the target base station is located and the historical channel data of the target base station, the preset configuration file is queried to obtain the usage environment scenario in which the target base station is located.

[0039] In one feasible embodiment, the historical channel state information, the historical channel data, and the historical base station prior information are classified and stored using Hive according to the usage environment scenario of their respective base stations.

[0040] Step S20: Generate multiple training samples based on the historical channel state information, historical channel data, base station prior information, and usage environment scenarios.

[0041] In one feasible embodiment, the historical channel state information, the historical channel data, and the base station prior information are classified according to their respective usage environment scenarios to obtain classification results; and training samples corresponding to each usage environment scenario are generated based on the classification results.

[0042] In another feasible embodiment, a training sample consists of input feature data and a real label corresponding to the input feature data. The input feature data includes historical channel data of a preset base station within the historical time period, prior information of the historical base station, and the corresponding usage environment scenario. The real label includes historical channel state information of the preset base station corresponding to the input feature data within the historical time period.

[0043] Step S30: Based on the multiple training samples, iteratively optimize the channel prediction model to be trained to obtain the channel prediction model.

[0044] In one feasible embodiment, based on the training samples corresponding to each of the usage environment scenarios, the channel prediction model to be trained for each of the usage environment scenarios is iteratively optimized to obtain the channel prediction model corresponding to each of the various usage environment scenarios.

[0045] It is understandable that, since the training samples corresponding to various usage scenarios are different, the structure of the channel prediction models corresponding to each usage scenario, trained from the training samples corresponding to each usage scenario, is different. That is, the hidden layer parameters of the models are different, and the number of multilayer perceptron layers in the multilayer perceptron mixer of the models may also be different.

[0046] In another feasible embodiment, the input feature data of the training samples are input into the channel prediction model to be trained to obtain the output label, the difference between the output label and the true label is determined, and the model loss corresponding to the channel prediction model to be trained is calculated based on the difference. Then, it is determined whether the model loss has converged. If the model loss has converged, the channel prediction model to be trained is used as the channel prediction model. If the model loss has not converged, the channel prediction model to be trained is updated based on the gradient calculated by the model loss using a preset model update method, wherein the preset model update method includes gradient descent and gradient ascent, etc.

[0047] This application provides a channel prediction model training method. It involves acquiring historical channel state information of at least one preset base station over a historical time period, prior information of each preset base station, and environmental information of each preset base station. Multiple training samples are generated based on these historical channel state information, prior information of each base station, and environmental information. The channel prediction model to be trained is iteratively optimized based on these training samples to obtain the channel prediction model. Since the training samples are generated from the historical channel state information, prior information of each base station, and environmental information, and are used to iteratively optimize the channel prediction model, the optimized channel prediction model uses historical channel state information, prior information of each base station, and environmental information as decision-making criteria to predict channel state information. Therefore, the generalization ability and prediction accuracy of the channel prediction model are improved.

[0048] Example 2

[0049] Furthermore, based on the first embodiment of this application, in another embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description and will not be repeated hereafter. Based on this, the step of determining the usage environment scenario of each preset base station in step S10 includes:

[0050] Step S11: Take any one of the preset base stations as the target base station, and obtain the environmental information of the environment where at least one classified base station is located, the corresponding channel data, and the environmental information of the environment where the target base station is located.

[0051] In this embodiment, it should be noted that the classified base station is a base station that has been classified according to its usage environment scenario.

[0052] Step S12: Based on the channel data corresponding to each of the classified base stations and the historical channel data of the target base station, predict the first probability that the target base station is in a classified environment scenario corresponding to any of the classified base stations.

[0053] In one feasible embodiment, a second similarity is obtained between the historical channel data of the target base station and the channel data of each of the classified base stations. The second similarity is normalized to obtain the normalized result corresponding to each second similarity. The normalized result corresponding to each second similarity is used as the first probability that the target base station is in the classified environment scene corresponding to any of the classified base stations.

[0054] In another feasible embodiment, a channel data environment classification model is obtained by training the channel data corresponding to each of the classified base stations, and the historical channel data of the target base station is input into the channel data environment classification model to obtain the first probability that the target base station is in a classified environment scenario corresponding to any of the classified base stations.

[0055] Step S13: Based on the environmental information of the environment where the classified base station is located and the environmental information of the environment where the target base station is located, predict the second probability that the target base station is in the classified environment scene corresponding to any of the classified base stations.

[0056] In one feasible embodiment, the first similarity between the environmental information of the environment where the target base station is located and the environmental information of the environment where each of the classified base stations is located is obtained. The first similarity is normalized to obtain the normalized result corresponding to each first similarity. The normalized result corresponding to each first similarity is used as the second probability that the target base station is in the classified environment scene corresponding to any of the classified base stations.

[0057] In step S13, the environmental information includes at least one of environmental sound data, base station usage information, environmental meteorological information, and environmental location information.

[0058] The step of predicting the second probability that the target base station is in a classified environment scenario corresponding to any one of the classified base stations, based on the environmental information of the environment where the classified base stations are located and the environmental information of the environment where the target base station is located, includes:

[0059] Step A10: Obtain the environmental scene prediction model obtained by iterative optimization of the environmental information of the environment where the classified base stations are located;

[0060] Step A20: Based on the environmental information of the environment in which the target base station is located, the second probability of the target base station being in any of the classified environmental scenarios corresponding to the classified base station is predicted by the environmental scenario prediction model.

[0061] In one feasible embodiment, the environmental information of the environment in which the target base station is located is mapped to a second probability that the target base station is in a classified environmental scene corresponding to any one of the classified base stations through the environmental scene prediction model.

[0062] Step S14: Take any one of the classified base stations as the classified selected base station, and calculate the total probability that the target base station is in the classified environment scene corresponding to the classified selected base station based on the first probability that the target base station is in the classified environment scene corresponding to the classified selected base station and the second probability that the target base station is in the classified environment scene corresponding to the classified selected base station.

[0063] In one feasible embodiment, the sum of the first probability and the second probability corresponding to the classified selected base station is taken as the total probability that the target base station is in the classified environment scenario corresponding to the classified selected base station.

[0064] In another feasible embodiment, if the first probability corresponding to the classified selected base station is greater than or equal to the second probability, then the first probability is taken as the total probability; if the first probability corresponding to the classified selected base station is less than the second probability, then the second probability is taken as the total probability.

[0065] It is understandable that there are multiple methods to determine the usage environment of a base station, such as judging based on the base station's environmental information or judging based on the base station's channel data. By using both environmental information and channel data as the basis for determining the usage environment of a base station, the accuracy of judging the usage environment of a base station is improved.

[0066] In step S14, the step of calculating the total probability that the target base station is in the classified environment scenario corresponding to the classified selected base station based on the first probability that the target base station is in the classified environment scenario corresponding to the classified selected base station and the second probability that the target base station is in the classified environment scenario corresponding to the classified selected base station includes:

[0067] Step B10: Determine the first weight corresponding to the first probability and the second weight corresponding to the second probability of the selected base station;

[0068] In one feasible embodiment, the first weight corresponding to the first probability and the second weight corresponding to the second probability are obtained under the classified environment scenario corresponding to the selected base station.

[0069] In one feasible embodiment, multiple weighted samples of the classified environment scenario corresponding to the selected base station are obtained. Each weighted sample includes input data and a true label corresponding to the input data. The input data includes a first probability and a second probability that the environment is such that the preset base station is in the classified environment scenario corresponding to the selected base station. The true label is the true probability that the preset base station corresponding to the input data is in the classified environment scenario corresponding to the selected base station. By inputting the input data from the multiple weighted samples into a linear regression model, the first weight corresponding to the first probability and the second weight corresponding to the second probability under the classified environment scenario corresponding to the selected base station are fitted according to the true label corresponding to the input data.

[0070] Step B20: Calculate the first product between the first probability and the first weight of the classified selected base station, and the second product between the second probability and the second weight, and use the sum of the first product and the second product as the total probability that the target base station is in the classified environment scenario corresponding to the classified selected base station.

[0071] Step S15: Select the target classification base station that maximizes the total probability among the classified base stations, and take the classified environment scene corresponding to the target classification base station as the usage environment scene of the target base station.

[0072] This application provides a channel prediction model training method. By using any one of the preset base stations as the target base station, environmental information of the environment of at least one classified base station, corresponding channel data, and environmental information of the target base station's environment are obtained. Based on the channel data corresponding to each classified base station and the historical channel data of the target base station, a first probability is predicted that the target base station is in a classified environment scenario corresponding to any one of the classified base stations. Based on the environmental information of the environment of the classified base stations and the environmental information of the target base station's environment, a second probability is predicted that the target base station is in a classified environment scenario corresponding to any one of the classified base stations. The first probability and each of the second probabilities are used to calculate the total probability that the target base station is in any of the classified base stations in the classified environment scenario. Among the classified base stations, the target classified base station that maximizes the corresponding total probability is selected. The classified environment scenario corresponding to the target classified base station is taken as the usage environment scenario of the target base station. Channel data and environmental information are used together as the decision basis for determining the usage environment scenario of the base station, which improves the accuracy of the judgment of the usage environment scenario of the preset base station. The usage environment scenario of the preset base station is used to generate training samples, and the training samples are used to train the channel prediction model to be trained. Therefore, the prediction accuracy of the channel prediction model is also improved.

[0073] Example 3

[0074] Furthermore, based on the first embodiment of this application, in another embodiment of this application, the content that is the same as or similar to the above-described embodiments one and / or two can be referred to the above description, and will not be repeated hereafter. Based on this, in step S20, the step of generating multiple training samples according to the historical channel state information, the historical channel data, the base station prior information, and the usage environment scenarios includes:

[0075] Step S21: Based on the historical channel state information, historical channel data, and historical base station prior information corresponding to each of the usage environment scenarios, generate multiple training samples corresponding to each usage environment scenario. Each training sample consists of input feature data and a real label corresponding to the input feature data. The input feature data includes historical channel data and historical base station prior information of a preset base station in a usage environment scenario within the historical time period. The real label includes historical channel state information of the preset base station in the usage environment scenario corresponding to the input feature data within the historical time period.

[0076] In step S30, the step of iteratively optimizing the channel prediction model to be trained based on the multiple training samples to obtain the channel prediction model includes:

[0077] Step S31: Based on the multiple training samples corresponding to each of the aforementioned usage environment scenarios, iteratively optimize the channel prediction model to be trained for each usage environment scenario to obtain the channel prediction model corresponding to each usage environment scenario.

[0078] In this embodiment, it should be noted that the channel prediction model to be trained can be a convolutional neural network or a recurrent neural network or other network with hidden layer structures. Its hidden layers can include convolutional layers, multilayer perceptrons, residual modules, feature fusion layers or multilayer perceptron mixers and other structures. The channel prediction model to be trained can also use fully connected layers as classifiers to output model prediction results.

[0079] In one feasible embodiment, any scenario in each of the usage environment scenarios is taken as the target environment scenario. Based on the training samples corresponding to the target environment scenario, the channel prediction model to be trained corresponding to the target environment scenario is iteratively optimized to obtain the channel prediction model corresponding to the target environment scenario.

[0080] In one feasible embodiment, reference is made to Figure 2 , Figure 2 This is a flowchart illustrating the training process of a channel prediction model involved in the channel prediction model training method described in this application. Figure 2 The channel prediction model to be trained includes a multilayer perceptron, a multilayer perceptron mixer, and a fully connected layer. By inputting training samples into the channel prediction model to be trained, the multilayer perceptron and multilayer perceptron mixer in the channel prediction model to be trained extract features from the training samples to obtain channel features. The fully connected layer classifies the channel features to obtain the model output results.

[0081] In another feasible embodiment, refer to Figure 3 , Figure 3 This is a flowchart illustrating the training process of another channel prediction model to be trained, as described in the channel prediction model training method of this application. Figure 3The channel prediction model to be trained comprises convolutional layers, residual modules, feature fusion layers, multilayer perceptron mixers, and fully connected layers. Training samples are input into the channel prediction model. The convolutional layers and residual modules in the model extract features from the training samples to obtain the first feature. The model then upsamples the training samples to obtain the sampling result. The convolutional layers in the model extract features from the sampling result to obtain the second feature. The feature fusion layer fuses the first and second features to obtain the fused feature. The multilayer perceptron mixer extracts features from the fused feature to obtain the channel feature. Finally, the fully connected layer classifies the channel feature to obtain the model output.

[0082] In step S30, after the step of iteratively optimizing the channel prediction model to be trained based on the multiple training samples to obtain the channel prediction model, the method further includes:

[0083] Step S40: Collect the channel state information, channel data and base station prior information of each of the preset base stations at preset time intervals within the preset time interval;

[0084] It is understandable that, since the channel state information, channel data, and prior information of the base station change over time, if a model trained with training samples generated from fixed data is used to predict the channel state information, the accuracy of the prediction is likely to be low.

[0085] In this embodiment, it should be noted that the preset time is a pre-set interval for model updates, which can be 30 days, 15 days, or other times.

[0086] Optionally, the specific implementation of collecting the channel state information, channel data and base station prior information corresponding to each of the preset base stations within the preset time period can refer to the specific implementation process of step S10 above, and will not be repeated here.

[0087] Step S50: Take any time period within each of the preset time periods as the target time period, generate adjustment samples based on the channel state information, channel data, base station prior information, and usage environment scenario of each preset base station within the target time period, and iteratively optimize the number of multilayer perceptron layers in the multilayer perceptron mixer in the channel prediction model based on the adjustment samples.

[0088] In one feasible embodiment, based on the channel state information, channel data, base station prior information, and usage environment scenario of each preset base station within the target time period, adjustment samples corresponding to each usage environment scenario are generated. Each adjustment sample consists of input feature data and a real label corresponding to the input feature data. The input feature data includes the channel data and base station prior information of a preset base station within the target time period under a usage environment scenario. The real label includes the channel state information of the preset base station within the target time period under the usage environment scenario corresponding to the input feature data.

[0089] In another feasible embodiment, an adjustment sample consists of input feature data and a real label corresponding to the input feature data. The input feature data includes channel data of a preset base station in the target time period, base station prior information, and the corresponding usage environment scenario. The real label includes channel state information of the preset base station corresponding to the input feature data in the target time period.

[0090] In one feasible embodiment, based on the adjustment samples corresponding to each of the aforementioned usage scenarios, the number of multilayer perceptron layers in the multilayer perceptron mixer in the channel prediction model corresponding to each of the aforementioned usage scenarios is iteratively optimized.

[0091] It is understandable that, since the channel state information, channel data and prior information of the base station may change in real time, the adjustment samples corresponding to the same usage environment scenario at different times may also be different. As a result, the number of multilayer perceptron layers in the multilayer perceptron mixer in the channel prediction model obtained by iterating the adjustment samples corresponding to the same usage environment scenario at different times may also be different.

[0092] In another feasible embodiment, by inputting the input feature data of the adjusted sample into the channel prediction model to obtain the output label, the difference between the output label and the true label is determined. Based on the difference, the model loss corresponding to the channel prediction model to be trained is calculated, and then it is determined whether the model loss has converged. If the model loss has converged, the number of multilayer perceptron layers in the multilayer perceptron mixer in the channel prediction model is kept unchanged. If the model loss has not converged, the number of multilayer perceptron layers is adjusted, and the process returns to the step of inputting the input feature data of the adjusted sample into the channel prediction model to obtain the output label, until the model loss converges.

[0093] Optionally, in a feasible embodiment, after the step of iteratively optimizing the channel prediction model to be trained based on the multiple training samples to obtain the channel prediction model, the method further includes: if historical channel state information, historical channel data, and historical base station prior information of a new base station are detected, then the usage environment scenario of the new base station is determined; the channel prediction model is optimized based on the augmented samples generated from the historical channel state information, historical channel data, historical base station prior information, and the usage environment scenario of the new base station; or, the channel prediction model to be trained corresponding to the usage environment of the new base station is optimized based on the augmented samples to obtain the channel prediction model corresponding to the usage environment of the new base station.

[0094] As a feasible embodiment, refer to Figure 4 and Figure 5 , Figure 4 This is a flowchart illustrating the overall process of training a channel prediction model to be trained, as described in the channel prediction model training method of this application. Figure 5 This is a flowchart illustrating the overall process involved in the channel prediction model training method in the embodiments of this application. Figure 4 The system includes: preset base stations (base stations 1, 2, 3, and 4 shown in the diagram), newly added base stations, Hive, and a Kafka cluster. Channel state information, channel data, and prior information of each preset base station are collected, and the usage environment of each preset base station is determined. This information is stored in the Kafka cluster. Hive is used to classify the preset base stations according to their usage environment, generating training samples. Deep learning is then applied to the channel prediction model based on these training samples to obtain the channel prediction model. Every preset time interval, a program is deployed to collect the corresponding channel state information, channel data, and prior information of each preset base station within that preset time interval, thereby updating the channel prediction model. If historical channel state information, historical channel data, and historical prior information of a newly added base station are detected, the usage environment of that newly added base station is determined. Based on the augmented samples generated from the historical channel state information, historical channel data, historical prior information, and the usage environment of the newly added base station, the channel prediction model is optimized.

[0095] This application provides a channel prediction model training method. It involves collecting channel state information, channel data, and base station prior information for each preset base station at preset time intervals. Any time period within each preset time interval is taken as a target time period. Adjustment samples are generated based on the channel state information, channel data, base station prior information, and the usage environment of each preset base station within the target time period. The number of multilayer perceptron layers in the multilayer perceptron mixer of the channel prediction model is iteratively optimized based on the adjustment samples. This iterative optimization of the number of multilayer perceptron layers in the multilayer perceptron mixer of the channel prediction model at regular intervals ensures that the adopted channel prediction model matches the adjustment samples in real time, thereby improving the prediction accuracy of the channel prediction model.

[0096] Example 4

[0097] This application embodiment also provides a channel prediction model training device, the channel prediction model training device comprising:

[0098] The acquisition module is used to acquire historical channel status information, historical channel data and historical base station prior information of at least one preset base station in a historical time period, and to determine the usage environment scenario of each preset base station.

[0099] The generation module is used to generate multiple training samples based on the historical channel state information, historical channel data, base station prior information, and usage environment scenarios.

[0100] The iteration module is used to iteratively optimize the channel prediction model to be trained based on the multiple training samples to obtain the channel prediction model.

[0101] Optionally, the acquisition module is further configured to:

[0102] Take any one of the preset base stations as the target base station, and obtain the environmental information of the environment where at least one classified base station is located, the corresponding channel data, and the environmental information of the environment where the target base station is located;

[0103] Based on the channel data corresponding to each of the classified base stations and the historical channel data of the target base station, the first probability that the target base station is in a classified environment scenario corresponding to any of the classified base stations is predicted.

[0104] Based on the environmental information of the environment where the classified base stations are located and the environmental information of the environment where the target base station is located, a second probability is predicted that the target base station is in a classified environmental scenario corresponding to any of the classified base stations.

[0105] Take any one of the classified base stations as the classified selected base station, and calculate the total probability that the target base station is in the classified environment scenario corresponding to the classified selected base station based on the first probability that the target base station is in the classified environment scenario corresponding to the classified selected base station and the second probability that the target base station is in the classified environment scenario corresponding to the classified selected base station.

[0106] Among the classified base stations, select the target classified base station that maximizes the total probability, and use the classified environment scene corresponding to the target classified base station as the usage environment scene of the target base station.

[0107] Optionally, the environmental information includes at least one of environmental sound data, base station usage information, environmental meteorological information, and environmental location information, and the acquisition module is further configured to:

[0108] Obtain an environmental scene prediction model obtained by iterative optimization based on the environmental information of the environment where the classified base stations are located;

[0109] Based on the environmental information of the environment in which the target base station is located, the second probability of the target base station being in any of the classified environmental scenarios corresponding to the classified base station is predicted by the environmental scenario prediction model.

[0110] Optionally, the acquisition module is further configured to:

[0111] Determine the first weight corresponding to the first probability and the second weight corresponding to the second probability of the selected base station;

[0112] Calculate the first product between the first probability and the first weight of the selected base station and the second product between the second probability and the second weight, and use the sum of the first product and the second product as the total probability that the target base station is in the classified environment scenario corresponding to the selected base station.

[0113] Optionally, the generation module is further configured to:

[0114] Based on the historical channel state information, historical channel data, and historical base station prior information corresponding to each of the aforementioned usage environment scenarios, multiple training samples corresponding to each usage environment scenario are generated. Each training sample consists of input feature data and a real label corresponding to the input feature data. The input feature data includes historical channel data and historical base station prior information of a preset base station in a usage environment scenario within the historical time period. The real label includes historical channel state information of the preset base station in the usage environment scenario corresponding to the input feature data within the historical time period.

[0115] Optionally, the iteration module is further configured to:

[0116] Based on multiple training samples corresponding to each of the aforementioned usage scenarios, the channel prediction models to be trained for each usage scenario are iteratively optimized to obtain the channel prediction models corresponding to each usage scenario.

[0117] Optionally, after the step of iteratively optimizing the channel prediction model to be trained based on the multiple training samples to obtain the channel prediction model, the channel prediction model training device is further configured to:

[0118] At preset time intervals, channel state information, channel data and base station prior information corresponding to each preset base station are collected within the preset time interval;

[0119] Any time period within each of the preset time periods is taken as the target time period. Based on the channel state information, channel data, base station prior information, and usage environment scenario of each preset base station within the target time period, adjustment samples are generated. Based on the adjustment samples, the number of multilayer perceptron layers in the multilayer perceptron mixer in the channel prediction model is iteratively optimized.

[0120] The channel prediction model training apparatus provided in this application employs the channel prediction model training method described in the above embodiments, thus solving the technical problem of low generalization of the channel prediction model. Compared with the prior art, the beneficial effects of the channel prediction model training apparatus provided in this application are the same as those of the channel prediction model training method described in the above embodiments, and other technical features in this channel prediction model training apparatus are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0121] Example 5

[0122] This application provides an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the channel prediction model training method in the above embodiments.

[0123] The following is for reference. Figure 6The diagram illustrates a structural schematic of an electronic device suitable for implementing embodiments of the present disclosure. The electronic devices in the embodiments of the present disclosure may include, but are not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers (PDAs), tablet computers, portable media players (PMPs), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 6 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0124] like Figure 6 As shown, an electronic device may include a processing unit (such as a central processing unit, graphics processing unit, etc.) that can perform various appropriate actions and processes based on programs stored in ROM (Read-Only Memory) or programs loaded from storage devices into RAM (Random Access Memory). RAM also stores various programs and data required for the operation of the electronic device. The processing unit, ROM, and RAM are interconnected via a bus. Input / output (I / O) ports are also connected to the bus.

[0125] Typically, the following systems can be connected to I / O ports: input devices including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices including, for example, magnetic tapes, hard disks, etc.; and communication devices. Communication devices allow electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although electronic devices with various systems are shown in the figures, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems may be implemented alternatively.

[0126] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device, or installed from a ROM. When the computer program is executed by a processing device, it performs the functions defined above in the methods of embodiments of this disclosure.

[0127] The electronic device provided in this application employs the channel prediction model training method in the above embodiments, thus solving the technical problem of low generalization of the channel prediction model. Compared with the prior art, the beneficial effects of the electronic device provided in this application are the same as those of the channel prediction model training method provided in the above embodiments, and other technical features of the electronic device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0128] It should be understood that various parts of this disclosure can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments or examples.

[0129] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0130] Example 6

[0131] This embodiment provides a computer-readable storage medium having computer-readable program instructions stored thereon, the computer-readable program instructions being used to execute the channel prediction model training method in the above embodiment.

[0132] The computer-readable storage medium provided in this application embodiment may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, RAM, ROM, EPROM (Erasable Programmable Read Only Memory) or flash memory, optical fiber, CD-ROM (compact disc read-only memory), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0133] The aforementioned computer-readable storage medium may be included in an electronic device or may exist independently without being assembled into an electronic device.

[0134] The aforementioned computer-readable storage medium carries one or more programs. When the aforementioned one or more programs are executed by an electronic device, the electronic device causes the electronic device to: acquire historical channel state information, historical channel data, and historical base station prior information of at least one preset base station in a historical time period; and determine the usage environment scenario of each preset base station; generate multiple training samples based on each of the historical channel state information, each of the historical channel data, each of the base station prior information, and each of the usage environment scenarios; and iteratively optimize the channel prediction model to be trained based on the multiple training samples to obtain a channel prediction model.

[0135] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a LAN (Local Area Network) or a WAN (Wide Area Network)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0136] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0137] The modules described in the embodiments of this disclosure can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0138] The computer-readable storage medium provided in this application stores computer-readable program instructions for executing the channel prediction model training method described above, thus solving the technical problem of low generalization of the channel prediction model. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the channel prediction model training method provided in the above implementation, and will not be repeated here.

[0139] Example 7

[0140] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the channel prediction model training method described above.

[0141] The computer program product provided in this application solves the technical problem of low generalization of channel prediction models. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as the beneficial effects of the channel prediction model training method provided in the above embodiments, and will not be repeated here.

[0142] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent scope of this application.

Claims

1. A channel prediction model training method, characterized in that, The channel prediction model training method includes: The system acquires historical channel state information, historical channel data, and historical base station prior information for at least one preset base station within a historical time period, and determines the usage environment scenario of each preset base station. The base station prior information includes at least one of channel data time-frequency correlation information, channel data frequency domain information, and channel data noise amplitude information. Multiple training samples are generated based on the historical channel state information, historical channel data, historical base station prior information, and usage environment scenarios. Based on the multiple training samples, the channel prediction model to be trained is iteratively optimized to obtain the channel prediction model. The step of determining the usage environment scenario of each of the preset base stations includes: Take any one of the preset base stations as the target base station, and obtain the environmental information of the environment where at least one classified base station is located, the corresponding channel data, and the environmental information of the environment where the target base station is located; Based on the channel data corresponding to each of the classified base stations and the historical channel data of the target base station, the first probability that the target base station is in a classified environment scenario corresponding to any of the classified base stations is predicted. Based on the environmental information of the environment where the classified base stations are located and the environmental information of the environment where the target base station is located, a second probability is predicted that the target base station is in a classified environmental scenario corresponding to any of the classified base stations. Take any one of the classified base stations as the classified selected base station, and calculate the total probability that the target base station is in the classified environment scenario corresponding to the classified selected base station based on the first probability that the target base station is in the classified environment scenario corresponding to the classified selected base station and the second probability that the target base station is in the classified environment scenario corresponding to the classified selected base station. Among the classified base stations, select the target classified base station that maximizes the total probability, and use the classified environment scene corresponding to the target classified base station as the usage environment scene of the target base station.

2. The channel prediction model training method as described in claim 1, characterized in that, The environmental information includes at least one of the following: environmental sound data, base station usage information, environmental meteorological information, and environmental location information. The step of predicting the second probability that the target base station is in a classified environment scenario corresponding to any one of the classified base stations, based on the environmental information of the environment where the classified base stations are located and the environmental information of the environment where the target base station is located, includes: Obtain an environmental scene prediction model obtained by iterative optimization based on the environmental information of the environment where the classified base stations are located; Based on the environmental information of the environment in which the target base station is located, the second probability of the target base station being in any of the classified environmental scenarios corresponding to the classified base station is predicted by the environmental scenario prediction model.

3. The channel prediction model training method as described in claim 1, characterized in that, The step of calculating the total probability that the target base station is in the classified environment scenario corresponding to the classified selected base station based on the first probability that the target base station is in the classified environment scenario corresponding to the classified selected base station, and the second probability that the target base station is in the classified environment scenario corresponding to the classified selected base station, includes: Determine the first weight corresponding to the first probability and the second weight corresponding to the second probability of the selected base station; Calculate the first product between the first probability and the first weight of the selected base station and the second product between the second probability and the second weight, and use the sum of the first product and the second product as the total probability that the target base station is in the classified environment scenario corresponding to the selected base station.

4. The channel prediction model training method as described in claim 1, characterized in that, The step of generating multiple training samples based on the historical channel state information, historical channel data, historical base station prior information, and usage environment scenarios includes: Based on the historical channel state information, historical channel data, and historical base station prior information corresponding to each of the aforementioned usage environment scenarios, multiple training samples corresponding to each usage environment scenario are generated. Each training sample consists of input feature data and a real label corresponding to the input feature data. The input feature data includes historical channel data and historical base station prior information of a preset base station in a usage environment scenario within the historical time period. The real label includes historical channel state information of the preset base station in the usage environment scenario corresponding to the input feature data within the historical time period.

5. The channel prediction model training method as described in claim 4, characterized in that, The step of iteratively optimizing the channel prediction model to be trained based on the multiple training samples to obtain the channel prediction model includes: Based on multiple training samples corresponding to each of the aforementioned usage scenarios, the channel prediction models to be trained for each usage scenario are iteratively optimized to obtain the channel prediction models corresponding to each usage scenario.

6. The channel prediction model training method according to any one of claims 1 to 5, characterized in that, After the step of iteratively optimizing the channel prediction model to be trained based on the multiple training samples to obtain the channel prediction model, the method further includes: At preset time intervals, channel state information, channel data and base station prior information corresponding to each preset base station are collected within the preset time interval; Any time period within each of the preset time periods is taken as the target time period. Based on the channel state information, channel data, base station prior information, and usage environment scenario of each preset base station within the target time period, adjustment samples are generated. Based on the adjustment samples, the number of multilayer perceptron layers in the multilayer perceptron mixer in the channel prediction model is iteratively optimized.

7. A channel prediction model training device, characterized in that, The channel prediction model training device includes: The acquisition module is used to acquire historical channel state information, historical channel data and historical base station prior information of at least one preset base station in a historical time period, and to determine the usage environment scenario of each preset base station, wherein the base station prior information includes at least one of channel data time-frequency correlation information, channel data frequency domain information and channel data noise amplitude information; The generation module is used to generate multiple training samples based on the historical channel state information, historical channel data, historical base station prior information, and usage environment scenarios. The iteration module is used to iteratively optimize the channel prediction model to be trained based on the multiple training samples to obtain the channel prediction model. The acquisition module is also used for: Take any one of the preset base stations as the target base station, and obtain the environmental information of the environment where at least one classified base station is located, the corresponding channel data, and the environmental information of the environment where the target base station is located; Based on the channel data corresponding to each of the classified base stations and the historical channel data of the target base station, the first probability that the target base station is in a classified environment scenario corresponding to any of the classified base stations is predicted. Based on the environmental information of the environment where the classified base stations are located and the environmental information of the environment where the target base station is located, a second probability is predicted that the target base station is in a classified environmental scenario corresponding to any of the classified base stations. Take any one of the classified base stations as the classified selected base station, and calculate the total probability that the target base station is in the classified environment scenario corresponding to the classified selected base station based on the first probability that the target base station is in the classified environment scenario corresponding to the classified selected base station and the second probability that the target base station is in the classified environment scenario corresponding to the classified selected base station. Among the classified base stations, select the target classified base station that maximizes the total probability, and use the classified environment scene corresponding to the target classified base station as the usage environment scene of the target base station.

8. An electronic device, characterized in that, The electronic device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the steps of the channel prediction model training method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program for implementing a channel prediction model training method, which is executed by a processor to implement the steps of the channel prediction model training method as described in any one of claims 1 to 6.

Images