Communication link screening method, apparatus, device, and medium

Abstract

Embodiments of the present application disclose a communication link screening method, device, equipment and medium, the method comprising: an application server acquires terminal communication information, the terminal communication information is used to represent the network state perceived by the terminal; the terminal communication information is aligned with the terminal communication information acquired by other application servers to obtain target terminal communication information; a performance prediction model is obtained by performing model training based on federated learning between the target terminal communication information and the NWDAF of the connected core network, the NWDAF has pre-acquired key communication parameters used to represent the network performance related to the core network; the performance prediction result is obtained by predicting the communication quality performance of the communication link corresponding to the core network according to the performance prediction model, and the optimal communication link is screened according to the performance prediction results of each application server. The technical scheme of the embodiments of the present application improves the accuracy of communication link selection.

Description

Technical Field

[0001] This application relates to the field of communication technology, and more specifically, to a communication link screening method, a communication link screening device, an electronic device, and a computer-readable storage medium. Background Technology

[0002] With the rapid development of communication technology, higher requirements have been placed on the reliability, efficiency and intelligence of communication networks. In communication networks, multi-link core network switching technology is one of the important means to achieve efficient and reliable communication. However, how to select the most suitable communication link from multiple links has become an urgent problem to be solved. Summary of the Invention

[0003] Embodiments of this application provide a communication link filtering method, a communication link filtering device, an electronic device, a computer-readable storage medium, and a computer program product, which improve the accuracy of communication link selection.

[0004] Other features and advantages of this application will become apparent from the following detailed description, or may be learned in part from practice of this application.

[0005] In a first aspect, embodiments of this application provide a communication link filtering method applied to a communication system. The communication system includes at least two core networks and at least two application servers corresponding to each of the at least two core networks. The core networks and application servers are connected via communication links. The core network includes a Network Data Analysis Function (NWDAF) network element. The method is applied to the application server and includes: acquiring terminal communication information, which is used to characterize the network state perceived by the terminal; aligning the terminal communication information with terminal communication information acquired by other application servers to obtain target terminal communication information; training a performance prediction model based on federated learning using the target terminal communication information and the NWDAF of the connected core network, wherein the NWDAF has pre-acquired key communication parameters used to characterize the network performance related to the core network; predicting the communication quality performance of the communication link corresponding to the core network based on the performance prediction model to obtain a performance prediction result; and filtering the optimal communication link based on the performance prediction results of each application server.

[0006] Secondly, embodiments of this application also provide a communication link filtering device applied to a communication system. The communication system includes at least two core networks and at least two application servers corresponding to each of the at least two core networks. The core networks and application servers are connected via communication links. The core networks include a Network Data Analysis Function (NWDAF) network element. The device is configured on the application server. The device includes: an acquisition module for acquiring terminal communication information, which is used to characterize the network state perceived by the terminal; an alignment module for aligning the terminal communication information with terminal communication information acquired by other application servers to obtain target terminal communication information; a learning module for training a performance prediction model based on federated learning using the target terminal communication information and the NWDAF of the connected core network, wherein the NWDAF has pre-acquired key communication parameters used to characterize the network performance related to the core network; and a filtering module for predicting the communication quality performance of the communication links corresponding to the core network based on the performance prediction model to obtain performance prediction results, and filtering the optimal communication link based on the performance prediction results of each application server.

[0007] Thirdly, embodiments of this application provide an electronic device, including one or more processors; and a storage device for storing one or more computer programs, which, when executed by the one or more processors, cause the electronic device to implement the communication link filtering method as described above.

[0008] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor of an electronic device, causes the electronic device to perform the communication link filtering method as described above.

[0009] Fifthly, embodiments of this application provide a computer program product, including a computer program stored in a computer-readable storage medium, wherein a processor of an electronic device reads from and executes the computer program from the computer-readable storage medium, causing the electronic device to perform the communication link filtering method as described above.

[0010] In the technical solution provided in the embodiments of this application, in a multi-link core network scenario, the application server obtains terminal communication information, which is used to characterize the network state perceived by the terminal. The terminal communication information is then aligned with the terminal communication information obtained by other application servers to obtain target terminal communication information, ensuring data feature consistency and providing a unified input for federated learning. Subsequently, using the aligned target terminal communication information and key communication parameters collected by the NWDAF of the core network, a performance prediction model is trained through federated learning. During the federated learning process, multi-source data from the terminal side and the core network side are effectively integrated, enabling the performance prediction model to describe and evaluate the performance and quality of the links in detail, while avoiding direct sharing of raw data. Based on the performance prediction model, the communication quality performance of the corresponding communication links in the core network is predicted to obtain performance prediction results. The optimal communication link is then selected based on the performance prediction results of each application server, improving the accuracy of communication link selection and thus enhancing the overall network service quality and reliability.

[0011] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of one implementation environment involved in this application;

[0013] Figure 2 This is a flowchart illustrating a communication link filtering method in an exemplary embodiment of this application;

[0014] Figure 3 This is a flowchart illustrating an alignment process between servers, as shown in an exemplary embodiment of this application;

[0015] Figure 4 This is a schematic diagram illustrating a communication link filtering method as shown in an exemplary embodiment of this application;

[0016] Figure 5 This is a flowchart illustrating another communication link filtering method as shown in an exemplary embodiment of this application;

[0017] Figure 6 This is a flowchart illustrating a federated learning method as shown in an exemplary embodiment of this application;

[0018] Figure 7 This is a flowchart illustrating another communication link filtering method as shown in an exemplary embodiment of this application;

[0019] Figure 8 This is a structural block diagram illustrating a communication link screening device according to an exemplary embodiment of this application;

[0020] Figure 9 A schematic diagram of the structure of a computer system suitable for implementing the electronic device of the present application is shown. Detailed Implementation

[0021] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0022] In this application embodiment, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0023] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0024] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0025] It should also be noted that "multiple" as mentioned in this application refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0026] Please see Figure 1 , Figure 1This is a schematic diagram of an implementation environment involved in this application. The implementation environment includes at least two core networks 10, and at least two application servers (AS) 20 corresponding to each of the at least two core networks 10; the core networks and application servers are connected through a communication link, and the core networks include network data analytics function (NWDAF) elements.

[0027] NWDAF is used to provide specific network data analysis services to the network. A single NWDAF element can integrate both an Analytics Logical Function (AnLF) and a Model Training Logical Function (MTLF). The AnLF is responsible for model inference, providing general NWDAF service interfaces such as Nnwdaf_AnalyticsSubscription and Nnwdaf_AnalyticsInfo, and can generate analysis results (including static statistical data and dynamic inference results) based on requests from consumer elements. The MTLF is responsible for model training and can provide the trained model to the AnLF. The AnLF is the only consumer element for the services provided by the MTLF.

[0028] Application servers can implement user plane functions, namely the AS-Internet Protocol (IP) transport network-User plane function (UPF) interface.

[0029] In this embodiment, the application server is used to acquire terminal communication information at the terminal level, which is used to characterize the network state perceived by the terminal; the terminal communication information is aligned with the terminal communication information acquired by other application servers to obtain target terminal communication information; a federated learning-based model is trained based on the target terminal communication information and the key communication parameters obtained by the NWDAF of the corresponding core network to obtain a performance prediction model, whereby the key communication parameters are used to characterize the network performance related to the core network; the communication quality performance of the communication link with the core network is predicted based on the performance prediction model to obtain a performance prediction result, and the optimal communication link is selected based on the performance prediction results of each application server.

[0030] It should be noted that, Figure 1 The number of core network and application servers shown is merely illustrative; any number of core network and application servers can be used depending on actual needs.

[0031] It should be noted that in the specific implementation of this application, if the information and / or data involve objects, when the embodiments of this application are applied to specific products or technologies, permission or consent from the objects is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0032] The communication link filtering method provided in the embodiments of this application can be applied to multi-link scenarios in 5G, 6G and other networks.

[0033] The following details the various implementation details of the technical solutions in the embodiments of this application:

[0034] like Figure 2 As shown, Figure 2 This is a flowchart illustrating a communication link filtering method according to an embodiment of this application. This method can be applied to... Figure 1 In the implementation environment shown, this method can be executed by an application server. This communication link filtering method may include steps S210 to S240, which are described in detail below:

[0035] S210. Obtain terminal communication information, which is used to characterize the network status perceived by the terminal.

[0036] In this embodiment, the application server obtains terminal communication information at the terminal level, which is used to characterize the network state perceived by the terminal. The terminal communication information includes signal strength, signal-to-noise ratio, and reception quality, reflecting the terminal's perception of the network environment at a specific point in time.

[0037] Optionally, the terminal communication information includes at least one of the following: Reference Signal Receiving Power (RSRP), Received Signal Strength Indication (RSSI), Reference Signal Receiving Quality (RSRQ), and Signal to Interference plus Noise Ratio (SINR).

[0038] S220. Align the terminal communication information with the terminal communication information obtained by other application servers to obtain the target terminal communication information.

[0039] In this embodiment, different application servers collect data from their respective coverage areas or different terminals. Consequently, the terminal communication information obtained by each application server may differ. To ensure that the data can be compared and calculated in the same feature space and that the model parameters can be effectively aggregated in subsequent federated learning, it is necessary to ensure that the feature parameters used are consistent. Therefore, information alignment processing is required between the application servers. Through alignment processing, the differences between the servers due to data format, sampling frequency, annotation method, etc., are eliminated, ensuring that all parties use the same feature parameters to represent the same type of information.

[0040] For example, application server A obtains terminal communication information set 1, application server B obtains terminal communication information set 2, and application server C obtains terminal communication information set 3. Application servers A, B, and C need to interact to find common feature parameters from terminal communication information set 1, terminal communication information set 2, and terminal communication information set 3, and then obtain the target terminal communication information corresponding to application service A through the common feature parameters.

[0041] In one example, the application server performs privacy intersection with other applications to align the target terminal communication information. Optionally, it obtains a list of identifiers corresponding to the terminal communication information and performs privacy intersection with the identifier lists of other application servers to obtain a list of intersection identifiers. Based on the list of intersection identifiers, it extracts the corresponding target terminal communication information from the terminal communication information.

[0042] Here, the identifier refers to a unique identifier for terminal communication information, such as a terminal ID. Terminal communication information can include multiple identifiers. Therefore, to obtain the identifier list, the application server performs a privacy intersection operation with other application servers to find identifiers shared by their data without revealing their unique data. Figure 3 As shown, application server A sends a list of identifiers to another application server B. The other application server B calculates the intersection of its local identifier list and the received identifier list to obtain an intersection identifier list. This intersection identifier list contains identifiers shared by both servers. Then, application server A receives the transaction identifier list sent by the other application server B and uses it to extract the corresponding target terminal communication information from its local terminal communication information. Alternatively, an application server can also receive identifier lists from other application servers to determine the transaction identifier list.

[0043] In this embodiment of the application, by using privacy-based intersection, the parties can find common data without exposing their own complete datasets, which meets the requirements of data privacy and security.

[0044] Optionally, in 5G and 6G networks, communication links and devices typically span multiple different application domains. This means that the terminal communication information between application server A and application server B includes information from different domains, and this information may differ to some extent. Therefore, adversarial training methods can be employed to align data from different domains using a unified feature space. For example, application servers A and B can use Generative Adversarial Networks (GANs) or Autoencoders Adversarial Networks (AAEs) to automatically convert the terminal communication information into a standardized feature representation, resulting in a list of identifiers.

[0045] Optionally, application server A and application server B may use different feature identifiers, which may lead to inconsistencies during data integration. In order to ensure data consistency among all parties and to protect data privacy during the alignment process, in this embodiment of the application, the terminal communication information is hashed to obtain a communication information identifier, and noise items are randomly added to the communication information identifier to obtain an identifier list.

[0046] For example, application server A obtains terminal communication information S. A Application server B obtains terminal communication information S B Application server A uses consistent hashing to obtain the communication information identifier H(S). A Application server A in H(S) A Randomly add forged data items to obtain the identifier list H(S′) A )=H(S A )∪N A ; where N A These are fake data points generated based on Laplace noise or the exponential mechanism to obfuscate real data. The intensity of the noise is controlled by the privacy budget ∈. The smaller the privacy budget, the stronger the protection but the lower the accuracy. The privacy budget ∈ can be dynamically adjusted, for example, by determining it based on the size of the intersection data. If the intersection data is small, the noise intensity is reduced to improve the intersection accuracy; if the data size is large, the noise intensity is increased to enhance privacy protection.

[0047] Similarly, application server B obtains H(S′) B Then, servers A and B exchange their respective identifier lists to obtain the intersection identifier list I. A∩B Because of the presence of noise, the exchanged list of identifiers does not directly expose the real dataset, thus ensuring privacy and security.

[0048] Since the intersection identifier list may contain some fake positive data, in this embodiment of the application, the intersection identifier list is compared with the local communication information identifier to remove noise items and obtain the target intersection identifier list; the target terminal communication information corresponding to the target intersection identifier list is extracted from the terminal communication information.

[0049] For example, server A compares data in its local database, removing false positive data items caused by noise and retaining the true target intersection identifier I. final =I A∩B ∩H(S A ), and then from S A Extract the corresponding target terminal communication information.

[0050] S230. Based on the communication information of the target terminal and the NWDAF of the connected core network, a federated learning-based model is trained to obtain a performance prediction model. The NWDAF has pre-acquired key communication parameters used to characterize the network performance related to the core network.

[0051] In this embodiment, NWDAF pre-acquires key communication parameters, which are used to characterize network performance related to the core network.

[0052] Optionally, key communication parameters include core network parameters and communication link parameters. Communication link parameters mainly involve the transmission characteristics of signals in the physical medium, including signal transmission and reception, spectrum resource utilization, signal quality, and transmission speed; these parameters directly affect the basic performance of communication. Optionally, communication link parameters include at least one of the following: signal beam transmission angle, reception angle, bandwidth, delay, signal power, noise power, bit error rate (BER), signal-to-noise ratio (SNR), and frequency utilization. Core network parameters mainly involve the processing and transmission efficiency of data in the core part of the network, including protocol design, data scheduling, and forwarding; these parameters affect the overall performance of communication and the terminal experience; optional, core network parameters include at least one of the following: protocol efficiency, throughput, data transmission rate, and communication efficiency.

[0053] Among these, the signal beam transmission angle refers to the angle at which the transmitting antenna transmits a signal in space, affecting signal coverage and directionality; the receiving angle refers to the ability of the receiving antenna to receive a signal from a specific angle, affecting signal reception quality and anti-interference capability; bandwidth refers to the frequency range that the signal can pass through, and a larger bandwidth means more data can be transmitted; latency refers to the time required for information to travel from the sender to the receiver, including propagation delay, processing delay, queuing delay, and transmission delay, etc., and a shorter latency means faster transmission speed and a better user experience; signal power is related to communication distance and represents the signal transmission strength, and a larger signal power usually means a longer transmission distance or stronger anti-interference capability; noise power refers to the average noise level on the communication line, and a lower noise power indicates a cleaner communication environment and less interference to signal transmission; bit error rate (BER) refers to the ratio of the number of erroneous bits in the transmitted bits to the total number of transmitted bits, and a lower BER means higher transmission accuracy; signal-to-noise ratio (SNR) refers to the ratio of signal power to noise power, representing the signal strength and clarity, and a higher SNR means a clearer signal and better transmission effect. Throughput refers to the number of bits successfully transmitted per unit of time. Higher throughput indicates a stronger data processing capability. Data transmission rate can include bit rate and baud rate. Bit rate refers to the number of binary bits of data transmitted per unit of time in a communication system; a higher bit rate indicates faster data transmission. Baud rate refers to the number of waveform changes in a transmitted signal per unit of time in a communication system, also known as the modulation rate. Frequency utilization refers to the transmission speed within a unit of frequency band, reflecting the efficiency of spectrum resource utilization. Protocol efficiency refers to the ratio of the effective data bits in the transmitted data packet to the total length of the data packet; higher protocol efficiency indicates higher data transmission efficiency. Communication efficiency refers to the ratio of data frame transmission time to the total time used for sending messages.

[0054] In this embodiment, the application server and the NWDAF of the connected core network perform federated learning. Federated learning refers to allowing multiple participants (such as the application server and NWDAF) to collaboratively train a global model, i.e., a performance prediction model, without sharing the original data. Since the target terminal communication information is the network state at the terminal level and the key communication parameters are the network performance on the core network side, the target terminal communication information and key communication parameters are combined in the federated learning. The model training results in a performance prediction model that can describe and evaluate the performance and quality of the link in detail.

[0055] To comprehensively evaluate the performance and quality of the communication link, in addition to terminal-level information and key communication parameters on the core network side, participating base stations can be added during federated learning to incorporate base station-side information. The base station connects to the application server through the core network, and the base station-side information includes resource usage, such as spectrum utilization and power usage. High base station load may affect communication quality. Subsequently, the application server, NWDAF, and base station perform federated learning-based training to obtain a performance prediction model.

[0056] S240. Based on the performance prediction model, predict the communication quality performance of the communication links corresponding to the core network to obtain the performance prediction results, and select the optimal communication link based on the performance prediction results of each application server.

[0057] In this embodiment of the application application, the application server is connected to the core network through a communication link. The application server then predicts the communication quality performance of the corresponding communication link to obtain a performance prediction result. The performance prediction result includes performance indicators such as throughput, latency, bit error rate, signal-to-noise ratio, etc. Each application server predicts the performance prediction result of its own communication link through steps S210 to S240. Then, the optimal communication link is selected from multiple communication links by combining the performance prediction results of each application server.

[0058] In one example, the central server selects the optimal communication link from multiple communication links. Optionally, the application server sends the performance prediction results to the central server so that the central server can sort the communication links according to the key performance indicators in the performance prediction results of each application server and determine the optimal communication link according to the sorting results; and receives the optimal communication link sent by the central server.

[0059] The central server can be selected from multiple application servers through competition, or it can be a server other than the application servers. Each application server sends its performance prediction results to the central server, which then obtains the performance prediction results for each communication link. Key performance indicators (KPIs) are extracted from these KPIs. Since different application servers may use different units of measurement or scales, the KPIs need to be standardized. The communication links are then ranked according to these KPIs. During ranking, the central server can obtain a link quality score through a weighted sum of multiple indicators, and then select the link with the highest quality score as the optimal communication link. The weights of the performance indicators can be dynamically adjusted based on current network requirements and service types. Alternatively, the central server can also identify links that perform well across multiple indicators as the optimal communication link, and then send the link information of this optimal communication link to each application server.

[0060] In one example, the application server itself selects the most suitable communication link for the terminals in the covered area from multiple communication links. Optionally, it obtains the communication needs of the terminals in the area covered by the application server; and filters each communication link according to the terminal communication needs and the performance prediction results of each application server to determine the optimal communication link for the area.

[0061] Terminal communication requirements refer to the network needs of terminals, which depend on factors such as object behavior, application type (e.g., video streaming, real-time communication, IoT data transmission), and terminal type. The application server receives performance prediction results from other application servers and comprehensively considers the terminal communication requirements and the performance prediction results of each link. For example, if terminals in a region are conducting high-definition video conferencing, the application server selects the link that can provide high bandwidth and low latency as the optimal communication link for the region based on various performance prediction results. Similarly, other application servers can also filter out the optimal communication link for their respective regions. The application server itself selects the most suitable communication link for the terminals in its coverage area from multiple communication links based on the terminal communication requirements, thus achieving personalized filtering.

[0062] In one example, after selecting the optimal communication link, the application server can update its local routing or resource allocation policy. Meanwhile, since the base station is connected to the application server through the core network, the application server can also send the optimal communication link to the core network, which in turn sends the optimal communication link to the base station. The base station then adjusts its communication policy to transmit data through the optimal communication link.

[0063] like Figure 4 As shown, taking application server 1 as an example, application server 1 collects terminal communication information at the terminal level in the covered area. Application server 1 performs alignment processing with application servers 2 and 3 to obtain target terminal communication information. Application server 1 interacts with the NWDAF network element of core network 1 through federated learning to obtain performance prediction model 1. Application server 1 uses performance prediction model 1 to predict the performance of communication link 1 to obtain the performance prediction result. The performance prediction result is sent to the central server (not shown), and the optimal communication link is obtained. The optimal communication link information is sent to the base station through core network 1, and the base station performs subsequent communication processes.

[0064] In this embodiment, in a multi-link core network scenario, the application server acquires terminal communication information, which is used to characterize the network state perceived by the terminal. This terminal communication information is then aligned with terminal communication information acquired by other application servers to obtain target terminal communication information, ensuring data feature consistency and providing unified input for federated learning. Subsequently, using the aligned target terminal communication information and key communication parameters collected by the core network NWDAF, a performance prediction model is trained through federated learning. During the federated learning process, multi-source data from both the terminal and core network sides are effectively integrated, enabling the performance prediction model to meticulously describe and evaluate the performance and quality of the links. This avoids direct sharing of raw data. Furthermore, the performance prediction model predicts the communication quality performance of the corresponding communication links in the core network, yielding performance prediction results. Finally, based on the performance prediction results of each application server, the optimal communication link is selected, improving the accuracy of communication link selection and thus enhancing the overall network service quality and reliability.

[0065] This application also provides another communication link filtering method, which can be applied to... Figure 1 The implementation environment shown is illustrated using the example of the method being executed by an application server. Figure 5 As shown, Figure 2 The S230 in the original text has been expanded to S510 to S530. Detailed descriptions of S510 to S530 are as follows:

[0066] S510. Align the target terminal communication information with the key communication parameters obtained by the NWDAF network element to obtain training communication information.

[0067] The application server sends the list of identifiers for the target terminal's communication information to the NWDAF network element. The NWDAF network element then performs intersection processing on the received list of identifiers and the list of identifiers corresponding to the key communication parameters. Based on the result of the intersection processing, it extracts training communication information from the target terminal's communication information. For details, please refer to the aforementioned alignment processing process.

[0068] Optionally, the target terminal communication information includes RSRP, RSSI, RSRQ, and SINR, and key communication parameters include bandwidth, latency, etc. The same identifier for the target terminal communication information and key communication parameters can be GUTI (Globally Unique Temporary Identifier). GUTI is a globally unique temporary identifier, a temporary ID assigned to the terminal. The same identifier can also be a session ID or flow ID, or a timestamp or cell ID, etc. The GUTI and session ID are used at different layers of the network, ensuring that both the application server and the NWDAF network element can obtain them.

[0069] S520: Train the local model according to the training communication information to obtain model parameters, and send the model parameters to the central server so that the central server can aggregate the model parameters of the application server and the model parameters of the NWDAF network element to obtain the global model.

[0070] S530 receives the global model sent by the central server and updates the global model to obtain the performance prediction model.

[0071] like Figure 6 As shown, the application server uses training communication information to update the local model and updates the model parameters through the backpropagation algorithm. Similarly, the NWDAF network element uses the aligned key communication parameters to obtain the model parameters of the local model. In each round of model training, the application server and the NWDAF network element send their respective trained model parameters to the central server. The central server aggregates the parameters of all local models and merges the model updates of the application server and the NWDAF network element into a global model. This global model integrates the knowledge of all local models. Then, the central server sends the aggregated global model to the application server. The application server then uses the global model as the basis to continue updating to obtain the final performance prediction model.

[0072] In this system, the central server can be a trusted third-party server. The aggregation process of the central server can employ a weighted average. For example, the model parameters of the application server and NWDAF network elements are determined based on the corresponding training communication information and the amount of data of the aligned key communication parameters. The weights corresponding to the model parameters of the application server and the NWDAF network elements are then calculated and averaged. The larger the amount of data, the greater its impact on the final global model, and therefore its corresponding weight is larger. Furthermore, in 5G and 6G networks, the real-time status of communication links and devices changes very frequently. Due to differences in device and network environments, some devices may not be able to stably participate in federated learning. To address these changes, the central server can dynamically adjust the contribution (i.e., weight) of the application server and NWDAF network elements to the global model update based on their network status. For instance, application servers with better network status are assigned larger weights to their model parameters, while NWDAF network elements with poorer network status are assigned smaller weights to their model parameters, thus avoiding the influence of abnormal data on the model.

[0073] Optionally, in 5G and 6G networks, with the dynamic changes in devices and network environments, data privacy and security are particularly important. Therefore, when the application server trains the local model based on training communication information, if the difference in the core network performance within a preset time period exceeds a preset difference threshold, it indicates that the application server is in an unstable network environment. In this case, random noise is added to the training communication information to obtain the target training communication information. The local model is then trained based on the target training communication information to obtain model parameters, preventing the inference of the original data of a specific device through model updates. In this way, even if the application server is in an unstable network environment, privacy and security can still be guaranteed. The difference in network performance within the preset time period can be the difference in bandwidth, throughput, etc., within the preset time period.

[0074] In one example, to avoid noise affecting the accuracy of model parameters, the amount of noise added is determined based on the amount of training communication data; for example, the less training communication data there is, the less noise is added. In another example, the intensity and amount of noise added are fixed.

[0075] If the difference in network performance of the core network within a preset time period is less than a preset difference threshold, the model parameters can be obtained by directly training the local model based on the training communication information.

[0076] Optionally, the model parameters are obtained by training the local model based on the training communication information, including: obtaining the derived information and label data of the training communication information, wherein the derived information includes the rate of change and statistics of the training communication information within a preset time period; inputting the training communication information and the derived information into the local model and obtaining the sample prediction results output by the local model; and calculating the model parameters based on the sample prediction results and label data.

[0077] As described above, the training communication information includes RSRP, RSSI, RSRQ, and SINR. The rate of change (such as difference or average value) of each parameter is calculated within a preset time period. This rate of change is used to capture the dynamic trend of network state changes. Statistics (such as mean, variance, peak value, etc.) are calculated for the parameters within the preset time period. These statistics are used to construct a comprehensive feature describing the network state within the time period. Optionally, the interaction relationship between two parameters (such as the combined feature of SINR and RSRQ, which better reflects the link quality) can be used as derived information. This derived information enables the local model to respond to network state changes in a timely manner and optimize the communication link quality.

[0078] Training communication information and derived information are input into a local model, which can be a neural network, such as a deep feedforward neural network (DNN), a convolutional neural network (CNN), a temporal model such as a recurrent neural network (RNN), or a long short-term memory network (LSTM). Features are extracted from the training communication information and derived information to obtain the training prediction results output by the local model based on the extracted features. The model loss function is calculated based on the training prediction results and label data, and the model parameters are updated through gradient descent or the Adam optimizer.

[0079] Optionally, in this embodiment, federated learning can be initialized to set hyperparameters, including the learning rate and the number of iterations. The learning rate determines the step size of each parameter update, affecting the convergence speed and accuracy of the model. The number of iterations is the number of training rounds of the model. Each training round uses local data to train the model and then uploads its updates to the central server. Step S520, which involves training the local model based on training communication information to obtain model parameters, includes: initializing the learning rate and number of training rounds of federated learning, and training the local model based on the training communication information and the learning rate to obtain model parameters. The application server uses the training communication information as input, or uses the training communication information and its derived information as input, to train the local model at the initial learning rate.

[0080] After receiving the global model from the central server, the global model initially combines the network states of the terminal side and the core network side. The application server can then continue training based on this global model. However, in 5G and 6G networks, the real-time states of communication links and devices change very frequently (e.g., changes in bandwidth, latency, signal-to-noise ratio). Traditional federated learning methods typically assume relatively stable network and device states, which may not be effective in handling these rapid changes. Therefore, the federated learning in this application adopts a more dynamic training strategy that can adapt to changes in the network environment in real time and integrate learning results from different devices and network environments.

[0081] In one example, during training, the learning rate can be adaptive to adapt to changes under different network conditions; optionally, updating the global model to obtain the performance prediction model in step S530 includes: adjusting the learning rate according to the network changes of the core network and the number of training iterations to obtain the target learning rate; acquiring incremental data of training communication information, and updating the global model according to the target learning rate and the incremental data to obtain the performance prediction model.

[0082] During training, the learning rate is dynamically adjusted based on changes in the network. If the network environment is relatively stable (e.g., low latency, high signal-to-noise ratio), a larger learning rate can be used to accelerate training. If the network state fluctuates significantly, the learning rate is reduced to ensure a more stable model. The learning rate is also adjusted based on the number of training iterations. For example, as training progresses, the learning rate is gradually reduced to promote model convergence. If the number of training iterations exceeds a first preset number of iterations, the initial learning rate is reduced to obtain the first learning rate. If the number of training iterations exceeds a second preset number of iterations, the first learning rate is reduced to obtain the second learning rate. The second preset number of training iterations must be greater than the first preset number of iterations.

[0083] In one example, the learning rate is adjusted based on changes in the network to obtain a first candidate learning rate, and the learning rate is adjusted based on the number of training iterations to obtain a second candidate learning rate. The first and second candidate learning rates are then weighted and averaged to obtain the target learning rate.

[0084] In this embodiment, the model parameters of the global model are updated through incremental learning. That is, during the federated learning process, new terminal communication information is constantly generated. Therefore, the application server will obtain incremental data of training communication information, input the incremental data into the global model, update the model parameters of the global model at the target learning rate, and send the updated model parameters to the central server to obtain the new global model. Similarly, NWDAF also uploads the updated model parameters to the central server to reduce latency, so that the central server can update the global model in a timely manner to obtain the performance prediction model, enabling the model to continuously learn as the network changes.

[0085] In this embodiment, by dynamically adjusting the learning rate and introducing incremental data updates, the global model can reflect changes in network status in real time, helping the core network to perform efficient resource scheduling and link optimization.

[0086] The process of updating the global model is repeated until the initial training iteration count is reached. This involves acquiring incremental data, inputting it into the new global model, updating its parameters at a dynamically adjusted target learning rate, and sending the updated global model to the central server. This process continues until the training iteration count is reached. If the new global model's parameters converge (i.e., the parameter changes are minimal) and the training process reaches a stable state, the federated learning process terminates, and the new global model is used as the performance prediction model. If the new global model's parameters do not converge, federated learning continues to update the model parameters.

[0087] It is understandable that the application server and NWDAF may train at different speeds in federated learning. Therefore, through asynchronous training, the application server and NWDAF can update model parameters according to their own progress, instead of waiting for other nodes to complete, which can avoid redundancy in training time.

[0088] The implementation details of the technical solutions in the embodiments of this application are described from the perspective of NWDAF, including:

[0089] NWDAF obtains the key communication parameters of the core network and the parameter identifier list of the key communication parameters. It receives the identifier list of target terminal communication information sent by the application server connected to the core network. NWDAF determines the intersection of the identifier list and the parameter identifier list and sends the intersection list to the application server. At the same time, it extracts the target key communication parameters corresponding to the intersection list from the local key communication parameters, trains the local model based on the target key communication parameters to obtain model parameters, sends the model parameters to the central server, and receives the global model sent by the central server.

[0090] NWDAF then trains the global model further, obtains incremental data of key communication parameters of the target, updates the global model based on the incremental data, and sends the updated model parameters to the central server.

[0091] For details on the training process of the local model and the update process of the global model using NWDAF, please refer to the execution steps of the application server; these will not be repeated here.

[0092] It is worth noting that this application provides a method for communication link selection based on a multi-core network and multiple application servers. By collecting key communication parameters such as signal beam transmission angle, reception angle, bandwidth, and delay, and utilizing federated learning among different servers, the method predicts the communication quality performance of each link, thereby selecting suitable links for communication. This method effectively solves the problem of multi-link core network switching in 5G, 6G, and other communication networks, improving the reliability and efficiency of the communication network.

[0093] like Figure 7 As shown, the method for filtering communication links includes:

[0094] S701, the data acquisition module collects key communication parameters in real time through sensors and measuring devices deployed in the 6G network.

[0095] The data acquisition module is located in the core network.

[0096] After the S702 data acquisition module collects this information, it transmits it to the NWDAF modules of multiple core networks through multiple links. At the same time, the application server on the core network side collects UE-level communication information.

[0097] S703: Different application servers use federated learning sample alignment to align sample features to ensure that feature parameters are the same.

[0098] a) Application server A transmits a feature ID to application server B; this feature ID is the feature ID obtained by hash calculation of the communication information of each terminal.

[0099] b) Application servers B and A perform a privacy-preserving intersection operation and pass the intersection to application server A; application servers A and B use encryption algorithms to encrypt their respective feature ID lists; exchange the encrypted ID lists; and calculate the intersection of the encrypted ID lists through a specific protocol.

[0100] c) Application server A receives the intersection and uses the intersection features as the feature ID for subsequent federated learning, while retrieving the information corresponding to the intersection from its own dataset.

[0101] S704: Different application servers interact with the core network NWDAF network elements on their respective links to perform federated learning.

[0102] A: The application server and NWDAF perform privacy intersection calculations, meaning the information IDs collected by NWDAF need to be consistent with the information collected by the application server; B: The process of training and inferring the federated learning model is executed; C: The inference results are stored on the application server side. The federated learning process is as follows:

[0103] Federated learning process:

[0104] a) Initialization: Set parameters such as the number of iterations and learning rate for federated learning. Determine the type of model to be trained and set the learning rate, batch size, number of iterations, etc.

[0105] b) Data Distribution: Local data is distributed to the NWDAFs participating in federated learning according to certain rules. This data distribution is optional. The core objective of data distribution is to rationally allocate computational load, improve model training efficiency, and ensure that each participant completes learning without directly accessing data from other servers. Data is distributed according to different strategies, such as by region, device type, or task type, ensuring that the participating NWDAFs can collaboratively learn without directly accessing application server data. For example, an application server responsible for a certain area distributes data for that area to the relevant NWDAF network elements in that area. The NWDAFs only learn the wireless signal characteristics within that area and do not acquire data from other areas. To ensure data privacy protection, random noise may be added during data distribution, or encrypted computation may be used for data distribution.

[0106] c) Model training: The application server and NWDAF train a local model based on local data and upload the model parameters to the central server.

[0107] d) Model aggregation: The central server aggregates the received model parameters to generate a global model.

[0108] e) Model update: Distribute the global model to the application server and NWDAF for the next round of iteration training.

[0109] f) Termination condition: The federated learning process is terminated when the set number of iterations is reached or the model performance no longer improves.

[0110] S705: Different application servers are compared using the central server, and the optimal communication link is selected based on the simulated link performance results.

[0111] After completing federated learning, each application server has trained a shared global model based on local data and NWDAF. This model can be used to extrapolate the performance of its respective links. In order to select the optimal link in the entire network, it is necessary to compare the extrapolation results of each application server. Therefore, a central server is used to coordinate and compare the link performance results of different application servers and notify the relevant application servers of the selection results.

[0112] S706, the application server notifies the corresponding NWDAF, and sends the information to the base station through the core network on this link, and the base station performs subsequent communication processes.

[0113] The application server sends optimal link information to the relevant base stations through the NWDAF network element in the core network. The base stations adjust their communication strategies according to the instructions and use the optimal link for subsequent communication.

[0114] The following describes an apparatus embodiment of this application, which can be used to execute the communication link filtering method in the above embodiments of this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the communication link filtering method described above.

[0115] This application provides a communication link filtering device applied to a communication system. The communication system includes at least two core networks and at least two application servers corresponding to each of the core networks. The core networks and application servers are connected via communication links. The core networks include a Network Data Analysis Function (NWDAF) network element. The device is configured on the application server. Figure 8 As shown, the device includes:

[0116] The acquisition module 810 is used to acquire terminal communication information, which is used to characterize the network status perceived by the terminal.

[0117] Alignment module 820 is used to align the terminal communication information with terminal communication information obtained by other application servers to obtain target terminal communication information;

[0118] The learning module 830 is used to perform federated learning-based model training on the NWDAF of the connected core network based on the communication information of the target terminal to obtain a performance prediction model. The NWDAF has pre-acquired key communication parameters for characterizing the network performance related to the core network.

[0119] The filtering module 840 is used to predict the communication quality performance of the communication links corresponding to the core network according to the performance prediction model to obtain the performance prediction result, and to filter the optimal communication link according to the performance prediction result of each application server.

[0120] In one embodiment of this application, based on the foregoing scheme, the alignment module is further configured to obtain a list of identifiers corresponding to the terminal communication information, and perform privacy intersection processing on the identifier list and the identifier list of the other application server to obtain an intersection identifier list; and extract the corresponding target terminal communication information from the terminal communication information according to the intersection identifier list.

[0121] In one embodiment of this application, based on the aforementioned scheme, the alignment module is further configured to perform hash calculation on the terminal communication information to obtain a communication information identifier, and randomly add noise items to the communication information identifier to obtain the identifier list; compare the intersection identifier list with the local communication information identifier to remove the noise items to obtain a target intersection identifier list; and extract the target terminal communication information corresponding to the target intersection identifier list from the terminal communication information.

[0122] In one embodiment of this application, based on the foregoing scheme, the learning module is further configured to align the target terminal communication information with the key communication parameters obtained by the NWDAF network element to obtain training communication information; train the local model according to the training communication information to obtain model parameters, and send the model parameters to the central server so that the central server can aggregate the model parameters of the application server and the model parameters of the NWDAF network element to obtain a global model; receive the global model sent by the central server, and update the global model to obtain a performance prediction model.

[0123] In one embodiment of this application, based on the foregoing scheme, the learning module is further configured to add random noise to the training communication information to obtain target training communication information if the difference in network performance of the core network within a preset time period is greater than a preset difference threshold; and to train the local model according to the target training communication information to obtain model parameters.

[0124] In one embodiment of this application, based on the foregoing scheme, the learning module is further configured to acquire derived information and label data of the training communication information, wherein the derived information includes the rate of change and statistics of the training communication information within a preset time period; input the training communication information and the derived information into the local model, and acquire the training prediction result output by the local model; and calculate the model parameters based on the training prediction result and the label data.

[0125] In one embodiment of this application, based on the foregoing scheme, the learning module is further configured to initialize the learning rate and training iterations of the federated learning, and train the local model to obtain model parameters according to the training communication information and the learning rate; adjust the learning rate according to the network changes of the core network and the training iterations to obtain a target learning rate; acquire incremental data of the training communication information, and update the global model according to the target learning rate and the incremental data to obtain a performance prediction model.

[0126] In one embodiment of this application, based on the foregoing scheme, the learning module is further configured to input the incremental data into the global model, update the model parameters of the global model at the target learning rate, and send the updated model parameters to the central server to obtain a new global model; repeat the step of updating the global model until the number of iterations reaches the initialized number of training iterations; if the model parameters of the new global model converge, then the new global model is used as the performance prediction model.

[0127] In one embodiment of this application, based on the aforementioned scheme, the terminal communication information includes at least one of reference signal received power, received signal strength indication, reference signal received quality, and signal-to-interference-plus-noise ratio; the key communication parameters include core network parameters and communication link parameters, wherein the communication link parameters include at least one of signal beam transmission angle, reception angle, bandwidth, delay, signal power, noise power, bit error rate, signal-to-noise ratio, and frequency utilization; and the core network parameters include at least one of protocol efficiency, throughput, data transmission rate, and communication efficiency.

[0128] In one embodiment of this application, based on the foregoing scheme, the filtering module is further configured to send the performance prediction results to the central server, so that the central server sorts each communication link according to the key performance indicators in the performance prediction results of each application server, and determines the optimal communication link according to the sorting results; and receives the optimal communication link sent by the central server.

[0129] In one embodiment of this application, based on the foregoing scheme, the filtering module is further configured to obtain the terminal communication requirements of the area covered by the application server; and to filter each communication link according to the terminal communication requirements and the performance prediction results of each application server to determine the optimal communication link for the area.

[0130] In one embodiment of this application, based on the foregoing scheme, the device further includes a transmitting module for transmitting the optimal communication link to the core network, so that the core network transmits the optimal communication link to the base station, and the base station is used to adjust the communication strategy to transmit data through the optimal communication link.

[0131] It should be noted that the apparatus provided in the above embodiments and the method provided in the above embodiments belong to the same concept, and the specific way in which each module and unit performs operations has been described in detail in the method embodiments, and will not be repeated here.

[0132] The apparatus of this application embodiment acquires terminal communication information in a multi-link core network scenario. This terminal communication information is used to characterize the network state perceived by the terminal. The terminal communication information is aligned with terminal communication information acquired by other application servers to obtain target terminal communication information, ensuring data feature consistency and providing unified input for federated learning. Then, using the aligned target terminal communication information and key communication parameters collected by the NWDAF of the core network, a performance prediction model is trained through federated learning. In the federated learning process, multi-source data from the terminal side and the core network side are effectively integrated, enabling the performance prediction model to describe and evaluate the performance and quality of the links in detail, while avoiding direct sharing of raw data. Based on the performance prediction model, the communication quality performance of the communication links corresponding to the core network is predicted to obtain performance prediction results. The optimal communication link is selected based on the performance prediction results of each application server, improving the accuracy of communication link selection and thus enhancing the overall network service quality and reliability.

[0133] Embodiments of this application also provide an electronic device, including one or more processors and a storage device, wherein the storage device is used to store one or more computer programs, which, when executed by one or more processors, cause the electronic device to implement the communication link filtering method as described above.

[0134] Figure 9 A schematic diagram of the structure of a computer system suitable for implementing the electronic device of the present application is shown.

[0135] It should be noted that, Figure 9 The computer system 900 of the electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0136] like Figure 9 As shown, the computer system 900 includes a central processing unit (CPU) 901, which can perform various appropriate actions and processes, such as executing the methods described in the above embodiments, based on a program stored in read-only memory (ROM) 902 or a program loaded from storage portion 908 into random access memory (RAM) 903. The RAM 903 also stores various programs and data required for system operation. The CPU 901, ROM 902, and RAM 903 are interconnected via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0137] In some embodiments, the following components are connected to the I / O interface 905: an input section 906 including a keyboard, mouse, etc.; an output section 907 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 908 including a hard disk, etc.; and a communication section 909 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to the I / O interface 905 as needed. A removable medium 911, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 910 as needed so that computer programs read from it can be installed into the storage section 908 as needed.

[0138] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 909, and / or installed from removable medium 911. When the computer program is executed by processor (CPU) 901, it performs various functions defined in the system of this application.

[0139] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory, flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable computer program. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0140] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains 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 a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and a computer program.

[0141] The units or modules described in the embodiments of this application can be implemented in software or hardware, and can also be located in a processor. The names of these units or modules do not necessarily limit the specific unit or module itself.

[0142] Another aspect of this application provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method described above. This computer-readable storage medium may be included in the electronic device described in the above embodiments, or it may exist independently and not assembled into the electronic device.

[0143] Another aspect of this application provides a computer program product comprising a computer program stored in a computer-readable storage medium. A processor of an electronic device reads the computer program from the computer-readable storage medium and executes the computer program, causing the electronic device to perform the methods described above in the various embodiments.

[0144] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0145] Other embodiments of this application will readily conceive of by considering the specification and practicing the embodiments disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0146] The above content is merely a preferred exemplary embodiment of this application and is not intended to limit the implementation of this application. Those skilled in the art can easily make corresponding modifications or alterations based on the main concept and spirit of this application. Therefore, the scope of protection of this application should be determined by the scope of protection claimed in the claims.

Claims

1. A communication link filtering method, characterized in that, The method is applied to a communication system, which includes at least two core networks and at least two application servers corresponding to each of the core networks. The core networks and application servers are connected via a communication link. The core networks include a Network Data Analysis Function (NWDAF) network element. The method is applied to the application servers and includes: Acquire terminal communication information, which is used to characterize the network status perceived by the terminal; The terminal communication information is aligned with the terminal communication information obtained by other application servers to obtain the target terminal communication information. A performance prediction model is obtained by training a federated learning model based on the target terminal communication information and the NWDAF of the connected core network. The NWDAF has pre-acquired key communication parameters for characterizing the network performance related to the core network. The performance prediction model is used to predict the communication quality performance of the communication links corresponding to the core network to obtain the performance prediction results, and the optimal communication link is selected based on the performance prediction results of each application server.

2. The method according to claim 1, characterized in that, The step of aligning the terminal communication information with the terminal communication information of other application servers to obtain the target terminal communication information includes: Obtain the identifier list corresponding to the terminal communication information, and perform privacy intersection processing on the identifier list and the identifier list of the other application servers to obtain the intersection identifier list; Extract the corresponding target terminal communication information from the terminal communication information based on the intersection identifier list.

3. The method according to claim 2, characterized in that, The step of obtaining the identifier list corresponding to the terminal communication information includes: The terminal communication information is hashed to obtain a communication information identifier, and noise items are randomly added to the communication information identifier to obtain the identifier list; Extracting the corresponding target terminal communication information from the terminal communication information based on the intersection identifier list includes: The intersection identifier list is compared with the local communication information identifier to remove the noise items and obtain the target intersection identifier list; Extract the target terminal communication information corresponding to the target intersection identifier list from the terminal communication information.

4. The method according to claim 1, characterized in that, The step of training a performance prediction model based on federated learning using the target terminal's communication information and the NWDAF network elements of the connected core network includes: The target terminal communication information is aligned with the key communication parameters obtained by the NWDAF network element to obtain training communication information; The local model is trained according to the training communication information to obtain model parameters, and the model parameters are sent to the central server so that the central server can aggregate the model parameters of the application server and the model parameters of the NWDAF network element to obtain a global model. The system receives the global model sent by the central server and updates the global model to obtain a performance prediction model.

5. The method according to claim 4, characterized in that, The step of training the local model based on the training communication information to obtain model parameters includes: If the difference in network performance of the core network within a preset time period is greater than a preset difference threshold, random noise is added to the training communication information to obtain the target training communication information. The local model is trained based on the target training communication information to obtain model parameters.

6. The method according to claim 4, characterized in that, The step of training the local model based on the training communication information to obtain model parameters includes: Obtain derived information and label data of the training communication information, wherein the derived information includes the rate of change and statistics of the training communication information within a preset time period; The training communication information and the derived information are input into the local model, and the training prediction results output by the local model are obtained; The model parameters are calculated based on the training prediction results and the label data.

7. The method according to claim 4, characterized in that, The step of training the local model based on the training communication information to obtain model parameters includes: Initialize the learning rate and training iterations for federated learning, and train the local model to obtain model parameters based on the training communication information and the learning rate; The global model is updated to obtain a performance prediction model, including: The target learning rate is obtained by adjusting the learning rate based on the network changes of the core network and the number of training iterations. The incremental data of the training communication information is obtained, and the global model is updated according to the target learning rate and the incremental data to obtain the performance prediction model.

8. The method according to claim 7, characterized in that, The step of updating the global model based on the target learning rate and the incremental data to obtain the performance prediction model includes: The incremental data is input into the global model, the model parameters of the global model are updated at the target learning rate, and the updated model parameters are sent to the central server to obtain a new global model. Repeat the steps to update the global model until the number of iterations reaches the initial number of training iterations; If the model parameters of the new global model converge, then the new global model will be used as the performance prediction model.

9. The method according to claim 1, characterized in that, The terminal communication information includes at least one of the following: reference signal received power, received signal strength indication, reference signal received quality, and signal-to-interference-plus-noise ratio; The key communication parameters include core network parameters and communication link parameters. The communication link parameters include at least one of the following: signal beam transmission angle, reception angle, bandwidth, delay, signal power, noise power, bit error rate, signal-to-noise ratio, and frequency utilization. The core network parameters include at least one of the following: protocol efficiency, throughput, data transmission rate, and communication efficiency.

10. The method according to any one of claims 1 to 9, characterized in that, The step of selecting the optimal communication link based on the prediction results of each application server includes: The performance prediction results are sent to the central server, so that the central server can sort the communication links according to the key performance indicators in the performance prediction results of each application server, and determine the optimal communication link according to the sorting results. Receive the optimal communication link sent by the central server.

11. The method according to any one of claims 1 to 9, characterized in that, The step of selecting the optimal communication link based on the prediction results of each application server includes: Obtain the terminal communication requirements of the area covered by the application server; Based on the terminal communication requirements and the performance prediction results of each application server, each communication link is screened to determine the optimal communication link for the region.

12. The method according to any one of claims 1 to 9, characterized in that, After selecting the optimal communication link based on the performance prediction results of each application server, the method further includes: The optimal communication link is sent to the core network, so that the core network sends the optimal communication link to the base station, and the base station is used to adjust the communication strategy to transmit data through the optimal communication link.

13. A communication link screening device, characterized in that, This device is applied to a communication system, which includes at least two core networks and at least two application servers corresponding to each of the core networks. The core networks and application servers are connected via a communication link. The core networks include a Network Data Analysis Function (NWDAF) network element. The device is configured on the application server and includes: The acquisition module is used to acquire terminal communication information, which is used to characterize the network status perceived by the terminal. The alignment module is used to align the terminal communication information with the terminal communication information obtained by other application servers to obtain the target terminal communication information. The learning module is used to train a performance prediction model based on federated learning with the NWDAF of the connected core network according to the communication information of the target terminal. The NWDAF has pre-acquired key communication parameters for characterizing the network performance related to the core network. The filtering module is used to predict the communication quality performance of the communication links corresponding to the core network according to the performance prediction model to obtain the performance prediction results, and to filter the optimal communication links according to the performance prediction results of each application server.

14. An electronic device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, cause the electronic device to perform the method of any one of claims 1 to 12.

15. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by the processor of the electronic device, causes the electronic device to perform the method of any one of claims 1 to 12.

Images