Model management method and apparatus, and computer device

Abstract

The present application relates to a model management method and apparatus, and a computer device. The method comprises: receiving a first configuration message sent by a base station, wherein the first configuration message includes user scheduling result indication information based on hierarchical federated learning, which user scheduling result indication information is obtained by the base station on the basis of user characteristic information of each terminal device, the user characteristic information comprising at least one of user data characteristic information and user channel state information; sending report information to a network, wherein the network stores a hierarchical federated learning model and a delivery policy, which have been generated by the base station, the report information at least comprising at least one of a terminal capability, an inference task type and model performance monitoring feedback information, and the report information being used for the network to generate a second configuration message, which is model inference policy delivery information or model update policy delivery information based on hierarchical federated learning; and on the basis of the second configuration information, obtaining a model inference result or a model update result.

Description

Model management methods, devices and computer equipment

[0001] Related applications

[0002] This application claims priority to Chinese patent application filed on September 25, 2024, with application number 2024113442757, entitled "Model Management Method, Apparatus and Computer Equipment Based on Hierarchical Federated Learning", the entire contents of which are incorporated herein by reference. Technical Field

[0003] This application relates to the field of federated learning technology, and in particular to a model management method, apparatus, and computer equipment. Background Technology

[0004] With the rapid development of 3GPP Radio Access Network (RAN) technology, the powerful representation and feature extraction capabilities of AI models can be combined with wireless communication scenarios to achieve time-frequency domain resource optimization and overhead reduction. Federated learning, as a distributed framework, can effectively improve the training efficiency of AI models while protecting the privacy and security of local data, reducing transmission overhead. Summary of the Invention

[0005] In a first aspect, this application provides a model management method applied to terminal devices in a mobile communication system. The mobile communication system further includes a network, which comprises network nodes and multiple base stations. Each base station corresponds to multiple terminal devices. The terminal devices are distributed nodes performing hierarchical federated learning, and the base stations are central nodes performing hierarchical federated learning. The method includes:

[0006] The base station receives a first configuration message, which includes user scheduling result indication information based on hierarchical federated learning. The user scheduling result indication information is obtained by the base station based on user feature information of each terminal device. The user feature information includes at least one of user data feature information and user channel state information.

[0007] The network sends reporting information to the network, which stores the hierarchical federated learning model and distribution strategy generated by the base station. The reporting information includes at least one of terminal capabilities, inference task type, and model performance monitoring feedback information. The reporting information is used by the network to generate a second configuration message, which is a model inference strategy distribution message or a model update strategy distribution message based on hierarchical federated learning.

[0008] Receive the second configuration message, and obtain the model inference result or model update result based on the second configuration message.

[0009] In one embodiment, the terminal capability includes one or more of terminal computing capability, terminal channel state, and inference features, wherein the terminal computing capability is used to characterize the model complexity supported by the terminal device, and the terminal channel state characterizes the uplink channel state of the terminal device.

[0010] In one embodiment, sending the reporting information to the network includes:

[0011] Based on the target transmission method, reporting information is sent to the base station so that the base station can send reporting information to the network node. The target transmission method includes at least one of system message, RRC signaling, MAC CE signaling, DCI signaling, and NAS message.

[0012] In one embodiment, the model performance monitoring feedback information includes one or more of the following: system performance indicators, accuracy indicators, complexity indicators, and generalization performance indicators.

[0013] In one embodiment, the method further includes:

[0014] If it is determined that the model performance monitoring feedback information and the preset indicator threshold meet the preset update conditions, a model update request is sent to the base station so that the network can respond to the model update request. The model update request includes one or more of the following: the model to be updated, the function identifier to be updated, and reference information.

[0015] In one embodiment, receiving the second configuration message includes:

[0016] The system receives a second configuration message sent by the base station based on a target transmission mode. The second configuration message is sent by the network node to the base station. The target transmission mode includes at least one of system messages, RRC signaling, MAC CE signaling, DCI signaling, and NAS messages.

[0017] In one embodiment, the second configuration message is a model update strategy distribution information, which includes the model to be updated and the update strategy; obtaining the model inference result or the model update result based on the second configuration message includes:

[0018] The model to be updated is updated based on the update strategy to obtain the model update result. The updated model includes the model to be updated / group of models to be updated. The update strategy includes one or more of the following: local update strategy, global update strategy, model replacement strategy, and model deletion strategy.

[0019] In one embodiment, the update process includes at least one of the following:

[0020] Update certain modules of the updated model to obtain the model update result;

[0021] Based on the target update algorithm, the model parameters of each of the models to be updated are updated to obtain the model update result. The target update algorithm includes one or more of the following: momentum gradient descent algorithm, stochastic gradient descent algorithm, and mini-batch gradient descent algorithm.

[0022] The model parameters of each of the models to be updated are updated based on the new model parameters to obtain the model update results;

[0023] Delete the model parameters and model content / function identifiers of the updated model to obtain the model update result.

[0024] In one embodiment, the second configuration message is model inference strategy delivery information, which includes the model identifier of the model to be inferred, the model file, and the identifier information of the terminal device to perform model inference; obtaining the model inference result or model update result based on the second configuration message includes:

[0025] Based on the model identifier, model file, and identifier information of the terminal device to be inferred, online or offline model inference is performed to obtain the model inference result.

[0026] Secondly, embodiments of this application provide a model management method applied to network nodes in a mobile communication system. The network includes network nodes and multiple base stations, each base station corresponding to multiple terminal devices. The method includes:

[0027] The terminal device receives reporting information sent through the base station, wherein the reporting information includes at least one of the following: terminal capability, inference task type, and model performance monitoring feedback information.

[0028] A second configuration message is generated based on the reported information and sent to the terminal device through the base station. The second configuration message is a model inference strategy distribution information or a model update strategy distribution information based on hierarchical federated learning. The second configuration message is used to enable the terminal device to obtain the model inference result or the model update result.

[0029] In one embodiment, the terminal capability includes one or more of terminal computing capability, terminal channel state, and inference features, wherein the terminal computing capability is used to characterize the model complexity supported by the terminal device, and the terminal channel state characterizes the uplink channel state on the terminal device side.

[0030] In one embodiment, the reported information includes the terminal capability, and receiving the reported information sent by the terminal device through the base station includes:

[0031] Based on the target transmission mode, the terminal capabilities reported by the terminal device through the base station are received. The target transmission mode includes at least one of system messages, RRC signaling, MAC CE signaling, DCI signaling, and NAS messages.

[0032] In one embodiment, the method further includes:

[0033] The system receives a model inference request signaling sent by the terminal device through a base station and returns a response message corresponding to the inference request, so that the terminal device can send reporting information based on the response message.

[0034] In one embodiment, the second configuration message includes information on model inference strategy distribution or model update strategy distribution based on hierarchical federated learning; generating the second configuration message based on the reported information includes:

[0035] Based on the reported information, among multiple first models, at least one inference model whose target dimension satisfies the preset matching conditions is determined.

[0036] Based on the model identifier, model file, and identifier information of the terminal device to be inferred, a model inference strategy distribution information based on hierarchical federated learning is generated.

[0037] In one embodiment, determining at least one inference model whose target dimension satisfies preset matching conditions among multiple first models based on the reported information includes:

[0038] Based on the reported information, among multiple first models, a preset number of inference models whose model structure, model complexity, and model accuracy meet preset matching conditions are determined.

[0039] In one embodiment, sending the second configuration message to the terminal device via the base station includes:

[0040] The second configuration message is sent to the base station through the user plane interface, so that the base station sends the second configuration message to the terminal device through the target transmission mode, wherein the target transmission mode includes at least one of system message, RRC signaling, MAC CE signaling, DCI signaling, and NAS message.

[0041] In one embodiment, the model performance monitoring feedback information includes one or more of the following: system performance indicators, accuracy indicators, complexity indicators, and generalization performance indicators.

[0042] In one embodiment, the reported information includes model performance monitoring feedback information, and the generation of a second configuration message based on the reported information includes:

[0043] Receive model performance monitoring feedback information;

[0044] Receive a model update request, the model update request includes the identifier of the model / function to be updated and reference information, the model update request is generated when the model performance monitoring feedback information of the terminal device and the preset index threshold meet the preset update conditions;

[0045] The system responds to the model update request by generating a second configuration message.

[0046] In one embodiment, the reported information includes model performance monitoring feedback information, and the generation of a second configuration message based on the reported information includes:

[0047] Receive model performance monitoring feedback information;

[0048] A model update event is triggered based on a preset time period; or, if the model performance monitoring feedback information and the preset indicator threshold meet the preset update conditions, an update event is triggered; the preset time period is associated with terminal capabilities and auxiliary information.

[0049] Based on the triggered update event, a second configuration message is generated.

[0050] In one embodiment, the second configuration message includes information on the model inference strategy based on hierarchical federated learning; generating the second configuration message includes:

[0051] Identify at least one model to be updated;

[0052] Based on the model to be updated, the identification information of the terminal device to be updated, and the update strategy, a model inference strategy distribution information based on hierarchical federated learning is generated.

[0053] In one embodiment, the update strategy includes one or more of the following: a local update strategy, a global update strategy, a model replacement strategy, and a model deletion strategy.

[0054] In one embodiment, the method further includes:

[0055] Receive model data sent by each of the base stations, wherein the model data includes model files and model content / function identifiers;

[0056] Each first model is determined based on the model data, and then arranged according to the evaluation index of each first model to obtain a first model list.

[0057] In one embodiment, the method further includes:

[0058] The system receives identification requests and model capabilities sent by the base station. The identification requests include model identification requests and function identification requests. The model capabilities are used to indicate the models / functions supported by the base station.

[0059] In one embodiment, the method further includes:

[0060] An identification request is sent to each of the base stations to enable each base station to report its model capabilities, which are used to indicate the models / functions supported by the base station.

[0061] Thirdly, embodiments of this application provide a model management method based on hierarchical federated learning, applied to base stations in a mobile communication system. The mobile communication system includes a network and multiple base stations, each base station corresponding to multiple terminal devices. The terminal devices are distributed nodes performing hierarchical federated learning, and the base stations are central nodes performing hierarchical federated learning. The method includes:

[0062] Based on the user feature information of each terminal device, user scheduling result indication information corresponding to each terminal is obtained, and a first configuration message is sent to each terminal device respectively. The user feature information includes at least one of user data feature information and user channel state information. The first configuration message contains the user scheduling result indication information.

[0063] The terminal device receives reporting information, which includes at least one of the following: terminal capabilities, inference task type, and model performance monitoring feedback information.

[0064] The reported information is sent to the network node so that the network node receives a second configuration message; the second configuration message is a model inference strategy distribution message or a model update strategy distribution message based on hierarchical federated learning, and the second configuration message is used to enable the terminal device to obtain the model inference result or the model update result.

[0065] In one embodiment, the method further includes:

[0066] The network node sends model data, which includes model files and model content / function identifiers, so that the network node can determine each first model and obtain a first model list based on the model data.

[0067] In one embodiment, the method further includes:

[0068] The network node is sent an identification request and a model capability request. The identification request includes a model identification request and a function identification request. The model capability is used to indicate the model / function supported by the base station.

[0069] In one embodiment, the method further includes:

[0070] The base station receives an identification request sent by the network node and reports model capabilities to the network node based on the identification request. The model capabilities are used to indicate the models / functions supported by the base station.

[0071] In one embodiment, the method further includes:

[0072] Based on the model training results sent by each of the terminal devices, a trained global model is obtained.

[0073] In one embodiment, the terminal device is a user scheduling terminal device, and the step of obtaining the trained global model based on the model training results sent by each of the terminal devices includes:

[0074] For the i-th round of training, the model parameters of the global model in the i-th round are obtained based on the training results sent by each terminal device in the (i-1)-th round.

[0075] If the preset training completion conditions are not met, the model parameters of the global model in the i-th round are broadcast to each of the terminal devices so that each of the terminal devices can train based on the model parameters of the global model in the i-th round until the base station determines that the preset training completion conditions are met and obtains the trained global model.

[0076] In one embodiment, the preset training completion condition includes one or more of the following: the number of training rounds reaches a preset upper limit, the loss function of the global model is less than or equal to a preset value, and the difference between the loss function of the global model in two adjacent rounds is less than or equal to a preset value.

[0077] In one embodiment, the method further includes:

[0078] Based on the user feature information of each terminal device, user scheduling result indication information corresponding to each terminal is obtained, wherein the user feature information includes at least one of user data feature information and user channel state information;

[0079] A first configuration message is sent to each of the terminal devices, the first configuration message containing the user scheduling result indication information.

[0080] In one embodiment, obtaining user scheduling result indication information corresponding to each terminal based on the user feature information of each terminal device includes:

[0081] Get the pre-configured number of scheduled users;

[0082] Among all users within the coverage area of ​​the base station, based on at least one of the user data feature information and user channel state information of each terminal device, the terminal devices that are scheduled for the specified number of users are selected, and the user scheduling result indication information is determined as the scheduled indication information.

[0083] Fourthly, embodiments of this application provide a model inference apparatus applied to a terminal device in a mobile communication system. The mobile communication system further includes a network, which comprises network nodes and multiple base stations. Each base station corresponds to multiple terminal devices. The terminal devices are distributed nodes performing hierarchical federated learning, and the base stations are central nodes performing hierarchical federated learning. The apparatus includes:

[0084] The first receiving module is configured to receive a first configuration message sent by the base station. The first configuration message includes user scheduling result indication information based on hierarchical federated learning. The user scheduling result indication information is obtained by the base station based on user feature information of each terminal device. The user feature information includes at least one of user data feature information and user channel state information.

[0085] The first sending module is used to send reporting information to the network. The network stores the hierarchical federated learning model and distribution strategy generated by the base station. The reporting information includes at least one of terminal capabilities, inference task type, and model performance monitoring feedback information. The reporting information is used by the network to generate a second configuration message. The second configuration message is a model inference strategy distribution message or a model update strategy distribution message based on hierarchical federated learning.

[0086] The second receiving module is used to receive the second configuration message and obtain the model inference result or the model update result based on the second configuration message.

[0087] Fifthly, embodiments of this application provide a model management device applied to a network node in a mobile communication system. The network includes network nodes and multiple base stations, each base station corresponding to multiple terminal devices. The device includes:

[0088] The third receiving module is used to receive the reporting information sent by the terminal device through the base station. The reporting information includes at least one of the following: terminal capability, inference task type, and model performance monitoring feedback information.

[0089] The first generation module is used to generate a second configuration message based on the reported information, and send the second configuration message to the terminal device through the base station; the second configuration message is a model inference strategy distribution information or a model update strategy distribution information based on hierarchical federated learning, and the second configuration message is used to enable the terminal device to obtain the model inference result or the model update result.

[0090] Sixthly, embodiments of this application provide a model management device applied to a base station in a mobile communication system. The mobile communication system includes a network and multiple base stations, each base station corresponding to multiple terminal devices. The terminal devices are distributed nodes performing hierarchical federated learning, and the base stations are central nodes performing hierarchical federated learning. The device includes:

[0091] The second sending module is used to obtain user scheduling result indication information corresponding to each terminal based on the user feature information of each terminal device, and send a first configuration message to each terminal device respectively. The user feature information includes at least one of user data feature information and user channel state information, and the first configuration message contains the user scheduling result indication information.

[0092] The fourth receiving module is used to receive the reporting information sent by the terminal device, wherein the reporting information includes at least one of the following: terminal capability, inference task type, and model performance monitoring feedback information.

[0093] The third sending module is used to send the reported information to the network node so that the network node receives the second configuration message; the second configuration message is a model inference strategy distribution information or a model update strategy distribution information based on hierarchical federated learning, and the second configuration message is used to enable the terminal device to obtain the model inference result or the model update result.

[0094] In a seventh aspect, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps described in the above embodiments.

[0095] Eighthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps described in the above embodiments.

[0096] Ninthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps described in the above embodiments. Attached Figure Description

[0097] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0098] Figure 1 shows the application environment of a model management method based on hierarchical federated learning in one embodiment;

[0099] Figure 2 is a flowchart illustrating a model management method based on hierarchical federated learning in one embodiment;

[0100] Figure 3 is a flowchart illustrating the steps for obtaining model update results in one embodiment;

[0101] Figure 4 is a flowchart illustrating the sending steps in one embodiment;

[0102] Figure 5 is a flowchart illustrating the step of generating the second configuration message in one embodiment;

[0103] Figure 6 is a flowchart illustrating the step of generating the second configuration message in one embodiment;

[0104] Figure 7 is a flowchart illustrating the step of generating the second configuration message in one embodiment;

[0105] Figure 8 is a flowchart illustrating the steps for obtaining the model list in one embodiment;

[0106] Figure 9 is a flowchart illustrating the information transmission steps in one embodiment;

[0107] Figure 10 is a flowchart illustrating the training steps in one embodiment;

[0108] Figure 11a is a flowchart illustrating a model management method based on hierarchical federated learning in another embodiment;

[0109] Figure 11b is a schematic diagram of the positioning accuracy curve of a scheme in one embodiment;

[0110] Figure 11c is a schematic diagram of the localization error variation trend of the federated learning algorithm in one embodiment;

[0111] Figure 11d is a schematic diagram of the CDF curve of the positioning error in one embodiment;

[0112] Figure 12 is a structural block diagram of a model inference device based on hierarchical federated learning in one embodiment;

[0113] Figure 13 is a structural block diagram of a model inference device based on hierarchical federated learning in one embodiment;

[0114] Figure 14 is a structural block diagram of a model inference device based on hierarchical federated learning in one embodiment;

[0115] Figure 15 is an internal structure diagram of a computer device in one embodiment. Detailed Implementation

[0116] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0117] Given that how to implement federated learning in 5G-A / 6G networks is an urgent technical problem to be solved, this application provides a model management method, device, computer equipment, computer-readable storage medium, and computer program product that can improve the coverage of algorithms and enrich the applicable scenarios of generated federated learning models by utilizing a cloud-edge-device collaborative network architecture.

[0118] The model management method provided in this application embodiment can be applied to a mobile communication system in the application environment shown in Figure 1. The model management method provided in this application embodiment is, for example, a model management method based on hierarchical federated learning. The mobile communication system includes a network, which includes network nodes and multiple base stations. Each base station corresponds to multiple terminal devices. Each terminal device is a distributed node performing hierarchical federated learning, and each base station is a central node performing hierarchical federated learning. Network nodes can communicate with each base station, and base stations can communicate with the terminal devices corresponding to that base station. The mobile communication system can be a distributed network architecture. The UE (User Equipment) can be a terminal node, a terminal device, or a local client. The base station can be an edge node or a server, specifically, for example, a CU / BBU in a gNB / CU-DU separation architecture. The network node can be in the cloud, specifically, for example, an LMF / OAM (Orchestration and Management) / Network Data Analysis Function NWDAF (NW side). The terminal devices can be, but are not limited to, various personal computers, laptops, smartphones, tablets, etc.

[0119] In an exemplary embodiment, as shown in Figure 2, a model management method based on hierarchical federated learning is provided. Taking the application of this method to a terminal device in the mobile communication system shown in Figure 1 as an example, the mobile communication system further includes a network, which includes network nodes and multiple base stations. Each base station corresponds to multiple terminal devices. The terminal devices are distributed nodes performing hierarchical federated learning, and the base stations are central nodes performing hierarchical federated learning. The method includes:

[0120] Step 202: Receive the first configuration message sent by the base station.

[0121] The first configuration message includes user scheduling result indication information based on hierarchical federated learning. This user scheduling result indication information is obtained by the base station based on user feature information of each terminal device. The user feature information includes at least one of user data feature information and user channel state information. The user data feature information may be feature information of the locally updated gradient of the terminal device, and the user channel state information may be information obtained based on the channel estimation feedback of the user CSI-RS (Channel State Information Reference Signal).

[0122] Specifically, the base station can identify all terminal devices within its coverage area, and filter them based on hierarchical federated learning to determine the selected terminal devices. Based on user scheduling indication information, the base station obtains a first configuration message. The base station can send the first configuration message to each selected terminal device. The user scheduling result indication information contained in the first configuration message can indicate that the terminal device has been selected by the base station.

[0123] In one example, the specific process of base station scheduling may include: the base station determines a first number of terminal devices that need to be scheduled (used), and filters all terminal devices within the base station's coverage area based on the user characteristic information of each terminal device to extract the first number of terminal devices, and determines the user scheduling result indication information of the first number of terminal devices as first indication information, which can indicate that the terminal device has been selected by the base station; the user scheduling indication information of other terminal devices may be empty or may be second indication information, which is used to indicate that the terminal device has not been selected by the base station. This disclosure does not limit the specific form of the first indication information and the second indication information, and those skilled in the art can determine them specifically based on the actual application scenario.

[0124] Step 204: Send the reporting information to the network.

[0125] The network stores the hierarchical federated learning model and distribution strategy generated by the base station. The reported information includes at least one of the following: terminal capabilities, inference task type, and model performance monitoring feedback information. The reported information is used by the network to generate a second configuration message. The second configuration message is a model inference strategy distribution message or a model update strategy distribution message based on hierarchical federated learning.

[0126] Specifically, terminal devices can send reporting information to the network. After receiving the reporting information, the network can generate a second configuration message based on the reporting information, and then send the second configuration message to each terminal device. During model inference, the network sends the second configuration message to each terminal device, which may contain information on the model inference strategy based on hierarchical federated learning. During model update, the network sends the second configuration message to each terminal device, which may contain information on the model update strategy based on hierarchical federated learning.

[0127] Step 206: Receive the second configuration message and obtain the model inference result or model update result based on the second configuration message.

[0128] Specifically, the terminal device can receive the second configuration message and send information to perform model inference based on the model inference strategy in the second configuration message to obtain the model inference result; the terminal device can also send information to update the model based on the model update strategy in the second configuration message to obtain the model update result.

[0129] In the aforementioned model management method based on hierarchical federated learning, a first configuration message is received from the base station. This first configuration message contains user scheduling result indication information based on hierarchical federated learning. The user scheduling result indication information is obtained by the base station based on user characteristic information of each terminal device. The user characteristic information includes at least one of user data characteristic information and user channel state information. Reporting information is sent to the network. The network stores the hierarchical federated learning model and distribution strategy generated by the base station. The reporting information includes at least one of terminal capabilities, inference task type, and model performance monitoring feedback information. The reporting information is used by the network to generate a second configuration message, which is either model inference strategy distribution information or model update strategy distribution information based on hierarchical federated learning. The second configuration message is received, and model inference results or model update results are obtained based on the second configuration message. By adopting this method, data interaction between the base station and the terminal can be realized, which can improve the management efficiency and generalization performance of the wireless intelligent federated learning model, reduce the additional consumption of communication resources and computational overhead, and flexibly configure the resources to be allocated and flexibly schedule the terminal devices performing federated learning, further expanding the applicable scenarios of the model.

[0130] In one embodiment, the terminal capabilities include one or more of terminal computing capabilities, terminal channel states, and inference features. The terminal computing capabilities are used to characterize the model complexity supported by the terminal device, and the terminal channel states characterize the uplink channel states of the terminal device.

[0131] Specifically, terminal computing power is an indicator of the strength of the local computing power of a terminal device. Terminal computing power can be used to determine the model complexity supported by the terminal device. There is a positive correlation between terminal computing power and the supported model complexity; that is, the stronger the terminal computing power, the higher the supported model complexity. Terminal computing power can include the CPU's clock speed, number of threads, number of cores, GPU's bit width, and memory size, etc. Terminal channel state is used to characterize uplink channel state (CSI). Inference features can be intelligent features used to characterize the local data of the terminal device. For example, these inference features can include hyperparameters such as the number of features, number of samples, dimensions, data type, learning rate, and data batch size. The data type is used to characterize whether it is a time series. Terminal computing power and inference features can be optional options for the terminal device to send reporting information to the network. The terminal device can determine whether the reporting information includes terminal computing power and inference features based on the needs of the actual application scenario; or it can determine whether the reporting information includes terminal computing power and inference features based on the network-side configuration.

[0132] Optionally, the reported information may include terminal capabilities, and may also include inference task type, auxiliary information, etc.; the inference task type may characterize the type of model inference task, that is, the model type or the type of model inference result. This type may include one or more of classification tasks, prediction tasks, and clustering tasks, and the model used for model inference may include one or more of classification models, prediction models, and clustering models.

[0133] In this embodiment, the terminal capabilities carried in the reported information can include various types, and the specific content of the terminal capabilities can be determined based on the actual needs of the terminal device, ensuring the flexibility of the information represented by the terminal capabilities.

[0134] In one embodiment, the specific implementation process of the step "sending reporting information to the network" may include:

[0135] Based on the target transmission method, reporting information is sent to the base station so that the base station can send reporting information to the network nodes.

[0136] The target transmission method includes at least one of system messages, RRC signaling, MAC CE signaling, DCI signaling, and NAS messages.

[0137] Specifically, the reporting information sent by the terminal device to the network can be terminal device capability reporting signaling. The terminal device can report its capability reporting signaling to the base station (edge ​​node) through at least one of the following transmission methods: system message, RRC (Radio Resource Control) signaling, MAC CE (Media Access Control Element) signaling, DCI (Downlink Control Information) signaling, and NAS (Non-Access Stratum) message. That is, it reports to the edge node through at least one of these transmission methods. After receiving the reporting information, the edge node can send the reporting information to the network node corresponding to it. For example, the edge node can send the reporting information to the network node through the NG-C interface.

[0138] In this embodiment, the terminal device can transmit data with the base station through various transmission methods, ensuring the diversity and reliability of data transmission.

[0139] In one embodiment, the model performance monitoring feedback information includes one or more of the following: system performance metrics, accuracy metrics, complexity metrics, and generalization performance metrics.

[0140] Specifically, the model performance monitoring feedback information can include a variety of metrics. System performance metrics can include one or more of throughput, block error rate, and latency. Accuracy metrics can include one or more of the following: accuracy of model output, MSE / NMSE / cosine similarity between model input and output, etc. Complexity metrics can include model size and computational complexity. The model size refers to the number of parameters contained in the model, and the computational complexity can be FLOPs (Floating Point Operations Per second). The model performance monitoring feedback information can also include generalization performance metrics, the specific content of which can be determined based on the actual application scenario.

[0141] Optionally, after the UE performs model inference, it can monitor model performance in real time, obtain model performance monitoring results, and obtain model performance monitoring feedback information based on these results. The UE can send the model performance monitoring feedback information to the edge node corresponding to the terminal device. After receiving the model performance monitoring feedback information, the edge node can send the model performance monitoring feedback information to the network node (NW side) corresponding to the edge node.

[0142] In this embodiment, by monitoring the performance indicators during the model inference process, the model performance can be monitored in real time, and timely adjustments can be made to ensure the timeliness of the model.

[0143] In one embodiment, the hierarchical federated learning-based model management method further includes:

[0144] If the model performance monitoring feedback information and the preset indicator threshold are determined to meet the preset update conditions, a model update request is sent to the base station so that the network can respond to the model update request.

[0145] The model update request includes one or more of the following: the model to be updated, the function identifier to be updated, and reference information; the preset update conditions are used to determine whether to trigger the model update event; if it is determined that the preset update conditions are met, the update event is triggered; if it is determined that none of the preset update conditions are met, the update event is not triggered.

[0146] Specifically, the terminal device can obtain model performance monitoring feedback information based on the model performance monitoring results, for example, the model performance monitoring results can be used as model performance monitoring feedback information; the terminal device can determine the model performance monitoring results as reporting information and send the reporting information to the base station to which the terminal device belongs, and the base station can send the reporting information to the network node corresponding to the base station.

[0147] After the base station sends the reporting information to the network node corresponding to the base station, a model update process can be triggered. In one example, the model update triggering process can be initiated by the terminal device. The terminal device can compare the indicators in the model performance monitoring feedback information with the preset indicator thresholds corresponding to the indicators. If it is determined that there are indicators that meet the preset update conditions, the terminal device can trigger a model update event. That is, the terminal device can generate a model update request and send the model update request to the base station corresponding to the terminal device. After receiving the model update request, the base station can report the model update request to the network node. The network node can respond to the model update request. The response process can include: the network node determines at least one model to be updated, generates model update policy distribution information based on the at least one model to be updated, that is, obtains a second configuration message, and the network node can send the model update policy distribution information to the terminal device to be updated through the base station. The terminal device can perform the model update based on the model update policy distribution information and obtain the model update result. The specific update process will be described in detail in the following embodiments.

[0148] In another example, the model update triggering process can be initiated by the network side. After receiving the reported information, the network node can trigger a model update event based on target triggering conditions. The specific process of the target triggering conditions can include periodic model updates and non-periodic model updates. One possible implementation is that the network node triggers an update event for the model to be updated / the group of models to be updated according to a preset time period, generating a second configuration message. The preset time period can be associated with terminal capabilities or with changes in auxiliary information, such as scene, site, channel, or data sampling rate, etc. Another possible triggering method is that the network node triggers a model update event based on the received model performance monitoring feedback information. Specifically, the network node can compare the indicators contained in the model performance monitoring feedback information with the preset indicator threshold corresponding to the indicator. If it is determined that the indicator and the preset indicator threshold meet the preset update conditions, then the model update event is triggered.

[0149] Optionally, the specific process by which a network node / UE determines whether the preset update condition is met can be as follows: if the indicator included in the model performance monitoring feedback information is latency, and the network node / UE determines that the latency is greater than or equal to the preset latency threshold corresponding to that latency, then the network node / UE can determine that the preset update condition is met and trigger a model update event; if the indicator included in the model performance monitoring feedback information is model accuracy, and the network node / UE determines that the model accuracy is less than or equal to the preset latency threshold corresponding to that latency, then the network node / UE can determine that the preset update condition is met and trigger a model update event; the network node / UE can determine whether the preset update condition is met based on the meaning of the indicator and the relationship between the indicator and the preset indicator threshold corresponding to that indicator.

[0150] In one example, a terminal device can report model performance monitoring feedback information to the base station to which it belongs. After receiving the model performance monitoring feedback information, the base station can send the feedback information to the network node (NW side) corresponding to the base station. The model update process is triggered by either the terminal device or the network side.

[0151] In one example, a network node can configure a preset indicator threshold through a target transmission method, which includes at least one of system messages, RRC signaling, MAC CE signaling, DCI signaling, and NAS messages. The network node can send an indicator threshold configuration message to the terminal device through a base station. The indicator threshold configuration message includes at least the specific value of the indicator threshold and the indicator identifier. The indicator threshold configuration message can be at least one of system messages, RRC signaling, MAC CE signaling, DCI signaling, and NAS messages.

[0152] Optionally, the model update request sent by the terminal device specifically includes the model to be updated, the identifier of the function to be updated, and reference information. The reference information may be model update reference information fed back by the terminal device, such as gradient reference information, loss function reference information, and one or more of the logs.

[0153] Optionally, the base station to which the terminal device belongs can be within the coverage area of ​​the base station, with the base station serving as the central node of the hierarchical federated learning, and each terminal device within the coverage area of ​​the base station serving as a distributed node of the hierarchical federated learning. The network node corresponding to the base station can be a network node used to manage each base station.

[0154] In this embodiment, the terminal device / network node can proactively determine whether a model update process is needed based on the monitored model indicator information, thereby improving the terminal device / network node's control over the model update process and enabling timely model updates when the model does not meet the requirements.

[0155] In one embodiment, the specific implementation process of the step "receiving the second configuration message" may include:

[0156] Receive the second configuration message sent by the base station based on the target transmission mode.

[0157] The second configuration message is sent by the network node to the base station, and the target transmission method includes at least one of the following: system message, RRC signaling, MAC CE signaling, DCI signaling, and NAS message.

[0158] Specifically, after receiving the reported information from the terminal device, the network node can process the reported information to obtain a second configuration message and send the second configuration message to the base station. After receiving the second configuration message, the base station can send the second configuration message to each terminal device corresponding to the base station.

[0159] In one example, a network node generates a second configuration message based on the reported information received from a terminal device. The network node can send the second configuration message to the base station to which the terminal device belongs, so that the base station can send the second configuration message to the terminal device. In another example, a network node can send the second configuration message to each base station managed by the network node, so that each base station can send the second configuration message to each terminal device within its range. In yet another example, a network node can filter the base stations managed by the network node and send the second configuration message to the filtered base stations. Each base station can schedule terminal devices based on the method described in the above embodiments to obtain the scheduled terminal devices and send the second configuration message to each scheduled terminal device. In this case, each base station can also send the second configuration message to the terminal device corresponding to the identifier of the terminal device carried in the second configuration message.

[0160] In this embodiment, network nodes, base stations, and terminal devices can transmit data through various transmission methods to ensure the reliability, diversity, and flexibility of data transmission.

[0161] In one embodiment, the second configuration message is a model update strategy distribution information, which includes the model to be updated and the update strategy. Specifically, during the model update process, the second configuration message generated by the network node may include the model update strategy distribution information; during the model inference process, the second configuration message generated by the network node may include the model inference strategy distribution information.

[0162] In one example, the second configuration message is a model update policy distribution information. The process by which a network node generates the second configuration message can be as follows: Based on reported information, the network node determines at least one model to be updated, encapsulates the model to be updated, the user identifier of the terminal device to be updated, and the update policy to obtain the model update policy distribution information. The network node can send this model update policy distribution information to the base station through the user plane interface. The base station can then send the second configuration message to the terminal device corresponding to the received user identifier in the model update policy distribution information. The user plane interface can be an NG-U interface. The model to be updated determined by the network node can be a single model or a group of models to be updated.

[0163] Accordingly, the specific implementation process of the step "obtaining the model inference result or model update result based on the second configuration message" may include:

[0164] The update strategy is used to update the model to be updated, and the updated model result is obtained.

[0165] The updated model includes the model to be updated / group of models to be updated, and the update strategy includes one or more of the following: local update strategy, global update strategy, model replacement strategy, and model deletion strategy. The update strategy can be the model update method used by the terminal device when updating the model.

[0166] Specifically, after receiving the second configuration message, the terminal device can obtain the model update policy distribution information carried in the second configuration message, and update the model to be updated based on the update policy in the model update policy distribution information to obtain the model update result. The model update result can be the updated model after the update processing. That is to say, the terminal device currently receiving the second configuration message is the terminal device selected by the base station based on the user identifier to be updated carried in the second configuration message. This terminal device is the terminal device to be updated. This terminal device can perform model update processing based on the user device to be updated, the model to be updated / model group to be updated in the model update policy distribution information, and the update policy indicated by the distribution information, to obtain the model update result. The obtained model update result can be the updated model / model group obtained by the terminal device after executing the model update method selected by the network node. The model update policy distribution information can be the model update policy distribution signaling.

[0167] In this embodiment, the base station can send model update policy distribution information to a specific terminal device based on the instructions of the network node, so that the selected terminal device can perform model update processing.

[0168] In one embodiment, as shown in FIG3, the specific implementation process of the step "perform update processing" may include at least one of the following:

[0169] Step 302: Update the relevant modules of the model to be updated to obtain the model update result.

[0170] Some of the models can be specific modules or modules pre-selected by network nodes.

[0171] Specifically, when the update strategy includes a partial update strategy, the terminal device can determine the model to be updated / model group to be updated, as well as some modules in the model to be updated / model group to be updated, based on the information issued by the model update strategy. The terminal device can also update the model parameters of some modules in the model to be updated / model group to be updated, and determine the updated model / model group as the model update result.

[0172] Step 304: Based on the target update algorithm, update the model parameters of each model to be updated to obtain the model update result.

[0173] The target update algorithm includes one or more of the following: momentum gradient descent, stochastic gradient descent, and mini-batch gradient descent.

[0174] Specifically, the terminal device can determine the model / model group to be updated based on the model update strategy information, as well as the global update strategy and target update algorithm included in the update strategy. Then, it updates the collected data and all model parameters of the model / model group to be updated using the target update algorithm to obtain the updated model / model group. Based on this, the terminal device can determine that the updated model / model group is the model update result. The target update algorithm included in the update strategy includes at least one of the following: momentum gradient descent algorithm, stochastic gradient descent algorithm, and mini-batch gradient descent algorithm.

[0175] Step 306: Update the model parameters of each model to be updated based on the new model parameters to obtain the model update results.

[0176] Specifically, the terminal device can issue information based on the model update strategy to determine the model / model group to be updated, as well as the model replacement strategy included in the update strategy. The terminal device can then determine new model parameters based on the model replacement strategy and replace the model parameters in the model / model group to be updated with the new model parameters, thus obtaining the updated model / model group. Based on this, the terminal device can determine that the updated model / model group is the model update result.

[0177] Step 308: Delete the model parameters and model content / function identifiers of the updated model to obtain the model update result.

[0178] Specifically, if the terminal device determines that the update strategy is the model deletion strategy, the terminal device can identify the model to be updated / the model group to be updated, delete the model parameters of the model to be updated / the model group to be updated, and delete the model content / function identifier corresponding to the model to be updated / the model group to be updated, thereby obtaining the model update result.

[0179] It should be noted that this disclosure does not limit the order or scope of execution of the above steps 302, 304, 306 and 308. Those skilled in the art can determine the specific execution steps based on the specific content of the update strategy contained in the second configuration message.

[0180] In this embodiment, the network node can configure the specific process of the terminal device performing model updates through the transmission of the second configuration message, thereby ensuring the degree of information synchronization between the network node and the terminal device.

[0181] In one embodiment, the second configuration message is model inference strategy distribution information, which includes the model identifier of the model to be inferred, the model file, and the identifier information of the terminal device to perform model inference. Specifically, during the model inference process, the second configuration message generated by the network node may include model inference strategy distribution information.

[0182] In one example, the second configuration message is a model inference policy distribution message. The process by which a network node generates the second configuration message can be as follows: the terminal device to be inferred can send a model inference request signaling to the network node through a base station. The network node can respond to the model inference request signaling, generate a model inference response message, and send the model inference response message to the terminal device to be inferred through the base station. The terminal device that receives the model inference response message can send a reporting message to the base station (edge ​​node) to which the terminal device belongs. The base station can send the reporting message to the network node through the NG-C interface. After receiving the reporting message, the network node can, based on the reporting message, determine at least one model to be inferred whose target dimension meets the preset matching conditions among multiple first models stored in the network node, as well as the model identifier, model file, and identifier information of the terminal device to be inferred based on the model identifier, model file, and terminal device to be inferred, generate a model inference policy distribution message based on hierarchical federated learning, i.e., generate the second configuration message, and transmit it to the edge node through the NG-U interface, and then the edge node transmits it to the terminal device to be inferred.

[0183] The process by which a network node determines the model / group of models to be inferred can involve multiple models stored locally by the network node, which can be multiple first models. Among these first models, the network node can match a preset number of models / groups of models to be inferred based on reported information and a pre-configured number, by maximizing model accuracy, minimizing model complexity, or other dimensions. This will be described in detail in the following embodiments and will not be repeated here. The first model can be a model that has already completed model recognition or a model that has already completed functional recognition.

[0184] Accordingly, the specific implementation process of the step "obtaining the model inference result or model update result based on the second configuration message" may include:

[0185] Based on the model identifier, model file, and identifier information of the terminal device to be inferred, online or offline model inference is performed to obtain the model inference result.

[0186] The model identifier of the model to be reasoned can be the identifier of the list of models to be reasoned, such as the model to be reasoned / list of models to be reasoned.

[0187] Specifically, network nodes can encapsulate the model / group of models to be inferred, the model files of each model to be inferred, and the identification information of the terminal device to which model inference is to be performed, to obtain model inference policy distribution information. The network nodes can then use this information as a second configuration message and transmit it to the base station via the NG-U interface. The base station can then transmit the second configuration message to each terminal device to which model inference is to be performed, based on the identifier of the terminal device in the second configuration message. Upon receiving the second configuration message, the terminal device can...

[0188] In one example, the base station can transmit the second configuration message to each terminal device to which model inference is to be performed via a target transmission method. The base station can transmit the second configuration message to the terminal device to which model inference is to be performed using at least one of the following message transmission methods: system message, RRC signaling, MAC CE signaling, DCI signaling, and NAS message. The form of model transmission may include one or more of the following: sending in a private format, sending in a public format where the user equipment knows the model structure, and sending in a public format where the user equipment does not know the model structure.

[0189] Based on this, the terminal device can perform model inference by issuing information based on the model inference strategy and obtain the model inference result. For application scenarios where the latency requirement is greater than the first latency threshold, the terminal device can adopt an online model inference strategy. After receiving the information issued by the model inference strategy, it collects real-time data and performs real-time model inference based on the collected real-time data and the second configuration message to obtain the online model inference result. For application scenarios where the latency requirement is less than the second latency threshold, the terminal device can obtain data that has been stored locally in advance and perform model inference based on the data and the second configuration message to obtain the offline model inference result.

[0190] In this embodiment, the terminal device can perform real-time model inference online or offline model inference based on pre-collected data, based on the configuration requirements of network nodes, thereby expanding the scenarios applicable to model inference.

[0191] In an exemplary embodiment, as shown in Figure 4, a model management method is provided, such as a model management method based on hierarchical federated learning, applied to network nodes in the mobile communication system shown in Figure 1. The network includes network nodes and multiple base stations, each base station corresponding to multiple terminal devices. The communication coverage area of ​​each base station can include multiple terminal devices. The method includes:

[0192] Step 402: Receive the reporting information sent by the terminal device through the base station.

[0193] The reported information shall include at least one of the following: terminal capabilities, inference task type, and model performance monitoring feedback information.

[0194] Specifically, the terminal device can send reporting information to the base station to which it belongs, and the base station can send the received reporting information to the network node.

[0195] Step 404: Generate a second configuration message based on the reported information, and send the second configuration message to the terminal device through the base station.

[0196] The second configuration message is a message for disseminating model inference strategies or model update strategies based on hierarchical federated learning. The second configuration message is used to enable the terminal device to obtain model inference results or model update results.

[0197] Specifically, based on this, network nodes can send the second configuration message to each terminal device. During model inference, the network node sends the second configuration message to each terminal device, which may contain information about the model inference strategy based on hierarchical federated learning. During model update, the network node sends the second configuration message to each terminal device, which may contain information about the model update strategy based on hierarchical federated learning. Terminal devices can perform model inference based on the model inference strategy information in the received second configuration message to obtain the model inference result; terminal devices can also update the model based on the model update strategy information in the second configuration message to obtain the model update result.

[0198] In this embodiment, data interaction between the base station and the terminal can be realized, which can improve the management efficiency and generalization performance in the wireless intelligent federated learning model, reduce the additional consumption of communication resources and computational overhead, and flexibly configure the resources to be allocated and flexibly schedule the terminal devices for federated learning, further expanding the applicable scenarios of the model.

[0199] In one embodiment, the terminal capabilities include one or more of terminal computing capabilities, terminal channel states, and inference features. The terminal computing capabilities are used to characterize the model complexity supported by the terminal device, and the terminal channel states characterize the uplink channel states on the terminal device side.

[0200] Specifically, terminal computing power is an indicator of the strength of the local computing power of a terminal device. Terminal computing power can be used to determine the model complexity supported by the terminal device. There is a positive correlation between terminal computing power and the supported model complexity; that is, the stronger the terminal computing power, the higher the supported model complexity. Terminal computing power can include the CPU's clock speed, number of threads, number of cores, GPU's bit width, and memory size, etc. Terminal channel state is used to characterize uplink channel state (CSI). Inference features can be intelligent features used to characterize the local data of the terminal device. For example, these inference features can include hyperparameters such as the number of features, number of samples, dimensions, data type, learning rate, and data batch size. The data type is used to characterize whether it is a time series. Terminal computing power and inference features can be optional options for the terminal device to send reporting information to the network. The terminal device can determine whether the reporting information includes terminal computing power and inference features based on the needs of the actual application scenario; or it can determine whether the reporting information includes terminal computing power and inference features based on the network-side configuration.

[0201] Optionally, the reported information may include terminal capabilities, and may also include inference task type, auxiliary information, etc.; the inference task type may characterize the type of model inference task, that is, the model type or the type of model inference result. This type may include one or more of classification tasks, prediction tasks, and clustering tasks, and the model used for model inference may include one or more of classification models, prediction models, and clustering models.

[0202] In this embodiment, the terminal capabilities carried in the reported information can include various types, and the specific content of the terminal capabilities can be determined based on the actual needs of the terminal device, ensuring the flexibility of the information represented by the terminal capabilities.

[0203] In one embodiment, the reported information includes terminal capabilities, and the specific implementation process of the step "receiving the reported information sent by the terminal device through the base station" may include:

[0204] Based on the target transmission method, the receiving terminal device reports its terminal capabilities through the base station.

[0205] The target transmission method includes at least one of system messages, RRC signaling, MAC CE signaling, DCI signaling, and NAS messages.

[0206] Specifically, the terminal device can report its capability reporting signaling to the base station (edge ​​node) through at least one of the following transmission methods: system message, RRC signaling, MAC CE signaling, DCI signaling, and NAS message. That is, it reports the information to the edge node using at least one of these methods. After receiving the reporting information, the edge node can send it to its corresponding network node. For example, the edge node can send the reporting information to the network node via the NG-C interface.

[0207] In this embodiment, the terminal device can transmit data with the base station through various transmission methods, ensuring the diversity and reliability of data transmission.

[0208] In one embodiment, the hierarchical federated learning-based model management method further includes:

[0209] The system receives a model inference request signaling sent by a terminal device through a base station and returns a response message corresponding to the inference request, so that the terminal device can send reporting information based on the response message; the response message can be a model inference response message.

[0210] Specifically, during the model inference process, the terminal device can be the terminal device to which model inference is to be performed. The terminal device can generate a model inference request signaling and send the model inference request signaling to the base station to which the terminal device belongs. After receiving the model inference request signaling, the base station can send the model inference request signaling to the network node. The network node can respond to the model inference request signaling, generate a model inference response message, and return the model inference response message to the terminal device to which model inference is to be performed through the base station. The terminal device that receives the model inference response message can send reporting information to the base station (edge ​​node) to which the terminal device belongs. The base station can send the reporting information to the network node through the NG-C interface. The network node can receive the reporting information and obtain the second configuration message.

[0211] In this embodiment, the terminal device can actively trigger the model inference process and receive relevant model inference information sent by the network node in a timely manner, thereby improving the efficiency of model inference.

[0212] In one embodiment, as shown in Figure 5, the second configuration message includes information on model inference strategy distribution or model update strategy distribution based on hierarchical federated learning. The specific implementation process of the step "generating the second configuration message based on the reported information" can be as follows:

[0213] Step 502: Based on the reported information, determine at least one inference model among multiple first models whose target dimension satisfies the preset matching conditions.

[0214] The multiple models stored locally on the network nodes can be multiple first models, which can be models that have completed model recognition or models that have completed functional recognition. The target dimension can include model structure dimension, model complexity dimension, and model accuracy, etc.; the preset matching conditions can be the least complex, the most complex, the highest accuracy, the lowest accuracy, or the most moderate, etc., and can be determined based on the actual application scenario.

[0215] Specifically, network nodes can select from multiple first models based on the reported information sent by the terminal device and the preset number of pre-configured models to be inferred, and extract the model whose model structure, model complexity, accuracy and other aspects best match the terminal capabilities carried in the reported information to be determined as the model to be inferred / group of models to be inferred, or compromise by selecting a preset number of models as the model to be inferred / group of models to be inferred.

[0216] Step 504: Based on the model identifier of the model to be inferred, the model file, and the identifier information of the terminal device to be inferred, generate model inference strategy distribution information based on hierarchical federated learning.

[0217] Specifically, network nodes can encapsulate the model identifiers of the models to be inferred / groups of models to be inferred, the model files of each model to be inferred, and the identifier information of the terminal devices to which model inference is to be performed, as selected in the above steps, to obtain an encapsulation result. This encapsulation result is then identified as model inference policy distribution information, which can also be model inference policy distribution signaling. Network nodes can transmit this model inference policy distribution information to the base station via NG-U, and the base station can then transmit this information to each terminal device to which model inference is to be performed.

[0218] In this embodiment, network nodes can determine the inference model to be executed based on the reported information, and synchronize data with the base station and the terminal device to be executed in a timely manner to ensure the timeliness of data and improve the efficiency of model inference.

[0219] In one embodiment, the specific implementation process of the step "based on the reported information, determining at least one inference model whose target dimension satisfies the preset matching conditions among multiple first models" may include:

[0220] Based on the reported information, among multiple first models, a preset number of inference models are determined that meet the preset matching conditions in terms of model structure, model complexity, and model accuracy.

[0221] Specifically, network nodes can select from multiple first models based on the reported information sent by the terminal device and the preset number of pre-configured models to be inferred, and extract the model whose model structure, model complexity, accuracy and other aspects best match the terminal capabilities carried in the reported information to be determined as the model to be inferred / group of models to be inferred, or compromise by selecting a preset number of models as the model to be inferred / group of models to be inferred.

[0222] In this embodiment, network nodes can filter the models to be inferred from multiple dimensions based on the reported information, ensuring the comprehensiveness and flexibility of the inference model selection.

[0223] In one embodiment, the specific implementation process of the step "sending the second configuration message to the terminal device via the base station" may include:

[0224] The second configuration message is sent to the base station through the user plane interface, so that the base station can send the second configuration message to the terminal device through the target transmission method.

[0225] The target transmission method includes at least one of system messages, RRC signaling, MAC CE signaling, DCI signaling, and NAS messages. The user plane interface can be an NG-U interface.

[0226] Specifically, after receiving the reported information from the terminal device, the network node can process the reported information to obtain a second configuration message and send the second configuration message to the base station. After receiving the second configuration message, the base station can send the second configuration message to each terminal device corresponding to that base station. The network node can transmit the second configuration message to each base station through the NG-U interface, and the base station can transmit the second configuration message to multiple terminal devices corresponding to that base station in a target transmission mode.

[0227] Optionally, the base station may transmit the second configuration message to the terminal device using at least one of the following message transmission methods: system message, RRC signaling, MAC CE signaling, DCI signaling, and NAS message.

[0228] In one example, a network node generates a second configuration message based on the reported information received from a terminal device. The network node can send the second configuration message to the base station to which the terminal device belongs, so that the base station can send the second configuration message to the terminal device. In another example, a network node can send the second configuration message to each base station managed by the network node, so that each base station can send the second configuration message to each terminal device within its range. In yet another example, a network node can filter the base stations managed by the network node and send the second configuration message to the filtered base stations. Each base station can schedule terminal devices based on the method described in the above embodiments to obtain the scheduled terminal devices and send the second configuration message to each scheduled terminal device. In this case, each base station can also send the second configuration message to the terminal device corresponding to the identifier of the terminal device carried in the second configuration message.

[0229] In this embodiment, network nodes, base stations, and terminal devices can transmit data through various transmission methods to ensure the reliability, diversity, and flexibility of data transmission.

[0230] In one embodiment, the model performance monitoring feedback information includes one or more of the following: system performance metrics, accuracy metrics, complexity metrics, and generalization performance metrics.

[0231] Specifically, the model performance monitoring feedback information can include a variety of metrics. System performance metrics can include one or more of throughput, block error rate, and latency. Accuracy metrics can include one or more of the following: accuracy of model output, MSE / NMSE / cosine similarity between model input and output, etc. Complexity metrics can include model size and computational complexity. The model size refers to the number of parameters contained in the model, and the computational complexity can be FLOPs. The model performance monitoring feedback information can also include generalization performance metrics, the specific content of which can be determined based on the actual application scenario.

[0232] Optionally, after the UE performs model inference, it can monitor model performance in real time, obtain model performance monitoring results, and obtain model performance monitoring feedback information based on these results. The UE can send the model performance monitoring feedback information to the edge node corresponding to the terminal device. After receiving the model performance monitoring feedback information, the edge node can send the model performance monitoring feedback information to the network node (NW side) corresponding to the edge node.

[0233] In this embodiment, by monitoring the performance indicators during the model inference process, the model performance can be monitored in real time, and timely adjustments can be made to ensure the timeliness of the model.

[0234] In one embodiment, the reported information includes model performance monitoring feedback information.

[0235] Accordingly, as shown in Figure 6, the specific implementation process of the step "generating a second configuration message based on the reported information" may include:

[0236] Step 602: Receive model performance monitoring feedback information.

[0237] Specifically, the terminal device can report model performance monitoring feedback information to the base station to which it belongs. After receiving the model performance monitoring feedback information, the base station can send the model performance monitoring feedback information to the network node (NW side) corresponding to the base station. The network node can receive the model performance monitoring feedback information sent by the base station.

[0238] Step 604: Receive a model update request. The model update request includes the identifier of the model / function to be updated and reference information.

[0239] Among them, the model update request is generated when the model performance monitoring feedback information of the terminal device meets the preset update conditions with the preset indicator threshold.

[0240] Specifically, the terminal device can compare the indicators contained in the model performance monitoring feedback information with the preset indicator threshold corresponding to the indicator. If it is determined that the indicator and the preset indicator threshold meet the preset update conditions, a model update event is triggered and a model update request is generated. The terminal device can send the model update request to the network node through the base station.

[0241] The specific process by which the UE determines whether the preset update condition is met can be as follows: If the indicator included in the model performance monitoring feedback information is latency, and the UE determines that the latency is greater than or equal to the preset latency threshold corresponding to that latency, then the UE can determine that the preset update condition is met and trigger a model update event; if the indicator included in the model performance monitoring feedback information is model accuracy, and the UE determines that the model accuracy is less than or equal to the preset latency threshold corresponding to that latency, then the UE can determine that the preset update condition is met and trigger a model update event; the UE can determine whether the preset update condition is met based on the meaning of the indicator and the relationship between the indicator and the preset indicator threshold corresponding to that indicator.

[0242] Step 606: Respond to the model update request and generate a second configuration message.

[0243] Specifically, the network node can respond to the model update request. The response process may include: the network node determining at least one model to be updated, generating model update policy distribution information based on the at least one model to be updated, i.e. obtaining a second configuration message, and the network node sending the model update policy distribution information to the terminal device to be updated through the base station. The terminal device can perform the model update based on the model update policy distribution information and obtain the model update result.

[0244] In this embodiment, the terminal device / network node can proactively determine whether a model update process is needed based on the monitored model indicator information, thereby improving the terminal device / network node's control over the model update process and enabling timely model updates when the model does not meet the requirements.

[0245] In one embodiment, the reported information includes model performance monitoring feedback information.

[0246] Accordingly, as shown in Figure 7, the specific implementation process of the step "generating a second configuration message based on the reported information" may include:

[0247] Step 702: Receive model performance monitoring feedback information.

[0248] Specifically, the terminal device can obtain model performance monitoring feedback information based on the model performance monitoring results, for example, the model performance monitoring results can be used as model performance monitoring feedback information; the terminal device can determine the model performance monitoring results as reporting information and send the reporting information to the base station to which the terminal device belongs, and the base station can send the reporting information to the network node corresponding to the base station.

[0249] Step 704: Trigger a model update event based on a preset time period. Alternatively, trigger an update event if the model performance monitoring feedback and preset indicator thresholds meet preset update conditions.

[0250] The preset time period is associated with the terminal's capabilities or auxiliary information.

[0251] Specifically, the model update triggering process can be initiated by the network side. After receiving the reported information, the network node can trigger a model update event based on target triggering conditions. The specific process of the target triggering conditions can include periodic model updates and non-periodic model updates. One possible implementation is that the network node triggers an update event for the model to be updated / the group of models to be updated according to a preset time period, generating a second configuration message. The preset time period can be related to terminal capabilities or changes in auxiliary information, such as scene, site, channel, or data sampling rate, etc. Another possible triggering method is that the network node triggers a model update event based on the received model performance monitoring feedback information. Specifically, the network node can compare the indicators contained in the model performance monitoring feedback information with the preset indicator threshold corresponding to the indicator. If it is determined that the indicator and the preset indicator threshold meet the preset update conditions, then the model update event is triggered.

[0252] Optionally, the specific process by which a network node determines whether the preset update condition is met can be as follows: if the indicator included in the model performance monitoring feedback information is latency, and the network node determines that the latency is greater than or equal to the preset latency threshold corresponding to that latency, then the network node can determine that the preset update condition is met and trigger a model update event; if the indicator included in the model performance monitoring feedback information is model accuracy, and the network node determines that the model accuracy is less than or equal to the preset latency threshold corresponding to that latency, then the network node can determine that the preset update condition is met and trigger a model update event; the network node can determine whether the preset update condition is met based on the meaning of the indicator and the relationship between the indicator and the preset indicator threshold corresponding to that indicator.

[0253] Step 706: Generate a second configuration message based on the triggered update event.

[0254] Specifically, the network node determines at least one model to be updated, generates model update policy distribution information based on the at least one model to be updated, and obtains a second configuration message. The network node can send the model update policy distribution information to the terminal device to be updated through the base station. The terminal device can perform the model update based on the model update policy distribution information and obtain the model update result.

[0255] In this embodiment, the terminal device / network node can proactively determine whether a model update process is needed based on the monitored model indicator information, thereby improving the terminal device / network node's control over the model update process and enabling timely model updates when the model does not meet the requirements.

[0256] In one embodiment, the second configuration message includes information on the model inference strategy based on hierarchical federated learning.

[0257] Accordingly, the specific implementation process of the step "generating the second configuration message" may include:

[0258] Identify at least one model to be updated. Based on the model to be updated, the identification information of the terminal device to be updated, and the update strategy, generate model inference strategy distribution information based on hierarchical federated learning.

[0259] Specifically, the second configuration message is the model inference strategy distribution information. The process of the network node generating the second configuration message can be as follows: the terminal device to be executed model inference can send a model inference request signaling to the network node through the base station. The network node can respond to the model inference request signaling, generate a model inference response message, and send the model inference response message to the terminal device to be executed model inference through the base station. The terminal device that receives the model inference response message can send reporting information to the base station (edge ​​node) to which the terminal device belongs. The base station can send the reporting information to the network node through the NG-C interface. After receiving the reporting information, the network node can, based on the reporting information, determine at least one model to be inferred from the multiple first models stored in the network node whose target dimension meets the preset matching conditions, and generate model inference strategy distribution information based on hierarchical federated learning, i.e., generate the second configuration message, based on the model identifier, model file, and identification information of the terminal device to be executed model inference, from the model identifier, model file, and identification information of the terminal device to be executed model inference, and then generate the second configuration message. The second configuration message is then transmitted to the edge node through the NG-U interface and transmitted to the terminal device to be executed model inference by the edge node.

[0260] In one example, network nodes can encapsulate the model identifier, model file, and identifier information of the terminal device to be inferred, obtain the encapsulation result, and determine that the encapsulation result is the information to be distributed by the model inference strategy based on hierarchical federated learning.

[0261] The process of determining the model to be inferred / group of models to be inferred by the network node can be as follows: the network node can store multiple models locally, which can be multiple first models. Among the multiple first models, the network node can match the model to be inferred / group of models to be inferred by maximizing model accuracy, minimizing model complexity, or other dimensions based on the reported information and a pre-configured preset number.

[0262] In this embodiment, network nodes can select suitable models to be inferred during the model inference process to ensure the efficiency of model inference.

[0263] In one embodiment, the update strategy includes one or more of the following: local update strategy, global update strategy, model replacement strategy, and model deletion strategy.

[0264] Specifically, the specific meaning of the update strategy and the specific process by which the terminal device performs model update processing based on the update strategy to obtain the model update result have been described in detail in the embodiments corresponding to steps 302 to 308 above, and will not be repeated here.

[0265] In one embodiment, as shown in Figure 8, the model management method based on hierarchical federated learning further includes:

[0266] Step 802: Receive model data sent by each base station.

[0267] The model data includes model files and model content / function identifiers.

[0268] Specifically, the base station can identify the federated learning model indicated by the base station and send the model data of the identified federated learning model to the network node through the NG-U interface. The network node can receive the model data sent by each base station. The model data represents each federated learning model identified by the base station. The model data includes model files, model content, or functional identifiers.

[0269] Step 804: Determine each first model based on the data of each model, and arrange them according to the evaluation index of each first model to obtain a list of first models.

[0270] The evaluation metrics may include one or more of the following: accuracy, convergence speed, model complexity, or other evaluation metrics.

[0271] Specifically, network nodes obtain multiple identified models, i.e., first models, based on model files in various model data. Network nodes can sort the first models according to the evaluation metrics of each first model to obtain a first model list, which can be a model list or a function list.

[0272] In this embodiment, network nodes can efficiently store the functions / models identified by each base station, thereby improving the storage efficiency of the models.

[0273] In one embodiment, the hierarchical federated learning-based model management method further includes:

[0274] The system receives identification requests and model capabilities sent by the base station. The identification requests include model identification requests and function identification requests. The model capabilities are used to indicate the models / functions supported by the base station.

[0275] Specifically, after the federated learning model is trained, an identification event can be triggered. This identification event can be a model identification event or a function identification event. The identification event can be triggered by the base station. The specific identification process can be as follows: the base station can send an identification request to the network node through the NG-C interface. The base station reports the model capabilities to the network node. The model capabilities reported by the base station are associated with specific configurations / specific conditions. The model capabilities reported by the base station to the network node can also be associated with auxiliary conditions. The specific configurations / specific conditions can be specific configurations / specific conditions under the capabilities of each terminal in the area where the base station is located. The area where the base station is located can be the cell / sector / cell group where the base station is located. Auxiliary conditions can be, but are not limited to, scenarios, sites, channels, datasets, etc.

[0276] In one example, the model capabilities reported by the base station may include model type, model input and output descriptions, AI (Artificial Intelligence) related features involved in the model, achievable accuracy on a standardized dataset, convergence speed, model complexity, and application scenarios. The interaction of model capabilities between the base station and network nodes can be carried out via user plane signaling. For all models reported by the base stations, the network node sorts and builds a model / function list according to accuracy, convergence speed, model complexity, or other evaluation metrics, and assigns a model / function ID to each base station. The network node sends the model / function ID indication information to the corresponding base station via the NG-U interface. The model / function ID is unique on the network side, and the ID format includes at least one of the following:

[0277] It consists of two parts of bits concatenated together, including: the identifier bit of the cell / sector / cell group, and the bit used to distinguish different models / functions within the same cell / sector / cell group;

[0278] Other bit encoding methods used to distinguish different models / functions.

[0279] Based on this, the base station can determine the identified federated learning model indicated by the base station and send the model data of the identified federated learning model to the network node through the NG-U interface. The network node can receive the model data sent by each base station. The model data represents each federated learning model identified by the base station. The model data includes model files, model content, or functional identifiers.

[0280] In this embodiment, the base station can actively trigger model recognition, thereby improving the flexibility of model recognition.

[0281] In one embodiment, the hierarchical federated learning-based model management method further includes:

[0282] An identification request is sent to each base station to enable each base station to report its model capabilities. Model capabilities are used to indicate the models / functions supported by the base station.

[0283] Specifically, after the federated learning model is trained, an identification event can be triggered. This identification event can be a model identification event or a function identification event. The identification event can be triggered by a network node. The specific identification process can be as follows: the network node can send an identification request to the base station through the NG-C interface. After receiving the identification request, the base station responds to the model identification request by reporting the model capabilities to the network node. The model capabilities reported by the base station are associated with specific configurations / specific conditions. The model capabilities reported by the base station to the network node can also be associated with auxiliary conditions. The specific configurations / specific conditions can be specific configurations / specific conditions under the capabilities of each terminal in the area where the base station is located. The area where the base station is located can be the cell / sector / cell group where the base station is located. Auxiliary conditions can be, but are not limited to, scenarios, sites, channels, datasets, etc.

[0284] The base station reports model capabilities to the network node. The model capabilities reported by the base station are associated with specific configurations / specific conditions. The model capabilities reported by the base station to the network node can also be associated with auxiliary conditions. The specific configurations / specific conditions can be specific configurations / specific conditions under the capabilities of each terminal in the area where the base station is located. The area where the base station is located can be the cell / sector / cell group where the base station is located. Auxiliary conditions can be, but are not limited to, scenarios, sites, channels, datasets, etc.

[0285] In one example, the model capabilities reported by the base station may include model type, model input and output descriptions, AI-related features involved in the model, achievable accuracy on a standardized dataset, convergence speed, model complexity, and application scenarios. The interaction of model capabilities between the base station and network nodes can be carried out via user plane signaling. For all models reported by the base stations, the network node sorts and builds a model / function list according to accuracy, convergence speed, model complexity, or other evaluation metrics, and assigns a model / function ID to each base station. The network node sends the model / function ID indication information to the corresponding base station via the NG-U interface. The model / function ID is unique on the network side, and the ID format includes at least one of the following:

[0286] It consists of two parts of bits concatenated together, including: the identifier bit of the cell / sector / cell group, and the bit used to distinguish different models / functions within the same cell / sector / cell group;

[0287] Other bit encoding methods used to distinguish different models / functions.

[0288] Based on this, the base station can determine the identified federated learning model indicated by the base station and send the model data of the identified federated learning model to the network node through the NG-U interface. The network node can receive the model data sent by each base station. The model data represents each federated learning model identified by the base station. The model data includes model files, model content, or functional identifiers.

[0289] In this embodiment, network nodes can actively trigger model recognition, thereby improving the flexibility of model recognition.

[0290] In an exemplary embodiment, as shown in FIG9, a model management method is provided, such as a model management method based on hierarchical federated learning, applied to a base station in a mobile communication system shown in FIG1. ​​The mobile communication system includes a network and multiple base stations, each base station corresponding to multiple terminal devices. The terminal devices are distributed nodes performing hierarchical federated learning, and the base station is the central node performing hierarchical federated learning. The method includes:

[0291] Step 902: Based on the user feature information of each terminal device, obtain the user scheduling result indication information corresponding to each terminal, and send the first configuration message to each terminal device respectively.

[0292] The user characteristic information includes at least one of user data characteristic information and user channel state information, and the first configuration message includes user scheduling result indication information.

[0293] Specifically, the base station can identify all terminal devices within its coverage area, and filter them based on hierarchical federated learning to determine the selected terminal devices. Based on user scheduling indication information, the base station obtains a first configuration message. The base station can send the first configuration message to each selected terminal device. The user scheduling result indication information contained in the first configuration message can indicate that the terminal device has been selected by the base station.

[0294] In one example, the specific process of base station scheduling may include: the base station determines a first number of terminal devices that need to be scheduled (used), and filters all terminal devices within the base station's coverage area based on the user characteristic information of each terminal device to extract the first number of terminal devices, and determines the user scheduling result indication information of the first number of terminal devices as first indication information, which can indicate that the terminal device has been selected by the base station; the user scheduling indication information of other terminal devices may be empty or may be second indication information, which is used to indicate that the terminal device has not been selected by the base station. This disclosure does not limit the specific form of the first indication information and the second indication information, and those skilled in the art can determine them specifically based on the actual application scenario.

[0295] Step 904: Receive the reporting information sent by the terminal device.

[0296] The reported information shall include at least one of the following: terminal capabilities, inference task type, and model performance monitoring feedback information.

[0297] Step 906: Send the reported information to the network node so that the network node receives the second configuration message.

[0298] The second configuration message is a message for disseminating model inference strategies or model update strategies based on hierarchical federated learning. The second configuration message is used to enable the terminal device to obtain model inference results or model update results.

[0299] Specifically, terminal devices can send reporting information to the base station. The base station can then forward the received reporting information to network nodes. Upon receiving the reporting information, the network nodes can generate a second configuration message based on it. Based on this second configuration message, the network nodes can send it to the base station, which in turn sends it to each terminal device. During model inference, the base station sends the second configuration message to each terminal device. This second configuration message may contain information about the model inference strategy based on hierarchical federated learning. During model update, the network nodes send the second configuration message to each terminal device. This second configuration message may contain information about the model update strategy based on hierarchical federated learning. Terminal devices can receive the second configuration message and perform model inference based on the model inference strategy information within it to obtain the model inference result. Terminal devices can also update the model based on the model update strategy information within the second configuration message to obtain the model update result.

[0300] In this embodiment, data interaction between the base station and the terminal can be realized, which can improve the management efficiency and generalization performance in the wireless intelligent federated learning model, reduce the additional consumption of communication resources and computational overhead, and flexibly configure the resources to be allocated and flexibly schedule the terminal devices for federated learning, further expanding the applicable scenarios of the model.

[0301] In one embodiment, the hierarchical federated learning-based model management method further includes:

[0302] The model data, including model files and model content / functional identifiers, is sent to network nodes so that the network nodes can determine each first model and obtain a list of first models based on the model data.

[0303] Specifically, the base station can identify the federated learning model indicated by the base station and send the model data of the identified federated learning model to the network node through the NG-U interface. The network node can receive the model data sent by each base station. The model data represents each federated learning model identified by the base station. The model data includes model files, model content, or functional identifiers.

[0304] Network nodes obtain multiple identified models, or first models, based on model files in various model data. These first models can be sorted according to evaluation metrics to obtain a first model list, which can be either a model list or a function list. Evaluation metrics may include one or more of the following: accuracy, convergence speed, model complexity, or other evaluation metrics.

[0305] In this embodiment, network nodes can efficiently store the functions / models identified by each base station, thereby improving the storage efficiency of the models.

[0306] In one embodiment, the hierarchical federated learning-based model management method further includes:

[0307] The system sends identification requests and model capabilities to network nodes. The identification requests include model identification requests and function identification requests. The model capabilities are used to indicate the models / functions supported by the base station.

[0308] Specifically, after the federated learning model is trained, an identification event can be triggered. This identification event can be a model identification event or a function identification event. The identification event can be triggered by the base station. The specific identification process can be as follows: the base station can send an identification request to the network node through the NG-C interface. The base station reports the model capabilities to the network node. The model capabilities reported by the base station are associated with specific configurations / specific conditions. The model capabilities reported by the base station to the network node can also be associated with auxiliary conditions. The specific configurations / specific conditions can be specific configurations / specific conditions under the capabilities of each terminal in the area where the base station is located. The area where the base station is located can be the cell / sector / cell group where the base station is located. Auxiliary conditions can be, but are not limited to, scenarios, sites, channels, datasets, etc.

[0309] In one example, the model capabilities reported by the base station may include model type, model input and output descriptions, AI-related features involved in the model, achievable accuracy on a standardized dataset, convergence speed, model complexity, and application scenarios. The interaction of model capabilities between the base station and network nodes can be carried out via user plane signaling. For all models reported by the base stations, the network node sorts and builds a model / function list according to accuracy, convergence speed, model complexity, or other evaluation metrics, and assigns a model / function ID to each base station. The network node sends the model / function ID indication information to the corresponding base station via the NG-U interface. The model / function ID is unique on the network side, and the ID format includes at least one of the following:

[0310] It consists of two parts of bits concatenated together, including: the identifier bit of the cell / sector / cell group, and the bit used to distinguish different models / functions within the same cell / sector / cell group;

[0311] Other bit encoding methods used to distinguish different models / functions.

[0312] Based on this, the base station can determine the identified federated learning model indicated by the base station and send the model data of the identified federated learning model to the network node through the NG-U interface. The network node can receive the model data sent by each base station. The model data represents each federated learning model identified by the base station. The model data includes model files, model content, or functional identifiers.

[0313] In this embodiment, the base station can actively trigger model recognition, thereby improving the flexibility of model recognition.

[0314] In one embodiment, the hierarchical federated learning-based model management method further includes:

[0315] It receives identification requests sent by network nodes and reports model capabilities to network nodes based on the identification requests. Model capabilities are used to indicate the models / functions supported by the base station.

[0316] Specifically, after the federated learning model is trained, an identification event can be triggered. This identification event can be a model identification event or a function identification event. The identification event can be triggered by a network node. The specific identification process can be as follows: the network node can send an identification request to the base station through the NG-C interface. After receiving the identification request, the base station responds to the model identification request by reporting the model capabilities to the network node. The model capabilities reported by the base station are associated with specific configurations / specific conditions. The model capabilities reported by the base station to the network node can also be associated with auxiliary conditions. The specific configurations / specific conditions can be specific configurations / specific conditions under the capabilities of each terminal in the area where the base station is located. The area where the base station is located can be the cell / sector / cell group where the base station is located. Auxiliary conditions can be, but are not limited to, scenarios, sites, channels, datasets, etc.

[0317] The base station reports model capabilities to the network node. The model capabilities reported by the base station are associated with specific configurations / specific conditions. The model capabilities reported by the base station to the network node can also be associated with auxiliary conditions. The specific configurations / specific conditions can be specific configurations / specific conditions under the capabilities of each terminal in the area where the base station is located. The area where the base station is located can be the cell / sector / cell group where the base station is located. Auxiliary conditions can be, but are not limited to, scenarios, sites, channels, datasets, etc.

[0318] In one example, the model capabilities reported by the base station may include model type, model input and output descriptions, AI-related features involved in the model, achievable accuracy on a standardized dataset, convergence speed, model complexity, and application scenarios. The interaction of model capabilities between the base station and network nodes can be carried out via user plane signaling. For all models reported by the base stations, the network node sorts and builds a model / function list according to accuracy, convergence speed, model complexity, or other evaluation metrics, and assigns a model / function ID to each base station. The network node sends the model / function ID indication information to the corresponding base station via the NG-U interface. The model / function ID is unique on the network side, and the ID format includes at least one of the following:

[0319] It consists of two parts of bits concatenated together, including: the identifier bit of the cell / sector / cell group, and the bit used to distinguish different models / functions within the same cell / sector / cell group;

[0320] Other bit encoding methods used to distinguish different models / functions.

[0321] Based on this, the base station can determine the identified federated learning model indicated by the base station and send the model data of the identified federated learning model to the network node through the NG-U interface. The network node can receive the model data sent by each base station. The model data represents each federated learning model identified by the base station. The model data includes model files, model content, or functional identifiers.

[0322] In this embodiment, network nodes can actively trigger model recognition, thereby improving the flexibility of model recognition.

[0323] In one embodiment, the hierarchical federated learning-based model management method further includes:

[0324] Based on the model training results sent by each terminal device, a fully trained global model is obtained.

[0325] Specifically, the base station can obtain a trained global model based on the model training results received from each terminal device.

[0326] In one embodiment, the terminal device is a terminal device for scheduling users.

[0327] Accordingly, as shown in Figure 10, the specific implementation process of the step "obtaining the trained global model based on the model training results sent by each terminal device" may include:

[0328] Step 1002: For the i-th round of training, based on the training results sent by each terminal device in the (i-1)-th round, obtain the model parameters of the global model in the i-th round.

[0329] Step 1004: If the preset training completion conditions are not met, broadcast the model parameters of the global model in the i-th round to each terminal device so that each terminal device can train based on the model parameters of the global model in the i-th round until the base station determines that the preset training completion conditions are met and obtains the trained global model.

[0330] In this process, each terminal device can send training results that are locally updated gradients. The base station can perform model aggregation based on the locally updated gradients sent by each terminal device to obtain updated global model parameters. If the base station determines that the preset training completion conditions are not met, it can broadcast the updated global model parameters to the terminal device corresponding to the base station through the Physical Downlink Shared Channel (PDSCH) so that the terminal device receiving the global model parameters can perform training until the preset training completion conditions are met and the trained global model is obtained.

[0331] In one example, the network can contain M cells / sectors / cell groups (with the same cell ID). Each cell / sector / cell group has one edge node (base station) and N user equipment (terminal devices). There is one network node in the network. For each cell / sector / cell group, the following steps are executed independently:

[0332] Step 1: User devices and edge nodes initiate model training. The model training process includes at least one or more of the following: periodic model training and non-periodic model training.

[0333] Periodic model training involves user equipment and edge nodes initiating the model training process according to a preset time period. This time period is associated with specific configurations / conditions within the cell / sector / cell group and the capabilities of each UE. The time period can also be associated with auxiliary conditions, which may include scenarios, sites, channels, data sampling rates, typical parameters, or other requirements. Optionally, network nodes can configure this time period through at least one of system messages, RRC signaling, MAC CE signaling, DCI signaling, and NAS messages.

[0334] Non-periodic model training involves the user equipment and edge nodes initiating the model training process in a non-periodic manner (triggered by the UE side or NW side).

[0335] The triggering methods for model training include at least one of the following:

[0336] The UE side triggers the signal to send a model training request to the edge node for one or more AI-related features. The edge node responds to the request and begins model training.

[0337] When triggered on the NW side, the NW side sends a model training request signaling to the user device through the edge node for one or more AI-related features. The user device responds to the request and begins model training.

[0338] Step 2: The model training is carried out in T rounds. During the i-th round of model training, 0≤i≤T, the user device trains the local model and obtains the local model update result.

[0339] Step 3: The edge node can determine the set of users to be scheduled during this round of model training based on the user scheduling method. Only the scheduled user equipment needs to execute the local model training process in step 102. Specifically, the edge node pre-sets the number of users to be scheduled for this round of model training. From all users in the cell / sector / cell group within its coverage area, the edge node determines the set of users to be scheduled based on the channel state information fed back by the user CSI-RS channel estimation, the feature information between the user's local update gradient, or by randomly selecting a portion of users. The edge node then broadcasts the set of IDs of the users to be scheduled to all users within the cell / sector / cell group through the downlink physical broadcast channel PBCH.

[0340] Step 4: The scheduled user equipment uploads the local model update results (training results) to the edge node through the Physical Uplink Shared Channel (PUSCH).

[0341] Step 5: The edge nodes perform model aggregation based on the local update gradients in the training results uploaded by all scheduled users to obtain the model parameters of the updated global model.

[0342] Step 6: The edge node broadcasts the updated global model parameters to all user equipment via the Physical Downlink Shared Channel (PDSCH).

[0343] If it is determined that the preset training completion condition is not currently met, repeat steps 1 to 6 until the global model converges, i.e., the preset training completion condition is met. The preset training completion condition can be a convergence condition, and determining whether the model meets the preset convergence condition can include at least one of the following:

[0344] The number of training rounds reaches a preset upper limit, the loss function of the global model on the validation set is less than or equal to a preset value, and the difference between the loss function of the global model on the validation set in two adjacent rounds is less than or equal to a preset value.

[0345] Specifically, within the same cell / sector / cell group, one or more models can be trained. The training modes include at least one of the following: Parallel mode: the training of multiple models is simultaneous or overlaps in time; Serial mode: the training of multiple models is sequential and non-overlapping in time.

[0346] In one embodiment, the preset training completion conditions include one or more of the following: the number of training rounds reaches a preset upper limit, the loss function of the global model is less than or equal to a preset value, and the difference between the loss function of the global model in two adjacent rounds is less than or equal to a preset value.

[0347] Specifically, the preset training completion condition can be a convergence condition. Determining whether the model meets the preset convergence condition can include at least one of the following: the number of training rounds reaches a preset upper limit, the loss function of the global model on the validation set is less than or equal to a preset value, and the difference between the loss function of the global model on the validation set in two adjacent rounds is less than or equal to a preset value.

[0348] In one embodiment, the hierarchical federated learning-based model management method further includes:

[0349] Based on the user characteristic information of each terminal device, user scheduling result indication information corresponding to each terminal is obtained. The user characteristic information includes at least one of user data characteristic information and user channel state information. A first configuration message containing the user scheduling result indication information is sent to each terminal device.

[0350] In one embodiment, the specific implementation process of the step "obtaining user scheduling result indication information corresponding to each terminal based on the user feature information of each terminal device" may include:

[0351] Obtain the pre-configured number of scheduled users. Among all users within the base station's coverage area, based on at least one of the user data characteristic information and user channel state information of each terminal device, select the terminal devices that will be scheduled, and determine the user scheduling result indication information as the scheduled indication information.

[0352] Specifically, the base station can identify all terminal devices within its coverage area, and filter them based on hierarchical federated learning to determine the selected terminal devices. Based on user scheduling indication information, the base station obtains a first configuration message. The base station can send the first configuration message to each selected terminal device. The user scheduling result indication information contained in the first configuration message can indicate that the terminal device has been selected by the base station.

[0353] In one example, the specific process of base station scheduling may include: the base station determines a first number of terminal devices that need to be scheduled (used), and filters all terminal devices within the base station's coverage area based on the user characteristic information of each terminal device to extract the first number of terminal devices, and determines the user scheduling result indication information of the first number of terminal devices as first indication information, which can indicate that the terminal device has been selected by the base station; the user scheduling indication information of other terminal devices may be empty or may be second indication information, which is used to indicate that the terminal device has not been selected by the base station. This disclosure does not limit the specific form of the first indication information and the second indication information, and those skilled in the art can determine them specifically based on the actual application scenario.

[0354] In this embodiment, the base station can perform user scheduling to ensure the reliability of terminal devices undergoing model training and improve the efficiency of model training.

[0355] The following describes in detail the specific implementation process of the above-mentioned model management method based on hierarchical federated learning, using a specific embodiment. Before introducing the specific implementation process, the technical terms involved in this embodiment will be explained:

[0356] 6G: 6th Generation Mobile Communication Technology

[0357] IoT: Internet of Things

[0358] LCM: Life Cycle Management

[0359] LMF: Location Management Function

[0360] SINR: Signal to Interference plus Noise Ratio

[0361] SGD: Stochastic Gradient Descent

[0362] SSB: Synchronization Signal Block

[0363] RSRP: Reference Signal Receiving Power

[0364] KPI: Key Performance Indicators

[0365] MSE / NMSE: (Normalized) Mean Squared Error

[0366] The method provided in this embodiment mainly implements lifecycle management based on hierarchical federated learning, and provides corresponding signaling interaction mechanisms and configuration information. It aims to improve transmission reliability and reduce the probability of interruption. Furthermore, it can improve the convergence speed of the federated learning model through user equipment scheduling. The method provided in this embodiment is a lifecycle management method that supports flexible user scheduling, which can improve the management efficiency and generalization performance of the intelligent air interface federated learning model, reduce communication and computational overhead, effectively combat the impact of data heterogeneity, improve the convergence speed and accuracy of model training, consider energy-saving strategies to reasonably reduce the cost of deploying federated learning algorithms in the communication system, and consider model generalization performance so that the federated learning model can be well adapted to new application scenarios or deployed on newly added network nodes. The method provided in this embodiment is based on the fact that the convergence of the federated learning model is closely related to the data distribution of each user equipment, and can effectively perform user scheduling to accelerate the convergence speed of the federated learning model.

[0367] The Lifecycle Management (LCM) involved in the method provided in this embodiment is an important concept proposed in the current 3GPP standardization research based on the wireless AI framework. As a comprehensive management solution covering the entire process of AI models, LCM includes data collection, model training, model inference, model management (based on functionality / model ID selection, activation / deactivation, model switching / rollback, model monitoring, model update), and model transmission. Therefore, the method provided in this embodiment is designed and optimized under the condition of satisfying the lifecycle management framework defined by 3GPP.

[0368] Specifically, the distributed network architecture in the method provided in this embodiment includes network nodes, base stations, and terminal devices. Network nodes can be central nodes / cloud-based systems, such as LMF / OAM / NWDAF (NW side); base stations can be edge nodes / server-side systems, such as CU / BBU in a gNB / CU-DU separation architecture; and terminal devices can be terminal nodes / user equipment / local clients, such as UE (UE side). As shown in Figure 11a, the lifecycle management part based on hierarchical federated learning in the method provided in this embodiment includes at least a model training process based on hierarchical federated learning, a model identification process based on hierarchical federated learning, a model inference process based on hierarchical federated learning, and a model update process based on hierarchical federated learning. Specifically, initialization involves initializing model parameters for all user equipment (UEs). For model training based on hierarchical federated learning, there are M cells / sectors / cell groups in the network. In each cell / sector / cell group, a UE or edge node triggers the start of the model training process. Any scheduled UE updates its local model parameters and transmits them to the edge nodes via the air interface. The edge nodes aggregate and update the global model parameters. This process is repeated several times until the global model parameters converge, forming M clustered federated learning models. For model / function identification based on hierarchical federated learning, network nodes identify the model / function of the federated learning models trained on all cells / sectors / cell groups. All edge nodes transmit the identified federated learning models via NG interfaces. The model is sent to the network node for model storage. The network node assigns model / function IDs to all identified models for model deployment, inference, and update operations. For model inference based on hierarchical federated learning, the user equipment to be inferred sends a model inference request signaling to the NW side, reporting UE capabilities and inference task-related information. The network node determines the matching inference model / model group based on the information reported by the UE side, generates a signaling containing the configuration information of the inference model / model group, and transmits it to the user equipment to be inferred. The user equipment to be inferred performs model inference and obtains the inference result, and monitors model performance in real time. For model update based on hierarchical federated learning, the model update process is triggered by the UE side or the NW side, and the network node executes the model update process.

[0369] The following describes the model training process based on hierarchical federated learning, including:

[0370] The network can contain M cells / sectors / cell groups (with the same cell ID). Each cell / sector / cell group has one edge node (base station) and N user equipment (terminal devices). There is one network node in the network. For each cell / sector / cell group, the following steps are executed independently:

[0371] Step 1: User devices and edge nodes initiate model training. The model training process includes at least one or more of the following: periodic model training and non-periodic model training.

[0372] Periodic model training involves user equipment and edge nodes initiating the model training process according to a preset time period. This time period is associated with specific configurations / conditions within the cell / sector / cell group and the capabilities of each UE. The time period can also be associated with auxiliary conditions, which may include scenarios, sites, channels, data sampling rates, typical parameters, or other requirements. Optionally, network nodes can configure this time period through at least one of system messages, RRC signaling, MAC CE signaling, DCI signaling, and NAS messages.

[0373] Non-periodic model training involves the user equipment and edge nodes initiating the model training process in a non-periodic manner (triggered by the UE side or NW side).

[0374] The triggering methods for model training include at least one of the following:

[0375] The UE side triggers the signal to send a model training request to the edge node for one or more AI-related features. The edge node responds to the request and begins model training.

[0376] When triggered on the NW side, the NW side sends a model training request signaling to the user device through the edge node for one or more AI-related features. The user device responds to the request and begins model training.

[0377] Step 2: The model training is carried out in T rounds. During the i-th round of model training, 0≤i≤T, the user device trains the local model and obtains the local model update result.

[0378] Specifically, the process of "user device training a local model and obtaining local model update results" can be as follows: the user device can be a terminal device, the local model update result (training result) can be the local update gradient, and the terminal device can obtain the local update gradient based on local data using the stochastic gradient descent (SGD) algorithm. For example, the local update gradient g can be obtained using the following formula. i (w i,t ):

[0379] Among them, w i,t For the local model parameters of the i-th user, This refers to data batches obtained by random sampling based on the user's local dataset. For w i,t The gradient of the loss function at the data point (x,y).

[0380] Step 3: The edge node can determine the set of users to be scheduled during this round of model training based on the user scheduling method. Only the scheduled user equipment needs to execute the local model training process in step 102. Specifically, the edge node pre-sets the number of users to be scheduled for this round of model training. From all users in the cell / sector / cell group within its coverage area, the edge node determines the set of users to be scheduled based on the channel state information fed back by the user CSI-RS channel estimation, the feature information between the user's local update gradient, or by randomly selecting a portion of users. The edge node then broadcasts the set of IDs of the users to be scheduled to all users within the cell / sector / cell group through the downlink physical broadcast channel PBCH.

[0381] Among them, the number of users pre-configured by the base station for scheduling is less than or equal to the number of all users in the RRC active state within this cell / sector / cell group.

[0382] Specifically, the base station can define the utility function for each terminal device as the intra-cell revenue minus the inter-cell revenue. The intra-cell revenue is the average of the cosine similarity between the user's local update gradient and the local update gradients of other users in the same cell / sector / cell group. The local update gradients of each user device can be fed back to the base station through the uplink data channel. For example, the local update gradients of user devices can be compressed to obtain a compressed result, which is then fed back to the base station through the uplink data channel to reduce transmission overhead. The inter-cell revenue is the average of the cosine similarity between the user's local update gradient and the local update gradients of other users outside the same cell / sector / cell group.

[0383] The base station can also define the optimization problem as maximizing the sum of the weighted utility functions of all users within the cell / sector / cell group, with the weights representing the proportion of each user's data volume. The base station can perform user scheduling based on a proportional fairness algorithm to determine which user equipment will be scheduled. For example, the true utility of each user equipment can be defined as the ratio of the utility function value in the current training round to the time-averaged utility function value over all past training rounds. The base station selects the K users with the highest true utility from all users within the cell / sector / cell group as the users scheduled in this training round. This determines the user scheduling result indication information corresponding to each scheduled user equipment, where K is cell / sector / cell group specific, configured by the NW side, and can be included in system messages or NAS messages.

[0384] Step 4: The scheduled user equipment uploads the local model update results (training results) to the edge node through the Physical Uplink Shared Channel (PUSCH).

[0385] Specifically, the physical uplink shared channel can be a flat Rayleigh fading channel, and the signal-to-interference-plus-noise ratio (SINR) can be calculated using the following formula:

[0386] Among them, P i,t h is the transmit power of the i-th user (user equipment). i,t Used to characterize small-scale flat Rayleigh channel fading, d i Φ represents the distance from the user equipment to the base station. o,i Let α represent the set of users that interfere with the i-th user, and σ represent the path loss coefficient. 2 This represents the power of additive white Gaussian noise.

[0387] Step 5: The edge nodes perform model aggregation based on the local update gradients in the training results uploaded by all scheduled users to obtain the model parameters of the updated global model.

[0388] Specifically, the base station updates the global model parameters based on a weighted summation method, with the weights being the amount of data from each scheduled user equipment, as shown in the following formula.

[0389] in, Let η represent the global model parameters of the m-th cell / sector / cell group, and β represent the learning rate. i λ represents the ratio of the data size of the i-th user to the total data size of all users. i,t This indicates whether the i-th user successfully transferred the local gradient update during the t-th round of model training. If the transfer was successful, λ represents the gradient. i,t =1, otherwise λ i,t =0.

[0390] Step 6: The edge node broadcasts the updated global model parameters to all user equipment via the Physical Downlink Shared Channel (PDSCH).

[0391] If it is determined that the preset training completion condition is not currently met, repeat steps 1 to 6 until the global model converges, i.e., the preset training completion condition is met. The preset training completion condition can be a convergence condition, and determining whether the model meets the preset convergence condition can include at least one of the following:

[0392] The number of training rounds reaches a preset upper limit, the loss function of the global model on the validation set is less than or equal to a preset value, and the difference between the loss function of the global model on the validation set in two adjacent rounds is less than or equal to a preset value.

[0393] Specifically, within the same cell / sector / cell group, one or more models can be trained. The training modes include at least one of the following: Parallel mode: the training of multiple models is simultaneous or overlaps in time; Serial mode: the training of multiple models is sequential and non-overlapping in time.

[0394] The following describes the model identification / function identification process based on hierarchical federated learning, including:

[0395] Step a: The network node performs model identification / function identification on all trained federated learning models across all cells / sectors / cell groups. The identification process can be triggered by at least one of the following methods: edge node triggering or network node triggering. Edge node triggering includes: the edge node sending a model / function identification request to the network node via the NG-C interface and reporting its model capabilities. These capabilities are the capabilities of the model on the edge node, indicating the models / functions supported by the edge node. Network node triggering includes: the network node initiating a model / function identification request to the edge node via the NG-C interface; the edge node responding to the request and reporting its model capabilities. These capabilities are the capabilities of the model on the edge node, indicating the models / functions supported by the capabilities of the model on the edge node.

[0396] Step b: All edge nodes send the federated learning model file indicated by the model / function ID to the network node via the NG-U interface for model deployment / inference / update / storage.

[0397] The following describes a specific implementation of the model inference process based on hierarchical federated learning. The process by which network nodes obtain model inference policy information can be as follows: network nodes can sequentially encapsulate the inference model / model group list and model file information in the order of "Model / function ID of Model 1, Model file of Model 1, Model / function ID of Model 2, Model file of Model 2, ...", to obtain the encapsulation result. The encapsulation result is determined to be the model inference policy information. The encapsulation format of the encapsulation result is an IP packet, and the format of the model inference policy information can be a user device-specific binary executable file format.

[0398] In one embodiment, a 5G NR indoor positioning algorithm based on hierarchical federated learning is proposed. The user terminal is primarily responsible for updating local model parameters. The base station is responsible for collecting local model parameter information, using an aggregation algorithm to aggregate local parameters to update the global model, and then sending the new model parameters back to the local terminal. Network nodes are responsible for distributing positioning strategies and controlling resources.

[0399] As an efficient positioning method for indoor scenarios, AI technology can be applied to indoor industrial internet scenarios to address the inaccuracy of traditional methods in heavy indoor NLOS scenarios. The distributed federated learning training architecture has certain advantages, and the proposed algorithm has the following benefits: high user privacy protection, cooperative positioning to increase positioning accuracy, reduced communication overhead, and faster training convergence speed.

[0400] Table 1 shows the positioning errors of different positioning schemes at the 50%, 67%, 80%, and 90% percentiles of positioning accuracy:

[0401] Table 1

[0402] Figure 11b shows the CDF (Current Positioning Accuracy) curve for the AI-based positioning solution, corresponding to the CDF of the positioning accuracy. The results listed in Table 1 and Figure 11b demonstrate that the AI-based method significantly improves positioning accuracy. When using traditional positioning algorithms for UE coordinate prediction, the positioning error exceeds 10 meters at the 90th percentile of the CDF. The AI-based method can reduce the positioning error to within 1 meter. Combining multiple measurement information as input to the AI ​​model further improves positioning accuracy. Simulation results show that when using RSRP to acquire features, the positioning error for 90% of users can be reduced to within 0.38 meters.

[0403] With 10 clients / base stations online in the simulation setup, the personalized federated learning algorithm FedProx and the federated averaging algorithm FedAvg were used as comparison algorithms. The error variation trends of the Per-FedAvg algorithm and centralized training were also compared, as shown in Figure 11c. Figure 11c shows that the federated learning distributed training method has better convergence compared to the centralized training algorithm.

[0404] As shown in Figure 11d, using the CDF of the positioning error for comparison, the 90th percentile of the Per-FedAvg positioning error corresponds to an error of 0.22m, and there are no prominent values ​​from the distribution function. In contrast, the 90th percentile of the positioning error reached after 300 iterations in the centralized training case is 0.54m. Compared with the centralized training case, distributed training significantly reduces the positioning error.

[0405] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0406] Based on the same concept, this application also provides a model inference apparatus for implementing the model management method described above. The solution provided by this apparatus is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more model inference apparatus embodiments provided below can be found in the limitations of the model management method described above, and will not be repeated here.

[0407] In an exemplary embodiment, as shown in FIG12, a model inference apparatus 1200 is provided, applied to a terminal device in a mobile communication system. The mobile communication system further includes a network, the network including network nodes and multiple base stations, each base station corresponding to multiple terminal devices. The terminal devices are distributed nodes performing hierarchical federated learning, and the base stations are central nodes performing hierarchical federated learning. The apparatus includes:

[0408] The first receiving module 1202 is configured to receive a first configuration message sent by the base station. The first configuration message includes user scheduling result indication information based on hierarchical federated learning. The user scheduling result indication information is obtained by the base station based on user feature information of each terminal device. The user feature information includes at least one of user data feature information and user channel state information.

[0409] The first sending module 1204 is used to send reporting information to the network. The network stores the hierarchical federated learning model and distribution strategy generated by the base station. The reporting information includes at least one of terminal capabilities, inference task type, and model performance monitoring feedback information. The reporting information is used by the network to generate a second configuration message. The second configuration message is a model inference strategy distribution message or a model update strategy distribution message based on hierarchical federated learning.

[0410] The second receiving module 1206 is used to receive the second configuration message and obtain the model inference result or the model update result based on the second configuration message.

[0411] In one embodiment, the terminal capability includes one or more of terminal computing capability, terminal channel state, and inference features, wherein the terminal computing capability is used to characterize the model complexity supported by the terminal device, and the terminal channel state characterizes the uplink channel state of the terminal device.

[0412] In one embodiment, the first sending module is specifically used for:

[0413] Based on the target transmission method, reporting information is sent to the base station so that the base station can send reporting information to the network node. The target transmission method includes at least one of system message, RRC signaling, MAC CE signaling, DCI signaling, and NAS message.

[0414] In one embodiment, the model performance monitoring feedback information includes one or more of the following: system performance indicators, accuracy indicators, complexity indicators, and generalization performance indicators.

[0415] In one embodiment, the device further includes:

[0416] The first determining module is specifically used to: if it is determined that the model performance monitoring feedback information and the preset indicator threshold meet the preset update conditions, send a model update request to the base station so that the network responds to the model update request. The model update request includes one or more of the following: the model to be updated, the function identifier to be updated, and reference information.

[0417] In one embodiment, the second receiving module is specifically used for:

[0418] The system receives a second configuration message sent by the base station based on a target transmission mode. The second configuration message is sent by the network node to the base station. The target transmission mode includes at least one of system messages, RRC signaling, MAC CE signaling, DCI signaling, and NAS messages.

[0419] In one embodiment, the second configuration message is a model update strategy distribution information, which includes the model to be updated and the update strategy. The second receiving module is specifically used for:

[0420] The model to be updated is updated based on the update strategy to obtain the model update result. The updated model includes the model to be updated / group of models to be updated. The update strategy includes one or more of the following: local update strategy, global update strategy, model replacement strategy, and model deletion strategy.

[0421] In one embodiment, the second receiving module is specifically used for at least one of the following:

[0422] Update certain modules of the updated model to obtain the model update results.

[0423] Based on the target update algorithm, the model parameters of each of the models to be updated are updated to obtain the model update result. The target update algorithm includes one or more of the momentum gradient descent algorithm, stochastic gradient descent algorithm, and mini-batch gradient descent algorithm.

[0424] The model parameters of each model to be updated are updated based on the new model parameters to obtain the model update results.

[0425] Delete the model parameters and model content / function identifiers of the updated model to obtain the model update result.

[0426] In one embodiment, the second configuration message is a model inference strategy delivery information, which includes the model identifier of the model to be inferred, the model file, and the identifier information of the terminal device to perform model inference. The second receiving module is specifically used for:

[0427] Based on the model identifier, model file, and identifier information of the terminal device to be inferred, online or offline model inference is performed to obtain the model inference result.

[0428] In an exemplary embodiment, as shown in FIG13, a model inference device 1300 is provided, applied to a network node in a mobile communication system. The network includes network nodes and multiple base stations, each base station corresponding to multiple terminal devices. The device includes:

[0429] The third receiving module 1302 is used to receive the reporting information sent by the terminal device through the base station, wherein the reporting information includes at least one of the following: terminal capability, inference task type, and model performance monitoring feedback information.

[0430] The first generation module 1304 is used to generate a second configuration message based on the reported information, and send the second configuration message to the terminal device through the base station. The second configuration message is a model inference strategy distribution information or a model update strategy distribution information based on hierarchical federated learning, and is used to enable the terminal device to obtain the model inference result or the model update result.

[0431] In one embodiment, the terminal capability includes one or more of terminal computing capability, terminal channel state, and inference features, wherein the terminal computing capability is used to characterize the model complexity supported by the terminal device, and the terminal channel state characterizes the uplink channel state on the terminal device side.

[0432] In one embodiment, the third receiving module is specifically used for:

[0433] Based on the target transmission mode, the terminal capabilities reported by the terminal device through the base station are received. The target transmission mode includes at least one of system messages, RRC signaling, MAC CE signaling, DCI signaling, and NAS messages.

[0434] In one embodiment, the device further includes:

[0435] The fourth receiving module is used to receive the model inference request signaling sent by the terminal device through the base station, and return the response message corresponding to the inference request, so that the terminal device can send reporting information based on the response message.

[0436] In one embodiment, the second configuration message includes information on model inference strategy distribution or model update strategy distribution based on hierarchical federated learning. The first generation module is specifically used for:

[0437] Based on the reported information, among multiple first models, at least one inference model whose target dimension satisfies the preset matching conditions is determined.

[0438] Based on the model identifier, model file, and identifier information of the terminal device to be inferred, a model inference strategy distribution information based on hierarchical federated learning is generated.

[0439] In one embodiment, the first generation module is specifically used to: based on the reported information, determine a preset number of inference models among multiple first models whose model structure, model complexity, and model accuracy meet preset matching conditions.

[0440] In one embodiment, sending the second configuration message to the terminal device via the base station includes:

[0441] The second configuration message is sent to the base station through the user plane interface, so that the base station sends the second configuration message to the terminal device through the target transmission mode, wherein the target transmission mode includes at least one of system message, RRC signaling, MAC CE signaling, DCI signaling, and NAS message.

[0442] In one embodiment, the model performance monitoring feedback information includes one or more of the following: system performance indicators, accuracy indicators, complexity indicators, and generalization performance indicators.

[0443] In one embodiment, the reported information includes model performance monitoring feedback information, and the first generating module is specifically used to: receive model performance monitoring feedback information.

[0444] The system receives a model update request, which includes an identifier for the model / function to be updated and reference information. The model update request is generated when the model performance monitoring feedback information from the terminal device and the preset indicator threshold meet preset update conditions.

[0445] The system responds to the model update request by generating a second configuration message.

[0446] In one embodiment, the reported information includes model performance monitoring feedback information, and the first generating module is specifically used to: receive model performance monitoring feedback information.

[0447] A model update event is triggered based on a preset time period. Alternatively, an update event is triggered if the model performance monitoring feedback information and preset indicator thresholds meet preset update conditions. The preset time period is associated with terminal capabilities and auxiliary information.

[0448] Based on the triggered update event, a second configuration message is generated.

[0449] In one embodiment, the second configuration message includes information on the model inference strategy based on hierarchical federated learning. The first generation module is specifically configured to: determine at least one model to be updated.

[0450] Based on the model to be updated, the identification information of the terminal device to be updated, and the update strategy, a model inference strategy distribution information based on hierarchical federated learning is generated.

[0451] In one embodiment, the update strategy includes one or more of the following: a local update strategy, a global update strategy, a model replacement strategy, and a model deletion strategy.

[0452] In one embodiment, the device further includes:

[0453] The fifth receiving module is used to receive model data sent by each of the base stations. The model data includes model files and model content / function identifiers.

[0454] The sorting module is used to determine each of the first models based on the model data, and sort them according to the evaluation index of each of the first models to obtain a list of first models.

[0455] In one embodiment, the device further includes:

[0456] The sixth receiving module is used to receive identification requests and model capabilities sent by the base station. The identification requests include model identification requests and function identification requests. The model capabilities are used to indicate the models / functions supported by the base station.

[0457] In one embodiment, the device further includes:

[0458] The second sending module is used to send an identification request to each of the base stations so that each base station can report its model capabilities, wherein the model capabilities are used to indicate the models / functions supported by the base station.

[0459] In an exemplary embodiment, as shown in FIG14, a model inference apparatus 1400 is provided, applied to a base station in a mobile communication system, the apparatus comprising:

[0460] The third sending module 1402 is used to obtain user scheduling result indication information corresponding to each terminal based on the user feature information of each terminal device, and send a first configuration message to each terminal device respectively. The user feature information includes at least one of user data feature information and user channel state information. The first configuration message contains the user scheduling result indication information.

[0461] The seventh receiving module 1404 is used to receive the reporting information sent by the terminal device, wherein the reporting information includes at least one of the following: terminal capability, inference task type, and model performance monitoring feedback information.

[0462] The fourth sending module 1406 is used to send the reported information to the network node so that the network node receives the second configuration message; the second configuration message is a model inference strategy distribution information or a model update strategy distribution information based on hierarchical federated learning, and the second configuration message is used to enable the terminal device to obtain the model inference result or the model update result.

[0463] In one embodiment, the device further includes:

[0464] The fifth sending module is used to send model data to the network node. The model data includes model files and model content / function identifiers, so that the network node can determine each first model and obtain a first model list based on the model data.

[0465] In one embodiment, the device further includes:

[0466] The sixth sending module is used to send an identification request and a model capability to the network node. The identification request includes a model identification request and a function identification request. The model capability is used to indicate the models / functions supported by the base station.

[0467] In one embodiment, the device further includes:

[0468] The eighth receiving module is used to receive the identification request sent by the network node and report the model capabilities to the network node based on the identification request. The model capabilities are used to indicate the models / functions supported by the base station.

[0469] In one embodiment, the device further includes:

[0470] The seventh sending module is used to obtain the trained global model based on the model training results sent by each of the terminal devices.

[0471] In one embodiment, the seventh sending module is specifically used to: for the i-th round of training, based on the training results sent by each of the terminal devices in the (i-1)-th round, obtain the model parameters of the global model in the i-th round;

[0472] If the preset training completion conditions are not met, the model parameters of the global model in the i-th round are broadcast to each of the terminal devices so that each of the terminal devices can train based on the model parameters of the global model in the i-th round until the base station determines that the preset training completion conditions are met and obtains the trained global model.

[0473] In one embodiment, the preset training completion condition includes one or more of the following: the number of training rounds reaches a preset upper limit, the loss function of the global model is less than or equal to a preset value, and the difference between the loss function of the global model in two adjacent rounds is less than or equal to a preset value.

[0474] In one embodiment, the device further includes:

[0475] The indication information determination module is used to obtain user scheduling result indication information corresponding to each terminal based on the user feature information of each terminal device, wherein the user feature information includes at least one of user data feature information and user channel state information.

[0476] A first configuration message is sent to each of the terminal devices, the first configuration message containing the user scheduling result indication information.

[0477] In one embodiment, the indication information determination module is specifically used to obtain the pre-configured number of scheduled users;

[0478] Among all users within the coverage area of ​​the base station, based on at least one of the user data feature information and user channel state information of each terminal device, the terminal devices that are scheduled for the specified number of users are selected, and the user scheduling result indication information is determined as the scheduled indication information.

[0479] The aforementioned model inference device can also be referred to as a model inference device based on hierarchical federated learning.

[0480] Each module in the aforementioned model inference device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0481] In an exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram is shown in Figure 15. The computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is connected to the system bus via the I / O interfaces. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device stores data for model lifecycle management. The I / O interfaces of the computer device are used for exchanging information between the processor and external devices. The communication interface of the computer device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a model management method.

[0482] Those skilled in the art will understand that the structure shown in Figure 15 is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or may combine certain components, or may have different component arrangements.

[0483] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps described in the above embodiments.

[0484] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, performs the steps described in the above embodiments.

[0485] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps described in the above embodiments.

[0486] It should be noted that the user information (including but not limited to terminal device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0487] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0488] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0489] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

A model management method applied to a terminal device in a mobile communication system, the mobile communication system further comprising a network, the network comprising a network node and a plurality of base stations, each of the base stations corresponding to a plurality of terminal devices, the terminal devices being distributed nodes performing hierarchical federated learning, and the base stations being central nodes performing hierarchical federated learning; the method comprising: receiving a first configuration message sent by the base station, the first configuration message containing user scheduling result indication information based on hierarchical federated learning, the user scheduling result indication information being obtained by the base station based on user feature information of each of the terminal devices, the user feature information including at least one of user data feature information and user channel state information; sending reporting information to the network, the network storing a hierarchical federated learning model generated by the base station and a distribution policy, the reporting information including at least one of terminal capability, inference task type, and model performance monitoring feedback information, the reporting information being used by the network to generate a second configuration message, the second configuration message being model inference policy distribution information or model update policy distribution information based on hierarchical federated learning; receiving the second configuration message and obtaining a model inference result or a model update result based on the second configuration message. The method of claim 1, wherein, The terminal capability includes one or more of terminal computing capability, terminal channel state, and inference feature, the terminal computing capability being used to represent the model complexity supported by the terminal device, and the terminal channel state representing the uplink channel state of the terminal device. The method of claim 1, wherein, The sending of the reporting information to the network comprises: sending the reporting information to the base station based on a target transmission mode, so that the base station sends the reporting information to the network node, the target transmission mode including at least one of system message, RRC signaling, MAC CE signaling, DCI signaling, and NAS message. The method of claim 1, wherein, The model performance monitoring feedback information includes one or more of system performance indicators, accuracy indicators, complexity indicators, and generalization performance indicators. The method of claim 4, further comprising: if it is determined that the model performance monitoring feedback information meets a preset update condition with a preset indicator threshold, sending a model update request to the base station, so that the network responds to the model update request, the model update request including one or more of a model to be updated, a function identifier to be updated, and reference information. The method of claim 1, wherein, The receiving of the second configuration message comprises: receiving the second configuration message sent by the base station based on a target transmission mode, the second configuration message being sent by the network node to the base station, and the target transmission mode including at least one of system message, RRC signaling, MAC CE signaling DCI signaling, and NAS message. The method of claim 1, wherein, The second configuration message is model update policy distribution information, the model update policy distribution information including a model to be updated and an update policy; and the obtaining of the model inference result or the model update result based on the second configuration message comprises: The updating strategy is used to update the to-be-updated model, and a model updating result is obtained, the to-be-updated model includes a to-be-updated model / to-be-updated model group, and the updating strategy includes one or more of a local updating strategy, a global updating strategy, a model replacement strategy, and a model deletion strategy. The method of claim 7, wherein, The updating process includes at least one of the following: updating part of the model to obtain a model updating result; updating the model parameters of each to-be-updated model based on a target updating algorithm to obtain a model updating result, the target updating algorithm including one or more of a momentum gradient descent algorithm, a stochastic gradient descent algorithm, and a small batch gradient descent algorithm; updating the model parameters of each to-be-updated model based on new model parameters to obtain a model updating result; deleting the model parameters, model content / function identifiers of the updated model to obtain a model updating result. The method of claim 1, wherein The second configuration message is a model inference strategy distribution information, and the model inference strategy distribution information includes a model identifier of a to-be-inferred model, a model file, and identifier information of a terminal device to be executed for model inference. The model inference result or the model updating result is obtained based on the second configuration message, including: performing online mode model inference or offline mode model inference based on the model identifier of the to-be-inferred model, the model file, and the identifier information of the terminal device to be executed for model inference to obtain a model inference result. A model management method applied to a network node in a mobile communication system, the network including a network node and a plurality of base stations, each base station corresponding to a plurality of terminal devices, the method including: receiving report information sent by the terminal device through the base station, the report information including at least one of terminal capability, inference task type, and model performance monitoring feedback information; generating a second configuration message based on the report information and sending the second configuration message to the terminal device through the base station; the second configuration message is a model inference strategy distribution information or a model updating strategy distribution information based on hierarchical federated learning, and the second configuration message is used to make the terminal device obtain a model inference result or a model updating result. The method of claim 10, wherein, The terminal capability includes one or more of terminal computing capability, terminal channel state, and inference feature, the terminal computing capability is used to represent the model complexity supported by the terminal device, and the terminal channel state represents the uplink channel state on the terminal device side. The method of claim 10, wherein, The report information includes the terminal capability, and the receiving of the report information sent by the terminal device through the base station includes: receiving terminal capability reported by the terminal device through the base station based on a target transmission mode, the target transmission mode including at least one of system message, RRC signaling, MAC CE signaling, DCI signaling, and NAS message. The method of claim 10 further includes: receiving a model inference request signaling sent by the terminal device through the base station, and returning a response message corresponding to the inference request to make the terminal device send report information based on the response message. The method of claim 10, wherein The second configuration message comprises model inference strategy issuing information or model update strategy issuing information based on hierarchical federated learning; The second configuration message is generated based on the reported information, comprising: Based on the reported information, at least one to-be-inferred model in the plurality of first models is determined, which satisfies a preset matching condition of a target dimension; Based on the model identifier of the to-be-inferred model, the model file, and the identifier information of the terminal device to be executed for model inference, the model inference strategy issuing information based on hierarchical federated learning is generated. The method of claim 11, wherein, The second configuration message is generated based on the reported information, comprising: Based on the reported information in the plurality of first models, a preset number of to-be-inferred models are determined, which satisfy a preset matching condition of model structure, model complexity, and model accuracy. The method of claim 14, wherein, The second configuration message is transmitted to the terminal device by the base station, comprising: The second configuration message is transmitted to the base station through a user plane interface, so that the base station transmits the second configuration message to the terminal device through the target transmission mode, and the target transmission mode comprises at least one of system message, RRC signaling, MAC CE signaling, DCI signaling, and NAS message. The method of claim 14, wherein, The model performance monitoring feedback information comprises one or more of system performance indicators, accuracy indicators, complexity indicators, and generalization performance indicators. The method of claim 14, wherein, The reported information comprises model performance monitoring feedback information, and the second configuration message is generated based on the reported information, comprising: Receiving model performance monitoring feedback information; Receiving a model update request, the model update request comprising a to-be-updated model / to-be-updated function identifier, reference information, and the model update request being generated in a case where the model performance monitoring feedback information of the terminal device and a preset indicator threshold satisfy a preset update condition; Responding to the model update request to generate a second configuration message. The method of claim 14, wherein, The reported information comprises model performance monitoring feedback information, and the second configuration message is generated based thereon, comprising: Receiving model performance monitoring feedback information; Based on a preset time period, triggering a model update event; or, if the model performance monitoring feedback information and a preset indicator threshold satisfy a preset update condition, triggering an update event; the preset time period is associated with terminal capability and auxiliary information; Based on the triggered update event, a second configuration message is generated. The method according to claim 18 or 19, wherein The second configuration message comprises model inference strategy issuing information based on hierarchical federated learning; The second configuration message is generated, comprising: Determining at least one to-be-updated model; Based on the to-be-updated model, the identifier information of the terminal device to be executed for model update, and the update strategy, the model inference strategy issuing information based on hierarchical federated learning is generated. The method of claim 20, wherein, The update strategy comprises one or more of a local update strategy, a global update strategy, a model replacement strategy, and a model deletion strategy. The method of claim 18, further comprising: Receiving model data transmitted by each base station, the model data comprising a model file, a model content / function identifier; Determine each first model based on each model data, and arrange each first model according to an evaluation index of each first model to obtain a first model list. The method of claim 22, further comprising: receiving an identification request and model capability sent by the base station, wherein the identification request comprises a model identification request and a function identification request, and the model capability is used to indicate a model supported by the base station and a function supported by the base station. The method of claim 22, further comprising: sending an identification request to each base station to make each base station report a model capability, wherein the model capability is used to indicate a model supported by the base station and a function supported by the base station. A model management method, wherein, A base station applied to a mobile communication system, the mobile communication system comprising a network and a plurality of base stations, each base station corresponding to a plurality of terminal devices, the terminal devices being distributed nodes performing hierarchical federated learning, and the base station being a central node performing hierarchical federated learning; the method comprising: obtaining user scheduling result indication information corresponding to each terminal device based on user feature information of each terminal device, and sending a first configuration message to each terminal device, wherein the user feature information comprises at least one of user data feature information and user channel state information, and the first configuration message comprises the user scheduling result indication information; receiving report information sent by the terminal device, wherein the report information comprises at least one of terminal capability, inference task type, and model performance monitoring feedback information; sending the report information to the network node to make the network node obtain a second configuration message; the second configuration message is model inference strategy or model update strategy information issued based on hierarchical federated learning, and the second configuration message is used to make the terminal device obtain a model inference result or a model update result. The method of claim 25, further comprising: sending model data to the network node, wherein the model data comprises a model file and a model content / function identifier, so that the network node determines each first model based on each model data and obtains a first model list. The method of claim 25, further comprising: sending an identification request and model capability to the network node, wherein the identification request comprises a model identification request and a function identification request, and the model capacity is used to indicate a model supported by the base station and a function supported by the base station; The method of claim 25, further comprising: receiving an identification request sent by the network node, and reporting a model capability to the network node based on the identification request, wherein the model capability is used to indicate a model supported by the base station and a function support by the base station. The method of claim 25, further comprising: obtaining a global model trained based on model training results sent by each terminal device. The method of claim 29, wherein, The terminal device is a user-scheduled terminal device, and obtaining a global model trained based on model training results sent by each terminal device comprises: for the i-th round of training, obtaining model parameters of the i-th round of global model based on training results sent by each terminal device in the i-1-th round of training. In a case where the preset training completion condition is not met, the base station broadcasts model parameters of the i-th round of global model to each terminal device, so that each terminal device is trained based on the model parameters of the i-th round of global model until the base station determines that the preset training completion condition is met, and a trained global model is obtained. The method of claim 30, wherein, The preset training completion condition includes one or more of the following: a training round number reaches a preset upper limit value, a loss function of the global model is less than or equal to a preset value, and a difference between loss functions of adjacent two rounds is less than or equal to a preset value. The method of claim 25, further comprising: obtaining user scheduling result indication information corresponding to each terminal device based on user feature information of each terminal device, the user feature information including at least one of user data feature information and user channel state information; sending a first configuration message to each terminal device, the first configuration message containing the user scheduling result indication information. The method of claim 32, wherein, The method of claim 25, wherein the obtaining user scheduling result indication information corresponding to each terminal device based on user feature information of each terminal device comprises: obtaining a preset number of scheduled users; from all users in a coverage range of the base station, filtering terminal devices of the number of scheduled users based on at least one of user data feature information and user channel state information of each terminal device, and determining the user scheduling result indication information as scheduled indication information. A model inference device applied to a terminal device in a mobile communication system, the mobile communication system further including a network, the network including a network node and a plurality of base stations, each base station corresponding to a plurality of terminal devices, the terminal device being a distributed node performing hierarchical federated learning, and the base station being a central node performing hierarchical federated learning; the device comprising: a first receiving module configured to receive a first configuration message sent by the base station, the first configuration message containing user scheduling result indication information based on hierarchical federated learning, the user scheduling result indication information being obtained by the base station based on user feature information of each terminal device, the user feature information including at least one of the user data feature information and the user channel state information; a first sending module configured to send reporting information to the network, the network storing a hierarchical federated learning model generated by the base station and a distribution strategy, the reporting information including at least one of terminal capability, inference task type, and model performance monitoring feedback information, the reporting information being used by the network to generate a second configuration message, the second configuration message being model inference strategy distribution information or model update strategy distribution information based on hierarchical federated learning; a second receiving module configured to receive the second configuration message, and obtain a model inference result or a model update result based on the second configuration message. A model management device applied to a network node in a mobile communication system, the network including a network node and a plurality of base stations, each base station corresponding to multiple terminal devices, the device comprising: The third receiving module is configured to receive report information sent by the terminal device through the base station, wherein the report information at least includes at least one of terminal capability, inference task type, and model performance monitoring feedback information. The first generating module is configured to generate a second configuration message based on the report information, and send the second configuration message to the terminal device through the base station; the second configuration message is model inference strategy or model update strategy issuing information based on hierarchical federated learning, and the second configuration message is used to make the terminal device obtain model inference result or model update result. A model management apparatus is applied to a base station in a mobile communication system, the mobile communication system includes a network and a plurality of base stations, each base station corresponds to a plurality of terminal devices, the terminal device is a distributed node performing hierarchical federated learning, and the base station is a central node performing hierarchical federated learning; the apparatus includes: The second sending module is configured to obtain user scheduling result indication information corresponding to each terminal device based on user feature information of each terminal device, and send a first configuration message to each terminal device, wherein the user feature information includes at least one of user data feature information and user channel state information, and the first configuration message includes the user scheduling result indication information. The fourth receiving module is configured to receive report information sent by the terminal device, wherein the report information at least includes at least one of terminal capability, inference task type, and model performance monitoring feedback information. The third sending module is configured to send the report information to the network node, so that the network node obtains a second configuration message; the second configuration message is model inference strategy or model update strategy issuing information based on hierarchical federated learning, and the second configuration message is used to make the terminal device obtain model inference result or model update result. A computer device comprising a memory and a processor, the memory storing a computer program, wherein, The processor executes the computer program to realize the steps of the method in any one of claims 1 to 33. A computer-readable storage medium having stored thereon a computer program, wherein, The computer program is executed by the processor to realize the steps of the method in any one of claims 1 to 33. A computer program product comprising a computer program, wherein, The computer program is implemented by the processor to realize the steps of the method in any one of claims 1 to 32. The computer program is implemented by the processor to realize the steps of the method in any one claim of 1 to 33.

Images