Method, device and readable storage medium for analyzing model transmission state in subscription network

Abstract

The application provides a model transmission state analysis method and device in a subscription network and a readable storage medium. The method comprises the following steps: an application function (AF) directly or through a network capability exposure function (NEF) sends a first message to a network data analysis function (NWDAF); the AF directly or through the NEF receives analysis information of an AI / ML model transmission state sent by the NWDAF, wherein the analysis information is determined by the NWDAF according to data of AI / ML model transmission states sent by other network functions (5GC NF(s)); and wherein the analysis information is used to adjust network policy parameters and / or application layer model information. The application can effectively analyze the AI / ML model transmission state, thereby enabling the network to effectively adjust the network transmission strategy based on the AI / ML model transmission state and enabling a third party to obtain the analysis of the AI / ML model transmission state to adjust the application layer information.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a method, apparatus and readable storage medium for analyzing the transmission state of a model in a subscription network. Background Technology

[0002] In recent years, due to technological breakthroughs in artificial intelligence, its applications have become increasingly widespread. However, mobile terminals, constrained by strict energy consumption, computing power, and memory costs, cannot run heavyweight artificial intelligence (AI) / machine learning (ML) models (hereinafter referred to as AI / ML models) on their own. Therefore, the current approach involves transmitting the inference of many AI / ML models from mobile terminals to the cloud or other terminals; in other words, it is necessary to transfer AI / ML models to the cloud or other terminals.

[0003] Furthermore, due to concerns about privacy protection of transmitted data and alleviating network data transmission pressure, the requirements for AI / ML model transmission are becoming increasingly stringent. As the channel for transmitting AI / ML models, 5G systems, in order to enhance the intelligence capabilities of 5G networks and meet the requirements for AI / ML model transmission in TS 22.261 (SA1#93e approved), need to support the exposure of monitoring and status information about AI-ML sessions to third parties.

[0004] However, existing technologies cannot effectively analyze the transmission status of AI / ML models, which prevents the network from effectively adjusting its transmission strategy based on the transmission status of AI / ML models, and prevents third parties from obtaining analysis of the transmission status of AI / ML models to adjust application layer information. Summary of the Invention

[0005] This application provides a method, apparatus, and readable storage medium for analyzing the transmission status of models in a subscription network, which solves the technical problem that the transmission status of AI / ML models cannot be effectively analyzed, thus preventing the network from effectively adjusting its transmission strategy based on the transmission status of AI / ML models and preventing third parties from adjusting application layer information based on the analysis of the transmission status of AI / ML models.

[0006] In a first aspect, this application provides a method for analyzing the transmission state of a model in a subscription network, the method being applied to an application function (AF), the method comprising:

[0007] The first message is sent directly or through the Network Capability Opening Function (NEF) to the Network Data Analysis Function (NWDAF); wherein the first message is used to request subscription to analysis information on the transmission status of artificial intelligence / machine learning (AI / ML) models in the network;

[0008] The NEF receives AI / ML model transmission status analysis information sent by the NWDAF directly or through the NEF. The analysis information is determined by the NWDAF based on data of AI / ML model transmission status sent by other network functions 5GC NF(s) of the 5G core network.

[0009] The analytical information is used to adjust network policy parameters and / or application layer model information.

[0010] Optionally, the data on the transmission status of the AI / ML model is obtained by the NWDAF by sending a second message to the 5GC NF(s) based on the parameters requested in the received first message. The second message is used to collect data for analyzing the transmission status of the AI / ML model in the network.

[0011] Optionally, the parameters requested in the first message include at least one of the following: network data analysis identifier, identifier of a user equipment (UE) or group of UEs receiving the AI / ML model or any UE that meets the analysis conditions, identifier of the application using the AI / ML model, region of AI / ML model transmission, network slice indicating the protocol data unit (PDU) session transmitting the AI / ML model quality of service flow, data network indicating the PDU session transmitting the AI / ML model quality of service flow, time period of AI / ML model transmission, start timestamp of AI / ML model transmission, end timestamp of AI / ML model transmission, size of AI / ML transmission model, quality of service requirements for indicating the quality of service flow of AI / ML model transmission, and / or specific quality of service requirements for indicating the transmission of AI / ML model.

[0012] The second message includes at least one of the following: the current location of the UE using the AI / ML model, the identifier of the application using the AI / ML model, the service quality flow identifier for transmitting the AI / ML model, the uplink bit rate and downlink bit rate for transmitting the AI / ML model, the uplink packet delay and downlink packet delay of the AI / ML model, the number of abnormal releases of the service quality flow during the AI / ML model transmission period, the number of packet transmissions of the AI / ML model, the number of packet retransmissions of the AI / ML model, the data acquisition time, the duration of the AI / ML model transmission, the start timestamp of the AI / ML model transmission, the end timestamp of the AI / ML model transmission, the size of the AI / ML transmission model, the network slice of the PDU session used for transmitting the AI / ML model service quality flow, the data network of the PDU session used for transmitting the AI / ML model service quality flow, and the service process for the AF.

[0013] The analysis information includes at least one of the following: network slice of the PDU session used to transmit the AI / ML model quality of service flow, identifier of the application using the AI / ML model, region information of the region using the AI / ML model, validity period of the analysis results, User Plane Function (UPF) providing AI / ML model transmission, data network name of the PDU session used to transmit the AI / ML model quality of service flow, size of the AI / ML transmission model, duration of AI / ML model transmission, start timestamp of AI / ML model transmission, end timestamp of AI / ML model transmission, quality of service flow identifier for transmitting the AI / ML model, uplink bit rate and downlink bit rate for transmitting the AI / ML model, uplink packet delay and downlink packet delay of the AI / ML model, number of abnormal releases of the quality of service flow during the AI / ML model transmission period, number of times the reporting threshold for abnormal release of the quality of service flow during the AI / ML model transmission period is reached, number of packet transmissions of the AI / ML model, and number of packet retransmissions of the AI / ML model.

[0014] If the AI / ML model performs federated learning, the parameters requested in the first message also include: federated learning group information, which includes at least one of the following: the identifier of the federated learning group used to indicate the analysis, the UE identifier or UE(s) identifier of the UE participating in the federated learning, and the application identifier of the UE participating in the federated learning.

[0015] Accordingly, the second message also includes at least one of the following: the identifier of the federated learning group for indicating analysis, the UE identifier or UE(s) identifier of the UE participating in the federated learning, and the application identifier of the UE participating in the federated learning;

[0016] Accordingly, the analysis information also includes at least one of the following: the identifier of the federated learning group used to indicate the analysis, the UE identifier or UE(s) identifier of the UE participating in the federated learning, and the identifier of each application that provides the AI / ML model or participates in the federated learning.

[0017] In this embodiment, by sending a request to NWDAF to subscribe to analysis information on the transmission status of AI / ML models in the network, and receiving analysis information from NWDAF determined based on the collected 5GC NF(s) data, the network policy parameters and / or application layer model information are adjusted according to the analysis information. This enables effective analysis of the transmission status of AI / ML models, thereby allowing the network to effectively adjust its transmission policy based on the transmission status of AI / ML models and enabling third parties to obtain the analysis of the transmission status of AI / ML models and adjust the application layer information accordingly.

[0018] Optionally, after receiving the analysis information, the method further includes:

[0019] Based on the analysis information, a first request is sent directly or through the NEF to the policy control function PCF;

[0020] The first request is used to request an update to the network policy parameters for AI / ML model transmission; the network policy parameters are used to optimize the AI / ML model transmission status.

[0021] Optionally, sending a first request to the Policy Control Function (PCF) directly or through the NEF based on the analysis information includes:

[0022] Based on at least one of the following analysis information: uplink bit rate and downlink bit rate of the transmitted AI / ML model, uplink packet delay and downlink packet delay of the AI / ML model, number of abnormal releases of the Quality of Service (QoS) stream during the AI / ML model transmission period, number of packet transmissions of the AI / ML model, number of packet retransmissions of the AI / ML model, and number of times the reporting threshold for abnormal release of the QoS stream during the AI / ML model transmission period is reached, new QoS parameters for the transmitted AI / ML model are determined; the new QoS parameters include at least one of the following: 5G QoS identifier, reflective QoS control, maximum uplink bit rate of the transmitted AI / ML model, maximum downlink bit rate of the transmitted AI / ML model, minimum uplink bit rate of the transmitted AI / ML model, minimum downlink bit rate of the transmitted AI / ML model, and priority of the QoS stream;

[0023] Based on the identifier of the application transmitting the AI / ML model, the region information of the AI / ML model, the IP address information of the application service using the AI / ML model, the network slice of the PDU session used to transmit the AI / ML model service quality flow, and the data network name of the PDU session used to transmit the AI / ML model service quality flow, the region information and address information of the UE(s) transmitting the AI / ML model and each AF are determined; or, if the AI / ML model performs federated learning, based on the identifier of the federated learning group indicating the analysis, the UE identifier or UE(s) identifier participating in the federated learning, and the identifiers of each application indicating the provision of the AI / ML model or participation in the federated learning, the region information and address information of the UE(s) transmitting the AI / ML model and each AF are determined.

[0024] Based on the area information and address information of the UE(s) transmitting the AI / ML model and each AF, the data network access identifier (DNAI) and the area information and address information of the UE(s) and each AF corresponding to the DNAI are determined. The area information and address information of the UE(s) and each AF corresponding to the DNAI are used to provide a path for optimizing the transmission status of the AI / ML model.

[0025] The new quality of service parameters, the DNAI, and the area information and address information of the UE(s) and each AF corresponding to the DNAI are sent to the PCF directly or through the NEF as parameters in the first request.

[0026] Optionally, determining the Data Network Access Identifier (DNAI) and the corresponding UE(s) and AF's area and address information based on the area and address information of the UE(s) transmitting the AI / ML model and each AF includes:

[0027] Based on the UE(s) transmitting AI / ML models and the area and address information of each AF, determine whether the current routing path is poor;

[0028] If the current routing path is not good, the destination address of both parties in the transmission AI / ML model is determined based on the address information of the UE(s) and each AF in the transmission AI / ML model, as well as the area information of the UE(s).

[0029] Determine the nearest path based on the destination address;

[0030] Based on the nearest path, determine the DNAI, as well as the region information and address information of the UE(s) and each AF corresponding to the DNAI.

[0031] Optionally, the first request is specifically used for:

[0032] The system requests the PCF to adjust the 5G Quality of Service identifier, reflective Quality of Service control, maximum uplink bit rate of the transmission AI / ML model, maximum downlink bit rate of the transmission AI / ML model, minimum uplink bit rate of the transmission AI / ML model, minimum downlink bit rate of the transmission AI / ML model, and priority of the Quality of Service flow in the PCC rules according to the new Quality of Service parameters, and instructs the PCF to provide the adjusted first update result directly or through NEF; wherein, the first update result is determined by the PCF based on the result of adjusting the PCC rules according to the new Quality of Service parameters;

[0033] Accordingly, the method further includes:

[0034] The first update result sent by the PCF is received directly or through the NEF, and the first update result includes whether the first request is accepted or the first request is rejected.

[0035] Optionally, the first request is specifically used for:

[0036] The PCF requests the SMF to determine whether the Session Management Function (SMF) needs to update the Session Management Policy. If it is determined that the SMF needs to update the Session Management Policy, the PCF sends a second request to the SMF. The parameters requested in the second request include at least one of the following: DNAI, Traffic Orientation Policy Identifier, and Traffic Route Information. The second request is used by the SMF to determine the selected User Plane Function (UPF) based on the new Session Management Policy and provide the corresponding DNAI, Traffic Orientation Policy Identifier, and Traffic Route Information.

[0037] Accordingly, the method further includes:

[0038] The PCF receives the second update result directly or through the NEF. The second update result is determined by the PCF based on whether to update the UPF path according to the new session management policy sent by the SMF.

[0039] The second update result includes whether the first request is accepted or rejected.

[0040] In this embodiment, new service quality parameters are determined based on the analysis information, and the new service quality parameters are sent to the PCF. The PCF then adjusts the PCC rules accordingly based on the new service quality parameters or updates the SM policy through the SMF and provides the DNAI, as well as the area information and address information of the UE(s) and each AF corresponding to the DNAI. By receiving the notification sent by the PCF indicating whether the first request is accepted or rejected, it is possible to request the adjustment of the network policy based on the NWDAF analysis results to optimize the transmission status of the AI / ML model.

[0041] Optionally, after receiving the analysis information, the method further includes:

[0042] Based on the analysis information, the information of the application layer model is adjusted. The information of the application layer model includes at least one of the following: model compression, model size, model transmission time period, and model encoding / decoding. The information of the application layer model is used to update the quality of service parameters.

[0043] Based on the information from the adjusted application layer model, new service quality parameters are determined. These new service quality parameters include: 5G service quality identifier, reflective service quality control, maximum uplink bit rate of the transmission AI / ML model, maximum downlink bit rate of the transmission AI / ML model, minimum uplink bit rate of the transmission AI / ML model, minimum downlink bit rate of the transmission AI / ML model, and priority of the service quality flow.

[0044] Send a third request directly or through the NEF to the Policy Control Function (PCF);

[0045] The parameters requested in the third request include the new quality of service parameters, and the third request is used to request an update to the quality of service parameters.

[0046] Optionally, the third request is specifically used for:

[0047] The PCF is requested to adjust the 5G Quality of Service identifier, reflective quality of service control, maximum uplink bit rate of transmission AI / ML model, maximum downlink bit rate of transmission AI / ML model, minimum uplink bit rate of transmission AI / ML model, minimum downlink bit rate of transmission AI / ML model, and priority in the PCC rules according to the new quality of service parameters.

[0048] Accordingly, the method further includes:

[0049] The third update result is received directly or through NEF from the PCF. The third update result is determined by the PCF based on the results of adjusting the PCC rules. The third update result includes whether the third request is accepted or rejected.

[0050] Optionally, after adjusting the information of the application layer model, the method further includes:

[0051] The adjusted application layer model information, including model compression, model size, and model encoding / decoding, is directly sent to the PCF. This adjusted application layer model information is used to support the PCF in adjusting the 5G service quality identifier, reflective service quality control, maximum uplink bit rate of the transmission AI / ML model, maximum downlink bit rate of the transmission AI / ML model, minimum uplink bit rate of the transmission AI / ML model, minimum downlink bit rate of the transmission AI / ML model, and service quality flow priority in the PCC rules.

[0052] Accordingly, the method further includes:

[0053] The system receives a fourth update result from the PCF, which is determined by the PCF adjusting the PCC rules based on the information from the adjusted application layer model. The fourth update result includes whether the third request is accepted or rejected.

[0054] Optionally, after adjusting the information of the application layer model, the method further includes:

[0055] The model transmission time in the adjusted application layer model information is directly sent to the PCF. The model transmission time in the adjusted application layer model information is used to support the PCF in adjusting the gate state parameters in the PCC rules. The gate state parameters are used to support the SMF in updating the session management policy based on the transmission start time and transmission end time in the gate state.

[0056] Accordingly, the method further includes:

[0057] The PCF receives a fifth update result, which is determined by the PCF based on the new session management policy received from the SMF. The fifth update result includes whether the third request is accepted or rejected.

[0058] In this embodiment, based on the analysis information, the information of the application layer model is adjusted. Based on the adjusted application layer model information, new service quality parameters are determined and sent to the PCF. This causes the PCF to adjust the PCC rules accordingly. Alternatively, the adjusted application layer model information is directly sent to the PCF, causing the PCF to adjust the aforementioned service quality parameters in the PCC rules based on model compression, model size, and model encoding / decoding. Or, the PCF adjusts the gate state parameters in the PCC rules based on the model transmission time, allowing the SMF to update the session management policy based on the transmission start time and transmission end time in the gate state. By receiving notifications from the PCF indicating whether a third request is accepted or rejected, the information of the application layer model can be adjusted based on the NWDAF analysis results, thereby updating QoS requirements and optimizing the AI / ML model transmission status.

[0059] Secondly, this application provides a method for analyzing model transmission state in a subscription network, the method being applied to the Network Data Analysis Function (NWDAF), the method comprising:

[0060] The application function (AF) receives a first message directly or through the Network Capability Opening Function (NEF); wherein the first message is used to request subscription to analysis information on the transmission status of artificial intelligence / machine learning (AI / ML) models in the network;

[0061] Based on the parameters requested in the first message, a second message is sent to other network functions 5GC NF(s) of the 5G core network. The second message is used to collect data for analyzing the transmission status of AI / ML models in the network.

[0062] It receives AI / ML model transmission status data sent by other network functions of the 5G core network, 5GC NF(s), and analyzes the AI / ML model transmission status data to obtain AI / ML model transmission status analysis information.

[0063] The analytical information is used to adjust network policy parameters and / or application layer model information through AF.

[0064] Optionally, the parameters requested in the first message include at least one of the following: network data analysis identifier, identifier of a user equipment (UE) or group of UEs receiving the AI / ML model or any UE that meets the analysis conditions, identifier of the application using the AI / ML model, region of AI / ML model transmission, network slice indicating the protocol data unit (PDU) session transmitting the AI / ML model quality of service flow, data network indicating the PDU session transmitting the AI / ML model quality of service flow, time period of AI / ML model transmission, start timestamp of AI / ML model transmission, end timestamp of AI / ML model transmission, size of AI / ML transmission model, quality of service requirements for indicating the quality of service flow of AI / ML model transmission, and / or specific quality of service requirements for indicating the transmission of AI / ML model.

[0065] The second message includes at least one of the following: the current location of the UE using the AI / ML model, the identifier of the application using the AI / ML model, the service quality flow identifier for transmitting the AI / ML model, the uplink bit rate and downlink bit rate for transmitting the AI / ML model, the uplink packet delay and downlink packet delay of the AI / ML model, the number of abnormal releases of the service quality flow during the AI / ML model transmission period, the number of packet transmissions of the AI / ML model, the number of packet retransmissions of the AI / ML model, the data acquisition time, the duration of the AI / ML model transmission, the start timestamp of the AI / ML model transmission, the end timestamp of the AI / ML model transmission, the size of the AI / ML transmission model, the network slice of the PDU session used for transmitting the AI / ML model service quality flow, the data network of the PDU session used for transmitting the AI / ML model service quality flow, and the service process for the AF.

[0066] The analysis information includes at least one of the following: network slice of the PDU session used to transmit the AI / ML model quality of service flow, identifier of the application using the AI / ML model, region information of the region using the AI / ML model, validity period of the analysis results, User Plane Function (UPF) providing AI / ML model transmission, data network name of the PDU session used to transmit the AI / ML model quality of service flow, size of the AI / ML transmission model, duration of AI / ML model transmission, start timestamp of AI / ML model transmission, end timestamp of AI / ML model transmission, quality of service flow identifier for transmitting the AI / ML model, uplink bit rate and downlink bit rate for transmitting the AI / ML model, uplink packet delay and downlink packet delay of the AI / ML model, number of abnormal releases of the quality of service flow during the AI / ML model transmission period, number of times the reporting threshold for abnormal release of the quality of service flow during the AI / ML model transmission period is reached, number of packet transmissions of the AI / ML model, and number of packet retransmissions of the AI / ML model.

[0067] If the AI / ML model performs federated learning, the parameters requested in the first message also include: federated learning group information, which includes at least one of the following: the identifier of the federated learning group used to indicate the analysis, the UE identifier or UE(s) identifier of the UE participating in the federated learning, and the application identifier of the UE participating in the federated learning.

[0068] Accordingly, the second message also includes at least one of the following: the identifier of the federated learning group for indicating analysis, the UE identifier or UE(s) identifier of the UE participating in the federated learning, and the application identifier of the UE participating in the federated learning;

[0069] Accordingly, the analysis information also includes at least one of the following: the identifier of the federated learning group used to indicate the analysis, the UE identifier or UE(s) identifier of the UE participating in the federated learning, and the identifier of each application that provides the AI / ML model or participates in the federated learning.

[0070] Thirdly, this application provides a model transmission state analysis device in a subscription network, the device comprising: a memory, a transceiver, and a processor.

[0071] A memory for storing computer programs; a transceiver for sending and receiving data under the control of the processor; and a processor for reading the computer programs from the memory and performing the following operations:

[0072] The first message is sent directly or through the Network Capability Opening Function (NEF) to the Network Data Analysis Function (NWDAF); wherein the first message is used to request subscription to analysis information on the transmission status of artificial intelligence / machine learning (AI / ML) models in the network;

[0073] The NEF receives AI / ML model transmission status analysis information sent by the NWDAF directly or through the NEF. The analysis information is determined by the NWDAF based on data of AI / ML model transmission status sent by other network functions 5GC NF(s) of the 5G core network.

[0074] The analytical information is used to adjust network policy parameters and / or application layer model information.

[0075] Fourthly, this application provides a model transmission state analysis device in a subscription network, the device comprising a memory, a transceiver, and a processor:

[0076] A memory for storing computer programs; a transceiver for sending and receiving data under the control of the processor; and a processor for reading the computer programs from the memory and performing the following operations:

[0077] The application function (AF) receives a first message directly or through the Network Capability Opening Function (NEF); wherein the first message is used to request subscription to analysis information on the transmission status of artificial intelligence / machine learning (AI / ML) models in the network;

[0078] Based on the parameters requested in the first message, a second message is sent to other network functions 5GC NF(s) of the 5G core network. The second message is used to collect data for analyzing the transmission status of AI / ML models in the network.

[0079] It receives AI / ML model transmission status data sent by other network functions of the 5G core network, 5GC NF(s), and analyzes the AI / ML model transmission status data to obtain AI / ML model transmission status analysis information.

[0080] The analytical information is used to adjust network policy parameters and / or application layer model information through AF.

[0081] Fifthly, this application provides a model transmission state analysis apparatus in a subscription network, the apparatus comprising:

[0082] The sending unit is used to send a first message directly or through the Network Capability Opening Function (NEF) to the Network Data Analysis Function (NWDAF); wherein, the first message is used to request subscription to analysis information on the transmission status of artificial intelligence / machine learning (AI / ML) models in the network;

[0083] The analysis unit is used to receive, directly or through the NEF, analysis information of the AI / ML model transmission status sent by the NWDAF. The analysis information is determined by the NWDAF based on data of the AI / ML model transmission status sent by other network functions 5GC NF(s) of the 5G core network.

[0084] The analytical information is used to adjust network policy parameters and / or application layer model information.

[0085] Sixthly, this application provides a model transmission state analysis apparatus in a subscription network, the apparatus comprising:

[0086] The receiving unit is used to receive a first message sent by the application function AF directly or through the Network Capability Opening Function (NEF); wherein the first message is used to request the subscription of analysis information on the transmission status of artificial intelligence / machine learning (AI / ML) models in the network;

[0087] The sending unit is configured to send a second message to other network functions 5GCNF(s) of the 5G core network according to the parameters requested in the first message. The second message is used to collect data for analyzing the transmission status of AI / ML models in the network.

[0088] The analysis unit is used to receive AI / ML model transmission status data sent by other network functions 5GC NF(s) of the 5G core network, and analyze the AI / ML model transmission status data to obtain analysis information of AI / ML model transmission status.

[0089] The analytical information is used to adjust network policy parameters and / or application layer model information through AF.

[0090] In a seventh aspect, this application provides a processor-readable storage medium storing a computer program for causing the processor to perform the method described in either the first or second aspect.

[0091] This application provides a method, apparatus, and readable storage medium for analyzing the transmission status of models in a subscription network. The method involves sending a first message directly or via the Network Capability Open Function (NEF) to the Network Data Analysis Function (NWDAF). The first message requests analysis information on the transmission status of artificial intelligence / machine learning (AI / ML) models in the subscription network. The method also involves receiving the AI / ML model transmission status analysis information sent by the NWDAF directly or via the NEF. This analysis information is determined by the NWDAF based on AI / ML model transmission status data received from other network functions (5GC NF(s)) of the 5G core network. The AI / ML model transmission status data is obtained by the NWDAF by sending a second message to the 5GC NF(s) based on parameters requested in the received first message. This second message is used to collect data for analyzing the transmission status of AI / ML models in the network. The analysis information is used to adjust network policy parameters and / or application layer model information. By sending a request to NWDAF to subscribe to analysis information on the transmission status of AI / ML models in the network, and receiving analysis information from NWDAF determined based on the collected 5GC NF(s) data, the network policy parameters and / or application layer model information are adjusted according to the analysis information. This enables effective analysis of the transmission status of AI / ML models, allowing the network to effectively adjust its transmission policy based on the transmission status of AI / ML models, and enabling third parties to obtain the analysis of the transmission status of AI / ML models and adjust application layer information accordingly.

[0092] It should be understood that the description in the foregoing summary section is not intended to limit the key or essential features of the embodiments of the present invention, nor is it intended to restrict the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0093] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0094] Figure 1 A network architecture diagram of the model transmission state analysis method in the subscription network provided in the embodiments of this application;

[0095] Figure 2 A network architecture diagram of 5GC supporting network data analysis is provided for embodiments of this application;

[0096] Figure 3A flowchart illustrating the model transmission state analysis method in a subscription network provided in Embodiment 1 of this application;

[0097] Figure 4 A schematic diagram of the signaling flow of the model transmission state analysis method in the subscription network provided in Embodiment 1 of this application;

[0098] Figure 5 A schematic diagram of the signaling flow of the model transmission state analysis method in the subscription network provided in Embodiment 2 of this application;

[0099] Figure 6 A schematic diagram of the signaling flow for the model transmission state analysis method in the subscription network provided in Embodiment 3 of this application;

[0100] Figure 7 A flowchart illustrating the model transmission state analysis method in a subscription network provided in Embodiment 4 of this application;

[0101] Figure 8 This is a schematic diagram of the structure of a model transmission state analysis device in a subscription network provided in an embodiment of this application;

[0102] Figure 9 A schematic diagram of the structure of a model transmission state analysis device in a subscription network provided in another embodiment of this application;

[0103] Figure 10 A schematic diagram of the structure of a model transmission state analysis device in a subscription network provided in yet another embodiment of this application;

[0104] Figure 11 A schematic diagram of the structure of a model transmission state analysis device in a subscription network provided in another embodiment of this application. Detailed Implementation

[0105] In this application, the term "and / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0106] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0107] To clearly understand the technical solution of this application, the solutions of the prior art will first be described in detail. In the prior art, the SA1 R18 requirements approved by SA#93e require AI / ML model transmission in at least the following scenarios:

[0108] Scenario 1: Distribution and sharing of AI / ML models. Due to changes in tasks or environments, mobile terminals have limited memory and cannot pre-load all models. Therefore, mobile terminals need to download new AI / ML models from the network in real time via the 5G system.

[0109] Scenario 2: Federated learning algorithm using 5GS. When training a global model on a cloud server, it is necessary to aggregate the models trained locally on various terminal devices. Each training iteration process: a terminal device downloads a global model from the cloud server and trains it using local data; the terminal reports the intermediate training results to the cloud server; the cloud server aggregates the intermediate training results from all terminals and updates the global model, then redistributes the global model to the terminals; the terminals then execute the next iteration.

[0110] Scenario 3: AI / ML Model Segmentation Between AI / ML Endpoints. An AI / ML model can be segmented into multiple parts based on the current task or environment. The trend is to perform computationally complex and energy-intensive inference on the network, while inference on the terminal for parts requiring privacy protection or sensitive to latency. For example, the terminal downloads / loads a model, infers specific layers / parts first, and then sends the intermediate results to the network; the network then executes the remaining layers / parts and feeds back the inference results to the terminal. This scenario involves transmitting part of the model in the first step or intermediate step, and therefore may include model transmission.

[0111] Therefore, as a channel for transmitting AI / ML models, in order to improve the intelligence capabilities of 5G networks and meet the requirements of TS 22.261 (passed by SA1#93e) for the transmission of AI / ML models in 5G systems, 5G systems need to support the exposure of monitoring and status information about AI-ML sessions to third parties. However, currently, there is no analysis of the transmission status of AI / ML models, third parties cannot effectively adjust their own behavior based on the transmission status of AI / ML models, and the network cannot effectively adjust its status based on the transmission status of AI / ML models.

[0112] Further research by the inventors revealed that effective analysis of the transmission status of AI / ML models requires interaction between Application Functions (AF), Network Exposure Functions (NEF), Network Data Analytics Functions (NWDAF), and various Network Functions (NF). For example... Figure 1 As shown, the AF can send a request to the NWDAF directly or through the NEF to indicate that it is subscribing to the AI / ML model transmission status analysis in the network. The NWDAF analyzes the AI / ML model transmission status in the network and feeds back the data by collecting data from various network functions (NFs) in the 5G core network (5GC). This enables the network to effectively analyze the AI / ML model transmission status, thereby allowing the network to effectively adjust its state based on the AI / ML model transmission status, and enabling third parties to adjust their own behavior data based on the analysis of the AI / ML model transmission status.

[0113] Therefore, based on the inventors' inventive research, this application proposes a method for analyzing the transmission status of models in a subscription network. In this application, a first message is sent directly or through the Network Capability Open Function (NEF) to the Network Data Analysis Function (NWDAF). The first message requests analysis information on the transmission status of AI / ML models in the subscription network. The NEF receives the analysis information on the transmission status of AI / ML models sent by the NWDAF. This analysis information is determined by the NWDAF based on data on the transmission status of AI / ML models sent by other network functions of the 5G core network (i.e., 5GC NF(s)). The data on the transmission status of AI / ML models is obtained by the NWDAF by sending a second message to the 5GC NF(s) based on the parameters requested in the received first message. This second message is used to collect data for analyzing the transmission status of AI / ML models in the network. The analysis information is used to adjust network policy parameters and / or application layer model information. By sending a request to NWDAF to subscribe to analysis information on the transmission status of AI / ML models in the network, and receiving analysis information from NWDAF determined based on the collected 5GC NF(s) data, the network policy parameters and / or application layer model information are adjusted according to the analysis information. This enables effective analysis of the transmission status of AI / ML models, allowing the network to effectively adjust its transmission policy based on the transmission status of AI / ML models, and enabling third parties to obtain the analysis of the transmission status of AI / ML models and adjust application layer information accordingly.

[0114] Figure 2 The network architecture diagram of 5GC supporting network data analysis provided in the embodiments of this application is as follows: Figure 2 As shown in this embodiment, NWDAF is a network analysis function managed by the operator. NWDAF can provide data analysis services to 5GC network functions, application functions (AF), and operation administration and maintenance (OAM). The analysis results can be historical statistics or predictive information. NWDAF can serve one or more network slices.

[0115] The 5GC also includes several other functions, namely: User Plane Function (UPF), Session Management Function (SMF), Access and Mobility Management Function (AMF), Unified Data Repository (UDR), Network Exposure Function (NEF), AF, Policy Control Function (PCF), and Online Charging System (OCS). These other functions are collectively referred to as NFs. The NWDAF communicates with other functional entities in the 5G core network, including 5GC NF(s) and OAM, through service-oriented interfaces.

[0116] In 5GC, different NWDAF instances can provide different types of specialized analytics. For consumer NFs to discover suitable NWDAF instances to provide specific types of analytics, NWDAF instances must provide their supported Analytic ID when registering with the Network Repository Function (NRF). The Analytic ID represents the analytics type (or analytics identifier). This allows consumer NFs to provide the Analytic ID when querying the NRF for NWDAF instances, indicating the type of analytics needed. 5GC network functions and OAM determine how to use the data analytics provided by the Network Data Analytics Function (NWDAF) to improve network performance.

[0117] In one application scenario, the Application Filter (AF) requests the Network Data Layer Filter (NWDAF) to provide AI / ML model transmission status analysis. The analysis results (or analysis information) include: the application identifier (i.e., Application ID) using the AI / ML model, the region information using the AI / ML model, the time period for transmitting the AI / ML model, the size of the transmitted model, the Quality of Service (QoS) related information for transmitting the AI / ML model, the network slice used for AI / ML model transmission, and the Data Network Name (DNN) information. If federated learning is involved, it also includes: the group identifier (i.e., federated learning group ID), the UE ID or UE group ID participating in federated learning, and the address information of the application server providing the model or participating in federated learning. Based on the data analysis provided by the NWDAF, the AF requests adjustments to the 5GS network policy or adjustments to the application layer AI / ML model parameters to optimize the AI / ML model transmission status. The 5GC NF(s) adjusts the network policy based on the request, or the AF adjusts the application layer AI / ML model parameters based on the data analysis.

[0118] When the AF sends an analysis request to the NWDAF, if the AF is in the trusted zone, the AF can send the request directly to the NWDAF; if the AF is in the trusted zone, the AF can send the request to the NWDAF through the NEF, that is, the AF sends the request to the NEF, and then the NEF sends the request to the NWDAF.

[0119] Therefore, by sending a request directly or through NEF to NWDAF to subscribe to analysis information on the transmission status of AI / ML models in the network, receiving analysis information from NWDAF determined based on the collected 5GC NF(s) data, and adjusting network policy parameters and / or application layer model information based on the analysis information, the system effectively analyzes the transmission status of AI / ML models. This enables the network to effectively adjust its transmission policy based on the transmission status of AI / ML models and allows third parties to obtain the analysis of the transmission status of AI / ML models and adjust application layer information accordingly.

[0120] The embodiments of this application will now be described with reference to the accompanying drawings.

[0121] Figure 3 This is a flowchart illustrating the model transmission state analysis method in a subscription network provided in Embodiment 1 of this application, as shown below. Figure 3 As shown, the execution entity of the model transmission state analysis method in the subscription network provided in this embodiment is AF. Therefore, the model transmission state analysis method in the subscription network provided in this embodiment includes the following steps:

[0122] Step 101: The AF sends the first message directly or through the Network Capability Opening Function (NEF) to the Network Data Analysis Function (NWDAF).

[0123] The first message is used to request the subscription of analysis information on the transmission status of artificial intelligence / machine learning (AI / ML) models in the network.

[0124] In this embodiment, the parameters requested in the first message include at least one of the following: network data analysis identifier (i.e., Analytics ID), identifier of a user equipment (UE) or a group of UEs receiving the AI / ML model, or any UE that meets the analysis conditions (i.e., Target of Analytics Reporting), identifier of the application using the AI / ML model (i.e., Application ID), area of ​​AI / ML model transmission (i.e., Area of ​​Interest (AoI)), network slice (i.e., S-NSSAI) indicating the protocol data unit (PDU) session transmitting the AI / ML model quality of service flow, data network (i.e., DNN) indicating the PDU session transmitting the AI / ML model quality of service flow, time period of AI / ML model transmission (i.e., Model transmission duration), start timestamp of AI / ML model transmission (i.e., Model transmission start), end timestamp of AI / ML model transmission (i.e., Model transmission stop), size of AI / ML transmission model (i.e., Model size), quality of service requirements (i.e., 5QI (5G QoS Identifier)) used to indicate the quality of service flow of AI / ML model transmission, and / or specific quality of service requirements (i.e., QoS Characteristics) used to indicate the transmission of AI / ML model.

[0125] Specific quality of service requirements, such as packet transmission latency and packet error rate, are specified. See Table 1 below for an example table of parameters requested in the first message.

[0126] Table 1: Example of parameters requested in the first message

[0127]

[0128] If federated learning exists, the parameters requested in the first message may also include: Federated Learning(FL) group information; the Federated Learning(FL) group information includes at least one of the following: the identifier of the federated learning group used to indicate the analysis (i.e., Federated Learning(FL) group ID), the UE identifier or UE(s) identifier participating in the federated learning (i.e., Federated Learning(FL) UE ID or UE group ID), and the application identifier participating in the federated learning (i.e., Federated Learning(FL) Application ID). See Table 2 below for an example table of parameters requested in the first message.

[0129] Table 2: Example of parameters requested in the first message

[0130]

[0131] In this embodiment, if the AF is untrusted (i.e., the AF is not in the trusted zone), the AF sends an AI / ML model transmission status subscription request to the NEF, such as Nnef_AnalyticsExposure_Subscribe (i.e., analysis open subscription) or Nnef_AnalyticsExposure_Fetch (i.e., analysis open fetch) request. The NEF sends a first message to the NWDAF, which can be an AI / ML model open transmission status subscription Nnwdaf_AnalyticsSubscription_Subscribe (i.e., analysis subscription subscription) or Nnwdaf_AnalyticsInfo_Request (i.e., analysis information) request. This request can carry parameters as shown in the table, requesting to subscribe to the analysis information of the AI / ML model transmission status in the network. If the AF is trusted (i.e., the AF is in the trusted zone), the AF directly sends the first message to the NWDAF.

[0132] The AI / ML model transmission status subscription request can carry parameters as shown in Table 1 or Table 2, requesting to subscribe to the analysis information of the AI / ML model transmission status in the network.

[0133] Step 102: The AF receives the analysis information of the AI / ML model transmission status sent by the NWDAF directly or through the NEF.

[0134] The analysis information is determined by the NWDAF based on the data of AI / ML model transmission status sent by other network functions 5GC NF(s) of the 5G core network.

[0135] Optionally, the data on the transmission status of the AI / ML model is obtained by the NWDAF by sending a second message to the 5GC NF(s) based on the parameters requested in the received first message. The second message is used to collect data for analyzing the transmission status of the AI / ML model in the network.

[0136] In this embodiment, the second message includes at least one of the following: the current location of the UE using the AI / ML model (i.e., UE location), the identifier of the application using the AI / ML model (i.e., Application ID, which can be the identifier of the server or the identifier of the AF), the Quality of Service Flow Identifier (i.e., QFI) for transmitting the AI / ML model, the bit rate for the uplink direction (i.e., UL direction) and the bit rate for the downlink direction (i.e., DL direction) of the AI / ML model, the packet delay for the uplink direction (i.e., UL direction) and the packet delay for the downlink direction (i.e., DL direction) of the AI / ML model, the number of abnormal releases of the Quality of Service Flow during the AI / ML model transmission period (QoSSustainability), the number of packet transmissions of the AI / ML model (packet transmission), and the number of packet retransmissions of the AI / ML model (packet retransmission). The data includes: retransmission, data acquisition time (i.e., timestamp), AI / ML model transmission duration (i.e., AI / ML model transmission time period), AI / ML model transmission start timestamp, AI / ML model transmission end timestamp, AI / ML transmission model size, network slice of the PDU session used to transmit the AI / ML model QoS flow, data network of the PDU session used to transmit the AI / ML model QoS flow, and service flow (i.e., IP filter information) used for the AF. See the example table of the second message shown in Table 3 below.

[0137] Table 3: Example Table of Second Message

[0138]

[0139] If federated learning is present, the second message also includes at least one of the following: an identifier for the federated learning group used to specify the analysis (i.e., Federated Learning(FL)group ID), the UE or UE(s) participating in the federated learning (i.e., Federated Learning(FL)UE ID or UE group ID), and the application identifier participating in the federated learning (i.e., Federated Learning(FL)Application ID). See Table 4 below for an example table of second messages.

[0140] Table 4: Example Table of Second Message

[0141]

[0142] In this embodiment, if the AF is untrusted, it receives the AI / ML model transmission status analysis information sent by the NEF to the NEF. If the AF is trusted, it directly receives the AI / ML model transmission status analysis information sent by the NWDAF.

[0143] The analysis information includes at least one of the following: network slices for PDU sessions used to transmit AI / ML model quality of service flows, identifiers of applications using AI / ML models, region information for using AI / ML models, validity period of analysis results, and user plane function UPF (i.e., UPF) providing AI / ML model transmission. The data network name of the PDU session used to transmit the AI / ML model Quality of Service (QoS) flow, the size of the AI / ML transmission model, the duration of the AI / ML model transmission, the start timestamp of the AI / ML model transmission, the end timestamp of the AI / ML model transmission, and the QoS requirements. The QoS requirements include: the QoS flow identifier (QFI) for transmitting the AI / ML model, the uplink and downlink bit rates for transmitting the AI / ML model, the uplink and downlink packet delays for the AI / ML model, the number of abnormal releases of the QoS flow during the AI / ML model transmission period, the number of times the reporting threshold for abnormal releases of the QoS flow during the AI / ML model transmission period was reached, the number of AI / ML model packet transmissions, and the number of AI / ML model packet retransmissions. See Table 5 below for an example of the analysis information.

[0144] Table 5: Example Table of Analysis Information

[0145]

[0146]

[0147] If federated learning is present, the analysis information also includes at least one of the following: an identifier for specifying the federated learning group for analysis, an identifier of the UE participating in the federated learning or an identifier of the UE(s) participating in the federated learning, and an identifier (i.e., Application Server Instance Address) indicating the application providing the AI / ML model or participating in the federated learning. See Table 6 below for an example table of analysis information.

[0148] Table 6: Example Table of Analysis Information

[0149]

[0150]

[0151] In this embodiment, by sending a request to NWDAF to subscribe to analysis information on the transmission status of AI / ML models in the network, and receiving analysis information from NWDAF determined based on the collected 5GC NF(s) data, the network policy parameters and / or application layer model information are adjusted according to the analysis information. This enables effective analysis of the transmission status of AI / ML models, allowing the network to effectively adjust its transmission policy based on the transmission status of AI / ML models, and enabling third parties to obtain the analysis of the transmission status of AI / ML models and adjust application layer information accordingly.

[0152] For example, see Figure 4 As shown, Figure 4 This is a schematic diagram of the signaling flow of the model transmission state analysis method in the subscription network provided in Embodiment 1 of this application. Figure 4 This is a signaling interaction diagram between AF and NWDAF, NEF, and NF in the model transmission state analysis method in a subscription network. The model transmission state analysis method in a subscription network provided in this embodiment includes the following steps (i.e., the signaling interaction process corresponding to Embodiment 1: AF requests NWDAF to provide AI / ML model transmission state analysis): (wherein, steps 4011 to 4016 are when AF is in an untrusted area, and steps 4021 to 4024 are when AF is in a trusted area.)

[0153] Step 4011: If the AF is untrusted, the AF sends an AI / ML model transfer status subscription request to the NEF, either Nnef_AnalyticsExposure_Subscribe or Nnef_AnalyticsExposure_Fetch.

[0154] In this embodiment, the request may carry parameters as shown in Table 1 or Table 2, requesting to subscribe to analysis information on the transmission status of AI / ML models in the network.

[0155] Step 4012: NEF sends an AI / ML model open transport state subscription request, either Nnwdaf_AnalyticsSubscription_Subscribe (analysis subscription) or Nnwdaf_AnalyticsInfo_Request (analysis information), to NWDAF.

[0156] Among them, the AI / ML model open transmission state subscription Nnwdaf_AnalyticsSubscription_Subscriber Nnwdaf_AnalyticsInfo_Request request can be used as the first message.

[0157] In this embodiment, the request may carry parameters as shown in Table 1 or Table 2, requesting to subscribe to analysis information on the transmission status of AI / ML models in the network.

[0158] Step 4013: NWDAF calls Nnf_EventExposure_Subscribe (i.e., event open subscription) to collect data from 5GC NF(s).

[0159] The collected data, as shown in Table 3 or Table 4, is used to analyze the transmission status of AI / ML models in the network. The NWDAF can send the second message to the 5GC NF(s) by calling Nnf_EventExposure_Subscribe.

[0160] Steps 4014 and 5GC NF(s) call Nnf_EventExposure_Notify (i.e., event exposure notification) to send the required data back to NWDAF.

[0161] Step 4015: NWDAF calls Nnwdaf_AnalyticsSubscription_Notify (analysis subscription notification) or Nnwdaf_AnalyticsInfo_Request response (analysis information request response) to send analysis information about the AI / ML model transmission status to NEF.

[0162] Step 4016: NEF calls Nnef_AnalyticsExposure_Notify (analysis open notification) or Nnef_AnalyticsExposure_Fetch response (analysis open fetch response) to send analysis information about the AI / ML model transmission status to AF.

[0163] The analysis information is shown in Table 5 or Table 6.

[0164] Step 4021: If the AF is trusted, the AF directly sends an AI / ML model transmission state subscription request to the NWDAF, performing the operation described in step 4012. Additionally, the consumer can also be a PCF or an SMF.

[0165] Step 4022: Perform the operation as described in step 4013. That is, step 4013.

[0166] Step 4023: Perform the operation as described in step 4014. That is, step 4014.

[0167] Step 4024: NWDAF directly sends the analysis information of the AI / ML model transmission status to AF, and performs the operation as described in step 4016.

[0168] Example 2: After receiving the analysis information, the method further includes:

[0169] Based on the analysis information, a first request is sent directly or through the NEF to the Policy Control Function (PCF).

[0170] The first request is used to request an update to the network policy parameters for AI / ML model transmission; the network policy parameters are used to optimize the AI / ML model transmission status.

[0171] Specifically, based on the NWDAF analysis results, the AF requests adjustments to the network policy to optimize the AI / ML model transmission status. Specifically, if the AF is in the trusted zone, it directly sends a first request to the PCF, requesting the PCF to update the network policy parameters used for AI / ML model transmission. If the AF is not in the trusted zone, it sends a first request to the PCF through the NEF, requesting the PCF to update the network policy parameters used for AI / ML model transmission.

[0172] Optionally, sending a first request to the Policy Control Function (PCF) directly or through the NEF based on the analysis information can be achieved through the following steps:

[0173] Step a1: Based on at least one of the following in the analysis information: uplink bit rate and downlink bit rate of the AI / ML model transmission, uplink packet delay and downlink packet delay of the AI / ML model, number of abnormal releases of the quality of service flow during the AI / ML model transmission period, number of packet transmissions of the AI / ML model, number of packet retransmissions of the AI / ML model, and number of times the reporting threshold for abnormal release of the quality of service flow during the AI / ML model transmission period is reached, determine new quality of service parameters for the transmission of the AI / ML model; the new quality of service parameters include at least one of the following: 5G quality of service identifier, reflective quality of service control, maximum uplink bit rate of the AI / ML model transmission, maximum downlink bit rate of the AI / ML model transmission, minimum uplink bit rate of the AI / ML model transmission, minimum downlink bit rate of the AI / ML model transmission, and priority of the quality of service flow.

[0174] Step a2: Based on the identifier of the application transmitting the AI / ML model, the region information of the AI / ML model, the IP address information of the application service using the AI / ML model, the network slice of the PDU session used to transmit the AI / ML model service quality flow, and the data network name of the PDU session used to transmit the AI / ML model service quality flow, determine the region information and address information of the UE(s) transmitting the AI / ML model and each AF; or, if the AI / ML model performs federated learning, based on the identifier of the federated learning group indicating the analysis, the UE identifier or UE(s) identifier participating in the federated learning, and the identifiers of each application indicating the provision of the AI / ML model or participation in the federated learning, determine the region information and address information of the UE(s) transmitting the AI / ML model and each AF.

[0175] Step a3: Based on the area information and address information of the UE(s) transmitting the AI / ML model and each AF, determine the data network access identifier (DNAI) and the area information and address information of the UE(s) and each AF corresponding to the DNAI. The area information and address information of the UE(s) and each AF corresponding to the DNAI are used to provide a path for optimizing the transmission status of the AI / ML model.

[0176] Step a4: Send the new quality of service parameters, the DNAI, and the area information and address information of the UE(s) and each AF corresponding to the DNAI to the PCF directly or through the NEF as parameters in the first request.

[0177] Specifically, the AF determines new QoS parameters for the transmission AI / ML model based on the analysis information about the transmission AI / ML model obtained from the NWDAF, such as: the uplink bit rate and downlink bit rate of the transmission AI / ML model, the uplink packet delay and downlink packet delay of the AI / ML model, the number of abnormal releases of the QoS flow during the transmission period of the AI / ML model, the number of packet transmissions of the AI / ML model, the number of packet retransmissions of the AI / ML model, and the number of times the reporting threshold for abnormal release of the QoS flow during the transmission period of the AI / ML model is reached, and provides the new QoS parameters to the PCF.

[0178] The AF determines the region and address information of the UE(s) transmitting the AI / ML model based on the information obtained from the NWDAF regarding the transmission of AI / ML model analysis, such as: the identifier of the application transmitting the AI / ML model, the region information of the AI / ML model, the IP address information of the application service using the AI / ML model, the network slice of the PDU session used to transmit the AI / ML model service quality flow, and the data network name of the PDU session used to transmit the AI / ML model service quality flow. Alternatively, the AF determines the region and address information of the UE(s) transmitting the AI / ML model and each AF based on the information obtained from the NWDAF regarding the transmission of AI / ML model analysis, such as: the identifier of the federated learning group indicating the analysis, the UE identifier or UE(s) identifier participating in the federated learning, and the identifiers of each application indicating the provision of the AI / ML model or participation in the federated learning. Then, the AF determines the data network access identifier (DNAI) and the region and address information of the UE(s) and each AF corresponding to the DNAI. The AF sends a first request, carrying new QoS parameters, DNAI, and the corresponding UE(s) and area and address information of each AF, directly or through the NEF to the PCF, requesting the PCF to update the relevant policy parameters for AI / ML model transmission based on the parameters carried in the first request, thereby optimizing the AI / ML model transmission status.

[0179] Alternatively, step a3 can be achieved through the following steps:

[0180] Step a31: Based on the area information and address information of the UE(s) transmitting the AI / ML model and each AF, determine whether the current routing path is unsuitable;

[0181] Step a32: If the current routing path is not good, determine the destination address of both parties in the transmission AI / ML model based on the address information of the UE(s) and each AF in the transmission AI / ML model and the area information of the UE(s).

[0182] Step a33: Determine the nearest path based on the destination address;

[0183] Step a34: Based on the nearest path, determine the DNAI, as well as the region information and address information of the UE(s) and each AF corresponding to the DNAI.

[0184] Specifically, if the current routing path is deemed unsuitable, a DNAI that can provide a better service experience or performance is selected, and the PCF is provided with the DNAI for transmitting the AI / ML model, as well as the corresponding UE(s) and AF(s) area information and address information.

[0185] Optionally, the first request is specifically used for:

[0186] The system requests the PCF to adjust the 5G Quality of Service identifier, reflective Quality of Service control, maximum uplink bit rate of the transmission AI / ML model, maximum downlink bit rate of the transmission AI / ML model, minimum uplink bit rate of the transmission AI / ML model, minimum downlink bit rate of the transmission AI / ML model, and priority of the Quality of Service flow in the PCC rules according to the new Quality of Service parameters, and instructs the PCF to provide the adjusted first update result directly or through NEF; wherein, the first update result is determined by the PCF based on the new Quality of Service parameters when adjusting the PCC rules;

[0187] Accordingly, the method further includes:

[0188] The first update result sent by the PCF is received directly or through the NEF, and the first update result includes whether the first request is accepted or the first request is rejected.

[0189] Specifically, the AF requests the PCF to adjust the 5G Quality of Service identifier, reflective quality of service control, maximum uplink bit rate of the transmission AI / ML model, maximum downlink bit rate of the transmission AI / ML model, minimum uplink bit rate of the transmission AI / ML model, minimum downlink bit rate of the transmission AI / ML model, and priority of the quality of service flow in the PCC rules according to the new QoS parameters provided. The AF then notifies the AF, either directly or through the NEF, of the first update result, i.e., whether the request is accepted or rejected. Optionally, the first request is specifically used for:

[0190] The PCF requests the SMF to determine whether the Session Management Function (SMF) needs to update the Session Management Policy. If it is determined that the SMF needs to update the Session Management Policy, the PCF sends a second request to the SMF. The parameters requested in the second request include at least one of the following: DNAI, Traffic Orientation Policy Identifier, and Traffic Route Information. The second request is used by the SMF to determine the selected User Plane Function (UPF) according to the new Session Management Policy and provide the corresponding DNAI, Traffic Orientation Policy Identifier, and Traffic Route Information.

[0191] Accordingly, the method further includes:

[0192] The PCF receives the second update result directly or through the NEF. The second update result is determined by the PCF based on whether to update the UPF path according to the new session management policy sent by the SMF.

[0193] The second update result includes whether the first request is accepted or rejected.

[0194] Specifically, the AF requests the PCF to determine whether the Session Management Function (SMF) needs to update its session management policy. If the PCF determines that the SMF needs to update its policy information, it will send a second request to the SMF carrying request parameters such as DNAI, Traffic Orientation Policy Identifier, and Traffic Route Information. The SMF then determines the selected User Plane Function (UPF) based on the new session management policy and provides the corresponding DNAI, Traffic Orientation Policy Identifier, and Traffic Route Information to update the session management policy, i.e., update the SM policy. The PCF then notifies the AF of the second update result, either directly or through the NEF, i.e., whether the request was accepted or rejected.

[0195] For example, see Figure 5 As shown, Figure 5 This is a schematic diagram of the signaling flow of the model transmission state analysis method in the subscription network provided in Embodiment 2 of this application. Figure 5 This is a signaling interaction diagram between AF, NEF, and PCF in the model transmission state analysis method in a subscription network. The model transmission state analysis method in a subscription network provided in this embodiment includes the following steps (i.e., the signaling interaction process corresponding to Embodiment 2: based on the NWDAF analysis results, AF requests an adjustment to the network strategy to optimize the AI / ML model transmission state): (wherein, steps 5011 to 5014 represent AF in an untrusted area, and steps 5021 to 5022 represent AF in a trusted area.)

[0196] Step 5010: AF signs up with NWDAF and retrieves AI / ML model transmission status analysis.

[0197] In this embodiment, AF obtains AI / ML model transmission status analysis by subscribing to NEF or directly from NWDAF, as described in steps 4011 to 4024 above.

[0198] Step 5011: AF sends an Nnef_AFsessionWithQoS_Update (i.e., AF session update based on quality of service) request to NEF.

[0199] In this embodiment, if the AF is not trusted, for an established AF session with QoS requirements, the AF can send an Nnef_AFsessionWithQoS_Update (i.e., AF session update based on quality of service) request to the NEF to update the relevant policy parameters used for AI / ML model transmission, thereby optimizing the AI / ML model transmission status.

[0200] Specifically, based on the information obtained from NWDAF regarding the transmission AI / ML model analysis, including: QoS flow Bit Rate (i.e., the uplink and downlink bit rates of the transmission AI / ML model), QoS flow Packet Delay (i.e., the uplink and downlink packet delays of the AI / ML model), QoS Sustainability, Packet Transmission, Packet Retransmission, QoS Sustainability, and QoS Sustainability Reporting Threshold(s), the AF determines new QoS parameters for the transmission AI / ML model and provides new QoS parameters to the PCF: 5G QoS Identifier (5QI), Reflective QoS Control, UL-maximum bitrate (the maximum uplink bit rate of the transmission AI / ML model), and DL-maximum bitrate. Bitrate (i.e., the maximum bit rate in the downlink direction for transmitting AI / ML models), UL-guaranteedbitrate (i.e., the minimum bit rate in the uplink direction for transmitting AI / ML models), DL-guaranteed bitrate (i.e., the minimum bit rate in the downlink direction for transmitting AI / ML models), and Priority Level (i.e., the priority level), which increases or decreases with the increase or decrease of prediction results to ensure that 5GS meets the QoS requirements for model transmission.

[0201] In addition, the AF determines the region and address information of the UE(s) and AF(s) transmitting the AI / ML model based on the analysis information obtained from the NWDAF regarding the transmission of the AI / ML model: the AFID of the transmitting AI / ML model, the region information of the UE using the AI / ML model, the IP address information of the application service using the AI / ML model, and the network slice and DNN information used for AI / ML model transmission. Alternatively, the analysis information may also contain information about federated learning groups: Federated Learning (FL) group ID, Federated Learning (FL) UE ID or UE group ID, and the Application ID of Federated Learning (FL), to determine the region and address information of the UE(s) and AF(s) transmitting the AI / ML model, determine that the current routing path is not good, select the DNAI that can provide a better service experience or performance, and provide the PCF with the DNAI of the transmitting AI / ML model and the corresponding region and address information of the UE(s) and AF(s).

[0202] The process of determining that the current routing path is undesirable can be as follows: The AF determines the current routing path based on the AF's IP address information and the UE(s)'s area information (which can correspond to the IDs of AMF, SMF, UPF, etc., and the N6 interface of UPF connects to DN). For example, if some UEs / servers join / leave (federation group) before the next transmission, some paths will be undesirable, or if the number of hops is too high, the path will be undesirable.

[0203] The process of selecting a DNAI that can provide a better service experience or performance can be as follows: Based on the received IP address information and the area information of the UE(s), the destination addresses of both parties can be determined, and a better (nearest) routing path can be found, corresponding to the DNAI.

[0204] Step 5012: NEF sends an Npcf_PolicyAuthorization_Update request to PCF.

[0205] In this embodiment, NEF sends the above information to PCF via Npcf_PolicyAuthorization_Update request (i.e., policy authorization update request) to update the relevant policy parameters used for AI / ML model transmission, thereby optimizing the AI / ML model transmission status.

[0206] The Npcf_PolicyAuthorization_Update request can be used as a second request.

[0207] Step 5013: PCF notifies NEF of the result (i.e., PCF sends Npcf_PolicyAuthorization_Update response to NEF).

[0208] In this embodiment, based on the information provided by NEF, specifically, the PCF adjusts the following parameters in the PCC rules accordingly: 5G QoS Identifier (5QI), Reflective QoS Control, UL-maximumbitrate, DL-maximum bitrate, UL-guaranteed bitrate, DL-guaranteed bitrate, and Priority Level, and sends an Npcf_PolicyAuthorization_Update response to NEF to notify the NEF of the results.

[0209] If the PCF determines that the SMF needs to update its policy information, the PCF will send an Npcf_SMPolicyControl_UpdateNotify request (i.e., Session Policy Control Update Notification Request) to the SMF (DNAI, Per DNAI: Traffic steering policy identifier, Per DNAI: N6 traffic routing information) to update the SM policy. The SMF will then select a UPF based on this policy and provide its own DNAI (Per DNAI: Traffic steering policy identifier, Per DNAI: N6 traffic routing information). The Npcf_SMPolicyControl_UpdateNotify request can be used as a second request.

[0210] Step 5014: NEF sends the Nnef_AFsessionWithQoS_Update response to AF.

[0211] In this embodiment, NEF sends an Nnef_AFsessionWithQoS_Update response (i.e., an AF session update response based on quality of service) to notify the AF whether the request is accepted or rejected.

[0212] Step 5021: AF directly sends an Npcf_PolicyAuthorization_Update request to PCF.

[0213] In this embodiment, if the AF is trusted, the AF directly sends an Npcf_PolicyAuthorization_Update request to the PCF to update the relevant policy parameters for AI / ML model transmission, thereby optimizing the AI / ML model transmission status, and performs the operation as described in step 5012.

[0214] Step 5022: PCF directly notifies AF of the result (i.e., PCF directly sends Npcf_PolicyAuthorization_Update response to AF).

[0215] In this embodiment, firstly, the PCF performs the operation described in step 5012 based on the information provided by the AF. The PCF directly notifies the AF whether the request is accepted or rejected.

[0216] In Example 3, after receiving the analysis information, the method can also be implemented through the following steps:

[0217] Step b1: Based on the analysis information, adjust the information of the application layer model, wherein the information of the application layer model includes at least one of the following: model compression, model size, model transmission time period, and model encoding / decoding; the information of the application layer model is used to update the quality of service parameters.

[0218] Step b2: Based on the information of the adjusted application layer model, determine the new quality of service parameters, which include: 5G quality of service identifier, reflective quality of service control, maximum bit rate in the uplink direction of the transmission AI / ML model, maximum bit rate in the downlink direction of the transmission AI / ML model, minimum bit rate in the uplink direction of the transmission AI / ML model, minimum bit rate in the downlink direction of the transmission AI / ML model, and priority of the quality of service flow.

[0219] Step b3: Send a third request directly or through the NEF to the Policy Control Function (PCF);

[0220] The parameters requested in the third request include the new quality of service parameters, and the third request is used to request an update to the quality of service parameters.

[0221] Specifically, the Application Layer AF (AF) updates the Quality of Service (QoS) parameters by adjusting the application layer model information based on the QoS requirement information provided by the NWDAF. This includes adjusting model compression, model size, model transmission time period, and model encoding / decoding. Based on the adjusted application layer model information, the AF determines the new QoS parameters and sends a third request carrying these new QoS parameters directly or via the NEF or PCF.

[0222] Optionally, the third request is specifically used for:

[0223] The PCF is requested to adjust the 5G Quality of Service identifier, reflective quality of service control, maximum uplink bit rate of transmission AI / ML model, maximum downlink bit rate of transmission AI / ML model, minimum uplink bit rate of transmission AI / ML model, minimum downlink bit rate of transmission AI / ML model, and priority in the PCC rules according to the new quality of service parameters.

[0224] Accordingly, the method further includes:

[0225] The third update result is received directly or through NEF from the PCF. The third update result is determined by the PCF based on the results of adjusting the PCC rules. The third update result includes whether the third request is accepted or rejected.

[0226] Specifically, the AF requests the PCF to adjust the 5G Quality of Service identifier, reflective quality of service control, maximum uplink bit rate of the transmission AI / ML model, maximum downlink bit rate of the transmission AI / ML model, minimum uplink bit rate of the transmission AI / ML model, minimum downlink bit rate of the transmission AI / ML model, and priority in the PCC (Policy and Payment Control) rules according to the new QoS parameters provided. The PCF adjusts the PCC rules and notifies the AF directly or through the NEF whether the request is accepted or rejected.

[0227] Optionally, after adjusting the information of the application layer model, the method may also be implemented through the following steps:

[0228] The adjusted application layer model information, including model compression, model size, and model encoding / decoding, is directly sent to the PCF. This adjusted application layer model information is used to support the PCF in adjusting the 5G service quality identifier, reflective service quality control, maximum uplink bit rate of the transmission AI / ML model, maximum downlink bit rate of the transmission AI / ML model, minimum uplink bit rate of the transmission AI / ML model, minimum downlink bit rate of the transmission AI / ML model, and service quality flow priority in the PCC rules.

[0229] Accordingly, the method further includes:

[0230] The system receives a fourth update result from the PCF, which is determined by the PCF adjusting the PCC rules based on the information from the adjusted application layer model. The fourth update result includes whether the third request is accepted or rejected.

[0231] Specifically, the PCF adjusts the aforementioned QoS parameters in the PCC rules based on model compression, model size, and model encoding / decoding. These parameters include: 5G Quality of Service identifier, reflective quality of service control, maximum uplink bit rate of the transmitted AI / ML model, maximum downlink bit rate of the transmitted AI / ML model, minimum uplink bit rate of the transmitted AI / ML model, minimum downlink bit rate of the transmitted AI / ML model, and the priority of the quality of service flow. The PCF adjusts the PCC rules and notifies the AF, directly or through the NEF, whether the request is accepted or rejected.

[0232] Optionally, after adjusting the information of the application layer model, the method further includes:

[0233] The model transmission time in the adjusted application layer model information is directly sent to the PCF. The model transmission time in the adjusted application layer model information is used to support the PCF in adjusting the gate state parameters in the PCC rules. The gate state parameters are used to support the SMF in updating the session management policy based on the transmission start time and transmission end time in the gate state.

[0234] Accordingly, the method further includes:

[0235] The PCF receives a fifth update result, which is determined by the PCF based on the new session management policy received from the SMF. The fifth update result includes whether the third request is accepted or rejected.

[0236] Specifically, the PCF adjusts the Gate status (i.e., gate state parameter) in the PCC rule based on the model transmission time, updates the SM policy, and the SMF feeds back to the PCF based on this, affecting the start and end times of the flow transmission. The PCF then notifies the AF, either directly or through the NEF, whether the request is accepted or rejected.

[0237] For example, see Figure 6 As shown, Figure 6 This is a schematic diagram of the signaling flow of the model transmission state analysis method in the subscription network provided in Embodiment 3 of this application. Figure 6This is a signaling interaction diagram between the AF, NEF, and PCF in the model transmission state analysis method in the subscription network. The model transmission state analysis method in the subscription network provided in this embodiment includes the following steps (i.e., the signaling interaction process corresponding to Embodiment 3: the AF adjusts the information of the application layer model based on the analysis information provided by the NWDAF, such as adjusting model compression, model size, model transmission time period, model encoding / decoding, etc., and then updates the QoS requirements, similar to the steps in Embodiment 2): (wherein, steps 6012 to 6014 are when the AF is in the untrusted area, and steps 6021 to 6022 are when the AF is in the trusted area.)

[0238] Step 6010: AF signs up with NWDAF and retrieves the AI / ML model transmission status analysis.

[0239] Specifically, AF obtains AI / ML model transmission status analysis by subscribing to NEF or directly from NWDAF, as described in steps 4011 to 4024 above.

[0240] Step 6011: Based on the analysis information, AF adjusts the application layer behavior.

[0241] In this embodiment, AF adjusts the information of the application layer model based on the QoS requirement information provided by NWDAF, such as adjusting model compression, model size, model transmission time period, model encoding and decoding, etc.

[0242] Step 6012: AF sends an Nnef_AFsessionWithQoS_Update request to NEF.

[0243] Step 6013: NEF sends an Npcf_PolicyAuthorization_Update request to PCF.

[0244] Among them, the Npcf_PolicyAuthorization_Update request can be used as a third request.

[0245] Step 6014: PCF notifies NEF of the result (i.e., PCF sends the Npcf_PolicyAuthorization_Update response to NEF).

[0246] Step 6015: NEF sends the Nnef_AFsessionWithQoS_Update response to AF.

[0247] Step 6021: AF directly sends an Npcf_PolicyAuthorization_Update request to PCF.

[0248] Step 6022: PCF directly notifies AF of the result (i.e., PCF directly sends Npcf_PolicyAuthorization_Update response to AF).

[0249] Specifically, the AF requests the session to update the QoS. Specifically, based on adjustments to model compression, model size, model transmission time period, and model encoding / decoding, the AF determines new QoS parameters for transmitting the AI / ML model and provides these new QoS parameters to the PCF: 5G QoS Identifier (5QI), Reflective QoS Control, UL-maximum bitrate, DL-maximum bitrate, UL-guaranteed bitrate, DL-guaranteed bitrate, and Priority Level. Alternatively, the AF can directly send the model compression, model size, model transmission time period, and model encoding / decoding adjustment information to the PCF.

[0250] Step 6013: Based on the provided new QoS parameters, the PCF adjusts the following QoS parameters in the PCC rules: 5G QoSIdentifier (5QI), Reflective QoS Control, UL-maximum bitrate, DL-maximum bitrate, UL-guaranteed bitrate, DL-guaranteed bitrate, and Priority Level; or the PCF adjusts the above QoS parameters in the PCC rules based on model compression, model size, and model encoding / decoding; or the PCF adjusts the Gate status in the PCC rules based on the model transmission time, updates the SM policy, and the SMF affects the start and end times of the flow transmission accordingly; finally, the PCF notifies the AF whether the request is accepted or rejected through steps 5014 or 5022 in Embodiment 2.

[0251] This application describes a network-wide AI / ML model transmission status analysis (AI / ML model transmission status analysis) request sent by the Application Controller (AF), along with the parameters included in the request. The NWDAF collects input data from the 5GC NF(s) to analyze the AI / ML model transmission status in the network. The NWDAF performs the analysis and sends the AI / ML model transmission status analysis information to the AF. Based on the AI / ML model transmission status analysis information, the AF requests a QoS update for the AF session and adjusts the policies of relevant network elements such as the Policy Control Function (PCF) and Session Management Function (SMF). Based on the AI / ML model transmission status analysis information, the AF adjusts the relevant parameters of the application layer model information, thereby adjusting the QoS and optimizing the AI / ML model transmission status. This allows third parties to obtain the AI / ML model transmission status and, based on the analysis results, adjust the network's behavior to meet AI / ML model transmission requirements. Furthermore, third parties can adjust the behavior of the model application layer based on the analysis results to achieve efficient AI / ML model transmission, ensuring a good service experience and performance.

[0252] Figure 7 This is a flowchart illustrating the model transmission state analysis method in a subscription network provided in Embodiment 4 of this application, as shown below. Figure 7 As shown, the execution entity of the model transmission state analysis method in the subscription network provided in this embodiment is NWDAF. Therefore, the model transmission state analysis method in the subscription network provided in this embodiment includes the following steps:

[0253] Step 701: The NWDAF receives a first message sent by the application function AF directly or through the Network Capability Opening Function (NEF); wherein, the first message is used to request subscription to analysis information on the transmission status of artificial intelligence / machine learning (AI / ML) models in the network;

[0254] Step 702: NWDAF sends a second message to other network functions 5GCNF(s) of the 5G core network according to the parameters requested in the first message. The second message is used to collect data for analyzing the transmission status of AI / ML models in the network.

[0255] Step 703: NWDAF receives AI / ML model transmission status data sent by other network functions of the 5G core network, 5GC NF(s), and analyzes the AI / ML model transmission status data to obtain AI / ML model transmission status analysis information.

[0256] The analytical information is used to adjust network policy parameters and / or application layer model information through AF.

[0257] Optionally, the parameters requested in the first message include at least one of the following: network data analysis identifier, identifier of a user equipment (UE) or group of UEs receiving the AI / ML model or any UE that meets the analysis conditions, identifier of the application using the AI / ML model, region of AI / ML model transmission, network slice indicating the protocol data unit (PDU) session transmitting the AI / ML model quality of service flow, data network indicating the PDU session transmitting the AI / ML model quality of service flow, time period of AI / ML model transmission, start timestamp of AI / ML model transmission, end timestamp of AI / ML model transmission, size of AI / ML transmission model, quality of service requirements for indicating the quality of service flow of AI / ML model transmission, and / or specific quality of service requirements for indicating the transmission of AI / ML model.

[0258] The second message includes at least one of the following: the current location of the UE using the AI / ML model, the identifier of the application using the AI / ML model, the service quality flow identifier for transmitting the AI / ML model, the uplink bit rate and downlink bit rate for transmitting the AI / ML model, the uplink packet delay and downlink packet delay of the AI / ML model, the number of abnormal releases of the service quality flow during the AI / ML model transmission period, the number of packet transmissions of the AI / ML model, the number of packet retransmissions of the AI / ML model, the data acquisition time, the duration of the AI / ML model transmission, the start timestamp of the AI / ML model transmission, the end timestamp of the AI / ML model transmission, the size of the AI / ML transmission model, the network slice of the PDU session used for transmitting the AI / ML model service quality flow, the data network of the PDU session used for transmitting the AI / ML model service quality flow, and the service process for the AF.

[0259] The analysis information includes at least one of the following: network slice of the PDU session used to transmit the AI / ML model quality of service flow, identifier of the application using the AI / ML model, region information of the region using the AI / ML model, validity period of the analysis results, User Plane Function (UPF) providing AI / ML model transmission, data network name of the PDU session used to transmit the AI / ML model quality of service flow, size of the AI / ML transmission model, duration of AI / ML model transmission, start timestamp of AI / ML model transmission, end timestamp of AI / ML model transmission, quality of service flow identifier for transmitting the AI / ML model, uplink bit rate and downlink bit rate for transmitting the AI / ML model, uplink packet delay and downlink packet delay of the AI / ML model, number of abnormal releases of the quality of service flow during the AI / ML model transmission period, number of times the reporting threshold for abnormal release of the quality of service flow during the AI / ML model transmission period is reached, number of packet transmissions of the AI / ML model, and number of packet retransmissions of the AI / ML model.

[0260] If the AI / ML model performs federated learning, the parameters requested in the first message also include: federated learning group information, which includes at least one of the following: the identifier of the federated learning group used to indicate the analysis, the UE identifier or UE(s) identifier of the UE participating in the federated learning, and the application identifier of the UE participating in the federated learning.

[0261] Accordingly, the second message also includes at least one of the following: the identifier of the federated learning group for specifying the analysis, the UE identifier or UE(s) identifier of the UE participating in the federated learning, and the application identifier of the UE participating in the federated learning;

[0262] Accordingly, the analysis information also includes at least one of the following: the identifier of the federated learning group for specifying the analysis, the UE identifier or UE(s) identifier for participating in the federated learning, and the identifier of each application that provides the AI / ML model or participates in the federated learning.

[0263] In this embodiment, by receiving a request from the AF (Agent Focused Atmosphere) for analysis information on the transmission status of AI / ML models in the subscription network, and according to the parameters requested in the first message, data is collected from other network functions (5GC NF(s)) of the 5G core network. The system receives AI / ML model transmission status data sent by other network functions (5GC NF(s)) of the 5G core network, analyzes the AI / ML model transmission status data, and obtains analysis information on the AI / ML model transmission status. This enables effective analysis of the AI / ML model transmission status, allowing the AF to adjust network policy parameters and / or application layer model information based on the analysis information. Consequently, the network can effectively adjust its transmission policy based on the AI / ML model transmission status, and third parties can obtain the analysis of the AI / ML model transmission status to adjust application information.

[0264] It should be noted that the model transmission state analysis method in the subscription network provided in this application embodiment can achieve Figure 4 All method steps implemented in the method embodiment shown are capable of achieving the same technical effect. Therefore, the parts that are the same as those in the method embodiment and their beneficial effects will not be described in detail here.

[0265] Figure 8 This is a schematic diagram of the structure of the model transmission state analysis device in the subscription network provided in the embodiments of this application, as shown below. Figure 8 As shown, the model transmission status analysis device in the subscription network provided in this embodiment is applied to AF. The model transmission status analysis device in the subscription network provided in this embodiment includes: a transceiver 800, used to receive and send data under the control of a processor 810.

[0266] Among them, Figure 8 In this context, the bus architecture can include any number of interconnected buses and bridges, specifically linking various circuits together, represented by one or more processors (processor 810) and memory (memory 820). The bus architecture can also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. The bus interface provides an interface. The transceiver 800 can be multiple elements, including transmitters and receivers, providing units for communicating with various other devices over transmission media, including wireless channels, wired channels, optical fibers, etc. The processor 810 is responsible for managing the bus architecture and general processing, and the memory 820 can store data used by the processor 810 during operation.

[0267] The processor 810 can be a central processing unit (CPU), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a complex programmable logic device (CPLD). The processor can also adopt a multi-core architecture.

[0268] In this embodiment, the memory 820 is used to store computer programs; the transceiver 800 is used to send and receive data under the control of the processor 810; the processor 810 is used to read the computer programs in the memory and perform the following operations:

[0269] The first message is sent directly or through the Network Capability Opening Function (NEF) to the Network Data Analysis Function (NWDAF); wherein the first message is used to request subscription to analysis information on the transmission status of artificial intelligence / machine learning (AI / ML) models in the network;

[0270] The NEF receives AI / ML model transmission status analysis information sent by the NWDAF directly or through the NEF. The analysis information is determined by the NWDAF based on AI / ML model transmission status data sent by other network functions of the 5G core network, 5GC NF(s). The AI / ML model transmission status data is obtained by the NWDAF by sending a second message to the 5GC NF(s) based on the parameters requested in the received first message. The second message is used to collect data for analyzing the AI / ML model transmission status in the network.

[0271] The analytical information is used to adjust network policy parameters and / or application layer model information.

[0272] Optionally, the parameters requested in the first message include at least one of the following: network data analysis identifier, identifier of a user equipment (UE) or group of UEs receiving the AI / ML model or any UE that meets the analysis conditions, identifier of the application using the AI / ML model, region of AI / ML model transmission, network slice indicating the protocol data unit (PDU) session transmitting the AI / ML model quality of service flow, data network indicating the PDU session transmitting the AI / ML model quality of service flow, time period of AI / ML model transmission, start timestamp of AI / ML model transmission, end timestamp of AI / ML model transmission, size of AI / ML transmission model, quality of service requirements for indicating the quality of service flow of AI / ML model transmission, and / or specific quality of service requirements for indicating the transmission of AI / ML model.

[0273] The second message includes at least one of the following: the current location of the UE using the AI / ML model, the identifier of the application using the AI / ML model, the service quality flow identifier for transmitting the AI / ML model, the uplink bit rate and downlink bit rate for transmitting the AI / ML model, the uplink packet delay and downlink packet delay of the AI / ML model, the number of abnormal releases of the service quality flow during the AI / ML model transmission period, the number of packet transmissions of the AI / ML model, the number of packet retransmissions of the AI / ML model, the data acquisition time, the duration of the AI / ML model transmission, the start timestamp of the AI / ML model transmission, the end timestamp of the AI / ML model transmission, the size of the AI / ML transmission model, the network slice of the PDU session used for transmitting the AI / ML model service quality flow, the data network of the PDU session used for transmitting the AI / ML model service quality flow, and the service process for the AF.

[0274] The analysis information includes at least one of the following: network slice of the PDU session used to transmit the AI / ML model quality of service flow, identifier of the application using the AI / ML model, region information of the region using the AI / ML model, validity period of the analysis results, User Plane Function (UPF) providing AI / ML model transmission, data network name of the PDU session used to transmit the AI / ML model quality of service flow, size of the AI / ML transmission model, duration of AI / ML model transmission, start timestamp of AI / ML model transmission, end timestamp of AI / ML model transmission, quality of service flow identifier for transmitting the AI / ML model, uplink bit rate and downlink bit rate for transmitting the AI / ML model, uplink packet delay and downlink packet delay of the AI / ML model, number of abnormal releases of the quality of service flow during the AI / ML model transmission period, number of times the reporting threshold for abnormal release of the quality of service flow during the AI / ML model transmission period is reached, number of packet transmissions of the AI / ML model, and number of packet retransmissions of the AI / ML model.

[0275] If the AI / ML model performs federated learning, the parameters requested in the first message also include: federated learning group information, which includes at least one of the following: the identifier of the federated learning group used to indicate the analysis, the UE identifier or UE(s) identifier participating in the federated learning, and the application identifier participating in the federated learning.

[0276] Accordingly, the second message also includes at least one of the following: the identifier of the federated learning group for indicating analysis, the UE identifier or UE(s) identifier of the UE participating in the federated learning, and the application identifier of the UE participating in the federated learning;

[0277] Accordingly, the analysis information also includes at least one of the following: the identifier of the federated learning group for specifying the analysis, the UE identifier or UE(s) identifier for participating in the federated learning, and the identifier of each application that provides the AI / ML model or participates in the federated learning.

[0278] Optionally, the processor 810 is also used for:

[0279] Upon receiving the analysis information, a first request is sent directly or through the NEF to the Policy Control Function (PCF) based on the analysis information.

[0280] The first request is used to request an update to the network policy parameters for AI / ML model transmission; the network policy parameters are used to optimize the AI / ML model transmission status.

[0281] Optionally, when the processor 810 sends a first request to the policy control function PCF directly or through the NEF based on the analysis information, it specifically includes:

[0282] Based on at least one of the following analysis information: uplink bit rate and downlink bit rate of the transmitted AI / ML model, uplink packet delay and downlink packet delay of the AI / ML model, number of abnormal releases of the Quality of Service (QoS) stream during the AI / ML model transmission period, number of packet transmissions of the AI / ML model, number of packet retransmissions of the AI / ML model, and number of times the reporting threshold for abnormal release of the QoS stream during the AI / ML model transmission period is reached, new QoS parameters for the transmitted AI / ML model are determined; the new QoS parameters include at least one of the following: 5G QoS identifier, reflective QoS control, maximum uplink bit rate of the transmitted AI / ML model, maximum downlink bit rate of the transmitted AI / ML model, minimum uplink bit rate of the transmitted AI / ML model, minimum downlink bit rate of the transmitted AI / ML model, and priority of the QoS stream;

[0283] Based on the identifier of the application transmitting the AI / ML model, the region information of the AI / ML model, the IP address information of the application service using the AI / ML model, the network slice of the PDU session used to transmit the AI / ML model service quality flow, and the data network name of the PDU session used to transmit the AI / ML model service quality flow, the region information and address information of the UE(s) transmitting the AI / ML model and each AF are determined; or, if the AI / ML model performs federated learning, based on the identifier of the federated learning group indicating the analysis, the UE identifier or UE(s) identifier participating in the federated learning, and the identifiers of each application indicating the provision of the AI / ML model or participation in the federated learning, the region information and address information of the UE(s) transmitting the AI / ML model and each AF are determined.

[0284] Based on the area information and address information of the UE(s) transmitting the AI / ML model and each AF, the data network access identifier (DNAI) and the area information and address information of the UE(s) and each AF corresponding to the DNAI are determined. The area information and address information of the UE(s) and each AF corresponding to the DNAI are used to provide a path for optimizing the transmission status of the AI / ML model.

[0285] The new quality of service parameters, the DNAI, and the area information and address information of the UE(s) and each AF corresponding to the DNAI are sent to the PCF directly or through the NEF as parameters in the first request.

[0286] Optionally, the processor 810, when determining the Data Network Access Identifier (DNAI) and the corresponding UE(s) and the area and address information of each AF based on the area and address information of the UE(s) transmitting the AI / ML model and each AF, specifically includes:

[0287] Based on the UE(s) transmitting AI / ML models and the area and address information of each AF, determine whether the current routing path is poor;

[0288] If the current routing path is not good, the destination address of both parties in the transmission AI / ML model is determined based on the address information of the UE(s) and each AF in the transmission AI / ML model, as well as the area information of the UE(s).

[0289] Determine the nearest path based on the destination address;

[0290] Based on the nearest path, determine the DNAI, as well as the region information and address information of the UE(s) and each AF corresponding to the DNAI.

[0291] Optionally, the first request is specifically used for:

[0292] The system requests the PCF to adjust the 5G Quality of Service identifier, reflective Quality of Service control, maximum uplink bit rate of the transmission AI / ML model, maximum downlink bit rate of the transmission AI / ML model, minimum uplink bit rate of the transmission AI / ML model, minimum downlink bit rate of the transmission AI / ML model, and priority of the Quality of Service flow in the PCC rules according to the new Quality of Service parameters, and instructs the PCF to provide the adjusted first update result directly or through NEF; wherein, the first update result is determined by the PCF based on the new Quality of Service parameters when adjusting the PCC rules;

[0293] Correspondingly, the processor 810 also specifically includes:

[0294] The first update result sent by the PCF is received directly or through the NEF, and the first update result includes whether the first request is accepted or the first request is rejected.

[0295] Optionally, the first request is specifically used for:

[0296] The PCF requests the SMF to determine whether the Session Management Function (SMF) needs to update the Session Management Policy. If it is determined that the SMF needs to update the Session Management Policy, the PCF sends a second request to the SMF. The parameters requested in the second request include at least one of the following: DNAI, Traffic Orientation Policy Identifier, and Traffic Route Information. The second request is used by the SMF to determine the selected User Plane Function (UPF) based on the new Session Management Policy and provide the corresponding DNAI, Traffic Orientation Policy Identifier, and Traffic Route Information.

[0297] Correspondingly, the processor 810 also specifically includes:

[0298] The PCF receives the second update result directly or through the NEF. The second update result is determined by the PCF based on whether to update the UPF path according to the new session management policy sent by the SMF.

[0299] The second update result includes whether the first request is accepted or rejected.

[0300] Optionally, the processor 810 is also used for:

[0301] Upon receiving the analysis information, the information of the application layer model is adjusted based on the analysis information. The information of the application layer model includes at least one of the following: model compression, model size, model transmission time period, and model encoding / decoding. The information of the application layer model is used to update the quality of service parameters.

[0302] Based on the information from the adjusted application layer model, the new service quality parameters are determined. These new service quality parameters include: 5G service quality identifier, reflective service quality control, maximum uplink bit rate of the transmission AI / ML model, maximum downlink bit rate of the transmission AI / ML model, minimum uplink bit rate of the transmission AI / ML model, minimum downlink bit rate of the transmission AI / ML model, and priority of the service quality flow.

[0303] Send a third request directly or through the NEF to the Policy Control Function (PCF);

[0304] The parameters requested in the third request include the new quality of service parameters, and the third request is used to request an update to the quality of service parameters.

[0305] Optionally, the third request is specifically used for:

[0306] The PCF is requested to adjust the 5G Quality of Service identifier, reflective quality of service control, maximum uplink bit rate of transmission AI / ML model, maximum downlink bit rate of transmission AI / ML model, minimum uplink bit rate of transmission AI / ML model, minimum downlink bit rate of transmission AI / ML model, and priority in the PCC rules according to the new quality of service parameters.

[0307] Correspondingly, the processor 810 is also used for:

[0308] The third update result is received directly or through NEF from the PCF. The third update result is determined by the PCF based on the results of adjusting the PCC rules. The third update result includes whether the third request is accepted or rejected.

[0309] Optionally, the processor 810 is also used for:

[0310] After adjusting the application layer model information, the model compression, model size, and model encoding / decoding information in the adjusted application layer model information are directly sent to the PCF. The adjusted application layer model information is used to support the PCF in adjusting the 5G service quality identifier, reflective service quality control, maximum uplink bit rate of the transmission AI / ML model, maximum downlink bit rate of the transmission AI / ML model, minimum uplink bit rate of the transmission AI / ML model, minimum downlink bit rate of the transmission AI / ML model, and service quality flow priority in the PCC rules.

[0311] Correspondingly, the processor 810 is also used for:

[0312] The system receives a fourth update result from the PCF, which is determined by the PCF adjusting the PCC rules based on the information from the adjusted application layer model. The fourth update result includes whether the third request is accepted or rejected.

[0313] Optionally, the processor 810 is also used for:

[0314] After adjusting the application layer model information, the model transmission time in the adjusted application layer model information is directly sent to the PCF. The model transmission time in the adjusted application layer model information is used to support the PCF in adjusting the gate state parameters in the PCC rules. The gate state parameters are used to support the SMF in updating the session management policy according to the transmission start time and transmission end time in the gate state.

[0315] Correspondingly, the processor 810 is also used for:

[0316] The PCF receives a fifth update result, which is determined by the PCF based on the new session management policy received from the SMF. The fifth update result includes whether the third request is accepted or rejected.

[0317] It should be noted that the model transmission state analysis device in the subscription network provided in this application is capable of achieving... Figures 3-6 All method steps implemented in the method embodiment shown are capable of achieving the same technical effect. Therefore, the parts that are the same as those in the method embodiment and their beneficial effects will not be described in detail here.

[0318] Figure 9 A schematic diagram of the structure of a model transmission state analysis device in a subscription network provided in another embodiment of this application is shown below. Figure 9 As shown, the model transmission state analysis device in the subscription network provided in this embodiment is applied to AF. Therefore, the model transmission state analysis device 900 in the subscription network provided in this embodiment includes:

[0319] The sending unit 901 is used to send a first message directly or through the Network Capability Opening Function (NEF) to the Network Data Analysis Function (NWDAF); wherein, the first message is used to request subscription to analysis information on the transmission status of artificial intelligence / machine learning (AI / ML) models in the network;

[0320] Analysis unit 902 is used to receive analysis information on the transmission status of AI / ML models sent by NWDAF directly or through NEF. The second message is used to collect data for analyzing the transmission status of AI / ML models in the network.

[0321] The analytical information is used to adjust network policy parameters and / or application layer model information.

[0322] Optionally, the parameters requested in the first message include at least one of the following: network data analysis identifier, identifier of a user equipment (UE) or group of UEs receiving the AI / ML model or any UE that meets the analysis conditions, identifier of the application using the AI / ML model, region of AI / ML model transmission, network slice indicating the protocol data unit (PDU) session transmitting the AI / ML model quality of service flow, data network indicating the PDU session transmitting the AI / ML model quality of service flow, time period of AI / ML model transmission, start timestamp of AI / ML model transmission, end timestamp of AI / ML model transmission, size of AI / ML transmission model, quality of service requirements for indicating the quality of service flow of AI / ML model transmission, and / or specific quality of service requirements for indicating the transmission of AI / ML model.

[0323] The second message includes at least one of the following: the current location of the UE using the AI / ML model, the identifier of the application using the AI / ML model, the service quality flow identifier for transmitting the AI / ML model, the uplink bit rate and downlink bit rate for transmitting the AI / ML model, the uplink packet delay and downlink packet delay of the AI / ML model, the number of abnormal releases of the service quality flow during the AI / ML model transmission period, the number of packet transmissions of the AI / ML model, the number of packet retransmissions of the AI / ML model, the data acquisition time, the duration of the AI / ML model transmission, the start timestamp of the AI / ML model transmission, the end timestamp of the AI / ML model transmission, the size of the AI / ML transmission model, the network slice of the PDU session used for transmitting the AI / ML model service quality flow, the data network of the PDU session used for transmitting the AI / ML model service quality flow, and the service process for the AF.

[0324] The analysis information includes at least one of the following: network slice of the PDU session used to transmit the AI / ML model quality of service flow, identifier of the application using the AI / ML model, region information using the AI / ML model, validity period of the analysis results, User Plane Function (UPF) providing AI / ML model transmission, data network name of the PDU session used to transmit the AI / ML model quality of service flow, size of the AI / ML transmission model, duration of AI / ML model transmission, start timestamp of AI / ML model transmission, end timestamp of AI / ML model transmission, quality of service requirements, quality of service flow identifier for transmitting the AI / ML model, uplink bit rate and downlink bit rate for transmitting the AI / ML model, uplink packet delay and downlink packet delay of the AI / ML model, number of abnormal releases of the quality of service flow during the AI / ML model transmission period, number of times the reporting threshold for abnormal release of the quality of service flow during the AI / ML model transmission period is reached, number of packet transmissions of the AI / ML model, and number of packet retransmissions of the AI / ML model.

[0325] If the AI / ML model performs federated learning, the parameters requested in the first message also include: federated learning group information, which includes at least one of the following: the identifier of the federated learning group used to indicate the analysis, the UE identifier or UE(s) identifier of the UE participating in the federated learning, and the application identifier of the UE participating in the federated learning.

[0326] Accordingly, the second message also includes at least one of the following: the identifier of the federated learning group for specifying the analysis, the UE identifier or UE(s) identifier of the UE participating in the federated learning, and the application identifier of the UE participating in the federated learning;

[0327] Accordingly, the analysis information also includes at least one of the following: the identifier of the federated learning group for specifying the analysis, the UE identifier or UE(s) identifier for participating in the federated learning, and the identifier of each application that provides the AI / ML model or participates in the federated learning.

[0328] Optionally, the transmitting unit is also used for:

[0329] Upon receiving the analysis information, a first request is sent directly or through the NEF to the Policy Control Function (PCF) based on the analysis information.

[0330] The first request is used to request an update to the network policy parameters for AI / ML model transmission; the network policy parameters are used to optimize the AI / ML model transmission status.

[0331] Optionally, the sending unit is specifically configured to: determine new service quality parameters for transmitting the AI / ML model based on at least one of the following in the analysis information: uplink bit rate and downlink bit rate of transmitting the AI / ML model, uplink packet delay and downlink packet delay of the AI / ML model, number of abnormal releases of the service quality flow during the transmission period of the AI / ML model, number of packet transmissions of the AI / ML model, number of packet retransmissions of the AI / ML model, and number of times the reporting threshold for abnormal release of the service quality flow during the transmission period of the AI / ML model is reached; the new service quality parameters include at least one of the following: 5G service quality identifier, reflective service quality control, maximum uplink bit rate of transmitting the AI / ML model, maximum downlink bit rate of transmitting the AI / ML model, minimum uplink bit rate of transmitting the AI / ML model, minimum downlink bit rate of transmitting the AI / ML model, and priority of the service quality flow;

[0332] Based on the identifier of the application transmitting the AI / ML model, the region information of the AI / ML model, the IP address information of the application service using the AI / ML model, the network slice of the PDU session used to transmit the AI / ML model service quality flow, and the data network name of the PDU session used to transmit the AI / ML model service quality flow, the region information and address information of the UE(s) transmitting the AI / ML model and each AF are determined; or, if the AI / ML model performs federated learning, based on the identifier of the federated learning group indicating the analysis, the UE identifier or UE(s) identifier participating in the federated learning, and the identifiers of each application indicating the provision of the AI / ML model or participation in the federated learning, the region information and address information of the UE(s) transmitting the AI / ML model and each AF are determined.

[0333] Based on the area information and address information of the UE(s) transmitting the AI / ML model and each AF, the data network access identifier (DNAI) and the area information and address information of the UE(s) and each AF corresponding to the DNAI are determined. The area information and address information of the UE(s) and each AF corresponding to the DNAI are used to provide a path for optimizing the transmission status of the AI / ML model.

[0334] The new quality of service parameters, the DNAI, and the area information and address information of the UE(s) and each AF corresponding to the DNAI are sent to the PCF directly or through the NEF as parameters in the first request.

[0335] Optionally, the transmitting unit is also specifically used for:

[0336] Based on the UE(s) transmitting AI / ML models and the area and address information of each AF, determine whether the current routing path is poor;

[0337] If the current routing path is not good, the destination address of both parties in the transmission AI / ML model is determined based on the address information of the UE(s) and each AF in the transmission AI / ML model, as well as the area information of the UE(s).

[0338] Determine the nearest path based on the destination address;

[0339] Based on the nearest path, determine the DNAI, as well as the region information and address information of the UE(s) and each AF corresponding to the DNAI.

[0340] Optionally, the first request is specifically used for:

[0341] The system requests the PCF to adjust the 5G Quality of Service identifier, reflective Quality of Service control, maximum uplink bit rate of the transmission AI / ML model, maximum downlink bit rate of the transmission AI / ML model, minimum uplink bit rate of the transmission AI / ML model, minimum downlink bit rate of the transmission AI / ML model, and priority of the Quality of Service flow in the PCC rules according to the new Quality of Service parameters, and instructs the PCF to provide the adjusted first update result directly or through NEF; wherein, the first update result is determined by the PCF based on the new Quality of Service parameters when adjusting the PCC rules;

[0342] Correspondingly, the transmitting unit is also used for:

[0343] The first update result sent by the PCF is received directly or through the NEF, and the first update result includes whether the first request is accepted or the first request is rejected.

[0344] Optionally, the first request is specifically used for:

[0345] The PCF requests the SMF to determine whether the Session Management Function (SMF) needs to update the Session Management Policy. If it is determined that the SMF needs to update the Session Management Policy, the PCF sends a second request to the SMF. The parameters requested in the second request include at least one of the following: DNAI, Traffic Orientation Policy Identifier, and Traffic Route Information. The second request is used by the SMF to determine the selected User Plane Function (UPF) based on the new Session Management Policy and provide the corresponding DNAI, Traffic Orientation Policy Identifier, and Traffic Route Information.

[0346] Correspondingly, the receiving unit is also used for:

[0347] The PCF receives the second update result directly or through the NEF. The second update result is determined by the PCF based on whether to update the UPF path according to the new session management policy sent by the SMF.

[0348] The second update result includes whether the first request is accepted or rejected.

[0349] Optionally, the device further includes: a determining unit; the determining unit is configured to:

[0350] Upon receiving the analysis information, the information of the application layer model is adjusted based on the analysis information. The information of the application layer model includes at least one of the following: model compression, model size, model transmission time period, and model encoding / decoding. The information of the application layer model is used to update the quality of service parameters.

[0351] Based on the information from the adjusted application layer model, the new service quality parameters are determined. These new service quality parameters include: 5G service quality identifier, reflective service quality control, maximum uplink bit rate of the transmission AI / ML model, maximum downlink bit rate of the transmission AI / ML model, minimum uplink bit rate of the transmission AI / ML model, minimum downlink bit rate of the transmission AI / ML model, and priority of the service quality flow.

[0352] Send a third request directly or through the NEF to the Policy Control Function (PCF);

[0353] The parameters requested in the third request include the new quality of service parameters, and the third request is used to request an update to the quality of service parameters.

[0354] Optionally, the third request is specifically used for:

[0355] The PCF is requested to adjust the 5G Quality of Service identifier, reflective quality of service control, maximum uplink bit rate of transmission AI / ML model, maximum downlink bit rate of transmission AI / ML model, minimum uplink bit rate of transmission AI / ML model, minimum downlink bit rate of transmission AI / ML model, and priority in the PCC rules according to the new quality of service parameters.

[0356] Correspondingly, the receiving unit is also used for:

[0357] The third update result is received directly or through NEF from the PCF. The third update result is determined by the PCF based on the results of adjusting the PCC rules. The third update result includes whether the third request is accepted or rejected.

[0358] Optionally, the transmitting unit is also used for:

[0359] After adjusting the application layer model information, the model compression, model size, and model encoding / decoding information in the adjusted application layer model information are directly sent to the PCF. The adjusted application layer model information is used to support the PCF in adjusting the 5G service quality identifier, reflective service quality control, maximum uplink bit rate of the transmission AI / ML model, maximum downlink bit rate of the transmission AI / ML model, minimum uplink bit rate of the transmission AI / ML model, minimum downlink bit rate of the transmission AI / ML model, and service quality flow priority in the PCC rules.

[0360] Correspondingly, the receiving unit is also used for:

[0361] The system receives a fourth update result from the PCF, which is determined by the PCF adjusting the PCC rules based on the information from the adjusted application layer model. The fourth update result includes whether the third request is accepted or rejected.

[0362] Optionally, the transmitting unit is also used for:

[0363] After adjusting the application layer model information, the model transmission time in the adjusted application layer model information is directly sent to the PCF. The model transmission time in the adjusted application layer model information is used to support the PCF in adjusting the gate state parameters in the PCC rules. The gate state parameters are used to support the SMF in updating the session management policy according to the transmission start time and transmission end time in the gate state.

[0364] Correspondingly, the receiving unit is also used for:

[0365] The PCF receives a fifth update result, which is determined by the PCF based on the new session management policy received from the SMF. The fifth update result includes whether the third request is accepted or rejected.

[0366] It should be noted that the model transmission state analysis device in the subscription network provided in this application is capable of achieving... Figures 3-6 All method steps implemented in the method embodiment can achieve the same technical effect. Therefore, the parts that are the same as those in the method embodiment and their beneficial effects will not be described in detail here.

[0367] Figure 10 A schematic diagram of the structure of a model transmission state analysis device in a subscription network provided in another embodiment of this application is shown below. Figure 10As shown, the model transmission status analysis device in the subscription network provided in this embodiment is applied to NWDAF. The model transmission status analysis device in the subscription network provided in this embodiment includes: a transceiver 1000, used to receive and send data under the control of a processor 1010.

[0368] Among them, Figure 10 In this context, the bus architecture can include any number of interconnected buses and bridges, specifically linking various circuits together, represented by one or more processors (processor 1010) and memory (memory 1020). The bus architecture can also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. The bus interface provides an interface. The transceiver 1000 can be multiple elements, including transmitters and receivers, providing units for communicating with various other devices over transmission media, including wireless channels, wired channels, optical fibers, etc. The processor 1010 is responsible for managing the bus architecture and general processing, and the memory 1020 can store data used by the processor 1010 during operation.

[0369] The processor 1010 can be a central processing unit (CPU), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a complex programmable logic device (CPLD). The processor can also adopt a multi-core architecture.

[0370] In this embodiment, the memory 1020 is used to store computer programs; the transceiver 1000 is used to send and receive data under the control of the processor; and the processor 1010 is used to read the computer programs from the memory and perform the following operations:

[0371] The application function (AF) receives a first message directly or through the Network Capability Opening Function (NEF); wherein the first message is used to request subscription to analysis information on the transmission status of artificial intelligence / machine learning (AI / ML) models in the network;

[0372] Based on the parameters requested in the first message, a second message is sent to other network functions 5GC NF(s) of the 5G core network. The second message is used to collect data for analyzing the transmission status of AI / ML models in the network.

[0373] It receives AI / ML model transmission status data sent by other network functions of the 5G core network, 5GC NF(s), and analyzes the AI / ML model transmission status data to obtain AI / ML model transmission status analysis information.

[0374] The analytical information is used to adjust network policy parameters and / or application layer model information through AF.

[0375] Optionally, the parameters requested in the first message include at least one of the following: network data analysis identifier, identifier of a user equipment (UE) or group of UEs receiving the AI / ML model or any UE that meets the analysis conditions, identifier of the application using the AI / ML model, region of AI / ML model transmission, network slice indicating the protocol data unit (PDU) session transmitting the AI / ML model quality of service flow, data network indicating the PDU session transmitting the AI / ML model quality of service flow, time period of AI / ML model transmission, start timestamp of AI / ML model transmission, end timestamp of AI / ML model transmission, size of AI / ML transmission model, quality of service requirements for indicating the quality of service flow of AI / ML model transmission, and / or specific quality of service requirements for indicating the transmission of AI / ML model.

[0376] The second message includes at least one of the following: the current location of the UE using the AI / ML model, the identifier of the application using the AI / ML model, the service quality flow identifier for transmitting the AI / ML model, the uplink bit rate and downlink bit rate for transmitting the AI / ML model, the uplink packet delay and downlink packet delay of the AI / ML model, the number of abnormal releases of the service quality flow during the AI / ML model transmission period, the number of packet transmissions of the AI / ML model, the number of packet retransmissions of the AI / ML model, the data acquisition time, the duration of the AI / ML model transmission, the start timestamp of the AI / ML model transmission, the end timestamp of the AI / ML model transmission, the size of the AI / ML transmission model, the network slice of the PDU session used for transmitting the AI / ML model service quality flow, the data network of the PDU session used for transmitting the AI / ML model service quality flow, and the service process for the AF.

[0377] The analysis information includes at least one of the following: network slice of the PDU session used to transmit the AI / ML model quality of service flow, identifier of the application using the AI / ML model, region information using the AI / ML model, validity period of the analysis results, User Plane Function (UPF) providing AI / ML model transmission, data network name of the PDU session used to transmit the AI / ML model quality of service flow, size of the AI / ML transmission model, duration of AI / ML model transmission, start timestamp of AI / ML model transmission, end timestamp of AI / ML model transmission, quality of service requirements, quality of service flow identifier for transmitting the AI / ML model, uplink bit rate and downlink bit rate for transmitting the AI / ML model, uplink packet delay and downlink packet delay of the AI / ML model, number of abnormal releases of the quality of service flow during the AI / ML model transmission period, number of times the reporting threshold for abnormal release of the quality of service flow during the AI / ML model transmission period is reached, number of packet transmissions of the AI / ML model, and number of packet retransmissions of the AI / ML model.

[0378] If the AI / ML model performs federated learning, the parameters requested in the first message also include: federated learning group information, which includes at least one of the following: the identifier of the federated learning group used to indicate the analysis, the UE identifier or UE(s) identifier of the UE participating in the federated learning, and the application identifier of the UE participating in the federated learning.

[0379] Accordingly, the second message also includes at least one of the following: the identifier of the federated learning group for specifying the analysis, the UE identifier or UE(s) identifier of the UE participating in the federated learning, and the application identifier of the UE participating in the federated learning;

[0380] Accordingly, the analysis information also includes at least one of the following: the identifier of the federated learning group for specifying the analysis, the UE identifier or UE(s) identifier for participating in the federated learning, and the identifier of each application that provides the AI / ML model or participates in the federated learning.

[0381] It should be noted that the model transmission state analysis device in the subscription network provided in this application is capable of achieving... Figure 4 , Figure 7 All method steps implemented in the method embodiment shown are capable of achieving the same technical effect. Therefore, the parts that are the same as those in the method embodiment and their beneficial effects will not be described in detail here.

[0382] Figure 11 A schematic diagram of the structure of a model transmission state analysis device in a subscription network provided in another embodiment of this application is shown below. Figure 11As shown, the subscription network model transmission state analysis device provided in this embodiment is applied to NWDAF. Therefore, the subscription network model transmission state analysis device 1100 provided in this embodiment includes:

[0383] The receiving unit 1101 is used to receive a first message sent by the application function AF directly or through the Network Capability Opening Function (NEF); wherein, the first message is used to request subscription to analysis information on the transmission status of artificial intelligence / machine learning (AI / ML) models in the network;

[0384] The sending unit 1102 is used to send a second message to other network functions 5GC NF(s) of the 5G core network according to the parameters requested in the first message. The second message is used to collect data for analyzing the transmission status of AI / ML models in the network.

[0385] The analysis unit 1103 is also used to receive data on the transmission status of the AI / ML model sent by other network functions 5GC NF(s) of the 5G core network, and to analyze the data on the transmission status of the AI / ML model to obtain analysis information on the transmission status of the AI / ML model.

[0386] The analytical information is used to adjust network policy parameters and / or application layer model information through AF.

[0387] Optionally, the parameters requested in the first message include at least one of the following: network data analysis identifier, identifier of a user equipment (UE) or group of UEs receiving the AI / ML model or any UE that meets the analysis conditions, identifier of the application using the AI / ML model, region of AI / ML model transmission, network slice indicating the protocol data unit (PDU) session transmitting the AI / ML model quality of service flow, data network indicating the PDU session transmitting the AI / ML model quality of service flow, time period of AI / ML model transmission, start timestamp of AI / ML model transmission, end timestamp of AI / ML model transmission, size of AI / ML transmission model, quality of service requirements for indicating the quality of service flow of AI / ML model transmission, and / or specific quality of service requirements for indicating the transmission of AI / ML model.

[0388] The second message includes at least one of the following: the current location of the UE using the AI / ML model, the identifier of the application using the AI / ML model, the service quality flow identifier for transmitting the AI / ML model, the uplink bit rate and downlink bit rate for transmitting the AI / ML model, the uplink packet delay and downlink packet delay of the AI / ML model, the number of abnormal releases of the service quality flow during the AI / ML model transmission period, the number of packet transmissions of the AI / ML model, the number of packet retransmissions of the AI / ML model, the data acquisition time, the duration of the AI / ML model transmission, the start timestamp of the AI / ML model transmission, the end timestamp of the AI / ML model transmission, the size of the AI / ML transmission model, the network slice of the PDU session used for transmitting the AI / ML model service quality flow, the data network of the PDU session used for transmitting the AI / ML model service quality flow, and the service process for the AF.

[0389] The analysis information includes at least one of the following: network slice of the PDU session used to transmit the AI / ML model quality of service flow, identifier of the application using the AI / ML model, region information of the region using the AI / ML model, validity period of the analysis results, User Plane Function (UPF) providing AI / ML model transmission, data network name of the PDU session used to transmit the AI / ML model quality of service flow, size of the AI / ML transmission model, duration of AI / ML model transmission, start timestamp of AI / ML model transmission, end timestamp of AI / ML model transmission, quality of service flow identifier for transmitting the AI / ML model, uplink bit rate and downlink bit rate for transmitting the AI / ML model, uplink packet delay and downlink packet delay of the AI / ML model, number of abnormal releases of the quality of service flow during the AI / ML model transmission period, number of times the reporting threshold for abnormal release of the quality of service flow during the AI / ML model transmission period is reached, number of packet transmissions of the AI / ML model, and number of packet retransmissions of the AI / ML model.

[0390] If the AI / ML model performs federated learning, the parameters requested in the first message also include: federated learning group information, which includes at least one of the following: the identifier of the federated learning group used to indicate the analysis, the UE identifier or UE(s) identifier of the UE participating in the federated learning, and the application identifier of the UE participating in the federated learning.

[0391] Accordingly, the second message also includes at least one of the following: the identifier of the federated learning group for indicating analysis, the UE identifier or UE(s) identifier of the UE participating in the federated learning, and the application identifier of the UE participating in the federated learning;

[0392] Accordingly, the analysis information also includes at least one of the following: the identifier of the federated learning group for specifying the analysis, the UE identifier or UE(s) identifier for participating in the federated learning, and the identifier of each application that provides the AI / ML model or participates in the federated learning.

[0393] It should be noted that the model transmission state analysis device in the subscription network provided in this application is capable of achieving... Figure 4 , Figure 7 All method steps implemented in the method embodiment can achieve the same technical effect. Therefore, the parts that are the same as those in the method embodiment and their beneficial effects will not be described in detail here.

[0394] It should be noted that the division of units in this application embodiment is illustrative and only represents one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units described above can be implemented in hardware or as software functional units.

[0395] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a processor-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0396] This application also provides a processor-readable storage medium. The processor-readable storage medium stores a computer program that causes the processor to execute any of the above-described method embodiments.

[0397] The processor-readable storage medium can be any available medium or data storage device that the processor can access, including but not limited to magnetic storage (e.g., floppy disks, hard disks, magnetic tapes, magneto-optical disks (MOs)), optical storage (e.g., CDs, DVDs, BDs, HVDs), and semiconductor storage (e.g., ROMs, EPROMs, EEPROMs, non-volatile memory (NAND flash), solid-state drives (SSDs)).

[0398] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0399] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-executable instructions. These computer-executable instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0400] These processor-executable instructions may also be stored in a processor-readable memory that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the processor-readable memory produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0401] These processors can execute instructions that can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0402] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for model transmission state analysis in a subscription network, characterized by, The method is applied to an application function (AF), and the method comprises: sending a first message to a network data analytics function (NWDAF) directly or through a network exposure function (NEF); wherein the first message is used to request subscription of analysis information of an artificial intelligence / machine learning (AI / ML) model transmission state in a network; receiving, directly or through the NEF, analysis information of the AI / ML model transmission state sent by the NWDAF, wherein the analysis information is determined by the NWDAF according to data of the AI / ML model transmission state received from one or more other network functions (5GC NFs) of a 5G core network (5GCN); wherein the analysis information is used to adjust network policy parameters and application layer model information; or the analysis information is used to adjust application layer model information; the application layer model information comprises at least one of the following: model compression, model size, model transmission time period, and model codec; and the application layer model information is used to update quality of service (QoS) parameters; after receiving the analysis information, the method further comprises: adjusting application layer model information according to the analysis information; determining new QoS parameters according to the adjusted application layer model information, wherein the new QoS parameters comprise: a 5G QoS identifier, a reflective QoS control, an uplink direction maximum bit rate for transmitting an AI / ML model, a downlink direction maximum bit rate for transmitting the AI / ML model, an uplink direction minimum bit rate for transmitting the AI / ML model, a downlink direction minimum bit rate for transmitting the AI / ML model, and a priority of a QoS flow; sending a third request to a policy control function (PCF) directly or through the NEF; wherein the parameters requested in the third request comprise the new QoS parameters, and the third request is used to request updating of the QoS parameters.

2. The method of claim 1, wherein, The data of the AI / ML model transmission state is obtained by the NWDAF according to parameters requested in the received first message by sending a second message to the one or more 5GC NFs, wherein the second message is used to collect data for analyzing the AI / ML model transmission state in the network.

3. The method of claim 2, wherein, The parameters requested in the first message comprise at least one of the following: a network data analytics identifier, an identifier of a user equipment (UE) or a group of UEs receiving an AI / ML model or any UE satisfying an analysis condition, an identifier of an application using the AI / ML model, an AI / ML model transmission area, a network slice indicating a protocol data unit (PDU) session of a QoS flow for transmitting the AI / ML model, a data network indicating a PDU session of a QoS flow for transmitting the AI / ML model, an AI / ML model transmission time period, a start timestamp of AI / ML model transmission, an end timestamp of AI / ML model transmission, a size of the AI / ML model, a QoS requirement for indicating a QoS flow for transmitting the AI / ML model, and / or a packet transmission delay and a packet loss rate for indicating transmission of the AI / ML model. The second message includes at least one of the following: a current location of the UE using the AI / ML model, an identification of an application using the AI / ML model, a quality of service flow identifier for transmitting the AI / ML model, an uplink direction bit rate and a downlink direction bit rate for transmitting the AI / ML model, an uplink direction packet delay and a downlink direction packet delay of the AI / ML model, a number of abnormal releases of the quality of service flow within a time period of AI / ML model transmission, a number of packet transmissions of the AI / ML model, a number of packet retransmissions of the AI / ML model, a data collection time, a duration of AI / ML model transmission, a start timestamp of AI / ML model transmission, an end timestamp of AI / ML model transmission, a size of the AI / ML transmission model, a network slice of a PDU session for transmitting the AI / ML model quality of service flow, a data network of the PDU session for transmitting the AI / ML model quality of service flow, a service flow for the AF; The analysis information includes at least one of the following: a network slice of a PDU session for transmitting the AI / ML model quality of service flow, an identification of an application using the AI / ML model, regional information using the AI / ML model, a valid time of the analysis result, a user plane function (UPF) providing AI / ML model transmission, a data network name of a PDU session for transmitting the AI / ML model quality of service flow, a size of the AI / ML transmission model, a duration of AI / ML model transmission, a start timestamp of AI / ML model transmission, an end timestamp of AI / ML model transmission, a quality of service flow identifier for transmitting the AI / ML model, an uplink direction bit rate and a downlink direction bit rate for transmitting the AI / ML model, an uplink direction packet delay and a downlink direction packet delay of the AI / ML model, a number of abnormal releases of the quality of service flow within a time period of AI / ML model transmission, a number of times reaching a reporting threshold of abnormal releases of the quality of service flow within a time period of AI / ML model transmission, a number of packet transmissions of the AI / ML model, a number of packet retransmissions of the AI / ML model; If the AI / ML model performs federated learning, the parameters requested in the first message further include federated learning group information, and the federated learning group information includes at least one of the following: an identification of a federated learning group for analysis, an identification of a UE participating in federated learning, or an identification of a UE group participating in federated learning, or an identification of an application participating in federated learning; Correspondingly, the second message further includes at least one of the following: an identification of a federated learning group for analysis, an identification of a UE participating in federated learning, or an identification of a UE group participating in federated learning, or an identification of an application participating in federated learning; Correspondingly, the analysis information further includes at least one of the following: an identification of a federated learning group for analysis, an identification of a UE participating in federated learning, or an identification of a UE group participating in federated learning, or an identification of each application providing the AI / ML model or participating in federated learning.

4. The method according to any one of claims 1 to 3, characterized in that, After receiving the analysis information, the method further includes: According to the analysis information, a first request is sent to a policy control function (PCF) directly or through the NEF; The first request is used to request updating of network policy parameters for AI / ML model transmission; and the network policy parameters are used to optimize AI / ML model transmission status.

5. The method of claim 4, wherein, The first request is sent to the PCF directly or through the NEF according to the analysis information, and includes: According to at least one of the uplink direction bit rate of AI / ML model transmission, the downlink direction bit rate of AI / ML model transmission, the uplink direction packet delay of AI / ML model, the downlink direction packet delay of AI / ML model, the number of abnormal releases of a quality of service (QoS) flow within a time period of AI / ML model transmission, the number of packet transmissions of AI / ML model, the number of packet retransmissions of AI / ML model, and the number of times reaching a report threshold of abnormal release of a QoS flow within a time period of AI / ML model transmission in the analysis information, a new QoS parameter for AI / ML model transmission is determined; the new QoS parameter includes at least one of a 5G QoS identifier, a reflective QoS control, an uplink direction maximum bit rate of AI / ML model transmission, a downlink direction maximum bit rate of AI / ML model transmission, an uplink direction minimum bit rate of AI / ML model transmission, a downlink direction minimum bit rate of AI / ML model transmission, and a priority of a QoS flow; According to the application identifier of AI / ML model transmission, the regional information of AI / ML model use, the IP address information of application service using AI / ML model, the network slice of a PDU session for AI / ML model transmission, and the data network name of a PDU session for AI / ML model transmission in the analysis information, regional information and address information of one or more UEs and each AF for AI / ML model transmission are determined; or, if AI / ML model performs federated learning, according to the identifier of a federated learning group indicated by the analysis, the UE identifier or UE group identifier participating in federated learning, and the application identifier of each application providing AI / ML model or participating in federated learning in the analysis information, regional information and address information of one or more UEs and each AF for AI / ML model transmission are determined. According to the regional information and address information of one or more UEs and each AF for AI / ML model transmission, data network access identifier (DNAI) and regional information and address information of one or more UEs and each AF corresponding to the DNAI are determined, and the regional information and address information of one or more UEs and each AF corresponding to the DNAI are used to provide a path for optimizing AI / ML model transmission status. The new QoS parameter, the DNAI, and the regional information and address information of one or more UEs and each AF corresponding to the DNAI are sent to the PCF as parameters in the first request directly or through the NEF.

6. The method of claim 5, wherein, The determining of the data network access identifier (DNAI) and the region information and address information of the one or more UEs and the respective AFs corresponding to the DNAI according to the region information and address information of the one or more UEs and the respective AFs of the transmission AI / ML model comprises: According to the region information and address information of the one or more UEs and the respective AFs of the transmission AI / ML model, it is judged whether the current routing path is poor; If the current routing path is poor, the destination address of the transmission AI / ML model is determined according to the address information of the one or more UEs and the respective AFs and the region information of the one or more UEs of the transmission AI / ML model; According to the destination address, the nearest path is determined; According to the nearest path, the DNAI and the region information and address information of the one or more UEs and the respective AFs corresponding to the DNAI are determined.

7. The method of claim 5, wherein, The first request is specifically used for: Requesting the PCF to adjust the 5G quality of service identifier, reflective quality of service control, uplink direction maximum bit rate of the transmission AI / ML model, downlink direction maximum bit rate of the transmission AI / ML model, uplink direction minimum bit rate of the transmission AI / ML model, downlink direction minimum bit rate of the transmission AI / ML model, and priority of the quality of service flow in the PCC rule according to the new quality of service parameter, and instructing the PCF to feed back the first update result after adjustment directly or through the NEF; wherein the first update result is determined by the PCF according to the result of adjusting the PCC rule according to the new quality of service parameter; Correspondingly, the method further comprises: Directly or through the NEF, receiving the first update result sent by the PCF, wherein the first update result comprises that the first request is accepted or the first request is rejected.

8. The method of claim 5, wherein, The first request is specifically used for: Requesting the PCF to determine whether the session management function network element (SMF) needs to update the session management policy, and if it is determined that the SMF needs to update the session management policy, determining that the PCF sends a second request to the SMF, wherein the requested parameters in the second request comprise at least one of the following: DNAI, traffic steering policy identifier, and traffic routing information; The second request is used for the SMF to determine the selected user plane function (UPF) according to the new session management policy and provide the corresponding DNAI, traffic steering policy identifier, and traffic routing information; Correspondingly, the method further comprises: Directly or through the NEF, receiving the second update result sent by the PCF, wherein the second update result is determined by the PCF according to whether the UPF path is updated according to the new session management policy sent by the SMF; Wherein the second update result comprises that the first request is accepted or rejected.

9. The method of claim 1, wherein, The third request is specifically used for: requesting the PCF to adjust the 5G quality of service identifier, the reflective quality of service control, the uplink direction maximum bit rate of the AI / ML model transmission, the downlink direction maximum bit rate of the AI / ML model transmission, the uplink direction minimum bit rate of the AI / ML model transmission, the downlink direction minimum bit rate of the AI / ML model transmission, and the priority in the PCC rule according to the new quality of service parameter; Correspondingly, the method further comprises: receiving, directly or through the NEF, a third update result sent by the PCF, the third update result being determined by the PCF based on a result of adjusting the PCC rule, and the third update result comprising that the third request is accepted or the third request is rejected.

10. The method of claim 2, wherein, After the information of the adjusted application layer model is sent, the method further comprises: sending, directly to the PCF, the model compression, the model size, and the model codec in the information of the adjusted application layer model, the information of the adjusted application layer model being used to support the PCF to adjust the 5G quality of service identifier, the reflective quality of service control, the uplink direction maximum bit rate of the AI / ML model transmission, the downlink direction maximum bit rate of the AI / ML model transmission, the uplink direction minimum bit rate of the AI / ML model transmission, the downlink direction minimum bit rate of the AI / ML model transmission, and the priority of the quality of service flow in the PCC rule; Correspondingly, the method further comprises: receiving, directly or through the NEF, a third update result sent by the PCF, the third update result being determined by the PCF based on a result of adjusting the PCC rule, and the third update result comprising that the third request is accepted or the third request is rejected.

11. The method of claim 1, wherein, After the information of the adjusted application layer model is sent, the method further comprises: sending, directly to the PCF, the model transmission time in the information of the adjusted application layer model, the model transmission time in the information of the adjusted application layer model being used to support the PCF to adjust the gate status parameter in the PCC rule; and the gate status parameter being used to support the SMF to update the session management policy according to the transmission start time and the transmission end time in the gate status; Correspondingly, the method further comprises: receiving, directly or through the NEF, a third update result sent by the PCF, the third update result being determined by the PCF based on a result of adjusting the PCC rule, and the third update result comprising that the third request is accepted or the third request is rejected.

12. A method for model transmission state analysis in a subscription network, the method comprising: The method is applied to a network data analysis function (NWDAF), and the method comprises: receiving, directly or through a network exposure function (NEF), a first message sent by an application function (AF); wherein the first message is used to request to subscribe to analysis information of an artificial intelligence / machine learning (AI / ML) model transmission state in a network; sending, according to parameters requested in the first message, a second message to one or more other network functions (5GC NFs) of a 5G core network (5GC), the second message being used to collect data for analyzing the AI / ML model transmission state in the network; receive data of an AI / ML model transmission status transmitted by one or more other network functions 5GC NFs of a 5G core network, and analyze the data of the AI / ML model transmission status to obtain analysis information of the AI / ML model transmission status; The analysis information is used to adjust network policy parameters and application layer model information by the AF, or the analysis information is used to adjust application layer model information; the application layer model information includes at least one of model compression, model size, model transmission time period, and model encoding and decoding; the application layer model information is used to update quality of service parameters; the adjusted application layer model information is used to determine new quality of service parameters, and the new quality of service parameters include a 5G quality of service identifier, a reflective quality of service control, an uplink direction maximum bit rate for transmitting an AI / ML model, a downlink direction maximum bit rate for transmitting the AI / ML model, an uplink direction minimum bit rate for transmitting the AI / ML model, a downlink direction minimum bit rate for transmitting the AI / ML model, and a priority of a quality of service flow; and the new quality of service parameters are used for a policy control function PCF to update quality of service parameters.

13. The method of claim 12, wherein, The parameters requested in the first message include at least one of a network data analysis identifier, an identifier of a user equipment UE or a group of UEs receiving an AI / ML model or any UE satisfying an analysis condition, an identifier of an application using the AI / ML model, an area of AI / ML model transmission, a network slice indicating a protocol data unit PDU session of a quality of service flow for transmitting the AI / ML model, a data network indicating the PDU session of the quality of service flow for transmitting the AI / ML model, a time period of AI / ML model transmission, a start timestamp of AI / ML model transmission, an end timestamp of AI / ML model transmission, a size of the AI / ML transmission model, quality of service requirements for indicating the quality of service flow for transmitting the AI / ML model and / or packet transmission delay and packet error rate for indicating transmission of the AI / ML model; The second message includes at least one of a current location of a UE using the AI / ML model, an identifier of an application using the AI / ML model, a quality of service flow identifier for transmitting the AI / ML model, uplink direction and downlink direction bit rates for transmitting the AI / ML model, uplink direction and downlink direction packet delays of the AI / ML model, a number of abnormal releases of the quality of service flow within a time period of AI / ML model transmission, a number of packet transmissions of the AI / ML model, a number of packet retransmissions of the AI / ML model, a data collection time, a duration of AI / ML model transmission, a start timestamp of AI / ML model transmission, an end timestamp of AI / ML model transmission, a size of the AI / ML transmission model, a network slice of a PDU session for transmitting the AI / ML model quality of service flow, a data network of the PDU session for transmitting the AI / ML model quality of service flow, and a service flow for the AF; and The analysis information includes at least one of the following: a network slice of a PDU session for transmitting an AI / ML model service quality flow, an identifier of an application using an AI / ML model, area information using an AI / ML model, a validity time of an analysis result, a user plane function (UPF) providing AI / ML model transmission, a data network name of a PDU session for transmitting an AI / ML model service quality flow, a size of an AI / ML transmission model, a duration of AI / ML model transmission, a start timestamp of AI / ML model transmission, an end timestamp of AI / ML model transmission, a quality of service (QoS) flow identifier for transmitting an AI / ML model, an uplink direction bitrate and a downlink direction bitrate for transmitting an AI / ML model, an uplink direction packet delay of an AI / ML model and a downlink direction packet delay of an AI / ML model, a number of abnormal releases of a QoS flow within a time period of AI / ML model transmission, a number of times reaching a reporting threshold of abnormal releases of a QoS flow within a time period of AI / ML model transmission, a number of packet transmissions of an AI / ML model, and a number of packet retransmissions of an AI / ML model. If the AI / ML model performs federated learning, the parameters requested in the first message further include federated learning group information, and the federated learning group information includes at least one of the following: an identifier of a federated learning group for indicating analysis, an identifier of a UE or a group identifier of UEs participating in federated learning, and an identifier of an application participating in federated learning. Correspondingly, the second message further includes at least one of the following: an identifier of a federated learning group for indicating analysis, an identifier of a UE or a group identifier of UEs participating in federated learning, and an identifier of an application participating in federated learning. Correspondingly, the analysis information further includes at least one of the following: an identifier of a federated learning group for indicating analysis, an identifier of a UE or a group identifier of UEs participating in federated learning, and an identifier of each application providing an AI / ML model or participating in federated learning.

14. A model transmission state analysis apparatus in a subscription network, characterized by, The device includes a memory, a transceiver, and a processor. The memory is configured to store a computer program; the transceiver is configured to transceive data under control of the processor; and the processor is configured to read the computer program in the memory and perform the following operations: sending a first message to a network data analytics function (NWDAF) directly or through a network exposure function (NEF); wherein the first message is used to request subscription of analysis information of an artificial intelligence / machine learning (AI / ML) model transmission state in a network; receiving, directly or through the NEF, analysis information of the AI / ML model transmission state sent by the NWDAF, wherein the analysis information is determined by the NWDAF based on AI / ML model transmission state data received from one or more other network functions (5GC NFs) of a 5G core network. The analysis information is used to adjust network policy parameters and application layer model information, or the analysis information is used to adjust application layer model information; the application layer model information includes at least one of model compression, model size, model transmission time period, and model codec; and the application layer model information is used to update quality of service parameters. The application layer model information is adjusted according to the analysis information. New quality of service parameters are determined according to the adjusted application layer model information, and the new quality of service parameters include a 5G quality of service identifier, a reflective quality of service control, an uplink direction maximum bit rate for transmitting an AI / ML model, a downlink direction maximum bit rate for transmitting the AI / ML model, an uplink direction minimum bit rate for transmitting the AI / ML model, a downlink direction minimum bit rate for transmitting the AI / ML model, and a priority of a quality of service flow. The third request is directly or through the NEF sent to a policy control function PCF. The third request includes the new quality of service parameters.

15. A model transmission state analysis apparatus in a subscription network, characterized by, The device includes a memory, a transceiver, and a processor. The memory is configured to store a computer program; the transceiver is configured to transceive data under control of the processor; and the processor is configured to read the computer program in the memory and perform the following operations: A first message is directly or through a network capability exposure function NEF received from an application function AF; the first message is used to request to subscribe to analysis information of an artificial intelligence / machine learning AI / ML model transmission state in a network. A second message is sent to one or more other network functions 5GC NFs of a 5G core network according to parameters requested in the first message; the second message is used to collect data for analyzing the AI / ML model transmission state in the network. Data of the AI / ML model transmission state is received from the one or more other network functions 5GC NFs of the 5G core network, and the data of the AI / ML model transmission state is analyzed to obtain analysis information of the AI / ML model transmission state. The analysis information is used to adjust network policy parameters and application layer model information, or the analysis information is used to adjust application layer model information; the application layer model information includes at least one of model compression, model size, model transmission time period, and model codec; and the application layer model information is used to update quality of service parameters; the adjusted application layer model information is used to determine new quality of service parameters, and the new quality of service parameters include a 5G quality of service identifier, a reflective quality of service control, an uplink direction maximum bit rate for transmitting an AI / ML model, a downlink direction maximum bit rate for transmitting the AI / ML model, an uplink direction minimum bit rate for transmitting the AI / ML model, a downlink direction minimum bit rate for transmitting the AI / ML model, and a priority of a quality of service flow; and the new quality of service parameters are used for a policy control function PCF to update quality of service parameters.

16. A model transmission state analysis apparatus in a subscription network, characterized by, The device includes: a sending unit configured to send, directly or through a network exposure function (NEF), a first message to a network data analytics function (NWDAF); wherein the first message is used to request subscription of analysis information of an artificial intelligence / machine learning (AI / ML) model transmission state in a network; an analyzing unit configured to receive, directly or through the NEF, the analysis information of the AI / ML model transmission state sent by the NWDAF, wherein the analysis information is determined by the NWDAF based on data of the AI / ML model transmission state received from one or more other network functions (5GC NFs) of a 5G core network (5GCN); wherein the analysis information is used to adjust network policy parameters and application layer model information; or the analysis information is used to adjust the application layer model information; the application layer model information includes at least one of the following: model compression, model size, model transmission time period, and model codec; and the application layer model information is used to update quality of service (QoS) parameters; a determining unit configured to, after receiving the analysis information, adjust the application layer model information according to the analysis information, and determine new QoS parameters according to the adjusted application layer model information, wherein the new QoS parameters include a 5G QoS identifier, a reflective QoS control, an uplink direction maximum bit rate for AI / ML model transmission, a downlink direction maximum bit rate for AI / ML model transmission, an uplink direction minimum bit rate for AI / ML model transmission, a downlink direction minimum bit rate for AI / ML model transmission, and a priority of a QoS flow; a sending unit configured to send, directly or through the NEF, a third request to a policy control function (PCF); wherein the parameters requested in the third request include the new QoS parameters, and the third request is used to request updating of the QoS parameters.

17. A model transmission state analysis apparatus in a subscription network, characterized by, The apparatus comprises: a receiving unit configured to receive, directly or through a network exposure function (NEF), a first message sent by an application function (AF); wherein the first message is used to request subscription of analysis information of an artificial intelligence / machine learning (AI / ML) model transmission state in a network; a sending unit configured to send, according to the parameters requested in the first message, a second message to one or more other network functions (5GC NFs) of a 5G core network (5GCN), wherein the second message is used to collect data for analyzing the AI / ML model transmission state in the network; an analyzing unit configured to receive data of the AI / ML model transmission state sent by the one or more 5GC NFs, and analyze the data of the AI / ML model transmission state to obtain analysis information of the AI / ML model transmission state. The analysis information is used for adjusting network policy parameters and application layer model information by the AF; or the analysis information is used for adjusting application layer model information; the application layer model information includes at least one of the following: model compression, model size, model transmission time period, model coding and decoding; the application layer model information is used for updating quality of service parameters; the adjusted application layer model information is used for determining new quality of service parameters, and the new quality of service parameters include: a 5G quality of service identifier, a reflective quality of service control, an uplink direction maximum bit rate for transmitting an AI / ML model, a downlink direction maximum bit rate for transmitting the AI / ML model, an uplink direction minimum bit rate for transmitting the AI / ML model, a downlink direction minimum bit rate for transmitting the AI / ML model, and a priority of a quality of service flow; and the new quality of service parameters are used for a policy control function (PCF) to update quality of service parameters.

18. A processor-readable storage medium, comprising: The processor readable storage medium stores a computer program, and the computer program is used for enabling the processor to execute the method in any one of claims 1 to 13.

19. A computer program product comprising a computer program, characterized in that, The computer program, when executed by a processor, is used for implementing the method in any one of claims 1 to 13.

Images