WO2023093416A1 - 订阅网络中模型传输状态分析方法、装置及可读存储介质 - Google Patents

订阅网络中模型传输状态分析方法、装置及可读存储介质 Download PDF

Info

Publication number
WO2023093416A1
WO2023093416A1 PCT/CN2022/127034 CN2022127034W WO2023093416A1 WO 2023093416 A1 WO2023093416 A1 WO 2023093416A1 CN 2022127034 W CN2022127034 W CN 2022127034W WO 2023093416 A1 WO2023093416 A1 WO 2023093416A1
Authority
WO
WIPO (PCT)
Prior art keywords
model
transmission
network
information
analysis
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2022/127034
Other languages
English (en)
French (fr)
Inventor
刘莹莹
王文慧
段小嫣
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Datang Mobile Communications Equipment Co Ltd
Original Assignee
Datang Mobile Communications Equipment Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Datang Mobile Communications Equipment Co Ltd filed Critical Datang Mobile Communications Equipment Co Ltd
Priority to JP2024529997A priority Critical patent/JP7743629B2/ja
Priority to US18/701,933 priority patent/US20240414063A1/en
Priority to EP22897499.4A priority patent/EP4440065A4/en
Priority to KR1020247018950A priority patent/KR20240099444A/ko
Publication of WO2023093416A1 publication Critical patent/WO2023093416A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/08—Configuration management of networks or network elements
    • H04L41/0894—Policy-based network configuration management
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/16—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using machine learning or artificial intelligence
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/34—Signalling channels for network management communication
    • H04L41/342—Signalling channels for network management communication between virtual entities, e.g. orchestrators, SDN or NFV entities
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/50—Network service management, e.g. ensuring proper service fulfilment according to agreements
    • H04L41/5003—Managing SLA; Interaction between SLA and QoS
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/14—Session management
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/02—Arrangements for optimising operational condition
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/10—Scheduling measurement reports ; Arrangements for measurement reports
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W28/00—Network traffic management; Network resource management
    • H04W28/02—Traffic management, e.g. flow control or congestion control
    • H04W28/0268—Traffic management, e.g. flow control or congestion control using specific QoS parameters for wireless networks, e.g. QoS class identifier [QCI] or guaranteed bit rate [GBR]
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W88/00—Devices specially adapted for wireless communication networks, e.g. terminals, base stations or access point devices
    • H04W88/14—Backbone network devices

Definitions

  • the present disclosure relates to the field of communication technologies, and in particular to a method, device and readable storage medium for analyzing model transmission status in a subscription network.
  • AI/ML model (hereinafter referred to as AI/ML model), therefore, the current method is to transfer the reasoning of many AI/ML models from the mobile terminal to the cloud or other terminals, that is, it is necessary to transfer the AI/ML model to the cloud or other terminals.
  • the 5G system is used as a channel to transmit the AI/ML model.
  • the 5G system needs to support the transfer of AI/ML models. - ML session monitoring and state information exposed to third parties.
  • This disclosure provides a model transmission state analysis method, device and readable storage medium in a subscription network, which solves the problem that the AI/ML model transmission state cannot be effectively analyzed, and the network cannot be effectively adjusted based on the AI/ML model transmission state.
  • the present disclosure provides a method for analyzing model transmission status in a subscription network, the method is applied to an application function AF, and the method includes:
  • the analysis information is used to adjust network policy parameters and/or application layer model information.
  • the data of the transmission state of the AI/ML model is obtained by the NWDAF by sending a second message to the 5GC NF(s) according to the parameters requested in the received first message, and the first The second message is used to collect data for analyzing the transmission status of the AI/ML model in the network.
  • the parameters requested in the first message include at least one of the following: a network data analysis identifier, an identifier of a user equipment UE or a group of UEs receiving the AI/ML model, or any UE that meets the analysis conditions, using The identification of the application of the AI/ML model, the area where the AI/ML model is transmitted, the network slice indicating the PDU session of the protocol data unit PDU session that transmits the AI/ML model quality of service flow, and the data indicating the PDU session that transmits the AI/ML model quality of service flow Network, 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 used to indicate the transmission of AI/ML model QoS requirements for streams and/or specific QoS requirements for delivery of AI/ML models;
  • the second message includes at least one of the following items: the current location of the UE using the AI/ML model, the identifier of the application using the AI/ML model, the quality of service flow identifier of the transmission AI/ML model, the transmission AI/ML model
  • the bit rate of the uplink direction of the model and the bit rate of the downlink direction of the transmission AI/ML model, the packet delay of the uplink direction of the AI/ML model and the packet delay of the downlink direction of the AI/ML model, the quality of service flow in The number of abnormal releases during the time period of AI/ML model transmission, the number of packet transmissions of AI/ML models, the number of packet retransmissions of AI/ML models, data collection time, the duration of AI/ML model transmission, and the transmission of AI/ML models start timestamp of AI/ML model transfer, size of AI/ML transfer model, network slice of PDU session for transfer of AI/ML model QoS stream, network slice for transfer of AI/ML model QoS stream.
  • the analysis information includes at least one of the following: network slices of PDU sessions used to transmit AI/ML model quality of service flows, identification of applications using AI/ML models, area information using AI/ML models, and analysis results.
  • Valid time user plane function UPF that provides AI/ML model transmission
  • data network name of PDU session used to transmit AI/ML model quality of service flow size of AI/ML transmission model, duration of AI/ML model transmission
  • AI /ML model transmission start timestamp AI/ML model transmission end timestamp
  • quality of service flow identifier for transmission AI/ML model
  • uplink direction bit rate for transmission AI/ML model and time stamp for transmission AI/ML model Bit rate in the downlink direction packet delay in the uplink direction of the AI/ML model and packet delay in the downlink direction of the AI/ML model, the number of abnormally released QoS streams during the transmission period of the AI/ML model, and the The number of reporting thresholds for abnormal release of quality of service flows during the AI/ML model transmission period, the number of packet transmissions for the AI/ML model, and
  • the parameters requested in the first message also include: federated learning group information, and the federated learning group information includes at least one of the following: the federated learning group used to indicate the analysis ID, UE ID or UE(s) ID participating in federated learning, application ID participating in federated learning;
  • the second message further includes at least one of the following items: the identity of the federated learning group used to indicate the analysis, the UE identity or UE(s) identity participating in the federated learning, and the application identity participating in the federated learning;
  • the analysis information also includes at least one of the following items: 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 user indicating to provide the AI/ML model or participate in the federated learning Individual application IDs.
  • the method further includes:
  • the first request is used to request to update the network policy parameters used for AI/ML model transmission; the network policy parameters are used to optimize the transmission status of the AI/ML model.
  • the sending the first request to the policy control function PCF directly or through the NEF according to the analysis information includes:
  • the packet delay in the uplink direction of the AI/ML model and the downlink direction of the AI/ML model Packet delay in the link direction, the number of abnormally released quality of service flows during the AI/ML model transmission period, the number of packet transmissions in the AI/ML model, the number of packet retransmissions in the AI/ML model, At least one of the number of reporting thresholds for abnormal release during the time period of ML model transmission, to determine the new quality of service parameters of the transmission 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 bit rate in uplink direction for transmitting AI/ML models, maximum bit rate in downlink direction for transmitting AI/ML models, minimum bit rate in uplink direction for transmitting AI/ML models , the lowest bit rate in the downlink direction of the transmission AI/ML model, and the priority of the quality
  • the identification of the application transmitting the AI/ML model the area information using the AI/ML model, the IP address information of the application service using the AI/ML model, and the quality of service flow used to transmit the AI/ML model in the analysis information
  • the network slice of the PDU session, the data network name of the PDU session used to transmit the AI/ML model quality of service flow determine the area information and address information of the UE(s) and each AF that transmits the AI/ML model; or, if the AI/ML model
  • the ML model performs federated learning, according to the identifier of the federated learning group indicated in the analysis information, the UE identifier or UE(s) identifier participating in the federated learning, and the identifier of each application that provides the AI/ML model or participates in the federated learning, Determine the UE(s) transmitting the AI/ML model and the area information and address information of each AF;
  • the area information and address information of the UE(s) and each AF transmitting the AI/ML model 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 said The UE(s) corresponding to DNAI and the area information and address information of each AF are used to provide a path to optimize the transmission status of the AI/ML model;
  • the new QoS parameter, 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.
  • the area information and address information of the UE(s) and each AF transmitting the AI/ML model determine the data network access identifier DNAI and the area information of the UE(s) and each AF corresponding to the DNAI and address information, including:
  • the first request is specifically used for:
  • the method also includes:
  • the first update result sent by the PCF is received directly or through the NEF, where the first update result includes that the first request is accepted or the first request is rejected.
  • the first request is specifically used for:
  • the PCF Requesting the PCF to determine whether the session management function network element SMF needs to update the session management strategy, if it is determined that the SMF needs to update the session management strategy, then determine that the PCF sends a second request to the SMF, and the parameters requested in the second request include at least one of the following : DNAI, traffic steering policy identifier, 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 corresponding DNAI, traffic steering policy identifier, and traffic route information;
  • the method also includes:
  • the second update result includes whether the first request is accepted or rejected.
  • new quality of service parameters are determined according to the analysis information, and the new quality of service parameters are sent to the PCF, so that the PCF adjusts the PCC rules according to the new quality of service parameters or updates the SM policy through the SMF and provides DNAI and
  • the UE(s) corresponding to the DNAI and the area information and address information of each AF can request to adjust the network policy based on the NWDAF analysis result by receiving the notification sent by the PCF to indicate whether the first request is accepted or rejected.
  • Optimize AI/ML model transfer status are provided to the analysis information, and the new quality of service parameters are sent to the PCF, so that the PCF adjusts the PCC rules according to the new quality of service parameters or updates the SM policy through the SMF and provides DNAI and
  • the UE(s) corresponding to the DNAI and the area information and address information of each AF can request to adjust the network policy based on the NWDAF analysis result by receiving the notification sent by the PCF to indicate whether the first request is accepted or rejected.
  • the method further includes:
  • the information of the application layer model includes at least one of the following: model compression, model size, model transmission time period, model encoding and decoding; the information of the application layer model is used for updating quality of service parameters;
  • new quality of service parameters are determined, and the new quality of service parameters include: 5G service quality identifier, reflective service quality control, maximum bit in the uplink direction of the transmission AI/ML model rate, the maximum bit rate in the downlink direction for AI/ML models, the minimum bit rate for the uplink direction for AI/ML models, the minimum bit rate for the downlink direction for AI/ML models, and the priority of quality of service streams ;
  • the parameters requested in the third request include the new QoS parameter, and the third request is used to request to update the QoS parameter.
  • the third request is specifically used for:
  • the method also includes:
  • the third update result is determined by the PCF based on the result of adjusting the PCC rules, the third update result includes the third request being accepted or the third request being rejected .
  • the method further includes:
  • the model compression, model size and model codec in the information of the adjusted application layer model are directly sent to the PCF, and the information of the adjusted application layer model is used to support the PCF to adjust the 5G service quality identifier in the PCC rule, Reflective quality of service control, maximum bit rate in uplink direction for transmitting AI/ML models, maximum bit rate in downlink direction for transmitting AI/ML models, minimum bit rate in uplink direction for transmitting AI/ML models, transmitting AI The lowest bit rate in the downlink direction of the /ML model, the priority of the quality of service flow;
  • the method also includes:
  • the method further includes:
  • the model transmission time in the information of the adjusted application layer model is directly sent to the PCF, and the model transmission time in the information of the adjusted application layer model is used to support the PCF to adjust the gate status parameter in the PCC rule;
  • the gate The state parameter is used to support SMF to update the session management policy according to the transmission start time and transmission end time in the gate state;
  • the method also includes:
  • the information of the application layer model is adjusted, and new service quality parameters are determined according to the information of the adjusted application layer model, and the new service quality parameters are sent to the PCF, which will make the PCF follow the new The quality of service parameters adjust the PCC rules accordingly; or, directly send the adjusted information of the application layer model to the PCF, so that the PCF can adjust the above-mentioned quality of service parameters in the PCC rules according to the model compression, model size, and model codec, or, make the PCF Adjust the gate state parameters in the PCC rules according to the model transmission time, so that the SMF updates the session management policy according to the transmission start time and transmission end time in the gate state, and receives the message sent by the PCF to indicate whether the third request is accepted or rejected Notification can adjust the information of the application layer model based on the NWDAF analysis results, and then update the QoS requirements, thereby optimizing the transmission status of the AI/ML model.
  • the present disclosure provides a method for analyzing model transmission status in a subscription network, the method is applied to the network data analysis function NWDAF, and the method includes:
  • the first message is used to request to subscribe to the analysis information of the transmission status of the artificial intelligence/machine learning AI/ML model in the network;
  • the second message is used to collect data for analyzing the transmission status of the AI/ML model in the network ;
  • the analysis information is used to adjust network policy parameters and/or application layer model information through AF.
  • the parameters requested in the first message include at least one of the following: a network data analysis identifier, an identifier of a user equipment UE or a group of UEs receiving the AI/ML model, or any UE that meets the analysis conditions, using The identification of the application of the AI/ML model, the area where the AI/ML model is transmitted, the network slice indicating the PDU session of the protocol data unit PDU session that transmits the AI/ML model quality of service flow, and the data indicating the PDU session that transmits the AI/ML model quality of service flow Network, 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 used to indicate the transmission of AI/ML model QoS requirements for streams and/or specific QoS requirements for delivery of AI/ML models;
  • the second message includes at least one of the following items: the current location of the UE using the AI/ML model, the identifier of the application using the AI/ML model, the quality of service flow identifier of the transmission AI/ML model, the transmission AI/ML model
  • the bit rate of the uplink direction of the model and the bit rate of the downlink direction of the transmission AI/ML model, the packet delay of the uplink direction of the AI/ML model and the packet delay of the downlink direction of the AI/ML model, the quality of service flow in The number of abnormal releases during the time period of AI/ML model transmission, the number of packet transmissions of AI/ML models, the number of packet retransmissions of AI/ML models, data collection time, the duration of AI/ML model transmission, and the transmission of AI/ML models start timestamp of AI/ML model transfer, size of AI/ML transfer model, network slice of PDU session for transfer of AI/ML model QoS stream, network slice for transfer of AI/ML model QoS stream.
  • the analysis information includes at least one of the following: network slices of PDU sessions used to transmit AI/ML model quality of service flows, identification of applications using AI/ML models, area information using AI/ML models, and analysis results.
  • Valid time user plane function UPF that provides AI/ML model transmission
  • data network name of PDU session used to transmit AI/ML model quality of service flow size of AI/ML transmission model, duration of AI/ML model transmission
  • AI /ML model transmission start timestamp AI/ML model transmission end timestamp
  • quality of service flow identifier for transmission AI/ML model
  • uplink direction bit rate for transmission AI/ML model and time stamp for transmission AI/ML model Bit rate in the downlink direction packet delay in the uplink direction of the AI/ML model and packet delay in the downlink direction of the AI/ML model, the number of abnormally released QoS streams during the transmission period of the AI/ML model, and the The number of reporting thresholds for abnormal release of quality of service flows during the AI/ML model transmission period, the number of packet transmissions for the AI/ML model, and
  • the parameters requested in the first message also include: federated learning group information, and the federated learning group information includes at least one of the following: the federated learning group used to indicate the analysis ID, UE ID or UE(s) ID participating in federated learning, application ID participating in federated learning;
  • the second message further includes at least one of the following items: the identity of the federated learning group used to indicate the analysis, the UE identity or UE(s) identity participating in the federated learning, and the application identity participating in the federated learning;
  • the analysis information also includes at least one of the following items: 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 user indicating to provide the AI/ML model or participate in the federated learning Individual application IDs.
  • the present disclosure provides an apparatus for analyzing model transmission status in a subscription network, the apparatus includes: a memory, a transceiver, and a processor:
  • the memory is used to store computer programs; the transceiver is used to send and receive data under the control of the processor; the processor is used to read the computer programs in the memory and perform the following operations:
  • the analysis information is used to adjust network policy parameters and/or application layer model information.
  • the present disclosure provides a device for analyzing model transmission status in a subscription network, the device includes a memory, a transceiver, and a processor:
  • the memory is used to store computer programs; the transceiver is used to send and receive data under the control of the processor; the processor is used to read the computer programs in the memory and perform the following operations:
  • the first message is used to request to subscribe to the analysis information of the transmission status of the artificial intelligence/machine learning AI/ML model in the network;
  • the second message is used to collect data for analyzing the transmission status of the AI/ML model in the network ;
  • the analysis information is used to adjust network policy parameters and/or application layer model information through AF.
  • the present disclosure provides an apparatus for analyzing model transmission status in a subscription network, the apparatus comprising:
  • the sending unit is configured to send a first message to the network data analysis function NWDAF directly or through the network capability exposure function NEF; wherein, the first message is used to request to subscribe to the analysis of the transmission status of the artificial intelligence/machine learning AI/ML model in the network information;
  • the analysis unit is used to receive the analysis information of the AI/ML model transmission status sent by the NWDAF directly or through the NEF, and the analysis information is sent by the NWDAF according to other network functions 5GC NF(s) receiving the 5G core network The data of the AI/ML model transmission status is determined;
  • the analysis information is used to adjust network policy parameters and/or application layer model information.
  • the present disclosure provides a device for analyzing model transmission status in a subscription network, the device comprising:
  • the receiving unit is configured to receive the first message sent by the application function AF directly or through the network capability exposure function NEF; wherein, the first message is used to request to subscribe to the analysis information of the transmission status of the artificial intelligence/machine learning AI/ML model in the network ;
  • the sending unit 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, and the second message is used to collect and analyze AI/ML in the network The data of the model transfer state;
  • the analysis unit is used to receive the data of the transmission state of the AI/ML model sent by other network functions 5GC NF(s) of the 5G core network, and analyze the data of the transmission state of the AI/ML model to obtain the data of the transmission state of the AI/ML model analyze information;
  • the analysis information is used to adjust network policy parameters and/or application layer model information through AF.
  • the present disclosure provides a processor-readable storage medium, the processor-readable storage medium stores a computer program, and the computer program is used to enable the processor to execute any one of the first aspect or the second aspect. method described in the item.
  • the disclosure provides a method, device, and readable storage medium for analyzing model transmission status in a subscription network, and sends a first message to the network data analysis function NWDAF directly or through the network capability exposure function NEF; wherein, the first message is used to request Subscribe to the analysis information of the transmission status of the artificial intelligence/machine learning AI/ML model in the network; receive the analysis information of the transmission status of the AI/ML model sent by the NWDAF directly or through the NEF, and the analysis information is obtained by the NWDAF according to the received
  • Other network functions 5GC NF(s) of the 5G core network are determined by the data of the transmission state of the AI/ML model sent by the NWDAF according to the request in the first message received Parameters obtained by sending a second message to the 5GC NF(s), the second message is used to collect data for analyzing the transmission status of the AI/ML model in the network; wherein the analysis information is used to adjust the network Information about policy parameters and/or application layer models.
  • the information of the layer model realizes the effective analysis of the transmission state of the AI/ML model, and then enables the network to effectively adjust the network transmission strategy based on the transmission state of the AI/ML model and enables the third party to obtain the analysis of the transmission state of the AI/ML model. Adjustment of application layer information.
  • FIG. 1 is a network architecture diagram of a model transmission state analysis method in a subscription network provided by an embodiment of the present disclosure
  • FIG. 2 is a network architecture diagram of a 5GC supporting network data analysis provided by an embodiment of the present disclosure
  • FIG. 3 is a schematic flowchart of a method for analyzing model transmission status in a subscription network provided by Embodiment 1 of the present disclosure
  • FIG. 4 is a schematic diagram of a signaling flow of a method for analyzing a model transmission state in a subscription network provided by Embodiment 1 of the present disclosure
  • FIG. 5 is a schematic diagram of a signaling flow of a method for analyzing a model transmission state in a subscription network provided by Embodiment 2 of the present disclosure
  • FIG. 6 is a schematic diagram of a signaling flow of a method for analyzing a model transmission state in a subscription network provided by Embodiment 3 of the present disclosure
  • FIG. 7 is a schematic flowchart of a method for analyzing model transmission status in a subscription network provided by Embodiment 4 of the present disclosure.
  • FIG. 8 is a schematic structural diagram of a model transmission state analysis device in a subscription network provided by an embodiment of the present disclosure.
  • FIG. 9 is a schematic structural diagram of a model transmission state analysis device in a subscription network provided by another embodiment of the present disclosure.
  • FIG. 10 is a schematic structural diagram of a device for analyzing model transmission status in a subscription network provided by another embodiment of the present disclosure.
  • Fig. 11 is a schematic structural diagram of an apparatus for analyzing model transmission status in a subscription network provided by another embodiment of the present disclosure.
  • Scenario 1 Distribution and sharing of AI/ML models. Due to changes in tasks or environments, the memory of mobile terminals is limited, and all models cannot be pre-loaded. Therefore, mobile terminals need to download new AI/ML models from the network in real time through the 5G system.
  • Scenario 2 Through the 5GS federated learning algorithm.
  • the cloud server trains a global model, it needs to aggregate the models trained locally by each terminal device.
  • Each training iteration process a terminal device downloads a global model from the cloud server and uses local data for training; 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, and then puts The global model is distributed to the terminal; the terminal executes the next iteration.
  • AI/ML model segmentation between AI/ML endpoints An AI/ML model can be divided into multiple parts based on the current task or environment. The trend is to infer the parts that are complex and energy-intensive by the network, and the parts that require privacy protection or delay sensitivity are inferred on the terminal. For example, the terminal downloads/onboards a model, first infers specific layers/parts, and then sends the intermediate results to the network; the network then executes the remaining layers/parts, and then feeds back the inference results to the terminal.
  • This scenario transfers part of the model in the first step or in the middle, so it may include the transfer of the model.
  • the 5G system is used as a channel for transmitting AI/ML models.
  • the 5G system needs to support the transfer of AI/ML models.
  • the monitoring and status information of ML sessions is exposed to third parties.
  • third parties cannot effectively adjust their own behavior based on the transmission status of AI/ML models, and the network cannot be based on AI/ML models.
  • ML model transfer state effectively adjusts network state.
  • AF Application Function
  • NEF Network Exposure Function
  • NWDAF Network data analysis function
  • NF Network Function
  • Each network function (English: Network Function, abbreviated as: NF) (that is, NF(s)) collects data to analyze the transmission status of the AI/ML model in the network and gives feedback, which can effectively analyze the transmission status of the AI/ML model , so that the network can effectively adjust the network status based on the AI/ML model transmission status and the third party can obtain the analysis of the AI/ML model transmission status to adjust its own behavior data.
  • NF Network Function
  • the method for analyzing the model transmission state in the subscription network proposed in this disclosure is proposed.
  • the first message is sent to the network data analysis function NWDAF directly or through the network capability exposure function NEF;
  • the first message is used to request to subscribe to the analysis information of the transmission status of the artificial intelligence/machine learning AI/ML model in the network;
  • the analysis information of the transmission status of the AI/ML model sent by the NWDAF is received directly or through the NEF, and the analysis
  • the information is determined by the NWDAF according to the data of the transmission state of the AI/ML model sent by other network functions of the 5G core network (that is, 5GC NF(s)), and the data of the transmission state of the AI/ML model is determined by the NWDAF according to
  • the parameters requested in the first message received are obtained by sending a second message to the 5GC NF(s), and the second message is used to collect data for analyzing the transmission status of the AI/ML model in the network;
  • the information of the layer model realizes the effective analysis of the transmission state of the AI/ML model, and then enables the network to effectively adjust the network transmission strategy based on the transmission state of the AI/ML model and enables the third party to obtain the analysis of the transmission state of the AI/ML model. Adjustment of application layer information.
  • FIG. 2 is a network architecture diagram of 5GC supporting network data analysis provided by the embodiment of the present disclosure.
  • NWDAF is the network analysis function managed by the operator, and NWDAF can provide 5GC network functions and applications Function (English: Application Function, referred to as: AF) and operation management and maintenance (English: Operation Administration and Maintenance, referred to as: OAM) provide data analysis services.
  • AF Application Function
  • OAM Operation Administration and Maintenance
  • the analysis result may be historical statistical information or forecast information.
  • NWDAF can serve one or more network slices.
  • 5GC also includes other functions. They are user plane function (English: User Plane Function, referred to as: UPF), session management function (English: Session Management Function, referred to as: SMF), access and mobility management function (English: Access and Mobility Management Function, referred to as: AMF), unified database (English: Unified Data Repository, referred to as: UDR), network capability exposure function (English: Network Exposure Function, referred to as: NEF), AF, policy control function (English: It is: Policy Control Function, referred to as: PCF) and online charging system (English: Online Charging System, referred to as: OCS). Wherein, these other functions may be collectively referred to as NF. NWDAF communicates with other functional entities 5GC NF(s) and OAM in the 5G core network based on the service interface.
  • UPF User Plane Function
  • SMF Session Management Function
  • AMF Access and Mobility Management Function
  • UDR Unified Data Repository
  • NEF Network Exposure Function
  • AF Policy Control Function
  • NWDAF instances There can be different NWDAF instances in 5GC providing different types of dedicated analysis.
  • the NWDAF instance needs to provide its support Analytic ID when registering with the network database function (English: Network Repository Function, abbreviated as: NRF).
  • the analysis type (or analysis ID) is specified.
  • the consumer NF can provide the Analytic ID to indicate what type of analysis is required when querying the NWDAF instance to the NRF.
  • the 5GC network function and OAM decide how to use the data analysis provided by the network data analysis function NWDAF to improve network performance.
  • the AF requests NWDAF to provide AI/ML model transmission status analysis
  • the analysis result includes: the application identification (i.e. Application ID) using the AI/ML model, the use AI Area information of the /ML model, the time period for transmitting the AI/ML model, the size of the transmitted model, the quality of service (ie QoS) related information for the transmission of the AI/ML model, the network slice used for the transmission of the AI/ML model, and the name of the data network (English: Data Network Name, abbreviated as: DNN) information; if there is federated learning, it also includes: group identification (that is, federated learning group ID), UE ID or UE group ID participating in federated learning, providing models or participating in federated learning The address information of the application server.
  • AF requests to adjust the 5GS network strategy or adjust the application layer AI/ML model parameters to optimize the AI/ML model transmission status
  • 5GC NF(s) adjusts the network strategy based on the request or AF adjusts the application layer based on data analysis AI/ML model parameters.
  • AF when AF sends an analysis request to NWDAF, if AF is in the trusted area, AF can directly send the request to NWDAF; NWDAF sends the request.
  • NEF directly or through NEF, send a request to NWDAF to subscribe to the analysis information of the AI/ML model transmission status in the network, receive the analysis information sent by NWDAF based on the collected 5GC NF(s) data, and adjust the network strategy based on the analysis information Parameters and/or application layer model information enables effective analysis of the AI/ML model transmission status, thereby enabling the network to effectively adjust the network transmission strategy based on the AI/ML model transmission status and enabling third parties to obtain the AI/ML model
  • the analysis of the transmission status adjusts the information of the application layer.
  • Figure 3 is a schematic flowchart of the method for analyzing the model transmission state in the subscription network provided by Embodiment 1 of the present disclosure.
  • the method for analyzing the state of model transmission in the subscription network provided by the embodiment includes the following steps:
  • step 101 the AF sends a first message to the network data analysis function NWDAF directly or through the network capability exposure function NEF.
  • the first message is used to request to subscribe to the analysis information of the transmission status of the artificial intelligence/machine learning AI/ML model in the network.
  • the parameters requested in the first message include at least one of the following: a network data analysis identifier (that is, Analytics ID), an identifier of a user equipment UE or a group of UEs that receive the AI/ML model, or an identifier that satisfies the analysis condition Any UE (i.e. Target of Analytics Reporting), the identification of the application using the AI/ML model (i.e. Application ID), the area of AI/ML model transmission (i.e.
  • a network data analysis identifier that is, Analytics ID
  • an identifier of a user equipment UE or a group of UEs that receive the AI/ML model or an identifier that satisfies the analysis condition Any UE (i.e. Target of Analytics Reporting)
  • the identification of the application using the AI/ML model i.e. Application ID
  • the area of AI/ML model transmission i.e.
  • AoI (Area of Interest)), indicating the quality of service of the AI/ML model transmission
  • the network slice of the protocol data unit PDU session of the flow ie S-NSSAI
  • the data network indicating the PDU session of the AI/ML model quality of service flow ie DNN
  • the time period of the AI/ML model transmission ie Model transmission duration
  • the start timestamp of AI/ML model transmission ie Model transmission start
  • the end timestamp of AI/ML model transmission ie Model transmission stop
  • the size of AI/ML transmission model ie Model size
  • the quality of service requirements ie, 5QI (5G QoS Identifier)
  • the quality of service requirements ie, QoS Characteristics
  • Table 1 Example table of parameters requested in the first message
  • the parameters requested in the first message may also include: federated learning group information (ie Federated Learning (FL) group information); federated learning group information includes at least one of the following: federated learning group used to indicate analysis The ID of the Federated Learning (FL) group ID), the UE ID or UE(s) ID participating in the federated learning (ie the Federated Learning (FL) UE ID or UE group ID), the application ID participating in the federated learning (ie the Federated Learning (FL) Application ID). See Table 2 below for an example table of parameters requested in the first message.
  • federated learning group information ie Federated Learning (FL) group information
  • federated learning group information includes at least one of the following: federated learning group used to indicate analysis The ID of the Federated Learning (FL) group ID), the UE ID or UE(s) ID participating in the federated learning (ie the Federated Learning (FL) UE ID or UE group ID), the application ID participating in the federated learning (ie the Federated Learning (FL
  • Table 2 Example table of parameters requested in the first message
  • the AF sends an AI/ML model transmission status subscription request to the NEF, such as Nnef_AnalyticsExposure_Subscribe (that is, analysis open subscription) or Nnef_AnalyticsExposure_Fetch (that is, analysis open access) request.
  • NEF sends the first message to NWDAF.
  • the first message can be an AI/ML model open transmission state subscription Nnwdaf_AnalyticsSubscription_Subscribe (that is, analysis subscription subscription) or Nnwdaf_AnalyticsInfo_Request (that is, analysis information) request.
  • the request can carry the parameters shown in the table, and the request Subscribe to the analysis information of the transmission status of AI/ML models in the network. If the AF is trusted (ie the AF is in the trusted zone), the AF sends the first message directly to the NWDAF.
  • the AI/ML model transmission status subscription request may carry the parameters shown in Table 1 or Table 2, requesting to subscribe to the analysis information of the AI/ML model transmission status in the network.
  • Step 102 the AF directly or through the NEF receives the analysis information of the AI/ML model transmission status sent by the NWDAF.
  • the analysis information is determined by the NWDAF according to the AI/ML model transmission status data sent by other network functions 5GC NF(s) of the 5G core network.
  • the data of the transmission state of the AI/ML model is obtained by the NWDAF by sending a second message to the 5GC NF(s) according to the parameters requested in the received first message, and the first The second message is used to collect data for analyzing the transmission status of the AI/ML model in the network.
  • the second message includes at least one of the following items: the current location of the UE using the AI/ML model (i.e. UE location), the identification of the application using the AI/ML model (i.e. Application ID, which may be the ID of the server) logo, which can also be the logo of AF), the quality of service flow identifier (ie QFI) of the transmission AI/ML model, the bit rate of the uplink direction of the transmission AI/ML model (ie bit rate for UL direction) and the transmission AI/ML model
  • the bit rate in the downlink direction of the ML model ie bit rate for DL direction
  • the packet delay in the uplink direction of the AI/ML model ie Packet delay for UL direction
  • the packet delay in the downlink direction of the AI/ML model That is, Packet delay for the DL direction
  • the number of abnormally released QoS flows during the AI/ML model transmission period QoS Sustainability
  • the packet transmission quantity of the AI/ML model packetet transmission
  • Table 3 Second message example table
  • the second message also includes at least one of the following items: indicating the identifier of the federated learning group used for the specified analysis (ie 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 identification for participating in federated learning (i.e. Federated Learning (FL) Application ID). See the second message example table shown in Table 4 below.
  • Table 4 Second message example table
  • the analysis information of the AI/ML model transmission status sent by the NEF is received, and the analysis information is sent by the NWDAF to the NEF. If the AF is trusted, it directly receives the analysis information of the AI/ML model transmission status sent by the NWDAF.
  • the analysis information includes at least one of the following: network slices of PDU sessions used to transmit AI/ML model quality of service flows, identification of applications using AI/ML models, area information using AI/ML models, analysis The effective time of the result (i.e. Validity period), the user plane function UPF that provides AI/ML model transmission (i.e. UPF Info), the data network name of the PDU session used to transmit the AI/ML model quality of service flow, and the AI/ML transmission model
  • the size of AI/ML model transmission, the start time stamp of AI/ML model transmission, the end time stamp of AI/ML model transmission, quality of service requirements (ie QoS requirements); quality of service requirements include: transmission of AI/ML model Quality of Service Flow Identifier (i.e.
  • bit rate in uplink direction for AI/ML model transmission and downlink direction bit rate for AI/ML model transmission bit rate in uplink direction for AI/ML model transmission and downlink direction bit rate for AI/ML model transmission
  • packet delay in uplink direction for AI/ML model and AI Packet delay in the downlink direction of the /ML model packet delay in uplink direction for AI/ML model and AI Packet delay in the downlink direction of the /ML model
  • the number of abnormal releases of QoS flows during the time period of AI/ML model transmission and the number of times the reporting threshold of abnormal release of QoS flows during the time period of AI/ML model transmission is reached
  • the number of packet transmissions for the AI/ML model and the number of packet retransmissions for the AI/ML model. See Table 5 below for an example table of analysis information.
  • the analysis information also includes at least one of the following: the identification of the federated learning group used for the specified analysis, the UE identification or UE(s) identification participating in the federated learning, an indication to provide AI/ML models or participate in the federation
  • the learned application identifiers that is, Application Server Instance Address. See Table 6 below for an example table of analysis information.
  • FIG. 4 is a schematic diagram of the signaling flow of the model transmission state analysis method in the subscription network provided by Embodiment 1 of the present disclosure.
  • FIG. 4 is the AF and NWDAF in the subscription network model transmission state analysis method. Signaling interaction diagram between NEF and NF.
  • the method for analyzing model transmission status in the subscription network provided by this embodiment includes the following steps (that is, the signaling interaction process corresponding to Embodiment 1: AF requests NWDAF to provide AI/ML model transmission status analysis): (wherein, steps 4011 to 4016 are AF In an untrusted zone, steps 4021 to 4024 are AF in a trusted zone.)
  • Step 4011 if the AF is untrusted, the AF sends an AI/ML model transfer status subscription Nnef_AnalyticsExposure_Subscribe (namely analysis open subscription) or Nnef_AnalyticsExposure_Fetch (namely analysis open access) request to the NEF.
  • Nnef_AnalyticsExposure_Subscribe namely analysis open subscription
  • Nnef_AnalyticsExposure_Fetch namely analysis open access
  • the request may carry the parameters shown in Table 1 or Table 2, requesting to subscribe to the analysis information of the transmission status of the AI/ML model in the network.
  • Step 4012 NEF sends AI/ML model open transmission status subscription Nnwdaf_AnalyticsSubscription_Subscribe (ie analysis subscription subscription) or Nnwdaf_AnalyticsInfo_Request (ie analysis information) request to NWDAF.
  • Nnwdaf_AnalyticsSubscription_Subscribe ie analysis subscription subscription
  • Nnwdaf_AnalyticsInfo_Request ie analysis information
  • the AI/ML model open transmission state subscription Nnwdaf_AnalyticsSubscription_Subscribe or Nnwdaf_AnalyticsInfo_Request request can be used as the first message.
  • the request may carry the parameters shown in Table 1 or Table 2, requesting to subscribe to the analysis information of the transmission status of the AI/ML model in the network.
  • Step 4013 NWDAF calls Nnf_EventExposure_Subscribe (event open subscription) to collect data from 5GC NF(s).
  • the collected data is shown in Table 3 or Table 4, which is used to analyze the transmission status of the AI/ML model in the network.
  • the way for NWDAF to send the second message to the 5GC NF(s) may be that NWDAF calls Nnf_EventExposure_Subscribe.
  • Step 4014 5GC NF(s) calls Nnf_EventExposure_Notify (event exposure notification) to feed back required data to NWDAF.
  • Step 4015 NWDAF calls Nnwdaf_AnalyticsSubscription_Notify (i.e. analysis subscription notification) or Nnwdaf_AnalyticsInfo_Request response (i.e. analysis information request response) to send analysis information of AI/ML model transmission status to NEF.
  • Nnwdaf_AnalyticsSubscription_Notify i.e. analysis subscription notification
  • Nnwdaf_AnalyticsInfo_Request response i.e. analysis information request response
  • Step 4016 NEF calls Nnef_AnalyticsExposure_Notify (that is, analysis open notification) or Nnef_AnalyticsExposure_Fetch response (that is, analysis open fetch response) to send analysis information of AI/ML model transmission status to AF.
  • Nnef_AnalyticsExposure_Notify that is, analysis open notification
  • Nnef_AnalyticsExposure_Fetch response that is, analysis open fetch response
  • Step 4021 if the AF is trusted, the AF directly sends an AI/ML model transmission status subscription request to the NWDAF, and the operation is performed as described in step 4012 .
  • the consumer can also be PCF, SMF.
  • Step 4022 perform operations as described in step 4013. Namely step 4013.
  • Step 4023 perform operations as described in step 4014. Namely step 4014.
  • Step 4024 NWDAF directly sends the analysis information of the AI/ML model transmission status to the AF, and the execution operation is as described in step 4016.
  • Embodiment 2 after receiving the analysis information, the method further includes:
  • a first request is sent to the policy control function PCF directly or through the NEF.
  • the first request is used to request to update the network policy parameters used for AI/ML model transmission; the network policy parameters are used to optimize the transmission status of the AI/ML model.
  • the AF requests to adjust the network policy to optimize the AI/ML model transmission status. Specifically, if the AF is in the trusted area, the AF 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 area, the AF 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.
  • sending the first request to the policy control function PCF directly or through the NEF may be implemented through the following steps:
  • Step a1 according to the bit rate in the uplink direction of the AI/ML model transmitted in the analysis information, the bit rate in the downlink direction of the AI/ML model transmitted, the packet delay in the uplink direction of the AI/ML model, and the AI/ML Packet delay in the downlink direction of the model, the number of abnormal releases of QoS flows during the time period of AI/ML model transmission, the number of packet transmissions of AI/ML models, the number of packet retransmissions of AI/ML models, and the number of QoS flows reached At least one of the number of reporting thresholds for abnormal release during the AI/ML model transmission time period determines a new quality of service parameter for transmitting the AI/ML model; the new quality of service parameter includes at least one of the following: 5G quality of service identifier, reflective quality of service control, maximum bit rate in uplink direction for transmitting AI/ML models, maximum bit rate for downlink direction for transmitting AI/ML models, uplink direction for transmitting AI/ML models Minimum bit rate, minimum bit rate in the downlink direction for
  • Step a2 according to the identification of the application transmitting the AI/ML model in the analysis information, the area information using the AI/ML model, the IP address information of the application service using the AI/ML model, and the service for transmitting the AI/ML model
  • Step a3 according to the area information and address information of the UE(s) and each AF transmitting the AI/ML model, 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 UE(s) corresponding to the DNAI and the area information and address information of each AF are used to provide a path for optimizing the transmission status of the AI/ML model.
  • Step a4 sending the new quality of service parameter, the DNAI, and the UE(s) corresponding to the DNAI and the area information and address information of each AF as parameters in the first request directly or through the NEF to PCF.
  • AF analyzes the information obtained from NWDAF on the transmission AI/ML model, such as: the bit rate of the uplink direction of the transmission AI/ML model and the downlink direction bit rate of the transmission AI/ML model, AI/ML model Packet delay in the uplink direction of the AI/ML model and packet delay in the downlink direction of the AI/ML model, the number of abnormally released QoS streams during the transmission period of the AI/ML model, the number of packet transmissions of the AI/ML model, AI/ML The number of packet retransmissions of the model, the number of times the reporting threshold of the abnormal release of the quality of service flow during the AI/ML model transmission period is reached, the new QoS parameters for the transmission AI/ML model are determined, and the new QoS parameters are provided to the PCF.
  • the bit rate of the uplink direction of the transmission AI/ML model and the downlink direction bit rate of the transmission AI/ML model such as: the bit rate of the uplink direction of the transmission AI/ML model and the downlink direction bit rate of the transmission AI
  • AF analyzes the information obtained from NWDAF about the transmission of AI/ML models, such as: the identification of the application that transmits the AI/ML model, the area information that uses the AI/ML model, the IP address information of the application service that uses the AI/ML model, the user The network slice of the PDU session used to transmit the AI/ML model quality of service flow, the data network name of the PDU session used to transmit the AI/ML model quality of service flow, or, AF based on the analysis information about the transmission AI/ML model obtained from NWDAF For example: the identification of the federated learning group that indicates the analysis, the UE identification or UE(s) identification that participates in the federated learning, the identification of each application that indicates the provision of the AI/ML model or participates in the federated learning, and determines the UE(s) that transmits the AI/ML model ) and the area information and address information of each AF, and then determine the data network access identifier DNAI and the UE(s) corresponding to the DNAI and the area information
  • the AF sends the first request carrying the new QoS parameters, the DNAI, and the UE(s) corresponding to the DNAI and the area information and address information of each AF to the PCF directly or through the NEF, and requests the PCF to The requested parameters are carried to update the relevant policy parameters for AI/ML model transmission to optimize the AI/ML model transmission status.
  • step a3 can be achieved through the following steps:
  • Step a31 according to the UE(s) transmitting the AI/ML model and the area information and address information of each AF, it is judged whether the current routing path is not good;
  • Step a32 if the current routing path is not good, then determine the destination addresses of both parties in the AI/ML model according to the address information of the UE(s) and each AF that is transmitting the AI/ML model, and the area information of the UE(s);
  • Step a33 Determine the shortest path according to the destination address
  • Step a34 Determine the DNAI and the area information and address information of the UE(s) and each AF corresponding to the DNAI according to the shortest path.
  • the first request is specifically used for:
  • the method also includes:
  • the first update result sent by the PCF is received directly or through the NEF, where the first update result includes that the first request is accepted or the first request is rejected.
  • the AF requests the PCF to adjust the PCC rules according to the new QoS parameters provided.
  • the request PCF adjusts the 5G service quality identifier, reflective service quality control, transmission AI/ Maximum bit rate in the uplink direction for ML models, maximum bit rate in the downlink direction for transmitting AI/ML models, minimum bit rate in the uplink direction for transmitting AI/ML models, minimum in the downlink direction for transmitting AI/ML models.
  • the bit rate, the priority of the quality of service flow, and the first update result is notified directly or through the NEF, that is, the AF is notified whether the request is accepted or rejected.
  • the first request is specifically used for:
  • the PCF Requesting the PCF to determine whether the session management function network element SMF needs to update the session management strategy, if it is determined that the SMF needs to update the session management strategy, then determine that the PCF sends a second request to the SMF, and the parameters requested in the second request include at least one of the following : DNAI, traffic steering policy identifier, traffic route 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 corresponding DNAI, traffic steering policy identifier, traffic route information;
  • the method also includes:
  • the second update result includes whether the first request is accepted or rejected.
  • the AF requests the PCF to determine whether the session management function network element SMF needs to update the session management policy. If the PCF determines that the SMF needs to update the policy information, the PCF will initiate a request to the SMF carrying DNAI, traffic steering policy identifier, traffic route information, etc. For the second request of parameters, the SMF determines the selected user plane function UPF according to the new session management policy and provides the corresponding DNAI, traffic steering policy identifier, and traffic route information to update the session management policy, that is, update the SM policy. And notify the second update result directly or through the NEF, that is, notify the AF whether the request is accepted or rejected.
  • FIG. 5 is a schematic flow diagram of the signaling flow of the model transmission state analysis method in the subscription network provided by Embodiment 2 of the present disclosure.
  • FIG. 5 is the AF, NEF and Signaling interaction diagram between PCFs.
  • the method for analyzing the model transmission state in the subscription network provided by this embodiment includes the following steps (that is, the signaling interaction process corresponding to Embodiment 2: based on the NWDAF analysis result, the AF requests to adjust the network policy to optimize the AI/ML model transmission state): (where , steps 5011 to 5014 are AF in an untrusted area, and steps 5021 to 5022 are AF in a trusted area.)
  • Step 5010 AF signs and retrieves AI/ML model transmission status analysis from NWDAF.
  • the AF subscribes to and obtains the AI/ML model transmission status analysis through the NEF or directly to the NWDAF, as described in steps 4011 to 4024 above.
  • Step 5011 AF sends Nnef_AFsessionWithQoS_Update (that is, AF session update based on QoS) request to NEF.
  • Nnef_AFsessionWithQoS_Update that is, AF session update based on QoS
  • the AF can send a Nnef_AFsessionWithQoS_Update (ie AF session update based on quality of service) request to the NEF to update the AF session used for AI/ ML model transfer related policy parameters to optimize the AI/ML model transfer status.
  • Nnef_AFsessionWithQoS_Update ie AF session update based on quality of service
  • AF analyzes the information about the transmission AI/ML model obtained from NWDAF: QoS flow Bit Rate (that is, the quality of service flow bit rate, such as: the uplink direction bit rate of the transmission AI/ML model and the transmission AI/ML model bit rate in the downlink direction), QoS flow Packet Delay (that is, the quality of service flow packet delay, such as: packet delay in the uplink direction of the AI/ML model and packet delay in the downlink direction of the AI/ML model), QoS Sustainability , Packet transmission, Packet retransmission, QoS Sustainability, QoS Sustainability Reporting Threshold(s), determine the new QoS parameters of the transmission AI/ML model, and provide new QoS parameters to PCF: 5G QoS Identifier (5QI, namely 5G service quality identification character), Reflective QoS Control (reflective quality of service control), UL-maximum bitrate (ie, the maximum bit rate in the uplink direction of the transmission AI/ML model), DL-max
  • AF analyzes the information obtained from NWDAF on the transmission of AI/ML models: AF ID of the transmission AI/ML model, UE region information using the AI/ML model, IP address information of the application service using the AI/ML model, AI /ML model transmission uses the network slice and DNN information, and determines the area information and address information of the UE(s) and AF(s) that transmit the AI/ML model.
  • federated learning group information there may also be information about federated learning group information in the analysis information: Federated Learning (FL) group ID, Federated Learning (FL) UE ID or UE group ID, Application ID of Federated Learning (FL), determine the transmission AI/ML
  • the area information and address information of the UE(s) and AF(s) of the model judge that the current routing path is not good, select the DNAI that can provide better service experience or performance, and provide the DNAI and corresponding information for the transmission AI/ML model to PCF Area information and address information of UE(s) and AF(s).
  • the process of judging that the current routing path is not good can be: AF is based on the IP address information of AF, the area information of UE(s) (it can correspond to the ID of AMF, SMF, UPF, etc., and the N6 interface of UPF is connected to DN), determine Current routing paths; for example, if some UEs/servers join/leave (federated group) before the next transmission, then some paths are not good, or if there are too many hops, the paths are not good.
  • 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 transmission parties can be determined, and a better (nearest ) routing path, corresponding to the DNAI.
  • Step 5012 NEF sends Npcf_PolicyAuthorization_Update request (namely policy authorization update request) to PCF.
  • the NEF sends the above information to the PCF through the Npcf_PolicyAuthorization_Update request (i.e., a policy authorization update request) to update relevant policy parameters for AI/ML model transmission so as to optimize the AI/ML model transmission status.
  • Npcf_PolicyAuthorization_Update request i.e., a policy authorization update request
  • Npcf_PolicyAuthorization_Update request can be used as the second request.
  • Step 5013 PCF notifies NEF of the result (that is, PCF sends Npcf_PolicyAuthorization_Update response (policy authorization update response) to NEF.
  • the PCF adjusts the PCC rules according to the information provided by the NEF, specifically, according to the new QoS parameters provided: 5G QoS Identifier (5QI), Reflective QoS Control, UL-maximum bitrate, DL-maximum bitrate, UL-guaranteed bitrate, DL-guaranteed bitrate, Priority Level, and send Npcf_PolicyAuthorization_Update response (policy authorization update response) to notify NEF of the result.
  • 5G QoS Identifier 5G QoS Identifier
  • Reflective QoS Control Reflective QoS Control
  • UL-maximum bitrate DL-maximum bitrate
  • UL-guaranteed bitrate UL-guaranteed bitrate
  • Priority Level Priority Level
  • Npcf_PolicyAuthorization_Update response policy authorization update response
  • Npcf_SMPolicyControl_UpdateNotify request (that is, session policy control update notification request) (DNAI, Per DNAI: Traffic steering policy identifier, Per DNAI: N6 traffic routing information) to SMF to update SM policy, SMF Select UPF according to this strategy, and provide DNAI, Per DNAI: Traffic steering policy identifier (traffic steering policy identifier), Per DNAI: N6 traffic routing information (traffic routing information).
  • Npcf_SMPolicyControl_UpdateNotify request can be used as the second request.
  • Step 5014 NEF sends Nnef_AFsessionWithQoS_Update response to AF.
  • the NEF sends Nnef_AFsessionWithQoS_Update response (that is, an AF session update response based on quality of service), and notifies the AF whether the request is accepted or rejected.
  • Step 5021 AF directly sends Npcf_PolicyAuthorization_Update request to PCF.
  • the AF directly sends the Npcf_PolicyAuthorization_Update request to the PCF to update the relevant policy parameters for AI/ML model transmission so as to optimize the AI/ML model transmission status, and the execution operation is as described in step 5012.
  • Step 5022 the PCF notifies the AF directly of the result (that is, the PCF directly sends an Npcf_PolicyAuthorization_Update response (that is, a policy authorization update response) to the AF).
  • the PCF performs operations as described in step 5012 according to the information provided by the AF.
  • the PCF notifies the AF directly whether the request is accepted or rejected.
  • Embodiment 3 after receiving the analysis information, the method can also be implemented through the following steps:
  • Step b1 adjust the information of the application layer model according to 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, model codec; the application layer model The information is used to update the quality of service parameters;
  • Step b2 according to the information of the adjusted application layer model, determine the new service quality parameters, the new service quality parameters include: 5G service quality identifier, reflective service quality control, uplink of the transmission AI/ML model Maximum bit rate in link direction, maximum bit rate in downlink direction for AI/ML model transmission, minimum bit rate in uplink direction for transmission AI/ML model, minimum bit rate in downlink direction for transmission AI/ML model, service priority of mass flow;
  • Step b3 sending a third request to the policy control function PCF directly or through the NEF;
  • the parameters requested in the third request include the new QoS parameter, and the third request is used to request to update the QoS parameter.
  • AF adjusts the application layer model information to update service quality parameters, such as: adjusting model compression, model size, model transmission time period, model encoding and decoding, etc.
  • service quality parameters such as: adjusting model compression, model size, model transmission time period, model encoding and decoding, etc.
  • the new QoS parameter is determined, and directly or through the NEF, the PCF sends a third request carrying the new QoS parameter.
  • the third request is specifically used for:
  • the method also includes:
  • the third update result is determined by the PCF based on the result of adjusting the PCC rules, the third update result includes the third request being accepted or the third request being rejected .
  • the AF requests the PCF to adjust the 5G service quality identifier, reflective service quality control, and the maximum uplink direction of the transmission AI/ML model in the PCC (policy and charging control) rules according to the new QoS parameters provided.
  • Bit rate, maximum bit rate in downlink direction for AI/ML model transmission, minimum bit rate for uplink direction for AI/ML model transmission, minimum bit rate in downlink direction for AI/ML model transmission, priority, PCF adjustment The PCC rules and notifies the AF directly or through the NEF whether this request is accepted or rejected.
  • the method may also be implemented through the following steps:
  • the model compression, model size and model codec in the information of the adjusted application layer model are directly sent to the PCF, and the information of the adjusted application layer model is used to support the PCF to adjust the 5G service quality identifier in the PCC rule, Reflective quality of service control, maximum bit rate in uplink direction for transmitting AI/ML models, maximum bit rate in downlink direction for transmitting AI/ML models, minimum bit rate in uplink direction for transmitting AI/ML models, transmitting AI The lowest bit rate in the downlink direction of the /ML model, the priority of the quality of service flow;
  • the method also includes:
  • the PCF adjusts the aforementioned QoS parameters in the PCC rules according to model compression, model size, and model encoding and decoding. For example: 5G service quality identifier, reflective quality of service control, maximum bit rate for the uplink direction of AI/ML model transmission, maximum bit rate for downlink direction of AI/ML model transmission, uplink direction for AI/ML model transmission The lowest bit rate in the downlink direction, the lowest bit rate in the downlink direction for transmitting AI/ML models, and the priority of quality of service flows.
  • the PCF adjusts the PCC rules and informs the AF directly or through the NEF whether the request is accepted or rejected.
  • the method further includes:
  • the model transmission time in the information of the adjusted application layer model is directly sent to the PCF, and the model transmission time in the information of the adjusted application layer model is used to support the PCF to adjust the gate status parameter in the PCC rule;
  • the gate The state parameter is used to support SMF to update the session management policy according to the transmission start time and transmission end time in the gate state;
  • the method also includes:
  • the PCF adjusts the Gate status (ie, the gate status parameter) in the PCC rule according to the model transmission time, and updates the SM policy.
  • the SMF affects the transmission start and end time of the flow based on this, and feeds back to the PCF, and the PCF connects or passes the NEF to the AF. Notifies whether this request was accepted or rejected.
  • FIG. 6 is a schematic diagram of the signaling flow of the model transmission state analysis method in the subscription network provided by Embodiment 3 of the present disclosure.
  • FIG. 6 is the AF and NEF and Signaling interaction diagram between PCFs.
  • the method for analyzing the state of model transmission in the subscription network provided by this embodiment includes the following steps (i.e.
  • AF adjusts the information of the application layer model based on the analysis information provided by NWDAF, such as adjusting model compression, model Size, model transmission time period, model codec, etc., and then update the QoS requirements, similar to the steps in the second embodiment): (wherein, steps 6012 to 6014 are AF in an untrusted area, and steps 6021 to 6022 are AF in a trusted area area.)
  • Step 6010 AF signs and retrieves AI/ML model transmission status analysis from NWDAF.
  • the AF subscribes to the NWDAF through the NEF or directly and obtains the analysis of the transmission status of the AI/ML model, as described in steps 4011 to 4024 above.
  • Step 6011 AF adjusts the behavior of the application layer based on the analysis information.
  • the AF adjusts the information of the application layer model based on the QoS requirement information provided by the NWDAF, such as adjusting model compression, model size, model transmission time period, and model encoding and decoding.
  • Step 6012 AF sends Nnef_AFsessionWithQoS_Update request to NEF.
  • Step 6013 NEF sends Npcf_PolicyAuthorization_Update request to PCF.
  • Npcf_PolicyAuthorization_Update request can be used as the third request.
  • Step 6014 PCF notifies NEF of the result (that is, PCF sends Npcf_PolicyAuthorization_Update response to NEF.
  • Step 6015 NEF sends Nnef_AFsessionWithQoS_Update response to AF.
  • Step 6021 AF directly sends Npcf_PolicyAuthorization_Update request to PCF.
  • step 6022 the PCF notifies the AF directly of the result (that is, the PCF directly sends an Npcf_PolicyAuthorization_Update response (that is, a policy authorization update response) to the AF).
  • the AF requests the session to update the QoS.
  • AF determines new QoS parameters for transmitting AI/ML models based on the adjusted model compression, model size, model transmission time period, model codec, etc., and provides new QoS parameters to PCF: 5G QoS Identifier (5QI) , Reflective QoS Control, UL-maximum bitrate, DL-maximum bitrate, UL-guaranteed bitrate, DL-guaranteed bitrate, Priority Level.
  • 5QI 5G QoS Identifier
  • Reflective QoS Control UL-maximum bitrate, DL-maximum bitrate, UL-guaranteed bitrate, DL-guaranteed bitrate, Priority Level.
  • the AF directly sends model compression, model size, model transmission time period, model codec adjustment information, etc. to the PCF.
  • Step 6013 PCF adjusts PCC rules according to the new QoS parameters provided: 5G QoS Identifier (5QI), Reflective QoS Control, UL-maximum bitrate, DL-maximum bitrate, UL-guaranteed bitrate, DL-guaranteed bitrate, Priority Level; or PCF adjusts the above QoS parameters in PCC rules based on model compression, model size, and model encoding and decoding; or PCF adjusts Gate status in PCC rules based on model transmission time, updates SM policy, and SMF affects the transmission of streams based on this. and the end time; finally, the AF is notified whether the request is accepted or rejected by means of step 5014 or 5022 in the second embodiment.
  • 5QI 5G QoS Identifier
  • Reflective QoS Control UL-maximum bitrate, DL-maximum bitrate, UL-guaranteed bitrate, DL-guaranteed bitrate, Priority Level
  • PCF adjusts the
  • NWDAF receives the AI/ML model transmission status analysis request sent by AF, as well as the parameters contained in the request; NWDAF collects input data from 5GC NF(s) to analyze the AI/ML model transmission status in the network; NWDAF conducts Analyze and send AI/ML model transmission status analysis information to AF; AF requests AF session QoS update based on AI/ML model transmission status analysis information, and adjusts corresponding policy control function PCF, session management function network element SMF and other network elements Strategy; AF adjusts the relevant parameters of the application layer model information based on the analysis information of the AI/ML model transmission status, and then adjusts the QoS, thereby optimizing the transmission status of the AI/ML model.
  • the third party can obtain the transmission status of the AI/ML model, and based on the analysis results of the model transmission, the network can adjust its own behavior to meet the transmission requirements of the AI/ML model, and the third party can also adjust the model application layer based on the analysis results of the model transmission To achieve efficient transmission of AI/ML models and ensure the business experience and performance of AI/ML model transmission.
  • Fig. 7 is a schematic flowchart of the method for analyzing the model transmission state in the subscription network provided by Embodiment 4 of the present disclosure.
  • the execution subject of the method for analyzing the state of model transmission in the subscription network provided in this embodiment is NWDAF, then the present disclosure
  • the method for analyzing the state of model transmission in the subscription network provided by the embodiment includes the following steps:
  • Step 701 NWDAF receives the first message sent by the application function AF directly or through the network capability exposure function NEF; wherein, the first message is used to request to subscribe to the analysis information of the transmission status of the artificial intelligence/machine learning AI/ML model in the network;
  • Step 702 NWDAF sends a second message to other network functions 5GC NF(s) of the 5G core network according to the parameters requested in the first message, and the second message is used to collect and analyze AI/ML models in the network Transfer status data.
  • Step 703 NWDAF receives the data of the transmission state of the AI/ML model sent by other network functions 5GC NF(s) of the 5G core network, and analyzes the data of the transmission state of the AI/ML model to obtain the analysis of the transmission state of the AI/ML model information.
  • the analysis information is used to adjust network policy parameters and/or application layer model information through AF.
  • the parameters requested in the first message include at least one of the following: a network data analysis identifier, an identifier of a user equipment UE or a group of UEs receiving the AI/ML model, or any UE that meets the analysis conditions, using The identification of the application of the AI/ML model, the area where the AI/ML model is transmitted, the network slice indicating the PDU session of the protocol data unit PDU session that transmits the AI/ML model quality of service flow, and the data indicating the PDU session that transmits the AI/ML model quality of service flow Network, 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 used to indicate the transmission of AI/ML model QoS requirements for streams and/or specific QoS requirements for delivery of AI/ML models;
  • the second message includes at least one of the following items: the current location of the UE using the AI/ML model, the identifier of the application using the AI/ML model, the quality of service flow identifier of the transmission AI/ML model, the transmission AI/ML model
  • the bit rate of the uplink direction of the model and the bit rate of the downlink direction of the transmission AI/ML model, the packet delay of the uplink direction of the AI/ML model and the packet delay of the downlink direction of the AI/ML model, the quality of service flow in The number of abnormal releases during the time period of AI/ML model transmission, the number of packet transmissions of AI/ML models, the number of packet retransmissions of AI/ML models, data collection time, the duration of AI/ML model transmission, and the transmission of AI/ML models start timestamp of AI/ML model transfer, size of AI/ML transfer model, network slice of PDU session for transfer of AI/ML model QoS stream, network slice for transfer of AI/ML model QoS stream.
  • the analysis information includes at least one of the following: network slices of PDU sessions used to transmit AI/ML model quality of service flows, identification of applications using AI/ML models, area information using AI/ML models, and analysis results.
  • Valid time user plane function UPF that provides AI/ML model transmission
  • data network name of PDU session used to transmit AI/ML model quality of service flow size of AI/ML transmission model, duration of AI/ML model transmission
  • AI /ML model transfer start timestamp AI/ML model transfer end timestamp
  • QoS flow identifier for transfer AI/ML model, uplink direction bit rate for transfer AI/ML model and Bit rate in the downlink direction, packet delay in the uplink direction of the AI/ML model and packet delay in the downlink direction of the AI/ML model, the number of abnormally released QoS streams during the transmission period of the AI/ML model, and the The number of reporting thresholds for abnormal release of quality of service flows during the AI/ML model transmission period, the number of packet transmissions for the AI/ML model, and the number of packet retransmission
  • the parameters requested in the first message also include: federated learning group information, and the federated learning group information includes at least one of the following: the federated learning group used to indicate the analysis ID, UE ID or UE(s) ID participating in federated learning, application ID participating in federated learning;
  • the second message further includes at least one of the following items: the identifier of the federated learning group used for the specified analysis, the UE identifier or UE(s) identifier participating in the federated learning, and the application identifier participating in the federated learning;
  • the analysis information also includes at least one of the following: the identifier of the federated learning group used for the specified analysis, the UE identifier or UE(s) identifier participating in the federated learning, the user indicating to provide the AI/ML model or participating in the federated learning Individual application IDs.
  • the other network functions 5GC NF(s) of the 5G core network are collected Data, receiving the data of the transmission state of the AI/ML model sent by other network functions 5GC NF(s) of the 5G core network, and analyzing the data of the transmission state of the AI/ML model to obtain the analysis information of the transmission state of the AI/ML model, Realize the effective analysis of AI/ML model transmission status, and enable AF to adjust network policy parameters and/or application layer model information according to the analysis information, so that the network can effectively adjust network transmission based on AI/ML model transmission status Policy and the analysis of the transmission status of the AI/ML model to the third party to adjust the application information.
  • the method for analyzing the model transmission state in the subscription network provided by the embodiment of the present disclosure can realize all the method steps realized by the method embodiment shown in FIG. 4 and can achieve the same technical effect, which is not repeated here Parts and beneficial effects in this embodiment that are the same as those in the method embodiment are specifically described in detail.
  • Fig. 8 is a schematic structural diagram of a model transmission state analysis device in a subscription network provided by an embodiment of the present disclosure. As shown in Fig. 8, the model transmission state analysis device in a subscription network provided by this embodiment is applied to AF. Then, the apparatus for analyzing the transmission status of models in a subscription network provided in this embodiment includes: a transceiver 800 configured to receive and send data under the control of a processor 810 .
  • the bus architecture may include any number of interconnected buses and bridges, specifically one or more processors represented by the processor 810 and various circuits of the memory represented by the memory 820 are linked together.
  • the bus architecture can also link together various other circuits such as peripherals, voltage regulators, and power management circuits, etc., which are well known in the art and therefore will not be further described herein.
  • the bus interface provides the interface.
  • the transceiver 800 may be a plurality of elements, including a transmitter and a receiver, providing a unit for communicating with various other devices over transmission media, including wireless channels, wired channels, optical cables, and other transmission media.
  • the processor 810 is responsible for managing the bus architecture and general processing, and the memory 820 may store data used by the processor 810 when performing operations.
  • the processor 810 can be a central processing device (CPU), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field-Programmable Gate Array, FPGA) or a complex programmable logic device (Comple6 Programmable Logic Device , CPLD), the processor can also adopt a multi-core architecture.
  • CPU central processing device
  • ASIC Application Specific Integrated Circuit
  • FPGA field programmable gate array
  • CPLD Complex programmable logic device
  • 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:
  • the analysis information is the AI/ML model sent by the NWDAF directly or through the NEF, and the analysis information is the AI/ML model sent by the NWDAF according to other network functions 5GC NF(s) receiving the 5G core network
  • the data of the transmission state is determined, and the data of the transmission state of the AI/ML model is obtained by the NWDAF by sending a second message to the 5GC NF(s) according to 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;
  • the analysis information is used to adjust network policy parameters and/or application layer model information.
  • the parameters requested in the first message include at least one of the following: a network data analysis identifier, an identifier of a user equipment UE or a group of UEs receiving the AI/ML model, or any UE that meets the analysis conditions, using The identification of the application of the AI/ML model, the area where the AI/ML model is transmitted, the network slice indicating the PDU session of the protocol data unit PDU session that transmits the AI/ML model quality of service flow, and the data indicating the PDU session that transmits the AI/ML model quality of service flow Network, 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 used to indicate the transmission of AI/ML model QoS requirements for streams and/or specific QoS requirements for delivery of AI/ML models;
  • the second message includes at least one of the following items: the current location of the UE using the AI/ML model, the identifier of the application using the AI/ML model, the quality of service flow identifier of the transmission AI/ML model, the transmission AI/ML model
  • the bit rate of the uplink direction of the model and the bit rate of the downlink direction of the transmission AI/ML model, the packet delay of the uplink direction of the AI/ML model and the packet delay of the downlink direction of the AI/ML model, the quality of service flow in The number of abnormal releases during the time period of AI/ML model transmission, the number of packet transmissions of AI/ML models, the number of packet retransmissions of AI/ML models, data collection time, the duration of AI/ML model transmission, and the transmission of AI/ML models start timestamp of AI/ML model transfer, size of AI/ML transfer model, network slice of PDU session for transfer of AI/ML model QoS stream, network slice for transfer of AI/ML model QoS stream.
  • the analysis information includes at least one of the following: network slices of PDU sessions used to transmit AI/ML model quality of service flows, identification of applications using AI/ML models, area information using AI/ML models, and analysis results.
  • Valid time user plane function UPF that provides AI/ML model transmission
  • data network name of PDU session used to transmit AI/ML model quality of service flow size of AI/ML transmission model, duration of AI/ML model transmission
  • AI /ML model transmission start timestamp AI/ML model transmission end timestamp
  • quality of service flow identifier for transmission AI/ML model
  • uplink direction bit rate for transmission AI/ML model and time stamp for transmission AI/ML model Bit rate in the downlink direction packet delay in the uplink direction of the AI/ML model and packet delay in the downlink direction of the AI/ML model, the number of abnormally released QoS streams during the transmission period of the AI/ML model, and the The number of reporting thresholds for abnormal release of quality of service flows during the AI/ML model transmission period, the number of packet transmissions for the AI/ML model, and
  • the parameters requested in the first message also include: federated learning group information, and the federated learning group information includes at least one of the following: federated learning group used to refer to the analysis ID, UE ID or UE(s) ID participating in federated learning, application ID participating in federated learning;
  • the second message further includes at least one of the following items: the identity of the federated learning group used to indicate the analysis, the UE identity or UE(s) identity participating in the federated learning, and the application identity participating in the federated learning;
  • the analysis information also includes at least one of the following: the identifier of the federated learning group used for the specified analysis, the UE identifier or UE(s) identifier participating in the federated learning, the user indicating to provide the AI/ML model or participating in the federated learning Individual application IDs.
  • processor 810 is also used for:
  • the first request is used to request to update the network policy parameters used for AI/ML model transmission; the network policy parameters are used to optimize the transmission status of the AI/ML model.
  • processor 810 when the processor 810 is configured to send the first request to the policy control function PCF directly or through the NEF according to the analysis information, it specifically includes:
  • the packet delay in the uplink direction of the AI/ML model and the downlink direction of the AI/ML model Packet delay in the link direction, the number of abnormally released quality of service flows during the AI/ML model transmission period, the number of packet transmissions in the AI/ML model, the number of packet retransmissions in the AI/ML model, At least one of the number of reporting thresholds for abnormal release during the time period of ML model transmission, to determine the new quality of service parameters of the transmission 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 bit rate in uplink direction for transmitting AI/ML models, maximum bit rate in downlink direction for transmitting AI/ML models, minimum bit rate in uplink direction for transmitting AI/ML models , the lowest bit rate in the downlink direction of the transmission AI/ML model, and the priority of the quality
  • the identification of the application transmitting the AI/ML model the area information using the AI/ML model, the IP address information of the application service using the AI/ML model, and the quality of service flow used to transmit the AI/ML model in the analysis information
  • the network slice of the PDU session, the data network name of the PDU session used to transmit the AI/ML model quality of service flow determine the area information and address information of the UE(s) and each AF that transmits the AI/ML model; or, if the AI/ML model
  • the ML model performs federated learning, according to the identifier of the federated learning group indicated in the analysis information, the UE identifier or UE(s) identifier participating in the federated learning, and the identifier of each application that provides the AI/ML model or participates in the federated learning, Determine the UE(s) transmitting the AI/ML model and the area information and address information of each AF;
  • the area information and address information of the UE(s) and each AF transmitting the AI/ML model 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 said The UE(s) corresponding to DNAI and the area information and address information of each AF are used to provide a path to optimize the transmission status of the AI/ML model;
  • the processor 810 is configured to determine the data network access identifier DNAI and the UE(s) and each AF corresponding to the DNAI according to the UE(s) transmitting the AI/ML model and the area information and address information of each AF
  • the area information and address information of AF include:
  • the first request is specifically used for:
  • processor 810 also specifically:
  • the first update result sent by the PCF is received directly or through the NEF, where the first update result includes that the first request is accepted or the first request is rejected.
  • the first request is specifically used for:
  • the PCF Requesting the PCF to determine whether the session management function network element SMF needs to update the session management strategy, if it is determined that the SMF needs to update the session management strategy, then determine that the PCF sends a second request to the SMF, and the parameters requested in the second request include at least one of the following : DNAI, traffic steering policy identifier, 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 corresponding DNAI, traffic steering policy identifier, and traffic route information;
  • processor 810 also specifically:
  • the second update result includes whether the first request is accepted or rejected.
  • processor 810 is also used for:
  • the information of the application layer model After receiving the analysis information, adjust the information of the application layer model according to the analysis information, and the information of the application layer model includes at least one of the following: model compression, model size, model transmission time period, model codec ;
  • the information of the application layer model is used to update the quality of service parameters;
  • the new service quality parameters include: 5G service quality identifier, reflective service quality control, uplink direction of transmission AI/ML model Maximum bit rate, maximum bit rate for AI/ML model transmission in downlink direction, minimum bit rate for transmission AI/ML model in uplink direction, minimum bit rate for transmission AI/ML model in downlink direction, quality of service flow priority;
  • the parameters requested in the third request include the new QoS parameter, and the third request is used to request to update the QoS parameter.
  • the third request is specifically used for:
  • processor 810 is also used for:
  • the third update result is determined by the PCF based on the result of adjusting the PCC rules, the third update result includes the third request being accepted or the third request being rejected .
  • processor 810 is also used for:
  • the model compression, model size and model encoding and decoding in the information of the adjusted application layer model are directly sent to the PCF, and the information of the adjusted application layer model is used to support the PCF Adjust the 5G service quality identifier in the PCC rules, reflective quality of service control, the maximum bit rate in the uplink direction of the transmission AI/ML model, the maximum bit rate in the downlink direction of the transmission AI/ML model, and the transmission AI/ML model The minimum bit rate in the uplink direction, the minimum bit rate in the downlink direction for the transmission of AI/ML models, and the priority of the quality of service flow;
  • processor 810 is also used for:
  • processor 810 is also used for:
  • the model transmission time in the information of the adjusted application layer model is directly sent to the PCF, and the model transmission time in the information of the adjusted application layer model is used to support PCF adjustment
  • the gate status parameter in the PCC rule is used to support the SMF to update the session management strategy according to the transmission start time and transmission end time in the gate status
  • processor 810 is also used for:
  • model transmission state analysis device in the subscription network provided by the present disclosure can realize all the method steps realized by the method embodiments shown in Fig. 3-Fig. 6, and can achieve the same technical effect.
  • the same parts and beneficial effects in this embodiment as those in the method embodiment will be described in detail.
  • Fig. 9 is a schematic structural diagram of a model transmission state analysis device in a subscription network provided by another embodiment of the present disclosure. As shown in Fig. 9, the model transmission state analysis device in a subscription network provided by this embodiment is applied to AF, then this embodiment
  • the device 900 for analyzing model transmission status in the subscription network provided includes:
  • the sending unit 901 is configured to send a first message to the network data analysis function NWDAF directly or through the network capability exposure function NEF; wherein, the first message is used to request to subscribe to the transmission status of the artificial intelligence/machine learning AI/ML model in the network analyze information;
  • the analysis unit 902 is configured to receive the analysis information of the transmission status of the AI/ML model sent by the NWDAF directly or through the NEF, and the second message is used to collect data for analyzing the transmission status of the AI/ML model in the network;
  • the analysis information is used to adjust network policy parameters and/or application layer model information.
  • the parameters requested in the first message include at least one of the following: a network data analysis identifier, an identifier of a user equipment UE or a group of UEs receiving the AI/ML model, or any UE that meets the analysis conditions, using The identification of the application of the AI/ML model, the area where the AI/ML model is transmitted, the network slice indicating the PDU session of the protocol data unit PDU session that transmits the AI/ML model quality of service flow, and the data indicating the PDU session that transmits the AI/ML model quality of service flow Network, 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 used to indicate the transmission of AI/ML model QoS requirements for streams and/or specific QoS requirements for delivery of AI/ML models;
  • the second message includes at least one of the following items: the current location of the UE using the AI/ML model, the identifier of the application using the AI/ML model, the quality of service flow identifier of the transmission AI/ML model, the transmission AI/ML model
  • the bit rate of the uplink direction of the model and the bit rate of the downlink direction of the transmission AI/ML model, the packet delay of the uplink direction of the AI/ML model and the packet delay of the downlink direction of the AI/ML model, the quality of service flow in The number of abnormal releases during the time period of AI/ML model transmission, the number of packet transmissions of AI/ML models, the number of packet retransmissions of AI/ML models, data collection time, the duration of AI/ML model transmission, and the transmission of AI/ML models start timestamp of AI/ML model transfer, size of AI/ML transfer model, network slice of PDU session for transfer of AI/ML model QoS stream, network slice for transfer of AI/ML model QoS stream.
  • the analysis information includes at least one of the following: network slices of PDU sessions used to transmit AI/ML model quality of service flows, identification of applications using AI/ML models, area information using AI/ML models, and analysis results.
  • Valid time user plane function UPF that provides AI/ML model transmission, data network name of PDU session used to transmit AI/ML model quality of service flow, size of AI/ML transmission model, duration of AI/ML model transmission, AI /ML model transmission start timestamp, AI/ML model transmission end timestamp, quality of service requirements, quality of service flow identifier for transmitting AI/ML models, uplink direction bit rate for transmitting AI/ML models and transmitting AI Bit rate in the downlink direction of the /ML model, packet delay in the uplink direction of the AI/ML model, packet delay in the downlink direction of the AI/ML model, and abnormal release of QoS streams during the transmission period of the AI/ML model
  • the parameters requested in the first message also include: federated learning group information, and the federated learning group information includes at least one of the following: the federated learning group used to indicate the analysis ID, UE ID or UE(s) ID participating in federated learning, application ID participating in federated learning;
  • the second message further includes at least one of the following items: the identifier of the federated learning group used for the specified analysis, the UE identifier or UE(s) identifier participating in the federated learning, and the application identifier participating in the federated learning;
  • the analysis information also includes at least one of the following: the identifier of the federated learning group used for the specified analysis, the UE identifier or UE(s) identifier participating in the federated learning, the user indicating to provide the AI/ML model or participating in the federated learning Individual application IDs.
  • the sending unit is also used for:
  • the first request is used to request to update the network policy parameters used for AI/ML model transmission; the network policy parameters are used to optimize the transmission status of the AI/ML model.
  • the sending unit is specifically configured to: transmit the bit rate in the uplink direction of the AI/ML model and the bit rate in the downlink direction of the transmission AI/ML model in the analysis information, and the uplink direction of the AI/ML model Packet delay in the road direction and packet delay in the downlink direction of the AI/ML model, the number of abnormally released QoS streams during the transmission period of the AI/ML model, the number of packet transmissions of the AI/ML model, and the number of packets of the AI/ML model At least one of the number of retransmissions and the number of times the report threshold of the abnormal release of the quality of service stream during the AI/ML model transmission period is reached, and the new quality of service parameters for the transmission AI/ML model are determined; the new quality of service The parameters include at least one of the following: 5G service quality 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, transmission AI The minimum bit rate in the up
  • the identification of the application transmitting the AI/ML model the area information using the AI/ML model, the IP address information of the application service using the AI/ML model, and the quality of service flow used to transmit the AI/ML model in the analysis information
  • the network slice of the PDU session, the data network name of the PDU session used to transmit the AI/ML model quality of service flow determine the area information and address information of the UE(s) and each AF that transmits the AI/ML model; or, if the AI/ML model
  • the ML model performs federated learning, according to the identifier of the federated learning group indicated in the analysis information, the UE identifier or UE(s) identifier participating in the federated learning, and the identifier of each application that provides the AI/ML model or participates in the federated learning, Determine the UE(s) transmitting the AI/ML model and the area information and address information of each AF;
  • the area information and address information of the UE(s) and each AF transmitting the AI/ML model 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 said The UE(s) corresponding to DNAI and the area information and address information of each AF are used to provide a path to optimize the transmission status of the AI/ML model;
  • the sending unit is also specifically used for:
  • the first request is specifically used for:
  • the sending unit is also used for:
  • the first update result sent by the PCF is received directly or through the NEF, where the first update result includes that the first request is accepted or the first request is rejected.
  • the first request is specifically used for:
  • the PCF Requesting the PCF to determine whether the session management function network element SMF needs to update the session management strategy, if it is determined that the SMF needs to update the session management strategy, then determine that the PCF sends a second request to the SMF, and the parameters requested in the second request include at least one of the following : DNAI, traffic steering policy identifier, 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 corresponding DNAI, traffic steering policy identifier, and traffic route information;
  • the receiving unit is also used for:
  • the second update result includes whether the first request is accepted or rejected.
  • the device further includes: a determination unit; the determination unit is used for:
  • the information of the application layer model After receiving the analysis information, adjust the information of the application layer model according to the analysis information, and the information of the application layer model includes at least one of the following: model compression, model size, model transmission time period, model codec ;
  • the information of the application layer model is used to update the quality of service parameters;
  • the new service quality parameters include: 5G service quality identifier, reflective service quality control, uplink direction of transmission AI/ML model Maximum bit rate, maximum bit rate for AI/ML model transmission in downlink direction, minimum bit rate for transmission AI/ML model in uplink direction, minimum bit rate for transmission AI/ML model in downlink direction, quality of service flow priority;
  • the parameters requested in the third request include the new QoS parameter, and the third request is used to request to update the QoS parameter.
  • the third request is specifically used for:
  • the receiving unit is also used for:
  • the third update result is determined by the PCF based on the result of adjusting the PCC rules, the third update result includes the third request being accepted or the third request being rejected .
  • the sending unit is also used for:
  • the model compression, model size and model encoding and decoding in the information of the adjusted application layer model are directly sent to the PCF, and the information of the adjusted application layer model is used to support the PCF Adjust the 5G service quality identifier in the PCC rules, reflective quality of service control, the maximum bit rate in the uplink direction of the transmission AI/ML model, the maximum bit rate in the downlink direction of the transmission AI/ML model, and the transmission AI/ML model The minimum bit rate in the uplink direction, the minimum bit rate in the downlink direction for the transmission of AI/ML models, and the priority of the quality of service flow;
  • the receiving unit is also used for:
  • the sending unit is also used for:
  • the model transmission time in the information of the adjusted application layer model is directly sent to the PCF, and the model transmission time in the information of the adjusted application layer model is used to support PCF adjustment
  • the gate status parameter in the PCC rule is used to support the SMF to update the session management strategy according to the transmission start time and transmission end time in the gate status
  • the receiving unit is also used for:
  • model transmission state analysis device in the subscription network provided by the present disclosure can realize all the method steps realized by the method embodiments in Fig. 3-Fig. 6, and can achieve the same technical effect. Parts and beneficial effects in the embodiment that are the same as those in the method embodiment are described in detail.
  • Fig. 10 is a schematic structural diagram of a model transmission state analysis device in a subscription network provided by another embodiment of the present disclosure.
  • the model transmission state analysis device in a subscription network provided by this embodiment is applied to NWDAF.
  • the apparatus for analyzing the transmission state of models in a subscription network provided in this embodiment includes: a transceiver 1000 configured to receive and send data under the control of a processor 1010 .
  • the bus architecture may include any number of interconnected buses and bridges, specifically one or more processors represented by the processor 1010 and various circuits of the memory represented by the memory 1020 are linked together.
  • the bus architecture can also link together various other circuits such as peripherals, voltage regulators, and power management circuits, etc., which are well known in the art and therefore will not be further described herein.
  • the bus interface provides the interface.
  • Transceiver 1000 may be a plurality of elements, including a transmitter and a receiver, providing a unit for communicating with various other devices over transmission media, including wireless channels, wired channels, optical cables, and other transmission media.
  • 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 when performing operations.
  • the processor 1010 can be a central processing device (CPU), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field-Programmable Gate Array, FPGA) or a complex programmable logic device (Comple8 Programmable Logic Device , CPLD), the processor can also adopt a multi-core architecture.
  • CPU central processing device
  • ASIC Application Specific Integrated Circuit
  • FPGA field programmable gate array
  • CPLD Complex programmable logic device
  • 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; the processor 1010 is used to read the computer programs in the memory and perform the following operations:
  • the first message is used to request to subscribe to the analysis information of the transmission status of the artificial intelligence/machine learning AI/ML model in the network;
  • the second message is used to collect data for analyzing the transmission status of the AI/ML model in the network ;
  • the analysis information is used to adjust network policy parameters and/or application layer model information through AF.
  • the parameters requested in the first message include at least one of the following: a network data analysis identifier, an identifier of a user equipment UE or a group of UEs receiving the AI/ML model, or any UE that meets the analysis conditions, using The identification of the application of the AI/ML model, the area where the AI/ML model is transmitted, the network slice indicating the PDU session of the protocol data unit PDU session that transmits the AI/ML model quality of service flow, and the data indicating the PDU session that transmits the AI/ML model quality of service flow Network, 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 used to indicate the transmission of AI/ML model QoS requirements for streams and/or specific QoS requirements for delivery of AI/ML models;
  • the second message includes at least one of the following items: the current location of the UE using the AI/ML model, the identifier of the application using the AI/ML model, the quality of service flow identifier of the transmission AI/ML model, the transmission AI/ML model
  • the bit rate of the uplink direction of the model and the bit rate of the downlink direction of the transmission AI/ML model, the packet delay of the uplink direction of the AI/ML model and the packet delay of the downlink direction of the AI/ML model, the quality of service flow in The number of abnormal releases during the time period of AI/ML model transmission, the number of packet transmissions of AI/ML models, the number of packet retransmissions of AI/ML models, data collection time, the duration of AI/ML model transmission, and the transmission of AI/ML models start timestamp of AI/ML model transfer, size of AI/ML transfer model, network slice of PDU session for transfer of AI/ML model QoS stream, network slice for transfer of AI/ML model QoS stream.
  • the analysis information includes at least one of the following: network slices of PDU sessions used to transmit AI/ML model quality of service flows, identification of applications using AI/ML models, area information using AI/ML models, and analysis results.
  • Valid time user plane function UPF that provides AI/ML model transmission, data network name of PDU session used to transmit AI/ML model quality of service flow, size of AI/ML transmission model, duration of AI/ML model transmission, AI /ML model transmission start timestamp, AI/ML model transmission end timestamp, quality of service requirements, quality of service flow identifier for transmitting AI/ML models, uplink direction bit rate for transmitting AI/ML models and transmitting AI Bit rate in the downlink direction of the /ML model, packet delay in the uplink direction of the AI/ML model, packet delay in the downlink direction of the AI/ML model, and abnormal release of QoS streams during the transmission period of the AI/ML model
  • the parameters requested in the first message also include: federated learning group information, and the federated learning group information includes at least one of the following: the federated learning group used to indicate the analysis ID, UE ID or UE(s) ID participating in federated learning, application ID participating in federated learning;
  • the second message further includes at least one of the following items: the identifier of the federated learning group used for the specified analysis, the UE identifier or UE(s) identifier participating in the federated learning, and the application identifier participating in the federated learning;
  • the analysis information also includes at least one of the following: the identifier of the federated learning group used for the specified analysis, the UE identifier or UE(s) identifier participating in the federated learning, the user indicating to provide the AI/ML model or participating in the federated learning Individual application IDs.
  • model transmission state analysis device in the subscription network provided by the present disclosure can realize all the method steps realized by the method embodiments shown in Fig. 4 and Fig. 7 , and can achieve the same technical effect.
  • the same parts and beneficial effects in this embodiment as those in the method embodiment will be described in detail.
  • Figure 11 is a schematic structural diagram of a device for analyzing model transmission status in a subscription network provided by another embodiment of the present disclosure.
  • the apparatus 1100 for analyzing the model transmission state in the subscription network provided by the example includes:
  • the receiving unit 1101 is configured to receive the first message sent by the application function AF directly or through the network capability exposure function NEF; wherein, the first message is used to request to subscribe to the analysis of the transmission status of the artificial intelligence/machine learning AI/ML model in the network information;
  • the sending unit 1102 is configured 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, and the second message is used to collect and analyze AI/ ML model transfer state data;
  • the analysis unit 1103 is also used to receive the data of the transmission state of the AI/ML model sent by other network functions 5GC NF(s) of the 5G core network, and analyze the data of the transmission state of the AI/ML model to obtain the transmission state of the AI/ML model Status analysis information;
  • the analysis information is used to adjust network policy parameters and/or application layer model information through AF.
  • the parameters requested in the first message include at least one of the following: a network data analysis identifier, an identifier of a user equipment UE or a group of UEs receiving the AI/ML model, or any UE that meets the analysis conditions, using The identification of the application of the AI/ML model, the area where the AI/ML model is transmitted, the network slice indicating the PDU session of the protocol data unit PDU session that transmits the AI/ML model quality of service flow, and the data indicating the PDU session that transmits the AI/ML model quality of service flow Network, 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 used to indicate the transmission of AI/ML model QoS requirements for streams and/or specific QoS requirements for delivery of AI/ML models;
  • the second message includes at least one of the following items: the current location of the UE using the AI/ML model, the identifier of the application using the AI/ML model, the quality of service flow identifier of the transmission AI/ML model, the transmission AI/ML model
  • the bit rate of the uplink direction of the model and the bit rate of the downlink direction of the transmission AI/ML model, the packet delay of the uplink direction of the AI/ML model and the packet delay of the downlink direction of the AI/ML model, the quality of service flow in The number of abnormal releases during the time period of AI/ML model transmission, the number of packet transmissions of AI/ML models, the number of packet retransmissions of AI/ML models, data collection time, the duration of AI/ML model transmission, and the transmission of AI/ML models start timestamp of AI/ML model transfer, size of AI/ML transfer model, network slice of PDU session for transfer of AI/ML model QoS stream, network slice for transfer of AI/ML model QoS stream.
  • the analysis information includes at least one of the following: network slices of PDU sessions used to transmit AI/ML model quality of service flows, identification of applications using AI/ML models, area information using AI/ML models, and analysis results.
  • Valid time user plane function UPF that provides AI/ML model transmission
  • data network name of PDU session used to transmit AI/ML model quality of service flow size of AI/ML transmission model, duration of AI/ML model transmission
  • AI /ML model transmission start timestamp AI/ML model transmission end timestamp
  • quality of service flow identifier for transmission AI/ML model
  • uplink direction bit rate for transmission AI/ML model and time stamp for transmission AI/ML model Bit rate in the downlink direction packet delay in the uplink direction of the AI/ML model and packet delay in the downlink direction of the AI/ML model, the number of abnormally released QoS streams during the transmission period of the AI/ML model, and the The number of reporting thresholds for abnormal release of quality of service flows during the AI/ML model transmission period, the number of packet transmissions for the AI/ML model, and
  • the parameters requested in the first message also include: federated learning group information, and the federated learning group information includes at least one of the following: the federated learning group used to indicate the analysis ID, UE ID or UE(s) ID participating in federated learning, application ID participating in federated learning;
  • the second message further includes at least one of the following items: the identity of the federated learning group used to indicate the analysis, the UE identity or UE(s) identity participating in the federated learning, and the application identity participating in the federated learning;
  • the analysis information also includes at least one of the following: the identifier of the federated learning group used for the specified analysis, the UE identifier or UE(s) identifier participating in the federated learning, the user indicating to provide the AI/ML model or participating in the federated learning Individual application IDs.
  • model transmission state analysis device in the subscription network provided by the present disclosure can realize all the method steps realized by the method embodiments in Fig. 4 and Fig. 7, and can achieve the same technical effect. Parts and beneficial effects in the embodiment that are the same as those in the method embodiment are described in detail.
  • each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, each unit may exist separately physically, or two or more units may be integrated into one unit.
  • the above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.
  • An integrated unit may be stored in a processor-readable storage medium if it is realized in the form of a software function unit and sold or used as an independent product.
  • the technical solution of the present disclosure is essentially or part of the contribution to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , including several instructions for enabling a computer device (which may be a personal computer, server, or network device, etc.) or a processor (processor) to execute all or part of the steps of the methods in various embodiments of the present disclosure.
  • the aforementioned storage media include: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disk or optical disc and other media that can store program codes. .
  • the embodiment of the present disclosure also provides a processor-readable storage medium.
  • the processor-readable storage medium stores a computer program, and the computer program is used to cause a processor to execute any one of the above method embodiments.
  • 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 (such as floppy disk, hard disk, magnetic tape, magneto-optical disk (MO), etc.), optical storage (such as CD, DVD, BD, HVD, etc.), and semiconductor memory (such as ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)), etc.
  • magnetic storage such as floppy disk, hard disk, magnetic tape, magneto-optical disk (MO), etc.
  • optical storage such as CD, DVD, BD, HVD, etc.
  • semiconductor memory such as ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)
  • the embodiments of the present disclosure may be provided as methods, systems, or computer program products. Accordingly, the present disclosure can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present disclosure may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, optical storage, etc.) having computer-usable program code embodied therein.
  • processor-executable instructions may also be stored in a processor-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, such that the instructions stored in the processor-readable memory produce a manufacturing product, the instruction device realizes the functions specified in one or more procedures of the flow chart and/or one or more blocks of the block diagram.
  • processor-executable instructions 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
  • the executed instructions provide steps for implementing the functions specified in the procedure or procedures of the flowchart and/or the block or blocks of the block diagrams.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Databases & Information Systems (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

本公开提供一种订阅网络中模型传输状态分析方法、装置及可读存储介质，该方法包括：应用功能AF直接或通过网络能力开放功能NEF向网络数据分析功能NWDAF发送第一消息；AF直接或通过所述NEF接收所述NWDAF发送的AI/ML模型传输状态的分析信息，所述分析信息是所述NWDAF根据接收其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据确定的；其中，所述分析信息用于调整网络策略参数和/或应用层模型的信息。本公开能够有效地对AI/ML模型传输状态进行分析，进而使得网络基于AI/ML模型传输状态有效地调整网络传输策略以及第三方获得AI/ML模型传输状态的分析进行应用层信息的调整。

Description

订阅网络中模型传输状态分析方法、装置及可读存储介质
本公开要求于2021年11月24日提交中国专利局、申请号为202111407051.2、申请名称为“订阅网络中模型传输状态分析方法、装置及可读存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本公开中。
技术领域
本公开涉及通信技术领域，尤其涉及一种订阅网络中模型传输状态分析方法、装置及可读存储介质。
背景技术
近年来，由于人工智能的技术突破，人工智能的应用越来越广泛。对于移动终端，由于其具有严格的能耗、计算和内存成本限制，无法在终端运行重量级的人工智能(英文为：Artificial Intelligence，简称为：AI)/机器学习(英文为：Machine Learning，简称为：ML)模型(以下称为AI/ML模型)，因此，目前采用的手段是将许多AI/ML模型的推理从移动终端传输到云或者其他终端，即需要把AI/ML模型传输给云或者其他终端。
此外，由于考虑到对传输数据的隐私保护以及缓解网络传输数据的压力等问题，对AI/ML模型的传输要求也越来越高。其中，5G系统作为传输AI/ML模型的通道，为了提高5G网络的智能化能力，满足SA1#93e通过的TS 22.261中AI/ML模型在5G系统中传输的要求，5G系统需要支持将关于AI-ML会话的监测和状态信息暴露给第三方。
但是，现有技术中，无法有效地对AI/ML模型传输状态进行分析，进而使得网络无法基于AI/ML模型传输状态有效地调整网络传输策略以及第三方无法获得AI/ML模型传输状态的分析进行应用层信息的调整。
发明内容
本公开提供一种订阅网络中模型传输状态分析方法、装置及可读存储介 质，解决了无法有效地对AI/ML模型传输状态进行分析，进而使得网络无法基于AI/ML模型传输状态有效地调整网络传输策略和第三方无法获得AI/ML模型传输状态的分析进行应用层信息的调整的技术问题。
第一方面，本公开提供一种订阅网络中模型传输状态分析方法，所述方法应用于应用功能AF，所述方法包括：
直接或通过网络能力开放功能NEF向网络数据分析功能NWDAF发送第一消息；其中，所述第一消息用于请求订阅网络中人工智能/机器学习AI/ML模型传输状态的分析信息；
直接或通过所述NEF接收所述NWDAF发送的AI/ML模型传输状态的分析信息，所述分析信息是所述NWDAF根据接收5G核心网的其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据确定的；
其中，所述分析信息用于调整网络策略参数和/或应用层模型的信息。
可选地，所述AI/ML模型传输状态的数据是所述NWDAF根据接收到的所述第一消息中请求的参数通过向所述5GC NF(s)发送第二消息得到的，所述第二消息用于采集用于分析网络中AI/ML模型传输状态的数据。
可选地，所述第一消息中请求的参数包括下述至少一项：网络数据分析标识、接收AI/ML模型的一个用户设备UE或一组UE的标识或满足分析条件的任意UE、使用AI/ML模型的应用的标识、AI/ML模型传输的区域、指示传输AI/ML模型服务质量流的协议数据单元PDU会话的网络切片、指示传输AI/ML模型服务质量流的PDU会话的数据网络、AI/ML模型传输的时间段、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于指示传输AI/ML模型的服务质量流的服务质量要求和/或用于指示传输AI/ML模型的特定的服务质量要求；
所述第二消息中包括下述至少一项：使用AI/ML模型的UE的当前位置、使用AI/ML模型的应用的标识、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量、数据采集时间、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、 AI/ML传输模型的大小、用于传输AI/ML模型服务质量流的PDU会话的网络切片、用于传输AI/ML模型服务质量流的PDU会话的数据网络、用于所述AF的服务流程；
所述分析信息包括下述至少一项：用于传输AI/ML模型服务质量流的PDU会话的网络切片、使用AI/ML模型的应用的标识、使用AI/ML模型的区域信息、分析结果的有效时间、提供AI/ML模型传输的用户面功能UPF、用于传输AI/ML模型服务质量流的PDU会话的数据网络名称、AI/ML传输模型的大小、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、达到服务质量流在AI/ML模型传输的时间段内异常释放的报告阈值的次数、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量；
其中，若AI/ML模型执行联邦学习，所述第一消息中请求的参数还包括：联邦学习群信息，所述联邦学习群信息包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
相应的，所述第二消息中还包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
相应的，所述分析信息还包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、指示提供AI/ML模型或者参与联邦学习的各个应用标识。
本公开实施例中，通过向NWDAF发送订阅网络中AI/ML模型传输状态的分析信息的请求，接收NWDAF发送的根据采集到的5GC NF(s)数据确定的分析信息，根据分析信息来调整网络策略参数和/或应用层模型的信息，实现了有效地对AI/ML模型传输状态进行分析，进而使得网络能够基于AI/ML模型传输状态有效地调整网络传输策略以及使得第三方获得AI/ML模型传输状态的分析进行应用层信息的调整。
可选地，在接收到所述分析信息后，所述方法还包括：
根据所述分析信息，直接或通过所述NEF向策略控制功能PCF发送第一请求；
其中，所述第一请求用于请求更新用于AI/ML模型传输的网络策略参数；所述网络策略参数用于优化AI/ML模型传输状态。
可选地，所述根据所述分析信息，直接或通过所述NEF向策略控制功能PCF发送第一请求，包括：
根据所述分析信息中传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量、达到服务质量流在AI/ML模型传输的时间段内异常释放的报告阈值的次数中的至少一项，确定传输AI/ML模型的新的服务质量参数；所述新的服务质量参数包括下述至少一项：5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级；
根据所述分析信息中的传输AI/ML模型的应用的标识、使用AI/ML模型的区域信息、使用AI/ML模型的应用服务的IP地址信息、用于传输AI/ML模型服务质量流的PDU会话的网络切片、用于传输AI/ML模型服务质量流的PDU会话的数据网络名称，确定传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息；或者，若AI/ML模型执行联邦学习，根据所述分析信息中的指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、指示提供AI/ML模型或者参与联邦学习的各个应用标识，确定传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息；
根据传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息，确定数据网络接入标识DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息，所述DNAI对应的UE(s)和各个AF的区域信息及地址信息用于提供优化AI/ML模型传输状态的路径；
将所述新的服务质量参数、所述DNAI以及所述DNAI对应的UE(s)和各 个AF的区域信息及地址信息作为所述第一请求中的参数直接或通过所述NEF发送给PCF。
可选地，所述根据传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息，确定数据网络接入标识DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息，包括：
根据传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息，判断当前路由路径是否不佳；
若当前路由路径不佳，则根据传输AI/ML模型的UE(s)和各个AF的地址信息以及UE(s)的区域信息，确定传输AI/ML模型中传输双方的目的地址；
根据所述目的地址，确定最近的路径；
根据最近的路径，确定DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息。
可选地，所述第一请求，具体用于：
请求PCF根据所述新的服务质量参数调整PCC规则中的5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级并指示所述PCF将调整后的第一更新结果直接或通过NEF反馈；其中，所述第一更新结果是PCF根据所述新的服务质量参数调整PCC规则的结果确定的；
相应的，所述方法还包括：
直接或通过NEF接收PCF发送的所述第一更新结果，所述第一更新结果包括所述第一请求被接受或所述第一请求被拒绝。
可选地，所述第一请求具体用于：
请求PCF确定会话管理功能网元SMF是否需要更新会话管理策略，若确定SMF需要更新会话管理策略，则确定PCF向SMF发送第二请求，所述第二请求中请求的参数包括下述至少一项：DNAI、流量导向政策标识符、流量路线信息；所述第二请求用于SMF根据新的会话管理策略确定选择的用户面功能UPF并提供相应的DNAI、流量导向政策标识符、流量路线信息；
相应的，所述方法还包括：
直接或通过NEF接收PCF发送的第二更新结果，所述第二更新结果是PCF根据SMF发送的新的会话管理策略是否更新UPF路径确定的；
其中，所述第二更新结果包括所述第一请求被接受或被拒绝。
本公开实施例中，根据分析信息，确定新的服务质量参数，并将新的服务质量参数发送给PCF，使得PCF依据新的服务质量参数相应调整PCC规则或通过SMF更新SM策略并提供DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息，通过接收PCF发送的用于表示第一请求被接受还是被拒绝的通知，能够实现基于NWDAF分析结果，请求调整网络策略从而优化AI/ML模型传输状态。
可选地，在接收到所述分析信息后，所述方法还包括：
根据所述分析信息，调整应用层模型的信息，所述应用层模型的信息包括下述至少一项：模型压缩、模型大小、模型传输时间段、模型编解码；所述应用层模型的信息用于更新服务质量参数；
根据调整后的应用层模型的信息，确定新的服务质量参数，所述新的服务质量参数包括：5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级；
直接或通过所述NEF向策略控制功能PCF发送第三请求；
其中，所述第三请求中请求的参数包括所述新的服务质量参数，所述第三请求用于请求更新服务质量参数。
可选地，所述第三请求，具体用于：
请求PCF根据所述新的服务质量参数调整PCC规则中的5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、优先级；
相应的，所述方法还包括：
直接或通过NEF接收PCF发送的第三更新结果，第三更新结果是PCF基于调整PCC规则的结果确定的，所述第三更新结果包括所述第三请求被接受或所述第三请求被拒绝。
可选地，在所述调整应用层模型的信息之后，所述方法还包括：
将调整后的应用层模型的信息中的模型压缩、模型大小以及模型编解码直接发送给PCF，所述调整后的应用层模型的信息用于支持PCF调整PCC规则中的5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级；
相应的，所述方法还包括：
接收PCF发送的第四更新结果，第四更新结果是PCF根据调整后的应用层模型的信息调整PCC规则的结果确定的，所述第四更新结果包括所述第三请求被接受或所述第三请求被拒绝。
可选地，在所述调整应用层模型的信息之后，所述方法还包括：
将调整后的应用层模型的信息中的模型传输时间直接发送给PCF，所述调整后的应用层模型的信息中的模型传输时间用于支持PCF调整PCC规则中的门状态参数；所述门状态参数用于支持SMF根据门状态中的传输开始时间和传输结束时间来更新会话管理策略；
相应的，所述方法还包括：
接收PCF发送的第五更新结果，第五更新结果是PCF通过接收SMF发送的新的会话管理策略的结果确定的，所述第五更新结果包括所述第三请求被接受或所述第三请求被拒绝。
本公开实施例中，根据分析信息，调整应用层模型的信息，依据调整后的应用层模型的信息，确定新的服务质量参数并将新的服务质量参数发送给PCF，将使得PCF依据新的服务质量参数相应调整PCC规则；或者，直接将调整后的应用层模型的信息发送给PCF，使得PCF依据模型压缩、模型大小、模型编解码调整PCC规则中上述的服务质量参数，或者，使得PCF依据模型传输时间调整PCC规则中的门状态参数，使得SMF根据门状态中的传输开始时间和传输结束时间来更新会话管理策略，通过接收PCF发送的用于表示第三请求被接受还是被拒绝的通知，能够实现基于NWDAF分析结果，调整应用层模型的信息，进而更新QoS要求，从而优化AI/ML模型传输状态。
第二方面，本公开提供一种订阅网络中模型传输状态分析方法，所述方 法应用于网络数据分析功能NWDAF，所述方法包括：
直接或通过网络能力开放功能NEF接收应用功能AF发送的第一消息；其中，所述第一消息用于请求订阅网络中人工智能/机器学习AI/ML模型传输状态的分析信息；
根据所述第一消息中请求的参数，向5G核心网的其他网络功能5GC NF(s)发送第二消息，所述第二消息用于采集用于分析网络中AI/ML模型传输状态的数据；
接收5G核心网的其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据，并对AI/ML模型传输状态的数据进行分析，得到AI/ML模型传输状态的分析信息；
其中，所述分析信息用于通过AF调整网络策略参数和/或应用层模型的信息。
可选地，所述第一消息中请求的参数包括下述至少一项：网络数据分析标识、接收AI/ML模型的一个用户设备UE或一组UE的标识或满足分析条件的任意UE、使用AI/ML模型的应用的标识、AI/ML模型传输的区域、指示传输AI/ML模型服务质量流的协议数据单元PDU会话的网络切片、指示传输AI/ML模型服务质量流的PDU会话的数据网络、AI/ML模型传输的时间段、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于指示传输AI/ML模型的服务质量流的服务质量要求和/或用于指示传输AI/ML模型的特定的服务质量要求；
所述第二消息中包括下述至少一项：使用AI/ML模型的UE的当前位置、使用AI/ML模型的应用的标识、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量、数据采集时间、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于传输AI/ML模型服务质量流的PDU会话的网络切片、用于传输AI/ML模型服务质量流的PDU会话的数据网络、用于所述AF的服务流程；
所述分析信息包括下述至少一项：用于传输AI/ML模型服务质量流的PDU会话的网络切片、使用AI/ML模型的应用的标识、使用AI/ML模型的区域信息、分析结果的有效时间、提供AI/ML模型传输的用户面功能UPF、用于传输AI/ML模型服务质量流的PDU会话的数据网络名称、AI/ML传输模型的大小、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、达到服务质量流在AI/ML模型传输的时间段内异常释放的报告阈值的次数、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量；
其中，若AI/ML模型执行联邦学习，所述第一消息中请求的参数还包括：联邦学习群信息，所述联邦学习群信息包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
相应的，所述第二消息中还包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
相应的，所述分析信息还包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、指示提供AI/ML模型或者参与联邦学习的各个应用标识。
第三方面，本公开提供一种订阅网络中模型传输状态分析装置，所述装置包括：存储器，收发机，处理器：
存储器，用于存储计算机程序；收发机，用于在所述处理器的控制下收发数据；处理器，用于读取所述存储器中的计算机程序并执行以下操作：
直接或通过网络能力开放功能NEF向网络数据分析功能NWDAF发送第一消息；其中，所述第一消息用于请求订阅网络中人工智能/机器学习AI/ML模型传输状态的分析信息；
直接或通过所述NEF接收所述NWDAF发送的AI/ML模型传输状态的分析信息，所述分析信息是所述NWDAF根据接收5G核心网的其他网络功 能5GC NF(s)发送的AI/ML模型传输状态的数据确定的；
其中，所述分析信息用于调整网络策略参数和/或应用层模型的信息。
第四方面，本公开提供一种订阅网络中模型传输状态分析装置，所述装置包括存储器，收发机，处理器：
存储器，用于存储计算机程序；收发机，用于在所述处理器的控制下收发数据；处理器，用于读取所述存储器中的计算机程序并执行以下操作：
直接或通过网络能力开放功能NEF接收应用功能AF发送的第一消息；其中，所述第一消息用于请求订阅网络中人工智能/机器学习AI/ML模型传输状态的分析信息；
根据所述第一消息中请求的参数，向5G核心网的其他网络功能5GC NF(s)发送第二消息，所述第二消息用于采集用于分析网络中AI/ML模型传输状态的数据；
接收5G核心网的其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据，并对AI/ML模型传输状态的数据进行分析，得到AI/ML模型传输状态的分析信息；
其中，所述分析信息用于通过AF调整网络策略参数和/或应用层模型的信息。
第五方面，本公开提供一种订阅网络中模型传输状态分析装置，所述装置包括：
发送单元，用于直接或通过网络能力开放功能NEF向网络数据分析功能NWDAF发送第一消息；其中，所述第一消息用于请求订阅网络中人工智能/机器学习AI/ML模型传输状态的分析信息；
分析单元，用于直接或通过所述NEF接收所述NWDAF发送的AI/ML模型传输状态的分析信息，所述分析信息是所述NWDAF根据接收5G核心网的其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据确定的；
其中，所述分析信息用于调整网络策略参数和/或应用层模型的信息。
第六方面，本公开提供一种订阅网络中模型传输状态分析装置，所述装置包括：
接收单元，用于直接或通过网络能力开放功能NEF接收应用功能AF发送的第一消息；其中，所述第一消息用于请求订阅网络中人工智能/机器学习 AI/ML模型传输状态的分析信息；
发送单元，用于根据所述第一消息中请求的参数，向5G核心网的其他网络功能5GC NF(s)发送第二消息，所述第二消息用于采集用于分析网络中AI/ML模型传输状态的数据；
分析单元，用于接收5G核心网的其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据，并对AI/ML模型传输状态的数据进行分析，得到AI/ML模型传输状态的分析信息；
其中，所述分析信息用于通过AF调整网络策略参数和/或应用层模型的信息。
第七方面，本公开提供一种处理器可读存储介质，所述处理器可读存储介质存储有计算机程序，所述计算机程序用于使所述处理器执行第一方面或第二方面任一项所述的方法。
本公开提供一种订阅网络中模型传输状态分析方法、装置及可读存储介质，直接或通过网络能力开放功能NEF向网络数据分析功能NWDAF发送第一消息；其中，所述第一消息用于请求订阅网络中人工智能/机器学习AI/ML模型传输状态的分析信息；直接或通过所述NEF接收所述NWDAF发送的AI/ML模型传输状态的分析信息，所述分析信息是所述NWDAF根据接收5G核心网的其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据确定的，所述AI/ML模型传输状态的数据是所述NWDAF根据接收到的所述第一消息中请求的参数通过向所述5GC NF(s)发送第二消息得到的，所述第二消息用于采集用于分析网络中AI/ML模型传输状态的数据；其中，所述分析信息用于调整网络策略参数和/或应用层模型的信息。通过向NWDAF发送订阅网络中AI/ML模型传输状态的分析信息的请求，接收NWDAF发送的根据采集到的5GC NF(s)数据确定的分析信息，根据分析信息来调整网络策略参数和/或应用层模型的信息，实现了有效地对AI/ML模型传输状态进行分析，进而使得网络能够基于AI/ML模型传输状态有效地调整网络传输策略以及使得第三方获得AI/ML模型传输状态的分析进行应用层信息的调整。
应当理解，上述发明内容部分中所描述的内容并非旨在限定本发明的实施例的关键或重要特征，亦非用于限制本发明的范围。本发明的其它特征将通过以下的描述变得容易理解。
附图说明
为了更清楚地说明本公开或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作一简单地介绍，显而易见地，下面描述中的附图是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1为本公开实施例提供的订阅网络中模型传输状态分析方法的网络架构图；
图2为本公开实施例提供的支持网络数据分析的5GC的网络架构图；
图3为本公开实施例一提供的订阅网络中模型传输状态分析方法的流程示意图；
图4为本公开实施例一提供的订阅网络中模型传输状态分析方法的信令流程示意图；
图5为本公开实施例二提供的订阅网络中模型传输状态分析方法的信令流程示意图；
图6为本公开实施例三提供的订阅网络中模型传输状态分析方法的信令流程示意图；
图7为本公开实施例四提供的订阅网络中模型传输状态分析方法的流程示意图；
图8为本公开实施例提供的订阅网络中模型传输状态分析装置的结构示意图；
图9为本公开另一实施例提供的订阅网络中模型传输状态分析装置的结构示意图；
图10为本公开再一实施例提供的订阅网络中模型传输状态分析装置的结构示意图；
图11为本公开又一实施例提供的订阅网络中模型传输状态分析装置的结构示意图。
具体实施方式
本公开中术语“和/或”，描述关联对象的关联关系，表示可以存在三种 关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。字符“/”一般表示前后关联对象是一种“或”的关系。
下面将结合本公开实施例中的附图，对本公开实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本公开一部分实施例，并不是全部的实施例。基于本公开中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本公开保护的范围。
为了清楚理解本公开的技术方案，首先对现有技术的方案进行详细介绍。现有技术中，在SA#93e通过的SA1R18需求中，需要AI/ML模型传输的场景至少存在以下几种情况：
场景1、AI/ML模型的分发和共享。由于任务或者环境等的变化，移动终端的内存等有限，不能预先机载所有的模型，因此需要移动终端通过5G系统从网络实时下载新的AI/ML模型。
场景2、通过5GS的联邦学习算法。云服务器训练一个全局的模型时，需要聚合各个终端设备在本地训练的模型。每次训练迭代过程：一个终端设备从云服务器下载一个全局模型，用本地数据进行训练；终端将中间训练结果上报给云服务器；云服务器聚合来自所有终端的中间训练结果并更新全局模型，然后把全局模型再分发给终端；终端再执行下一次迭代。
场景3、AI/ML端点之间的AI/ML模型切分。一个AI/ML模型可以基于当前的任务或者环境切分成多个部分。趋势是将计算复杂，能耗大的部分由网络进行推理，需要隐私保护或者时延敏感的部分在终端推理。例如，终端下载/机载一个模型，先推理特定几层/部分，然后把中间结果发送给网络；网络再执行剩余层/部分，然后把推理结果反馈给终端。该场景在第一步或者中间传输部分模型，因此可能包含模型的传输。
因此，5G系统作为传输AI/ML模型的通道，为了提高5G网络的智能化能力，满足SA1#93e通过的TS 22.261中AI/ML模型在5G系统中传输的要求，5G系统需要支持将关于AI-ML会话的监测和状态信息暴露给第三方，但是，目前，没有对AI/ML模型传输状态的分析，第三方也不能基于AI/ML模型传输状态有效地调整自身行为以及网络无法基于AI/ML模型传输状态有效地调整网络状态。
发明人进一步研究发现，要对AI/ML模型传输状态进行有效地分析，需 要应用功能(英文为：Application Function，简称为：AF)、网络能力开放功能(英文为：Network Exposure Function，简称为：NEF)、网络数据分析功能(英文为：Network data analytic function，简称为：NWDAF)以及各个网络功能(英文为：Network Function，简称为：NF)之间进行交互。如图1所示，AF可以直接或通过NEF向NWDAF发送用于表示订阅网络中AI/ML模型传输状态分析的请求，NWDAF通过向5G核心网(英文为：5G Core Network，简称为：5GC)中各个网络功能(英文为：Network Function，简称为：NF)(即NF(s))采集数据来分析网络中AI/ML模型传输状态并反馈，能够实现对AI/ML模型传输状态有效地分析，进而使得网络能够基于AI/ML模型传输状态有效地调整网络状态和第三方能够获得AI/ML模型传输状态的分析进行自身行为数据的调整。
所以基于上述发明人的创造性研究，提出了本公开提出的订阅网络中模型传输状态分析方法，本公开中，直接或通过网络能力开放功能NEF向网络数据分析功能NWDAF发送第一消息；其中，所述第一消息用于请求订阅网络中人工智能/机器学习AI/ML模型传输状态的分析信息；直接或通过所述NEF接收所述NWDAF发送的AI/ML模型传输状态的分析信息，所述分析信息是所述NWDAF根据接收5G核心网的其他网络功能(即5GC NF(s))发送的AI/ML模型传输状态的数据确定的，所述AI/ML模型传输状态的数据是所述NWDAF根据接收到的所述第一消息中请求的参数通过向所述5GC NF(s)发送第二消息得到的，所述第二消息用于采集用于分析网络中AI/ML模型传输状态的数据；其中，所述分析信息用于调整网络策略参数和/或应用层模型的信息。通过向NWDAF发送订阅网络中AI/ML模型传输状态的分析信息的请求，接收NWDAF发送的根据采集到的5GC NF(s)数据确定的分析信息，根据分析信息来调整网络策略参数和/或应用层模型的信息，实现了有效地对AI/ML模型传输状态进行分析，进而使得网络能够基于AI/ML模型传输状态有效地调整网络传输策略以及使得第三方获得AI/ML模型传输状态的分析进行应用层信息的调整。
图2为本公开实施例提供的支持网络数据分析的5GC的网络架构图，如图2所示，本公开实施例中，NWDAF是运营商管理的网络分析功能，NWDAF能够向5GC网络功能、应用功能(英文为：Application Function，简称为： AF)和操作管理维护(英文为：Operation Administration and Maintenance，简称为：OAM)提供数据分析服务。其中，分析结果可以是历史统计信息或者预测信息。NWDAF可以服务一个或多个网络切片。
其中，在5GC中还包括其他多种功能。分别为用户平面功能(英文为：User Plane Function，简称为：UPF)、会话管理功能(英文为：Session Management Function，简称为：SMF)、接入和移动性管理功能(英文为：Access and Mobility Management Function，简称为：AMF)、统一数据库(英文为：Unified Data Repository，简称为：UDR)、网络能力开放功能(英文为：Network Exposure Function，简称为：NEF)、AF、策略控制功能(英文为：Policy Control Function，简称为：PCF)及在线计费系统(英文为：Online Charging System，简称为：OCS)。其中，这些其他功能均可统称为NF。NWDAF基于服务化接口与5G核心网中其他功能实体5GC NF(s)及OAM进行通信。
5GC中可以有不同NWDAF实例提供不同类型的专用分析。为了让消费者NF能够发现合适的NWDAF实例来提供特定类型的分析，NWDAF实例需在向网络数据库功能(英文为：Network Repository Function，简称为：NRF)注册时提供其支持Analytic ID，Analytic ID代表了分析类型(或分析标识)。这样消费者NF可以在向NRF查询NWDAF实例时，提供Analytic ID来指示需要何种类型的分析。5GC网络功能和OAM决定如何使用网络数据分析功能NWDAF提供的数据分析来提高网络性能。
本公开实施例中，在一种应用场景中，AF请求NWDAF提供AI/ML模型传输状态分析，分析结果(或分析信息)包含：使用AI/ML模型的应用标识(即Application ID)、使用AI/ML模型的区域信息、传输AI/ML模型的时间段、所传输模型的大小、传输AI/ML模型的服务质量(即QoS)相关信息、AI/ML模型传输使用的网络切片、数据网络名称(英文为：Data Network Name，简称为：DNN)信息；如果有联邦学习，还包括：组标识(即联邦学习group ID)、参与联邦学习的UE ID or UE group ID、提供模型或者参与联邦学习的应用服务器的地址信息。AF根据NWDAF提供的数据分析，请求调整5GS的网络策略或者调整应用层AI/ML模型参数来优化AI/ML模型传输状态，5GC NF(s)基于请求调整网络策略或者AF基于数据分析调整应用层AI/ML模型参数。
其中，AF向NWDAF发送分析请求时，若AF在受信区，AF可以直接向NWDAF发送请求；若AF在受信区，AF可以通过NEF向NWDAF发送请求，即AF向NEF发送请求，再由NEF向NWDAF发送请求。
因此，直接或通过NEF向NWDAF发送订阅网络中AI/ML模型传输状态的分析信息的请求，接收NWDAF发送的根据采集到的5GC NF(s)数据确定的分析信息，根据分析信息来调整网络策略参数和/或应用层模型的信息，实现了有效地对AI/ML模型传输状态进行分析，进而使得网络能够基于AI/ML模型传输状态有效地调整网络传输策略以及使得第三方获得AI/ML模型传输状态的分析进行应用层信息的调整。
以下将参照附图来描述本公开的实施例。
图3为本公开实施例一提供的订阅网络中模型传输状态分析方法的流程示意图，如图3所示，本实施例提供的订阅网络中模型传输状态分析方法的执行主体为AF，则本公开实施例提供的订阅网络中模型传输状态分析方法包括以下步骤：
步骤101、AF直接或通过网络能力开放功能NEF向网络数据分析功能NWDAF发送第一消息。
其中，所述第一消息用于请求订阅网络中人工智能/机器学习AI/ML模型传输状态的分析信息。
本实施例中，第一消息中请求的参数包括下述至少一项：网络数据分析标识(即Analytics ID)、接收AI/ML模型的一个用户设备UE或一组UE的标识或满足分析条件的任意UE(即Target of Analytics Reporting)、使用AI/ML模型的应用的标识(即Application ID)、AI/ML模型传输的区域(即AoI(Area of Interest))、指示传输AI/ML模型服务质量流的协议数据单元PDU会话的网络切片(即S-NSSAI)、指示传输AI/ML模型服务质量流的PDU会话的数据网络(即DNN)、AI/ML模型传输的时间段(即Model transmission duration)、AI/ML模型传输的开始时间戳(即Model transmission start)、AI/ML模型传输的结束时间戳(即Model transmission stop)、AI/ML传输模型的大小(即Model size)、用于指示传输AI/ML模型的服务质量流的服务质量要求(即5QI(5G QoS Identifier))和/或用于指示传输AI/ML模型的特定的服务质量要求(即QoS Characteristics)。
其中，特定的服务质量要求，例如包传输时延、误包率等。参见下述表1所示的第一消息中请求的参数示例表。
表1：第一消息中请求的参数示例表
Figure PCTCN2022127034-appb-000001
如果存在联邦学习，第一消息中请求的参数还可以包括：联邦学习群信息(即Federated Learning(FL)group information)；联邦学习群信息包括下述 至少一项：用于指示分析的联邦学习组的标识(即Federated Learning(FL)group ID)、参加联邦学习的UE标识或UE(s)标识(即Federated Learning(FL)UE ID or UE group ID)、参加联邦学习的应用标识(即Federated Learning(FL)Application ID)。参见下述表2所示的第一消息中请求的参数示例表。
表2：第一消息中请求的参数示例表
Figure PCTCN2022127034-appb-000002
Figure PCTCN2022127034-appb-000003
本实施例中，如果AF是不受信的(即AF不在受信区)，AF向NEF发送AI/ML模型传输状态订阅请求，比如Nnef_AnalyticsExposure_Subscribe(即分析开放订阅)or Nnef_AnalyticsExposure_Fetch(即分析开放获取)请求。NEF向NWDAF发送第一消息，该第一消息可以为AI/ML模型开放传输状态订阅Nnwdaf_AnalyticsSubscription_Subscribe(即分析订阅订阅)or Nnwdaf_AnalyticsInfo_Request(即分析信息)请求，该请求中可以携带如表所示参数，请求订阅网络中AI/ML模型传输状态的分析信息。如果AF是受信的(即AF在受信区)，AF直接向NWDAF发送第一消息。
其中，该AI/ML模型传输状态订阅请求中可以携带如表1或表2所示参数，请求订阅网络中AI/ML模型传输状态的分析信息。
步骤102、AF直接或通过所述NEF接收所述NWDAF发送的AI/ML模型传输状态的分析信息。
其中，所述分析信息是所述NWDAF根据接收5G核心网的其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据确定的。
可选地，所述AI/ML模型传输状态的数据是所述NWDAF根据接收到的所述第一消息中请求的参数通过向所述5GC NF(s)发送第二消息得到的，所述第二消息用于采集用于分析网络中AI/ML模型传输状态的数据。
本实施例中，第二消息中包括下述至少一项：使用AI/ML模型的UE的当前位置(即UE location)、使用AI/ML模型的应用的标识(即Application ID，可以是服务器的标识，也可以是AF的标识)、传输AI/ML模型的服务质量流标识符(即QFI)、传输AI/ML模型的上行链路方向比特率(即bit rate for UL direction)及传输AI/ML模型的下行链路方向比特率(即bit rate for DL direction)、AI/ML模型的上行链路方向分组延迟(即Packet delay for UL direction)及AI/ML模型的下行链路方向分组延迟(即Packet delay for the DL direction)、服务质量流在AI/ML模型传输的时间段内异常释放的数量(QoS Sustainability)、AI/ML模型的分组传输数量(packet transmission)、AI/ML模型的分组重传数量(即packet  retransmission)、数据采集时间(即Timestamp)、AI/ML模型传输的时长(即AI/ML模型传输的时间段)、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于传输AI/ML模型服务质量流的PDU会话的网络切片、用于传输AI/ML模型服务质量流的PDU会话的数据网络、用于所述AF的服务流程(即IP filter information)。参见下述表3所示的第二消息示例表。
表3：第二消息示例表
Figure PCTCN2022127034-appb-000004
如果存在联邦学习，所述第二消息中还包括下述至少一项：指示用于指定分析的联邦学习组的标识(即Federated Learning(FL)group ID)、参加联邦学习的UE或UE(s)(即Federated Learning(FL)UE ID or UE group ID)、参加联邦学习的应用标识(即Federated Learning(FL)Application ID)。参见下述表4所示的第二消息示例表。
表4：第二消息示例表
Figure PCTCN2022127034-appb-000005
本实施例中，如果AF是不受信的，接收NEF发送的AI/ML模型传输状态的分析信息，该分析信息是NWDAF向NEF发送的。如果AF是受信的，直接接收NWDAF发送的AI/ML模型传输状态的分析信息。
其中，所述分析信息包括下述至少一项：用于传输AI/ML模型服务质量流的PDU会话的网络切片、使用AI/ML模型的应用的标识、使用AI/ML模型的区域信息、分析结果的有效时间(即Validity period)、提供AI/ML模型传输的用户面功能UPF(即UPF Info)、用于传输AI/ML模型服务质量流的 PDU会话的数据网络名称、AI/ML传输模型的大小、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、服务质量要求(即QoS requirements)；服务质量要求包括：传输AI/ML模型的服务质量流标识符(即QFI)、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、达到服务质量流在AI/ML模型传输的时间段内异常释放的报告阈值的次数、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量。参见下述表5所示的分析信息示例表。
表5：分析信息示例表
Figure PCTCN2022127034-appb-000006
Figure PCTCN2022127034-appb-000007
如果存在联邦学习，所述分析信息还包括下述至少一项：用于指定分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、指示提供AI/ML模型或者参与联邦学习的各个应用标识(即Application Server Instance Address)。参见下述表6所示的分析信息示例表。
表6：分析信息示例表
Figure PCTCN2022127034-appb-000008
Figure PCTCN2022127034-appb-000009
本实施例中，通过向NWDAF发送订阅网络中AI/ML模型传输状态的分析信息的请求，接收NWDAF发送的根据采集到的5GC NF(s)数据确定的分析信息，根据分析信息来调整网络策略参数和/或应用层模型的信息，实现了有效地对AI/ML模型传输状态进行分析，进而使得网络能够基于AI/ML模型传输状态有效地调整网络传输策略以及使得第三方获得AI/ML模型传输状态的分析进行应用层信息的调整。
示例性地，参见图4所示，图4为本公开实施例一提供的订阅网络中模型传输状态分析方法的信令流程示意图，图4为订阅网络中模型传输状态分析方法中AF与NWDAF、NEF和NF间的信令交互图。本实施例提供的订阅网络中模型传输状态分析方法包括以下步骤(即实施例一对应的信令交互过程：AF请求NWDAF提供AI/ML模型传输状态分析)：(其中，步骤4011至4016是AF在不受信任区域，步骤4021至4024是AF在受信任区域。)
步骤4011、如果AF是不受信的，AF向NEF发送AI/ML模型传输状态订阅Nnef_AnalyticsExposure_Subscribe(即分析开放订阅)or Nnef_AnalyticsExposure_Fetch(即分析开放获取)请求。
本实施例中，该请求中可以携带如表1或表2所示参数，请求订阅网络中AI/ML模型传输状态的分析信息。
步骤4012、NEF向NWDAF发送AI/ML模型开放传输状态订阅Nnwdaf_AnalyticsSubscription_Subscribe(即分析订阅订阅)or  Nnwdaf_AnalyticsInfo_Request(即分析信息)请求。
其中，AI/ML模型开放传输状态订阅Nnwdaf_AnalyticsSubscription_Subscribe or Nnwdaf_AnalyticsInfo_Request请求可以作为第一消息。
本实施例中，该请求中可以携带如表1或表2所示参数，请求订阅网络中AI/ML模型传输状态的分析信息。
步骤4013、NWDAF调用Nnf_EventExposure_Subscribe(即事件开放订阅)向5GC NF(s)采集数据。
其中，采集数据如表3或表4所示，用于分析网络中AI/ML模型传输状态。NWDAF向所述5GC NF(s)发送第二消息的方式可以是NWDAF调用Nnf_EventExposure_Subscribe。
步骤4014、5GC NF(s)调用Nnf_EventExposure_Notify(即事件开放通知)向NWDAF反馈所需数据。
步骤4015、NWDAF调用Nnwdaf_AnalyticsSubscription_Notify(即分析订阅通知)or Nnwdaf_AnalyticsInfo_Request response(即分析信息请求响应)向NEF发送AI/ML模型传输状态的分析信息。
步骤4016、NEF调用Nnef_AnalyticsExposure_Notify(即分析开放通知)or Nnef_AnalyticsExposure_Fetch response(即分析开放获取响应)向AF发送AI/ML模型传输状态的分析信息。
其中，分析信息如表5或表6所示。
步骤4021、如果AF是可信的，AF直接向NWDAF发送AI/ML模型传输状态订阅请求，执行操作如步骤4012描述。此外，消费者还可以是PCF、SMF。
步骤4022、执行操作如步骤4013描述。即步骤4013。
步骤4023、执行操作如步骤4014描述。即步骤4014。
步骤4024、NWDAF直接向AF发送AI/ML模型传输状态的分析信息，执行操作如步骤4016描述。
实施例二，在接收到所述分析信息后，所述方法还包括：
根据所述分析信息，直接或通过所述NEF向策略控制功能PCF发送第一请求。
其中，所述第一请求用于请求更新用于AI/ML模型传输的网络策略参数；所述网络策略参数用于优化AI/ML模型传输状态。
具体地，AF基于NWDAF分析结果，AF请求调整网络策略从而优化AI/ML模型传输状态。具体地，如果AF在受信区，AF直接向PCF发送第一请求，请求PCF更新用于AI/ML模型传输的网络策略参数。如果AF不在受信区，AF通过NEF向PCF发送第一请求，请求PCF更新用于AI/ML模型传输的网络策略参数。
可选地，根据所述分析信息，直接或通过所述NEF向策略控制功能PCF发送第一请求，可以通过以下步骤实现：
步骤a1、根据所述分析信息中传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量、达到服务质量流在AI/ML模型传输的时间段内异常释放的报告阈值的次数中的至少一项，确定传输AI/ML模型的新的服务质量参数；所述新的服务质量参数包括下述至少一项：5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级。
步骤a2、根据所述分析信息中的传输AI/ML模型的应用的标识、使用AI/ML模型的区域信息、使用AI/ML模型的应用服务的IP地址信息、用于传输AI/ML模型服务质量流的PDU会话的网络切片、用于传输AI/ML模型服务质量流的PDU会话的数据网络名称，确定传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息；或者，若AI/ML模型执行联邦学习，根据所述分析信息中的指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、指示提供AI/ML模型或者参与联邦学习的各个应用标识，确定传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息。
步骤a3、根据传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息，确定数据网络接入标识DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息，所述DNAI对应的UE(s)和各个AF的区域信息及地址 信息用于提供优化AI/ML模型传输状态的路径。
步骤a4、将所述新的服务质量参数、所述DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息作为所述第一请求中的参数直接或通过所述NEF发送给PCF。
具体地，AF依据从NWDAF获取的关于传输AI/ML模型分析信息，比如：传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量、达到服务质量流在AI/ML模型传输的时间段内异常释放的报告阈值的次数，确定传输AI/ML模型的新的QoS参数，并向PCF提供新的QoS参数。
AF依据从NWDAF获取的关于传输AI/ML模型分析信息，比如：传输AI/ML模型的应用的标识、使用AI/ML模型的区域信息、使用AI/ML模型的应用服务的IP地址信息、用于传输AI/ML模型服务质量流的PDU会话的网络切片、用于传输AI/ML模型服务质量流的PDU会话的数据网络名称，或者，AF依据从NWDAF获取的关于传输AI/ML模型分析信息，比如：指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、指示提供AI/ML模型或者参与联邦学习的各个应用标识，确定传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息，进而确定数据网络接入标识DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息。AF将携带有新的QoS参数、DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息的第一请求直接或通过所述NEF发送给PCF，请求PCF依据第一请求中携带的请求的参数，来更新用于AI/ML模型传输的相关策略参数从而优化AI/ML模型传输状态。
可选地，步骤a3可以通过以下步骤实现：
步骤a31、根据传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息，判断当前路由路径是否不佳；
步骤a32、若当前路由路径不佳，则根据传输AI/ML模型的UE(s)和各个AF的地址信息以及UE(s)的区域信息，确定传输AI/ML模型中传输双方的目的地址；
步骤a33、根据所述目的地址，确定最近的路径；
步骤a34、根据最近的路径，确定DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息。
具体地，判断当前路由路径不佳，选择能够提供更优的服务体验或性能的DNAI，向PCF提供传输AI/ML模型的DNAI以及对应的UE(s)和AF(s)的区域信息、地址信息。
可选地，所述第一请求，具体用于：
请求PCF根据所述新的服务质量参数调整PCC规则中的5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级并指示所述PCF将调整后的第一更新结果直接或通过NEF反馈；其中，所述第一更新结果是PCF根据所述新的服务质量参数调整PCC规则确定的；
相应的，所述方法还包括：
直接或通过NEF接收PCF发送的所述第一更新结果，所述第一更新结果包括所述第一请求被接受或所述第一请求被拒绝。
具体地，AF请求PCF依据提供的新的QoS参数，相应调整PCC规则中的请求PCF根据所述新的服务质量参数调整PCC规则中的5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级，并直接或通过NEF通知第一更新结果，即向AF通知这个请求是被接受还是被拒绝。可选地，所述第一请求具体用于：
请求PCF确定会话管理功能网元SMF是否需要更新会话管理策略，若确定SMF需要更新会话管理策略，则确定PCF向SMF发送第二请求，所述第二请求中请求的参数包括下述至少一项：DNAI、流量导向政策标识符、流量路线信息；所述第二请求用于用于SMF根据新的会话管理策略确定选择的用户面功能UPF并提供相应的DNAI、流量导向政策标识符、流量路线信息；
相应的，所述方法还包括：
直接或通过NEF接收PCF发送的第二更新结果，所述第二更新结果是 PCF根据SMF发送的新的会话管理策略是否更新UPF路径确定的；
其中，所述第二更新结果包括所述第一请求被接受或被拒绝。
具体地，AF请求PCF确定会话管理功能网元SMF是否需要更新会话管理策略，如果PCF确定SMF需要更新策略信息，PCF将会向SMF发起携带有DNAI、流量导向政策标识符、流量路线信息等请求参数的第二请求，SMF根据新的会话管理策略确定选择的用户面功能UPF并提供相应的DNAI、流量导向政策标识符、流量路线信息来更新会话管理策略，即更新SM策略。并直接或通过NEF通知第二更新结果，即向AF通知这个请求是被接受还是被拒绝。
示例性地，参见图5所示，图5为本公开实施例二提供的订阅网络中模型传输状态分析方法的信令流程示意图，图5为订阅网络中模型传输状态分析方法中AF与NEF和PCF间的信令交互图。本实施例提供的订阅网络中模型传输状态分析方法包括以下步骤(即实施例二对应的信令交互过程：基于NWDAF分析结果，AF请求调整网络策略从而优化AI/ML模型传输状态)：(其中，步骤5011至5014是AF在不受信任区域，步骤5021至5022是AF在受信任区域。)
步骤5010、AF从NWDAF签约并检索到AI/ML模型传输状态分析。
本实施例中，AF通过NEF或者直接向NWDAF订阅并获得AI/ML模型传输状态分析，如上述步骤4011至4024所述。
步骤5011、AF向NEF发送Nnef_AFsessionWithQoS_Update(即基于服务质量的AF会话更新)请求。
本实施例中，如果AF不受信，对于已经建立和有QoS要求的AF session(即AF会话)，AF可以向NEF发送Nnef_AFsessionWithQoS_Update(即基于服务质量的AF会话更新)请求，来更新用于AI/ML模型传输的相关策略参数从而优化AI/ML模型传输状态。
具体的，AF依据从NWDAF获取的关于传输AI/ML模型分析信息：QoS flow Bit Rate(即服务质量流比特率，如：传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率)、QoS flow Packet Delay(即服务质量流分组延迟，如：AI/ML模型的上行链路方向分组延迟及AI/ML模 型的下行链路方向分组延迟)、QoS Sustainability、Packet transmission、Packet retransmission、QoS Sustainability、QoS Sustainability Reporting Threshold(s)，确定传输AI/ML模型的新的QoS参数，并向PCF提供新的QoS参数：5G QoS Identifier(5QI，即5G服务质量标识符)、Reflective QoS Control(即反射式服务质量控制)、UL-maximum bitrate(即传输AI/ML模型的上行链路方向最大比特率)、DL-maximum bitrate(即传输AI/ML模型的下行链路方向最大比特率)、UL-guaranteed bitrate(即传输AI/ML模型的上行链路方向最低比特率)、DL-guaranteed bitrate(传输AI/ML模型的下行链路方向最低比特率)，Priority Level(即优先级)，即随着预测结果的增减而增减使5GS满足模型传输的QoS要求。
此外，AF依据从NWDAF获取的关于传输AI/ML模型分析信息：传输AI/ML模型的AF ID，使用AI/ML模型的UE区域信息，使用AI/ML模型的应用服务的IP地址信息，AI/ML模型传输使用的网络切片、DNN信息，确定传输AI/ML模型的UE(s)和AF(s)的区域信息、地址信息。或者，分析信息中也可能有关于联邦学习group信息的信息：Federated Learning(FL)group ID、Federated Learning(FL)UE ID or UE group ID、Federated Learning(FL)的Application ID，确定传输AI/ML模型的UE(s)和AF(s)的区域信息、地址信息，判断当前路由路径不佳，选择能够提供更优的服务体验或性能的DNAI，向PCF提供传输AI/ML模型的DNAI以及对应的UE(s)和AF(s)的区域信息、地址信息。
其中，判断当前路由路径不佳的过程可以为：AF基于AF的IP地址信息，UE(s)的区域信息(它可对应AMF、SMF、UPF等的ID，UPF的N6接口连接DN)，确定当前的路由路径；例如，有些UE/服务器如果下次传输前加入/离开(联邦群组)，则有些路径则不佳，或者跳数过多，路径就不佳。
选择能够提供更优的服务体验或性能的DNAI的过程可以为：基于收到的IP地址信息，UE(s)的区域信息，可以确定传输双方的目的地址，就能找出更优的(最近的)路由路径，对应出DNAI。
步骤5012、NEF向PCF发送Npcf_PolicyAuthorization_Update request(即策略授权更新请求)。
本实施例中，NEF将上述信息通过Npcf_PolicyAuthorization_Update request(即策略授权更新请求)发送给PCF，来更新用于AI/ML模型传输的相关策略参数从而优化AI/ML模型传输状态。
其中，Npcf_PolicyAuthorization_Update request可以作为第二请求。
步骤5013、PCF向NEF通知结果(即PCF向NEF发送Npcf_PolicyAuthorization_Update response(策略授权更新响应)。
本实施例中，PCF依据NEF提供的信息，具体的，PCF依据提供的新的QoS参数，相应调整PCC规则中的：5G QoS Identifier(5QI)、Reflective QoS Control、UL-maximum bitrate、DL-maximum bitrate、UL-guaranteed bitrate、DL-guaranteed bitrate，Priority Level，并发送Npcf_PolicyAuthorization_Update response(策略授权更新响应)向NEF通知结果。
如果PCF确定SMF需要更新策略信息，PCF将会向SMF发起Npcf_SMPolicyControl_UpdateNotify request(即会话策略控制更新通知请求)(DNAI，Per DNAI:Traffic steering policy identifier，Per DNAI:N6traffic routing information)，更新SM策略，SMF依据此策略选择UPF，并提供DNAI，Per DNAI:Traffic steering policy identifier(即流量导向政策标识符)，Per DNAI:N6traffic routing information(流量路线信息)。其中，Npcf_SMPolicyControl_UpdateNotify request可以作为第二请求。
步骤5014、NEF向AF发送Nnef_AFsessionWithQoS_Update response。
本实施例中，NEF发送Nnef_AFsessionWithQoS_Update response(即基于服务质量的AF会话更新响应)，向AF通知这个请求是被接受还是被拒绝。
步骤5021、AF直接向PCF发送Npcf_PolicyAuthorization_Update request。
本实施例中，如果AF是受信的，AF直接向PCF发送Npcf_PolicyAuthorization_Update request来更新用于AI/ML模型传输的相关策略参数从而优化AI/ML模型传输状态，执行操作如步骤5012描述。
步骤5022、PCF直接向AF通知结果(即PCF直接向AF发送Npcf_PolicyAuthorization_Update response(即策略授权更新响应))。
本实施例中，首先，PCF依据AF提供的信息，执行操作如步骤5012描 述。PCF直接向AF通知这个请求是被接受还是被拒绝。
实施例三，在接收到所述分析信息后，所述方法还可以通过以下步骤实现：
步骤b1、根据所述分析信息，调整应用层模型的信息，所述应用层模型的信息包括下述至少一项：模型压缩、模型大小、模型传输时间段、模型编解码；所述应用层模型的信息用于更新服务质量参数；
步骤b2、根据调整后的应用层模型的信息，确定所述新的服务质量参数，所述新的服务质量参数包括：5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级；
步骤b3、直接或通过所述NEF向策略控制功能PCF发送第三请求；
其中，所述第三请求中请求的参数包括所述新的服务质量参数，所述第三请求用于请求更新服务质量参数。
具体地，AF基于NWDAF提供的QoS要求信息，调整应用层模型的信息来更新服务质量参数，比如：调整模型压缩、模型大小、模型传输时间段、模型编解码等。根据调整后的应用层模型的信息，确定所述新的服务质量参数，直接或通过NEF，PCF发送携带有新的服务质量参数的第三请求。
可选地，所述第三请求，具体用于：
请求PCF根据所述新的服务质量参数调整PCC规则中的5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、优先级；
相应的，所述方法还包括：
直接或通过NEF接收PCF发送的第三更新结果，第三更新结果是PCF基于调整PCC规则的结果确定的，所述第三更新结果包括所述第三请求被接受或所述第三请求被拒绝。
具体地，AF请求PCF依据提供的新的QoS参数，相应调整PCC(即策略和付费控制)规则中的5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最 大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、优先级，PCF调整PCC规则并直接或通过NEF向AF通知这个请求是被接受还是被拒绝。
可选地，在所述调整应用层模型的信息之后，所述方法还可以通过以下步骤实现：
将调整后的应用层模型的信息中的模型压缩、模型大小以及模型编解码直接发送给PCF，所述调整后的应用层模型的信息用于支持PCF调整PCC规则中的5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级；
相应的，所述方法还包括：
接收PCF发送的第四更新结果，第四更新结果是PCF根据调整后的应用层模型的信息调整PCC规则的结果确定的，所述第四更新结果包括所述第三请求被接受或所述第三请求被拒绝。
具体地，PCF依据模型压缩、模型大小、模型编解码调整PCC规则中的上述QoS参数。比如：5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级。PCF调整PCC规则并直接或通过NEF向AF通知这个请求是被接受还是被拒绝。
可选地，在所述调整应用层模型的信息之后，所述方法还包括：
将调整后的应用层模型的信息中的模型传输时间直接发送给PCF，所述调整后的应用层模型的信息中的模型传输时间用于支持PCF调整PCC规则中的门状态参数；所述门状态参数用于支持SMF根据门状态中的传输开始时间和传输结束时间来更新会话管理策略；
相应的，所述方法还包括：
接收PCF发送的第五更新结果，第五更新结果是PCF通过接收SMF发送的新的会话管理策略的结果确定的，所述第五更新结果包括所述第三请求被接受或所述第三请求被拒绝。
具体地，PCF依据模型传输时间调整PCC规则中的Gate status(即门状态参数)，更新SM策略，SMF依据此影响流的传输开始和结束时间，反馈给PCF，并PCF接或通过NEF向AF通知这个请求是被接受还是被拒绝。
示例性地，参见图6所示，图6为本公开实施例三提供的订阅网络中模型传输状态分析方法的信令流程示意图，图6为订阅网络中模型传输状态分析方法中AF与NEF和PCF间的信令交互图。本实施例提供的订阅网络中模型传输状态分析方法包括以下步骤(即实施例三对应的信令交互过程：AF基于NWDAF提供的分析信息，调整应用层模型的信息，如，调整模型压缩、模型大小、模型传输时间段、模型编解码等，进而更新QoS要求，类似实施例二的步骤流程)：(其中，步骤6012至6014是AF在不受信任区域，步骤6021至6022是AF在受信任区域。)
步骤6010、AF从NWDAF签约并检索到AI/ML模型传输状态分析。
具体地AF通过NEF或者直接向NWDAF订阅并获得AI/ML模型传输状态分析，如上述步骤4011至4024所述。
步骤6011、AF基于分析信息，调整应用层行为。
本实施例中，AF基于NWDAF提供的QoS要求信息，调整应用层模型的信息，如，调整模型压缩、模型大小、模型传输时间段、模型编解码等。
步骤6012、AF向NEF发送Nnef_AFsessionWithQoS_Update请求。
步骤6013、NEF向PCF发送Npcf_PolicyAuthorization_Update request。
其中，Npcf_PolicyAuthorization_Update request可以作为第三请求。
步骤6014、PCF向NEF通知结果(即PCF向NEF发送Npcf_PolicyAuthorization_Update response。
步骤6015、NEF向AF发送Nnef_AFsessionWithQoS_Update response。
步骤6021、AF直接向PCF发送Npcf_PolicyAuthorization_Update request。
步骤6022、PCF直接向AF通知结果(即PCF直接向AF发送Npcf_PolicyAuthorization_Update response(即策略授权更新响应))。
具体地，AF请求session进行QoS更新。具体的，AF基于调整的模型压缩、模型大小、模型传输时间段、模型编解码等，确定传输AI/ML模型的 新的QoS参数，并向PCF提供新的QoS参数：5G QoS Identifier(5QI)、Reflective QoS Control、UL-maximum bitrate、DL-maximum bitrate、UL-guaranteed bitrate、DL-guaranteed bitrate，Priority Level。或者，AF将模型压缩、模型大小、模型传输时间段、模型编解码调整信息等直接发送给PCF。
步骤6013、PCF依据提供的新的QoS参数，相应调整PCC规则中的：5G QoS Identifier(5QI)、Reflective QoS Control、UL-maximum bitrate、DL-maximum bitrate、UL-guaranteed bitrate、DL-guaranteed bitrate，Priority Level；或者PCF依据模型压缩、模型大小、模型编解码调整PCC规则中的上述QoS参数；或者PCF依据模型传输时间调整PCC规则中的Gate status，更新SM策略，SMF依据此影响流的传输开始和结束时间；最后，通过实施例二中步骤5014或5022的方式向AF通知这个请求是被接受还是被拒绝。
本公开，NWDAF接收AF发送的AI/ML模型传输状态分析请求，以及请求中包含的参数；NWDAF从5GC NF(s)采集的输入数据，用于分析网络中AI/ML模型传输状态；NWDAF进行分析并向AF发送AI/ML模型传输状态分析信息；AF基于AI/ML模型传输状态分析信息请求AF session的QoS更新，并调整相应的策略控制功能PCF、会话管理功能网元SMF等网元的策略；AF基于AI/ML模型传输状态分析信息调整应用层模型信息的相关参数，进而调整QoS，从而优化AI/ML模型传输状态。使得第三方可以获得AI/ML模型传输状态，还可以基于模型传输的分析结果，网络调整自身行为来满足AI/ML模型传输要求，并且第三方还可以基于模型传输的分析结果，调整模型应用层的行为来实现AI/ML模型的高效传输，保证AI/ML模型传输的业务体验和业务性能。
图7为本公开实施例四提供的订阅网络中模型传输状态分析方法的流程示意图，如图7所示，本实施例提供的订阅网络中模型传输状态分析方法的执行主体为NWDAF，则本公开实施例提供的订阅网络中模型传输状态分析方法包括以下步骤：
步骤701、NWDAF直接或通过网络能力开放功能NEF接收应用功能AF发送的第一消息；其中，所述第一消息用于请求订阅网络中人工智能/机器学习AI/ML模型传输状态的分析信息；
步骤702、NWDAF根据所述第一消息中请求的参数，向5G核心网的其 他网络功能5GC NF(s)发送第二消息，所述第二消息用于采集用于分析网络中AI/ML模型传输状态的数据。
步骤703、NWDAF接收5G核心网的其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据，并对AI/ML模型传输状态的数据进行分析，得到AI/ML模型传输状态的分析信息。
其中，所述分析信息用于通过AF调整网络策略参数和/或应用层模型的信息。
可选地，所述第一消息中请求的参数包括下述至少一项：网络数据分析标识、接收AI/ML模型的一个用户设备UE或一组UE的标识或满足分析条件的任意UE、使用AI/ML模型的应用的标识、AI/ML模型传输的区域、指示传输AI/ML模型服务质量流的协议数据单元PDU会话的网络切片、指示传输AI/ML模型服务质量流的PDU会话的数据网络、AI/ML模型传输的时间段、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于指示传输AI/ML模型的服务质量流的服务质量要求和/或用于指示传输AI/ML模型的特定的服务质量要求；
所述第二消息中包括下述至少一项：使用AI/ML模型的UE的当前位置、使用AI/ML模型的应用的标识、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量、数据采集时间、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于传输AI/ML模型服务质量流的PDU会话的网络切片、用于传输AI/ML模型服务质量流的PDU会话的数据网络、用于所述AF的服务流程；
所述分析信息包括下述至少一项：用于传输AI/ML模型服务质量流的PDU会话的网络切片、使用AI/ML模型的应用的标识、使用AI/ML模型的区域信息、分析结果的有效时间、提供AI/ML模型传输的用户面功能UPF、用于传输AI/ML模型服务质量流的PDU会话的数据网络名称、AI/ML传输模型的大小、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML 模型传输的结束时间戳、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、达到服务质量流在AI/ML模型传输的时间段内异常释放的报告阈值的次数、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量；
其中，若AI/ML模型执行联邦学习，所述第一消息中请求的参数还包括：联邦学习群信息，所述联邦学习群信息包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
相应的，所述第二消息中还包括下述至少一项：用于指定分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
相应的，所述分析信息还包括下述至少一项：用于指定分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、指示提供AI/ML模型或者参与联邦学习的各个应用标识。
本实施例中，通过接收AF发送的订阅网络中AI/ML模型传输状态的分析信息的请求，根据所述第一消息中请求的参数，向5G核心网的其他网络功能5GC NF(s)采集数据，接收5G核心网的其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据，并对AI/ML模型传输状态的数据进行分析，得到AI/ML模型传输状态的分析信息，实现了有效地对AI/ML模型传输状态进行分析，并使得AF根据分析信息来调整网络策略参数和/或应用层模型的信息，进而使得网络能够基于AI/ML模型传输状态有效地调整网络传输策略以及使得第三方获得AI/ML模型传输状态的分析进行应用程信息的调整。
在此需要说明的是，本公开实施例提供的订阅网络中模型传输状态分析方法，能够实现图4所示方法实施例所实现的所有方法步骤，且能够达到相同的技术效果，在此不再对本实施例中与方法实施例相同的部分及有益效果进行具体赘述。
图8为本公开实施例提供的订阅网络中模型传输状态分析装置的结构示意图，如图8所示，本实施例提供的订阅网络中模型传输状态分析装置应用 于AF。则本实施例提供的订阅网络中模型传输状态分析装置包括：收发机800，用于在处理器810的控制下接收和发送数据。
其中，在图8中，总线架构可以包括任意数量的互联的总线和桥，具体由处理器810代表的一个或多个处理器和存储器820代表的存储器的各种电路链接在一起。总线架构还可以将诸如外围设备、稳压器和功率管理电路等之类的各种其他电路链接在一起，这些都是本领域所公知的，因此，本文不再对其进行进一步描述。总线接口提供接口。收发机800可以是多个元件，即包括发送机和接收机，提供用于在传输介质上与各种其他装置通信的单元，这些传输介质包括无线信道、有线信道、光缆等传输介质。处理器810负责管理总线架构和通常的处理，存储器820可以存储处理器810在执行操作时所使用的数据。
处理器810可以是中央处埋器(CPU)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现场可编程门阵列(Field－Programmable Gate Array，FPGA)或复杂可编程逻辑器件(Comple6 Programmable Logic Device，CPLD)，处理器也可以采用多核架构。
本实施例中，存储器820，用于存储计算机程序；收发机800，用于在处理器810的控制下收发数据；处理器810，用于读取存储器中的计算机程序并执行以下操作：
直接或通过网络能力开放功能NEF向网络数据分析功能NWDAF发送第一消息；其中，所述第一消息用于请求订阅网络中人工智能/机器学习AI/ML模型传输状态的分析信息；
直接或通过所述NEF接收所述NWDAF发送的AI/ML模型传输状态的分析信息，所述分析信息是所述NWDAF根据接收5G核心网的其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据确定的，所述AI/ML模型传输状态的数据是所述NWDAF根据接收到的所述第一消息中请求的参数通过向所述5GC NF(s)发送第二消息得到的，所述第二消息用于采集用于分析网络中AI/ML模型传输状态的数据；
其中，所述分析信息用于调整网络策略参数和/或应用层模型的信息。
可选地，所述第一消息中请求的参数包括下述至少一项：网络数据分析 标识、接收AI/ML模型的一个用户设备UE或一组UE的标识或满足分析条件的任意UE、使用AI/ML模型的应用的标识、AI/ML模型传输的区域、指示传输AI/ML模型服务质量流的协议数据单元PDU会话的网络切片、指示传输AI/ML模型服务质量流的PDU会话的数据网络、AI/ML模型传输的时间段、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于指示传输AI/ML模型的服务质量流的服务质量要求和/或用于指示传输AI/ML模型的特定的服务质量要求；
所述第二消息中包括下述至少一项：使用AI/ML模型的UE的当前位置、使用AI/ML模型的应用的标识、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量、数据采集时间、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于传输AI/ML模型服务质量流的PDU会话的网络切片、用于传输AI/ML模型服务质量流的PDU会话的数据网络、用于所述AF的服务流程；
所述分析信息包括下述至少一项：用于传输AI/ML模型服务质量流的PDU会话的网络切片、使用AI/ML模型的应用的标识、使用AI/ML模型的区域信息、分析结果的有效时间、提供AI/ML模型传输的用户面功能UPF、用于传输AI/ML模型服务质量流的PDU会话的数据网络名称、AI/ML传输模型的大小、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、达到服务质量流在AI/ML模型传输的时间段内异常释放的报告阈值的次数、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量；
其中，若AI/ML模型执行联邦学习，所述第一消息中请求的参数还包括：联邦学习群信息，所述联邦学习群信息包括下述至少一项：用于指指示分析 的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
相应的，所述第二消息中还包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
相应的，所述分析信息还包括下述至少一项：用于指定分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、指示提供AI/ML模型或者参与联邦学习的各个应用标识。
可选地，处理器810，还用于：
在接收到所述分析信息后，根据所述分析信息，直接或通过所述NEF向策略控制功能PCF发送第一请求；
其中，所述第一请求用于请求更新用于AI/ML模型传输的网络策略参数；所述网络策略参数用于优化AI/ML模型传输状态。
可选地，处理器810，用于根据所述分析信息，直接或通过所述NEF向策略控制功能PCF发送第一请求时，具体包括：
根据所述分析信息中传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量、达到服务质量流在AI/ML模型传输的时间段内异常释放的报告阈值的次数中的至少一项，确定传输AI/ML模型的新的服务质量参数；所述新的服务质量参数包括下述至少一项：5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级；
根据所述分析信息中的传输AI/ML模型的应用的标识、使用AI/ML模型的区域信息、使用AI/ML模型的应用服务的IP地址信息、用于传输AI/ML模型服务质量流的PDU会话的网络切片、用于传输AI/ML模型服务质量流的PDU会话的数据网络名称，确定传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息；或者，若AI/ML模型执行联邦学习，根据所述分析信息 中的指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、指示提供AI/ML模型或者参与联邦学习的各个应用标识，确定传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息；
根据传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息，确定数据网络接入标识DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息，所述DNAI对应的UE(s)和各个AF的区域信息及地址信息用于提供优化AI/ML模型传输状态的路径；
将所述新的服务质量参数、所述DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息作为所述第一请求中的参数直接或通过所述NEF发送给PCF。
可选地，处理器810，用于根据传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息，确定数据网络接入标识DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息时，具体包括：
根据传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息，判断当前路由路径是否不佳；
若当前路由路径不佳，则根据传输AI/ML模型的UE(s)和各个AF的地址信息以及UE(s)的区域信息，确定传输AI/ML模型中传输双方的目的地址；
根据所述目的地址，确定最近的路径；
根据最近的路径，确定DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息。
可选地，所述第一请求，具体用于：
请求PCF根据所述新的服务质量参数调整PCC规则中的5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级并指示所述PCF将调整后的第一更新结果直接或通过NEF反馈；其中，所述第一更新结果是PCF根据所述新的服务质量参数调整PCC规则确定的；
相应的，处理器810，还具体：
直接或通过NEF接收PCF发送的所述第一更新结果，所述第一更新结果包括所述第一请求被接受或所述第一请求被拒绝。
可选地，所述第一请求具体用于：
请求PCF确定会话管理功能网元SMF是否需要更新会话管理策略，若确定SMF需要更新会话管理策略，则确定PCF向SMF发送第二请求，所述第二请求中请求的参数包括下述至少一项：DNAI、流量导向政策标识符、流量路线信息；所述第二请求用于SMF根据新的会话管理策略确定选择的用户面功能UPF并提供相应的DNAI、流量导向政策标识符、流量路线信息；
相应的，处理器810，还具体：
直接或通过NEF接收PCF发送的第二更新结果，所述第二更新结果是PCF根据SMF发送的新的会话管理策略是否更新UPF路径确定的；
其中，所述第二更新结果包括所述第一请求被接受或被拒绝。
可选地，处理器810，还用于：
在接收到所述分析信息后，根据所述分析信息，调整应用层模型的信息，所述应用层模型的信息包括下述至少一项：模型压缩、模型大小、模型传输时间段、模型编解码；所述应用层模型的信息用于更新服务质量参数；
根据调整后的应用层模型的信息，确定所述新的服务质量参数，所述新的服务质量参数包括：5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级；
直接或通过所述NEF向策略控制功能PCF发送第三请求；
其中，所述第三请求中请求的参数包括所述新的服务质量参数，所述第三请求用于请求更新服务质量参数。
可选地，所述第三请求，具体用于：
请求PCF根据所述新的服务质量参数调整PCC规则中的5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、优先级；
相应的，处理器810，还用于：
直接或通过NEF接收PCF发送的第三更新结果，第三更新结果是PCF基于调整PCC规则的结果确定的，所述第三更新结果包括所述第三请求被接 受或所述第三请求被拒绝。
可选地，处理器810，还用于：
在所述调整应用层模型的信息之后，将调整后的应用层模型的信息中的模型压缩、模型大小以及模型编解码直接发送给PCF，所述调整后的应用层模型的信息用于支持PCF调整PCC规则中的5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级；
相应的，处理器810，还用于：
接收PCF发送的第四更新结果，第四更新结果是PCF根据调整后的应用层模型的信息调整PCC规则的结果确定的，所述第四更新结果包括所述第三请求被接受或所述第三请求被拒绝。
可选地，处理器810，还用于：
在所述调整应用层模型的信息之后，将调整后的应用层模型的信息中的模型传输时间直接发送给PCF，所述调整后的应用层模型的信息中的模型传输时间用于支持PCF调整PCC规则中的门状态参数；所述门状态参数用于支持SMF根据门状态中的传输开始时间和传输结束时间来更新会话管理策略；
相应的，处理器810，还用于：
接收PCF发送的第五更新结果，第五更新结果是PCF通过接收SMF发送的新的会话管理策略的结果确定的，所述第五更新结果包括所述第三请求被接受或所述第三请求被拒绝。
在此需要说明的是，本公开提供的订阅网络中模型传输状态分析装置，能够实现图3-图6所示方法实施例所实现的所有方法步骤，且能够达到相同的技术效果，在此不再对本实施例中与方法实施例相同的部分及有益效果进行具体赘述。
图9为本公开另一实施例提供的订阅网络中模型传输状态分析装置的结构示意图，如图9所示，本实施例提供的订阅网络中模型传输状态分析装置应用于AF，则本实施例提供的订阅网络中模型传输状态分析装置900包括：
发送单元901，用于直接或通过网络能力开放功能NEF向网络数据分析功能NWDAF发送第一消息；其中，所述第一消息用于请求订阅网络中人工 智能/机器学习AI/ML模型传输状态的分析信息；
分析单元902，用于直接或通过所述NEF接收所述NWDAF发送的AI/ML模型传输状态的分析信息，所述第二消息用于采集用于分析网络中AI/ML模型传输状态的数据；
其中，所述分析信息用于调整网络策略参数和/或应用层模型的信息。
可选地，所述第一消息中请求的参数包括下述至少一项：网络数据分析标识、接收AI/ML模型的一个用户设备UE或一组UE的标识或满足分析条件的任意UE、使用AI/ML模型的应用的标识、AI/ML模型传输的区域、指示传输AI/ML模型服务质量流的协议数据单元PDU会话的网络切片、指示传输AI/ML模型服务质量流的PDU会话的数据网络、AI/ML模型传输的时间段、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于指示传输AI/ML模型的服务质量流的服务质量要求和/或用于指示传输AI/ML模型的特定的服务质量要求；
所述第二消息中包括下述至少一项：使用AI/ML模型的UE的当前位置、使用AI/ML模型的应用的标识、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量、数据采集时间、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于传输AI/ML模型服务质量流的PDU会话的网络切片、用于传输AI/ML模型服务质量流的PDU会话的数据网络、用于所述AF的服务流程；
所述分析信息包括下述至少一项：用于传输AI/ML模型服务质量流的PDU会话的网络切片、使用AI/ML模型的应用的标识、使用AI/ML模型的区域信息、分析结果的有效时间、提供AI/ML模型传输的用户面功能UPF、用于传输AI/ML模型服务质量流的PDU会话的数据网络名称、AI/ML传输模型的大小、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、服务质量要求、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方 向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、达到服务质量流在AI/ML模型传输的时间段内异常释放的报告阈值的次数、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量；
其中，若AI/ML模型执行联邦学习，所述第一消息中请求的参数还包括：联邦学习群信息，所述联邦学习群信息包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
相应的，所述第二消息中还包括下述至少一项：用于指定分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
相应的，所述分析信息还包括下述至少一项：用于指定分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、指示提供AI/ML模型或者参与联邦学习的各个应用标识。
可选地，发送单元，还用于：
在接收到所述分析信息后，根据所述分析信息，直接或通过所述NEF向策略控制功能PCF发送第一请求；
其中，所述第一请求用于请求更新用于AI/ML模型传输的网络策略参数；所述网络策略参数用于优化AI/ML模型传输状态。
可选地，发送单元，具体用于：根据所述分析信息中传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量、达到服务质量流在AI/ML模型传输的时间段内异常释放的报告阈值的次数中的至少一项，确定传输AI/ML模型的新的服务质量参数；所述新的服务质量参数包括下述至少一项：5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级；
根据所述分析信息中的传输AI/ML模型的应用的标识、使用AI/ML模型的区域信息、使用AI/ML模型的应用服务的IP地址信息、用于传输AI/ML模型服务质量流的PDU会话的网络切片、用于传输AI/ML模型服务质量流的PDU会话的数据网络名称，确定传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息；或者，若AI/ML模型执行联邦学习，根据所述分析信息中的指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、指示提供AI/ML模型或者参与联邦学习的各个应用标识，确定传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息；
根据传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息，确定数据网络接入标识DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息，所述DNAI对应的UE(s)和各个AF的区域信息及地址信息用于提供优化AI/ML模型传输状态的路径；
将所述新的服务质量参数、所述DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息作为所述第一请求中的参数直接或通过所述NEF发送给PCF。
可选地，发送单元，还具体用于：
根据传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息，判断当前路由路径是否不佳；
若当前路由路径不佳，则根据传输AI/ML模型的UE(s)和各个AF的地址信息以及UE(s)的区域信息，确定传输AI/ML模型中传输双方的目的地址；
根据所述目的地址，确定最近的路径；
根据最近的路径，确定DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息。
可选地，所述第一请求，具体用于：
请求PCF根据所述新的服务质量参数调整PCC规则中的5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级并指示所述PCF将调整后的第一更新结果直接或通过NEF反馈；其中，所述第一更新结果是PCF根据所述新的服务质量参数调整PCC规则确定的；
相应的，发送单元，还用于：
直接或通过NEF接收PCF发送的所述第一更新结果，所述第一更新结果包括所述第一请求被接受或所述第一请求被拒绝。
可选地，所述第一请求具体用于：
请求PCF确定会话管理功能网元SMF是否需要更新会话管理策略，若确定SMF需要更新会话管理策略，则确定PCF向SMF发送第二请求，所述第二请求中请求的参数包括下述至少一项：DNAI、流量导向政策标识符、流量路线信息；所述第二请求用于SMF根据新的会话管理策略确定选择的用户面功能UPF并提供相应的DNAI、流量导向政策标识符、流量路线信息；
相应的，接收单元，还用于：
直接或通过NEF接收PCF发送的第二更新结果，所述第二更新结果是PCF根据SMF发送的新的会话管理策略是否更新UPF路径确定的；
其中，所述第二更新结果包括所述第一请求被接受或被拒绝。
可选地，该装置还包括：确定单元；确定单元，用于：
在接收到所述分析信息后，根据所述分析信息，调整应用层模型的信息，所述应用层模型的信息包括下述至少一项：模型压缩、模型大小、模型传输时间段、模型编解码；所述应用层模型的信息用于更新服务质量参数；
根据调整后的应用层模型的信息，确定所述新的服务质量参数，所述新的服务质量参数包括：5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级；
直接或通过所述NEF向策略控制功能PCF发送第三请求；
其中，所述第三请求中请求的参数包括所述新的服务质量参数，所述第三请求用于请求更新服务质量参数。
可选地，所述第三请求，具体用于：
请求PCF根据所述新的服务质量参数调整PCC规则中的5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、优先级；
相应的，接收单元，还用于：
直接或通过NEF接收PCF发送的第三更新结果，第三更新结果是PCF基于调整PCC规则的结果确定的，所述第三更新结果包括所述第三请求被接受或所述第三请求被拒绝。
可选地，发送单元，还用于：
在所述调整应用层模型的信息之后，将调整后的应用层模型的信息中的模型压缩、模型大小以及模型编解码直接发送给PCF，所述调整后的应用层模型的信息用于支持PCF调整PCC规则中的5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级；
相应的，接收单元，还用于：
接收PCF发送的第四更新结果，第四更新结果是PCF根据调整后的应用层模型的信息调整PCC规则的结果确定的，所述第四更新结果包括所述第三请求被接受或所述第三请求被拒绝。
可选地，发送单元，还用于：
在所述调整应用层模型的信息之后，将调整后的应用层模型的信息中的模型传输时间直接发送给PCF，所述调整后的应用层模型的信息中的模型传输时间用于支持PCF调整PCC规则中的门状态参数；所述门状态参数用于支持SMF根据门状态中的传输开始时间和传输结束时间来更新会话管理策略；
相应的，接收单元，还用于：
接收PCF发送的第五更新结果，第五更新结果是PCF通过接收SMF发送的新的会话管理策略的结果确定的，所述第五更新结果包括所述第三请求被接受或所述第三请求被拒绝。
在此需要说明的是，本公开提供的订阅网络中模型传输状态分析装置，能够实现图3-图6方法实施例所实现的所有方法步骤，且能够达到相同的技术效果，在此不再对本实施例中与方法实施例相同的部分及有益效果进行具体赘述。
图10为本公开再一实施例提供的订阅网络中模型传输状态分析装置的结构示意图，如图10所示，本实施例提供的订阅网络中模型传输状态分析装 置应用于NWDAF。则本实施例提供的订阅网络中模型传输状态分析装置包括：收发机1000，用于在处理器1010的控制下接收和发送数据。
其中，在图10中，总线架构可以包括任意数量的互联的总线和桥，具体由处理器1010代表的一个或多个处理器和存储器1020代表的存储器的各种电路链接在一起。总线架构还可以将诸如外围设备、稳压器和功率管理电路等之类的各种其他电路链接在一起，这些都是本领域所公知的，因此，本文不再对其进行进一步描述。总线接口提供接口。收发机1000可以是多个元件，即包括发送机和接收机，提供用于在传输介质上与各种其他装置通信的单元，这些传输介质包括无线信道、有线信道、光缆等传输介质。处理器1010负责管理总线架构和通常的处理，存储器1020可以存储处理器1010在执行操作时所使用的数据。
处理器1010可以是中央处埋器(CPU)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现场可编程门阵列(Field－Programmable Gate Array，FPGA)或复杂可编程逻辑器件(Comple8 Programmable Logic Device，CPLD)，处理器也可以采用多核架构。
本实施例中，存储器1020，用于存储计算机程序；收发机1000，用于在处理器的控制下收发数据；处理器1010，用于读取存储器中的计算机程序并执行以下操作：
直接或通过网络能力开放功能NEF接收应用功能AF发送的第一消息；其中，所述第一消息用于请求订阅网络中人工智能/机器学习AI/ML模型传输状态的分析信息；
根据所述第一消息中请求的参数，向5G核心网的其他网络功能5GC NF(s)发送第二消息，所述第二消息用于采集用于分析网络中AI/ML模型传输状态的数据；
接收5G核心网的其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据，并对AI/ML模型传输状态的数据进行分析，得到AI/ML模型传输状态的分析信息；
其中，所述分析信息用于通过AF调整网络策略参数和/或应用层模型的信息。
可选地，所述第一消息中请求的参数包括下述至少一项：网络数据分析 标识、接收AI/ML模型的一个用户设备UE或一组UE的标识或满足分析条件的任意UE、使用AI/ML模型的应用的标识、AI/ML模型传输的区域、指示传输AI/ML模型服务质量流的协议数据单元PDU会话的网络切片、指示传输AI/ML模型服务质量流的PDU会话的数据网络、AI/ML模型传输的时间段、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于指示传输AI/ML模型的服务质量流的服务质量要求和/或用于指示传输AI/ML模型的特定的服务质量要求；
所述第二消息中包括下述至少一项：使用AI/ML模型的UE的当前位置、使用AI/ML模型的应用的标识、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量、数据采集时间、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于传输AI/ML模型服务质量流的PDU会话的网络切片、用于传输AI/ML模型服务质量流的PDU会话的数据网络、用于所述AF的服务流程；
所述分析信息包括下述至少一项：用于传输AI/ML模型服务质量流的PDU会话的网络切片、使用AI/ML模型的应用的标识、使用AI/ML模型的区域信息、分析结果的有效时间、提供AI/ML模型传输的用户面功能UPF、用于传输AI/ML模型服务质量流的PDU会话的数据网络名称、AI/ML传输模型的大小、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、服务质量要求、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、达到服务质量流在AI/ML模型传输的时间段内异常释放的报告阈值的次数、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量；
其中，若AI/ML模型执行联邦学习，所述第一消息中请求的参数还包括：联邦学习群信息，所述联邦学习群信息包括下述至少一项：用于指示分析的 联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
相应的，所述第二消息中还包括下述至少一项：用于指定分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
相应的，所述分析信息还包括下述至少一项：用于指定分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、指示提供AI/ML模型或者参与联邦学习的各个应用标识。
在此需要说明的是，本公开提供的订阅网络中模型传输状态分析装置，能够实现图4、图7所示方法实施例所实现的所有方法步骤，且能够达到相同的技术效果，在此不再对本实施例中与方法实施例相同的部分及有益效果进行具体赘述。
图11为本公开又一实施例提供的订阅网络中模型传输状态分析装置的结构示意图，如图11所示，本公开实施例提供的订阅网络中模型传输状态分析装置应用于NWDAF，则本实施例提供的订阅网络中模型传输状态分析装置1100包括：
接收单元1101，用于直接或通过网络能力开放功能NEF接收应用功能AF发送的第一消息；其中，所述第一消息用于请求订阅网络中人工智能/机器学习AI/ML模型传输状态的分析信息；
发送单元1102，用于根据所述第一消息中请求的参数，向5G核心网的其他网络功能5GC NF(s)发送第二消息，所述第二消息用于采集用于分析网络中AI/ML模型传输状态的数据；
分析单元1103，还用于接收5G核心网的其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据，并对AI/ML模型传输状态的数据进行分析，得到AI/ML模型传输状态的分析信息；
其中，所述分析信息用于通过AF调整网络策略参数和/或应用层模型的信息。
可选地，所述第一消息中请求的参数包括下述至少一项：网络数据分析标识、接收AI/ML模型的一个用户设备UE或一组UE的标识或满足分析条件的任意UE、使用AI/ML模型的应用的标识、AI/ML模型传输的区域、指 示传输AI/ML模型服务质量流的协议数据单元PDU会话的网络切片、指示传输AI/ML模型服务质量流的PDU会话的数据网络、AI/ML模型传输的时间段、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于指示传输AI/ML模型的服务质量流的服务质量要求和/或用于指示传输AI/ML模型的特定的服务质量要求；
所述第二消息中包括下述至少一项：使用AI/ML模型的UE的当前位置、使用AI/ML模型的应用的标识、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量、数据采集时间、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于传输AI/ML模型服务质量流的PDU会话的网络切片、用于传输AI/ML模型服务质量流的PDU会话的数据网络、用于所述AF的服务流程；
所述分析信息包括下述至少一项：用于传输AI/ML模型服务质量流的PDU会话的网络切片、使用AI/ML模型的应用的标识、使用AI/ML模型的区域信息、分析结果的有效时间、提供AI/ML模型传输的用户面功能UPF、用于传输AI/ML模型服务质量流的PDU会话的数据网络名称、AI/ML传输模型的大小、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、达到服务质量流在AI/ML模型传输的时间段内异常释放的报告阈值的次数、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量；
其中，若AI/ML模型执行联邦学习，所述第一消息中请求的参数还包括：联邦学习群信息，所述联邦学习群信息包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
相应的，所述第二消息中还包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
相应的，所述分析信息还包括下述至少一项：用于指定分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、指示提供AI/ML模型或者参与联邦学习的各个应用标识。
在此需要说明的是，本公开提供的订阅网络中模型传输状态分析装置，能够实现图4、图7方法实施例所实现的所有方法步骤，且能够达到相同的技术效果，在此不再对本实施例中与方法实施例相同的部分及有益效果进行具体赘述。
需要说明的是，本公开实施例对单元的划分是示意性的，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式。另外，在本公开各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个处理器可读取存储介质中。基于这样的理解，本公开的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)或处理器(processor)执行本公开各个实施例方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(Read-Only Memory，ROM)、随机存取存储器(Random Access Memory，RAM)、磁碟或者光盘等各种可以存储程序代码的介质。
本公开实施例还提供一种处理器可读存储介质。处理器可读存储介质存储有计算机程序，计算机程序用于使处理器执行上述任一种方法实施例。
其中，处理器可读存储介质可以是处理器能够存取的任何可用介质或数据存储设备，包括但不限于磁性存储器(例如软盘、硬盘、磁带、磁光盘(MO)等)、光学存储器(例如CD、DVD、BD、HVD等)、以及半导体存储器(例如ROM、EPROM、EEPROM、非易失性存储器(NAND FLASH)、固态硬 盘(SSD))等。
本领域内的技术人员应明白，本公开的实施例可提供为方法、系统、或计算机程序产品。因此，本公开可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且，本公开可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器和光学存储器等)上实施的计算机程序产品的形式。
本公开是参照根据本公开实施例的方法、设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机可执行指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机可执行指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
这些处理器可执行指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的处理器可读存储器中，使得存储在该处理器可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。
这些处理器可执行指令也可装载到计算机或其他可编程数据处理设备上，使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。
显然，本领域的技术人员可以对本公开进行各种改动和变型而不脱离本公开的精神和范围。这样，倘若本公开的这些修改和变型属于本公开权利要求及其等同技术的范围之内，则本公开也意图包含这些改动和变型在内。

Claims (23)

  1. 一种订阅网络中模型传输状态分析方法，其特征在于，所述方法应用于应用功能AF，所述方法包括：
    直接或通过网络能力开放功能NEF向网络数据分析功能NWDAF发送第一消息；其中，所述第一消息用于请求订阅网络中人工智能/机器学习AI/ML模型传输状态的分析信息；
    直接或通过所述NEF接收所述NWDAF发送的AI/ML模型传输状态的分析信息，所述分析信息是所述NWDAF根据接收5G核心网的其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据确定的；
    其中，所述分析信息用于调整网络策略参数和/或应用层模型的信息。
  2. 根据权利要求1所述的方法，其特征在于，所述AI/ML模型传输状态的数据是所述NWDAF根据接收到的所述第一消息中请求的参数通过向所述5GC NF(s)发送第二消息得到的，所述第二消息用于采集用于分析网络中AI/ML模型传输状态的数据。
  3. 根据权利要求2所述的方法，其特征在于，所述第一消息中请求的参数包括下述至少一项：网络数据分析标识、接收AI/ML模型的一个用户设备UE或一组UE的标识或满足分析条件的任意UE、使用AI/ML模型的应用的标识、AI/ML模型传输的区域、指示传输AI/ML模型服务质量流的协议数据单元PDU会话的网络切片、指示传输AI/ML模型服务质量流的PDU会话的数据网络、AI/ML模型传输的时间段、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于指示传输AI/ML模型的服务质量流的服务质量要求和/或用于指示传输AI/ML模型的特定的服务质量要求；
    所述第二消息中包括下述至少一项：使用AI/ML模型的UE的当前位置、使用AI/ML模型的应用的标识、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量、数据采集时间、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、 AI/ML传输模型的大小、用于传输AI/ML模型服务质量流的PDU会话的网络切片、用于传输AI/ML模型服务质量流的PDU会话的数据网络、用于所述AF的服务流程；
    所述分析信息包括下述至少一项：用于传输AI/ML模型服务质量流的PDU会话的网络切片、使用AI/ML模型的应用的标识、使用AI/ML模型的区域信息、分析结果的有效时间、提供AI/ML模型传输的用户面功能UPF、用于传输AI/ML模型服务质量流的PDU会话的数据网络名称、AI/ML传输模型的大小、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、达到服务质量流在AI/ML模型传输的时间段内异常释放的报告阈值的次数、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量；
    其中，若AI/ML模型执行联邦学习，所述第一消息中请求的参数还包括：联邦学习群信息，所述联邦学习群信息包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
    相应的，所述第二消息中还包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
    相应的，所述分析信息还包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、指示提供AI/ML模型或者参与联邦学习的各个应用标识。
  4. 根据权利要求1-3任一项所述的方法，其特征在于，在接收到所述分析信息后，所述方法还包括：
    根据所述分析信息，直接或通过所述NEF向策略控制功能PCF发送第一请求；
    其中，所述第一请求用于请求更新用于AI/ML模型传输的网络策略参数；所述网络策略参数用于优化AI/ML模型传输状态。
  5. 根据权利要求4所述的方法，其特征在于，所述根据所述分析信息，直接或通过所述NEF向策略控制功能PCF发送第一请求，包括：
    根据所述分析信息中传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量、达到服务质量流在AI/ML模型传输的时间段内异常释放的报告阈值的次数中的至少一项，确定传输AI/ML模型的新的服务质量参数；所述新的服务质量参数包括下述至少一项：5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级；
    根据所述分析信息中的传输AI/ML模型的应用的标识、使用AI/ML模型的区域信息、使用AI/ML模型的应用服务的IP地址信息、用于传输AI/ML模型服务质量流的PDU会话的网络切片、用于传输AI/ML模型服务质量流的PDU会话的数据网络名称，确定传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息；或者，若AI/ML模型执行联邦学习，根据所述分析信息中的指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、指示提供AI/ML模型或者参与联邦学习的各个应用标识，确定传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息；
    根据传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息，确定数据网络接入标识DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息，所述DNAI对应的UE(s)和各个AF的区域信息及地址信息用于提供优化AI/ML模型传输状态的路径；
    将所述新的服务质量参数、所述DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息作为所述第一请求中的参数直接或通过所述NEF发送给PCF。
  6. 根据权利要求5所述的方法，其特征在于，所述根据传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息，确定数据网络接入标识DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息，包括：
    根据传输AI/ML模型的UE(s)和各个AF的区域信息及地址信息，判断当前路由路径是否不佳；
    若当前路由路径不佳，则根据传输AI/ML模型的UE(s)和各个AF的地址信息以及UE(s)的区域信息，确定传输AI/ML模型中传输双方的目的地址；
    根据所述目的地址，确定最近的路径；
    根据最近的路径，确定DNAI以及所述DNAI对应的UE(s)和各个AF的区域信息及地址信息。
  7. 根据权利要求5所述的方法，其特征在于，所述第一请求，具体用于：
    请求PCF根据所述新的服务质量参数调整PCC规则中的5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级并指示所述PCF将调整后的第一更新结果直接或通过NEF反馈；其中，所述第一更新结果是PCF根据所述新的服务质量参数调整PCC规则的结果确定的；
    相应的，所述方法还包括：
    直接或通过NEF接收PCF发送的所述第一更新结果，所述第一更新结果包括所述第一请求被接受或所述第一请求被拒绝。
  8. 根据权利要求5所述的方法，其特征在于，所述第一请求具体用于：
    请求PCF确定会话管理功能网元SMF是否需要更新会话管理策略，若确定SMF需要更新会话管理策略，则确定PCF向SMF发送第二请求，所述第二请求中请求的参数包括下述至少一项：DNAI、流量导向政策标识符、流量路线信息；所述第二请求用于SMF根据新的会话管理策略确定选择的用户面功能UPF并提供相应的DNAI、流量导向政策标识符、流量路线信息；
    相应的，所述方法还包括：
    直接或通过NEF接收PCF发送的第二更新结果，所述第二更新结果是PCF根据SMF发送的新的会话管理策略是否更新UPF路径确定的；
    其中，所述第二更新结果包括所述第一请求被接受或被拒绝。
  9. 根据权利要求1-3任一项所述的方法，其特征在于，在接收到所述分析信息后，所述方法还包括：
    根据所述分析信息，调整应用层模型的信息，所述应用层模型的信息包括下述至少一项：模型压缩、模型大小、模型传输时间段、模型编解码；所述应用层模型的信息用于更新服务质量参数；
    根据调整后的应用层模型的信息，确定新的服务质量参数，所述新的服务质量参数包括：5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级；
    直接或通过所述NEF向策略控制功能PCF发送第三请求；
    其中，所述第三请求中请求的参数包括所述新的服务质量参数，所述第三请求用于请求更新服务质量参数。
  10. 根据权利要求9所述的方法，其特征在于，所述第三请求，具体用于：
    请求PCF根据所述新的服务质量参数调整PCC规则中的5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、优先级；
    相应的，所述方法还包括：
    直接或通过NEF接收PCF发送的第三更新结果，第三更新结果是PCF基于调整PCC规则的结果确定的，所述第三更新结果包括所述第三请求被接受或所述第三请求被拒绝。
  11. 根据权利要求9所述的方法，其特征在于，在所述调整应用层模型的信息之后，所述方法还包括：
    将调整后的应用层模型的信息中的模型压缩、模型大小以及模型编解码直接发送给PCF，所述调整后的应用层模型的信息用于支持PCF调整PCC规则中的5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链路方向最低比特率、服务质量流的优先级；
    相应的，所述方法还包括：
    接收PCF发送的第四更新结果，第四更新结果是PCF根据调整后的应用层模型的信息调整PCC规则的结果确定的，所述第四更新结果包括所述第三请求被接受或所述第三请求被拒绝。
  12. 根据权利要求9所述的方法，其特征在于，在所述调整应用层模型的信息之后，所述方法还包括：
    将调整后的应用层模型的信息中的模型传输时间直接发送给PCF，所述调整后的应用层模型的信息中的模型传输时间用于支持PCF调整PCC规则中的门状态参数；所述门状态参数用于支持SMF根据门状态中的传输开始时间和传输结束时间来更新会话管理策略；
    相应的，所述方法还包括：
    接收PCF发送的第五更新结果，第五更新结果是PCF通过接收SMF发送的新的会话管理策略的结果确定的，所述第五更新结果包括所述第三请求被接受或所述第三请求被拒绝。
  13. 一种订阅网络中模型传输状态分析方法，其特征在于，所述方法应用于网络数据分析功能NWDAF，所述方法包括：
    直接或通过网络能力开放功能NEF接收应用功能AF发送的第一消息；其中，所述第一消息用于请求订阅网络中人工智能/机器学习AI/ML模型传输状态的分析信息；
    根据所述第一消息中请求的参数，向5G核心网的其他网络功能5GC NF(s)发送第二消息，所述第二消息用于采集用于分析网络中AI/ML模型传输状态的数据；
    接收5G核心网的其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据，并对AI/ML模型传输状态的数据进行分析，得到AI/ML模型传输状态的分析信息；
    其中，所述分析信息用于通过AF调整网络策略参数和/或应用层模型的信息。
  14. 根据权利要求13所述的方法，其特征在于，所述第一消息中请求的参数包括下述至少一项：网络数据分析标识、接收AI/ML模型的一个用户设备UE或一组UE的标识或满足分析条件的任意UE、使用AI/ML模型的应用的标识、AI/ML模型传输的区域、指示传输AI/ML模型服务质量流的协议数 据单元PDU会话的网络切片、指示传输AI/ML模型服务质量流的PDU会话的数据网络、AI/ML模型传输的时间段、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于指示传输AI/ML模型的服务质量流的服务质量要求和/或用于指示传输AI/ML模型的特定的服务质量要求；
    所述第二消息中包括下述至少一项：使用AI/ML模型的UE的当前位置、使用AI/ML模型的应用的标识、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量、数据采集时间、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于传输AI/ML模型服务质量流的PDU会话的网络切片、用于传输AI/ML模型服务质量流的PDU会话的数据网络、用于所述AF的服务流程；
    所述分析信息包括下述至少一项：用于传输AI/ML模型服务质量流的PDU会话的网络切片、使用AI/ML模型的应用的标识、使用AI/ML模型的区域信息、分析结果的有效时间、提供AI/ML模型传输的用户面功能UPF、用于传输AI/ML模型服务质量流的PDU会话的数据网络名称、AI/ML传输模型的大小、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、达到服务质量流在AI/ML模型传输的时间段内异常释放的报告阈值的次数、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量；
    其中，若AI/ML模型执行联邦学习，所述第一消息中请求的参数还包括：联邦学习群信息，所述联邦学习群信息包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
    相应的，所述第二消息中还包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
    相应的，所述分析信息还包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、指示提供AI/ML模型或者参与联邦学习的各个应用标识。
  15. 一种订阅网络中模型传输状态分析装置，其特征在于，所述装置包括存储器，收发机，处理器：
    存储器，用于存储计算机程序；收发机，用于在所述处理器的控制下收发数据；处理器，用于读取所述存储器中的计算机程序并执行以下操作：
    直接或通过网络能力开放功能NEF向网络数据分析功能NWDAF发送第一消息；其中，所述第一消息用于请求订阅网络中人工智能/机器学习AI/ML模型传输状态的分析信息；
    直接或通过所述NEF接收所述NWDAF发送的AI/ML模型传输状态的分析信息，所述分析信息是所述NWDAF根据接收5G核心网的其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据确定的；
    其中，所述分析信息用于调整网络策略参数和/或应用层模型的信息。
  16. 一种订阅网络中模型传输状态分析装置，其特征在于，所述装置包括存储器，收发机，处理器：
    存储器，用于存储计算机程序；收发机，用于在所述处理器的控制下收发数据；处理器，用于读取所述存储器中的计算机程序并执行以下操作：
    直接或通过网络能力开放功能NEF接收应用功能AF发送的第一消息；其中，所述第一消息用于请求订阅网络中人工智能/机器学习AI/ML模型传输状态的分析信息；
    根据所述第一消息中请求的参数，向5G核心网的其他网络功能5GC NF(s)发送第二消息，所述第二消息用于采集用于分析网络中AI/ML模型传输状态的数据；
    接收5G核心网的其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据，并对AI/ML模型传输状态的数据进行分析，得到AI/ML模型传输状态的分析信息；
    其中，所述分析信息用于通过AF调整网络策略参数和/或应用层模型的信息。
  17. 一种订阅网络中模型传输状态分析装置，其特征在于，所述装置包括：
    发送单元，用于直接或通过网络能力开放功能NEF向网络数据分析功能NWDAF发送第一消息；其中，所述第一消息用于请求订阅网络中人工智能/机器学习AI/ML模型传输状态的分析信息；
    分析单元，用于直接或通过所述NEF接收所述NWDAF发送的AI/ML模型传输状态的分析信息，所述分析信息是所述NWDAF根据接收5G核心网的其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据确定的；
    其中，所述分析信息用于调整网络策略参数和/或应用层模型的信息。
  18. 根据权利要求17所述的装置，其特征在于，所述AI/ML模型传输状态的数据是所述NWDAF根据接收到的所述第一消息中请求的参数通过向所述5GC NF(s)发送第二消息得到的，所述第二消息用于采集用于分析网络中AI/ML模型传输状态的数据。
  19. 根据权利要求17或18所述的装置，其特征在于，发送单元，还用于：
    在接收到所述分析信息后，根据所述分析信息，直接或通过所述NEF向策略控制功能PCF发送第一请求；
    其中，所述第一请求用于请求更新用于AI/ML模型传输的网络策略参数；所述网络策略参数用于优化AI/ML模型传输状态。
  20. 根据权利要求17或18所述的装置，其特征在于，所述装置还包括：确定单元；确定单元，用于：
    在接收到所述分析信息后，根据所述分析信息，调整应用层模型的信息，所述应用层模型的信息包括下述至少一项：模型压缩、模型大小、模型传输时间段、模型编解码；所述应用层模型的信息用于更新服务质量参数；
    根据调整后的应用层模型的信息，确定所述新的服务质量参数，所述新的服务质量参数包括：5G服务质量标识符、反射式服务质量控制、传输AI/ML模型的上行链路方向最大比特率、传输AI/ML模型的下行链路方向最大比特率、传输AI/ML模型的上行链路方向最低比特率、传输AI/ML模型的下行链 路方向最低比特率、服务质量流的优先级；
    直接或通过所述NEF向策略控制功能PCF发送第三请求；
    其中，所述第三请求中请求的参数包括所述新的服务质量参数，所述第三请求用于请求更新服务质量参数。
  21. 一种订阅网络中模型传输状态分析装置，其特征在于，所述装置包括：
    接收单元，用于直接或通过网络能力开放功能NEF接收应用功能AF发送的第一消息；其中，所述第一消息用于请求订阅网络中人工智能/机器学习AI/ML模型传输状态的分析信息；
    发送单元，用于根据所述第一消息中请求的参数，向5G核心网的其他网络功能5GC NF(s)发送第二消息，所述第二消息用于采集用于分析网络中AI/ML模型传输状态的数据；
    分析单元，用于接收5G核心网的其他网络功能5GC NF(s)发送的AI/ML模型传输状态的数据，并对AI/ML模型传输状态的数据进行分析，得到AI/ML模型传输状态的分析信息；
    其中，所述分析信息用于通过AF调整网络策略参数和/或应用层模型的信息。
  22. 根据权利要求21所述的装置，其特征在于，所述第一消息中请求的参数包括下述至少一项：网络数据分析标识、接收AI/ML模型的一个用户设备UE或一组UE的标识或满足分析条件的任意UE、使用AI/ML模型的应用的标识、AI/ML模型传输的区域、指示传输AI/ML模型服务质量流的协议数据单元PDU会话的网络切片、指示传输AI/ML模型服务质量流的PDU会话的数据网络、AI/ML模型传输的时间段、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于指示传输AI/ML模型的服务质量流的服务质量要求和/或用于指示传输AI/ML模型的特定的服务质量要求；
    所述第二消息中包括下述至少一项：使用AI/ML模型的UE的当前位置、使用AI/ML模型的应用的标识、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下 行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量、数据采集时间、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、AI/ML传输模型的大小、用于传输AI/ML模型服务质量流的PDU会话的网络切片、用于传输AI/ML模型服务质量流的PDU会话的数据网络、用于所述AF的服务流程；
    所述分析信息包括下述至少一项：用于传输AI/ML模型服务质量流的PDU会话的网络切片、使用AI/ML模型的应用的标识、使用AI/ML模型的区域信息、分析结果的有效时间、提供AI/ML模型传输的用户面功能UPF、用于传输AI/ML模型服务质量流的PDU会话的数据网络名称、AI/ML传输模型的大小、AI/ML模型传输的时长、AI/ML模型传输的开始时间戳、AI/ML模型传输的结束时间戳、传输AI/ML模型的服务质量流标识符、传输AI/ML模型的上行链路方向比特率及传输AI/ML模型的下行链路方向比特率、AI/ML模型的上行链路方向分组延迟及AI/ML模型的下行链路方向分组延迟、服务质量流在AI/ML模型传输的时间段内异常释放的数量、达到服务质量流在AI/ML模型传输的时间段内异常释放的报告阈值的次数、AI/ML模型的分组传输数量、AI/ML模型的分组重传数量；
    其中，若AI/ML模型执行联邦学习，所述第一消息中请求的参数还包括：联邦学习群信息，所述联邦学习群信息包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
    相应的，所述第二消息中还包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、参加联邦学习的应用标识；
    相应的，所述分析信息还包括下述至少一项：用于指示分析的联邦学习组的标识、参加联邦学习的UE标识或UE(s)标识、指示提供AI/ML模型或者参与联邦学习的各个应用标识。
  23. 一种处理器可读存储介质，其特征在于，所述处理器可读存储介质存储有计算机程序，所述计算机程序用于使所述处理器执行权利要求1至14任一项所述的方法。
PCT/CN2022/127034 2021-11-24 2022-10-24 订阅网络中模型传输状态分析方法、装置及可读存储介质 Ceased WO2023093416A1 (zh)

Priority Applications (4)

Application Number Priority Date Filing Date Title
JP2024529997A JP7743629B2 (ja) 2021-11-24 2022-10-24 サブスクリプションネットワークにおけるモデル伝送状態の解析方法、装置及び読み取り可能な記憶媒体
US18/701,933 US20240414063A1 (en) 2021-11-24 2022-10-24 Method and apparatus for subscribing to analytics of model transfer status in network, and readable storage medium
EP22897499.4A EP4440065A4 (en) 2021-11-24 2022-10-24 Method and apparatus for model transmission state analysis in subscription network, and readable storage medium
KR1020247018950A KR20240099444A (ko) 2021-11-24 2022-10-24 네트워크에서의 모델 전송 상태 분석을 구독하는 방법, 장치 및 판독 가능 저장매체

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111407051.2 2021-11-24
CN202111407051.2A CN116170820B (zh) 2021-11-24 2021-11-24 订阅网络中模型传输状态分析方法、装置及可读存储介质

Publications (1)

Publication Number Publication Date
WO2023093416A1 true WO2023093416A1 (zh) 2023-06-01

Family

ID=86409974

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/127034 Ceased WO2023093416A1 (zh) 2021-11-24 2022-10-24 订阅网络中模型传输状态分析方法、装置及可读存储介质

Country Status (6)

Country Link
US (1) US20240414063A1 (zh)
EP (1) EP4440065A4 (zh)
JP (1) JP7743629B2 (zh)
KR (1) KR20240099444A (zh)
CN (1) CN116170820B (zh)
WO (1) WO2023093416A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2025150877A1 (en) * 2024-01-11 2025-07-17 Samsung Electronics Co., Ltd. Method and apparatus for supporting energy usage in communication network

Families Citing this family (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20230244995A1 (en) * 2022-01-11 2023-08-03 Electronics And Telecommunications Research Institute Apparatus and method for evaluating machine learning model
KR20240096197A (ko) * 2022-12-19 2024-06-26 삼성전자주식회사 무선 통신 시스템에서 인공지능 모델을 이용하여 트래픽을 처리하기 위한 방법 및 장치
US20240334222A1 (en) * 2023-03-31 2024-10-03 Ofinno, Llc Packet Performance Measurement
CN116723537A (zh) * 2023-06-07 2023-09-08 中国联合网络通信集团有限公司 一种基于nwdaf的数据分析方法、装置及可读存储介质
CN119172816B (zh) * 2023-06-20 2026-01-13 中国电信股份有限公司 间接路径切换方法、装置、电子设备及存储介质
CN119233325A (zh) * 2023-06-29 2024-12-31 华为技术有限公司 一种QoS参数配置方法和通信装置
CN119449611A (zh) * 2023-08-02 2025-02-14 维沃软件技术有限公司 策略协商方法、策略生成方法及设备
CN119485380A (zh) * 2023-08-11 2025-02-18 华为技术有限公司 一种通信方法及装置
CN118827438A (zh) * 2023-09-26 2024-10-22 中国移动通信有限公司研究院 一种信息处理方法、装置、设备及可读存储介质
US20250141950A1 (en) * 2023-10-30 2025-05-01 Verizon Patent And Licensing Inc. Systems and methods for receiving service identifiers for quality of service decisions
CN120238912A (zh) * 2023-12-29 2025-07-01 华为技术有限公司 通信方法及相关装置
CN120935610A (zh) * 2024-05-08 2025-11-11 中兴通讯股份有限公司 信息传输方法、通信节点、存储介质及计算机程序产品
WO2025236136A1 (zh) * 2024-05-11 2025-11-20 北京小米移动软件有限公司 通信方法、实体、通信系统及存储介质
CN121418886A (zh) * 2024-07-24 2026-01-27 华为技术有限公司 一种通信方法及相关装置
WO2026077736A1 (en) * 2024-10-07 2026-04-16 Nokia Technologies Oy Nw initiated pdu session management for ml model data transfer

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20210014141A1 (en) * 2019-07-12 2021-01-14 Verizon Patent And Licensing Inc. System and method of closed loop analytics for network automation
US20210021494A1 (en) * 2019-10-03 2021-01-21 Intel Corporation Management data analytics
WO2021032495A1 (en) * 2019-08-16 2021-02-25 Telefonaktiebolaget Lm Ericsson (Publ) Methods, apparatus and machine-readable media relating to machine-learning in a communication network
US20210099367A1 (en) * 2019-09-27 2021-04-01 Samsung Electronics Co., Ltd. Method and apparatus for detecting service and analyzing service characteristic using nwdaf in mobile communication system
WO2021141291A1 (ko) * 2020-01-06 2021-07-15 삼성전자 주식회사 무선 통신 시스템에서 네트워크 트래픽을 수집하는 방법 및 장치

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7233583B2 (en) * 2004-06-28 2007-06-19 Nokia Corporation Method and apparatus providing context transfer for inter-BS and inter-PCF handoffs in a wireless communication system
CN110300006B (zh) * 2018-03-21 2022-10-21 中国移动通信有限公司研究院 数据处理方法及装置、功能实体及存储介质
WO2020049181A1 (en) * 2018-09-07 2020-03-12 NEC Laboratories Europe GmbH System and method for network automation in slice-based network using reinforcement learning
CN111770490B (zh) * 2019-04-02 2022-08-05 大唐移动通信设备有限公司 一种确定终端行为分析的方法和设备
CN110677299B (zh) * 2019-09-30 2024-06-28 中兴通讯股份有限公司 网络数据采集方法、装置和系统
JP7457146B2 (ja) * 2020-02-27 2024-03-27 華為技術有限公司 モバイルネットワークにおける分析の生成及び消費
CN113676846B (zh) * 2020-05-15 2023-04-11 大唐移动通信设备有限公司 一种传输方式确定方法、装置、设备及存储介质
CN112423382A (zh) * 2020-11-09 2021-02-26 江苏第二师范学院（江苏省教育科学研究院） 基于5g网络的模型管理方法及使用nrf的注册和更新方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20210014141A1 (en) * 2019-07-12 2021-01-14 Verizon Patent And Licensing Inc. System and method of closed loop analytics for network automation
WO2021032495A1 (en) * 2019-08-16 2021-02-25 Telefonaktiebolaget Lm Ericsson (Publ) Methods, apparatus and machine-readable media relating to machine-learning in a communication network
US20210099367A1 (en) * 2019-09-27 2021-04-01 Samsung Electronics Co., Ltd. Method and apparatus for detecting service and analyzing service characteristic using nwdaf in mobile communication system
US20210021494A1 (en) * 2019-10-03 2021-01-21 Intel Corporation Management data analytics
WO2021141291A1 (ko) * 2020-01-06 2021-07-15 삼성전자 주식회사 무선 통신 시스템에서 네트워크 트래픽을 수집하는 방법 및 장치

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
"3rd Generation Partnership Project; Technical Specification Group Services and System Aspects; Study on traffic characteristics and performance requirements for AI/ML model transfer in 5GS (Release 18)", 3GPP STANDARD; TECHNICAL REPORT; 3GPP TR 22.874, 3RD GENERATION PARTNERSHIP PROJECT (3GPP), MOBILE COMPETENCE CENTRE ; 650, ROUTE DES LUCIOLES ; F-06921 SOPHIA-ANTIPOLIS CEDEX ; FRANCE, vol. SA WG1, no. V1.0.0, 17 March 2021 (2021-03-17), Mobile Competence Centre ; 650, route des Lucioles ; F-06921 Sophia-Antipolis Cedex ; France , pages 1 - 103, XP052000078 *
See also references of EP4440065A4

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2025150877A1 (en) * 2024-01-11 2025-07-17 Samsung Electronics Co., Ltd. Method and apparatus for supporting energy usage in communication network

Also Published As

Publication number Publication date
JP2024539470A (ja) 2024-10-28
CN116170820B (zh) 2026-03-17
EP4440065A4 (en) 2025-04-02
CN116170820A (zh) 2023-05-26
US20240414063A1 (en) 2024-12-12
KR20240099444A (ko) 2024-06-28
EP4440065A1 (en) 2024-10-02
JP7743629B2 (ja) 2025-09-24

Similar Documents

Publication Publication Date Title
WO2023093416A1 (zh) 订阅网络中模型传输状态分析方法、装置及可读存储介质
WO2023093430A1 (zh) 基于模型传输状态分析的信息处理方法及装置
CA3112926C (en) Slice information processing method and apparatus
CN111565418B (zh) 一种o-ran与mec的通信方法及系统
US20260121988A1 (en) Management of traffic over a communication channel
US10084665B1 (en) Resource selection using quality prediction
CN106534333B (zh) 一种基于mec和mcc的双向选择计算卸载方法
WO2021012697A1 (zh) 应用mos模型的训练方法、设备及系统
WO2009138000A1 (zh) 一种网络流量控制方法、装置及系统
CN109451534B (zh) 一种用于5G系统会话管理中QoS流的动态控制方法和装置
CN114422451A (zh) 一种网络流量识别方法及相关设备
CN108886703A (zh) 在5g系统中的情境感知qos/qoe策略供应和适配
CN103634409A (zh) 实现移动互联网应用永远在线的方法及系统
CN110769466A (zh) 信息发送和接收、业务和网络质量调控方法、装置及设备
Zhang et al. MoWIE: toward systematic, adaptive network information exposure as an enabling technique for cloud-based applications over 5G and beyond
CN118138479A (zh) 一种模型训练方法、系统、通信实体及存储介质
CN110289971B (zh) 共享业务管理方法及系统
CN112099871A (zh) 一种服务质量配置方法及装置
CN107786371B (zh) 一种数据的加速方法、装置及存储介质
CN114554554A (zh) 信息传输方法、装置、pcf实体、终端及业务运营支撑系统
CN119449712A (zh) 流特定的拥塞处理
WO2024032031A1 (zh) 一种数据分析方法及装置
CN116800614A (zh) 网络切片分配方法及装置、网络切片申请方法、电子设备
CN117061420A (zh) 路由路径确定方法、装置、设备、介质以及程序产品
Pencheva et al. Programmability of policy control at the edge of the mobile network

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22897499

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2024529997

Country of ref document: JP

Kind code of ref document: A

WWE Wipo information: entry into national phase

Ref document number: 1020247018950

Country of ref document: KR

WWE Wipo information: entry into national phase

Ref document number: 2022897499

Country of ref document: EP

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2022897499

Country of ref document: EP

Effective date: 20240624