WO2023143255A1 - 一种通信方法及装置 - Google Patents

一种通信方法及装置 Download PDF

Info

Publication number
WO2023143255A1
WO2023143255A1 PCT/CN2023/072639 CN2023072639W WO2023143255A1 WO 2023143255 A1 WO2023143255 A1 WO 2023143255A1 CN 2023072639 W CN2023072639 W CN 2023072639W WO 2023143255 A1 WO2023143255 A1 WO 2023143255A1
Authority
WO
WIPO (PCT)
Prior art keywords
terminal device
network element
device group
data
qos
Prior art date
Application number
PCT/CN2023/072639
Other languages
English (en)
French (fr)
Inventor
封召
辛阳
王远
Original Assignee
华为技术有限公司
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
Priority claimed from CN202210326157.8A external-priority patent/CN116567608A/zh
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Publication of WO2023143255A1 publication Critical patent/WO2023143255A1/zh

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W8/00—Network data management
    • H04W8/18—Processing of user or subscriber data, e.g. subscribed services, user preferences or user profiles; Transfer of user or subscriber data

Definitions

  • the embodiments of the present application relate to the field of wireless communication, and in particular, to a communication method and device.
  • Federated learning is a machine learning framework that can effectively help multiple users perform data usage and machine learning modeling while meeting the requirements of user privacy protection, data security, and government regulations.
  • federated learning can effectively solve the problem of data islands, perform joint modeling on the basis of not sharing user data, and then technically break data islands and realize artificial intelligence (AI) collaboration.
  • AI artificial intelligence
  • QoS quality of service
  • the present application provides a communication method and device for improving the efficiency of federated learning model training.
  • the present application provides a communication method, the method includes: a first network element sends a first request message to a second network element, and the first request message is used to request data of a terminal device group, wherein the The terminal devices in the terminal device group are terminal devices participating in horizontal federated learning; the first network element receives the data of the terminal device group from the second network element, and the data of the terminal device group includes the measurement of QoS parameters Aggregation information of results; the first network element adjusts the QoS parameters of the first terminal device in the terminal device group according to the data of the terminal device group; the first network element sends the adjusted QoS parameters to the policy control network element Describe the QoS parameters of the first terminal device.
  • the first network element subscribes the data of the terminal equipment group to the second network element, the second network element sends the data of the terminal equipment group to the first network element, and the first network element adjusts the terminal equipment according to the data of the terminal equipment group
  • the QoS parameters of at least one terminal device in the group can improve the efficiency of federated learning model training.
  • the first request message includes indication information, and the indication information is used to indicate the measurement result of the QoS parameter of the feedback terminal equipment;
  • the data of the terminal equipment group also includes A measurement result of a QoS parameter of a second terminal device, where the second terminal device includes the first terminal device.
  • the second network element can also send the measurement results of the QoS parameters of the terminal devices in the terminal device group to the first network element.
  • the first request message further includes a feedback condition, and the feedback condition is used to indicate a condition to be met for feeding back the measurement result of the QoS parameter of the terminal device; QoS parameters The measurement result of satisfies the feedback condition.
  • the second network element can feed back the measurement results of the QoS parameters of the terminal devices that meet the feedback conditions.
  • the aggregated information includes an aggregated bit rate, and the aggregated bit rate is used to characterize the sum of the bit rates of the QoS flows of the terminal devices in the terminal device group; in the first network element When adjusting the QoS parameters of the first terminal device in the terminal device group according to the data of the terminal device group, if the aggregated bit rate is greater than the contracted maximum bit rate of the terminal device group, the first network element according to the Adjusting the bit rate of the QoS flow of the first terminal device according to the data of the terminal device group.
  • the first network element may adjust the bit rate of the QoS flow of at least one terminal device according to the data of the terminal device group, the at least one terminal device includes the first terminal device, and when adjusting the at least one terminal device After the bit rate of the QoS flow of the device, the sum of the bit rates of the QoS flows of each terminal device in the terminal device group is less than or equal to the contracted maximum bit rate.
  • the aggregated information includes statistical values of transmission delays of terminal devices in the terminal device group; the first network element adjusts the terminal device according to the data of the terminal device group When the QoS parameter of the first terminal device in the group is set, if the statistical value of the transmission delay is greater than the first preset threshold, the first network element adjusts the QoS parameter of the first terminal device according to the data of the terminal device group.
  • the bit rate of the QoS flow is set.
  • the statistical value of the transmission delays of the terminal devices in the terminal device group may be a variance determined according to the transmission delays of the terminal devices in the terminal device group.
  • the statistical value of the transmission delays of the terminal devices in the terminal device group refers to the variance determined by the weighted transmission delays or differential transmission delays of the terminal devices in the terminal device group.
  • the first terminal device is a terminal device with a transmission delay greater than a second preset threshold or less than a third preset threshold, where the second preset threshold is greater than the third preset threshold Set the threshold.
  • the first request message further includes the first preset threshold and/or the second preset threshold.
  • the first request message includes at least one of identifiers of terminal devices included in the terminal device group, identifiers of the terminal device group, identification information of analysis types, and types of the aggregation information. A sort of.
  • the first request message further includes at least one of the following: a trigger condition for sending the aggregation information; an application identifier, used to indicate an application corresponding to a measurement result of the QoS parameter of the terminal device.
  • the trigger condition for sending aggregated information may include periodically feeding aggregated information, or feeding aggregated information that meets a preset condition.
  • the preset conditions here may include that the aggregated bit rate is greater than the contracted maximum bit rate of the terminal device group, and/or the statistical value of the transmission delay of the terminal devices in the terminal device group is greater than a first preset threshold, and the like.
  • the first network element is an application function network element
  • the second network element is a data analysis function network element or a policy control network element.
  • the present application provides a communication method, the method includes: a second network element receives a first request message from a first network element, and the first request message is used to request data of a terminal device group, wherein the The terminal devices in the terminal device group are terminal devices participating in horizontal federated learning; the second network element sends the data of the terminal device group, and the data of the terminal device group includes aggregation information of measurement results of QoS parameters.
  • the first network element subscribes the data of the terminal equipment group to the second network element, and the second network element sends the data of the terminal equipment group to the first network element, so that the first network element adjusts the data according to the data of the terminal equipment group
  • the QoS parameter of at least one terminal device in the terminal device group can further improve the efficiency of federated learning model training.
  • the first request message includes indication information, and the indication information is used to indicate the measurement result of the QoS parameter of the feedback terminal equipment;
  • the data of the terminal equipment group also includes A measurement result of a QoS parameter of a second terminal device, where the second terminal device includes the first terminal device.
  • the second network element can also send the measurement results of the QoS parameters of the terminal devices in the terminal device group to the first network element.
  • the first request message further includes a feedback condition, and the feedback condition is used to indicate a condition to be met for feeding back the measurement result of the QoS parameter of the terminal device; The measurement result of the QoS parameter satisfies the feedback condition.
  • the second network element can feed back the measurement results of the QoS parameters of the terminal devices that meet the feedback conditions.
  • the aggregated information includes an aggregated bit rate, where the aggregated bit rate is used to represent a sum of bit rates of QoS flows of terminal devices in the terminal device group.
  • the aggregation information includes statistical values of transmission delays of terminal devices in the terminal device group.
  • the first request message includes at least one of an identifier of a terminal device in the terminal device group, an identifier of the terminal device group, identification information of an analysis type, and a type of the aggregation information. kind.
  • the first request message further includes at least one of the following: a trigger condition for sending the aggregation information; an application identifier, used to indicate an application corresponding to a measurement result of the QoS parameter of the terminal device.
  • it further includes: the second network element sending a second request message to a user plane network element, where the second request message is used to request data of the terminal device group; The element receives the data of the terminal device group from the user plane network element.
  • the second network element can directly obtain the data of the terminal device group from the user plane network element.
  • the method further includes: the second network element sending a third request message to the user plane network element, where the third request message is used to request the QoS parameters of the terminal devices in the terminal device group measurement results; the second network element receives the measurement results of the QoS parameters of the terminal devices in the terminal device group from the user plane network element; the second network element receives the QoS parameter measurement results of the terminal devices in the terminal device group Measurement results of QoS parameters and said first request message determine data of said group of terminal devices.
  • the second network element can aggregate the measurement results of the QoS parameters of the terminal equipment obtained from the user plane network elements to obtain the data of the terminal equipment group.
  • the second network element is a policy control network element; the method further includes: the second network element receives each terminal included in the terminal device group from the first network element The bit rates of the QoS flows corresponding to the devices respectively; the second network element receives the contracted maximum bit rate of the terminal equipment group from the third network element, and the third network element is a unified data storage network element or a unified data management network element; the second network element according to the contract maximum bit of the terminal equipment group rate and the bit rate of the QoS stream corresponding to each terminal device included in the terminal device group, and determine that the sum of the bit rates of the QoS streams corresponding to each terminal device included in the terminal device group is less than the subscription rate of the terminal device group The maximum bit rate; the second network element generates PCC rules corresponding to each terminal device included in the terminal device group; the second network element sends the PCC rule to a session management network element.
  • the second network element is a policy control network element; the method further includes: the second network element receives each terminal included in the terminal device group from the first network element The bit rate reference ranges of the QoS flows corresponding to the devices respectively; the second network element receives the contracted maximum bit rate of the terminal device group from the third network element, and the third network element is a unified data storage network element or a unified data storage network element management network element; the second network element determines that the terminal equipment group includes The sum of the lower bounds of the bit rate reference ranges of the QoS streams corresponding to each terminal device in the terminal device group is less than the contracted maximum bit rate of the terminal device group; PCC rules: the second network element sends the PCC rules to the session management network element.
  • the policy control network element can be based on the value of the bit rate of the QoS flow corresponding to each terminal device included in the terminal device group, or the value of the bit rate of the QoS flow corresponding to each terminal device included in the terminal device group
  • the range determines whether the contracted maximum bit rate for the group of end devices is met.
  • the method further includes: the second network element sending second information to the first network element, the second information indicating that corresponding PCC rules.
  • the first network element is an application function network element
  • the second network element is a data analysis function network element or a policy control network element.
  • an embodiment of the present application provides a communication device, the device is a first network element, or a device for realizing the functions of the first network element, and the device includes a processing module and a transceiver module; the transceiver module , for sending a first request message to a second network element, where the first request message is used to request data of a terminal device group, where the terminal devices in the terminal device group are terminal devices participating in horizontal federated learning; from The second network element receives data of the terminal device group, where the data of the terminal device group includes aggregation information of measurement results of QoS parameters; the processing module is configured to adjust the The QoS parameter of the first terminal device in the terminal device group; the transceiver module is configured to send the adjusted QoS parameter of the first terminal device to a policy control network element.
  • the first request message includes indication information, and the indication information indicates the measurement result of the QoS parameter of the feedback terminal device;
  • the data of the terminal device group further includes a measurement result of a QoS parameter of a second terminal device in the terminal device group, where the second terminal device includes the first terminal device.
  • the first request message further includes a feedback condition, and the feedback condition indicates a condition to be satisfied for feeding back the measurement result of the QoS parameter of the terminal device; the QoS parameter of the second terminal device The measurement result of satisfies the feedback condition.
  • the aggregated information includes an aggregated bit rate, and the aggregated bit rate is used to characterize the sum of bit rates of the QoS flows of the terminal devices in the terminal device group;
  • the processing module is configured to When adjusting the QoS parameters of the first terminal device in the terminal device group according to the data of the terminal device group, if the aggregated bit rate is greater than the contracted maximum bit rate of the terminal device group, according to the The data adjusts the bit rate of the QoS flow of the first terminal device.
  • the aggregation information includes statistical values of transmission delays of terminal devices in the terminal device group; the processing module is configured to adjust the terminal according to the data of the terminal device group The first terminal in the equipment group When the QoS parameter of the terminal device is selected, if the statistical value of the transmission delay is greater than the first preset threshold, the bit rate of the QoS flow of the first terminal device is adjusted according to the data of the terminal device group.
  • the first terminal device is a terminal device with a transmission delay greater than a second preset threshold.
  • the first request message further includes the first preset threshold and/or the second preset threshold.
  • the first request message includes at least one of an identifier of a terminal device in the terminal device group, an identifier of the terminal device group, identification information of an analysis type, and a type of the aggregation information. kind.
  • the first request message further includes at least one of the following: a trigger condition for sending the aggregation information; an application identifier, used to indicate an application corresponding to a measurement result of the QoS parameter of the terminal device.
  • the first network element is an application function network element
  • the second network element is a data analysis function network element or a policy control network element.
  • the present application provides a communication device, the device is a second network element, or a device for realizing the functions of the second network element, the device includes a processing module and a transceiver module; the processing module calls the The transceiver module executes: receiving a first request message from the first network element, the first request message is used to request data of a terminal device group, wherein the terminal devices in the terminal device group are terminal devices participating in horizontal federated learning; Sending the data of the terminal device group, where the data of the terminal device group includes aggregation information of measurement results of QoS parameters.
  • the first request message includes indication information, and the indication information indicates the measurement result of the QoS parameter of the feedback terminal device; the data of the terminal device group also includes the first Measurement results of QoS parameters of two terminal devices, where the second terminal device includes the first terminal device.
  • the first request message further includes a feedback condition, and the feedback condition indicates a condition to be satisfied for feeding back the measurement result of the QoS parameter of the terminal device; the QoS parameter of the second terminal device The measurement result of satisfies the feedback condition.
  • the aggregated information includes an aggregated bit rate, where the aggregated bit rate is used to represent a sum of bit rates of QoS flows of terminal devices in the terminal device group.
  • the aggregation information includes statistical values of transmission delays of terminal devices in the terminal device group.
  • the first request message includes at least one of an identifier of a terminal device in the terminal device group, an identifier of the terminal device group, identification information of an analysis type, and a type of the aggregation information. kind.
  • the first request message further includes at least one of the following: a trigger condition for sending the aggregation information; an application identifier, used to indicate an application corresponding to a measurement result of the QoS parameter of the terminal device.
  • the transceiver module is further configured to send a second request message to a user plane network element, where the second request message is used to request the data of the terminal device group; receive from the user plane network element The data of the terminal device group.
  • the transceiver module is further configured to send a third request message to a user plane network element, where the third request message is used to request the measurement of the QoS parameters of the terminal devices in the terminal device group Result; receive the measurement result of the QoS parameter of the terminal device in the terminal device group from the user plane network element; the processing module is further configured to according to the measurement result of the QoS parameter of the terminal device in the terminal device group and the first request message to determine the data of the terminal device group.
  • the first network element is an application function network element
  • the second network element is a data analysis function network element or a policy control network element.
  • the present application provides a communication method, the method includes: applying a functional network element to obtain the information in the terminal device group The transmission delay and local calculation delay of each terminal device.
  • the terminal devices in the terminal device group are terminal devices participating in horizontal federated learning.
  • the application function network element is based on the transmission delay and local calculation delay of each terminal device. delay, adjusting the QoS parameter of the first terminal device in the terminal device group; the application function network element sending the adjusted QoS parameter of the first terminal device to the policy control network element.
  • the application function network element can also not collect data through NWDAF or PCF, but can count or obtain the transmission delay and local calculation delay of each terminal device in the terminal device group, and then determine the terminal device in the terminal device group
  • the QoS parameter adjustment strategy reduces the number of messages exchanged between network elements, reduces the network element load, and improves the QoS parameter adjustment efficiency of terminal equipment, so that the sum of local calculation delay and transmission delay of different terminal equipment is approximately The same, thereby improving the model training efficiency of horizontal federated learning.
  • the application function network element when the application function network element adjusts the QoS parameter of the first terminal device in the terminal device group according to the transmission delay and local calculation delay of each terminal device, the application The functional network element determines the statistical value of the total delay of each terminal device according to the transmission delay and local calculation delay of each terminal device, wherein the total delay of each terminal device is the transmission delay of the terminal device The sum of delay and local calculation delay; when the total delay of the first terminal device is greater than a first preset threshold or less than a second preset threshold, the application function network element adjusts the first terminal A QoS parameter of the device, wherein the first preset threshold and the second preset threshold are determined according to the statistics.
  • the adjusted QoS parameters of the first terminal device include resource type, priority, packet delay budget, guaranteed stream bit rate, maximum stream bit rate, allocation and preemption priority at least one of the .
  • the local calculation delay is the time required to determine the update parameters of the model of the horizontal federated learning
  • the transmission delay is the time required to transmit the update parameters to the application function network element How long it takes.
  • the present application provides a communication device, the device is an application function network element, or a device for realizing the function of the application function network element, and the device includes a processing module and a transceiver module; the processing module uses To obtain the transmission delay and local calculation delay of each terminal device in the terminal device group, the terminal devices in the terminal device group are terminal devices participating in horizontal federated learning, according to the transmission delay and local calculation delay of each terminal device The time delay is for adjusting the QoS parameter of the first terminal device in the terminal device group; the transceiver module is used for sending the adjusted QoS parameter of the first terminal device to a policy control network element.
  • the processing module when adjusting the QoS parameter of the first terminal device in the terminal device group according to the transmission delay and local calculation delay of each terminal device, the processing module is configured to: The transmission delay and local calculation delay of each terminal device are determined, and the statistical value of the total delay of each terminal device is determined, wherein the total delay of each terminal device is the transmission delay of the terminal device and the local calculation delay.
  • the adjusted QoS parameters of the first terminal device include resource type, priority, packet delay budget, guaranteed stream bit rate, maximum stream bit rate, allocation and preemption priority at least one of the .
  • the local calculation delay is the time required to determine the update parameters of the model of the horizontal federated learning
  • the transmission delay is the time required to transmit the update parameters to the application function network element How long it takes.
  • the present application further provides a device.
  • the device can perform the method design described above.
  • the apparatus may be a chip or a circuit capable of performing the function corresponding to the above method, or a device including the chip or circuit.
  • the apparatus includes: a memory, configured to store computer executable program codes; and a processor, and the processor is coupled to the memory.
  • the program code stored in the memory includes instructions, when the processor executes When the instruction is executed, the device or the device installed with the device is made to execute the method in any one of the above possible designs.
  • the device may also include a communication interface, which may be a transceiver, or, if the device is a chip or a circuit, the communication interface may be an input/output interface of the chip, such as an input / output pins, etc.
  • a communication interface which may be a transceiver, or, if the device is a chip or a circuit, the communication interface may be an input/output interface of the chip, such as an input / output pins, etc.
  • the device includes corresponding functional units for respectively implementing the steps in the above method.
  • the functions may be implemented by hardware, or may be implemented by executing corresponding software through hardware.
  • Hardware or software includes one or more units corresponding to the functions described above.
  • the present application provides a computer-readable storage medium, where a computer program is stored in the computer-readable storage medium, and when the computer program is run on a device, the method in any one of the above possible designs is executed.
  • the present application provides a computer program product, where the computer program product includes a computer program, and when the computer program is run on a device, the method in any one of the above possible designs is executed.
  • the present application provides a communication system, the system includes a first network element and a second network element, and the first network element is used to execute any one of the possible design methods in the first aspect , the second network element is configured to execute the method in any possible design of the second aspect.
  • Fig. 1 is a schematic diagram of the architecture of the mobile communication system used in the present application.
  • FIG. 2 is a schematic diagram of data sets participating in horizontal federated learning in the embodiment of the present application
  • Fig. 3 is a schematic diagram of the model training process of horizontal federated learning in the embodiment of the present application.
  • FIG. 4 is an overview flowchart of a communication method in an embodiment of the present application.
  • FIG. 5A is a schematic diagram of the transmission delay of the terminal devices in the terminal device group before adjustment in the embodiment of the present application;
  • FIG. 5B is a schematic diagram of the transmission delay of the terminal devices in the terminal device group after adjustment in the embodiment of the present application.
  • FIG. 6 is a flow chart of interaction between the first network element and other network elements before the first network element sends the first request message to the second network element in the embodiment of the present application;
  • FIG. 7 is an overview flowchart of another communication method in the embodiment of the present application.
  • FIG. 8 is an overview flowchart of another communication method in the embodiment of the present application.
  • FIG. 9 is an overview flowchart of another communication method in the embodiment of the present application.
  • FIG. 10 is a flowchart of interaction between AF and PCF in the embodiment of the present application.
  • FIG. 11 is one of the structural schematic diagrams of a communication device in the embodiment of the present application.
  • FIG. 12 is a second structural schematic diagram of a communication device in an embodiment of the present application.
  • At least one item (piece) of a, b, or c can represent: a, b, c, a-b, a-c, b-c, or a-b-c, where a, b, c can be single or multiple .
  • the technical solutions provided by the embodiments of the present application can be applied to various communication systems. For example: it can be applied to LTE system or 5G system, and can also be applied to other future-oriented new systems, etc. This embodiment of the present application does not specifically limit it. Also, the term “system” and “network” can be used interchangeably. The following only takes the 5G communication system architecture as an example for illustration.
  • the 5G communication system formulated by the 3rd generation partnership project (3rd generation partnership project, 3GPP) standard.
  • the communication system includes a terminal device (for example, user equipment (user equipment, UE)), a radio access network (radio access network, RAN), and a core network (Core network, CN).
  • Data network data network, DN
  • the network elements of the core network can be divided into two parts: the user plane and the control plane.
  • the control plane is responsible for the management of the mobile network
  • the user plane is responsible for the transmission of business data.
  • the terminal device is the entrance for the mobile user to interact with the network, and can provide basic computing power and storage capacity, display service windows to the user, and receive user operation input.
  • the next-generation terminal equipment (NextGen UE) can adopt new air interface technology to establish signal connection and data connection with RAN, so as to transmit control signals and business data to the mobile network.
  • Terminal devices may include various handheld devices with wireless communication functions, vehicle-mounted devices, wearable devices, computing devices or other processing devices connected to wireless modems, as well as various forms of terminals, mobile stations (mobile station, MS), terminal (terminal), soft terminals, etc., such as water meters, electricity meters, sensors, etc.
  • RAN Deployed close to terminal equipment, it provides network access functions for authorized users in specific areas, and can determine transmission tunnels of different qualities to transmit user data according to user levels and business requirements. RAN can manage its own resources, use them rationally, provide access services for terminal equipment on demand, and be responsible for forwarding control signals and user data between terminal equipment and the core network.
  • Core network responsible for maintaining the subscription data of the mobile network, managing network elements of the mobile network, and providing functions such as session management, mobility management, policy management, and security authentication for terminal devices.
  • the terminal device When the terminal device is attached, it provides network access authentication for the terminal device; when the terminal device has a service request, it allocates network resources for the terminal device; when the terminal device moves, it updates the network resources for the terminal device; when the terminal device is idle, Provide a fast recovery mechanism for the terminal device; release network resources for the terminal device when the terminal device is detached; provide data routing functions for the terminal device when the terminal device has business data, such as forwarding uplink data to the data network; or from the data
  • the network receives the downlink data of the terminal device, forwards it to the RAN, and then the RAN sends it to the terminal device.
  • Data network a data network that provides business services for users.
  • the client is located in the terminal device, and the server is located in the data network.
  • the data network can be a private network, such as a local area network, or an external network not controlled by the operator, such as the Internet (Internet), or a proprietary network jointly deployed by the operator, such as providing an IP multimedia network subsystem (IP multimedia network subsystem). core network subsystem (IMS) service network.
  • IP multimedia network subsystem IP multimedia network subsystem
  • IMS core network subsystem
  • the core network user plane includes user plane function (user plane function, UPF); core network control plane includes access and mobility management function (access and mobility management function, AMF), session management function (session management function, SMF), network exposure function (network exposure function, NEF), network function storage function (NF repository function, NRF), unified data management (unified data management, UDM), unified data storage (unified data repository, UDR), policy control function (policy control function, PCF), application function (application function, AF).
  • UPF user plane function
  • core network control plane includes access and mobility management function (access and mobility management function, AMF), session management function (session management function, SMF), network exposure function (network exposure function, NEF), network function storage function (NF repository function, NRF), unified data management (unified data management, UDM), unified data storage (unified data repository, UDR), policy control function (policy control function, PCF), application function (application function, AF).
  • the control plane of the core network adopts a service-oriented architecture, and the interaction between network elements on the control plane adopts the method of service invocation to replace the point-to-point communication method in the traditional architecture.
  • the control plane NEs will open services to other control plane NEs for other control plane NEs to call; in point-to-point communication, the communication interface between control plane NEs will store a set of specific messages, only It can be used by control plane network elements at both ends of the interface during communication.
  • Session management network element mainly used for session management, IP address allocation and management of terminal equipment, selection of endpoints that can manage user equipment plane functions, policy control, or charging function interfaces, and downlink data notification.
  • the session management network element may be an SMF network element.
  • the session management function network element may still be an SMF network element, or have other names, which are not limited in this application.
  • Nsmf is a service-based interface provided by SMF. SMF can communicate with other network functions through Nsmf.
  • Access management network element mainly used for mobility management and access management, etc., for example, it can be a mobility management entity (mobility management entity, MME) function in a 4G communication network or an AMF network element in a 5G network. In future communications such as 6G communications, the access management network element may still be an AMF network element, or have other names, which are not limited in this application.
  • Namf is a service-based interface provided by AMF. AMF can communicate with other network functions through Namf.
  • Network opening network element used to securely open services and capabilities provided by 3GPP network functions to the outside.
  • network open network elements may be NEF network elements.
  • network open function network elements may still be NEF network elements, or have other names, which are not limited in this application.
  • Nnef is a service-based interface provided by NEF, and NEF can communicate with other network functions through Nnef.
  • Network storage network element used to provide service registration, discovery and authorization, and maintain available network function (network function, NF) instance information, which can realize the on-demand configuration of network functions and services and the interconnection between NFs.
  • the network storage network element can be an NRF network element.
  • the network storage function network element can still be an NRF network element, or have other names, which are not limited in this application.
  • Nnrf is a service-based interface provided by NRF. NRF can communicate with other network functions through Nnrf.
  • Policy control network element a unified policy framework for guiding network behavior, providing policy rule information, etc. for control plane functional network elements (such as AMF, SMF, etc.).
  • the policy control network element may be a PCF network element.
  • future communications such as 6G communication, the policy control network element may still be a PCF network element, or have other names, which are not limited in this application.
  • Npcf is a service-based interface provided by PCF, and PCF can communicate with other network functions through Npcf.
  • Data management network element used to process user identification, subscription, access authentication, registration, or mobility management.
  • the data management network element may be a UDM network element.
  • future communications such as 6G communication, the data management network element may still be a UDM network element, or have other names, which are not limited in this application.
  • Nudm is a service-based interface provided by UDM, and UDM can communicate with other network functions through Nudm.
  • Data storage network element used to implement the access function of contract data, policy data, application data and other types of data.
  • the data storage network element can be a UDR network element.
  • future communication such as 6G communication
  • the data storage network element It can still be a UDR network element, or have other names, which are not limited in this application.
  • Nudr is a service-based interface provided by UDR, and UDR can communicate with other network functions through Nudr.
  • Application network element It is used to route the data affected by the application, access the open function of the network, or interact with the policy framework for policy control, etc.
  • the application network element may be an AF network element.
  • future communications such as 6G communication, the application network element may still be an AF network element, or have other names, which are not limited in this application.
  • Naf is a service-based interface provided by AF. AF can communicate with other network functions through Naf.
  • User plane network element used for packet routing and forwarding, or quality of service (QoS) processing of user plane data, etc.
  • the user plane network element may be a user plane function (UPF) network element.
  • UPF user plane function
  • future communications such as 6G communication
  • the user plane network element may still be a UPF network element, or have other names. This application There is no limit to this.
  • the core network may also include other network elements, which is not limited in this application.
  • federated learning can be divided into three categories: horizontal federated learning, vertical federated learning, and federated transfer learning.
  • horizontal federated learning vertical federated learning
  • federated transfer learning mainly introduces the principle of horizontal federated learning.
  • the data set includes multiple sets of data, each set of data corresponds to a user, and each set of data records at least one user characteristic of the user corresponding to the set of data.
  • data set A includes UE1
  • the user characteristics included in the data of UE1 are X1, X2 and X3
  • the user characteristics included in the data of UE2 are X1, X2 and X3
  • the data set B includes the data of UE4 and the data of UE5, the data of UE4
  • the included user features are X1, X2, and X3
  • the UE5 data includes the user features X1, X2, and X3.
  • dataset A and dataset B can select the same user features from their respective datasets for model training of horizontal federated learning.
  • the participants of horizontal federated learning are not limited to two parties.
  • the participants of horizontal federated learning may be very large.
  • each participant (Client 1, Client2...ClientK shown in Figure 3) needs to download the latest model from the server (Server) through 5GS for the model training process
  • the specific iteration process generally includes the following steps:
  • each participant uses the local dataset to calculate the gradient of the model, and then encrypts and uploads the gradient to the server.
  • the gradient here includes the change direction and magnitude of the parameters of the model.
  • the server performs gradient aggregation according to the gradient uploaded by each participant, and updates the parameters of the model.
  • the gradient aggregation here generally refers to averaging the collected gradients.
  • 3Server distributes the parameters of the updated model to each participant.
  • QoS Quality of service
  • QoS describes a set of service requirements that a network must meet in order to ensure an appropriate level of service for data transmission.
  • 5G networks need to support various business transmissions, such as video, mobile payment, web browsing, and factory automation control etc.
  • Different services require different QoS.
  • video services require large bandwidth
  • automation control services generally require low delay and high reliability.
  • operators can provide different QoS guarantees for different services.
  • QoS Flow is the finest granularity for end-to-end QoS control in a protocol data unit (PDU) session.
  • the QoS control of QoS Flow is mainly determined by the QoS parameters associated with it.
  • the QoS framework based on 5G network supports guaranteed bandwidth QoS Flow (QoS Flows that require guaranteed flow bit rate, GBR QoS Flow) and non-guaranteed bandwidth QoS Flow (QoS Flows that do not require guaranteed flow bit rate, Non-GBR QoS Flow) .
  • QoS Flow QoS Flows that do not require guaranteed flow bit rate, Non-GBR QoS Flow
  • GBR QoS Flow the network needs to reserve resources to guarantee its bandwidth.
  • Non-GBR QoS Flow there is no need to reserve resources.
  • the corresponding QoS parameters of GBR and non-GBR QoS Flow are shown in Table 1.
  • 5QI is used to identify a set of 5G QoS characteristics.
  • 5G QoS characteristics are used to describe the end-to-end packet forwarding process of QoS flow received between UE and UPF, including resource type (such as Non-GBR and GBR), priority, packet delay budget (packet delay budget, PDB ) and packet error rate, etc.
  • QoS parameters are all for a single QoS Flow.
  • some aggregated QoS parameters are defined in the 5G network.
  • Aggregated QoS parameters are used to characterize the statistical characteristics of QoS parameters of multiple QoS Flows, for example, aggregate maximum bit rate (AMBR), which is used to limit the maximum value of the total bit rate of a group of QoS Flows.
  • AMBR aggregate maximum bit rate
  • PDU session-level aggregate maximum bit rate (Session-AMBR): Limits the aggregate bit rate expected to be provided in all Non-GBR QoS Flows of a specific PDU session.
  • UE-level aggregated maximum bit rate (UE-AMBR): applicable to Non-GBR QoS Flow.
  • UE-Slice-MBR Slice-level maximum bit rate per UE
  • Slice-level maximum bit rate (Slice-MBR per S-NSSAI): applicable to GBRQoS Flow and Non-GBR QoS Flow.
  • the method provided in this application can not only be applied to the scenario of horizontal federated learning, but also can be applied to other model training processes or business processes that require interaction between the server and the terminal devices in the terminal device group.
  • the following horizontal federated learning is performed
  • the scenario is described as an example.
  • the present application provides a communication method, as shown in FIG. 4 , wherein the following first network element may be an application function network element, and the second network element may be a data analysis function network element or a policy control network element.
  • the data analysis function network element can be It refers to a network data analysis function (network data analytics Function, NWDAF) network element
  • the policy control network element may refer to a PCF network element.
  • the method includes:
  • Step 400 the first network element sends a first request message to the second network element, the first request message is used to request data of a terminal device group, and the data of the terminal device group includes aggregation information of measurement results of QoS parameters.
  • the first request message may also be called a first subscription message, and this application does not limit the name of the first request message.
  • the first request message may be an analysis subscription message (Nnwdaf_AnalyticsSubscription_Subscribe).
  • the first request message may be an event open subscription message (Nnwdaf_EventExposure Subscribe).
  • the terminal device group includes at least two terminal devices.
  • the terminal devices in the terminal device group may be terminal devices participating in horizontal federated learning.
  • the terminal devices in the terminal device group jointly participate in the same horizontal federated learning model training process.
  • the first network element sends the first request message to the second network element
  • all terminal devices in the terminal device group have established service sessions associated with horizontal federated learning.
  • the first service is a model training service of some kind of horizontal federated learning
  • the terminal device group includes 3 terminal devices. These 3 terminal devices are the participants of the first service. session.
  • the aggregated information may also be referred to as aggregated QoS information, and the aggregated information is aggregated and determined according to measurement results of QoS parameters of each terminal device in the terminal device group.
  • the aggregation information is used to characterize the statistical characteristics of the measurement results of the QoS parameters of each terminal device included in the terminal device group.
  • QoS parameters of the terminal device here, reference may be made to the relevant description about the QoS parameters above.
  • the first request message may include but not limited to the following:
  • Identification information of an analysis type used to indicate that the aggregated information is aggregated information of a terminal device group, that is, aggregated information at a terminal device group level.
  • Analytics ID Aggregated QoS for FL group, indicating that the aggregation information of the FL group (group) is requested.
  • the type of the aggregation information is used to indicate the parameters included in the aggregation information.
  • the type of the aggregated information may be represented by aggregated QoS parameters to be measured (Aggregated QoS parameters to be measured).
  • the type of the aggregated information may specifically include aggregated bit rates, statistical values of transmission delays of terminal devices in the terminal device group, and the like. It can be understood that, according to the model training business requirements of horizontal federated learning, the parameter information may also include other parameters, which are not limited here.
  • the aggregated bit rate is used to represent the sum of the bit rates of the QoS flows of the terminal devices in the terminal device group.
  • the aggregated bit rate can be converted into the aggregated bandwidth, and the aggregated bandwidth is used to represent the sum of the corresponding bandwidths of the terminal devices in the terminal device group.
  • the QoS flow here is the QoS flow corresponding to the horizontal federated learning service. If there is no special statement, the QoS flow involved in the following refers to the QoS flow corresponding to the horizontal federated learning service.
  • the bit rate of the QoS flow here may refer to GFBR.
  • each participant needs to complete local calculations and upload the calculation results to the server, so a large amount of bandwidth resources will be consumed.
  • a maximum available bandwidth is generally set for the model training service of horizontal federated learning. to ensure that The sum of the bandwidth of each participant does not exceed the set maximum available bandwidth, the first network element can subscribe to the aggregated bandwidth from the second network element, and adjust the bandwidth of at least one participant when the aggregated bandwidth is greater than the set maximum available bandwidth .
  • the relevant description in step 420 for a specific adjustment manner, reference may be made to the relevant description in step 420 below.
  • the statistical value of the transmission delays of the terminal devices in the terminal device group may be a variance determined according to the transmission delays of the terminal devices in the terminal device group.
  • the statistical value of the transmission delay of the terminal devices in the terminal device group refers to the variance determined by the weighted transmission delay of the terminal devices in the terminal device group (hereinafter referred to as the variance of the weighted transmission delay) or the variance determined by the terminal device
  • the determined variance of the differential transmission delay of the terminal devices in the device group (hereinafter referred to as the variance of the differential transmission delay).
  • the weighted transmission delay of terminal device A is the ratio of the actual transmission delay of terminal device A to the expected transmission delay of terminal device A.
  • the differential transmission delay of device A is the difference between the actual transmission delay of terminal device A and the expected transmission delay of terminal device A
  • the actual transmission delay of terminal device A is the difference between the sending time of QoS flow of terminal device A and terminal device
  • the time interval between when the QoS flow of A arrives at the server, and the expected transmission delay of terminal device A is the time interval between the sending time of the QoS flow of terminal device A and the preset time.
  • the actual transmission delay of terminal device A can be obtained through the measurement result of the QoS parameter of terminal device A, and the expected transmission delay of terminal device A can be sent by the first network element to the second network element through step 400, or can be advanced Configured to the second network element or the user plane network element.
  • the server here can be co-located with any network element, or can be an independent server.
  • the actual transmission delay of the first terminal device may be the PDB of the first terminal device.
  • the terminal device group includes UE1, UE2, and UE3.
  • Each UE corresponds to a time axis.
  • the time from the start point of the time axis to the triangle mark is the local calculation completion time of the UE.
  • From the triangle mark to The duration corresponding to the circular mark is the actual transmission delay of the UE, and the duration corresponding to the intersection point of the triangle mark to the first dotted line and the time axis is the expected transmission delay of the UE.
  • the position of the first dotted line can be understood as the time when the server receives the calculation result of the last UE (that is, UE2).
  • the local calculation completion time of each UE may be different, and it is difficult to change, that is, the positions of the triangle marks in Figure 5A are different, and the transmission delay of each UE may also be different, that is, the circular mark in Figure 5A The locations are different, but the transmission delay of the UE can be realized by adjusting the bandwidth of the UE.
  • the time difference between the QoS flows arriving at the server is relatively large, and the smaller the variance, it means that the time difference between the arrival times of the QoS flows sent by the three respectively at the server is relatively small.
  • the server can update the model only after receiving the calculation results of each participant.
  • the calculation results of each participant are transmitted to the server with a large difference, that is, when the variance determined by the weighted transmission delay or differential transmission delay of each participant is large, it will cause the overall training cycle to become longer and reduce model training. s efficiency.
  • the calculation results of each participant are transmitted to the server at about the same time, that is, when the variance determined by the weighted transmission delay or differential transmission delay of each participant is small, the waiting time of the server can be reduced, thereby improving the efficiency of model training. Therefore, the first network element may subscribe the variance to the second network element, and adjust the transmission delay of at least one participant when the variance is greater than the set variance threshold. For the specific adjustment method, refer to the relevant description in step 420 below.
  • the server can update the model only after receiving the calculation results of each participant.
  • the calculation results of each participant are transmitted to the server with a large difference, that is, when the variance determined by the weighted transmission delay of each participant is large, the overall training cycle will become longer and the efficiency of model training will be reduced.
  • the calculation results of each participant are transmitted to the server at about the same time, that is, when the variance determined by the weighted transmission delay of each participant is small, the waiting time of the server can be reduced, thereby improving the efficiency of model training. Therefore, the first network element may subscribe the variance to the second network element, and adjust the transmission delay of at least one participant when the variance is greater than the set variance threshold. For the specific adjustment method, refer to the relevant description in step 420 below.
  • Information for indicating the terminal device group may include an identifier of each terminal device in the terminal device group, or a list of terminal devices corresponding to the terminal device group, or an identifier of the terminal device group.
  • the information used to indicate a terminal device group may be identified by a target of analytics reporting (Target of analytics reporting).
  • the terminal device group includes UE1, UE2, and UE3, and the information used to indicate the terminal device group includes the identifier of UE1, the identifier of UE2, and the identifier of UE3.
  • the second network element may determine which terminal device's QoS parameter measurement results to collect, or aggregate the terminal device's QoS parameter measurement results, and then determine the aggregation information.
  • the trigger condition for sending aggregated information may include periodically feeding aggregated information, or feeding aggregated information when a preset condition is met.
  • the preset conditions here may include that the aggregated bit rate is greater than the contracted maximum bit rate of the terminal device group, and/or the statistical value of the transmission delay of the terminal devices in the terminal device group is greater than a first preset threshold, and the like.
  • the second network element feeds back the aggregation information according to the trigger condition for sending the aggregation information.
  • the measurement result indicated by the indication information is the measurement result at the terminal device level.
  • the feedback condition indicates the condition to be met when feeding back the measurement result of the QoS parameter of the terminal device.
  • the feedback condition may indicate that a terminal device whose transmission delay is greater than a second preset threshold feeds back a QoS parameter measurement result, or a terminal device whose transmission delay is smaller than a third preset threshold feeds back a QoS parameter measurement
  • the terminal equipment whose QoS flow bit rate is greater than the preset bit rate feeds back the measurement result of the QoS parameter.
  • the second preset threshold is greater than the third preset threshold.
  • the data of the terminal device group when the first request message only includes the indication information and does not include the feedback condition, the data of the terminal device group also includes the measurement results of the QoS parameters of each terminal device in the terminal device group, that is, the second network element sends One NE The QoS parameter measurement results of all terminal devices in the terminal device group are fed back.
  • the data of the terminal device group when the first request message includes the indication information and the feedback condition, the data of the terminal device group further includes the measurement result of the QoS parameter of at least one terminal device in the terminal device group, and the at least one terminal device's QoS parameter measurement result
  • the measurement result of the QoS parameter satisfies the feedback condition, that is, the second network element feeds back the measurement result of the QoS parameter of some terminal devices in the terminal device group to the first network element.
  • the application identifier which is used to indicate the scope of the application that collects the measurement results of the QoS parameters of the terminal equipment.
  • the application identifier is an identifier of an application for horizontal federated learning.
  • the second network element instructs the user plane network element to collect the measurement result of the QoS parameter of the terminal device according to the application identifier.
  • Step 410 the second network element sends the data of the terminal device group to the first network element.
  • the second network element when the second network element is an NWDAF network element, the second network element sends an analysis notification message (Nnwdaf_AnalyticsSubscription_Notify) to the first network element, and the message carries the data of the terminal device group.
  • the first request message may be an event exposure notification message (Nnwdaf_EventExposure Notify), and the message carries data of the terminal device group. It should be noted that each communication between the first network element and the second network element needs to be forwarded by the NEF network element.
  • the second network element sends a second request message to the user plane network element, the second request message is used to request data of the terminal device group, and the second network element receives the terminal device group data from the user plane network element The data.
  • the second request message may include the content included in the first request message except the type information associated with the aggregation information.
  • the first request message includes the above items 1-7
  • the second request message includes the above items 2-7.
  • the user plane network element completes the collection of the measurement results of the QoS parameters of the terminal equipment in the terminal equipment group, and determines the data of the terminal equipment group according to the collected measurement results, for example, according to the parameter information associated with the aggregated information
  • the aggregation information is determined, and the measurement result of the QoS parameter of the terminal device that needs to be fed back is determined according to the indication information and/or the feedback condition.
  • the user plane network element sends the data of the determined terminal device group to the second network element, and the second network element sends the data of the terminal device group to the first network element.
  • the terminal equipment group includes UE1, UE2 and UE3, and the user plane network element receives the second request message, and collects the measurement result of the QoS parameter of UE1, the measurement result of the QoS parameter of UE2, and the QoS parameter of UE3 according to the second request message measurement results. Further, the user plane network element determines, according to the parameter information associated with the aggregated information in the second request message, that the aggregated information includes aggregated bit rate and statistical value of transmission delay, then according to the bit rate of the QoS flow corresponding to UE1, The sum of the bit rate of the QoS flow corresponding to UE2 and the bit rate of the QoS flow corresponding to UE3 is used as the aggregate bit rate of the terminal device group.
  • the user plane network element is also based on the actual transmission delay of UE1, the actual transmission delay of UE2, the actual transmission delay of UE3, and the expected transmission delay of UE1 received from the second network element or configured in advance, and the expected transmission delay of UE2
  • the delay and the expected transmission delay of UE3 calculate the statistical value of the transmission delay of the terminal device group.
  • the user plane network element sends the data of the terminal device group to the second network element, where the data of the terminal device group includes the aggregated bit rate of the terminal device group and the statistical value of the transmission delay of the terminal device group.
  • the measurement result of the QoS parameter of UE1 includes the bit rate of the QoS flow corresponding to UE1 and the actual transmission delay of UE1
  • the measurement result of the QoS parameter of UE2 includes the bit rate of the QoS flow corresponding to UE2 and the actual transmission delay of UE2
  • the measurement result of the QoS parameter of UE3 includes the bit rate of the QoS flow corresponding to UE3 and the actual transmission delay of UE3.
  • the second network element sends a third request message to the user plane network element, and the third request message is used to request the measurement result of the QoS parameter of the terminal device in the terminal device group, and the second network element receives from User plane network element receiving terminal The second network element determines the data of the terminal device group according to the measurement result of the QoS parameter of the terminal device in the terminal device group and the first request message.
  • the user plane network element collects the measurement results of the QoS parameters of the terminal devices in the terminal device group, and sends the collected measurement results to the second network element, and the second network element determines the terminal according to the collected results.
  • aggregate information is determined according to parameter information associated with the aggregate information
  • measurement results of QoS parameters of terminal devices that need to be fed back are determined according to indication information and/or feedback conditions.
  • the terminal equipment group includes UE1, UE2, and UE3, and the user plane network element receives the third request message, and collects the measurement result of the QoS parameter of UE1, the measurement result of the QoS parameter of UE2, and the QoS parameter of UE3 according to the third request message measurement results.
  • the user plane network element sends the measurement result of the QoS parameter of UE1, the measurement result of the QoS parameter of UE2, and the measurement result of the QoS parameter of UE3 to the second network element.
  • the second network element determines that the aggregation information includes aggregated bit rate and statistical value of transmission delay, then according to the bit rate of the QoS flow corresponding to UE1, the corresponding QoS flow of UE2
  • the bit rate of the QoS flow, and the sum of the bit rates of the QoS flows corresponding to UE3 are used as the aggregate bit rate of the terminal device group.
  • the second network element also based on the actual transmission delay of UE1, the actual transmission delay of UE2, the actual transmission delay of UE3, and the expected transmission delay of UE1, the expected transmission delay of UE2, and the expected transmission time of UE3 configured in advance Delay calculates the statistical value of the transmission delay of the terminal device group.
  • the second network element sends the data of the terminal device group to the first network element, wherein the data of the terminal device group includes the aggregate bit rate of the terminal device group and the statistical value of the transmission delay of the terminal device group.
  • the measurement result of the QoS parameter of UE1 includes the bit rate of the QoS flow corresponding to UE1 and the actual transmission delay of UE1
  • the measurement result of the QoS parameter of UE2 includes the bit rate of the QoS flow corresponding to UE2 and the actual transmission delay of UE2
  • the measurement result of the QoS parameter of UE3 includes the bit rate of the QoS flow corresponding to UE3 and the actual transmission delay of UE3.
  • Step 420 The first network element adjusts the QoS parameter of the first terminal device in the terminal device group according to the data of the terminal device group.
  • the first network element may adjust the QoS parameter of at least one terminal device in the terminal device group, for example, the QoS parameter of the first terminal device, or the QoS parameters of multiple terminal devices, or the QoS parameters of all terminal devices .
  • the first network element may adjust the value of the QoS parameter, or the value range of the QoS parameter. For example, the first network element may adjust the value of the bit rate of the QoS flow of the first terminal device, or the value range of the bit rate.
  • a possible implementation manner of adjusting the QoS parameter of at least one terminal device in the terminal device group will be described below in combination with data of a specific terminal device group, wherein the at least one terminal device includes the first terminal device.
  • Mode 1 When the aggregation information includes the aggregated bit rate, if the aggregated bit rate is greater than the contracted maximum bit rate of the terminal device group, the first network element adjusts the bit rate of the QoS flow of at least one terminal device, and adjusts the QoS flow of at least one terminal device. After the bit rate of the stream, the sum of the bit rates of the QoS streams of each terminal device in the terminal device group is less than or equal to the contracted maximum bit rate, or the value range of the bit rate of the QoS streams of each terminal device in the terminal device group The sum of the lower bounds of is less than or equal to the contracted maximum bit rate.
  • the second The network element or the user plane network element judges whether the aggregated bit rate is greater than the contracted maximum bit rate of the terminal device group according to the contracted maximum bit rate of the terminal device group, and if so, the second network element sends the data of the terminal device group to the first network element,
  • the data of the terminal equipment group includes the statistical value of the transmission delay.
  • the first network element or according to the maximum contracted bit rate of the terminal device group determines whether the aggregated bit rate is greater than the maximum contracted rate of the terminal device group. bit rate.
  • the first network element when the first network element adjusts the bit rate of the QoS flow of at least one terminal device, the first network element may configure the bit rate of the QoS flow for each terminal device in the terminal device group according to the previous configuration. Determine the adjustment scheme, that is, to adjust the bit rate of the QoS flow of which terminal devices, as well as the direction and size of the adjustment, so that after adjusting the bit rate of the QoS flow of at least one terminal device, the QoS flow of each terminal device in the terminal device group
  • the sum of the bit rates is less than or equal to the contracted maximum bit rate, or the sum of the lower bounds of the bit rates of the QoS flows of each terminal device in the terminal device group is less than or equal to the contracted maximum bit rate.
  • the first network element when the first network element adjusts the bit rate of the QoS flow of at least one terminal device, if the first request message includes indication information, the second network element also feeds back the information in the terminal device group to the first network element.
  • the first network element can determine the adjustment scheme in combination with the measurement results of the QoS parameters of each terminal device in the terminal device group; or, the first network element adjusts the bit of the QoS flow of at least one terminal device rate, if the first request message includes indication information and feedback conditions, the second network element also feeds back to the first network element the measurement results of the QoS parameters of the terminal devices in the terminal device group that meet the feedback conditions, and the first network element may combine the terminal
  • the measurement results of the QoS parameters of the terminal devices satisfying the feedback condition in the device group determine the adjustment scheme.
  • the feedback condition may be that a terminal device whose QoS stream bit rate is greater than a preset bit rate feeds back a QoS parameter measurement
  • the terminal device group includes UE1, UE2 and UE3, and the aggregate bit rate of the terminal device group is the sum of the bit rate of the QoS flow corresponding to UE1, the bit rate of the QoS flow corresponding to UE2, and the bit rate of the QoS flow corresponding to UE3.
  • the first network element may adjust the value of at least one of UE1, UE2 and or UE3, so that the bit rate of the QoS flow corresponding to UE1,
  • the bit rate of the QoS flow corresponding to UE2 and the bit rate of the QoS flow corresponding to UE3 are less than or equal to the contracted maximum bit rate of the terminal device group, or the lower bound of the bit rate range of the QoS flow corresponding to UE1.
  • the lower bound of the value range of the bit rate of the QoS stream, and the sum of the lower bounds of the lower bound of the bit rate range of the QoS stream corresponding to UE3 is less than or equal to the contracted maximum bit rate of the terminal device group.
  • Method 2 When the aggregation information includes the statistical value of the transmission delay of the terminal devices in the terminal device group, if the statistical value of the transmission delay is greater than the first preset threshold, the first network element adjusts at least The bit rate of a QoS flow for an end device.
  • At least one terminal device is a terminal device whose transmission delay is greater than a second preset threshold or smaller than a third preset threshold, where the second preset threshold is greater than the third preset threshold.
  • the sum of the bit rates of the QoS flows of each terminal device in the terminal device group is less than or equal to the contracted maximum bit rate, or the terminal device group
  • the sum of the lower bounds of the value ranges of the bit rates of the QoS streams of each terminal device in is less than or equal to the contracted maximum bit rate.
  • the second network element or user The plane network element judges whether the statistical value of the transmission delay is greater than the first preset threshold according to the first preset threshold, and if so, the second network element sends the data of the terminal device group to the first network element, and the data of the terminal device group includes transmission Statistical value of delay.
  • the first request message does not include a trigger condition for sending aggregation information
  • the first network element judges whether the statistical value of the transmission delay is greater than the first preset threshold according to the first preset threshold.
  • the first network element when the first network element adjusts the bit rate of the QoS flow of at least one terminal device, if the first request message includes indication information, the second network element also feeds back the information in the terminal device group to the first network element. Based on the measurement results of the QoS parameters of each terminal device, the first network element can determine the adjustment scheme based on the measurement results of the QoS parameters of each terminal device in the terminal device group, that is, which terminal devices to adjust the bit rate of the QoS flow, and the direction and direction of the adjustment. size.
  • the bit rate is increased to reduce its transmission delay, and/or for a terminal device with a relatively long transmission delay
  • a short terminal device for example, a terminal device with a transmission delay smaller than the third preset threshold increases its transmission delay by reducing the bit rate.
  • the first network element when the first network element adjusts the bit rate of the QoS flow of at least one terminal device, if the first request message includes indication information and feedback conditions, the second network element also feeds back to the first network element that the terminal device group meets the feedback conditions
  • the first network element may determine the adjustment scheme in combination with the measurement results of the QoS parameters of the terminal devices in the terminal device group that satisfy the feedback condition.
  • the feedback condition is that a terminal device whose transmission delay is greater than a second preset threshold feeds back a QoS parameter measurement result, and/or a terminal device whose transmission delay is smaller than a third preset threshold feeds back a QoS parameter measurement result.
  • the first network element lowers the bit rate of the QoS flow of UE1 to increase the actual transmission delay of UE1, and increases the bit rate of the QoS flow of UE2 to shorten the The actual transmission delay of UE2, lower the bit rate of the QoS flow of UE3, and increase the actual transmission delay of UE3.
  • the time length corresponding to the triangle mark to the hollow circle mark is the actual transmission delay of the UE before adjustment
  • the time length corresponding to the triangle mark to the solid circle mark is the adjusted transmission delay of the UE.
  • the period from the triangle mark to the intersection point of the first dotted line and the time axis corresponds to the adjusted expected transmission delay of the UE, from the triangle mark to the intersection point of the second dotted line and the time axis
  • the corresponding duration is the expected transmission delay of the UE before adjustment.
  • the server receives the calculation results of all UEs in the terminal device group at the time indicated by the second dotted line, and after the adjustment, the server receives the calculation results of all UEs in the terminal device group According to the calculation result of UE, the time shown by the first dotted line is earlier than the time shown by the second dotted line, so the waiting time of the server can be reduced, and the efficiency of model training can be improved.
  • Step 430 the first network element sends the adjusted QoS parameter of the first terminal device to the policy control network element.
  • the policy control network element will also perform corresponding policy control according to the aggregation information of the terminal device group to ensure that the restrictions of the aggregation information are met.
  • the aggregated bit rate must be less than or equal to the contracted maximum bit rate of the terminal device group, and the The statistical value needs to be less than or equal to the first preset threshold. Taking the aggregated bit rate as an example, the relevant policy control method is briefly introduced below.
  • the policy control method includes:
  • the policy control network element checks whether the aggregated bit rate of the terminal device group is higher than the contracted maximum bit rate of the terminal device group. If yes, the policy control network element should refuse to establish SM policy association, otherwise, the policy control network element normally establishes SM policy association.
  • session management session management
  • the policy control network element can limit the protocol data unit ( protocol data unit, PDU) session or PCC rule traffic, and interact with the session management network element accordingly, where the preset threshold is less than the maximum bit rate of the contract of the terminal device group, and the preset threshold and the contract maximum of the terminal device group
  • the bit rate is close, for example, the preset threshold is 95% of the contracted maximum bit rate of the terminal device group; if the policy control network element finds that the aggregate bit rate of the terminal device group is less than the preset threshold, the policy control network element can relax the PDU session or PCC The traffic limit of the rule.
  • the first network element subscribes the data of the terminal equipment group to the second network element, the second network element sends the data of the terminal equipment group to the first network element, and the first network element adjusts the terminal according to the data of the terminal equipment group
  • the QoS parameters of at least one terminal device in the device group can improve the efficiency of model training.
  • the first network element may also execute the operation process shown in FIG. 6 .
  • Step 600 the first network element sends the contracted maximum bit rate of the terminal equipment group to the third network element, and the third network element is a unified data storage network element or a unified data management network element.
  • the first network element may determine the contracted maximum bit rate of the terminal device group according to the model training service requirements of the horizontal federated learning, and send the contracted maximum bit rate of the terminal device group to the third network element.
  • the contracted maximum bit rate of the terminal device group can be carried by a data management creation request (Nudr_DM_Create Request) or a data management update request (Nudr_DM_Update Request).
  • the third network element When the contracted maximum bit rate of the terminal device group is carried by Nudr_DM_Create Request, the third network element creates the contracted maximum bit rate of the terminal device group in the subscription information, and sends a data management creation response (Nudr_DM_Create Response) to the first network element, which The message indicates that the subscription information has been created.
  • Nudr_DM_Create Response a data management creation response
  • the third network element updates the contracted maximum bit rate of the terminal device group in the subscription information, and sends a data management update response (Nudr_DM_Update Response) to the first network element, which The message indicates that the subscription information has been updated.
  • Step 610 The first network element sends first information to the policy control function, where the first information includes reference information of QoS parameters corresponding to respective terminal devices included in the terminal device group.
  • the reference information of the QoS parameters corresponding to each terminal device included in the terminal device group includes at least one of the following:
  • the value or value range of the bit rate of the QoS flow corresponding to each terminal device included in the terminal device group, or the value or value range of the packet loss rate corresponding to each terminal device included in the terminal device group, or the value or value range of the terminal device The value or value range of the transmission delay corresponding to each terminal device included in the group.
  • bit rate of the QoS flow 20 Mbps
  • packet loss rate 0.01%
  • the first network element may determine the first information according to the model training service requirements of horizontal federated learning, and the first information may be carried by a policy authorization creation request (Npcf_PolicyAuthorization_Create Request) or a policy authorization update request (Npcf_PolicyAuthorization_Update Request).
  • Npcf_PolicyAuthorization_Create Request a policy authorization creation request
  • Npcf_PolicyAuthorization_Update Request a policy authorization update request
  • Step 620 The policy control network element requests the contracted maximum bit rate of the terminal equipment group from the third network element.
  • the policy control network element requests the contracted maximum bit rate of the terminal device group from the third network element through a data management query request (Nudr_DM_QueryRequest) service operation.
  • Nudr_DM_QueryRequest a data management query request
  • Step 630 The third network element sends the contracted maximum bit rate of the terminal device group to the policy control network element.
  • the third network element sends the contracted maximum bit rate of the terminal device group to the policy control network element through a data management query response (Nudr_DM_QueryResponse) service operation.
  • Nudr_DM_QueryResponse data management query response
  • Step 640 The policy control network element verifies the first information according to the contracted maximum bit rate of the terminal device group.
  • the first information includes bits of QoS flows corresponding to each terminal device included in the terminal device group rate
  • the policy control network element determines each terminal device included in the terminal device group respectively Whether the sum of the bit rates of the corresponding QoS flows is less than or equal to the contracted maximum bit rate of the terminal device group, if so, the verification succeeds, otherwise the verification fails.
  • the policy control network element bases the contracted maximum bit rate on the terminal device group and each terminal included in the terminal device group
  • the value range of the bit rate of the QoS flow corresponding to the device respectively determine whether the sum of the lower bounds of the bit rate range of the QoS flow corresponding to each terminal device included in the terminal device group is less than or equal to the maximum contracted bit rate of the terminal device group rate, if so, the verification succeeds, otherwise the verification fails.
  • Step 650 When the verification is successful, the policy control network element generates policy and charging control (policy and charging control, PCC) rules corresponding to each terminal device included in the terminal device group according to the first information, and the second network element sends the session management The network element sends PCC rules corresponding to each terminal device.
  • policy and charging control policy and charging control
  • the session management network element may establish or modify a corresponding session according to the PCC rules corresponding to each terminal device sent by the policy control network element.
  • Step 660 The policy control network element sends second information to the first network element, and the second information indicates that corresponding PCC rules have been generated for each terminal device included in the terminal device group.
  • the second information may be carried by policy authorization creation response (Npcf_PolicyAuthorization_Create Response).
  • the second information may be carried by the policy authorization update response (Npcf_PolicyAuthorization_Update Response).
  • the present application provides another communication method, as shown in FIG. 7 , wherein the following first network element may be an application function network element.
  • the first network element subscribes the data of the terminal device group to the data analysis function network element, and provides the address information of the policy control network element, and the second network element feeds back the data of the terminal device group to the policy Control network elements.
  • the policy control network element adjusts the QoS parameter of the first terminal device in the terminal device group according to the data of the terminal device group. Compared with the embodiment shown in FIG.
  • the first network element is responsible for receiving the data of the terminal device group, and adjusting the QoS parameters of the first terminal device in the terminal device group according to the data of the terminal device group.
  • the policy control network element is responsible for receiving the data of the terminal device group, and adjusting the QoS parameter of the first terminal device in the terminal device group according to the data of the terminal device group.
  • the method includes:
  • Step 700 The first network element sends a first request message to the data analysis function network element, the first request message is used to request data of a terminal device group, and the data of the terminal device group includes aggregation information of measurement results of QoS parameters.
  • step 700 reference may be made to the relevant description in the above-mentioned step 400, which will not be repeated here.
  • the first request message needs to include the address information of the policy control network element in addition to the contents of items 1-8 above.
  • Step 710 The data analysis function network element sends the data of the terminal device group to the policy control network element according to the address information of the policy control network element.
  • the specific implementation manner of determining the data of the terminal device group by the network element with the data analysis function can refer to the relevant description in the above step 410 .
  • Step 720 The policy control network element receives the information of the terminal device group, and the policy control network element adjusts the QoS parameter of the first terminal device in the terminal device group according to the data of the terminal device group.
  • the specific implementation manner of adjusting the QoS parameters of the first terminal device in the terminal device group by the policy control network element according to the data of the terminal device group can refer to the first network element adjusting the QoS parameter of the first terminal device in the terminal device group according to the data of the terminal device group in step 420.
  • the first network element subscribes to the data analysis function network element for the data of the terminal equipment group, and notifies the second network element to send the data of the terminal equipment group to the policy control network element, and the policy control network element according to the data of the terminal equipment group
  • the QoS parameters of at least one terminal device in the terminal device group are adjusted according to the data, so as to improve the efficiency of federated learning model training.
  • the first network element may also perform the operation process shown in FIG. 6 .
  • the present application provides another communication method, as shown in FIG. 8 , wherein the following first network element may be an application function network element.
  • the first network element subscribes the data of the terminal device group to the policy control network element, and the policy control network element obtains the data of the terminal device group, and adjusts the data in the terminal device group according to the data of the terminal device group.
  • QoS parameters of the first terminal device QoS parameters of the first terminal device. Therefore, compared with the embodiment shown in FIG. 4, the first network element is responsible for receiving the data of the terminal device group, and adjusting the QoS parameters of the first terminal device in the terminal device group according to the data of the terminal device group.
  • the policy control network element is responsible for obtaining the data of the terminal device group, and adjusting the QoS parameter of the first terminal device in the terminal device group according to the data of the terminal device group.
  • the method includes:
  • Step 800 The first network element sends a first request message to the policy control network element.
  • the first request message is used to request data of a terminal device group, and the data of the terminal device group includes aggregation information of measurement results of QoS parameters.
  • step 800 reference may be made to the relevant description in the above-mentioned step 400, which will not be repeated here.
  • Step 810 The policy control network element determines the data of the terminal equipment group.
  • Step 820 The policy control network element adjusts the QoS parameter of the first terminal device in the terminal device group according to the data of the terminal device group.
  • the specific implementation manner of adjusting the QoS parameters of the first terminal device in the terminal device group by the policy control network element according to the data of the terminal device group can refer to the first network element adjusting the QoS parameter of the first terminal device in the terminal device group according to the data of the terminal device group in step 420.
  • the first network element subscribes the data of the terminal device group to the policy control network element, and the policy control network element obtains the data of the terminal device group, and adjusts the QoS of at least one terminal device in the terminal device group according to the data of the terminal device group parameters, which can improve the efficiency of model training.
  • the first network element may also perform the operation process shown in FIG. 6 .
  • the application function network element may not collect data through NWDAF or PCF, but collects statistics or obtains the transmission delay and local calculation delay of each terminal device in the terminal device group, and then determines the time delay in the terminal device group.
  • the QoS parameter adjustment strategy of terminal equipment reduces the number of messages exchanged between network elements, reduces the load of network elements, and improves the QoS parameter adjustment efficiency of terminal equipment, making the difference between local calculation delay and transmission delay of different terminal equipment is roughly the same as , thus improving the model training efficiency of horizontal federated learning.
  • the method includes:
  • Step 900 The application function network element obtains the transmission delay and local calculation delay of each terminal device in the terminal device group, and the terminal devices in the terminal device group are terminal devices participating in horizontal federated learning.
  • the local computing delay is the time required to determine the update parameters of the horizontal federated learning model.
  • the transmission delay is the time required to transmit the updated parameters to the application function network element. It can be understood that, for different terminal devices in the terminal device group, the local calculation delays of different terminal devices may be different, and the transmission delays of different terminal devices may also be different. Exemplarily, as shown in FIG.
  • the terminal device group includes UE1, UE2, and UE3, each UE corresponds to a time axis, and the duration from the starting point of the time axis (that is, the local calculation start time) to the triangle mark is the The UE's local calculation completion time, that is, the local calculation delay, and the corresponding time length from the triangle mark to the circle mark is the transmission delay of the UE.
  • the local calculation start time is the start time of determining the update parameters of the model of the horizontal federated learning.
  • the application function network element may determine the transmission delay of the terminal device sending the data packet according to the sending time of the data packet carrying the update parameter and the receiving time of the data packet. For details, refer to the relevant description in FIG. 10 .
  • the application function network element can obtain the local computing delay of the terminal device participating in the horizontal federated learning before the horizontal federated learning starts, or obtain the local computing delay of the terminal device participating in the horizontal federated learning during the model training phase of the horizontal federated learning delay.
  • the relevant description in FIG. 10 refer to the relevant description in FIG. 10 .
  • Step 910 The application function network element adjusts the QoS parameter of the first terminal device in the terminal device group according to the transmission delay and local calculation delay of each terminal device.
  • the application function network element adjusts the QoS parameter of the first terminal device in the terminal device group according to the transmission delay and local calculation delay of each terminal device
  • the application function network element adjusts the QoS parameters of the first terminal device in the terminal device group according to the transmission delay and The local calculation delay determines the statistical value of the total delay of each terminal device, wherein the total delay of each terminal device is the sum of the transmission delay of the terminal device and the local calculation delay.
  • the application function network element adjusts the QoS parameter of the first terminal device, where the first preset threshold and the second preset Let the threshold be determined according to the statistical value.
  • the statistical value may be an arithmetic mean, or a geometric mean, or a root mean square mean, or a harmonic mean, or a weighted mean, etc., which are not limited in the present application.
  • the first preset threshold and the second preset threshold are determined by the application function network element.
  • the first preset threshold may be the statistical value of the total delay+20ms, or the statistical value of the total delay*1.12, etc.
  • the second preset threshold may be the statistical value of the total delay - 20 ms, or the statistical value of the total delay * 0.88, and so on.
  • the adjusted QoS parameters of the first terminal device include at least one of resource type, priority, packet delay budget, guaranteed stream bit rate, maximum stream bit rate, allocation and preemption priority.
  • resource types may include GBR, Non-GBR, and delay-critical (Delay-critical) GBR.
  • the priority may refer to the priority of resource scheduling, that is, the priority level (priority level) parameter in the 5QI.
  • Allocation and preemption priority refers to the priority of resource allocation and preemption.
  • the transmission delay of the QoS flow of the first terminal device can be reduced, and by increasing the packet delay budget, the transmission delay of the QoS flow of the first terminal device can be increased.
  • the stability of the QoS Flow bit rate can be guaranteed, thereby ensuring the stability of the time delay.
  • the application function network element can also adjust parameters such as packet error rate, average window, and maximum data burst volume.
  • Step 920 The application function network element sends the adjusted QoS parameter of the first terminal device to the policy control network element.
  • the application function network element also adjusts the QoS reference of the first terminal device in the terminal device group according to the transmission delay and local calculation delay of each terminal device.
  • the QoS reference does not specify specific QoS parameters, and is mapped to specific QoS parameters by the policy control network element. That is to say, the policy control network element can adjust the QoS parameter of the first terminal device in the terminal device group according to the QoS reference of the first terminal device.
  • the application function network element sends the adjusted QoS reference of the first terminal device to the policy control network element.
  • the application function network element sends the adjusted expected transmission delay of the first terminal device to the policy control network element, and the policy control network element maps it to specific QoS parameters according to the expected transmission delay of the first terminal device, for example, At least one of resource type, priority, packet delay budget, guaranteed stream bit rate, maximum stream bit rate, allocation and preemption priority.
  • FIG. 9 The embodiment shown in FIG. 9 is described below in conjunction with FIG. 10:
  • the application function network element before the model training phase of horizontal federated learning, can set different expected transmission delays for each terminal device according to the local computing delays of the terminal devices participating in horizontal federated learning, so that The sum of the local computing delay and the expected transmission delay of different terminal devices is roughly the same, which is used to improve the model training efficiency of horizontal federated learning.
  • the application function network element can obtain the sum of the local computing delay and the actual transmission delay of different terminal devices, that is, the total delay.
  • the application function network element can calculate the statistical value of the total delay of different terminal devices, that is, the expected total delay, and filter out the terminal devices with a large difference between the total delay and the expected total delay, and adjust the QoS of these terminal devices parameters, so that the total delay of these terminal devices can be roughly the same as the expected total delay.
  • S1001.AF obtains the expected transmission delay of each UE according to the local calculation delay and the initial total delay of each UE in the terminal device group, and accordingly sets the QoS reference for the QoS Flow of each UE for the federated learning service value or QoS parameter value.
  • the AF can obtain the local computing delay and the initial total delay of the UE participating in the horizontal federated learning during the test phase.
  • the testing phase refers to the time period before the AF determines the UE participating in the horizontal federated learning, establishes a session connection with the UE participating in the horizontal federated learning, and before the AF and the UE perform horizontal federated learning model training.
  • the testing phase can also be called the preparation phase.
  • the AF interacts with the UE through application layer data packets to obtain the UE's local calculation delay and initial total delay, and then obtain the local calculation delay and initial total delay of each UE participating in horizontal federated learning. delay.
  • the AF may set a deadline, so that the intermediate results (ie update parameters) of the UEs all arrive at the AF before the deadline. Furthermore, AF can calculate the expected transmission delay of different UEs according to the deadline and the local calculation delay of each UE, and set the QoS reference value or QoS of the UE's QoS Flow according to the expected transmission delay of each UE Parameter value, for example, PDB.
  • a terminal device group includes UE1, UE2, and UE3.
  • the local calculation delay of UE1 is 20ms, the total delay of UE1 is 90ms, the local calculation delay of UE2 is 40ms, the total delay of UE2 is 70ms, and the total delay of UE3 is 90ms.
  • the local calculation delay is 30ms, and the total delay of UE3 is 80ms.
  • AF sets the cut-off time according to the above three total delays. If the average value of the above three total delays calculated by AF is 80ms, the cut-off time is set to 80ms, that is, from the local calculation start time to the receipt of all UE data packets, The required time is 80ms.
  • the expected transmission delay of UE1 is 60ms
  • the expected transmission delay of UE2 is 40ms
  • the expected transmission delay of UE3 is 50ms.
  • AF sets UE1 according to UE1's expected transmission delay The PDB of the QoS Flow.
  • the AF sets the PDB of the QoS Flow of UE2 according to the expected transmission delay of UE2, and the AF sets the PDB of the QoS Flow of UE3 according to the expected transmission delay of UE3.
  • the PDB of the QoS Flow of UE2 ⁇ the PDB of the QoS Flow of UE3 ⁇ the PDB of the QoS Flow of UE1.
  • the AF sends a request message to the PCF, where the request message includes a QoS reference value or a QoS parameter value.
  • the AF sends the QoS reference value or QoS parameter value of each UE in the terminal device group to the PCF through the PCF service interface policy authorization creation request (Npcf_PolicyAuthorization_Creat Request), or the AF sends the PCF service interface policy authorization update request (Npcf_PolicyAuthorization_Update Request) sends the QoS reference value or QoS parameter value of the QoS parameter of each UE in the terminal device group to the PCF.
  • Npcf_PolicyAuthorization_Creat Request the PCF service interface policy authorization creation request
  • Npcf_PolicyAuthorization_Update Request sends the QoS reference value or QoS parameter value of the QoS parameter of each UE in the terminal device group to the PCF.
  • the request message includes UE identification (UE ID), application session identification (App session ID), QoS reference or QoS parameters (QoS reference or QoS parameters), wherein the UE identification is used to indicate the UE whose QoS parameters are adjusted.
  • the application session identifier (App session ID) is used to indicate the QoS parameters of the adjusted application session.
  • the QoS reference or QoS parameter indicates the adjusted QoS parameter, for example, the QoS reference is the transmission delay of the terminal equipment, and the QoS parameter is the PDB.
  • the PCF sets corresponding QoS parameters according to the request message, and generates PCC rules.
  • the PCF sends the PCC rule to the SMF, so that the SMF binds the PCC rule to the corresponding QoS Flow.
  • the PCF sends a response message to the AF.
  • the response message indicates that the QoS parameter setting is successful.
  • the PCF when the AF sends the QoS reference value or QoS parameter value of each UE in the terminal device group to the PCF through the PCF service interface policy authorization creation request, the PCF sends the PCF service interface policy authorization creation response (Npcf_PolicyAuthorization_Create Response) to the PCF.
  • the AF notifies that the QoS parameter setting is successful.
  • the PCF authorizes the update response (Npcf_PolicyAuthorization_Update Response) through the PCF service interface policy Notify the AF that the QoS parameters are set successfully.
  • the PCF may also notify the AF that the setting of QoS parameters fails, for example, the network resources are insufficient, and the QoS parameters specified by the AF cannot be satisfied.
  • AF acquires the transmission delay and local calculation delay of each UE in the terminal device group.
  • the UE when the UE sends a data packet to the AF, it will carry the sending time of the data packet as a timestamp in the data packet.
  • the AF After the AF receives the data packet, it can The transmission delay of the data packet is obtained by calculating the above timestamp.
  • the data packet carries the updated parameters of the horizontal federated learning model.
  • the transmission delay of each UE acquired by the AF may refer to an average value of transmission delays of all data packets in a certain training round.
  • the UE sends a total of 10 data packets in a certain training round, and the AF determines the transmission delay of each data packet according to the above method, and then determines the average value of the transmission delay of the 10 data packets as the transmission delay of the UE. delay.
  • the AF can obtain the UE's local calculation delay during the test phase.
  • the sending time of the data packet will be carried in the data packet as a timestamp.
  • the AF can start the calculation according to the timestamp and local calculation. Time, to determine the local calculation delay of the UE.
  • the data packet carries the updated parameters of the horizontal federated learning model.
  • the local calculation delay of each UE that may be acquired by the AF may refer to an average value of the local calculation delays of all data packets in a certain training round. It can be understood that the local computation start time of each UE participating in horizontal federated learning is roughly the same.
  • AF adjusts the QoS reference value or QoS parameter value of UE1 in the terminal device group according to the transmission delay and local calculation delay of each UE.
  • the AF calculates the sum of the transmission delay and the local calculation delay of each UE as the total delay of the UE, and then determines the statistical value of the total delay of each UE according to the total delay of each UE, that is, the expected total delay.
  • the statistical value of the total delay of each UE may be an arithmetic mean, a geometric mean, a root mean square mean, a harmonic mean, or a weighted mean of the total delay of each UE.
  • the AF adjusts the QoS reference value or QoS parameter value of UE1.
  • the AF determines that the arithmetic mean value of the total delay of each UE is 50 ms, and the AF sets the first preset threshold value to the arithmetic mean value + 10 ms, that is, 60 ms, and sets the second preset threshold
  • the threshold is the arithmetic mean value -10ms, that is, 40ms, and the AF screens out UEs whose total delay is less than 40ms and UEs whose total delay is greater than 60ms, and adjusts the QoS reference value or QoS parameter value of these UEs.
  • the AF can increase the PDB corresponding to the UE, and/or decrease the ARP of the UE's QoS Flow.
  • the AF may lower the PDB corresponding to the UE, and/or increase the ARP of the UE's QoS Flow.
  • the probability of achieving the set expected transmission delay can be increased.
  • AF can set QoS parameters such as GBR and/or MBR of QoS Flow to ensure the stability of QoS Flow bit rate, and then ensure the stability of delay.
  • GBR can also be GFBR
  • MBR can also be MFBR.
  • the AF sends a request message to the PCF, where the request message is used to request to adjust the QoS parameter of UE1.
  • the AF sends the adjusted QoS reference value or QoS parameter value of UE1 to the PCF through a PCF service interface policy authorization update request (Npcf_PolicyAuthorization_Update Request).
  • Npcf_PolicyAuthorization_Update Request a PCF service interface policy authorization update request
  • the request message includes UE1ID, App session ID, adjusted QoS reference value or QoS parameter value.
  • the PCF sets corresponding QoS parameters according to the request message, and generates PCC rules.
  • the PCF updates the corresponding QoS parameters according to the request message.
  • the PCF may change the type of QoS Flow from Non-GBR to GBR or Delay-critical GBR, and generate updated PCC rules. For example, only GBR and Delay-critical GBR types of QoS Flow have GBR and/or MBR parameters, so if AF specifies the value of GBR and/or MBR, or AF sets the reference value of the flow bit rate, it means If the type of the QoS Flow is Non-GBR, the PCF will change the type of the QoS Flow to GBR or Delay-critical GBR.
  • the PCF sends the updated PCC rule to the SMF, so that the SMF binds the updated PCC rule to the corresponding QoS Flow.
  • the PCF sends a response message to the AF.
  • the response message indicates that the QoS parameter setting is successful.
  • the PCF notifies the AF that the QoS parameter setting is successful through the PCF service interface policy authorization update response (Npcf_PolicyAuthorizationUpdate Response) service operation.
  • the application function network element when determining the QoS parameter adjustment strategy of the terminal equipment, the application function network element does not need to rely on the NWDAF or PCF network element to obtain the data of the terminal equipment from the UPF, which can reduce the burden on other network elements (such as NWDAF, PCF, UPF), The QoS parameter adjustment efficiency of the terminal equipment is improved.
  • Fig. 11 shows a possible exemplary block diagram of a communication device involved in the embodiment of the present application.
  • the device 1100 includes: a transceiver module 1120 and a processing module 1110, and the transceiver module 1120 may include a receiving unit and a sending unit.
  • the processing module 1110 is used to control and manage the actions of the device 1100 .
  • the transceiver module 1120 is used to support communication between the device 1100 and other network entities.
  • the device 1100 may further include a storage unit for storing program codes and data of the device 1100 .
  • each module in the apparatus 1100 may be implemented by software.
  • the processing module 1110 may be a processor or a controller, such as a general-purpose central processing unit (central processing unit, CPU), a general-purpose processor, digital signal processing (digital signal processing, DSP), an application-specific integrated circuit (application specific integrated circuits, ASIC), field programmable gate array (field programmable gate array, FPGA) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It can realize or execute various exemplary logical blocks, modules and circuits described in conjunction with the disclosed content of the embodiments of the present application.
  • the processor may also be a combination of computing functions, for example, a combination of one or more microprocessors, a combination of DSP and a microprocessor, and so on.
  • the transceiver module 1120 may be a communication interface, a transceiver or a transceiver circuit, etc., wherein the communication interface is collectively referred to as, in a specific implementation, the communication interface may include multiple interfaces, and the storage unit may be a memory.
  • the processing module 1110 in the device 1100 can support the device 1100 to execute the actions of the first network element in the method examples above, for example, it can support the device 1100 to execute the Step 400 or step 420 in 4.
  • the transceiver module 1120 may support the device 1100 to communicate with the second network element, for example, the transceiver module 1120 may support the device 1100 to execute step 400, step 410, and step 430 in FIG. 4 .
  • the transceiver module 1120 is configured to send a first request message to a second network element, where the first request message is used to request data of a terminal device group, where the terminal devices in the terminal device group are participating in the horizontal A terminal device for federated learning; receiving data of the terminal device group from the second network element, where the data of the terminal device group includes aggregation information of measurement results of QoS parameters; the processing module 1110 is configured to, according to the The data of the terminal device group adjusts the QoS parameter of the first terminal device in the terminal device group; the transceiver module 1120 is configured to send the adjusted QoS parameter of the first terminal device to a policy control network element.
  • the first request message includes indication information, and the indication information is used to indicate the measurement result of the QoS parameter of the feedback terminal equipment;
  • the data of the terminal equipment group also includes A measurement result of a QoS parameter of a second terminal device, where the second terminal device includes the first terminal device.
  • the first request message further includes a feedback condition, and the feedback condition is used to indicate a condition to be met for feeding back the measurement result of the QoS parameter of the terminal device; The measurement result of the QoS parameter satisfies the feedback condition.
  • the aggregated information includes an aggregated bit rate, and the aggregated bit rate is used to characterize the sum of the bit rates of the QoS flows of the terminal devices in the terminal device group; the processing module 1110 uses When adjusting the QoS parameter of the first terminal device in the terminal device group according to the data of the terminal device group, if the aggregated bit rate is greater than the contracted maximum bit rate of the terminal device group, according to the terminal device group The data adjusts the bit rate of the QoS flow of the first terminal device.
  • the aggregation information includes statistical values of transmission delays of terminal devices in the terminal device group; the processing module 1110 is configured to adjust the When the QoS parameter of the first terminal device in the terminal device group, if the statistical value of the transmission delay is greater than the first preset threshold, according to the terminal The data of the device group adjusts the bit rate of the QoS flow of the first terminal device.
  • the first terminal device is a terminal device with a transmission delay greater than a second preset threshold.
  • the device 1100 may correspond to the first network element in the foregoing method embodiments, and the operations and/or functions of the various modules in the device 1100 are respectively to implement the first network element in the foregoing method embodiments
  • the corresponding steps of the method can therefore also achieve the beneficial effects of the foregoing method embodiments, and for the sake of brevity, details are not described here.
  • the processing module 1110 in the device 1100 may support the device 1100 to execute the actions of the second network element in the above method examples.
  • the transceiver module 1120 may support the device 1100 to communicate with the first network element, for example, the transceiver module 1120 may support the device 1100 to execute step 400, step 410, and step 430 in FIG. 4 .
  • the processing module 1110 invokes the transceiver module 1120 to perform: receiving a first request message from a first network element, where the first request message is used to request data of a terminal device group, where the terminal device group
  • the terminal device is a terminal device participating in the horizontal federated learning; the data of the terminal device group is sent, and the data of the terminal device group includes aggregation information of measurement results of QoS parameters.
  • the first request message includes indication information, and the indication information is used to indicate the measurement result of the QoS parameter of the feedback terminal equipment;
  • the data of the terminal equipment group also includes A measurement result of a QoS parameter of a second terminal device, where the second terminal device includes the first terminal device.
  • the first request message further includes a feedback condition, and the feedback condition is used to indicate a condition to be met for feeding back the measurement result of the QoS parameter of the terminal device; The measurement result of the QoS parameter satisfies the feedback condition.
  • the aggregated information includes an aggregated bit rate, where the aggregated bit rate is used to represent a sum of bit rates of QoS flows of terminal devices in the terminal device group.
  • the aggregation information includes statistical values of transmission delays of terminal devices in the terminal device group.
  • the transceiver module 1120 is further configured to send a second request message to a user plane network element, where the second request message is used to request data of the terminal device group; from the user plane network element Data of the group of terminal devices is received.
  • the transceiver module 1120 is further configured to send a third request message to a user plane network element, where the third request message is used to request the QoS parameters of the terminal devices in the terminal device group Measurement results; receiving the measurement results of the QoS parameters of the terminal devices in the terminal device group from the user plane network element; the processing module 1110 is further configured to, according to the Measurement results and said first request message determine data of said group of terminal devices.
  • the device 1100 may correspond to the second network element in the foregoing method embodiments, and the operations and/or functions of the various modules in the device 1100 are respectively to implement the second network element in the foregoing method embodiments
  • the corresponding steps of the method can therefore also achieve the beneficial effects of the foregoing method embodiments, and for the sake of brevity, details are not described here.
  • the processing module 1110 in the device 1100 can support the device 1100 to execute the actions of the application function network element in the above method examples.
  • the processing module 1110 may support the apparatus 1100 to execute step 900 and step 910 in FIG. 9 .
  • the transceiver module 1120 may support the device 1100 to communicate with the policy control network element, for example, the transceiver module 1120 may support the device 1100 to execute step 920 in FIG. 9 .
  • the processing module 1110 is configured to obtain the transmission delay and the local calculation delay of each terminal device in the terminal device group.
  • the terminal devices in the terminal device group are terminal devices participating in horizontal federated learning, and according to the transmission delay and local calculation delay of each terminal device, adjust the time delay of the first terminal device in the terminal device group QoS parameters;
  • the transceiver module 1120 is configured to send the adjusted QoS parameters of the first terminal device to a policy control network element.
  • the processing module 1110 when adjusting the QoS parameter of the first terminal device in the terminal device group according to the transmission delay and local calculation delay of each terminal device, the processing module 1110 is configured to: The transmission delay and local calculation delay of each terminal device determine the statistical value of the total delay of each terminal device, wherein the total delay of each terminal device is the transmission delay of the terminal device and the local calculating the sum of delays; when the total delay of the first terminal device is greater than a first preset threshold or less than a second preset threshold, adjusting the QoS parameter of the first terminal device, wherein the first A preset threshold and the second preset threshold are determined according to the statistics.
  • the adjusted QoS parameters of the first terminal device include resource type, priority, packet delay budget, guaranteed stream bit rate, maximum stream bit rate, allocation and preemption priority at least one of the .
  • the local calculation delay is the time required to determine the update parameters of the model of the horizontal federated learning
  • the transmission delay is the time required to transmit the update parameters to the application function network element How long it takes.
  • the device 1100 may correspond to the application function network element in the foregoing method embodiments, and the operations and/or functions of each module in the device 1100 are to realize the application function network element in the foregoing method embodiments.
  • the corresponding steps of the method can therefore also achieve the beneficial effects of the foregoing method embodiments, and for the sake of brevity, details are not described here.
  • Fig. 12 shows a schematic structural diagram of a communication device 1200 according to an embodiment of the present application.
  • the apparatus 1200 includes: a processor 1201 .
  • the processor 1201 when the device 1200 is the first network element or a chip in the first network element, in a possible implementation manner, when the processor 1201 is used to call the interface to perform the following actions: send the first request message to the second network element , the first request message is used to request data of a terminal device group, where the terminal devices in the terminal device group are terminal devices participating in horizontal federated learning; receiving the data of the terminal device group from the second network element Data, the data of the terminal equipment group includes aggregation information of measurement results of QoS parameters; adjust the QoS parameters of the first terminal equipment in the terminal equipment group according to the data of the terminal equipment group; send the adjusted QoS parameter to the policy control network element QoS parameters of the first terminal device.
  • apparatus 1200 may also be used to perform other steps and/or operations on the first network element side in the foregoing embodiments, and details are not described here for brevity.
  • the processor 1201 is used to call the interface to perform the following actions:
  • apparatus 1200 may also be used to perform other steps and/or operations on the second network element side in the foregoing embodiments, and details are not described here for brevity.
  • the device 1200 is an application function network element or a chip in an application function network element, in a possible implementation manner, when the processor 1201 is used to call an interface to perform the following actions:
  • the terminal devices in the terminal device group are terminal devices participating in horizontal federated learning. According to the transmission delay and local calculation time of each terminal device Delayed, adjusting the QoS parameter of the first terminal device in the terminal device group; sending the adjusted QoS parameter of the first terminal device to the policy control network element.
  • apparatus 1200 may also be used to perform other steps and/or operations on the application function network element side in the foregoing embodiments, and details are not described here for brevity.
  • the processor 1201 may call an interface to perform the above sending and receiving actions, where the called interface may be a logical interface or a physical interface, which is not limited thereto.
  • the physical interface can be implemented through a transceiver.
  • the apparatus 1200 further includes a transceiver 1203 .
  • the apparatus 1200 further includes a memory 1202, and the memory 1202 may store the program codes in the foregoing method embodiments, so as to be called by the processor 1201.
  • the apparatus 1200 includes a processor 1201, a memory 1202, and a transceiver 1203, the processor 1201, the memory 1202, and the transceiver 1203 communicate with each other through an internal connection path to transmit control and/or data signals.
  • the processor 1201, the memory 1202, and the transceiver 1203 may be implemented by a chip, and the processor 1201, the memory 1202, and the transceiver 1203 may be implemented in the same chip, or may be implemented in different chips respectively, Or a combination of any two of these functions can be implemented in one chip.
  • the memory 1202 may store program codes, and the processor 1201 calls the program codes stored in the memory 1202 to implement corresponding functions of the apparatus 1200 .
  • a processor may be an integrated circuit chip with signal processing capabilities.
  • each step of the above-mentioned method embodiments may be completed by an integrated logic circuit of hardware in a processor or instructions in the form of software.
  • the above-mentioned processor can be a general-purpose processor, a digital signal processor (digital signal processor, DSP), an application specific integrated circuit (application specific integrated circuit, ASIC), an off-the-shelf programmable gate array (field programmable gate array, FPGA) or other available Programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, system on chip (system on chip, SoC), central processing unit (central processor unit, CPU), or network processor (network processor, NP), can also be a digital signal processing circuit (digital signal processor, DSP), can also be a microcontroller (micro controller unit, MCU), can also be a programmable controller (programmable logic device, PLD) or other Integrated chip.
  • DSP digital signal processor
  • ASIC application specific integrated circuit
  • FPGA off-the-shelf programmable gate array
  • FPGA field programmable gate array
  • Programmable logic devices discrete gate or transistor logic devices, discrete hardware components, system on chip (system on chip, SoC), central processing unit
  • a general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like.
  • the steps of the method disclosed in connection with the embodiments of the present application may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor.
  • the software module can be located in a mature storage medium in the field such as random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, register.
  • the storage medium is located in the memory, and the processor reads the information in the memory, and completes the steps of the above method in combination with its hardware.
  • the memory in the embodiments of the present application may be a volatile memory or a nonvolatile memory, or may include both volatile and nonvolatile memories.
  • the non-volatile memory can be read-only memory (read-only memory, ROM), programmable read-only memory (programmable ROM, PROM), erasable programmable read-only memory (erasable PROM, EPROM), electrically programmable Erases programmable read-only memory (electrically EPROM, EEPROM) or flash memory.
  • Volatile memory can be random access memory (RAM), which acts as external cache memory.
  • RAM random access memory
  • static RAM static random access memory
  • dynamic RAM dynamic random access memory
  • DRAM dynamic random access memory
  • synchronous dynamic random access memory synchronous DRAM, SDRAM
  • double data rate synchronous dynamic random access memory double data rate SDRAM, DDR SDRAM
  • enhanced synchronous dynamic random access memory enhanced SDRAM, ESDRAM
  • serial link DRAM SLDRAM
  • direct memory bus random access memory direct rambus RAM, DR RAM
  • sequence numbers of the above-mentioned processes do not mean the order of execution, and the order of execution of the processes should be determined by their functions and internal logic.
  • the various numerical numbers or serial numbers involved in the above-mentioned various processes are only for convenience of description, and shall not constitute any limitation to the implementation process of the embodiment of the present application.
  • the above is an example of the three elements of A, B and C to illustrate the optional items of the project.
  • the expression includes at least one of the following: A, B, ..., and X"
  • the applicable entries for this item can also be obtained according to the aforementioned rules.
  • the disclosed systems, devices and methods may be implemented in other ways.
  • the device embodiments described above are only illustrative.
  • the division of the units is only a logical function division. In actual implementation, there may be other division methods.
  • multiple units or components can be combined or May be integrated into another system, or some features may be ignored, or not implemented.
  • the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be in electrical, mechanical or other forms.
  • the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
  • each functional unit in each embodiment of the present application 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 functions described above are realized in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium.
  • the technical solution of the present application is essentially or the part that contributes to the prior art or the 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 A number of instructions are used to make a computer device (which may be a personal computer) computer, server, or network equipment, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application.
  • the aforementioned storage medium includes: U disk, mobile hard disk, read-only memory ROM, random access memory RAM, magnetic disk or optical disk, and other media capable of storing program codes.

Landscapes

  • Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

一种通信方法及装置，该方法包括：第一网元向第二网元发送第一请求消息，第一请求消息用于请求终端设备组的数据，其中，终端设备组中的终端设备为参与横向联邦学习的终端设备。第一网元从第二网元接收终端设备组的数据，终端设备组的数据包括QoS参数的测量结果的聚合信息，第一网元根据终端设备组的数据调整终端设备组中第一终端设备的QoS参数，第一网元向策略控制网元发送调整后的第一终端设备的QoS参数。采用上述方法可以提高联邦学习的模型训练的效率。

Description

一种通信方法及装置
相关申请的交叉引用
本申请要求在2022年01月28日提交中国专利局、申请号为202210108437.1、申请名称为“一种通信方法及装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中；本申请要求在2022年03月29日提交中国专利局、申请号为202210326157.8、申请名称为“一种通信方法及装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请实施例涉及无线通信领域，尤其涉及一种通信方法及装置。
背景技术
联邦学习(federated learning，FL)是一个机器学习框架，能有效帮助多个用户在满足用户隐私保护、数据安全和政府法规的要求下，进行数据使用和机器学习建模。联邦学习作为分布式的机器学习范式，可以有效解决数据孤岛问题，在不共享用户数据的基础上进行联合建模，进而从技术上打破数据孤岛，实现人工智能(artificial intelligence，AI)协作。
在联邦学习的模型训练过程中，现有的服务质量(quality of service，QoS)参数调整方法可能会影响联邦学习的模型训练的效率。
发明内容
本申请提供一种通信方法及装置，用以提高联邦学习的模型训练的效率。
第一方面，本申请提供一种通信方法，该方法包括：第一网元向第二网元发送第一请求消息，所述第一请求消息用于请求终端设备组的数据，其中，所述终端设备组中的终端设备为参与横向联邦学习的终端设备；所述第一网元从所述第二网元接收所述终端设备组的数据，所述终端设备组的数据包括QoS参数的测量结果的聚合信息；所述第一网元根据所述终端设备组的数据调整所述终端设备组中第一终端设备的QoS参数；所述第一网元向策略控制网元发送调整后的所述第一终端设备的QoS参数。
采用上述方法，第一网元向第二网元订阅终端设备组的数据，第二网元将终端设备组的数据发送至第一网元，第一网元根据终端设备组的数据调整终端设备组中至少一个终端设备的QoS参数，进而可以实现提高联邦学习的模型训练的效率。
在一种可能的设计中，所述第一请求消息包括指示信息，所述指示信息用于指示反馈终端设备的QoS参数的测量结果；所述终端设备组的数据还包括所述终端设备组中的第二终端设备的QoS参数的测量结果，所述第二终端设备包括所述第一终端设备。
采用上述设计，第二网元还可以向第一网元发送终端设备组中的终端设备的QoS参数的测量结果。
在一种可能的设计中，所述第一请求消息还包括反馈条件，所述反馈条件用于指示反馈所述终端设备的QoS参数的测量结果所需满足的条件；所述第二终端设备的QoS参数 的测量结果满足所述反馈条件。
采用上述设计，可以实现第二网元反馈满足反馈条件的终端设备的QoS参数的测量结果。
在一种可能的设计中，所述聚合信息包括聚合比特率，所述聚合比特率用于表征所述终端设备组中的终端设备的QoS流的比特率之和；在所述第一网元根据所述终端设备组的数据调整所述终端设备组中第一终端设备的QoS参数时，若所述聚合比特率大于所述终端设备组的签约最大比特率，所述第一网元根据所述终端设备组的数据调整所述第一终端设备的QoS流的比特率。
示例性地，第一网元可以根据所述终端设备组的数据调整至少一个终端设备的QoS流的比特率，所述至少一个终端设备包括所述第一终端设备，在调整所述至少一个终端设备的QoS流的比特率之后，所述终端设备组中的各个终端设备的QoS流的比特率之和小于或者等于所述签约最大比特率。
采用上述设计，可以实现调整终端设备QoS参数，使得终端设备组中的各个终端设备的QoS流的比特率之和小于或者等于所述签约最大比特率。
在一种可能的设计中，所述聚合信息包括所述终端设备组中的终端设备的传输时延的统计值；在所述第一网元根据所述终端设备组的数据调整所述终端设备组中第一终端设备的QoS参数时，若所述传输时延的统计值大于第一预设阈值，则所述第一网元根据所述终端设备组的数据调整所述第一终端设备的QoS流的比特率。
其中，终端设备组中的终端设备的传输时延的统计值可以是根据终端设备组中的终端设备的传输时延确定的方差。示例性地，终端设备组中的终端设备的传输时延的统计值是指由终端设备组中的终端设备的加权传输时延或差分传输时延确定的方差。
采用上述设计，可以实现调整终端设备QoS参数，使得传输时延的统计值小于或等于第一预设阈值。
在一种可能的设计中，所述第一终端设备为传输时延大于第二预设阈值或小于第三预设阈值的终端设备，其中，所述第二预设阈值大于所述第三预设阈值。
采用上述设计，可以实现调低传输时延较长的终端设备的传输时延，以及为了满足终端设备组中的各个终端设备的QoS流的比特率之和小于或者等于所述签约最大比特率，还可以调高传输时延较短的终端设备的传输时延，实现终端设备组中各个终端设备发送至服务器的数据几乎同时到达服务器，从而提高联邦学习的模型训练的效率。
在一种可能的设计中，所述第一请求消息还包括所述第一预设阈值和/或所述第二预设阈值。
在一种可能的设计中，所述第一请求消息包括所述终端设备组包括的终端设备的标识、所述终端设备组的标识、分析类型的标识信息、所述聚合信息的类型中的至少一种。
在一种可能的设计中，所述第一请求消息还包括以下至少一项：发送所述聚合信息的触发条件；应用标识，用于指示终端设备的QoS参数的测量结果对应的应用。
其中，发送聚合信息的触发条件可以包括周期性反馈聚合信息，或者满足预设条件反馈聚合信息。例如，这里的预设条件可以包括聚合比特率大于终端设备组的签约最大比特率，和/或终端设备组中的终端设备的传输时延的统计值大于第一预设阈值等。
在一种可能的设计中，所述第一网元为应用功能网元，所述第二网元为数据分析功能网元或策略控制网元。
第二方面，本申请提供一种通信方法，该方法包括：第二网元从第一网元接收第一请求消息，所述第一请求消息用于请求终端设备组的数据，其中，所述终端设备组中的终端设备为参与横向联邦学习的终端设备；所述第二网元发送所述终端设备组的数据，所述终端设备组的数据包括QoS参数的测量结果的聚合信息。
采用上述方法，第一网元向第二网元订阅终端设备组的数据，第二网元将终端设备组的数据发送至第一网元，以使第一网元根据终端设备组的数据调整终端设备组中至少一个终端设备的QoS参数，进而可以实现提高联邦学习的模型训练的效率。
在一种可能的设计中，所述第一请求消息包括指示信息，所述指示信息用于指示反馈终端设备的QoS参数的测量结果；所述终端设备组的数据还包括所述终端设备组中的第二终端设备的QoS参数的测量结果，所述第二终端设备包括所述第一终端设备。
采用上述设计，第二网元还可以向第一网元发送终端设备组中的终端设备的QoS参数的测量结果。
在一种可能的设计中，所述第一请求消息还包括反馈条件，所述反馈条件用于指示反馈所述终端设备的QoS参数的测量结果所需满足的条件；所述第二终端设备的QoS参数的测量结果满足所述反馈条件。
采用上述设计，可以实现第二网元反馈满足反馈条件的终端设备的QoS参数的测量结果。
在一种可能的设计中，所述聚合信息包括聚合比特率，所述聚合比特率用于表征所述终端设备组中的终端设备的QoS流的比特率之和。
在一种可能的设计中，所述聚合信息包括所述终端设备组中的终端设备的传输时延的统计值。
在一种可能的设计中，所述第一请求消息包括所述终端设备组中终端设备的标识、所述终端设备组的标识、分析类型的标识信息、所述聚合信息的类型中的至少一种。
在一种可能的设计中，所述第一请求消息还包括以下至少一项：发送所述聚合信息的触发条件；应用标识，用于指示终端设备的QoS参数的测量结果对应的应用。
在一种可能的设计中，还包括：所述第二网元向用户面网元发送第二请求消息，所述第二请求消息用于请求所述终端设备组的数据；所述第二网元从用户面网元接收所述终端设备组的数据。
采用上述设计，第二网元可以从用户面网元直接获取终端设备组的数据。
在一种可能的设计中，还包括：所述第二网元向用户面网元发送第三请求消息，所述第三请求消息用于请求所述终端设备组中的终端设备的QoS参数的测量结果；所述第二网元从所述用户面网元接收所述终端设备组中的终端设备的QoS参数的测量结果；所述第二网元根据所述终端设备组中的终端设备的QoS参数的测量结果和所述第一请求消息确定所述终端设备组的数据。
采用上述设计，第二网元可以对从用户面网元获得的终端设备的QoS参数的测量结果进行聚合，获得终端设备组的数据。
在一种可能的设计中，所述第二网元为策略控制网元；所述方法，还包括：所述第二网元从所述第一网元接收所述终端设备组包括的各个终端设备分别对应的QoS流的比特率；所述第二网元从第三网元接收所述终端设备组的签约最大比特率，所述第三网元为统一数据存储网元或统一数据管理网元；所述第二网元根据所述终端设备组的签约最大比特 率和所述终端设备组包括的各个终端设备分别对应的QoS流的比特率，确定所述终端设备组包括的各个终端设备分别对应的QoS流的比特率之和小于所述终端设备组的签约最大比特率；所述第二网元生成所述终端设备组包括的各个终端设备分别对应的PCC规则；所述第二网元向会话管理网元发送所述PCC规则。
在一种可能的设计中，所述第二网元为策略控制网元；所述方法，还包括：所述第二网元从所述第一网元接收所述终端设备组包括的各个终端设备分别对应的QoS流的比特率参考范围；所述第二网元从第三网元接收所述终端设备组的签约最大比特率，所述第三网元为统一数据存储网元或统一数据管理网元；所述第二网元根据所述终端设备组的签约最大比特率和所述终端设备组包括的各个终端设备分别对应的QoS流的比特率参考范围，确定所述终端设备组包括的各个终端设备分别对应的QoS流的比特率参考范围的下界之和小于所述终端设备组的签约最大比特率；所述第二网元生成所述终端设备组包括的各个终端设备分别对应的PCC规则；所述第二网元向会话管理网元发送所述PCC规则。
采用上述设计，策略控制网元可以根据终端设备组包括的各个终端设备分别对应的QoS流的比特率的取值，或终端设备组包括的各个终端设备分别对应的QoS流的比特率的取值范围确定是否满足终端设备组的签约最大比特率。
在一种可能的设计中，还包括：所述第二网元向所述第一网元发送第二信息，所述第二信息指示已为所述终端设备组包括的各个终端设备分别生成相应的PCC规则。
在一种可能的设计中，所述第一网元为应用功能网元，所述第二网元为数据分析功能网元或策略控制网元。
第三方面，本申请实施例提供一种通信装置，所述装置为第一网元，或用于实现第一网元的功能的装置，所述装置包括处理模块和收发模块；所述收发模块，用于向第二网元发送第一请求消息，所述第一请求消息用于请求终端设备组的数据，其中，所述终端设备组中的终端设备为参与横向联邦学习的终端设备；从所述第二网元接收所述终端设备组的数据，所述终端设备组的数据包括QoS参数的测量结果的聚合信息；所述处理模块，用于根据所述终端设备组的数据调整所述终端设备组中第一终端设备的QoS参数；所述收发模块，用于向策略控制网元发送调整后的所述第一终端设备的QoS参数。
在一种可能的设计中，所述第一请求消息包括指示信息，所述指示信息指示反馈终端设备的QoS参数的测量结果；
所述终端设备组的数据还包括所述终端设备组中的第二终端设备的QoS参数的测量结果，所述第二终端设备包括所述第一终端设备。
在一种可能的设计中，所述第一请求消息还包括反馈条件，所述反馈条件指示反馈所述终端设备的QoS参数的测量结果所需满足的条件；所述第二终端设备的QoS参数的测量结果满足所述反馈条件。
在一种可能的设计中，所述聚合信息包括聚合比特率，所述聚合比特率用于表征所述终端设备组中的终端设备的QoS流的比特率之和；所述处理模块，用于在根据所述终端设备组的数据调整所述终端设备组中第一终端设备的QoS参数时，若所述聚合比特率大于所述终端设备组的签约最大比特率，根据所述终端设备组的数据调整所述第一终端设备的QoS流的比特率。
在一种可能的设计中，所述聚合信息包括所述终端设备组中的终端设备的传输时延的统计值；所述处理模块，用于在根据所述终端设备组的数据调整所述终端设备组中第一终 端设备的QoS参数时，若所述传输时延的统计值大于第一预设阈值，则根据所述终端设备组的数据调整所述第一终端设备的QoS流的比特率。
在一种可能的设计中，所述第一终端设备为传输时延大于第二预设阈值的终端设备。
在一种可能的设计中，所述第一请求消息还包括所述第一预设阈值和/或所述第二预设阈值。
在一种可能的设计中，所述第一请求消息包括所述终端设备组中终端设备的标识、所述终端设备组的标识、分析类型的标识信息、所述聚合信息的类型中的至少一种。
在一种可能的设计中，所述第一请求消息还包括以下至少一项：发送所述聚合信息的触发条件；应用标识，用于指示终端设备的QoS参数的测量结果对应的应用。
在一种可能的设计中，所述第一网元为应用功能网元，所述第二网元为数据分析功能网元或策略控制网元。
第四方面，本申请提供一种通信装置，所述装置为第二网元，或用于实现第二网元的功能的装置，该装置包括处理模块和收发模块；所述处理模块调用所述收发模块执行：从第一网元接收第一请求消息，所述第一请求消息用于请求终端设备组的数据，其中，所述终端设备组中的终端设备为参与横向联邦学习的终端设备；发送所述终端设备组的数据，所述终端设备组的数据包括QoS参数的测量结果的聚合信息。
在一种可能的设计中，所述第一请求消息包括指示信息，所述指示信息指示反馈终端设备的QoS参数的测量结果；所述终端设备组的数据还包括所述终端设备组中的第二终端设备的QoS参数的测量结果，所述第二终端设备包括所述第一终端设备。
在一种可能的设计中，所述第一请求消息还包括反馈条件，所述反馈条件指示反馈所述终端设备的QoS参数的测量结果所需满足的条件；所述第二终端设备的QoS参数的测量结果满足所述反馈条件。
在一种可能的设计中，所述聚合信息包括聚合比特率，所述聚合比特率用于表征所述终端设备组中的终端设备的QoS流的比特率之和。
在一种可能的设计中，所述聚合信息包括所述终端设备组中的终端设备的传输时延的统计值。
在一种可能的设计中，所述第一请求消息包括所述终端设备组中终端设备的标识、所述终端设备组的标识、分析类型的标识信息、所述聚合信息的类型中的至少一种。
在一种可能的设计中，所述第一请求消息还包括以下至少一项：发送所述聚合信息的触发条件；应用标识，用于指示终端设备的QoS参数的测量结果对应的应用。
在一种可能的设计中，所述收发模块，还用于向用户面网元发送第二请求消息，所述第二请求消息用于请求所述终端设备组的数据；从用户面网元接收所述终端设备组的数据。
在一种可能的设计中，所述收发模块，还用于向用户面网元发送第三请求消息，所述第三请求消息用于请求所述终端设备组中的终端设备的QoS参数的测量结果；从所述用户面网元接收所述终端设备组中的终端设备的QoS参数的测量结果；所述处理模块，还用于根据所述终端设备组中的终端设备的QoS参数的测量结果和所述第一请求消息确定所述终端设备组的数据。
在一种可能的设计中，所述第一网元为应用功能网元，所述第二网元为数据分析功能网元或策略控制网元。
第五方面，本申请提供一种通信方法，该方法包括：应用功能网元获取终端设备组中 各个终端设备的传输时延和本地计算时延，所述终端设备组中的终端设备为参与横向联邦学习的终端设备，所述应用功能网元根据所述各个终端设备的传输时延和本地计算时延，调整所述终端设备组中第一终端设备的QoS参数；所述应用功能网元向策略控制网元发送调整后的所述第一终端设备的QoS参数。
采用上述方法，应用功能网元也可以不通过NWDAF或者PCF收集数据，而是自身统计或获取终端设备组中各个终端设备的传输时延和本地计算时延，进而确定终端设备组中的终端设备的QoS参数的调整策略，减少了网元间交互的消息数量，减轻了网元负载，并提升了终端设备的QoS参数调整效率，使得不同终端设备的本地计算时延和传输时延之和大致相同，从而提升横向联邦学习的模型训练效率。
在一种可能的设计中，在所述应用功能网元根据所述各个终端设备的传输时延和本地计算时延，调整所述终端设备组中第一终端设备的QoS参数时，所述应用功能网元根据所述各个终端设备的传输时延和本地计算时延，确定所述各个终端设备的总时延的统计值，其中，每个终端设备的总时延为所述终端设备的传输时延与本地计算时延之和；在所述第一终端设备的总时延大于第一预设阈值或者小于第二预设阈值的情况下，所述应用功能网元调整所述第一终端设备的QoS参数，其中，所述第一预设阈值与所述第二预设阈值是根据所述统计值确定的。
在一种可能的设计中，所述调整后的所述第一终端设备的QoS参数包括资源类型、优先级、包时延预算、保证流比特率、最大流比特率、分配与抢占优先级中的至少一个。
在一种可能的设计中，所述本地计算时延为确定所述横向联邦学习的模型的更新参数所需的时长；所述传输时延为传输所述更新参数到所述应用功能网元所需的时长。
第六方面，本申请提供一种通信装置，所述装置为应用功能网元，或用于实现应用功能网元的功能的装置，所述装置包括处理模块和收发模块；所述处理模块，用于获取终端设备组中各个终端设备的传输时延和本地计算时延，所述终端设备组中的终端设备为参与横向联邦学习的终端设备，根据所述各个终端设备的传输时延和本地计算时延，调整所述终端设备组中第一终端设备的QoS参数；所述收发模块，用于向策略控制网元发送调整后的所述第一终端设备的QoS参数。
在一种可能的设计中，在根据所述各个终端设备的传输时延和本地计算时延，调整所述终端设备组中第一终端设备的QoS参数时，所述处理模块，用于根据所述各个终端设备的传输时延和本地计算时延，确定所述各个终端设备的总时延的统计值，其中，每个终端设备的总时延为所述终端设备的传输时延与本地计算时延之和；在所述第一终端设备的总时延大于第一预设阈值或者小于第二预设阈值的情况下，调整所述第一终端设备的QoS参数，其中，所述第一预设阈值与所述第二预设阈值是根据所述统计值确定的。
在一种可能的设计中，所述调整后的所述第一终端设备的QoS参数包括资源类型、优先级、包时延预算、保证流比特率、最大流比特率、分配与抢占优先级中的至少一个。
在一种可能的设计中，所述本地计算时延为确定所述横向联邦学习的模型的更新参数所需的时长；所述传输时延为传输所述更新参数到所述应用功能网元所需的时长。
第七方面，本申请还提供一种装置。该装置可以执行上述方法设计。该装置可以是能够执行上述方法对应的功能的芯片或电路，或者是包括该芯片或电路的设备。
在一种可能的实现方式中，该装置包括：存储器，用于存储计算机可执行程序代码；以及处理器，处理器与存储器耦合。其中存储器所存储的程序代码包括指令，当处理器执 行所述指令时，使该装置或者安装有该装置的设备执行上述任意一种可能的设计中的方法。
在一种可能的实现方式中，该装置还可以包括通信接口，该通信接口可以是收发器，或者，如果该装置为芯片或电路，则通信接口可以是该芯片的输入/输出接口，例如输入/输出管脚等。
在一种可能的设计中，该装置包括相应的功能单元，分别用于实现以上方法中的步骤。功能可以通过硬件实现，也可以通过硬件执行相应的软件实现。硬件或软件包括一个或多个与上述功能相对应的单元。
第八方面，本申请提供一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，当所述计算机程序在装置上运行时，执行上述任意一种可能的设计中的方法。
第九方面，本申请提供一种计算机程序产品，所述计算机程序产品包括计算机程序，当所述计算机程序在装置上运行时，执行上述任意一种可能的设计中的方法。
第十方面，本申请提供一种通信系统，该系统包括第一网元和第二网元，所述第一网元用于执行所述第一方面中的任意一种可能的设计中的方法，所述第二网元用于执行所述第二方面中的任意一种可能的设计中的方法。
附图说明
图1为本申请应用的移动通信系统的架构示意图；
图2为本申请实施例中参与横向联邦学习的数据集的示意图；
图3为本申请实施例中横向联邦学习的模型训练过程的示意图；
图4为本申请实施例中一种通信方法的概述流程图；
图5A为本申请实施例中调整前的终端设备组中的终端设备的传输时延的示意图；
图5B为本申请实施例中调整后的终端设备组中的终端设备的传输时延的示意图；
图6为本申请实施例中在第一网元向第二网元发送第一请求消息之前第一网元与其他网元的交互流程图；
图7为本申请实施例中另一种通信方法的概述流程图；
图8为本申请实施例中另一种通信方法的概述流程图；
图9为本申请实施例中另一种通信方法的概述流程图；
图10为本申请实施例中AF与PCF的交互流程图；
图11为本申请实施例中一种通信装置的结构示意图之一；
图12为本申请实施例中一种通信装置的结构示意图之二。
具体实施方式
下面将结合本申请实施例中的附图，对本申请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申请一部分实施例，而不是全部的实施例。本申请的说明书和权利要求书及上述附图中的术语“第一”、第二”以及相应术语标号等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的术语在适当情况下可以互换，这仅仅是描述本申请的实施例中对相同属性的对象在描述时所采用的区分方式。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，以 便包含一系列单元的过程、方法、系统、产品或设备不必限于那些单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它单元。
在本申请的描述中，除非另有说明，“/”表示或的意思，例如，A/B可以表示A或B；本申请中的“和/或”仅仅是一种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。另外，在本申请的描述中，“至少一项”是指一项或者多项，“多项”是指两项或两项以上。“以下至少一项(个)”或其类似表达，是指的这些项中的任意组合，包括单项(个)或复数项(个)的任意组合。例如，a,b,或c中的至少一项(个)，可以表示：a,b,c,a-b,a-c,b-c,或a-b-c，其中a,b,c可以是单个，也可以是多个。
本申请实施例提供的技术方案可以应用于各种通信系统。例如：可以适用于LTE系统或5G系统，也可以适用于其它面向未来的新系统等。本申请实施例对此不作具体限定。此外，术语“系统”可以和“网络”相互替换。以下仅以5G的通信系统架构为例进行说明。
如图1所示，第三代合作伙伴计划(3rd generation partnership project，3GPP)标准制定的5G的通信系统。该通信系统包括终端设备(例如，用户设备(user equipment，UE))，无线接入网(radio access network，RAN)，核心网(Core network，CN)。数据网络(data network，DN)在逻辑上，核心网的网元它们可以分为用户面和控制面两部分，控制面负责移动网络的管理，用户面负责业务数据的传输。
其中，终端设备是移动用户与网络交互的入口，能够提供基本的计算能力，存储能力，向用户显示业务窗口，接收用户操作输入。下一代终端设备(NextGen UE)可以采用新空口技术，与RAN建立信号连接，数据连接，从而传输控制信号和业务数据到移动网络。终端设备可以包括各种具有无线通信功能的手持设备、车载设备、可穿戴设备、计算设备或连接到无线调制解调器的其它处理设备，以及各种形式的终端，移动台(mobile station，MS)，终端(terminal)，软终端等等，例如水表、电表、传感器等。
RAN：部署在靠近终端设备的位置，为特定区域的授权用户提供入网功能，并能够根据用户的级别，业务的需求等确定不同质量的传输隧道来传输用户数据。RAN能够管理自身的资源，合理利用，按需为终端设备提供接入服务，并负责把控制信号和用户数据在终端设备和核心网之间转发。
核心网：负责维护移动网络的签约数据，管理移动网络的网元，为终端设备提供会话管理，移动性管理，策略管理，安全认证等功能。在终端设备附着的时候，为终端设备提供入网认证；在终端设备有业务请求时，为终端设备分配网络资源；在终端设备移动的时候，为终端设备更新网络资源；在终端设备空闲的时候，为终端设备提供快恢复机制；在终端设备去附着的时候，为终端设备释放网络资源；在终端设备有业务数据时，为终端设备提供数据路由功能，如转发上行数据到数据网络；或者从数据网络接收终端设备的下行数据，转发到RAN，从而由RAN发送给终端设备。
数据网络(data network，DN)：为用户提供业务服务的数据网络，一般客户端位于终端设备，服务端位于数据网络。数据网络可以是私有网络，如局域网，也可以是不受运营商管控的外部网络，例如互联网(Internet)，还可以是运营商共同部署的专有网络，例如提供IP多媒体网络子系统(IP multimedia core network subsystem，IMS)服务的网络。
其中，核心网用户面包括用户面功能(user plane function，UPF)；核心网控制面包括接入和移动性管理功能(access and mobility management function，AMF)，会话管理功能 (session management function，SMF)，网络开放功能(network exposure function，NEF)，网络功能仓储功能(NF repository function，NRF)，统一数据管理(unified data management，UDM)，统一数据存储(unified data repository，UDR)，策略控制功能(policy control function，PCF)，应用功能(application function，AF)。
核心网控制面采用服务化架构，控制面网元之间的交互采用服务调用的方式，来替换传统架构中的点对点通信方式。在服务化架构中，控制面网元会向其他控制面网元开放服务，供其他控制面网元调用；在点对点通信中，控制面网元之间通信接口会存储一套特定的消息，只能由接口两端的控制面网元在通信时使用。
以下对核心网中的功能实体的功能进行简单介绍：
1、会话管理网元：主要用于会话管理、终端设备的IP地址分配和管理、选择可管理用户设备平面功能、策略控制、或收费功能接口的终结点以及下行数据通知等。在5G通信中，会话管理网元可以是SMF网元，在未来通信如6G通信中，会话管理功能网元仍可以是SMF网元，或者有其它名称，本申请对此不作限定。Nsmf是SMF提供的基于服务的接口，SMF可以通过Nsmf与其他的网络功能通信。
2、接入管理网元：主要用于移动性管理和接入管理等，例如可以是4G通信网络中的移动性管理实体(mobility management entity，MME)功能或者5G网络中的AMF网元。在未来通信如6G通信中，接入管理网元仍可以是AMF网元，或者有其它名称，本申请对此不作限定。Namf是AMF提供的基于服务的接口，AMF可以通过Namf与其他的网络功能通信。
3、网络开放网元：用于安全地向外部开放由3GPP网络功能提供的业务和能力等。在5G通信中，网络开放网元可以是NEF网元，在未来通信如6G通信中，网络开放功能网元仍可以是NEF网元，或者有其它名称，本申请对此不作限定。其中Nnef是NEF提供的基于服务的接口，NEF可以通过Nnef与其他的网络功能通信。
4、网络存储网元：用于提供服务注册、发现和授权，并维护可用的网络功能(network function，NF)实例信息，可以实现网络功能和服务的按需配置以及NF之间的互连。在5G通信中，网络存储网元可以是NRF网元，在未来通信如6G通信中，网络存储功能网元仍可以是NRF网元，或者有其它名称，本申请对此不作限定。Nnrf是NRF提供的基于服务的接口，NRF可以通过Nnrf与其他的网络功能通信。
5、策略控制网元：用于指导网络行为的统一策略框架，为控制平面功能网元(例如AMF，SMF等)提供策略规则信息等。在5G通信中，策略控制网元可以是PCF网元，在未来通信如6G通信中，策略控制网元仍可以是PCF网元，或者有其它名称，本申请对此不作限定。其中Npcf是PCF提供的基于服务的接口，PCF可以通过Npcf与其他的网络功能通信。
6、数据管理网元：用于处理用户标识、签约、接入鉴权、注册、或移动性管理等。在5G通信中，数据管理网元可以是UDM网元，在未来通信如6G通信中，数据管理网元仍可以是UDM网元，或者有其它名称，本申请对此不作限定。其中Nudm是UDM提供的基于服务的接口，UDM可以通过Nudm与其他的网络功能通信。
7、数据存储网元：用于执行签约数据、策略数据、应用数据等类型数据的存取功能。在5G通信中，数据存储网元可以是UDR网元，在未来通信如6G通信中，数据存储网元 仍可以是UDR网元，或者有其它名称，本申请对此不作限定。其中Nudr是UDR提供的基于服务的接口，UDR可以通过Nudr与其他的网络功能通信。
8、应用网元：用于进行应用影响的数据路由，接入网络开放功能，或与策略框架交互进行策略控制等。在5G通信中，应用网元可以是AF网元，在未来通信如6G通信中，应用网元仍可以是AF网元，或者有其它名称，本申请对此不作限定。Naf是AF提供的基于服务的接口，AF可以通过Naf与其他的网络功能通信。
9、用户面网元：用于分组路由和转发、或用户面数据的服务质量(quality of service，QoS)处理等。在5G通信中，用户面网元可以是用户面功能(user plane function，UPF)网元，在未来通信如6G通信中，用户面网元仍可以是UPF网元，或者有其它名称，本申请对此不作限定。
可以理解的是，核心网还可以包括其他网元，本申请对此不作限定。
为便于理解本申请实施例，对本申请实施例中涉及的几个基本概念做简单说明。
1.横向联邦学习
根据联邦学习的参与各方数据源特征的不同，联邦学习可以被分为三类：横向联邦学习、纵向联邦学习和联邦迁移学习。下面主要介绍一下横向联邦学习的原理。
其中，数据集包括多组数据，每组数据对应一个用户，每组数据记录该组数据对应的用户的至少一个用户特征。
其中，在横向联邦学习的数据集之间，用户特征(X1、X2、X3……)重合度很高，用户(U1、U2、U3……)重合度很低，例如，数据集A包括UE1的数据和UE2的数据，UE1的数据包括的用户特征为X1、X2和X3，UE2的数据包括的用户特征为X1、X2和X3，数据集B包括UE4的数据和UE5的数据，UE4的数据包括的用户特征为X1、X2和X3，UE5的数据包括的用户特征为X1、X2和X3。如图2所示，对于数据集A和数据集B，用户的交集较小，而用户特征的交集较大。因此，数据集A和数据集B可以从各自的数据集中挑选出相同的用户特征进行横向联邦学习的模型训练。需要注意的是，横向联邦学习的参与方并不限于两方，比如在物联网(internet of things，IoT)场景下，横向联邦学习的参与方可能非常庞大。
如图3所示为借助5G系统(5G System，5GS)进行中间结果传输的横向联邦学习模型训练过程。在横向联邦学习的模型训练过程中，首先各个参与方(如图3所示的用户(Client)1、Client2……ClientK)需要通过5GS从服务器(Server)下载最新的模型，以进行模型训练过程的迭代，具体迭代过程一般包括以下几个步骤：
①每个参与方利用本地数据集计算模型的梯度，然后将梯度加密上传到Server。这里的梯度包括模型的参数的变化方向和大小。
②Server根据各个参与方上传的梯度进行梯度聚合，更新模型的参数。这里的梯度聚合一般是指将收集到的梯度求均值。
③Server将更新后的模型的参数分发给各个参与方。
④各个参与方根据更新后的模型的参数更新本地的模型。
2.服务质量(quality of service，QoS)
QoS描述了一组服务需求，网络必须满足这些服务需求才能确保数据传输的适当服务级别。
5G网络需要支持各种各样的业务传输，例如视频、移动支付、网页浏览，工厂自动化 控制等。不同的业务所需要的QoS是不同的。例如，视频业务需要较大的带宽，而自动化控制业务则一般需要较低的时延和较高的可靠性。基于5G网络的QoS框架，运营商能够针对不同的业务提供不同的QoS保证。
在5G网络中，服务质量流(QoS Flow)是协议数据单元(protocol data unit，PDU)会话中进行端到端QoS控制的最细粒度。QoS Flow的QoS控制主要由与其关联的QoS参数确定。基于5G网络的QoS框架支持保证带宽QoS Flow(QoS Flows that require guaranteed flow bit rate，GBR QoS Flow)与非保证带宽的QoS Flow(QoS Flows that do not require guaranteed flow bit rate，Non-GBR QoS Flow)。对于GBR QoS Flow，网络需要预留资源以保证其带宽。对于Non-GBR QoS Flow则不需要预留资源。GBR与non-GBR QoS Flow的对应的QoS参数见表1所示。
表1：不同类型QoS Flow的QoS参数
其中，5QI用于标识一组5G QoS特征。5G QoS特征用于描述QoS流在UE和UPF之间接收端到端的报文转发处理的情况，包括资源类型(如Non-GBR和GBR)、优先级、包时延预算(packet delay budget，PDB)和错包率等。
需要说明的是，上述QoS参数均是针对单个QoS Flow的。除此之外，5G网络中还定义了一些聚合QoS参数。聚合QoS参数用来表征多个QoS Flow的QoS参数的统计特性，例如，聚合最大比特率(aggregate maximum bit rate,AMBR)，它用于限制一组QoS Flow总比特率的最大值。
按照数据粒度的不同，现有5G标准中也定义了一些不同级别的AMBR参数：
PDU会话级的聚合最大比特率(Session-AMBR)：限制了在特定PDU会话的所有Non-GBR QoS Flow中预期提供的聚合比特率。
UE级的聚合最大比特率(UE-AMBR)：适用于Non-GBR QoS Flow。
每个UE的切片级的最大比特率(UE-Slice-MBR)：限制了在特定切片特定UE的PDU会话的所有GBR QoS Flow和Non-GBR QoS Flow中预期提供的聚合比特率。
切片级的最大比特率(Slice-MBR per S-NSSAI)：适用于GBRQoS Flow和Non-GBR QoS Flow。
可以理解的是，本申请提供的方法除了可以应用于横向联邦学习的场景，还可以适用于其他需要服务器与终端设备组中的终端设备进行交互的模型训练过程或业务流程，以下进行横向联邦学习场景为例进行说明。
在联邦学习的模型训练过程中，为了实现调整不同参与方的QoS参数，提升模型训练效率。本申请提供一种通信方法，如图4所示，其中，下述第一网元可以为应用功能网元，第二网元可以为数据分析功能网元或策略控制网元。示例性地，数据分析功能网元可以是 指网络数据分析功能(network data analytics Function，NWDAF)网元，策略控制网元可以是指PCF网元。
该方法包括：
步骤400：第一网元向第二网元发送第一请求消息，第一请求消息用于请求终端设备组的数据，终端设备组的数据包括QoS参数的测量结果的聚合信息。
其中，第一请求消息又可称为第一订阅消息，本申请对第一请求消息的名称不作限定。示例性地，在第二网元为NWDAF网元时，第一请求消息可以为分析订阅消息(Nnwdaf_AnalyticsSubscription_Subscribe)。在第二网元为PCF网元时，第一请求消息可以为事件开放订阅消息(Nnwdaf_EventExposure Subscribe)。
终端设备组包括至少两个终端设备。终端设备组中的终端设备可以为参与横向联邦学习的终端设备。示例性地，终端设备组中的终端设备共同参与同一个横向联邦学习的模型训练过程。
此外，需要说明的是，在第一网元向第二网元发送第一请求消息之前，终端设备组中的终端设备均已建立与横向联邦学习相关联的业务的会话。例如，第一业务为某种横向联邦学习的模型训练业务，终端设备组包括3个终端设备，这3个终端设备为第一业务的参与方，这3个终端设备均已建立关于第一业务的会话。
其中，所述聚合信息又可称为聚合QoS信息，所述聚合信息是根据终端设备组中的各个终端设备的QoS参数的测量结果进行聚合确定的。所述聚合信息用于表征终端设备组包括的各个终端设备的QoS参数的测量结果的统计特性。这里的终端设备的QoS参数可以参考上述关于QoS参数的相关描述。
示例性地，第一请求消息可以包括但不限于以下内容：
1、分析类型的标识信息，用于指示所述聚合信息为终端设备组的聚合信息，即终端设备组级别的聚合信息。
示例性地，该标识信息可以用分析标识(Analytics ID)表示，例如，Analytics ID＝Aggregated QoS for FL group，表示请求FL组(group)的聚合信息。
2、所述聚合信息的类型，用于指示所述聚合信息包括的参数。示例性地，所述聚合信息的类型可以用需要被测量的聚合QoS参数(Aggregated QoS parameters to be measured)表示。
所述聚合信息的类型具体可以包括聚合比特率、终端设备组中的终端设备的传输时延的统计值等。可以理解的是，根据横向联邦学习的模型训练业务需求，所述参数信息还可以包括其他参数，此处不作限定。
其中，聚合比特率用于表征终端设备组中的终端设备的QoS流的比特率之和。此外，根据比特率与带宽的换算关系，聚合比特率又可换算为聚合带宽，聚合带宽用于表征终端设备组中的终端设备对应的带宽之和。
可以理解的是，这里的QoS流为横向联邦学习业务对应的QoS流，如无特别声明，下文中所涉及的QoS流均是指横向联邦学习业务对应的QoS流。这里的QoS流的比特率可以是指GFBR。
示例性地，在横向联邦学习的模型训练过程中，每个参与方需要完成本地计算，并将计算结果上传至服务器，因此，将会消耗大量的带宽资源。为避免对其他终端设备或其他业务产生影响，一般会给横向联邦学习的模型训练业务设定一个最大可用带宽。为了保证 各个参与方的带宽之和不超过设定的最大可用带宽，第一网元可以向第二网元订阅聚合带宽，并在聚合带宽大于设定的最大可用带宽时，调整至少一个参与方的带宽。具体调整方式可以参考下述步骤420中的相关描述。
其中，终端设备组中的终端设备的传输时延的统计值可以是根据终端设备组中的终端设备的传输时延确定的方差。
示例性地，终端设备组中的终端设备的传输时延的统计值是指由终端设备组中的终端设备的加权传输时延确定的方差(以下简称为加权传输时延的方差)或者由终端设备组中的终端设备的差分传输时延确定的方差(以下简称为差分传输时延的方差)。以终端设备组中的任意一个终端设备(记为终端设备A)为例，终端设备A的加权传输时延为终端设备A的实际传输时延与终端设备A的预期传输时延的比值，终端设备A的差分传输时延为终端设备A的实际传输时延与终端设备A的预期传输时延的差值，终端设备A的实际传输时延为终端设备A的QoS流的发送时间与终端设备A的QoS流到达服务器的时间之间的时间间隔，终端设备A的预期传输时延为终端设备A的QoS流的发送时间与预设时间之间的时间间隔。其中，终端设备A的实际传输时延可以通过终端设备A的QoS参数的测量结果获得，终端设备A的预期传输时延可以通过步骤400由第一网元发送给第二网元，也可以提前配置到第二网元或者用户面网元。
可以理解的是，这里的服务器可以与任意一个网元合设，也可以是一个独立的服务器。上述第一终端设备的实际传输时延可以为第一终端设备的PDB。
如图5A所示，终端设备组包括UE1、UE2和UE3，每个UE对应一条时间轴，从该时间轴的起点至三角形标记所对应的时长为该UE的本地计算完成时间，从三角形标记至圆形标记对应的时长为该UE的实际传输时延，从三角形标记至第一条虚线与该时间轴的交点对应的时长为该UE的预期传输时延。其中，第一条虚线位置可以理解为服务器接收到最后一个UE(即UE2)的计算结果的时间。其中，各个UE的本地计算完成时间可能不同，且难以改变，即如图5A中的三角形标记的位置各不相同，而各个UE的传输时延也可能不同，即如图5A中的圆形标记的位置各不相同，但是UE的传输时延可以通过调整UE的带宽实现。
示例性地，UE1的加权传输时延＝UE1的实际传输时延/UE1的预期传输时延，UE2的加权传输时延＝UE2的实际传输时延/UE2的预期传输时延，UE3的加权传输时延＝UE3的实际传输时延/UE3的预期传输时延；或者，UE1的加权传输时延＝UE1的预期传输时延/UE1的实际传输时延，UE2的加权传输时延＝UE2的预期传输时延/UE2的实际传输时延，UE3的加权传输时延＝UE3的预期传输时延/UE3的实际传输时延。由UE1的加权传输时延、UE2的加权传输时延以及UE3的加权传输时延确定的方差越大，则表明三者分别发送的QoS流到达服务器的时间相差较大，而方差越小，则表明三者分别发送的QoS流到达服务器的时间相差较小。
示例性地，UE1的差分传输时延＝UE1的实际传输时延-UE1的预期传输时延，UE2的差分传输时延＝UE2的实际传输时延-UE2的预期传输时延，UE3的差分传输时延＝UE3的实际传输时延-UE3的预期传输时延；或者，UE1的差分传输时延＝UE1的预期传输时延-UE1的实际传输时延，UE2的差分传输时延＝UE2的预期传输时延-UE2的实际传输时延，UE3的差分传输时延＝UE3的预期传输时延-UE3的实际传输时延。由UE1的差分传输时延、UE2的差分传输时延以及UE3的差分传输时延确定的方差越大，则表明三者分别发送的 QoS流到达服务器的时间相差较大，而方差越小，则表明三者分别发送的QoS流到达服务器的时间相差较小。
示例性地，根据横向联邦学习的原理，服务器需要在接收到各个参与方的计算结果之后才能更新模型。当各个参与方的计算结果传输到服务器的时间相差较大，即由各个参与方的加权传输时延或者差分传输时延确定的方差较大时，将会造成整体训练周期变长，降低模型训练的效率。当各个参与方的计算结果传输到服务器的时间大致相同，即由各个参与方的加权传输时延或者差分传输时延确定的方差较小时，可以降低服务器的等待时间，进而提升模型训练效率。因此，第一网元可以向第二网元订阅方差，并在方差大于设定的方差阈值时，调整至少一个参与方的传输时延，具体调整方式可以参考下述步骤420中的相关描述。
示例性地，根据横向联邦学习的原理，服务器需要在接收到各个参与方的计算结果之后才能更新模型。当各个参与方的计算结果传输到服务器的时间相差较大，即由各个参与方的加权传输时延确定的方差较大时，将会造成整体训练周期变长，降低模型训练的效率。当各个参与方的计算结果传输到服务器的时间大致相同，即由各个参与方的加权传输时延确定的方差较小时，可以降低服务器的等待时间，进而提升模型训练效率。因此，第一网元可以向第二网元订阅方差，并在方差大于设定的方差阈值时，调整至少一个参与方的传输时延，具体调整方式可以参考下述步骤420中的相关描述。
3、用于指示终端设备组的信息。示例性地，用于指示终端设备组的信息可以包括终端设备组中的各个终端设备的标识，或终端设备组对应的终端设备列表，或者终端设备组的标识等。例如，用于指示终端设备组的信息可以用分析报告的目标(Target of analytics reporting)标识。
例如，终端设备组包括UE1、UE2和UE3，则用于指示终端设备组的信息包括UE1的标识、UE2的标识和UE3的标识。
第二网元根据用于指示终端设备组的信息，可以确定收集哪些终端设备的QoS参数的测量结果，或者对哪些终端设备的QoS参数的测量结果进行聚合，进而确定聚合信息。
4、发送聚合信息的触发条件
发送聚合信息的触发条件可以包括周期性反馈聚合信息，或者满足预设条件反馈聚合信息。
例如，这里的预设条件可以包括聚合比特率大于终端设备组的签约最大比特率，和/或终端设备组中的终端设备的传输时延的统计值大于第一预设阈值等。
第二网元根据发送聚合信息的触发条件反馈聚合信息。
5、指示信息，指示信息指示反馈终端设备的QoS参数的测量结果。
可以理解的是，通过指示信息指示反馈的测量结果为终端设备级的测量结果。
6、反馈条件，反馈条件指示反馈终端设备的QoS参数的测量结果所需满足的条件。
例如，反馈条件可以指示终端设备的传输时延大于第二预设阈值的终端设备反馈QoS参数的测量结果，或者，终端设备的传输时延小于第三预设阈值的终端设备反馈QoS参数的测量结果，或者，终端设备的QoS流的比特率大于预设比特率的终端设备反馈QoS参数的测量结果。其中，第二预设阈值大于第三预设阈值。
在一些实施例中，当第一请求消息仅包括指示信息不包括反馈条件时，终端设备组的数据还包括终端设备组中的各个终端设备的QoS参数的测量结果，即第二网元向第一网元 反馈终端设备组中的全部终端设备的QoS参数的测量结果。
在一些实施例中，当第一请求消息包括指示信息和反馈条件时，终端设备组的数据还包括终端设备组中的至少一个终端设备的QoS参数的测量结果，且所述至少一个终端设备的QoS参数的测量结果满足反馈条件，即第二网元向第一网元反馈终端设备组中的部分终端设备的QoS参数的测量结果。
7、应用标识，用于指示收集终端设备的QoS参数的测量结果的应用的范围。
例如，在终端设备组中的终端设备为参与横向联邦学习的终端设备时，该应用标识为横向联邦学习的应用的标识。
第二网元指示用户面网元根据应用标识收集终端设备的QoS参数的测量结果。
步骤410：第二网元向第一网元发送终端设备组的数据。
示例性地，在第二网元为NWDAF网元时，第二网元向第一网元发送分析通知消息(Nnwdaf_AnalyticsSubscription_Notify),该消息携带终端设备组的数据。在第二网元为PCF网元时，第一请求消息可以为事件开放通知消息(Nnwdaf_EventExposure Notify),该消息携带终端设备组的数据。需要说明的是，第一网元与第二网元的每次通信均需经过NEF网元转发。
在一种可能的实现方式中，第二网元向用户面网元发送第二请求消息，第二请求消息用于请求终端设备组的数据，第二网元从用户面网元接收终端设备组的数据。
其中，第二请求消息可以包括除了与所述聚合信息关联的类型信息之外的第一请求消息所包括的内容。例如，第一请求消息包括上述1-7项内容，则第二请求消息包括上述2-7项内容。
采用上述实现方式，用户面网元完成收集终端设备组中的终端设备的QoS参数的测量结果，并根据收集到的测量结果确定终端设备组的数据，例如根据与所述聚合信息关联的参数信息确定聚合信息，根据指示信息和/或反馈条件确定需要反馈的终端设备的QoS参数的测量结果。用户面网元将确定的终端设备组的数据发送至第二网元，并由第二网元将终端设备组的数据发送至第一网元。
例如，终端设备组包括UE1、UE2和UE3，用户面网元接收第二请求消息，并根据第二请求消息收集UE1的QoS参数的测量结果，UE2的QoS参数的测量结果，以及UE3的QoS参数的测量结果。进一步地，用户面网元根据第二请求消息中的与所述聚合信息关联的参数信息，确定聚合信息包括聚合比特率和传输时延的统计值，则根据UE1对应的QoS流的比特率，UE2对应的QoS流的比特率，UE3对应的QoS流的比特率求和，作为终端设备组的聚合比特率。用户面网元还根据UE1的实际传输时延、UE2的实际传输时延、UE3的实际传输时延，和从第二网元接收的或者提前配置的UE1的预期传输时延、UE2的预期传输时延、UE3的预期传输时延计算终端设备组的传输时延的统计值。用户面网元向第二网元发送终端设备组的数据，其中，终端设备组的数据包括终端设备组的聚合比特率和终端设备组的传输时延的统计值。其中，UE1的QoS参数的测量结果包括UE1对应的QoS流的比特率和UE1的实际传输时延，UE2的QoS参数的测量结果包括UE2对应的QoS流的比特率和UE2的实际传输时延，UE3的QoS参数的测量结果包括UE3对应的QoS流的比特率和UE3的实际传输时延。
在一种可能的实现方式中，第二网元向用户面网元发送第三请求消息，第三请求消息用于请求终端设备组中的终端设备的QoS参数的测量结果，第二网元从用户面网元接收终 端设备组中的终端设备的QoS参数的测量结果，第二网元根据终端设备组中的终端设备的QoS参数的测量结果和第一请求消息确定终端设备组的数据。
采用上述实现方式，用户面网元完成收集终端设备组中的终端设备的QoS参数的测量结果，并将收集到的测量结果发送至第二网元，第二网元根据收集到的结果确定终端设备组的数据，例如根据与所述聚合信息关联的参数信息确定聚合信息，根据指示信息和/或反馈条件确定需要反馈的终端设备的QoS参数的测量结果。
例如，终端设备组包括UE1、UE2和UE3，用户面网元接收第三请求消息，并根据第三请求消息收集UE1的QoS参数的测量结果，UE2的QoS参数的测量结果，以及UE3的QoS参数的测量结果。用户面网元向第二网元发送UE1的QoS参数的测量结果，UE2的QoS参数的测量结果，以及UE3的QoS参数的测量结果。第二网元根据第一请求消息中的与所述聚合信息关联的参数信息，确定聚合信息包括聚合比特率和传输时延的统计值，则根据UE1对应的QoS流的比特率，UE2对应的QoS流的比特率，UE3对应的QoS流的比特率求和，作为终端设备组的聚合比特率。第二网元还根据UE1的实际传输时延、UE2的实际传输时延、UE3的实际传输时延，和提前配置的UE1的预期传输时延、UE2的预期传输时延、UE3的预期传输时延计算终端设备组的传输时延的统计值。第二网元向第一网元发送终端设备组的数据，其中，终端设备组的数据包括终端设备组的聚合比特率和终端设备组的传输时延的统计值。其中，UE1的QoS参数的测量结果包括UE1对应的QoS流的比特率和UE1的实际传输时延，UE2的QoS参数的测量结果包括UE2对应的QoS流的比特率和UE2的实际传输时延，UE3的QoS参数的测量结果包括UE3对应的QoS流的比特率和UE3的实际传输时延。
步骤420：第一网元根据终端设备组的数据调整终端设备组中第一终端设备的QoS参数。
可以理解的是，第一网元可以调整终端设备组中的至少一个终端设备的QoS参数，例如，第一终端设备的QoS参数，或者多个终端设备的QoS参数，或者全部终端设备的QoS参数。
在调整终端设备的QoS参数时，第一网元可以调整该QoS参数的取值，或者该QoS参数的取值范围。例如，第一网元可以调整第一终端设备的QoS流的比特率的取值，或者比特率的取值范围。
下面结合具体的终端设备组的数据说明调整终端设备组中至少一个终端设备的QoS参数的可能的实现方式，其中，至少一个终端设备包括第一终端设备。
方式1：在聚合信息包括聚合比特率时，若聚合比特率大于终端设备组的签约最大比特率，第一网元调整至少一个终端设备的QoS流的比特率，在调整至少一个终端设备的QoS流的比特率之后，终端设备组中的各个终端设备的QoS流的比特率之和小于或者等于签约最大比特率，或者，终端设备组中的各个终端设备的QoS流的比特率的取值范围的下界之和小于或者等于签约最大比特率。
在一些可能的实施例中，若第一请求消息包括发送聚合信息的触发条件，且发送聚合信息的触发条件包括终端设备组的聚合比特率大于终端设备组的签约最大比特率，则由第二网元或用户面网元根据终端设备组的签约最大比特率判断聚合比特率是否大于终端设备组的签约最大比特率，若是，则第二网元向第一网元发送终端设备组的数据，终端设备组的数据包括传输时延的统计值。
在一些可能的实施例中，若第一请求消息未包括发送聚合信息的触发条件，则由第一网元或根据终端设备组的签约最大比特率判断聚合比特率是否大于终端设备组的签约最大比特率。
在一些可能的实施例中，在第一网元调整至少一个终端设备的QoS流的比特率时，第一网元可以根据前一次为终端设备组中的各个终端设备配置的QoS流的比特率确定调整方案，即调整哪些终端设备的QoS流的比特率，以及调整的方向和大小，以使在调整至少一个终端设备的QoS流的比特率之后，终端设备组中的各个终端设备的QoS流的比特率之和小于或者等于签约最大比特率，或者，终端设备组中的各个终端设备的QoS流的比特率的取值范围的下界之和小于或者等于签约最大比特率。
在一些可能的实施例中，在第一网元调整至少一个终端设备的QoS流的比特率时，若第一请求消息包括指示信息，第二网元还向第一网元反馈终端设备组中各个终端设备的QoS参数的测量结果，第一网元可以结合终端设备组中各个终端设备的QoS参数的测量结果确定调整方案；或者，在第一网元调整至少一个终端设备的QoS流的比特率时，若第一请求消息包括指示信息和反馈条件，第二网元还向第一网元反馈终端设备组中满足反馈条件的终端设备的QoS参数的测量结果，第一网元可以结合终端设备组中满足反馈条件的终端设备的QoS参数的测量结果确定调整方案。例如，反馈条件可以为终端设备的QoS流的比特率大于预设比特率的终端设备反馈QoS参数的测量结果。
例如，终端设备组包括UE1、UE2和UE3，终端设备组的聚合比特率为UE1对应的QoS流的比特率，UE2对应的QoS流的比特率，UE3对应的QoS流的比特率之和。在终端设备组的聚合比特率大于终端设备组的签约最大比特率时，第一网元可以调整UE1、UE2和或UE3中的至少一个UE的取值，使得UE1对应的QoS流的比特率，UE2对应的QoS流的比特率，UE3对应的QoS流的比特率之和小于或等于终端设备组的签约最大比特率，或者UE1对应的QoS流的比特率的取值范围的下界，UE2对应的QoS流的比特率的取值范围的下界，UE3对应的QoS流的比特率的取值范围的下界之和小于或等于终端设备组的签约最大比特率。
方式2：在聚合信息包括终端设备组中的终端设备的传输时延的统计值时，若传输时延的统计值大于第一预设阈值，则第一网元根据终端设备组的数据调整至少一个终端设备的QoS流的比特率。
示例性地，至少一个终端设备为传输时延大于第二预设阈值或小于第三预设阈值的终端设备，其中，第二预设阈值大于第三预设阈值。
此外，与方式1类似，在调整至少一个终端设备的QoS流的比特率之后，终端设备组中的各个终端设备的QoS流的比特率之和小于或者等于签约最大比特率，或者，终端设备组中的各个终端设备的QoS流的比特率的取值范围的下界之和小于或者等于签约最大比特率。
在一些可能的实施例中，若第一请求消息包括发送聚合信息的触发条件，且发送聚合信息的触发条件包括传输时延的统计值大于第一预设阈值，则由第二网元或用户面网元根据第一预设阈值判断传输时延的统计值是否大于第一预设阈值，若是，则第二网元向第一网元发送终端设备组的数据，终端设备组的数据包括传输时延的统计值。在一些可能的实施例中，若第一请求消息未包括发送聚合信息的触发条件，则由第一网元根据第一预设阈值判断传输时延的统计值是否大于第一预设阈值。
在一些可能的实施例中，在第一网元调整至少一个终端设备的QoS流的比特率时，若第一请求消息包括指示信息，第二网元还向第一网元反馈终端设备组中各个终端设备的QoS参数的测量结果，第一网元可以结合终端设备组中各个终端设备的QoS参数的测量结果确定调整方案，即调整哪些终端设备的QoS流的比特率，以及调整的方向和大小。示例性地，对于传输时延较长的终端设备(例如传输时延大于第二预设阈值的终端设备)采用调高比特率的方式降低它的传输时延，和/或对于传输时延较短的终端设备(例如传输时延小于第三预设阈值的终端设备)采用调低比特率的方式增加它的传输时延。
或者，在第一网元调整至少一个终端设备的QoS流的比特率时，若第一请求消息包括指示信息和反馈条件，第二网元还向第一网元反馈终端设备组中满足反馈条件的终端设备的QoS参数的测量结果，第一网元可以结合终端设备组中满足反馈条件的终端设备的QoS参数的测量结果确定调整方案。例如，反馈条件为终端设备的传输时延大于第二预设阈值的终端设备反馈QoS参数的测量结果，和/或终端设备的传输时延小于第三预设阈值的终端设备反馈QoS参数的测量结果。
示例性地，如图5A所示，假设UE1的加权传输时延、UE2的加权传输时延以及UE3的加权传输时延确定的方差或者由UE1的差分传输时延、UE2的差分传输时延以及UE3的差分传输时延确定的方差大于第一预设阈值，则第一网元调低UE1的QoS流的比特率，增加UE1的实际传输时延，调高UE2的QoS流的比特率，缩短UE2的实际传输时延，调低UE3的QoS流的比特率，增加UE3的实际传输时延。如图5B所示，对于每个UE，从三角形标记至空心圆形标记对应的时长为该UE的调整前的实际传输时延，从三角形标记至实心圆形标记对应的时长为该UE的调整后的实际传输时延，从三角形标记至第一条虚线与该时间轴的交点对应的时长为该UE的调整后的预期传输时延，从三角形标记至第二条虚线与该时间轴的交点对应的时长为该UE的调整前的预期传输时延。可见，在第一网元调整UE1的比特率，UE2的比特率以及UE3的比特率之后，UE1、UE2和UE3分别对应的QoS流到达服务器的时间大致相同，比较接近。进一步地，在调整前，服务器在第二条虚线所示的时间接收到终端设备组中全部UE的计算结果，在调整后，服务器在第一条虚线所示的时间接收到终端设备组中全部UE的计算结果，第一条虚线所示的时间早于第二条虚线所示的时间，因此可以实现降低服务器的等待时间，进而提升模型训练效率。
步骤430：第一网元向策略控制网元发送调整后的第一终端设备的QoS参数。
此外，策略控制网元还会根据终端设备组的聚合信息进行相应的策略控制，以保证满足聚合信息的限制，例如聚合比特率需要小于或等于终端设备组的签约最大比特率，传输时延的统计值需要小于或等于第一预设阈值。下面以聚合比特率为例，简要介绍一下相关的策略控制方法。
示例性地，策略控制方法包括：
1、在建立会话管理(session management，SM)策略关联时，策略控制网元检查终端设备组的聚合比特率是否高于终端设备组的签约最大比特率。如果是，策略控制网元应拒绝建立SM策略关联，否则，策略控制网元正常建立SM策略关联。
2、当该终端设备组的聚合比特率小于终端设备组的签约最大比特率，同时该终端设备组的聚合比特率大于预设阈值时，策略控制网元可以通过策略控制方法限制协议数据单元(protocol data unit，PDU)会话或PCC规则的流量，并与会话管理网元进行相应的交互，其中，预设阈值小于终端设备组的签约最大比特率，且预设阈值与终端设备组的签约最大 比特率接近，例如预设阈值为终端设备组的签约最大比特率的95％；如果策略控制网元发现终端设备组的聚合比特率小于预设阈值时，策略控制网元可以放宽PDU会话或PCC规则的流量限制。
采用上述实施例，第一网元向第二网元订阅终端设备组的数据，第二网元将终端设备组的数据发送至第一网元，第一网元根据终端设备组的数据调整终端设备组中至少一个终端设备的QoS参数，进而可以实现提升模型训练效率。
此外，在第一网元向第二网元发送第一请求消息之前，第一网元还可执行如图6所示的操作流程。
步骤600：第一网元向第三网元发送终端设备组的签约最大比特率，第三网元为统一数据存储网元或统一数据管理网元。
示例性地，第一网元可以根据横向联邦学习的模型训练业务需求确定终端设备组的签约最大比特率，并向第三网元发送终端设备组的签约最大比特率。终端设备组的签约最大比特率可以由数据管理创建请求(Nudr_DM_Create Request)或数据管理更新请求(Nudr_DM_Update Request)携带。
在终端设备组的签约最大比特率由Nudr_DM_Create Request携带时，第三网元在签约信息中创建终端设备组的签约最大比特率，并向第一网元发送数据管理创建响应(Nudr_DM_Create Response)，该消息指示签约信息已创建完成。
在终端设备组的签约最大比特率由Nudr_DM_Update Request携带时，第三网元在签约信息中更新终端设备组的签约最大比特率，并向第一网元发送数据管理更新响应(Nudr_DM_Update Response)，该消息指示签约信息已更新完成。
步骤610：第一网元向策略控制功能发送第一信息，第一信息包括终端设备组包括的各个终端设备分别对应的QoS参数的参考信息。
终端设备组包括的各个终端设备分别对应的QoS参数的参考信息包括以下至少一项：
终端设备组包括的各个终端设备分别对应的QoS流的比特率的取值或取值范围，或终端设备组包括的各个终端设备分别对应的丢包率的取值或取值范围，或终端设备组包括的各个终端设备分别对应的传输时延的取值或取值范围。
例如，对于某个终端设备，QoS流的比特率＝20Mbps，丢包率＝0.01％。或者，10Mbps<QoS流的比特率<20Mbps，丢包率<0.01％。
示例性地，第一网元可以根据横向联邦学习的模型训练业务需求确定第一信息，第一信息可以由政策授权创建请求(Npcf_PolicyAuthorization_Create Request)或政策授权更新请求(Npcf_PolicyAuthorization_Update Request)携带。
步骤620：策略控制网元向第三网元请求终端设备组的签约最大比特率。
示例性地，策略控制网元通过数据管理查询请求(Nudr_DM_QueryRequest)服务操作向第三网元请求终端设备组的签约最大比特率。
步骤630：第三网元向策略控制网元发送终端设备组的签约最大比特率。
示例性地，第三网元将终端设备组的签约最大比特率通过数据管理查询响应(Nudr_DM_QueryResponse)服务操作发送给策略控制网元。
步骤640：策略控制网元根据终端设备组的签约最大比特率对第一信息进行验证。
示例性地，在第一信息包括终端设备组包括的各个终端设备分别对应的QoS流的比特 率的取值时，策略控制网元根据终端设备组的签约最大比特率和终端设备组包括的各个终端设备分别对应的QoS流的比特率的取值，确定终端设备组包括的各个终端设备分别对应的QoS流的比特率之和是否小于或等于终端设备组的签约最大比特率，若是则验证成功，否则验证失败。
或者，在第一信息包括终端设备组包括的各个终端设备分别对应的QoS流的比特率的取值范围时，策略控制网元根据终端设备组的签约最大比特率和终端设备组包括的各个终端设备分别对应的QoS流的比特率的取值范围，确定终端设备组包括的各个终端设备分别对应的QoS流的比特率的取值范围的下界之和是否小于或等于终端设备组的签约最大比特率，若是则验证成功，否则验证失败。
步骤650：在验证成功时，策略控制网元根据第一信息生成终端设备组包括的各个终端设备分别对应的策略与计费控制(policy and charging control，PCC)规则，第二网元向会话管理网元发送各个终端设备分别对应的PCC规则。
进一步地，会话管理网元可以根据策略控制网元发送的各个终端设备分别对应的PCC规则，建立或修改相应的会话。
步骤660：策略控制网元向第一网元发送第二信息，第二信息指示已为终端设备组包括的各个终端设备分别生成相应的PCC规则。
示例性地，当第一信息由Npcf_PolicyAuthorization_Create Request携带时，第二信息可以由政策授权创建响应(Npcf_PolicyAuthorization_Create Response)携带。
当第一信息由Npcf_PolicyAuthorization_Update Request携带时，第二信息可以由政策授权更新响应(Npcf_PolicyAuthorization_Update Response)携带。
此外，为了实现调整不同参与方的QoS参数，提升联邦学习的模型训练的效率。本申请提供还一种通信方法，如图7所示，其中，下述第一网元可以为应用功能网元。在图7所示的实施例中，第一网元向数据分析功能网元订阅终端设备组的数据，并提供策略控制网元的地址信息，第二网元将终端设备组的数据反馈给策略控制网元。策略控制网元根据终端设备组的数据调整终端设备组中第一终端设备的QoS参数。相较于图4所示的实施例中，第一网元负责接收终端设备组的数据，并根据终端设备组的数据调整终端设备组中第一终端设备的QoS参数，在图7所示的实施例中，策略控制网元负责接收终端设备组的数据，并根据终端设备组的数据调整终端设备组中第一终端设备的QoS参数。
具体的，该方法包括：
步骤700：第一网元向数据分析功能网元发送第一请求消息，第一请求消息用于请求终端设备组的数据，终端设备组的数据包括QoS参数的测量结果的聚合信息。
其中，步骤700可以参考上述步骤400中的相关描述，此处不再赘述。
需要说明的是，第一请求消息除了可以包括上述1-8项的内容之外，还需要包括策略控制网元的地址信息。
步骤710：数据分析功能网元根据策略控制网元的地址信息向策略控制网元发送终端设备组的数据。
其中，数据分析功能网元确定终端设备组的数据的具体实现方式可以参考上述步骤410中的相关描述。
步骤720：策略控制网元接收终端设备组的信息，策略控制网元根据终端设备组的数据调整终端设备组中第一终端设备的QoS参数。
其中，策略控制网元根据终端设备组的数据调整终端设备组中第一终端设备的QoS参数的具体实现方式，可以参考步骤420中第一网元根据终端设备组的数据调整终端设备组中第一终端设备的QoS参数的具体实现方式。
采用上述实施例，第一网元向数据分析功能网元订阅终端设备组的数据，并通知第二网元将终端设备组的数据发送至策略控制网元，由策略控制网元根据终端设备组的数据调整终端设备组中至少一个终端设备的QoS参数，进而可以实现提升联邦学习的模型训练的效率。
可以理解的是，在执行步骤700之前，第一网元还可执行如图6所示的操作流程。
本申请提供还一种通信方法，如图8所示，其中，下述第一网元可以为应用功能网元。在图8所示的实施例中，第一网元向策略控制网元订阅终端设备组的数据，并策略控制网元获取终端设备组的数据，并根据终端设备组的数据调整终端设备组中第一终端设备的QoS参数。因此，相较于图4所示的实施例中，第一网元负责接收终端设备组的数据，并根据终端设备组的数据调整终端设备组中第一终端设备的QoS参数，在图8所示的实施例中，策略控制网元负责获取终端设备组的数据，并根据终端设备组的数据调整终端设备组中第一终端设备的QoS参数。
该方法包括：
步骤800：第一网元向策略控制网元发送第一请求消息，第一请求消息用于请求终端设备组的数据，终端设备组的数据包括QoS参数的测量结果的聚合信息。
其中，步骤800可以参考上述步骤400中的相关描述，此处不再赘述。
步骤810：策略控制网元确定终端设备组的数据。
其中，策略控制网元确定终端设备组的数据的具体实现方式可以参考上述步骤410中第二网元确定终端设备组的数据的相关描述。
步骤820：策略控制网元根据终端设备组的数据调整终端设备组中第一终端设备的QoS参数。
其中，策略控制网元根据终端设备组的数据调整终端设备组中第一终端设备的QoS参数的具体实现方式，可以参考步骤420中第一网元根据终端设备组的数据调整终端设备组中第一终端设备的QoS参数的具体实现方式。
采用上述实施例，第一网元向策略控制网元订阅终端设备组的数据，策略控制网元获取终端设备组的数据，并根据终端设备组的数据调整终端设备组中至少一个终端设备的QoS参数，进而可以实现提升模型训练效率。
可以理解的是，在执行步骤800之前，第一网元还可执行如图6所示的操作流程。
本申请提供还一种通信方法。区别与上述实施例，应用功能网元也可以不通过NWDAF或者PCF收集数据，而是自身统计或获取终端设备组中各个终端设备的传输时延和本地计算时延，进而确定终端设备组中的终端设备的QoS参数的调整策略，减少了网元间交互的消息数量，减轻了网元负载，并提升了终端设备的QoS参数调整效率，使得不同终端设备的本地计算时延和传输时延之和大致相同，从而提升横向联邦学习的模型训练效率。
如图9所示，该方法，包括：
步骤900：应用功能网元获取终端设备组中各个终端设备的传输时延和本地计算时延，终端设备组中的终端设备为参与横向联邦学习的终端设备。
其中，本地计算时延为确定横向联邦学习的模型的更新参数所需的时长。传输时延为传输更新参数到应用功能网元所需的时长。可以理解的是，对于终端设备组中的不同终端设备，不同终端设备的本地计算时延可能各不相同，不同终端设备的传输时延也可能各不相同。示例性地，如图5A所示，终端设备组包括UE1、UE2和UE3，每个UE对应一条时间轴，从该时间轴的起点(即本地计算开始时间)至三角形标记所对应的时长为该UE的本地计算完成时间，即本地计算时延，从三角形标记至圆形标记对应的时长为该UE的传输时延。其中，本地计算开始时间为确定横向联邦学习的模型的更新参数的起始时间。
示例性地，应用功能网元可以根据携带更新参数的数据包的发送时间和该数据包的接收时间，确定发送该数据包的终端设备的传输时延。具体可以参考图10中的相关描述。
示例性地，应用功能网元可以在横向联邦学习开始之前，获取参与横向联邦学习的终端设备的本地计算时延，或者在横向联邦学习的模型训练阶段获取参与横向联邦学习的终端设备的本地计算时延。具体可以参考图10中的相关描述。
步骤910：应用功能网元根据各个终端设备的传输时延和本地计算时延，调整终端设备组中第一终端设备的QoS参数。
示例性地，在应用功能网元根据各个终端设备的传输时延和本地计算时延，调整终端设备组中第一终端设备的QoS参数时，应用功能网元根据各个终端设备的传输时延和本地计算时延，确定各个终端设备的总时延的统计值，其中，每个终端设备的总时延为终端设备的传输时延与本地计算时延之和。在第一终端设备的总时延大于第一预设阈值或者小于第二预设阈值的情况下，应用功能网元调整第一终端设备的QoS参数，其中，第一预设阈值与第二预设阈值是根据统计值确定的。
其中，统计值可以为算术平均值，或者几何平均值，或者均方根平均值，或者调和平均值，或者加权平均值等，本申请对此不作限定。
其中，第一预设阈值和第二预设阈值是由应用功能网元确定的。例如，第一预设阈值可以为总时延的统计值+20ms，或者总时延的统计值*1.12等。又例如，第二预设阈值可以为总时延的统计值-20ms，或者总时延的统计值*0.88等。
其中，调整后的第一终端设备的QoS参数包括资源类型、优先级、包时延预算、保证流比特率、最大流比特率、分配与抢占优先级中的至少一个。
示例性地，资源类型可以包括GBR,Non-GBR，时延关键型(Delay-critical)GBR。通过将第一终端设备的QoS flow的类型从Non-GBR调整为GBR或时延关键型GBR，可以保证网络为第一终端设备的QoS flow预留足够的资源以保证其带宽，进而提升设定的传输时延得到保障的概率。
示例性地，优先级可以是指资源调度的优先级，即5QI中的优先级水平(priority level)参数。分配与抢占优先级是指资源分配和抢占的优先级。通过调整第一终端设备的QoS flow的优先级或分配与抢占优先级可以实现在拥塞场景下保证网络会优先处理优先级或分配与抢占优先级高的QoS Flow，因此在网络资源有限的情况下，调高QoS Flow的优先级或分配与抢占优先级会提升设定的传输时延得到保障的概率。
示例性地，通过调低包时延预算，可以降低第一终端设备的QoS flow的传输时延，通过调高包时延预算，可以增加第一终端设备的QoS flow的传输时延。
示例性地，通过调整保证流比特率和最大流比特率可以保证QoS Flow比特率的稳定性，进而保证时延的稳定性。
此外，应用功能网元还可以调整错包率、平均窗口、最大数据突发量等参数。
步骤920：应用功能网元向策略控制网元发送调整后的第一终端设备的QoS参数。
此外，在一种可能的实现方式中，应用功能网元还根据各个终端设备的传输时延和本地计算时延，调整终端设备组中第一终端设备的QoS参考。其中，QoS参考不指定具体的QoS参数，由策略控制网元映射为具体的QoS参数。也就是说，策略控制网元可以根据第一终端设备的QoS参考调整终端设备组中第一终端设备的QoS参数。此时，应用功能网元向策略控制网元发送调整后的第一终端设备的QoS参考。
例如，应用功能网元向策略控制网元发送调整后的第一终端设备的期望传输时延，则策略控制网元根据第一终端设备的期望传输时延，映射为具体的QoS参数，例如，资源类型、优先级、包时延预算、保证流比特率、最大流比特率、分配与抢占优先级中的至少一个。
以下结合图10对图9所示实施例进行说明：
在本实施例中，在横向联邦学习的模型训练阶段之前，应用功能网元可以根据参与横向联邦学习的终端设备的本地计算时延，为每个终端设备设定不同的期望传输时延，使得不同终端设备的本地计算时延和期望传输时延之和大致相同，用以提升横向联邦学习的模型训练效率。在横向联邦学习的模型训练阶段，应用功能网元可以获取不同终端设备的本地计算时延和实际传输时延之和，即总时延。进而，应用功能网元可以计算不同终端设备的总时延的统计值，即期望总时延，并筛选出总时延与期望总时延相差较大的终端设备，通过调整这些终端设备的QoS参数，使得这些终端设备的总时延能够与期望总时延大致相同。
S1001.AF根据终端设备组中每个UE的本地计算时延和初始总时延，得到每个UE的期望传输时延，并据此设置每个UE的针对联邦学习业务的QoS Flow的QoS参考值或者QoS参数值。
其中，AF可以在测试阶段获取参与横向联邦学习的UE的本地计算时延和初始总时延。其中，测试阶段是指在AF确定参与横向联邦学习的UE，并与参与横向联邦学习的UE建立了会话连接，且在AF与UE执行横向联邦学习的模型训练之前的时间段。此外，测试阶段又可称为准备阶段。示例性地，在测试阶段，AF与UE通过应用层数据包交互，可以获取UE的本地计算时延和初始总时延，进而获得参与横向联邦学习的每个UE的本地计算时延和初始总时延。
示例性地，为了实现不同UE的本地计算时延和传输时延之和大致相同，AF可以设置一个截止时间，用以实现UE的中间结果(即更新参数)均在该截止时间之前到达AF。进而，AF可以根据该截止时间和每个UE的本地计算时延，计算得到不同UE的期望传输时延，并根据每个UE的期望传输时延设置该UE的QoS Flow的QoS参考值或QoS参数值，例如，PDB。
例如，终端设备组包括UE1、UE2和UE3，其中，UE1的本地计算时延为20ms，UE1的总时延为90ms，UE2的本地计算时延为40ms,UE2的总时延为70ms，UE3的本地计算时延为30ms，UE3的总时延为80ms。AF根据上述3个总时延设置截止时间，如AF计算上述3个总时延的平均值为80ms，则设置截止时间为80ms，即从本地计算开始时间开始至收到全部UE的数据包，所需时长为80ms。则UE1的期望传输时延为60ms，UE2的期望传输时延为40ms，UE3的期望传输时延为50ms。AF根据UE1的期望传输时延设置UE1 的QoS Flow的PDB。AF根据UE2的期望传输时延设置UE2的QoS Flow的PDB，AF根据UE3的期望传输时延设置UE3的QoS Flow的PDB。例如，UE2的QoS Flow的PDB＜UE3的QoS Flow的PDB＜UE1的QoS Flow的PDB。
S1002.AF向PCF发送请求消息，该请求消息包括QoS参考值或者QoS参数值。
示例性地，AF通过PCF服务化接口策略授权创建请求(Npcf_PolicyAuthorization_Creat Request)向PCF发送终端设备组中每个UE的QoS参考值或者QoS参数值，或者AF通过PCF服务化接口策略授权更新请求(Npcf_PolicyAuthorization_Update Request)向PCF发送终端设备组中每个UE的QoS参数的QoS参考值或者QoS参数值。
其中，该请求消息包括UE标识(UE ID)，应用会话标识(App session ID)，QoS参考或QoS参数(QoS reference or QoS parameters)，其中，UE标识用于指示被调整QoS参数的UE。应用会话标识(App session ID)，用于指示被调整的应用的会话的QoS参数。QoS参考或QoS参数指示被调整的QoS参数，例如QoS参考为终端设备传输时延，QoS参数为PDB。
S1003.PCF根据请求消息设置相应的QoS参数，并生成PCC规则。
S1004.PCF向SMF发送PCC规则，用以SMF将PCC规则绑定到相应的QoS Flow上。
S1005.PCF向AF发送响应消息。响应消息指示QoS参数设置成功。
示例性地，在AF通过PCF服务化接口策略授权创建请求向PCF发送终端设备组中每个UE的QoS参考值或者QoS参数值时，PCF通过PCF服务化接口策略授权创建响应(Npcf_PolicyAuthorization_Create Response)向AF通知QoS参数设置成功。或者，在AF通过PCF服务化接口策略授权更新请求向PCF发送终端设备组中每个UE的QoS参数的QoS参考值或者QoS参数值时，PCF通过PCF服务化接口策略授权更新响应(Npcf_PolicyAuthorization_Update Response)向AF通知QoS参数设置成功。
此外，在一些实施例中，PCF也可以向AF通知QoS参数设置失败，例如网络资源不足，无法满足AF指定的QoS参数等。
在S1005之后，当终端设备的横向联邦学习的QoS Flow的QoS参数设置完成后，AF与终端设备组中各个UE执行横向联邦学习的模型训练。
S1006.AF获取终端设备组中每个UE的传输时延和本地计算时延。
在一种可能的设计中，在UE向AF发送数据包时，会将数据包的发送时间作为时间戳携带在数据包中，AF接收到数据包后，就可以根据数据包的接收时间和所述时间戳计算得到数据包的传输时延。其中，数据包携带横向联邦学习的模型的更新参数。示例性地，AF获取的每个UE的传输时延，可以是指某一训练轮次的全部数据包的传输时延的平均值。例如，UE在某一训练轮次共发送10个数据包，AF根据上述方法确定每个数据包的传输时延，进而确定这个10个数据包的传输时延的平均值，作为该UE的传输时延。
在一种可能的设计中，AF可以在测试阶段获取UE的本地计算时延。
在一种可能的设计中，在UE向AF发送数据包时，会将数据包的发送时间作为时间戳携带在数据包中，AF接收到数据包后，AF可以根据该时间戳和本地计算开始时间，确定该UE的本地计算时延。其中，数据包携带横向联邦学习的模型的更新参数。示例性地，AF可以获取的每个UE的本地计算时延，可以是指某一训练轮次的全部数据包的本地计算时延的平均值。可以理解的是，参与横向联邦学习的每个UE的本地计算开始时间大致相同。
S1007.AF根据每个UE的传输时延和本地计算时延，调整终端设备组中UE1的QoS参考值或者QoS参数值。
AF计算每个UE的传输时延和本地计算时延之和，作为该UE的总时延，进而根据各个UE的总时延，确定各个UE总时延的统计值，即期望总时延。各个UE总时延的统计值可以是各个UE的总时延的算术平均值，几何平均值，均方根平均值，调和平均值，加权平均值。
进一步，在UE1的总时延大于第一预设阈值或小于第二预设阈值的情况下，AF调整UE1的QoS参考值或者QoS参数值。
示例性地，假设统计值为算术平均值，AF确定各个UE总时延的算术平均值为50ms，AF设定第一预设阈值为算术平均值+10ms，即60ms，设定第二预设阈值为算术平均值-10ms，即40ms，则AF根据筛选出总时延小于40ms的UE和总时延大于60ms的UE，并调整这些UE的QoS参考值或者QoS参数值。
例如，对于总时延小于40ms的UE，AF可以调高该UE对应的PDB，和/或调低该UE的QoS Flow的ARP。
又例如，对于总时延大于60ms的UE，AF可以调低该UE对应的PDB，和/或调高该UE的QoS Flow的ARP。其中，通过调高该UE的QoS Flow的ARP，可以提升达到设定的期望传输时延得到保障的概率。此外，AF可以设定QoS Flow的GBR和/或MBR等QoS参数，来保证QoS Flow比特率的稳定性，进而保证时延的稳定性。GBR也可以为GFBR，MBR也可以为MFBR。
S1008.AF向PCF发送请求消息，该请求消息用于请求调整UE1的QoS参数。
示例性地，AF通过PCF服务化接口策略授权更新请求(Npcf_PolicyAuthorization_Update Request)向PCF发送UE1的调整后的QoS参考值或者QoS参数值。
其中，该请求消息包括UE1ID，App session ID，调整后的QoS参考值或者QoS参数值。
S1009.PCF根据请求消息设置相应的QoS参数，并生成PCC规则。
此外，PCF根据请求消息更新相应的QoS参数。此外，在一些可能的实施例中，PCF可能将QoS Flow的类型由Non-GBR变为GBR或Delay-critical GBR，并生成更新后的PCC规则。例如，只有GBR和Delay-critical GBR类型的QoS Flow才有GBR和/或MBR参数，所以如果AF指定了GBR和/或MBR的值，或者AF设定了流比特率的参考值，则意味着如果之前该QoS Flow的类型是Non-GBR，则PCF会将该QoS Flow的类型变为GBR或Delay-critical GBR。
S1010.PCF向SMF发送更新后的PCC规则，用以SMF将更新后的PCC规则绑定到相应的QoS Flow上。
S1011.PCF向AF发送响应消息。响应消息指示QoS参数设置成功。
示例性地，PCF通过PCF服务化接口策略授权更新响应(Npcf_PolicyAuthorizationUpdate Response)服务操作向AF通知QoS参数设置成功。
采用上述方法，在确定终端设备的QoS参数调整策略时，应用功能网元不用依赖NWDAF或者PCF网元从UPF获取终端设备的数据，可以减少其他网元(例如NWDAF，PCF，UPF)的负担，提升了终端设备的QoS参数调整效率。
图11示出了本申请实施例中所涉及的一种通信装置的可能的示例性框图，该装置1100包括：收发模块1120和处理模块1110，收发模块1120可以包括接收单元和发送单元。处理模块1110用于对装置1100的动作进行控制管理。收发模块1120用于支持装置1100与其他网络实体的通信。可选地，装置1100还可以包括存储单元，所述存储单元用于存储装置1100的程序代码和数据。
可选地，所述装置1100中各个模块可以是通过软件来实现。
可选地，处理模块1110可以是处理器或控制器，例如可以是通用中央处理器(central processing unit，CPU)，通用处理器，数字信号处理(digital signal processing，DSP)，专用集成电路(application specific integrated circuits，ASIC)，现场可编程门阵列(field programmable gate array，FPGA)或者其他可编程逻辑器件、晶体管逻辑器件、硬件部件或者其任意组合。其可以实现或执行结合本申请实施例公开内容所描述的各种示例性的逻辑方框，模块和电路。所述处理器也可以是实现计算功能的组合，例如包含一个或多个微处理器组合，DSP和微处理器的组合等等。收发模块1120可以是通信接口、收发器或收发电路等，其中，该通信接口是统称，在具体实现中，该通信接口可以包括多个接口，存储单元可以是存储器。
当装置1100为第一网元或第一网元中的芯片时，装置1100中的处理模块1110可以支持装置1100执行上文中各方法示例中第一网元的动作，例如可以支持装置1100执行图4中的步骤400或步骤420。
收发模块1120可以支持装置1100与第二网元进行通信，例如，收发模块1120可以支持装置1100执行图4中的步骤400、步骤410、步骤430。
例如，所述收发模块1120，用于向第二网元发送第一请求消息，所述第一请求消息用于请求终端设备组的数据，其中，所述终端设备组中的终端设备为参与横向联邦学习的终端设备；从所述第二网元接收所述终端设备组的数据，所述终端设备组的数据包括QoS参数的测量结果的聚合信息；所述处理模块1110，用于根据所述终端设备组的数据调整所述终端设备组中第一终端设备的QoS参数；所述收发模块1120，用于向策略控制网元发送调整后的所述第一终端设备的QoS参数。
在一种可能的设计中，所述第一请求消息包括指示信息，所述指示信息用于指示反馈终端设备的QoS参数的测量结果；所述终端设备组的数据还包括所述终端设备组中的第二终端设备的QoS参数的测量结果，所述第二终端设备包括所述第一终端设备。
在一种可能的设计中，所述第一请求消息还包括反馈条件，所述反馈条件用于指示反馈所述终端设备的QoS参数的测量结果所需满足的条件；所述第二终端设备的QoS参数的测量结果满足所述反馈条件。
在一种可能的设计中，所述聚合信息包括聚合比特率，所述聚合比特率用于表征所述终端设备组中的终端设备的QoS流的比特率之和；所述处理模块1110，用于在根据所述终端设备组的数据调整所述终端设备组中第一终端设备的QoS参数时，若所述聚合比特率大于所述终端设备组的签约最大比特率，根据所述终端设备组的数据调整所述第一终端设备的QoS流的比特率。
在一种可能的设计中，所述聚合信息包括所述终端设备组中的终端设备的传输时延的统计值；所述处理模块1110，用于在根据所述终端设备组的数据调整所述终端设备组中第一终端设备的QoS参数时，若所述传输时延的统计值大于第一预设阈值，则根据所述终端 设备组的数据调整所述第一终端设备的QoS流的比特率。
在一种可能的设计中，所述第一终端设备为传输时延大于第二预设阈值的终端设备。
应理解，根据本申请实施例的装置1100可对应于前述方法实施例中第一网元，并且装置1100中的各个模块的操作和/或功能分别为了实现前述方法实施例中第一网元的方法的相应步骤，因此也可以实现前述方法实施例中的有益效果，为了简洁，这里不作赘述。
当装置1100为第二网元或第二网元中的芯片时，装置1100中的处理模块1110可以支持装置1100执行上文中各方法示例中第二网元的动作。
收发模块1120可以支持装置1100与第一网元进行通信，例如，收发模块1120可以支持装置1100执行图4中的步骤400、步骤410、步骤430。
例如，所述处理模块1110调用所述收发模块1120执行：从第一网元接收第一请求消息，所述第一请求消息用于请求终端设备组的数据，其中，所述终端设备组中的终端设备为参与横向联邦学习的终端设备；发送所述终端设备组的数据，所述终端设备组的数据包括QoS参数的测量结果的聚合信息。
在一种可能的设计中，所述第一请求消息包括指示信息，所述指示信息用于指示反馈终端设备的QoS参数的测量结果；所述终端设备组的数据还包括所述终端设备组中的第二终端设备的QoS参数的测量结果，所述第二终端设备包括所述第一终端设备。
在一种可能的设计中，所述第一请求消息还包括反馈条件，所述反馈条件用于指示反馈所述终端设备的QoS参数的测量结果所需满足的条件；所述第二终端设备的QoS参数的测量结果满足所述反馈条件。
在一种可能的设计中，所述聚合信息包括聚合比特率，所述聚合比特率用于表征所述终端设备组中的终端设备的QoS流的比特率之和。
在一种可能的设计中，所述聚合信息包括所述终端设备组中的终端设备的传输时延的统计值。
在一种可能的设计中，所述收发模块1120，还用于向用户面网元发送第二请求消息，所述第二请求消息用于请求所述终端设备组的数据；从用户面网元接收所述终端设备组的数据。
在一种可能的设计中，所述收发模块1120，还用于向用户面网元发送第三请求消息，所述第三请求消息用于请求所述终端设备组中的终端设备的QoS参数的测量结果；从所述用户面网元接收所述终端设备组中的终端设备的QoS参数的测量结果；所述处理模块1110，还用于根据所述终端设备组中的终端设备的QoS参数的测量结果和所述第一请求消息确定所述终端设备组的数据。
应理解，根据本申请实施例的装置1100可对应于前述方法实施例中第二网元，并且装置1100中的各个模块的操作和/或功能分别为了实现前述方法实施例中第二网元的方法的相应步骤，因此也可以实现前述方法实施例中的有益效果，为了简洁，这里不作赘述。
当装置1100为应用功能网元或应用功能网元中的芯片时，装置1100中的处理模块1110可以支持装置1100执行上文中各方法示例中应用功能网元的动作。例如，处理模块1110可以支持装置1100执行图9中的步骤900和步骤910。
收发模块1120可以支持装置1100与策略控制网元进行通信，例如，收发模块1120可以支持装置1100执行图9中的步骤920。
例如，所述处理模块1110，用于获取终端设备组中各个终端设备的传输时延和本地计 算时延，所述终端设备组中的终端设备为参与横向联邦学习的终端设备，根据所述各个终端设备的传输时延和本地计算时延，调整所述终端设备组中第一终端设备的QoS参数；所述收发模块1120，用于向策略控制网元发送调整后的所述第一终端设备的QoS参数。
在一种可能的设计中，在根据所述各个终端设备的传输时延和本地计算时延，调整所述终端设备组中第一终端设备的QoS参数时，所述处理模块1110，用于根据所述各个终端设备的传输时延和本地计算时延，确定所述各个终端设备的总时延的统计值，其中，每个终端设备的总时延为所述终端设备的传输时延与本地计算时延之和；在所述第一终端设备的总时延大于第一预设阈值或者小于第二预设阈值的情况下，调整所述第一终端设备的QoS参数，其中，所述第一预设阈值与所述第二预设阈值是根据所述统计值确定的。
在一种可能的设计中，所述调整后的所述第一终端设备的QoS参数包括资源类型、优先级、包时延预算、保证流比特率、最大流比特率、分配与抢占优先级中的至少一个。
在一种可能的设计中，所述本地计算时延为确定所述横向联邦学习的模型的更新参数所需的时长；所述传输时延为传输所述更新参数到所述应用功能网元所需的时长。
应理解，根据本申请实施例的装置1100可对应于前述方法实施例中应用功能网元，并且装置1100中的各个模块的操作和/或功能分别为了实现前述方法实施例中应用功能网元的方法的相应步骤，因此也可以实现前述方法实施例中的有益效果，为了简洁，这里不作赘述。
图12示出了根据本申请实施例的通信装置1200的示意性结构图。如图12所示，所述装置1200包括：处理器1201。
当装置1200为第一网元或第一网元中的芯片时，一种可能的实现方式中，当所述处理器1201用于调用接口执行以下动作：向第二网元发送第一请求消息，所述第一请求消息用于请求终端设备组的数据，其中，所述终端设备组中的终端设备为参与横向联邦学习的终端设备；从所述第二网元接收所述终端设备组的数据，所述终端设备组的数据包括QoS参数的测量结果的聚合信息；根据所述终端设备组的数据调整所述终端设备组中第一终端设备的QoS参数；向策略控制网元发送调整后的所述第一终端设备的QoS参数。
应理解，所述装置1200还可用于执行前文实施例中第一网元侧的其他步骤和/或操作，为了简洁，这里不作赘述。
当装置1200为第二网元或第二网元中的芯片时，一种可能的实现方式中，当所述处理器1201用于调用接口执行以下动作：
从第一网元接收第一请求消息，所述第一请求消息用于请求终端设备组的数据，其中，所述终端设备组中的终端设备为参与横向联邦学习的终端设备；发送所述终端设备组的数据，所述终端设备组的数据包括QoS参数的测量结果的聚合信息。
应理解，所述装置1200还可用于执行前文实施例中第二网元侧的其他步骤和/或操作，为了简洁，这里不作赘述。
当装置1200为应用功能网元或应用功能网元中的芯片时，一种可能的实现方式中，当所述处理器1201用于调用接口执行以下动作：
获取终端设备组中各个终端设备的传输时延和本地计算时延，所述终端设备组中的终端设备为参与横向联邦学习的终端设备，根据所述各个终端设备的传输时延和本地计算时延，调整所述终端设备组中第一终端设备的QoS参数；向策略控制网元发送调整后的所述第一终端设备的QoS参数。
应理解，所述装置1200还可用于执行前文实施例中应用功能网元侧的其他步骤和/或操作，为了简洁，这里不作赘述。
应理解，所述处理器1201可以调用接口执行上述收发动作，其中，调用的接口可以是逻辑接口或物理接口，对此不作限定。可选地，物理接口可以通过收发器实现。可选地，所述装置1200还包括收发器1203。
可选地，所述装置1200还包括存储器1202，存储器1202中可以存储上述方法实施例中的程序代码，以便于处理器1201调用。
具体地，若所述装置1200包括处理器1201、存储器1202和收发器1203，则处理器1201、存储器1202和收发器1203之间通过内部连接通路互相通信，传递控制和/或数据信号。在一个可能的设计中，处理器1201、存储器1202和收发器1203可以通过芯片实现，处理器1201、存储器1202和收发器1203可以是在同一个芯片中实现，也可能分别在不同的芯片实现，或者其中任意两个功能组合在一个芯片中实现。该存储器1202可以存储程序代码，处理器1201调用存储器1202存储的程序代码，以实现装置1200的相应功能。
上述本申请实施例揭示的方法可以应用于处理器中，或者由处理器实现。处理器可能是一种集成电路芯片，具有信号的处理能力。在实现过程中，上述方法实施例的各步骤可以通过处理器中的硬件的集成逻辑电路或者软件形式的指令完成。上述的处理器可以是通用处理器、数字信号处理器(digital signal processor，DSP)、专用集成电路(application specific integrated circuit，ASIC)、现成可编程门阵列(field programmable gate array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件，还可以是系统芯片(system on chip，SoC)，还可以是中央处理器(central processor unit，CPU)，还可以是网络处理器(network processor，NP)，还可以是数字信号处理电路(digital signal processor，DSP)，还可以是微控制器(micro controller unit，MCU)，还可以是可编程控制器(programmable logic device，PLD)或其他集成芯片。可以实现或者执行本申请实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。结合本申请实施例所公开的方法的步骤可以直接体现为硬件译码处理器执行完成，或者用译码处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存储器，闪存、只读存储器，可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。该存储介质位于存储器，处理器读取存储器中的信息，结合其硬件完成上述方法的步骤。
可以理解，本申请实施例中的存储器可以是易失性存储器或非易失性存储器，或可包括易失性和非易失性存储器两者。其中，非易失性存储器可以是只读存储器(read-only memory，ROM)、可编程只读存储器(programmable ROM，PROM)、可擦除可编程只读存储器(erasable PROM，EPROM)、电可擦除可编程只读存储器(electrically EPROM，EEPROM)或闪存。易失性存储器可以是随机存取存储器(random access memory，RAM)，其用作外部高速缓存。通过示例性但不是限制性说明，许多形式的RAM可用，例如静态随机存取存储器(static RAM，SRAM)、动态随机存取存储器(dynamic RAM，DRAM)、同步动态随机存取存储器(synchronous DRAM，SDRAM)、双倍数据速率同步动态随机存取存储器(double data rate SDRAM，DDR SDRAM)、增强型同步动态随机存取存储器(enhanced SDRAM，ESDRAM)、同步连接动态随机存取存储器(synchlink DRAM， SLDRAM)和直接内存总线随机存取存储器(direct rambus RAM，DR RAM)。应注意，本文描述的系统和方法的存储器旨在包括但不限于这些和任意其它适合类型的存储器。
应理解，在本申请实施例中，编号“第一”、“第二”…仅仅为了区分不同的对象，比如为了区分不同的参数信息或者消息，并不对本申请实施例的范围构成限制，本申请实施例并不限于此。
还应理解，在本申请的各种实施例中，上述各过程的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定。上述各个过程涉及的各种数字编号或序号仅为描述方便进行的区分，而不应对本申请实施例的实施过程构成任何限定。
还应理解，本文中术语“和/或”，仅仅是一种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。另外，本文中字符“/”，一般表示前后关联对象是一种“或”的关系。
本申请中出现的类似于“项目包括如下中的一项或多项：A，B，以及C”表述的含义，如无特别说明，通常是指该项目可以为如下中任一个：A；B；C；A和B；A和C；B和C；A，B和C；A和A；A，A和A；A，A和B；A，A和C，A，B和B；A，C和C；B和B，B，B和B，B，B和C，C和C；C，C和C，以及其他A，B和C的组合。以上是以A，B和C共3个元素进行举例来说明该项目的可选用条目，当表达为“项目包括如下中至少一种：A，B，……，以及X”时，即表达中具有更多元素时，那么该项目可以适用的条目也可以按照前述规则获得。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统、装置和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
在本申请所提供的几个实施例中，应该理解到，所揭露的系统、装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述单元的划分，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，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请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。
所述功能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，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请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)执行本申请各个实施例所述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ROM、随机存取存储器RAM、磁碟或者光盘等各种可以存储程序代码的介质。

Claims (28)

  1. 一种通信方法，其特征在于，该方法包括：
    第一网元向第二网元发送第一请求消息，所述第一请求消息用于请求终端设备组的数据，其中，所述终端设备组中的终端设备为参与横向联邦学习的终端设备；
    所述第一网元从所述第二网元接收所述终端设备组的数据，所述终端设备组的数据包括服务质量QoS参数的测量结果的聚合信息；
    所述第一网元根据所述终端设备组的数据调整所述终端设备组中第一终端设备的QoS参数；
    所述第一网元向策略控制网元发送调整后的所述第一终端设备的QoS参数。
  2. 如权利要求1所述的方法，其特征在于，所述第一请求消息包括指示信息，所述指示信息用于指示反馈终端设备的QoS参数的测量结果；
    所述终端设备组的数据还包括所述终端设备组中的第二终端设备的QoS参数的测量结果，所述第二终端设备包括所述第一终端设备。
  3. 如权利要求2所述的方法，其特征在于，所述第一请求消息还包括反馈条件，所述反馈条件用于指示反馈所述终端设备的QoS参数的测量结果所需满足的条件；
    所述第二终端设备的QoS参数的测量结果满足所述反馈条件。
  4. 如权利要求1-3任一项所述的方法，其特征在于，所述聚合信息包括聚合比特率，所述聚合比特率用于表征所述终端设备组中的终端设备的QoS流的比特率之和；
    所述第一网元根据所述终端设备组的数据调整所述终端设备组中第一终端设备的QoS参数，包括：
    在所述聚合比特率大于所述终端设备组的签约最大比特率时，所述第一网元根据所述终端设备组的数据调整所述第一终端设备的QoS流的比特率。
  5. 如权利要求1-4任一项所述的方法，其特征在于，所述聚合信息包括所述终端设备组中的终端设备的传输时延的统计值；
    所述第一网元根据所述终端设备组的数据调整所述终端设备组中第一终端设备的QoS参数，包括：
    若所述传输时延的统计值大于第一预设阈值，则所述第一网元根据所述终端设备组的数据调整所述第一终端设备的QoS流的比特率。
  6. 如权利要求5所述的方法，其特征在于，所述第一终端设备为传输时延大于第二预设阈值的终端设备。
  7. 如权利要求6所述的方法，其特征在于，所述第一请求消息还包括所述第一预设阈值和/或所述第二预设阈值。
  8. 如权利要求1-7任一项所述的方法，其特征在于，所述第一请求消息包括所述终端设备组中终端设备的标识、所述终端设备组的标识、分析类型的标识信息、所述聚合信息的类型中的至少一种。
  9. 如权利要求8所述的方法，其特征在于，所述第一请求消息还包括以下至少一项：
    发送所述聚合信息的触发条件；
    应用标识，用于指示终端设备的QoS参数的测量结果对应的应用。
  10. 如权利要求1-9任一项所述的方法，其特征在于，所述第一网元为应用功能网元， 所述第二网元为数据分析功能网元或策略控制网元。
  11. 一种通信方法，其特征在于，该方法包括：
    第二网元从第一网元接收第一请求消息，所述第一请求消息用于请求终端设备组的数据，其中，所述终端设备组中的终端设备为参与横向联邦学习的终端设备；
    所述第二网元发送所述终端设备组的数据，所述终端设备组的数据包括QoS参数的测量结果的聚合信息。
  12. 如权利要求11所述的方法，其特征在于，所述第一请求消息包括指示信息，所述指示信息用于指示反馈终端设备的QoS参数的测量结果；
    所述终端设备组的数据还包括所述终端设备组中的第二终端设备的QoS参数的测量结果，所述第二终端设备包括所述第一终端设备。
  13. 如权利要求12所述的方法，其特征在于，所述第一请求消息还包括反馈条件，所述反馈条件用于指示反馈所述终端设备的QoS参数的测量结果所需满足的条件；
    所述第二终端设备的QoS参数的测量结果满足所述反馈条件。
  14. 如权利要求11-13任一项所述的方法，其特征在于，所述聚合信息包括聚合比特率，所述聚合比特率用于表征所述终端设备组中的终端设备的QoS流的比特率之和。
  15. 如权利要求11-14任一项所述的方法，其特征在于，所述聚合信息包括所述终端设备组中的终端设备的传输时延的统计值。
  16. 如权利要求11-15任一项所述的方法，其特征在于，所述第一请求消息包括所述终端设备组中终端设备的标识、所述终端设备组的标识、分析类型的标识信息、所述聚合信息的类型中的至少一种。
  17. 如权利要求16所述的方法，其特征在于，所述第一请求消息还包括以下至少一项：
    发送所述聚合信息的触发条件；
    应用标识，用于指示终端设备的QoS参数的测量结果对应的应用。
  18. 如权利要求11-17任一项所述的方法，其特征在于，还包括：
    所述第二网元向用户面网元发送第二请求消息，所述第二请求消息用于请求所述终端设备组的数据；
    所述第二网元从用户面网元接收所述终端设备组的数据。
  19. 如权利要求11-17任一项所述的方法，其特征在于，还包括：
    所述第二网元向用户面网元发送第三请求消息，所述第三请求消息用于请求所述终端设备组中的终端设备的QoS参数的测量结果；
    所述第二网元从所述用户面网元接收所述终端设备组中的终端设备的QoS参数的测量结果；
    所述第二网元根据所述终端设备组中的终端设备的QoS参数的测量结果和所述第一请求消息确定所述终端设备组的数据。
  20. 如权利要求11-19任一项所述的方法，其特征在于，所述第一网元为应用功能网元，所述第二网元为数据分析功能网元或策略控制网元。
  21. 一种通信方法，其特征在于，该方法包括：
    应用功能网元获取终端设备组中各个终端设备的传输时延和本地计算时延，所述终端设备组中的终端设备为参与横向联邦学习的终端设备；
    所述应用功能网元根据所述各个终端设备的传输时延和本地计算时延，调整所述终端 设备组中第一终端设备的QoS参数；
    所述应用功能网元向策略控制网元发送调整后的所述第一终端设备的QoS参数。
  22. 如权利要求21所述的方法，其特征在于，所述应用功能网元根据所述各个终端设备的传输时延和本地计算时延，调整所述终端设备组中第一终端设备的QoS参数，包括：
    所述应用功能网元根据所述各个终端设备的传输时延和本地计算时延，确定所述各个终端设备的总时延的统计值，其中，每个终端设备的总时延为所述终端设备的传输时延与本地计算时延之和；
    在所述第一终端设备的总时延大于第一预设阈值或者小于第二预设阈值的情况下，所述应用功能网元调整所述第一终端设备的QoS参数，其中，所述第一预设阈值与所述第二预设阈值是根据所述统计值确定的。
  23. 如权利要求22所述的方法，其特征在于，所述调整后的所述第一终端设备的QoS参数包括资源类型、优先级、包时延预算、保证流比特率、最大流比特率、分配与抢占优先级中的至少一个。
  24. 如权利要求21-23任一项所述的方法，其特征在于，所述本地计算时延为确定所述横向联邦学习的模型的更新参数所需的时长；
    所述传输时延为传输所述更新参数到所述应用功能网元所需的时长。
  25. 一种通信装置，其特征在于，包括：
    存储器，用于存储计算机指令；
    处理器，用于执行所述计算机指令，以使得所述通信装置执行如权利要求1至10以及11至20以及21至24中任一项所述的方法。
  26. 一种通信装置，其特征在于，包括处理器和接口电路，所述接口电路用于接收来自所述通信装置之外的其它通信装置的信号并传输至所述处理器或将来自所述处理器的信号发送给所述通信装置之外的其它通信装置，所述处理器通过逻辑电路或执行代码指令用于实现如权利要求1至10以及11至20以及21至24中任一项所述的方法。
  27. 一种计算机可读存储介质，其特征在于，所述存储介质中存储有计算机程序或指令，当所述计算机程序或指令被通信装置执行时，实现如权利要求1至10以及11至20以及21至24中任一项所述的方法。
  28. 一种通信系统，其特征在于，该系统包括第一网元和第二网元；
    所述第一网元，用于执行如权利要求1至10中任一项所述的方法；以及
    所述第二网元，用于执行如权利要求11至20中任一项所述的方法。
PCT/CN2023/072639 2022-01-28 2023-01-17 一种通信方法及装置 WO2023143255A1 (zh)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
CN202210108437 2022-01-28
CN202210108437.1 2022-01-28
CN202210326157.8A CN116567608A (zh) 2022-01-28 2022-03-29 一种通信方法及装置
CN202210326157.8 2022-03-29

Publications (1)

Publication Number Publication Date
WO2023143255A1 true WO2023143255A1 (zh) 2023-08-03

Family

ID=87470684

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/072639 WO2023143255A1 (zh) 2022-01-28 2023-01-17 一种通信方法及装置

Country Status (1)

Country Link
WO (1) WO2023143255A1 (zh)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111294224A (zh) * 2018-12-10 2020-06-16 华为技术有限公司 用于测量服务质量信息的方法和装置
WO2021233195A1 (zh) * 2020-05-22 2021-11-25 华为技术有限公司 一种通信方法及装置
CN113938911A (zh) * 2020-07-13 2022-01-14 华为技术有限公司 一种通信方法、设备及系统

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111294224A (zh) * 2018-12-10 2020-06-16 华为技术有限公司 用于测量服务质量信息的方法和装置
WO2020119564A1 (zh) * 2018-12-10 2020-06-18 华为技术有限公司 用于测量服务质量信息的方法和装置
WO2021233195A1 (zh) * 2020-05-22 2021-11-25 华为技术有限公司 一种通信方法及装置
CN113938911A (zh) * 2020-07-13 2022-01-14 华为技术有限公司 一种通信方法、设备及系统

Similar Documents

Publication Publication Date Title
CN109314710B (zh) 用于通信网络中的服务质量监测、策略执行和计费的系统和方法
US11956157B2 (en) Activation of PDU session and QOS flows in 3GPP-based ethernet bridges
US20210337404A1 (en) Service continuity implementation method, apparatus, and system
CN110099370B (zh) 服务层南向接口和服务质量
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
RU2571377C2 (ru) Система и способ управления службами, развитый nodeb и шлюз сети пакетной передачи данных
WO2022001640A1 (zh) 数据传输的方法和装置
US20230019215A1 (en) TSC-5G QoS MAPPING WITH CONSIDERATION OF ASSISTANCE TRAFFIC INFORMATION AND PCC RULES FOR TSC TRAFFIC MAPPING AND 5G QoS FLOWS BINDING
WO2021238331A1 (zh) 获取网络资源的方法、设备及系统
WO2014101505A1 (zh) 一种WLAN和分组核心网之间的QoS对齐的方法
CN111436030A (zh) 数据用量上报的方法、装置及系统
CN113079527A (zh) 用于控制无线通信系统中的网络切片数据速率的设备和方法
EP3962157A1 (en) Mdbv determining method, apparatus, and system
WO2022012361A1 (zh) 一种通信方法及装置
WO2022047803A1 (zh) 通信方法及装置
WO2023207970A1 (zh) 通信方法及通信装置
WO2023143255A1 (zh) 一种通信方法及装置
US11647419B2 (en) Adjusting window size based on quality of experience
WO2017028681A1 (zh) 一种数据传输状态的报告、确定传输数据量的方法及装置
CN115734282A (zh) 一种服务质量的处理方法、装置和通信系统
CN116567608A (zh) 一种通信方法及装置
WO2024027633A1 (zh) 一种业务流属性配置方法、装置和系统
WO2023169011A1 (zh) 一种通信方法、装置及设备
WO2024022158A1 (zh) 通信方法及装置
WO2023000305A1 (zh) Qos调整方法、装置、设备及介质

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: 23746137

Country of ref document: EP

Kind code of ref document: A1