Artificial intelligence (AI) model distribution method, apparatus, storage medium, and program product

Abstract

The present disclosure relates to an artificial intelligence (AI) model distribution method, an apparatus, a storage medium, and a program product. The method comprises: a first network function entity receives a first request from a terminal, the first request being used for requesting an AI operation to be performed on first service data; the first network function entity determines, on the basis of the first request, a second network function entity that stores a target AI model, the target AI model being used for the AI operation; and the first network function entity acquires the target AI model from the second network function entity, and sends the target AI model to at least one of a third network function entity and the terminal; or, the first network function entity instructs at least one of the third network function entity and the terminal to acquire the target AI model from the second network function entity.

Description

Methods, devices, storage media, and program products for distributing artificial intelligence (AI) models Cross-references

[0001] This application incorporates Chinese Patent Application No. 2025103679277, filed on March 26, 2025, entitled “Method, Apparatus, Storage Medium and Program Product for Distributing Artificial Intelligence (AI) Models”, which is incorporated herein by reference in its entirety. Technical Field

[0002] This disclosure relates to the field of communication technology, and in particular to a method, apparatus, storage medium, and program product for distributing artificial intelligence (AI) models. Background Technology

[0003] In 6G technology, in addition to the existing control plane and user plane, the 6G network introduces intelligent plane, computing plane, and data plane. At the same time, the artificial intelligence (AI) model has also been transformed from the external AI of the 5G network to the endogenous AI, and the various network elements of the 6G network have endogenous AI capabilities.

[0004] In related technologies, AI models are usually deployed on the control plane. When the user plane needs to call the functions of the AI ​​model, it usually needs to send the relevant business data to the control plane as input parameters, and then call the relevant AI model on the control plane to process the business data, obtain the processing results, and then feed the processing results back to the user plane. This greatly increases the overhead of data transmission and the related latency. Summary of the Invention

[0005] This disclosure provides a method, apparatus, storage medium, and program product for distributing artificial intelligence (AI) models.

[0006] In a first aspect, this disclosure provides a method for distributing artificial intelligence (AI) models, applied to a first network functional entity, including:

[0007] The first network function entity receives a first request from the terminal, the first request being used to request the execution of AI operations on the first service data;

[0008] The first network functional entity determines a second network functional entity that stores the target AI model based on the first request; the target AI model is used for the AI ​​operation.

[0009] The first network function entity obtains the target AI model from the second network function entity and sends the target AI model to at least one of the third network function entity and the terminal; or, the first network function entity instructs at least one of the third network function entity and the terminal to obtain the target AI model from the second network function entity.

[0010] In some embodiments, the AI ​​operation is a user plane management-related AI operation for the first service data, and the third network function entity is a user plane function entity, a user plane management function entity, or a session management function entity; or...

[0011] The AI ​​operation is an AI operation related to data plane management of the first business data, and the third network function entity is a data plane function entity, a data plane management function entity, or a task management function entity.

[0012] In some embodiments, the user plane function entity includes at least one of the following: N3 interface user plane function entity, intermediate user plane function entity, and protocol data unit session anchor point function entity.

[0013] In some embodiments, the first request includes at least one of business type information of the first business data and information of the target AI model, wherein both the business type information and the target AI model information are used to obtain the target AI model.

[0014] In some embodiments, the method further includes:

[0015] The first network functional entity selects a fourth network functional entity, which is used to manage the AI ​​services corresponding to the target AI model.

[0016] The first network function entity adds at least one of the service type information and the target AI model information from the first request to the first session context establishment request;

[0017] The first network function entity sends the first session context establishment request to the fourth network function entity;

[0018] The first network function entity receives the target AI model or the address information of the target AI model sent by the fourth network function entity.

[0019] In some embodiments, the first network functional entity determines a second network functional entity storing the target AI model based on the first request, including:

[0020] The first network function entity determines the second network function entity that stores the target AI model based on at least one of the service type information and the target AI model information.

[0021] In some embodiments, the first network function entity instructs at least one of the third network function entity and the terminal to obtain the target AI model from the second network function entity, including:

[0022] The first network function entity sends model acquisition instruction information to at least one of the third network function entity and the terminal, the model acquisition instruction information including the address information of the target AI model.

[0023] In some embodiments, the method further includes:

[0024] The first network functional entity selects a third network functional entity to distribute the target AI model.

[0025] In some embodiments, where the training mechanism of the target AI model is distributed learning or federated learning, the method further includes:

[0026] The first network functional entity sends the convergence instruction information of the target AI model to the third network functional entity;

[0027] The convergence indication information is used to indicate the central node entity in the third network functional entity, and the central node entity is the convergence node of the model training information of the target AI model on the third network functional entity.

[0028] In some embodiments, the model training information of the target AI model includes at least one of the model parameters and gradient parameters of the target AI model.

[0029] In some embodiments, the method further includes:

[0030] The first network functional entity receives the updated AI model sent by the central node entity;

[0031] The first network functional entity sends the updated AI model to the non-central node entity in the third network functional entity.

[0032] Secondly, this disclosure provides a method for distributing artificial intelligence (AI) models, applied to third network functional entities, including:

[0033] The third network functional entity receives the target AI model sent by the first network functional entity; or...

[0034] The third network function entity receives model acquisition instruction information sent by the first network function entity. The model acquisition instruction information is used to instruct at least one of the third network function entity and the terminal to acquire the target AI model from the second network function entity. The target AI model is used to perform AI operations on the first business data.

[0035] In some embodiments, the AI ​​operation is a user plane management-related AI operation for the first service data, and the third network function entity is a user plane function entity, a user plane management function entity, or a session management function entity; or...

[0036] The AI ​​operation is an AI operation related to data plane management of the first business data, and the third network function entity is a data plane function entity, a data plane management function entity, or a task management function entity.

[0037] In some embodiments, the user plane function entity includes at least one of the following: N3 interface user plane function entity, intermediate user plane function entity, and protocol data unit session anchor point function entity.

[0038] In some embodiments, the model acquisition indication information includes the address information of the target AI model.

[0039] In some embodiments, where the training mechanism of the target AI model is distributed learning or federated learning, the method further includes:

[0040] The third network functional entity receives the convergence indication information of the target AI model sent by the first network functional entity;

[0041] The convergence indication information is used to indicate the central node entity in the third network functional entity, and the central node entity is the convergence node of the model training information of the target AI model on the third network functional entity.

[0042] In some embodiments, the method further includes:

[0043] The third network functional entity trains the target AI model to obtain the model training information;

[0044] The non-central node entity in the third network functional entity sends the model training information to the central node;

[0045] The central node entity aggregates the model training information to obtain the updated AI model.

[0046] In some embodiments, the method further includes:

[0047] The central node entity sends the updated AI model to the non-central node entity; or...

[0048] The central node entity sends the updated AI model to the first network function entity.

[0049] In some embodiments, the method further includes:

[0050] The third network function entity sends the target AI model to the terminal, and the target AI model is used to optimize the first service data on the terminal.

[0051] In some embodiments, the method further includes:

[0052] The third network function entity receives the first service data from the terminal.

[0053] The third network function entity uses the target AI model to optimize the first business data.

[0054] In some embodiments, the target AI model information includes at least one of the following: the identification information of the AI ​​model, the storage location information of the AI ​​model, and the model size information of the AI ​​model.

[0055] Thirdly, this disclosure provides a method for distributing artificial intelligence (AI) models, applied to a terminal, including:

[0056] The terminal sends a first request to the first network function entity, the first request being used to request the execution of AI operations on the first service data;

[0057] The terminal receives a target AI model sent by the first network function entity; or, the terminal receives model acquisition instruction information from the first network function, the model acquisition instruction information being used to instruct at least one of the third network function entity and the terminal to acquire the target AI model from the second network function entity, the target AI model being used for the AI ​​operation.

[0058] In some embodiments, the first request carries at least one of the business type information of the first business data and the information of the target AI model, wherein both the business type information and the information of the target AI model are used to obtain the target AI model.

[0059] In some embodiments, the AI ​​operation is a user plane management-related AI operation for the first service data, and the third network function entity is a user plane function entity, a user plane management function entity, or a session management function entity; or...

[0060] The AI ​​operation is an AI operation related to data plane management of the first business data, and the third network function entity is a data plane function entity, a data plane management function entity, or a task management function entity.

[0061] In some embodiments, the user plane function entity includes at least one of the following: N3 interface user plane function entity, intermediate user plane function entity, and protocol data unit session anchor point function entity.

[0062] In some embodiments, the method further includes:

[0063] The terminal sends the terminal device's service data to the third network function entity.

[0064] Fourthly, this disclosure provides a method for distributing artificial intelligence (AI) models, applied to access network equipment, including:

[0065] The access network device receives a first request sent by the terminal, the first request being used to request the execution of an AI operation on the first service data, the AI ​​operation being executed through a target AI model;

[0066] The access network device sends the first request to the first network function entity.

[0067] In some embodiments, the method further includes:

[0068] The access network device selects the first network function entity according to the type of the access network device.

[0069] In some embodiments, the first request includes at least one of business type information of the first business data and information of the target AI model, wherein both the business type information and the target AI model information are used to obtain the target AI model.

[0070] In some embodiments, the method further includes:

[0071] The access network device receives the first service data sent by the terminal;

[0072] The access network device sends the first service data to a third network function entity, which includes the target AI model.

[0073] Fifthly, this disclosure provides an artificial intelligence (AI) model distribution device, applied to a first network functional entity, including a memory, a transceiver, and a processor:

[0074] A memory for storing computer programs; a transceiver for sending and receiving data under the control of the processor; and a processor for reading the computer programs from the memory and performing the following operations:

[0075] The transceiver receives a first request from the terminal, the first request being used to request the execution of an AI operation on the first business data;

[0076] Based on the first request, a second network functional entity is determined to store the target AI model, the target AI model being used for the AI ​​operation;

[0077] The target AI model is obtained from the second network functional entity and sent to at least one of the third network functional entity and the terminal; or, at least one of the third network functional entity and the terminal is instructed to obtain the target AI model from the second network functional entity.

[0078] Sixthly, this disclosure provides an artificial intelligence (AI) model distribution device, applied to a third network functional entity, including a memory, a transceiver, and a processor:

[0079] A memory for storing computer programs; a transceiver for sending and receiving data under the control of the processor; and a processor for reading the computer programs from the memory and performing the following operations:

[0080] The transceiver receives the target AI model sent by the first network function entity; or...

[0081] The transceiver receives model acquisition instruction information sent by the first network function entity. The model acquisition instruction information is used to instruct at least one of the third network function entity and the terminal to acquire the target AI model from the second network function entity. The target AI model is used to perform AI operations on the first service data.

[0082] Seventhly, this disclosure provides an artificial intelligence (AI) model distribution device for a terminal, including a memory, a transceiver, and a processor:

[0083] A memory for storing computer programs; a transceiver for sending and receiving data under the control of the processor; and a processor for reading the computer programs from the memory and performing the following operations:

[0084] The transceiver sends a first request to the first network function entity, the first request being used to request the execution of an AI operation on the first service data;

[0085] The transceiver receives the target AI model sent by the first network function entity; or, the transceiver receives model acquisition instruction information from the first network function, the model acquisition instruction information being used to instruct at least one of the third network function entity and the terminal to acquire the target AI model from the second network function entity, the target AI model being used for the AI ​​operation.

[0086] Eighthly, this disclosure provides an artificial intelligence (AI) model distribution device for use in access network equipment, including a memory, a transceiver, and a processor:

[0087] A memory for storing computer programs; a transceiver for sending and receiving data under the control of the processor; and a processor for reading the computer programs from the memory and performing the following operations:

[0088] The transceiver receives a first request sent by the terminal, the first request being used to request the execution of an AI operation on the first business data, the AI ​​operation being executed through a target AI model;

[0089] The first request is sent to the first network function entity via the transceiver.

[0090] Ninthly, this disclosure provides an artificial intelligence (AI) model distribution device, applied to a first network functional entity, comprising:

[0091] The first receiving module is used to receive a first request from the terminal, the first request being used to request the execution of AI operations on the first business data;

[0092] A first processing module is configured to, according to the first request, determine a second network functional entity storing a target AI model, the target AI model being used for the AI ​​operation; obtain the target AI model from the second network functional entity and send the target AI model to at least one of a third network functional entity and the terminal; or, instruct at least one of the third network functional entity and the terminal to obtain the target AI model from the second network functional entity.

[0093] In a tenth aspect, this disclosure provides an artificial intelligence (AI) model distribution device, applied to a third network functional entity, comprising:

[0094] The second receiving module is used to receive the target AI model sent by the first network functional entity; or...

[0095] The second receiving module is used to receive model acquisition instruction information sent by the first network function entity. The model acquisition instruction information is used to instruct at least one of the third network function entity and the terminal to acquire the target AI model from the second network function entity. The target AI model is used to perform AI operations on the first business data.

[0096] Eleventhly, this disclosure provides an artificial intelligence (AI) model distribution device for use in a terminal, comprising:

[0097] The third sending module is used to send a first request to the first network function entity, the first request being used to request the execution of AI operations on the first business data;

[0098] The third receiving module is used to receive the target AI model sent by the first network function entity; or, to receive model acquisition instruction information from the first network function, wherein the model acquisition instruction information is used to instruct at least one of the third network function entity and the terminal to acquire the target AI model from the second network function entity, and the target AI model is used for the AI ​​operation.

[0099] In a twelfth aspect, this disclosure provides an artificial intelligence (AI) model distribution device for use in access network equipment, comprising:

[0100] The fourth receiving module is used to receive a first request sent by the terminal, the first request being used to request the execution of an AI operation on the first business data, the AI ​​operation being executed through a target AI model;

[0101] The fourth sending module is used to send the first request to the first network function entity.

[0102] In a thirteenth aspect, this disclosure also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the distribution method for the artificial intelligence (AI) model described in the first, second, third, or fourth aspects above.

[0103] In a fourteenth aspect, this disclosure also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the distribution method for the artificial intelligence (AI) model described in the first, second, third, or fourth aspects above.

[0104] The above description is merely an overview of the technical solution of this disclosure. In order to better understand the technical means of this disclosure and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this disclosure more apparent and understandable, specific embodiments of this disclosure are described below. Attached Figure Description

[0105] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the embodiments described below. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this disclosure. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0106] Figure 1 is a schematic diagram of the logical architecture of 6G multidimensional converged network in related technologies;

[0107] Figure 2 is an application scenario diagram of an AI model distribution method provided in an embodiment of this disclosure;

[0108] Figure 3 is a flowchart illustrating an AI model distribution method provided in this embodiment of the present disclosure;

[0109] Figure 4 is a flowchart illustrating another AI model distribution method provided in an embodiment of this disclosure;

[0110] Figure 5 is a flowchart illustrating another AI model distribution method provided in an embodiment of this disclosure;

[0111] Figure 6 is one of the signaling interaction diagrams of an AI model distribution method provided in an embodiment of this disclosure;

[0112] Figure 7 is a second signaling interaction diagram of an AI model distribution method provided in an embodiment of this disclosure;

[0113] Figure 8 is a third signaling interaction diagram of an AI model distribution method provided in an embodiment of this disclosure;

[0114] Figure 9 is a signaling interaction diagram of an AI model distribution method provided in an embodiment of this disclosure;

[0115] Figure 10 is the fifth signaling interaction diagram of an AI model distribution method provided in an embodiment of this disclosure;

[0116] Figure 11 is a signaling interaction diagram of an AI model distribution method provided in an embodiment of this disclosure;

[0117] Figure 12 is the seventh signaling interaction diagram of an AI model distribution method provided in an embodiment of this disclosure;

[0118] Figure 13 is the eighth signaling interaction diagram of an AI model distribution method provided in an embodiment of this disclosure;

[0119] Figure 14 is one of the structural block diagrams of an artificial intelligence (AI) model distribution device provided in an embodiment of this disclosure;

[0120] Figure 15 is a second structural block diagram of an artificial intelligence (AI) model distribution device provided in an embodiment of this disclosure;

[0121] Figure 16 is a third structural block diagram of an artificial intelligence (AI) model distribution device provided in an embodiment of this disclosure;

[0122] Figure 17 is a fourth structural block diagram of an artificial intelligence (AI) model distribution device provided in an embodiment of this disclosure;

[0123] Figure 18 is a fifth structural block diagram of an artificial intelligence (AI) model distribution device provided in an embodiment of this disclosure;

[0124] Figure 19 is a structural block diagram of an artificial intelligence (AI) model distribution device provided in an embodiment of this disclosure. Detailed Implementation

[0125] The embodiments of the technical solutions disclosed herein will now be described in detail with reference to the accompanying drawings. These embodiments are merely illustrative of the technical solutions disclosed herein and are therefore intended to limit the scope of protection of this disclosure.

[0126] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of this disclosure; the terms “comprising” and “having”, and any variations thereof, in the specification, claims and foregoing description of the drawings of this disclosure are intended to cover non-exclusive inclusion.

[0127] In the description of the embodiments of this disclosure, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary or secondary relationship of the indicated technical features. In the description of the embodiments of this disclosure, "a plurality of" means two or more, unless otherwise explicitly defined.

[0128] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this disclosure. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0129] In the description of the embodiments of this disclosure, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0130] In the description of the embodiments of this disclosure, the term "multiple" refers to two or more (including two), similarly, "multiple groups" refers to two or more (including two groups), and "multiple pieces" refers to two or more (including two pieces).

[0131] In the description of the embodiments of this disclosure, the technical terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the embodiments of this disclosure and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of this disclosure.

[0132] In the description of the embodiments of this disclosure, unless otherwise expressly specified and limited, technical terms such as "installation," "connection," "joining," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in the embodiments of this disclosure according to the specific circumstances.

[0133] The artificial intelligence (AI) model distribution method, apparatus, storage medium, and program product provided in this disclosure can reduce data transmission overhead and processing latency.

[0134] The method and apparatus are based on the same concept of the application. Since the methods and apparatus solve problems in similar ways, the implementation of the apparatus and methods can refer to each other, and the repeated parts will not be described again.

[0135] The relevant technologies will be explained below.

[0136] In 6G technology, in addition to the existing control plane and user plane, the 6G network introduces intelligent plane, computing plane, and data plane. At the same time, the artificial intelligence (AI) model has also been transformed from the external AI of the 5G network to the endogenous AI, and the various network elements of the 6G network have endogenous AI capabilities.

[0137] The following section will first explain the AI ​​capabilities of 6G networks.

[0138] The AI ​​capabilities of 6G networks typically refer to AI services offered to the network itself, users, or externally. This is achieved through the intrinsic construction of AI elements such as computing power, algorithms, and data. Under the orchestration, management, and control of integrated AI and connectivity elements, it enables on-demand customization of AI services and guarantees of high Quality of Service (QoS). Native AI capabilities in 6G networks will be a core capability, including functions such as training, inference, and model validation. The network side will provide lifecycle management functions for AI models and data, such as AI data, AI models, and AI computing power-related capabilities.

[0139] Among them, AI data-related AI capabilities can include training data reporting, gradient reporting, etc.; AI model-related AI capabilities can include model generation, model storage, model selection, model distribution, model update, model replacement, model activation and deactivation, model performance monitoring, etc.; AI computing power-related AI capabilities can include the allocation of computing power resources such as training execution and inference execution.

[0140] Furthermore, the inherent AI capabilities in 6G networks can further enhance the intelligent communication experience, transforming "symbol-level" communication services focused on bit-level information transmission into "meaningful" communication services oriented towards high-level semantic delivery, thus breaking through the traditional "Shannon limit." By combining AI task orchestration and distributed computing power scheduling across the edge, cloud, and device layers, multimodal and barrier-free communication among natural persons, digital persons, and intelligent and non-intelligent devices can ultimately be achieved.

[0141] In related technologies, the AI ​​capabilities of 6G networks can be realized through AI models. An AI model is a system trained using computer algorithms and data, capable of simulating human intelligent behavior and performing a range of tasks, including image recognition, speech recognition, natural language processing, and machine translation. An AI model typically consists of an architecture, parameters, and training methods. The model architecture defines the data processing flow, the parameters are the objects of the AI ​​model's learning and optimization, and the training methods guide how the model learns and improves.

[0142] In the training process of an AI model, data preprocessing can begin, including cleaning and formatting the raw data to remove noise and outliers, ensuring data quality and accuracy. Subsequently, appropriate model structures and algorithms can be selected based on the characteristics and requirements of the task, and parameters can be adjusted and optimized to achieve better performance. Finally, the AI ​​model is trained using the training set. Through repeated iterations and learning, the weights and features of the AI ​​model are optimized to improve its performance.

[0143] It should be understood that AI models can be applied to multiple fields. For example, they can be used in the medical field for disease diagnosis, drug development, and gene analysis; in the financial field for credit assessment, risk prediction, and investment advice; in the transportation field for intelligent driving and traffic flow prediction; and in the education field for personalized teaching and learning assistance. The application of AI models in various fields can not only improve efficiency and accuracy but also create new value.

[0144] The logical architecture of 6G networks will be explained below.

[0145] Figure 1 illustrates the logical architecture of a 6G multi-dimensional converged network in related technologies. As shown in Figure 1, horizontally, the 6G network can be divided into a resource layer, a network function layer, and an application and service layer. The resource layer provides underlying resources such as wireless, computing, and storage, and provides corresponding support and services for the functional generation of the network function layer. The network function layer forms specific network functions, or combines one or more network functions, providing the most basic network service capabilities to meet the needs of the application and service layer. The application and service layer provides corresponding support for user services and applications, enabling service customization.

[0146] Referring back to Figure 1, vertically, the 6G network includes a communication plane carrying traditional communication services, as well as newly added data, computing, and intelligence planes. The data plane is responsible for the collection, cleaning, processing, and storage of data in the end-to-end network, and provides data services to other layers and planes. The computing plane provides a unified computing power repository, senses computing power needs, orchestrates computing tasks, provides computing power routing, computing power modeling, and state awareness, and provides computing services to other layers and planes. The intelligence plane provides the complete operating environment required for the entire lifecycle of endogenous AI, calls upon the services provided by the data and computing planes, and provides intelligent services to other layers and planes. The management plane manages all other layers and planes.

[0147] It should be understood that the network function layer may include control plane functions and user plane functions (UPF).

[0148] For example, control plane functions may include a Network Data Analytics Function (NWDAF). The NWDAF collects, analyzes, and processes data from the network and makes corresponding control decisions based on the analysis and processing results. The NWDAF can collect, analyze, and process network data according to NWDAF control rules. By formulating reasonable NWDAF control rules, it can be ensured that the NWDAF can quickly and accurately analyze and judge the network status based on real-time data from the network and take corresponding control measures to optimize network performance.

[0149] It should be understood that NWDAF control rules can include various conditional judgments and control strategies, which can be customized and adjusted according to different network needs and scenarios. For example, firstly, NWDAF control rules can include data acquisition rules, which specify how NWDAF collects data from the network. Data acquisition rules can include the frequency of data acquisition, the data type collected, and the data source. Secondly, NWDAF control rules can also include data analysis rules, which specify how NWDAF analyzes and processes the collected data. Data analysis rules can include data analysis algorithms, analysis indicators, and thresholds. Finally, NWDAF control rules can include decision control rules, which specify how NWDAF makes corresponding decisions and controls based on the data analysis results. Decision control rules can include decision conditions, decision strategies, and decision actions.

[0150] For example, the UPF is used for routing and forwarding of user plane data packets in the core network, playing a crucial role in low-latency, high-bandwidth edge computing and network slicing technologies. The UPF can support routing and forwarding of terminal service data, data and service identification, and action and policy execution. The UPF can also interact with the Session Management Function (SMF) through the N4 interface, directly controlled and managed by the SMF, and execute service flow processing according to various policies issued by the SMF.

[0151] The main functions of UPF can include the following four items:

[0152] In the first functional item, the UPF can be the interconnection point between the wireless access network and the data network (DN), used for the encapsulation and decapsulation of the user plane GTP tunneling protocol (GPRS Tunneling Protocol for User Plane, GTP-U).

[0153] In the second function item, UPF can be a Protocol Data Unit Session Anchor (PSA) used to provide mobility during radio access.

[0154] In the third function, the UPF can route and distribute packets locally, acting as an uplink classifier (UL-CL) or a branching point UPF as an intermediate UPF (I-UPF).

[0155] In the fourth functional category, UPF can include application monitoring, data stream QoS processing, traffic usage reporting, Internet Protocol (IP) management, mobility adaptation, policy control, and billing. In addition to network functional requirements, UPF can also consider metrics such as data security, physical environment requirements, and deployment power consumption.

[0156] In addition, the UPF, as the connection point between mobile and data networks, has important interfaces including N3, N4, N6, N9, and N19. Among them, the N3 interface is the interface between the Next Generation Radio Access Network (NG RAN) and the UPF, using the GTP-U protocol for user data tunneling. The N4 interface is the interface between the SMF and the UPF. The control plane is used to transmit node messages and session messages, using the Packet Forwarding Control Protocol (PFCP). The user plane is used to transmit messages that the SMF needs to receive or send through the UPF, using the GTP-U protocol. The N6 interface is the interface between the UPF and the external DN. In specific scenarios, the N6 interface supports leased lines or L2 / L3 layer tunnels, enabling communication between the IP and DN networks. The N9 interface is the interface between UPFs. In mobile scenarios, an I-UPF is inserted between the terminal and the PSA UPF for traffic forwarding, and the two UPFs use the GTP-U protocol for user plane packet transmission. The N19 interface is the user plane interface between two PSA UPFs when using Local Area Network (LAN) services. It directly routes traffic between different PDU sessions without using the N6 interface.

[0157] The following describes the session establishment process for a PDU.

[0158] It should be understood that a PDU session is similar to a Packet Data Network (PDN) connection, used to enable data interaction with an external Data Network Name (DNN). In the PDU session establishment request message sent to the network, the terminal may provide Single Network Slice Selection Assistance Information (S-NSSAI), DNN information, etc.

[0159] The aforementioned S-NSSAI includes the S-NSSAI of the home network and the S-NSSAI of the serving network. During the terminal's session establishment process, the slice selection is based on the Network Slice Selection Policy (NSSP), which determines the S-NSSAI. If the terminal cannot determine the S-NSSAI during PDU session establishment, it can omit the S-NSSAI information in the PDU Session Establishment Request message.

[0160] It should be noted that the S-NSSAI of the home network cannot be changed during the life of the PDU session, while the S-NSSAI of the serving network can be changed during the life of the PDU session.

[0161] The Data Network Name (DNN) is used to distinguish external data networks, similar to the Access Point Name (APN). For a user's specific S-NSSAI, the subscription data in the UDM may contain a list of subscribed DNNs and a default DNN. Therefore, when a terminal establishes a PDU session, if the request message does not carry a DNN, the AMF will establish the PDU session for the user based on the subscribed default DNN. If no subscribed default DNN is available, the AMF will select a DNN for the terminal based on its local configuration.

[0162] In related technologies, AI models are typically deployed on the control plane (e.g., NWDAF). When the user plane needs to call the functions of the AI ​​model, it usually needs to send the relevant business data to the control plane as input parameters, and then call the relevant AI model on the control plane to process the business data, obtain the processing results, and then feed the processing results back to the user plane. This greatly increases the overhead of data transmission and the related latency.

[0163] To address the aforementioned issues, this disclosure provides a method, apparatus, storage medium, and program product for distributing artificial intelligence (AI) models. A first network functional entity determines a second network functional entity that stores the target AI model. The first network functional entity then sends the target AI model from the second network functional entity to at least one of a third network functional entity and a terminal, or instructs at least one of the third network functional entity and the terminal to obtain the target AI model from the second network functional entity. This enables flexible deployment of the target AI model, thereby reducing the transmission overhead and processing latency of related first business data.

[0164] The following describes the application scenarios of the AI ​​model distribution method provided in this disclosure.

[0165] Figure 2 illustrates an application scenario of an AI model distribution method provided in this embodiment. As shown in Figure 2, terminal 101 sends a first request to first network function entity 103 via access network device 102. The first request requests the execution of an AI operation on first service data. Subsequently, based on the first request, first network function entity 103 determines a second network function entity 104 that stores the target AI model, which is used for the AI ​​operation. Finally, first network function entity 103 obtains the target AI model from second network function entity 104 and sends it to at least one of third network function entity 105 and terminal 101; or, first network function entity 103 instructs at least one of third network function entity 105 and terminal 101 to obtain the target AI model from second network function entity 104.

[0166] The terminal disclosed in this embodiment can be a device that provides at least one of voice and data connectivity to a user, a handheld device with wireless connectivity, or other processing devices connected to a wireless modem. The name of the terminal may differ in different systems; for example, in a 5G system, the terminal may be called User Equipment (UE). The wireless terminal can be a USB storage device, other personal computer memory devices, or a dongle. It can also communicate with one or more core networks (CNs) via a Radio Access Network (RAN). The wireless terminal can be a mobile terminal, such as a mobile phone (or "cellular" phone) and a computer with a mobile terminal, for example, a portable, pocket-sized, handheld, computer-embedded, or vehicle-mounted mobile device that exchanges at least one of voice and data with the radio access network. Examples of such devices include Personal Communication Service (PCS) phones, cordless phones, Session Initiated Protocol (SIP) phones, Wireless Local Loop (WLL) stations, Personal Digital Assistants (PDAs), personal computers, tablets, and Machine-type Communication (MTC) terminals. Wireless terminals can also be referred to as systems, subscriber units, subscriber stations, mobile stations, mobile devices, remote stations, access points, remote terminals, access terminals, user terminals, user agents, user devices, and wireless access devices and routers / modems that meet the limitations of this definition, but are not limited to these in the embodiments of this disclosure.

[0167] The first, second, and third network function entities involved in the embodiments of this disclosure can be network elements in core network equipment. The access network equipment involved in the embodiments of this disclosure can be a base station, which may include multiple cells providing services to terminal devices. Depending on the specific application, a base station may also be called an access point, or a device in the access network that communicates with wireless terminal devices through one or more sectors on the air interface, or other names. The access network equipment can be used to exchange received air frames with Internet Protocol (IP) packets, acting as a router between the wireless terminal device and the rest of the access network, where the rest of the access network may include an Internet Protocol (IP) communication network. The access network equipment can also coordinate the attribute management of the air interface. For example, the access network equipment involved in this disclosure can be an evolved Node B (eNB or e-NodeB) in a long term evolution (LTE) system, a 5G base station (gNB) in a next generation system, or a Home evolved Node B (HeNB), relay node, femto, pico, network testing equipment, etc., and is not limited in this disclosure. In some network structures, network equipment may include centralized unit (CU) nodes and distributed unit (DU) nodes, and the centralized unit and distributed unit may be geographically separated.

[0168] It should be understood that the technical solutions provided in this disclosure are applicable to a variety of communication systems. For example, applicable systems may include Long Term Evolution (LTE) systems, LTE Frequency Division Duplex (FDD) systems, LTE Time Division Duplex (TDD) systems, Long Term Evolution Advanced (LTE-A) systems, Universal Mobile Telecommunications System (UMTS), Worldwide Interoperability for Microwave Access (WiMAX) systems, 5G New Radio (NR) systems, 6G systems, and their evolved communication systems. These systems may include terminal equipment and network equipment. The systems may also include a core network component, such as Evolved Packet System (EPS) or 5G system (5GS).

[0169] The technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this disclosure.

[0170] In one embodiment, as shown in Figure 3, a flowchart illustrating a method for distributing AI models is provided to demonstrate how AI models are distributed. This AI model distribution method is applied to a first network functional entity and includes steps S201-S203:

[0171] S201, The first network function entity receives the first request from the terminal.

[0172] In this disclosure, when a terminal needs to perform AI operations on the first service data, it can send a first request to the first network function entity.

[0173] The aforementioned first network function entity may include a Session Management Function (SMF) entity.

[0174] It should be understood that the embodiments of this disclosure do not limit how the first network function entity receives the first request from the terminal. In some embodiments, the terminal may first send the first request to the access network device, and then the access network device may select the first network function entity according to the type of the access network device and send the first request to the first network function entity.

[0175] The types of access network equipment include service-oriented and non-service-oriented equipment.

[0176] For example, in the case of service-oriented access network equipment, after receiving the first request, the access network equipment can directly select the first network function entity and send the first request to the first network function entity by signaling interaction with the first network function entity.

[0177] For example, in the case of non-service access network equipment, after receiving the first request, the access network equipment can first select the Access and Mobility Management Function (AMF) entity, then the AMF entity selects the first network function entity, and then the access network equipment sends the first request to the first network function entity.

[0178] In some embodiments, the first request is used to request the execution of an AI operation on the first business data. For example, the first request may be a request to establish an AI service corresponding to the first business data.

[0179] The aforementioned first service data can be service data of any service type corresponding to the terminal. For example, the first service data may include service data of service types such as traffic identification, QoS optimization, and latency control.

[0180] It should be understood that the embodiments of this disclosure do not limit the AI ​​operation requested to be performed in the first request. In some embodiments, the AI ​​operation is a user-plane management-related AI operation for the first business data. For example, the AI ​​operation may include business recognition, intelligent interaction, behavior prediction, etc. In other embodiments, the AI ​​operation is a data-plane management-related AI operation for the first business data. For example, the AI ​​operation may include data integration, data cleaning, data analysis, etc.

[0181] In some embodiments, the first request includes at least one of business type information and target AI model information, both of which are used to obtain the target AI model. The target AI model is used for AI operations.

[0182] The information of the target AI model may include at least one of the following: the identifier of the target AI model, the storage location of the target AI model, and the model size of the target AI model.

[0183] It should be noted that the embodiments disclosed herein do not limit the type of the target AI model. In some embodiments, the target AI model may include machine learning models, deep learning models, knowledge graph models, reinforcement learning models, etc.

[0184] S202. The first network functional entity determines the second network functional entity that stores the target AI model according to the first request.

[0185] In this step, after the first network functional entity receives the first request from the terminal, it can determine the second network functional entity that stores the target AI model based on the first request.

[0186] The aforementioned second network functional entity may include an AI model storage functional entity. The AI ​​model storage functional entity may be a collective term for multiple network functional entities, which may store multiple AI models respectively.

[0187] It should be understood that this disclosure does not limit how the first network function entity determines the second network function entity to store the target AI model based on the first request. In some embodiments, the first network function entity may determine the second network function entity to store the target AI model based on at least one of service type information and AI model information.

[0188] For example, the first network function entity can select an AI model for the corresponding AI operation as the target AI model based on the business type information. Subsequently, the first network function entity determines the second network function entity to store the target AI model.

[0189] For example, the first network functional entity can select the AI ​​model corresponding to the AI ​​model information as the target AI model. Subsequently, the first network functional entity determines the second network functional entity that stores the target AI model.

[0190] For example, the first network function entity may also select the target AI model by referring to both business type information and AI model information. Subsequently, the first network function entity determines the second network function entity to store the target AI model.

[0191] In other embodiments, the first network functional entity may first select the fourth network functional entity. Next, the first network functional entity adds at least one of the service type information and the target AI model information from the first request to the first session context establishment request, and sends the first session context establishment request to the fourth network functional entity. Finally, the fourth network functional entity may send the address information of the target AI model to the first network functional entity. Alternatively, the fourth network functional entity may obtain the target AI model from the second network functional entity and then send the target AI model to the first network functional entity.

[0192] The aforementioned fourth network functional entity is used to manage the AI ​​services corresponding to the target AI model. For example, the aforementioned fourth network functional entity can be an AI control functional entity.

[0193] The address information of the target AI model can be the Uniform Resource Locator (URL) of the target AI model or the address of the second network function entity.

[0194] S203. The first network functional entity obtains the target AI model from the second network functional entity and sends the target AI model to at least one of the third network functional entity and the terminal.

[0195] In this step, once the first network functional entity determines the second network functional entity storing the target AI model, it can obtain the target AI model from the second network functional entity and send the target AI model to at least one of the third network functional entity and the terminal.

[0196] Among them, the type of the aforementioned third network functional entity can correspond to AI operations.

[0197] For example, the AI ​​operation is an AI operation related to user plane management for the first service data, and the third network function entity is a user plane function entity, a user plane management function entity, or a session management function entity.

[0198] For example, the AI ​​operation is an AI operation related to data plane management of the first business data, and the third network function entity is a data plane function entity, a data plane management function entity, or a task management function entity.

[0199] The user plane function entity includes at least one of the following: N3 interface user plane function (N3 UPF) entity, intermediate user plane function (I-UPF) entity, and protocol data unit session anchor (PDU Session Anchor, PSA) entity.

[0200] It should be noted that there may be one or more third network functional entities, and this disclosure does not impose any restrictions on this. In some embodiments, when there are multiple third network functional entities, the first network functional entity may select the third network functional entity from which to distribute the target AI model.

[0201] In some embodiments, if the third network function entity obtains the target AI model from the second network function entity, the terminal may also send the first service data to the third network function entity so that the third network function entity can use the target AI model to optimize the first service data.

[0202] For example, the terminal can first send the first service data to the access network device, and then the access network device can send the first service data to the third network function entity.

[0203] For example, a third network function entity can use a target AI model to optimize first service data to achieve services such as traffic identification, QoS optimization, and latency control.

[0204] Figure 4 is a flowchart illustrating another method for distributing an AI model according to an embodiment of this disclosure, used to explain another way of distributing an AI model to a third network functional entity and a terminal. As shown in Figure 4, this AI model distribution method is applied to a first network functional entity and includes steps S301-S303:

[0205] S301, The first network function entity receives the first request from the terminal.

[0206] In this disclosure, when a terminal needs to perform AI operations on the first service data, it can send a first request to the first network function entity.

[0207] The aforementioned first network function entity may include a Session Management Function (SMF) entity.

[0208] In some embodiments, the first request is used to request the execution of an AI operation on the first business data. For example, the first request may be a request to establish an AI service corresponding to the first business data.

[0209] The first service data mentioned above can be any type of service data of the terminal. For example, the first service data can include service data such as traffic identification, QoS optimization, and latency control.

[0210] It should be understood that the embodiments of this disclosure do not limit the AI ​​operation requested to be performed in the first request. In some embodiments, the AI ​​operation is a user-plane management-related AI operation for the first business data, such as business recognition, intelligent interaction, behavior prediction, etc. In other embodiments, the AI ​​operation is a data-plane management-related AI operation for the first business data, such as data integration, data cleaning, data analysis, etc.

[0211] In some embodiments, the first request includes at least one of business type information and target AI model information, both of which are used to obtain the target AI model. The target AI model is used for AI operations.

[0212] S302. The first network functional entity determines the second network functional entity that stores the target AI model according to the first request.

[0213] The aforementioned second network functional entity may include an AI model storage functional entity, which can store the target AI model.

[0214] It should be understood that this disclosure does not limit how the first network function entity determines the second network function entity to store the target AI model based on the first request. In some embodiments, the first network function entity may determine the second network function entity to store the target AI model based on at least one of service type information and AI model information.

[0215] In other embodiments, the first network functional entity may first select the fourth network functional entity. Next, the first network functional entity adds at least one of the service type information and the target AI model information from the first request to the first session context establishment request, and sends the first session context establishment request to the fourth network functional entity. Finally, the fourth network functional entity may send the address information of the target AI model to the first network functional entity. Alternatively, the fourth network functional entity may obtain the target AI model from the second network functional entity and then send the target AI model to the first network functional entity.

[0216] The aforementioned fourth network functional entity is used to manage the AI ​​services corresponding to the target AI model. For example, the aforementioned fourth network functional entity can be an AI control functional entity.

[0217] S303, the first network function entity instructs at least one of the third network function entity and the terminal to obtain the target AI model from the second network function entity.

[0218] In this step, once the first network functional entity determines the second network functional entity storing the target AI model, it can instruct at least one of the third network functional entity and the terminal to obtain the target AI model from the second network functional entity.

[0219] Among them, the type of the aforementioned third network functional entity can correspond to AI operations.

[0220] For example, the AI ​​operation is an AI operation related to user plane management for the first service data, and the third network function entity is a user plane function entity, a user plane management function entity, or a session management function entity.

[0221] For example, the AI ​​operation is an AI operation related to data plane management of the first business data, and the third network function entity is a data plane function entity, a data plane management function entity, or a task management function entity.

[0222] The user plane function entity includes at least one of the following: N3 interface user plane function (N3 UPF) entity, intermediate user plane function (I-UPF) entity, and protocol data unit session anchor (PDU Session Anchor, PSA) entity.

[0223] It should be noted that there may be one or more third network functional entities, and this disclosure does not impose any restrictions on this. In some embodiments, when there are multiple third network functional entities, the first network functional entity may select the third network functional entity from which to distribute the target AI model.

[0224] In some embodiments, a first network functional entity sends model acquisition instruction information to at least one of a third network functional entity and a terminal. The model acquisition instruction information includes the address information of the target AI model. After receiving the acquisition instruction information, at least one of the third network functional entity and the terminal can acquire the target AI model from the second network functional entity based on the address information of the target AI model.

[0225] The address information of the target AI model can be the Uniform Resource Locator (URL) of the target AI model or the address of the second network function entity.

[0226] In some embodiments, if the third network function entity obtains the target AI model from the second network function entity, the terminal may also send the first service data to the third network function entity so that the third network function entity can use the target AI model to optimize the first service data.

[0227] The AI ​​model distribution method provided in this disclosure involves a first network functional entity receiving a first request from a terminal, requesting AI operations to be performed on first service data. Subsequently, based on the first request, the first network functional entity determines a second network functional entity storing a target AI model, which is used for the AI ​​operation. Finally, the first network functional entity retrieves the target AI model from the second network functional entity and sends it to at least one of a third network functional entity and the terminal; or, the first network functional entity instructs at least one of the third network functional entity and the terminal to retrieve the target AI model from the second network functional entity. Because the first network functional entity determines the second network functional entity storing the target AI model, and then sends the target AI model from the second network functional entity to at least one of the third network functional entity and the terminal, or instructs at least one of the third network functional entity and the terminal to retrieve the target AI model from the second network functional entity, flexible deployment of the target AI model is achieved, thereby reducing the transmission overhead and processing latency of the related first service data.

[0228] The following describes model aggregation when the training mechanism of the target AI model is distributed learning or federated learning. Figure 5 is a flowchart illustrating another AI model distribution method provided in this embodiment. As shown in Figure 5, the AI ​​model distribution method includes S401-S411:

[0229] S401, The first network function entity receives a first request from the terminal, the first request being used to request the execution of an AI operation on the first service data.

[0230] S402. The first network functional entity determines the second network functional entity that stores the target AI model according to the first request. The target AI model is used for AI operations.

[0231] After S402, execute S403; or, after S402, execute S404.

[0232] S403, the first network functional entity obtains the target AI model from the second network functional entity and sends the target AI model to at least one of the third network functional entity and the terminal.

[0233] S404, The first network function entity instructs at least one of the third network function entity and the terminal to obtain the target AI model from the second network function entity.

[0234] S405, The first network functional entity sends the convergence instruction information of the target AI model to the third network functional entity.

[0235] The aggregation indication information is used to indicate the central node entity in the third network functional entity. The central node entity is the aggregation node for the model training information of the target AI model on the third network functional entity.

[0236] In some embodiments, the model training information of the target AI model includes at least one of the model parameters and gradient parameters of the target AI model.

[0237] It should be understood that distributed learning is a machine learning mechanism that distributes learning tasks across multiple computing nodes for parallel processing; federated learning is a machine learning mechanism that allows multiple computing nodes to participate in model training without directly sharing the original data. Both distributed learning and federated learning involve multiple computing nodes, and correspondingly, the target AI model can be distributed as an initial model to multiple third-party network functional entities.

[0238] For example, when the training mechanism of the target AI model is distributed learning or federated learning, the third network functional entities distributed to the AI ​​model include central node entities and non-central node entities. By aggregating instruction information, the central node entities in the third network functional entities can be designated as the aggregation object of the model training information of the target AI model.

[0239] S406, the third network functional entity trains the target AI model to obtain model training information.

[0240] In some embodiments, the third network functional entity can train and update the model based on the target AI model and the first business data it has acquired, thereby obtaining model training information.

[0241] S407. Non-central node entities send model training information to the central node.

[0242] S408, the central node entity aggregation model training information is used to obtain the updated AI model.

[0243] It should be understood that the embodiments of this disclosure do not limit how the central node entity aggregates model training information. In some embodiments, the central node entity can aggregate model training information by averaging, taking the median, weighted averaging, etc.

[0244] After S408, execute S409; or, after S408, execute S410-S411.

[0245] S409. The central node entity sends the updated AI model to the non-central node entities.

[0246] S410, the central node entity sends the updated AI model to the first network function entity.

[0247] S411, The first network functional entity sends the updated AI model to the non-central node entity in the third network functional entity.

[0248] The signaling interaction during the AI ​​model distribution process is described below. Figure 6 is a signaling interaction diagram of an AI model distribution method provided in an embodiment of this disclosure. As shown in Figure 6, the AI ​​model distribution method includes S501-S509:

[0249] S501, The terminal sends a first request to the access network device. The first request is used to request the execution of AI operations on the first service data.

[0250] The first request is used to request the execution of AI operations on the first business data. The first request includes at least one of the business type information and the target AI model information of the first business data, both of which are used to obtain the target AI model.

[0251] S502, The access network device sends a first request to the first network function entity.

[0252] In some embodiments, before the access network device sends a first request to the first network function entity, the access network device may select the first network function entity based on the type of the access network device.

[0253] The types of access network equipment include service-oriented and non-service-oriented equipment.

[0254] For example, in the case of service-oriented access network equipment, after receiving the first request, the access network equipment can directly select the first network function entity and send the first request to the first network function entity by signaling interaction with the first network function entity.

[0255] For example, in the case of non-service access network equipment, after receiving the first request, the access network equipment can first select the Access and Mobility Management Function (AMF) entity, then the AMF entity selects the first network function entity, and then the access network equipment sends the first request to the first network function entity.

[0256] S503. The first network functional entity determines the second network functional entity that stores the target AI model according to the first request. The target AI model is used for AI operations.

[0257] After 503, execute S504-S505; or, after S503, execute S506.

[0258] S504. The first network functional entity obtains the target AI model from the second network functional entity.

[0259] S505, the first network functional entity sends the target AI model to at least one of the third network functional entity and the terminal.

[0260] S506, The first network function entity instructs at least one of the third network function entity and the terminal to obtain the target AI model from the second network function entity.

[0261] S507. The terminal sends the first service data to the access network equipment.

[0262] S508, the access network device sends the first service data to the third network function entity, which includes the target AI model.

[0263] S509, The third network functional entity uses the target AI model to optimize the first business data.

[0264] For example, a third network function entity can use a target AI model to optimize first service data to achieve services such as traffic identification, QoS optimization, and latency control.

[0265] The following describes how to distribute AI models when dealing with a single third-party network functional entity. Figure 7 is a signaling interaction diagram of another AI model distribution method provided in this embodiment of the disclosure. As shown in Figure 7, the AI ​​model distribution method includes S601-S616:

[0266] S601. The terminal sends a first request to the access network device. The first request is used to request the execution of AI operations on the first service data.

[0267] The first request can be a request to establish an AI business.

[0268] In some embodiments, the first request may include at least one of business type information and target AI model information. Additionally, the first request may also include parameters such as S-NSSAI, Session and Service Continuity Mode (SSC Mode).

[0269] S602. The access network device selects the first network function entity according to the type of the access network device.

[0270] In some embodiments, when the access network device is non-service-oriented, the access network device first selects an AMF (Access Function Entity), and then the AMF selects a first network function entity. In other embodiments, when the access network device is non-service-oriented, the access network device directly selects the first network function entity.

[0271] S603. The access network device sends a first request to the first network function entity.

[0272] S604. The first network function entity selects the fourth network function entity based on at least one of the service type information and the target AI model information.

[0273] S605, The first network function entity sends a first session context establishment request to the fourth network function entity.

[0274] The first session context establishment request includes at least one of the following: business type information and target AI model information.

[0275] S606, the fourth network function entity determines the second network function entity that stores the target AI model based on at least one of the service type information and the target AI model information.

[0276] S607. The fourth network functional entity sends the address information of the target AI model to the first network functional entity.

[0277] After S607, execute S608-S610; or, after S607, execute S611-S613.

[0278] S608. The first network function entity sends an AI model request message to the second network function entity. The AI ​​model request message includes the address information of the target AI model.

[0279] S609, The second network functional entity sends the target AI model to the first network functional entity.

[0280] S610, the first network function entity sends a second session context establishment request to the third network function entity.

[0281] The second session context establishment request is the N4 session context establishment request, which includes the target AI model.

[0282] In some embodiments, the second session context establishment request may also include information such as forward action rule (FAR), packet detection rule (PDR), QoS, and radio-side tunnel identifier.

[0283] S611. The first network functional entity sends model acquisition instruction information to the third network functional entity. The model acquisition instruction information includes the address information of the target AI model.

[0284] S612. The third network functional entity obtains the target AI model based on the address information of the target AI model.

[0285] S613, the third network functional entity deploys the target AI model.

[0286] S614. The first network function entity sends an AI service session establishment response to the access network device.

[0287] S615, The access network device sends an AI service session establishment response to the terminal.

[0288] S616. The third network functional entity uses the target AI model to process AI business.

[0289] AI services may include traffic identification, QoS optimization, and latency control.

[0290] It should be noted that this embodiment does not restrict the order of the deployment of the target AI model in S613 and the establishment of the AI ​​service session in S614-S615; both can occur within S616. That is, before the terminal sends the first piece of first service data, the third network function entity completes the deployment of the target AI model, thereby enabling the processing of the first service data based on the target AI model.

[0291] Furthermore, a terminal can initiate multiple first requests. If model 1 is deployed when the terminal initiates the first first request, and model 1 can still meet the requirements of the second first request, then no new model will be deployed. That is, the same model may serve multiple first requests from the same terminal or different first requests from different terminals.

[0292] The following describes how to distribute AI models when there are multiple third-party network functional entities. Figure 8 is a signaling interaction diagram of another AI model distribution method provided in this embodiment of the disclosure. As shown in Figure 8, the AI ​​model distribution method includes S701-S717:

[0293] S701, The terminal sends a first request to the access network device. The first request is used to request the execution of AI operations on the first service data.

[0294] The first request can be a request to establish an AI business.

[0295] In some embodiments, the first request may include at least one of business type information and target AI model information. Additionally, the first request may also include parameters such as S-NSSAI, Session and Service Continuity Mode (SSC Mode).

[0296] S702. The access network device selects the first network function entity according to the type of the access network device.

[0297] In some embodiments, when the access network device is non-service-oriented, the access network device first selects an AMF (Access Function Entity), and then the AMF selects a first network function entity. In other embodiments, when the access network device is non-service-oriented, the access network device directly selects the first network function entity.

[0298] S703, The access network device sends a first request to the first network function entity.

[0299] S704. The first network function entity selects the fourth network function entity based on at least one of the service type information and the target AI model information.

[0300] S705, The first network function entity sends a first session context establishment request to the fourth network function entity.

[0301] The first session context establishment request includes at least one of the following: business type information and target AI model information.

[0302] S706, the fourth network function entity determines the second network function entity that stores the target AI model based on at least one of the service type information and the target AI model information.

[0303] S707, the fourth network functional entity sends the address information of the target AI model to the first network functional entity.

[0304] S708, First Network Functional Entity Selection N3 UPF Entity Distribution Target AI Model.

[0305] For example, the N3 UPF entity in the third network function entity is connected to the access network device and serves as the anchor point for receiving terminal services sent by the access network device. It can optimize the first service data based on the target AI model in the first instance, thereby generating a strategy, which is then sent to other third network function entities (e.g., at least one of the I-UPF entity and PSA entity), thereby reducing latency and improving reliability.

[0306] After S708, execute S709-S711; or, after S708, execute S712-S713.

[0307] S709. The first network function entity sends an AI model request message to the second network function entity. The AI ​​model request message includes the address information of the target AI model.

[0308] S710, the second network functional entity sends the target AI model to the first network functional entity.

[0309] S711, The first network function entity sends a second session context establishment request to the N3 UPF entity.

[0310] The second session context establishment request is the N4 session context establishment request, which includes the target AI model.

[0311] In some embodiments, the second session context establishment request may also include information such as forward action rule (FAR), packet detection rule (PDR), QoS, and radio-side tunnel identifier.

[0312] S712, The first network function entity sends model acquisition instruction information to the N3 UPF entity. The model acquisition instruction information includes the address information of the target AI model.

[0313] The S713 and N3 UPF entities obtain the target AI model based on the address information of the target AI model.

[0314] S714, N3 UPF entity deployment target AI model.

[0315] S715, The first network function entity sends an AI service session establishment response to the access network equipment.

[0316] S716, The access network device sends an AI service session establishment response to the terminal.

[0317] S717 and N3 UPF entities use the target AI model to process AI business.

[0318] AI services may include traffic identification, QoS optimization, and latency control.

[0319] Figure 9 is a signaling interaction diagram of another AI model distribution method provided in this embodiment of the present disclosure. As shown in Figure 9, the AI ​​model distribution method includes S801-S817:

[0320] S801, The terminal sends a first request to the access network device. The first request is used to request the execution of AI operations on the first service data.

[0321] The first request can be a request to establish an AI business.

[0322] In some embodiments, the first request may include at least one of business type information and target AI model information. Additionally, the first request may also include parameters such as S-NSSAI, Session and Service Continuity Mode (SSC Mode).

[0323] S802. The access network device selects the first network function entity according to the type of the access network device.

[0324] In some embodiments, when the access network device is non-service-oriented, the access network device first selects an AMF (Access Function Entity), and then the AMF selects a first network function entity. In other embodiments, when the access network device is non-service-oriented, the access network device directly selects the first network function entity.

[0325] S803, The access network device sends a first request to the first network function entity.

[0326] S804. The first network function entity selects the fourth network function entity based on at least one of the service type information and the target AI model information.

[0327] S805, the first network function entity sends a first session context establishment request to the fourth network function entity.

[0328] The first session context establishment request includes at least one of the following: business type information and target AI model information.

[0329] S806, the fourth network function entity determines the second network function entity that stores the target AI model based on at least one of the service type information and the target AI model information.

[0330] S807, the fourth network functional entity sends the address information of the target AI model to the first network functional entity.

[0331] S808, First Network Functional Entity Selects I-UPF Entity Distribution Target AI Model.

[0332] For example, the I-UPF entity in the third network function entity is connected to other user function plane entities (e.g., N3 UPF entity and PSA entity). It can optimize the first service data based on the target AI model in the first time, thereby generating a policy, which is then sent to other third network function entities (e.g., N3 UPF entity and PSA entity), thereby reducing latency and improving reliability.

[0333] It should be noted that when there are multiple I-UPF entities, one of the I-UPF entities can be selected to distribute the target AI model.

[0334] After S808, execute S809-S811; or, after S808, execute S812-S813.

[0335] S809. The first network function entity sends an AI model request message to the second network function entity. The AI ​​model request message includes the address information of the target AI model.

[0336] S810, the second network functional entity sends the target AI model to the first network functional entity.

[0337] S811, The first network function entity sends a second session context establishment request to the I-UPF entity.

[0338] The second session context establishment request is the N4 session context establishment request, which includes the target AI model.

[0339] In some embodiments, the second session context establishment request may also include information such as forward action rule (FAR), packet detection rule (PDR), QoS, and radio-side tunnel identifier.

[0340] S812, The first network function entity sends model acquisition instruction information to the I-UPF entity, which includes the address information of the target AI model.

[0341] S813, the I-UPF entity obtains the target AI model based on the address information of the target AI model.

[0342] S814, I-UPF entity deployment target AI model.

[0343] S815, The first network function entity sends an AI service session establishment response to the access network device.

[0344] S816, The access network device sends an AI service session establishment response to the terminal.

[0345] S817, I-UPF entities use the target AI model to process AI business.

[0346] AI services may include traffic identification, QoS optimization, and latency control.

[0347] Figure 10 is a signaling interaction diagram of another AI model distribution method provided in this embodiment of the present disclosure. As shown in Figure 10, the AI ​​model distribution method includes S901-S917:

[0348] S901, The terminal sends a first request to the access network device. The first request is used to request the execution of AI operations on the first service data.

[0349] The first request can be a request to establish an AI business.

[0350] In some embodiments, the first request may include at least one of business type information and target AI model information. Additionally, the first request may also include parameters such as S-NSSAI, Session and Service Continuity Mode (SSC Mode).

[0351] S902. The access network device selects the first network function entity according to the type of the access network device.

[0352] In some embodiments, when the access network device is non-service-oriented, the access network device first selects an AMF (Access Function Entity), and then the AMF selects a first network function entity. In other embodiments, when the access network device is non-service-oriented, the access network device directly selects the first network function entity.

[0353] S903, The access network device sends a first request to the first network function entity.

[0354] S904. The first network function entity selects the fourth network function entity based on at least one of the service type information and the target AI model information.

[0355] S905, The first network function entity sends a first session context establishment request to the fourth network function entity.

[0356] The first session context establishment request includes at least one of the following: business type information and target AI model information.

[0357] S906, the fourth network function entity determines the second network function entity that stores the target AI model based on at least one of the service type information and the target AI model information.

[0358] S907, the fourth network functional entity sends the address information of the target AI model to the first network functional entity.

[0359] S908, the first network functional entity selects the PSA entity to distribute the target AI model.

[0360] For example, the PSA entity in the third network functional entity is directly connected to the external network. It can optimize the first service data based on the target AI model in real time, thereby generating a strategy for better interaction with the external network. At the same time, the PSA entity can also receive the downlink first service data sent by the external network in real time, process it based on the target AI model, and reduce latency and improve reliability.

[0361] After S908, execute S909-S911; or, after S908, execute S912-S913.

[0362] S909, The first network functional entity sends an AI model request message to the second network functional entity. The AI ​​model request message includes the address information of the target AI model.

[0363] S910, the second network functional entity sends the target AI model to the first network functional entity.

[0364] S911, The first network function entity sends a second session context establishment request to the PSA entity.

[0365] The second session context establishment request is the N4 session context establishment request, which includes the target AI model.

[0366] In some embodiments, the second session context establishment request may also include information such as forward action rule (FAR), packet detection rule (PDR), QoS, and radio-side tunnel identifier.

[0367] S912, The first network functional entity sends model acquisition instruction information to the PSA entity, which includes the address information of the target AI model.

[0368] S913, the PSA entity obtains the target AI model based on the address information of the target AI model.

[0369] S914, PSA entity deployment target AI model.

[0370] S915, The first network function entity sends an AI service session establishment response to the access network equipment.

[0371] S916, The access network device sends an AI service session establishment response to the terminal.

[0372] S917, PSA entities use target AI models to handle AI business.

[0373] AI services may include traffic identification, QoS optimization, and latency control.

[0374] The distribution of AI models using distributed or federated learning mechanisms is described below. Figure 11 is a signaling interaction diagram of another AI model distribution method provided in this embodiment of the disclosure. As shown in Figure 11, the AI ​​model distribution method includes S1001-S1023:

[0375] S1001, The terminal sends a first request to the access network device. The first request is used to request the execution of AI operations on the first service data.

[0376] The first request can be a request to establish an AI business.

[0377] In some embodiments, the first request may include at least one of business type information and target AI model information. Additionally, the first request may also include parameters such as S-NSSAI, Session and Service Continuity Mode (SSC Mode).

[0378] S1002. The access network device selects the first network function entity according to the type of the access network device.

[0379] In some embodiments, when the access network device is non-service-oriented, the access network device first selects an AMF (Access Function Entity), and then the AMF selects a first network function entity. In other embodiments, when the access network device is non-service-oriented, the access network device directly selects the first network function entity.

[0380] S1003, The access network device sends a first request to the first network function entity.

[0381] S1004. The first network function entity selects the fourth network function entity based on at least one of the service type information and the target AI model information.

[0382] S1005, The first network function entity sends a first session context establishment request to the fourth network function entity.

[0383] The first session context establishment request includes at least one of the following: business type information and target AI model information.

[0384] S1006. The fourth network function entity determines the second network function entity that stores the target AI model based on at least one of the service type information and the target AI model information.

[0385] S1007. The fourth network functional entity sends the address information of the target AI model to the first network functional entity.

[0386] S1008. The first network function entity sends an AI model request message to the second network function entity. The AI ​​model request message includes the address information of the target AI model.

[0387] S1009, The second network functional entity sends the target AI model to the first network functional entity.

[0388] S1010, the first network function entity sends a second session context establishment request to the N3 UPF entity.

[0389] The second session context establishment request is an N4 session context establishment request. The second session context establishment request includes the target AI model and the convergence indication information of the target AI model. The convergence indication information is used to indicate that the central node entity in the third network functional entity is an N3 UPF entity. The central node entity is the convergence node of the model training information of the target AI model on the third network functional entity.

[0390] In some embodiments, the second session context establishment request may also include information such as forward action rule (FAR), packet detection rule (PDR), QoS, and radio-side tunnel identifier.

[0391] S1011, The first network function entity sends a third session context establishment request to the I-UPF entity.

[0392] The third session context establishment request is the N4 session context establishment request. The third session context establishment request includes the target AI model and the convergence indication information of the target AI model. The convergence indication information is used to indicate that the central node entity in the third network function entity is the N3 UPF entity.

[0393] S1012, The first network function entity sends a fourth session context establishment request to the PSA entity.

[0394] The fourth session context establishment request is the N4 session context establishment request. The fourth session context establishment request includes the target AI model and the convergence indication information of the target AI model. The convergence indication information is used to indicate that the central node entity in the third network function entity is the N3 UPF entity.

[0395] S1013 and N3 UPF entities send a second session context response request to the first network function entity.

[0396] In some embodiments, the N3 UPF entity can also allocate user plane resources, deploy the target AI model, and train and update the target AI model.

[0397] S1014, the I-UPF entity sends a fourth session context response request to the first network function entity.

[0398] In some embodiments, the I-UPF entity can also allocate user plane resources, deploy the target AI model, and train and update the target AI model.

[0399] S1015, the PSA entity sends a fourth session context response request to the first network function entity.

[0400] In some embodiments, the PSA entity may also allocate user plane resources, deploy the target AI model, and train and update the target AI model.

[0401] S1016, the I-UPF entity sends the model training information of the target AI model to the N3 UPF entity.

[0402] The model training information of the target AI model includes at least one of the model parameters and gradient parameters of the target AI model.

[0403] S1017, the PSA entity sends the model training information of the target AI model to the N3 UPF entity.

[0404] The training information of the S1018 and N3 UPF entity aggregation model is used to obtain the updated AI model.

[0405] S1019 and N3 UPF entities send the updated AI model to the I-UPF entity.

[0406] S1020 and N3 UPF entities send the updated AI model to the PSA entity.

[0407] S1021 and N3 UPF entities send a model aggregation and distribution completion message to the first network function entity.

[0408] S1022, The first network function entity sends an AI service session establishment response to the access network device.

[0409] S1023, The access network device sends an AI service session establishment response to the terminal.

[0410] Figure 12 is a signaling interaction diagram of another AI model distribution method provided in this embodiment of the present disclosure. As shown in Figure 12, the AI ​​model distribution method includes S1101-S1123:

[0411] S1101, The terminal sends a first request to the access network device. The first request is used to request the execution of AI operations on the first service data.

[0412] The first request can be a request to establish an AI business.

[0413] In some embodiments, the first request may include at least one of business type information and target AI model information. Additionally, the first request may also include parameters such as S-NSSAI, Session and Service Continuity Mode (SSC Mode).

[0414] S1102. The access network device selects the first network function entity according to the type of the access network device.

[0415] In some embodiments, when the access network device is non-service-oriented, the access network device first selects an AMF (Access Function Entity), and then the AMF selects a first network function entity. In other embodiments, when the access network device is non-service-oriented, the access network device directly selects the first network function entity.

[0416] S1103. The access network device sends a first request to the first network function entity.

[0417] S1104. The first network function entity selects the fourth network function entity based on at least one of the service type information and the target AI model information.

[0418] S1105, The first network function entity sends a first session context establishment request to the fourth network function entity.

[0419] The first session context establishment request includes at least one of the following: business type information and target AI model information.

[0420] S1106. The fourth network function entity determines the second network function entity that stores the target AI model based on at least one of the service type information and the target AI model information.

[0421] S1107. The fourth network functional entity sends the address information of the target AI model to the first network functional entity.

[0422] S1108. The first network function entity sends an AI model request message to the second network function entity. The AI ​​model request message includes the address information of the target AI model.

[0423] S1109, The second network functional entity sends the target AI model to the first network functional entity.

[0424] S1110, The first network function entity sends a second session context establishment request to the N3 UPF entity.

[0425] The second session context establishment request is an N4 session context establishment request. The second session context establishment request includes the target AI model and the convergence indication information of the target AI model. The convergence indication information is used to indicate that the central node entity in the third network functional entity is an I-UPF entity. The central node entity is the convergence node of the model training information of the target AI model on the third network functional entity.

[0426] In some embodiments, the second session context establishment request may also include information such as forward action rule (FAR), packet detection rule (PDR), QoS, and radio-side tunnel identifier.

[0427] S1111, The first network function entity sends a third session context establishment request to the I-UPF entity.

[0428] The third session context establishment request is the N4 session context establishment request. The third session context establishment request includes the target AI model and the convergence indication information of the target AI model. The convergence indication information is used to indicate that the central node entity in the third network functional entity is the I-UPF entity.

[0429] S1112, The first network function entity sends a fourth session context establishment request to the PSA entity.

[0430] The fourth session context establishment request is the N4 session context establishment request. The fourth session context establishment request includes the target AI model and the convergence indication information of the target AI model. The convergence indication information is used to indicate that the central node entity in the third network function entity is the I-UPF entity.

[0431] S1113 and N3 UPF entities send a second session context response request to the first network function entity.

[0432] In some embodiments, the N3 UPF entity can also allocate user plane resources, deploy the target AI model, and train and update the target AI model.

[0433] S1114, the I-UPF entity sends a fourth session context response request to the first network function entity.

[0434] In some embodiments, the I-UPF entity can also allocate user plane resources, deploy the target AI model, and train and update the target AI model.

[0435] S1115, the PSA entity sends a fourth session context response request to the first network function entity.

[0436] In some embodiments, the PSA entity may also allocate user plane resources, deploy the target AI model, and train and update the target AI model.

[0437] S1116 and N3 UPF entities send model training information of the target AI model to the I-UPF entity.

[0438] The model training information of the target AI model includes at least one of the model parameters and gradient parameters of the target AI model.

[0439] S1117. The PSA entity sends the model training information of the target AI model to the I-UPF entity.

[0440] The S1118 and I-UPF entity aggregation model training information are used to obtain the updated AI model.

[0441] S1119, the I-UPF entity sends the updated AI model to the N3 UPF entity.

[0442] S1120, the I-UPF entity sends the updated AI model to the PSA entity.

[0443] S1121, the I-UPF entity sends a model aggregation and distribution completion message to the first network function entity.

[0444] S1122. The first network function entity sends an AI service session establishment response to the access network device.

[0445] S1123. The access network device sends an AI service session establishment response to the terminal.

[0446] Figure 13 is a signaling interaction diagram of another AI model distribution method provided in this embodiment of the present disclosure. As shown in Figure 13, the AI ​​model distribution method includes S1201-S1223:

[0447] S1201, The terminal sends a first request to the access network device. The first request is used to request the execution of AI operations on the first service data.

[0448] The first request can be a request to establish an AI business.

[0449] In some embodiments, the first request may include at least one of business type information and target AI model information. Additionally, the first request may also include parameters such as S-NSSAI, Session and Service Continuity Mode (SSC Mode).

[0450] S1202. The access network device selects the first network function entity according to the type of the access network device.

[0451] In some embodiments, when the access network device is non-service-oriented, the access network device first selects an AMF (Access Function Entity), and then the AMF selects a first network function entity. In other embodiments, when the access network device is non-service-oriented, the access network device directly selects the first network function entity.

[0452] S1203, The access network device sends a first request to the first network function entity.

[0453] S1204. The first network function entity selects the fourth network function entity based on at least one of the service type information and the target AI model information.

[0454] S1205, The first network function entity sends a first session context establishment request to the fourth network function entity.

[0455] The first session context establishment request includes at least one of the following: business type information and target AI model information.

[0456] S1206, The fourth network function entity determines the second network function entity that stores the target AI model based on at least one of the service type information and the target AI model information.

[0457] S1207. The fourth network functional entity sends the address information of the target AI model to the first network functional entity.

[0458] S1208. The first network function entity sends an AI model request message to the second network function entity. The AI ​​model request message includes the address information of the target AI model.

[0459] S1209, The second network functional entity sends the target AI model to the first network functional entity.

[0460] S1210, The first network function entity sends a second session context establishment request to the N3 UPF entity.

[0461] The second session context establishment request is an N4 session context establishment request. The second session context establishment request includes the target AI model and the convergence indication information of the target AI model. The convergence indication information is used to indicate that the central node entity in the third network functional entity is a PSA entity. The central node entity is the convergence node of the model training information of the target AI model on the third network functional entity.

[0462] In some embodiments, the second session context establishment request may also include information such as forward action rule (FAR), packet detection rule (PDR), QoS, and radio-side tunnel identifier.

[0463] S1211, The first network function entity sends a third session context establishment request to the I-UPF entity.

[0464] The third session context establishment request is the N4 session context establishment request. The third session context establishment request includes the target AI model and the convergence indication information of the target AI model. The convergence indication information is used to indicate that the central node entity in the third network functional entity is the PSA entity.

[0465] S1212, The first network function entity sends a fourth session context establishment request to the PSA entity.

[0466] The fourth session context establishment request is the N4 session context establishment request. The fourth session context establishment request includes the target AI model and the convergence indication information of the target AI model. The convergence indication information is used to indicate that the central node entity in the third network functional entity is the PSA entity.

[0467] S1213 and N3 UPF entities send a second session context response request to the first network function entity.

[0468] In some embodiments, the N3 UPF entity can also allocate user plane resources, deploy the target AI model, and train and update the target AI model.

[0469] S1214, the I-UPF entity sends a fourth session context response request to the first network function entity.

[0470] In some embodiments, the I-UPF entity can also allocate user plane resources, deploy the target AI model, and train and update the target AI model.

[0471] S1215, the PSA entity sends a fourth session context response request to the first network function entity.

[0472] In some embodiments, the PSA entity may also allocate user plane resources, deploy the target AI model, and train and update the target AI model.

[0473] S1216, The I-UPF entity sends the model training information of the target AI model to the PSA entity.

[0474] The model training information of the target AI model includes at least one of the model parameters and gradient parameters of the target AI model.

[0475] S1217 and N3 UPF entities send model training information of the target AI model to the PSA entity.

[0476] The S1218 and PSA entity aggregation model training information are used to obtain the updated AI model.

[0477] S1219, The PSA entity sends the updated AI model to the I-UPF entity.

[0478] S1220, the PSA entity sends the updated AI model to the PSA entity.

[0479] S1221, The PSA entity sends a model aggregation and distribution completion message to the first network function entity.

[0480] S1222, The first network function entity sends an AI service session establishment response to the access network device.

[0481] S1223, The access network device sends an AI service session establishment response to the terminal.

[0482] The AI ​​model distribution method provided in this disclosure involves a first network functional entity receiving a first request from a terminal, requesting AI operations to be performed on first service data. Subsequently, based on the first request, the first network functional entity determines a second network functional entity storing a target AI model, which is used for the AI ​​operation. Finally, the first network functional entity retrieves the target AI model from the second network functional entity and sends it to at least one of a third network functional entity and the terminal; or, the first network functional entity instructs at least one of the third network functional entity and the terminal to retrieve the target AI model from the second network functional entity. Because the first network functional entity determines the second network functional entity storing the target AI model, and then sends the target AI model from the second network functional entity to at least one of the third network functional entity and the terminal, or instructs at least one of the third network functional entity and the terminal to retrieve the target AI model from the second network functional entity, flexible deployment of the target AI model is achieved. This reduces the transmission overhead and processing latency of the related first service data, while also improving the ability to invoke and process network-native AI.

[0483] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0484] Based on the same concept, this disclosure also provides an AI model distribution apparatus for implementing the AI ​​model distribution method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more AI model distribution apparatus embodiments provided below can be found in the limitations of the AI ​​model distribution method described above, and will not be repeated here.

[0485] In one embodiment, as shown in FIG14, an artificial intelligence (AI) model distribution device 1300 is provided, which is applied to a first network functional entity. The AI ​​model distribution device 1300 includes a first receiving module 1301 and a first processing module 1302.

[0486] The first receiving module 1301 is used to receive a first request from the terminal, the first request being used to request the execution of AI operations on the first business data.

[0487] The first processing module 1302 is configured to, according to the first request, determine a second network functional entity storing a target AI model, the target AI model being used for AI operations; obtain the target AI model from the second network functional entity and send the target AI model to at least one of a third network functional entity and a terminal; or, instruct at least one of the third network functional entity and the terminal to obtain the target AI model from the second network functional entity.

[0488] In some embodiments, the AI ​​operation is a user plane management-related AI operation for the first service data, and the third network function entity is a user plane function entity, a user plane management function entity, or a session management function entity; or...

[0489] AI operations refer to AI operations related to data plane management of the first business data, and the third network function entity is a data plane function entity, a data plane management function entity, or a task management function entity.

[0490] In some embodiments, the user plane function entity includes at least one of the following: N3 interface user plane function entity, intermediate user plane function entity, and protocol data unit session anchor point function entity.

[0491] In some embodiments, the first request includes at least one of business type information and target AI model information, both of which are used to obtain the target AI model.

[0492] In some embodiments, the first processing module 1302 is further configured to select a fourth network function entity, the fourth network function entity being used to manage the AI ​​services corresponding to the target AI model; the first network function entity adds at least one of the service type information and the target AI model information in the first request to the first session context establishment request.

[0493] The distribution of the artificial intelligence (AI) model 1300 also includes a first sending module 1303 for sending a first session context establishment request to a fourth network functional entity.

[0494] The first receiving module 1301 is also used to receive the target AI model or the address information of the target AI model sent by the fourth network functional entity.

[0495] In some embodiments, the first processing module 1302 is further configured to determine, based on at least one of service type information and AI model information, a second network function entity storing the target AI model.

[0496] In some embodiments, the first sending module 1303 is further configured to send model acquisition indication information to at least one of the third network function entity and the terminal, wherein the model acquisition indication information includes the address information of the target AI model.

[0497] In some embodiments, the first processing module 1302 is further configured to select a third network functional entity to distribute the target AI model.

[0498] In some embodiments, when the training mechanism of the target AI model is distributed learning or federated learning, the first processing module 1302 is further configured to send the convergence instruction information of the target AI model to the third network functional entity.

[0499] Among them, the aggregation indication information is used to indicate the central node entity in the third network functional entity. The central node entity is the aggregation node of the model training information of the target AI model on the third network functional entity.

[0500] In some embodiments, the model training information of the target AI model includes at least one of the model parameters and gradient parameters of the target AI model.

[0501] In some embodiments, the method further includes:

[0502] The first receiving module 1301 is also used to receive the updated AI model sent by the central node entity;

[0503] The first sending module 1303 is also used to send the updated AI model to the non-central node entity in the third network functional entity.

[0504] In one embodiment, as shown in FIG15, an artificial intelligence (AI) model distribution device 1400 is provided, applied to a third network functional entity. The AI ​​model distribution 1400 includes:

[0505] The second receiving module 1401 is used to receive the target AI model sent by the first network functional entity; or...

[0506] The second receiving module 1401 is further configured to receive model acquisition instruction information sent by the first network function entity. The model acquisition instruction information is used to instruct at least one of the third network function entity and the terminal to acquire a target AI model from the second network function entity. The target AI model is used to perform AI operations on the first service data.

[0507] In some embodiments, the AI ​​operation is a user plane management-related AI operation for the first service data, and the third network function entity is a user plane function entity, a user plane management function entity, or a session management function entity; or...

[0508] AI operations refer to AI operations related to data plane management of the first business data, and the third network function entity is a data plane function entity, a data plane management function entity, or a task management function entity.

[0509] In some embodiments, the user plane function entity includes at least one of the following: N3 interface user plane function entity, intermediate user plane function entity, and protocol data unit session anchor point function entity.

[0510] In some embodiments, the model acquisition instruction information includes the address information of the target AI model.

[0511] In some embodiments, when the training mechanism of the target AI model is distributed learning or federated learning, the second receiving module 1401 is further configured to receive the convergence instruction information of the target AI model sent by the first network functional entity.

[0512] Among them, the aggregation indication information is used to indicate the central node entity in the third network functional entity. The central node entity is the aggregation node of the model training information of the target AI model on the third network functional entity.

[0513] In some embodiments, the distribution of the artificial intelligence (AI) model 1400 further includes:

[0514] The second processing module 1402 is used to train the target AI model and obtain model training information; the non-central node entity in the third network functional entity sends the model training information to the central node; the central node entity aggregates the model training information to obtain the updated AI model.

[0515] In some embodiments, the second processing module 1402 is further configured to send the updated AI model to a non-central node entity; or, the central node entity sends the updated AI model to a first network function entity.

[0516] In some embodiments, the distribution of the artificial intelligence (AI) model 1400 further includes:

[0517] The second sending module 1403 is used to send the target AI model to the terminal, and the target AI model is used to optimize the first business data on the terminal.

[0518] In some embodiments, the second receiving module 1401 is further configured to receive the first service data of the terminal;

[0519] The second processing module 1402 is also used to optimize the first business data using the target AI model.

[0520] In some embodiments, the target AI model information includes at least one of the following: AI model identification information, AI model storage location information, and AI model size information.

[0521] In one embodiment, as shown in FIG16, an artificial intelligence (AI) model distribution device 1500 is provided, applied to a terminal. The AI ​​model distribution device 1500 includes:

[0522] The third sending module 1501 is used to send a first request to the first network function entity. The first request is used to request the execution of AI operations on the first business data.

[0523] The third receiving module 1502 is used to receive the target AI model sent by the first network function entity; or, to receive model acquisition instruction information from the first network function, wherein the model acquisition instruction information is used to instruct at least one of the third network function entity and the terminal to acquire the target AI model from the second network function entity, wherein the target AI model is used for AI operation.

[0524] In some embodiments, the first request carries at least one of business type information and target AI model information of the first business data, and both the business type information and the target AI model information are used to obtain the target AI model.

[0525] In some embodiments, the AI ​​operation is a user plane management-related AI operation for the first service data, and the third network function entity is a user plane function entity, a user plane management function entity, or a session management function entity; or...

[0526] AI operations refer to AI operations related to data plane management of the first business data, and the third network function entity is a data plane function entity, a data plane management function entity, or a task management function entity.

[0527] In some embodiments, the user plane function entity includes at least one of the following: N3 interface user plane function entity, intermediate user plane function entity, and protocol data unit session anchor point function entity.

[0528] In some embodiments, the third sending module 1501 is used to send service data of the terminal device to a third network function entity.

[0529] In one embodiment, as shown in FIG17, an artificial intelligence (AI) model distribution device 1600 is provided, applied to an access network device. The AI ​​model distribution device 1600 includes:

[0530] The fourth receiving module 1601 is used to receive a first request sent by the terminal. The first request is used to request the execution of an AI operation on the first business data. The AI ​​operation is executed through the target AI model.

[0531] The fourth sending module 1602 is used to send a first request to the first network function entity.

[0532] In some embodiments, the distribution of the artificial intelligence (AI) model 1600 further includes a fourth processing module 1603 for selecting a first network function entity based on the type of access network device.

[0533] In some embodiments, the first request includes at least one of business type information and target AI model information, both of which are used to obtain the target AI model.

[0534] In some embodiments, the fourth receiving module 1601 is further configured to receive first service data sent by the terminal;

[0535] The fourth sending module 1602 is also used to send the first service data to the third network function entity, which includes the target AI model.

[0536] It should be noted that the AI ​​model distribution device provided in this embodiment can implement all the method steps implemented in the AI ​​model distribution method embodiment and achieve the same technical effect. Therefore, the parts and beneficial effects that are the same as those in the method embodiment will not be described in detail here.

[0537] It should be noted that the module division in this embodiment is illustrative and represents only one logical functional division; in actual implementation, other division methods may be used. Furthermore, the functional modules in the various embodiments of this disclosure can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0538] If the integrated module is implemented as a software functional module and sold or used as an independent product, it can be stored in a processor-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the related technology, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this disclosure.

[0539] As shown in Figure 18, this embodiment of the present disclosure also provides an artificial intelligence (AI) model distribution device to implement the AI ​​model distribution method on the first network functional entity side described above. The AI ​​model distribution device includes a processor 1701, a memory 1702, and a transceiver 1703.

[0540] Memory 1702 is used to store computer programs; transceiver 1703 is used to receive and send data under the control of processor 1701. Processor 1701 is used to read the computer program from memory and perform the following operations:

[0541] The transceiver receives a first request from the terminal, which is used to request the execution of an AI operation on the first business data.

[0542] Based on the first request, a second network functional entity is determined to store the target AI model, which is used for AI operations;

[0543] Obtain the target AI model from the second network functional entity and send the target AI model to at least one of the third network functional entity and the terminal; or, instruct at least one of the third network functional entity and the terminal to obtain the target AI model from the second network functional entity.

[0544] In some embodiments, the AI ​​operation is a user plane management-related AI operation for the first service data, and the third network function entity is a user plane function entity, a user plane management function entity, or a session management function entity; or...

[0545] AI operations refer to AI operations related to data plane management of the first business data, and the third network function entity is a data plane function entity, a data plane management function entity, or a task management function entity.

[0546] In some embodiments, the user plane function entity includes at least one of the following: N3 interface user plane function entity, intermediate user plane function entity, and protocol data unit session anchor point function entity.

[0547] In some embodiments, the first request includes at least one of business type information and target AI model information, both of which are used to obtain the target AI model.

[0548] In some embodiments, the processor 1701 also performs the following operations:

[0549] Select the fourth network function entity, which is used to manage the AI ​​business corresponding to the target AI model;

[0550] Add at least one of the business type information and the target AI model information from the first request to the first session context establishment request;

[0551] Send the first session context establishment request to the fourth network function entity via transceiver;

[0552] The transceiver receives the target AI model or the address information of the target AI model sent by the fourth network function entity.

[0553] In some embodiments, the processor 1701 also performs the following operations:

[0554] Based on at least one of the business type information and AI model information, determine the second network functional entity that stores the target AI model.

[0555] In some embodiments, the processor 1701 also performs the following operations:

[0556] Send model acquisition instruction information to at least one of the transceiver third network function entities and terminals, the model acquisition instruction information including the address information of the target AI model.

[0557] In some embodiments, the processor 1701 also performs the following operations:

[0558] The third network functional entity is selected to distribute the target AI model.

[0559] In some embodiments, where the training mechanism of the target AI model is distributed learning or federated learning, the method further includes:

[0560] The target AI model is sent to the third network functional entity via a transceiver.

[0561] Among them, the aggregation indication information is used to indicate the central node entity in the third network functional entity. The central node entity is the aggregation node of the model training information of the target AI model on the third network functional entity.

[0562] In some embodiments, the model training information of the target AI model includes at least one of the model parameters and gradient parameters of the target AI model.

[0563] In some embodiments, the processor 1701 also performs the following operations:

[0564] Receive the updated AI model sent by the central node entity through the transceiver;

[0565] The updated AI model is sent to the non-central node entity in the third network functional entity via a transceiver.

[0566] The AI ​​model distribution device shown in Figure 18 can also implement the AI ​​model distribution method on the third network functional entity side mentioned above. The processor 1701 is used to read the computer program in the memory and perform the following operations:

[0567] Receive the target AI model sent by the first network function entity via a transceiver; or...

[0568] The transceiver receives model acquisition instruction information sent by the first network function entity. The model acquisition instruction information is used to instruct at least one of the third network function entity and the terminal to acquire the target AI model from the second network function entity. The target AI model is used to perform AI operations on the first business data.

[0569] In some embodiments, the AI ​​operation is a user plane management-related AI operation for the first service data, and the third network function entity is a user plane function entity, a user plane management function entity, or a session management function entity; or...

[0570] AI operations refer to AI operations related to data plane management of the first business data, and the third network function entity is a data plane function entity, a data plane management function entity, or a task management function entity.

[0571] In some embodiments, the user plane function entity includes at least one of the following: N3 interface user plane function entity, intermediate user plane function entity, and protocol data unit session anchor point function entity.

[0572] In some embodiments, the model acquisition instruction information includes the address information of the target AI model.

[0573] In some embodiments, when the training mechanism of the target AI model is distributed learning or federated learning, the processor 1701 also performs the following operations:

[0574] The transceiver receives the convergence instruction information of the target AI model sent by the first network function entity.

[0575] Among them, the aggregation indication information is used to indicate the central node entity in the third network functional entity. The central node entity is the aggregation node of the model training information of the target AI model on the third network functional entity.

[0576] In some embodiments, the processor 1701 also performs the following operations:

[0577] Train the target AI model to obtain model training information;

[0578] Non-central node entities send model training information to the central node;

[0579] The central node entity aggregates the model training information to obtain the updated AI model.

[0580] In some embodiments, the processor 1701 also performs the following operations:

[0581] The central node entity sends the updated AI model to the non-central node entities; or, the central node entity sends the updated AI model to the first network functional entity.

[0582] In some embodiments, the processor 1701 also performs the following operations:

[0583] The target AI model is sent to the terminal via a transceiver, and the target AI model is used to optimize the primary business data on the terminal.

[0584] In some embodiments, the processor 1701 also performs the following operations:

[0585] Receive the first service data from the terminal via transceiver;

[0586] Use the target AI model to optimize the primary business data.

[0587] In some embodiments, the target AI model information includes at least one of the following: AI model identification information, AI model storage location information, and AI model size information.

[0588] The AI ​​model distribution device shown in Figure 17 can also implement the AI ​​model distribution method on the access network device side described above. The processor 1701 is used to read the computer program in the memory and perform the following operations:

[0589] The transceiver receives a first request sent by the terminal. The first request is used to request the execution of an AI operation on the first business data. The AI ​​operation is executed through the target AI model.

[0590] The first request is sent to the first network function entity via the transceiver.

[0591] In some embodiments, the processor 1701 also performs the following operations:

[0592] Select the first network function entity based on the type of access network equipment.

[0593] In some embodiments, the first request includes at least one of business type information and target AI model information, both of which are used to obtain the target AI model.

[0594] In some embodiments, the processor 1701 also performs the following operations:

[0595] The transceiver receives the first service data sent by the terminal.

[0596] The first service data is sent to the third network function entity via a transceiver. The third network function entity includes the target AI model.

[0597] In Figure 18, the bus architecture can include any number of interconnected buses and bridges, specifically linking various circuits of one or more processors 1701 (represented by processor 1701) and memory 1702 (represented by memory 1702). The bus architecture can also link various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. The bus interface provides an interface. The transceiver 1703 can be multiple elements, including transmitters and receivers, providing a unit for communicating with various other devices over a transmission medium, including wireless channels, wired channels, optical fibers, etc. For different user equipment, the user interface 1704 can also be an interface capable of connecting external or internal devices, including but not limited to keypads, displays, speakers, microphones, joysticks, etc.

[0598] Processor 1701 is responsible for managing the bus architecture and general processing, while memory 1702 can store the data used by processor 1701 when performing operations.

[0599] In some embodiments, the processor 1701 may be a CPU (Central Processing Unit 1301), an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a CPLD (Complex Programmable Logic Device), and the processor 1701 may also adopt a multi-core architecture.

[0600] As shown in Figure 19, this embodiment of the present disclosure also provides an artificial intelligence (AI) model distribution device to implement the aforementioned terminal-side AI model distribution method. The AI ​​model distribution device includes a processor 1801, a memory 1802, and a transceiver 1803.

[0601] Memory 1802 is used to store computer programs; transceiver 1803 is used to receive and send data under the control of processor 1801. Processor 1801 is used to read the computer program from memory and perform the following operations:

[0602] The transceiver sends a first request to the first network function entity, the first request being used to request the execution of AI operations on the first service data;

[0603] The terminal receives a target AI model sent by a first network function entity via a transceiver; or, the terminal receives model acquisition instruction information from the first network function, which instructs at least one of the third network function entity and the terminal to acquire the target AI model from the second network function entity, and the target AI model is used for AI operations.

[0604] In some embodiments, the first request carries at least one of business type information and target AI model information of the first business data, and both the business type information and the target AI model information are used to obtain the target AI model.

[0605] In some embodiments, the AI ​​operation is a user plane management-related AI operation for the first service data, and the third network function entity is a user plane function entity, a user plane management function entity, or a session management function entity; or...

[0606] AI operations refer to AI operations related to data plane management of the first business data, and the third network function entity is a data plane function entity, a data plane management function entity, or a task management function entity.

[0607] In some embodiments, the user plane function entity includes at least one of the following: N3 interface user plane function entity, intermediate user plane function entity, and protocol data unit session anchor point function entity.

[0608] In some embodiments, the processor 1801 also performs the following operations:

[0609] The terminal device's service data is sent to a third network function entity via a transceiver.

[0610] In Figure 19, the bus architecture can include any number of interconnected buses and bridges, specifically linking various circuits of one or more processors 1801 (represented by a processor) and memory 1802 (represented by a memory). The bus architecture can also link various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. The bus interface provides the interface. The transceiver can be multiple components, including a transmitter and a receiver, providing a unit for communicating with various other devices over a transmission medium, including wireless channels, wired channels, optical fibers, etc. The processor is responsible for managing the bus architecture and general processing, and the memory can store data used by the processor 1801 during operation.

[0611] The processor 1801 can be a central processing unit (CPU), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a complex programmable logic device (CPLD). The processor can also adopt a multi-core architecture.

[0612] This disclosure also provides a processor-readable storage medium, which can be any available medium or data storage device that can be accessed by a processor, including but not limited to magnetic memory (e.g., floppy disk, hard disk, magnetic tape, magneto-optical disk (MO)), optical memory (e.g., CD, DVD, BD, HVD), and semiconductor memory (e.g., ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)).

[0613] In one embodiment, this disclosure also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described sensing signal transmission method, sensing signal reception method, and sensing signal configuration method.

[0614] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this disclosure can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this disclosure may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this disclosure may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0615] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0616] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-executable instructions. These computer-executable instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more flowchart illustrations and / or one or more block diagrams.

[0617] These processor-executable instructions may also be stored in a processor-readable memory that can instruct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the processor-readable memory produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0618] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0619] Obviously, those skilled in the art can make various modifications and variations to this disclosure without departing from its spirit and scope. Therefore, if such modifications and variations fall within the scope of the claims of this disclosure and their equivalents, this disclosure is also intended to include such modifications and variations.

[0620] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this disclosure, and not to limit them. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this disclosure, and they should all be covered within the scope of the claims and specification of this disclosure. In particular, as long as there is no structural conflict, the various technical features mentioned in the various embodiments can be combined in any way. This disclosure is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method of distributing an artificial intelligence (AI) model, wherein, Applied to the first network functional entity, including: The first network function entity receives a first request from the terminal, the first request being used to request the execution of AI operations on the first service data; The first network functional entity determines a second network functional entity that stores the target AI model based on the first request; the target AI model is used for the AI ​​operation. The first network function entity obtains the target AI model from the second network function entity and sends the target AI model to at least one of the third network function entity and the terminal; or, the first network function entity instructs at least one of the third network function entity and the terminal to obtain the target AI model from the second network function entity.

2. The method of claim 1, wherein, The AI ​​operation is a user plane management-related AI operation for the first service data, and the third network function entity is a user plane function entity, a user plane management function entity, or a session management function entity; or... The AI ​​operation is an AI operation related to data plane management of the first business data, and the third network function entity is a data plane function entity, a data plane management function entity, or a task management function entity.

3. The method of claim 2, wherein, The user plane functional entity includes at least one of the following: N3 interface user plane functional entity, intermediate user plane functional entity, and protocol data unit session anchor point functional entity.

4. The method of claim 1, wherein, The first request includes at least one of the business type information of the first business data and the information of the target AI model, both of which are used to obtain the target AI model.

5. The method of claim 4, wherein, The method further includes: The first network functional entity selects a fourth network functional entity, which is used to manage the AI ​​services corresponding to the target AI model. The first network function entity adds at least one of the service type information and / or the target AI model information from the first request to the first session context establishment request; The first network function entity sends the first session context establishment request to the fourth network function entity; The first network function entity receives the target AI model or the address information of the target AI model sent by the fourth network function entity.

6. The method of claim 4, wherein, The first network functional entity determines, based on the first request, the second network functional entity that stores the target AI model, including: The first network function entity determines the second network function entity that stores the target AI model based on at least one of the service type information and the target AI model information.

7. The method of claim 5 or 6, wherein, The first network function entity instructs at least one of the third network function entity and the terminal to obtain the target AI model from the second network function entity, including: The first network function entity sends model acquisition instruction information to at least one of the third network function entity and the terminal, the model acquisition instruction information including the address information of the target AI model.

8. The method of claim 1, wherein, The method further includes: The first network functional entity selects a third network functional entity to distribute the target AI model.

9. The method of claim 8, wherein, If the training mechanism of the target AI model is distributed learning or federated learning, then the method further includes: The first network functional entity sends the convergence instruction information of the target AI model to the third network functional entity; The convergence indication information is used to indicate the central node entity in the third network functional entity, and the central node entity is the convergence node of the model training information of the target AI model on the third network functional entity.

10. The method of claim 9, wherein, The model training information of the target AI model includes at least one of the model parameters and gradient parameters of the target AI model.

11. The method of claim 9 or 10, wherein, The method further includes: The first network functional entity receives the updated AI model sent by the central node entity; The first network functional entity sends the updated AI model to the non-central node entity in the third network functional entity. 12.A method of distributing an artificial intelligence (AI) model, wherein, Applied to third network functional entities, including: The third network functional entity receives the target AI model sent by the first network functional entity; or... The third network function entity receives model acquisition instruction information sent by the first network function entity. The model acquisition instruction information is used to instruct at least one of the third network function entity and the terminal to acquire the target AI model from the second network function entity. The target AI model is used to perform AI operations on the first business data.

13. The method of claim 12, wherein, The AI ​​operation is a user plane management-related AI operation for the first service data, and the third network function entity is a user plane function entity, a user plane management function entity, or a session management function entity; or... The AI ​​operation is an AI operation related to data plane management of the first business data, and the third network function entity is a data plane function entity, a data plane management function entity, or a task management function entity.

14. The method of claim 13, wherein, The user plane functional entity includes at least one of the following: N3 interface user plane functional entity, intermediate user plane functional entity, and protocol data unit session anchor point functional entity.

15. The method of claim 13, wherein, The model acquisition instruction information includes the address information of the target AI model.

16. The method of claim 12, wherein, If the training mechanism of the target AI model is distributed learning or federated learning, then the method further includes: The third network functional entity receives the convergence indication information of the target AI model sent by the first network functional entity; The convergence indication information is used to indicate the central node entity in the third network functional entity, and the central node entity is the convergence node of the model training information of the target AI model on the third network functional entity.

17. The method of claim 16, wherein, The method further includes: The third network functional entity trains the target AI model to obtain the model training information; The non-central node entity in the third network functional entity sends the model training information to the central node; The central node entity aggregates the model training information to obtain the updated AI model.

18. The method of claim 17, wherein, The method further includes: The central node entity sends the updated AI model to the non-central node entity; or... The central node entity sends the updated AI model to the first network function entity.

19. The method of claim 12, wherein, The method further includes: The third network function entity sends the target AI model to the terminal, and the target AI model is used to optimize the first service data on the terminal.

20. The method of claim 12 or 19, wherein, The method further includes: The third network function entity receives the first service data from the terminal. The third network function entity uses the target AI model to optimize the first business data.

21. The method of claim 12, wherein, The target AI model information includes at least one of the following: the identification information of the AI ​​model, the storage location information of the AI ​​model, and the model size information of the AI ​​model.

22. A method of distributing an artificial intelligence (AI) model, wherein, Applied to terminals, including: The terminal sends a first request to the first network function entity, the first request being used to request the execution of AI operations on the first service data; The terminal receives a target AI model sent by the first network function entity; or, the terminal receives model acquisition instruction information from the first network function, the model acquisition instruction information being used to instruct at least one of the third network function entity and the terminal to acquire the target AI model from the second network function entity, the target AI model being used for the AI ​​operation.

23. The method of claim 22, wherein, The first request carries at least one of the business type information of the first business data and the information of the target AI model, wherein both the business type information and the information of the target AI model are used to obtain the target AI model.

24. The method of claim 22, wherein, The AI ​​operation refers to user plane management-related AI operations for the first service data, and the third network function entity is a user plane function entity, a user plane management function entity, or a session management function entity; or... The AI ​​operation is an AI operation related to data plane management of the first business data, and the third network function entity is a data plane function entity, a data plane management function entity, or a task management function entity.

25. The method of claim 24, wherein, The user plane functional entity includes at least one of the following: N3 interface user plane functional entity, intermediate user plane functional entity, and protocol data unit session anchor point functional entity.

26. The method of any one of claims 22-25, wherein, The method further includes: The terminal sends the terminal device's service data to the third network function entity.

27. A method of distribution of an artificial intelligence (AI) model, wherein, Applied to access network equipment, including: The access network device receives a first request sent by the terminal, the first request being used to request the execution of an AI operation on the first service data, the AI ​​operation being executed through a target AI model; The access network device sends the first request to the first network function entity.

28. The method of claim 27, wherein, The method further includes: The access network device selects the first network function entity according to the type of the access network device.

29. The method of claim 27, wherein, The first request includes at least one of the business type information of the first business data and the information of the target AI model, both of which are used to obtain the target AI model.

30. The method of any one of claims 27-29, wherein, The method further includes: The access network device receives the first service data sent by the terminal; The access network device sends the first service data to a third network function entity, which includes the target AI model.

31. A distribution apparatus of an artificial intelligence (AI) model, wherein, Applied to the first network functional entity, including memory, transceiver, and processor: A memory for storing computer programs; a transceiver for sending and receiving data under the control of the processor; and a processor for reading the computer programs from the memory and performing the following operations: The transceiver receives a first request from the terminal, the first request being used to request the execution of an AI operation on the first business data; Based on the first request, a second network functional entity is determined to store the target AI model, the target AI model being used for the AI ​​operation; The target AI model is obtained from the second network functional entity and sent to at least one of the third network functional entity and the terminal; or, at least one of the third network functional entity and the terminal is instructed to obtain the target AI model from the second network functional entity.

32. A distribution apparatus of an artificial intelligence (AI) model, wherein, Applied to third-party network functional entities, including memory, transceiver, and processor: A memory for storing computer programs; a transceiver for sending and receiving data under the control of the processor; and a processor for reading the computer programs from the memory and performing the following operations: The transceiver receives the target AI model sent by the first network function entity; or... The transceiver receives model acquisition instruction information sent by the first network function entity. The model acquisition instruction information is used to instruct at least one of the third network function entity and the terminal to acquire the target AI model from the second network function entity. The target AI model is used to perform AI operations on the first service data.

33. A distribution apparatus of an artificial intelligence (AI) model, wherein, Applications in terminals, including memory, transceiver, and processor: The memory is used to store computer programs; the transceiver is used to send and receive data under the control of the processor. Processor, configured to read the computer program in the memory and perform the following operations: The transceiver sends a first request to the first network function entity, the first request being used to request the execution of an AI operation on the first service data; The transceiver receives the target AI model sent by the first network function entity; or, the transceiver receives model acquisition instruction information from the first network function, the model acquisition instruction information being used to instruct at least one of the third network function entity and the terminal to acquire the target AI model from the second network function entity, the target AI model being used for the AI ​​operation.

34. A distribution apparatus of an artificial intelligence (AI) model, wherein, Applied to access network equipment, including memory, transceivers, and processors: The memory is used to store computer programs; the transceiver is used to send and receive data under the control of the processor. Processor, configured to read the computer program in the memory and perform the following operations: The transceiver receives a first request sent by the terminal, the first request being used to request the execution of an AI operation on the first business data, the AI ​​operation being executed through a target AI model; The first request is sent to the first network function entity via the transceiver.

35. A processor-readable storage medium, wherein, The processor-readable storage medium stores a program for causing the processor to perform the method according to any one of claims 1 to 30.

36. A computer program product, wherein, The computer program product, when invoked by a computer, causes the computer to perform the method of any one of claims 1 to 30.

Images