Communication network architecture

Abstract

The present disclosure relates to a communication network architecture. In at least one embodiment of the present disclosure, on the basis of a user plane and a control plane of a 5G network architecture, a core network newly adds a data plane and an intelligent plane, a radio access network can receive an intelligent service request sent by a terminal and send intelligent service information carried by the intelligent service request to an intelligent network function of the intelligent plane, the intelligent network function of the intelligent plane can acquire data required by the intelligent service information from an intelligent network function of the data plane, and the intelligent network function of the intelligent plane uses the data required by the intelligent service information to determine an intelligent service result and sends the intelligent service result to the terminal by means of the radio access network, so as to realize interaction and functional collaboration between planes within the communication network architecture, enhance the AI processing capabilities in the core network, and improve the feasibility of a layered and multi-plane architecture for 6G communication networks.

Description

A communication network architecture

[0001] Cross-reference to related applications

[0002] This disclosure claims priority to Chinese Patent Application No. 202410563424.2, filed on May 8, 2024, entitled “A Communication Network Architecture”, the entire contents of which are incorporated herein by reference. Technical Field

[0003] This disclosure relates to the field of communication technology, specifically to a communication network architecture. Background Technology

[0004] Currently, the industry has proposed the concept of layers for the network architecture of the sixth-generation mobile communication standard (6G), but has not proposed the concept of surfaces, the specific network functions of each surface, or the coordination schemes between surfaces. As a result, it is impossible to effectively carry out the interaction between logical functional surfaces, and it is impossible to efficiently realize the functional coordination and interaction of logical functional surfaces. Summary of the Invention

[0005] At least one embodiment of this disclosure provides a communication network architecture.

[0006] This disclosure proposes a communication network architecture, which includes a wireless access network and a core network. The core network includes intelligent network functions of the data plane and intelligent network functions of the intelligent plane.

[0007] The wireless access network is used to receive intelligent service requests and send the intelligent service information carried in the intelligent service requests to the intelligent network function of the intelligent plane.

[0008] The intelligent network function of the intelligent plane is used to obtain the data required for intelligent service information from the intelligent network function of the data plane, determine the intelligent service result based on the data required for intelligent service information, and send the intelligent service response through the radio access network.

[0009] In some embodiments, the interaction between the intelligent network function of the data plane and the intelligent network function of the intelligent plane includes at least one of the following:

[0010] Point-to-point interface interaction;

[0011] Service-oriented interface interaction;

[0012] Interaction is based on anchor point functionality, with different anchor point functions set for the data plane and the intelligence plane.

[0013] In some embodiments, the core network also includes intelligent network functionality on the computing plane;

[0014] The intelligent network function of the intelligent plane is also used to send a computing power request to the intelligent network function of the computing plane when determining the intelligent business result based on the data required by the intelligent business information if the computing power is insufficient.

[0015] The intelligent network function of the computing plane is used to schedule and interconnect computing power based on computing power requests, and to perform intelligent business calculations based on the scheduled and interconnected computing power to obtain intelligent business results.

[0016] In some embodiments, the core network also includes intelligent network functionality for the orchestration management plane;

[0017] Wireless access network, used for intelligent network functions that send intelligent service requests and / or computing power requests to the orchestration management plane;

[0018] The intelligent network function of the orchestration management plane is used to parse the intelligent service information carried in the intelligent service request and send the intelligent service information to the intelligent network function of the intelligent plane; and / or, to parse the computing power demand information corresponding to the computing power request, perform computing power orchestration based on the computing power demand information, obtain computing power policy information and computing power routing information, and send the computing power policy information and computing power routing information to the intelligent network function of the computing plane.

[0019] In some embodiments, the intelligent network function of the intelligent plane includes at least one of the following functions:

[0020] The Intelligent Orchestration and Management Function (AIMOF) is used to decompose AI business into at least one AI task based on intelligent business information, and map at least one AI task to the corresponding requirement information.

[0021] The Intelligent Model Training Function (AIMTF) is used to obtain the training data required for training AI models from the intelligent network function of the data plane, and to train AI models based on the training data.

[0022] The Intelligent Model Validation Function (AIMAF) is used to validate AI models trained by the Intelligent Model Training Function (AIMTF).

[0023] The Intelligent Model Registration Function (AIMRF) is used to register AI models that have passed the Intelligent Model Validation Function (AIMAF) to the network.

[0024] The AIMSF (Intelligent Model Storage Function) is used to store AI models registered by the AIMRF (Intelligent Model Registration Function).

[0025] In some embodiments, the Intelligent Model Training Function (AIMTF) is also used to send a computing power request to the intelligent network function of the computing plane if computing power is insufficient.

[0026] In some embodiments, the intelligent network function of the data plane includes at least one of the following functions:

[0027] Data Management Function (DRF) is used to manage terminal data and network data;

[0028] The Data Orchestration Function (DOF) is used to receive data service requests, convert them into corresponding data pipeline construction requests, and send the data pipeline construction requests to the Data Control Function (DCF).

[0029] Data Control Function (DCF) is used to build data pipelines based on data pipeline build requests.

[0030] Data Agent (DA) is used to provide data service capabilities for data in the data pipeline built by Data Control Function (DCF).

[0031] In some embodiments, the data agent (DA) is also used to report data service capabilities to the data orchestration function (DOF) and the data control function (DCF);

[0032] The Data Orchestration Function (DOF) is also used to select a Data Proxy DA to provide data service capabilities based on data service requests and data service capabilities reported by at least one Data Proxy DA.

[0033] The Data Control Function (DCF) is also used to register or deregister the Data Agent (DA) based on the data service capabilities reported by the DA.

[0034] In some embodiments, the intelligent network functionality of the computing plane includes at least one of the following functions:

[0035] The Computation Control Function (CPCF) is used to parse the computing power requirement information corresponding to the computing power request, perform computing power orchestration based on the computing power requirement information, obtain computing power strategy information and computing power routing information, and send the computing power strategy information and computing power routing information to the Computation Interconnect Function (CPIF).

[0036] The Compute Interconnect Function (CPIF) is used to schedule and interconnect at least one Compute Execution Function (CPEF) based on computing power policy information and computing power routing information.

[0037] The Computation Execution Function (CPEF) is used for scheduling and interconnection based on the Computation Interconnect Function (CPIF) to perform intelligent service computation and obtain intelligent service results.

[0038] In some embodiments, the intelligent network functionality of the orchestration management plane includes at least one of the following functions:

[0039] Computational Control Function (CPCF);

[0040] Data orchestration function DOF;

[0041] The Service Requirement Adaptation Function (SRAF) is used to parse the intelligent service information carried in the intelligent service request sent by the radio access network and send the intelligent service information to the intelligent network function of the intelligent plane.

[0042] AIMOF (Intelligent Orchestration and Management Function)

[0043] The functions that are identical to those in the orchestration and management plane's intelligent network functions in the data plane, intelligent plane, and computing plane have been removed.

[0044] In some embodiments, the AI ​​task is an AI model training task;

[0045] The intelligent orchestration management function AIMOF sends an AI model training request to the intelligent model training function AIMTF. The AI ​​model training request carries the training parameters required to train the AI ​​model.

[0046] The AIMTF intelligent model training function obtains the training data and AI model identifiers required for training the AI ​​model from the intelligent network function of the data plane based on the training parameters.

[0047] The AIMTF (AI Model Training Function) sends an AI model retrieval request to the AIMSF (AI Model Storage Function) based on the AI ​​model's identifier. The AIMSF then sends an AI model retrieval response to the AIMTF, which includes the AI ​​model.

[0048] The AIMTF intelligent model training function retrains the AI ​​model based on the training data;

[0049] The AIMTF intelligent model training function sends the trained AI model to the AIMAF intelligent model verification function for verification.

[0050] The AIMTF intelligent model training function will send the verified AI model to the AIMRF intelligent model registration function for registration.

[0051] The AIMTF intelligent model training function sends the registered AI model to the AIMSF intelligent model storage function for storage.

[0052] In at least one embodiment of this disclosure, based on the user plane and control plane of the 5G network architecture, a data plane and an intelligent plane are added to the core network. The radio access network can receive intelligent service requests sent by the terminal and send the intelligent service information carried in the intelligent service request to the intelligent network function of the intelligent plane. The intelligent network function of the intelligent plane can obtain the data required for the intelligent service information from the intelligent network function of the data plane. The intelligent network function of the intelligent plane uses the data required for the intelligent service information to determine the intelligent service result and sends the intelligent service result to the terminal through the radio access network. This realizes the interaction and functional collaboration between planes in the communication network architecture, enhances the AI ​​processing capability of the core network, and improves the feasibility of the layered and planened architecture of the 6G communication network. Attached Figure Description

[0053] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings.

[0054] Figure 1 is a schematic diagram of a communication network architecture provided in an embodiment of this disclosure;

[0055] Figure 2 is a schematic diagram of the intelligent network functions included in different aspects of the communication network architecture shown in Figure 1;

[0056] Figure 3 is a schematic diagram of another communication network architecture provided in an embodiment of this disclosure;

[0057] Figure 4 is a schematic diagram of the intelligent network functions included in different aspects of the communication network architecture shown in Figure 3;

[0058] Figure 5 is a schematic diagram of another communication network architecture provided in the embodiments of this disclosure;

[0059] Figure 6 is a schematic diagram of the intelligent network functions included in different aspects of the communication network architecture shown in Figure 5;

[0060] Figure 7 illustrates a communication network architecture based on wireless access network control plane service provisioning according to an embodiment of this disclosure.

[0061] Figure 8 illustrates a communication network architecture based on a 5G communication network that integrates the functions of the wireless access network and the core network, according to an embodiment of this disclosure.

[0062] Figure 9 is a schematic diagram of the AI ​​model training process based on the communication network architecture shown in Figure 4;

[0063] Figure 10 is a schematic diagram of the AI ​​model training process based on the communication network architecture shown in Figure 6;

[0064] Figure 11 is a schematic diagram of the AI ​​model training process based on the communication network architecture shown in Figure 7;

[0065] Figure 12 is a schematic diagram of the AI ​​model training process based on the communication network architecture shown in Figure 8. Detailed Implementation

[0066] To better understand the above-described objectives, features, and advantages of this disclosure, the present disclosure will be described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the described embodiments are only some, not all, of the embodiments of this disclosure. The specific embodiments described herein are merely for explaining this disclosure and are not intended to limit it. All other embodiments obtained by those skilled in the art based on the described embodiments of this disclosure are within the scope of protection of this disclosure.

[0067] It should be noted that in this article, relational terms such as “first” and “second” are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0068] Figure 1 is a schematic diagram of a communication network architecture provided in an embodiment of this disclosure, which is applicable to 6G communication networks. As shown in Figure 1, the communication network architecture includes a radio access network and a core network. The radio access network retains the radio access network capabilities of the 5G network architecture. The core network retains the core network capabilities of the 5G network architecture. The core network includes intelligent network functions in the data plane and intelligent network functions in the intelligent plane. The data plane may include multiple intelligent network functions, and the intelligent plane may include multiple intelligent network functions. Intelligent network functions refer to network functions (NFs) that can participate in or assist in the processing of artificial intelligence (AI) services, so as to enhance the AI ​​processing capabilities of the core network.

[0069] The radio access network (RAN) receives intelligent service requests and transmits the intelligent service information carried in the requests to the intelligent network function of the intelligent plane. The intelligent service request is initiated by the terminal. The RAN can parse the intelligent service request to obtain the intelligent service information carried in the request and then transmit the intelligent service information to the intelligent network function of the intelligent plane.

[0070] The intelligent network function of the intelligent plane is used to obtain the data required for intelligent service information from the intelligent network function of the data plane, determine the intelligent service result based on the required data, and send the intelligent service response through the radio access network. The intelligent service response is sent to the terminal by the radio access network. The intelligent service response indicates that the intelligent service request has been completed and / or carries the intelligent service result. The intelligent network function of the intelligent plane can determine the intelligent service result based on the data required for intelligent service information, using conventional AI algorithms or self-developed AI algorithms.

[0071] As can be seen, the communication network architecture of this embodiment, based on the user plane and control plane of the 5G network architecture, includes a core network comprising a data plane and an intelligent plane. The radio access network can receive intelligent service requests sent by terminals and send the intelligent service information carried in the intelligent service requests to the intelligent network function of the intelligent plane. The intelligent network function of the intelligent plane can obtain the data required for intelligent service information from the intelligent network function of the data plane. The intelligent network function of the intelligent plane uses the data required for intelligent service information to determine the intelligent service result and sends the intelligent service result to the terminal through the radio access network. This realizes the interaction and functional collaboration between planes in the communication network architecture, enhances the AI ​​processing capability of the core network, and improves the feasibility of the layered and planened architecture of the 6G communication network.

[0072] In some embodiments, the interaction between the intelligent network function of the data plane and the intelligent network function of the intelligent plane includes at least one of the following:

[0073] Method 1: Point-to-point interface interaction.

[0074] For example, the intelligent network function of the intelligent plane sends data requests directly to the intelligent network function of the data plane using a point-to-point interface, while the intelligent network function of the data plane sends data responses directly to the intelligent network function of the intelligent plane using a point-to-point interface.

[0075] Method 2: Service-based interface interaction.

[0076] For example, the intelligent network function of the intelligent plane uses a service-oriented interface to send data requests to the intelligent network function of the data plane, and the intelligent network function of the data plane uses a service-oriented interface to send data responses to the intelligent network function of the intelligent plane.

[0077] Method 3: Interaction is based on anchor point functions, with different anchor point functions set for the data plane and the intelligence plane. Different planes have different anchor point functions for external interaction, and different intelligent network functions within the same plane need to use the anchor point functions of that plane to interact with externally.

[0078] For example, the intelligent network function of the intelligent plane uses the anchor point function of the intelligent plane to send data requests to the intelligent network function of the data plane. In turn, the intelligent network function of the data plane uses the anchor point function of the data plane to send data responses to the intelligent network function of the intelligent plane.

[0079] Based on the communication network architecture shown in Figure 1, Figure 2 is a schematic diagram of intelligent network functions included in different aspects according to an embodiment of this disclosure. In Figure 2, the intelligent network functions of the intelligent aspect include at least one of the following functions:

[0080] The system includes intelligent orchestration management function AIMOF, intelligent model training function AIMTF, intelligent model validation function AIMAF, intelligent model registration function AIMRF, and intelligent model storage function AIMSF.

[0081] The Intelligent Orchestration and Management (AIMOF) function is used to decompose AI business into at least one AI task based on intelligent business information, and map at least one AI task to corresponding requirement information. Requirement information includes, but is not limited to, at least one of the following: data requirements, algorithm requirements, computing power requirements, and AI model requirements.

[0082] The Intelligent Model Training Function (AIMTF) is used to obtain the training data needed to train the AI ​​model from the intelligent network function on the data plane, and then train the AI ​​model based on the training data to obtain the AI ​​model. For example, in Figure 2, the AIMTF obtains the training data needed to train the AI ​​model from the data management function (DRF) on the data plane. The AI ​​model training algorithm can use existing conventional model training algorithms or a self-developed model training algorithm.

[0083] The Intelligent Model Validation Function (AIMAF) is used to validate AI models trained by the Intelligent Model Training Function (AIMTF). Validation includes, but is not limited to, security validation and performance validation, and can utilize existing conventional model validation methods or self-developed model validation methods.

[0084] The Intelligent Model Registration Function (AIMRF) is used to register AI models that have passed the Intelligent Model Validation Function (AIMAF) to the network. When 6G communication networks and users need to use AI models, they can select the registered AI models.

[0085] The AIMSF (Intelligent Model Storage Function) is used to store AI models registered by the AIMRF (Intelligent Model Registration Function).

[0086] In Figure 2, the intelligent network functionality of the data plane includes at least one of the following functions:

[0087] Data Management Function (DRF), Data Orchestration Function (DOF), Data Control Function (DCF), and Data Agent (DA).

[0088] The Data Management Function (DRF) manages terminal and network data. Network data includes radio access network (RAN) data and core network data. The DRF supports data acquisition, transmission, storage, and collaboration, resolving the long-standing challenges of migrating and securely iterating network-wide data (including terminal and network data), and providing trusted data services.

[0089] The Data Orchestration Function (DOF) receives data service requests, converts them into corresponding data pipeline construction requests, and sends these requests to the Data Control Function (DCF). The DOF handles non-real-time data orchestration. The DOF can also collaborate with other network services; for example, it assists the computing power network service in data orchestration during computing power orchestration. Furthermore, the DOF selects a data proxy DA to provide data service capabilities based on the data service request and the data service capabilities reported by at least one data proxy DA.

[0090] The Data Control Function (DCF) is used to build data pipelines based on data pipeline construction requests. The DCF is responsible for real-time orchestration tasks, combining data pipelines within the local domain based on the capabilities of Data Agents (DAs) and data service requests. The Data Orchestration Function (DOF) and the DCF work together to achieve elasticity and programmability in the data pipeline. Additionally, the DCF is used to register or deregister Data Agents (DAs) based on the data service capabilities reported by the DAs. The DCF provides real-time monitoring of DAs by detecting their heartbeats. The DCF can also be used as a storage extension component for large-scale or long-term data storage, serving as a storage extension component for Data Agents (DAs).

[0091] The Data Agent (DA) provides data service capabilities to the data pipeline built by the Data Control Function (DCF). The DA can perform data services such as data acquisition, data preprocessing, data storage, data analysis, and data sharing within the orchestrated data pipeline. Data storage is responsible for the local storage of small amounts of data, short-term data, or data with privacy protection requirements; it can be built into the network function or deployed separately. The DA also reports data service capabilities to the Data Orchestration Function (DOF) and the Data Control Function (DCF).

[0092] In Figure 2, the communication network architecture also includes network functions on the user plane and network functions on the control plane.

[0093] Among them, the network functions of the user plane include the Intelligent User Plane Function (iUPF). In addition to the existing UPF functions of 5G communication networks, the iUPF also adds the original related functions of the Network Data Analytics Function (NWDAF), such as communication analysis (related to traffic routing processing), user data congestion analysis, wireless local area network (WLAN) performance analysis, packet flow description (PFD) measurement and analysis, end-to-end data transmission time analysis, and protocol data unit (PDU) session traffic analysis.

[0094] The network functions of the control plane include at least one of the following:

[0095] The functions include Intelligent Network Repository Function (iNRF), Intelligent Policy Control Function (iPCF), Intelligent Network Openness Function (iNEF), Intelligent Access and Mobility Management Function (iAMF), Intelligent Session Management Function (iSMF), Intelligent Unified Data Management Function (iUDM), and Intelligent Application Function (AIAF).

[0096] The Intelligent Network Repository Function (iNRF) not only has the existing NRF functions of 5G communication networks, but also adds the network data analysis function (NWDAF) and related functions, such as network element load analysis.

[0097] The Intelligent Policy Control Function (iPCF) not only has the existing PCF functions of 5G communication networks, but also adds the original related functions of the Network Data Analysis Function (NWDAF), such as network data analysis related to abnormal behavior.

[0098] The Intelligent Network Open Function (iNEF) not only has the existing NEF functions of 5G communication networks, but also adds the original related functions of Network Data Analysis Function (NWDAF), such as performance analysis of data network DN (e.g., operator services, Internet or third-party services).

[0099] The Intelligent Access and Mobility Management Function (iAMF) not only has the existing AMF functions of 5G communication networks, but also adds the original related functions of the Network Data Analysis Function (NWDAF), such as network data related to expected terminal behavior parameters, terminal mobility analysis, communication analysis (mobility management related), terminal dispersion analysis, positioning accuracy analysis, relative proximity analysis, and mobile behavior analysis.

[0100] The Intelligent Session Management Function (iSMF) not only has the existing SMF functions of 5G communication networks, but also adds the original related functions of the Network Data Analysis Function (NWDAF), such as communication analysis (related to traffic routing processing), session management congestion control experience analysis, and related analysis with redundant transmission experience.

[0101] The Intelligent Unified Data Management (iUDM) function, in addition to the existing UDM functions of 5G communication networks, also adds the original related functions of the Network Data Analysis (NWDAF) function, such as abnormal behavior-related network data analysis (identifying a group of UEs with abnormal behavior (such as being abused or hijacked)).

[0102] The Intelligent Application Function (AIAF) is used to pull the required AI models from the Intelligent Model Training Function (AIMTF) of the Intelligent Surface, and to inject the AI ​​models and model parameters; to perform distributed inference based on the injected models and to generate control parameters; to perform joint learning, joint inference, and joint generation with the Intelligent Model Training Function (AIMTF); and to collect and provide various parameters to the Intelligent Model Training Function (AIMTF). Internally, the AIAF generates various control parameters, algorithms, and policies, and provides various control and collaboration functions such as Radio Resource Management (RRM).

[0103] In Figure 2, the Radio Access Network (RAN) of the communication network architecture includes Distributed Units (DU), Centralized Unit-User Plane (CU-UP), Centralized Unit-Control Plane (CU-CP), and Intelligent Application Functions (AIAF). DU, CU-UP, and CU-CP possess the existing functions of DU, CU-UP, and CU-CP in 5G communication networks. The AIAF can obtain intelligent service requests sent by terminals from DU or CU-UP.

[0104] Figure 3 is a schematic diagram of another communication network architecture provided in an embodiment of this disclosure. The difference from Figure 1 is that in this embodiment, the core network also includes intelligent network functions of the computing plane, which is used to provide computing power services.

[0105] In Figure 3, the intelligent network function of the intelligent plane is also used to send a computing power request to the intelligent network function of the computing plane when determining the intelligent service result based on the data required for intelligent service information, if the computing power is insufficient. The intelligent network function of the computing plane is used to perform computing power scheduling and interconnection based on the computing power request, and to perform intelligent service calculation based on the scheduled and interconnected computing power to obtain the intelligent service result.

[0106] In Figure 3, the interaction methods between the intelligent network functions of the data plane, the intelligent network functions of the intelligent plane, and the intelligent network functions of the computing plane include at least one of the following:

[0107] Method 1: Point-to-point interface interaction.

[0108] For example, the intelligent network function of the intelligent plane sends computing power requests directly to the intelligent network function of the computing plane through a point-to-point interface. The intelligent network function of the computing plane then performs computing power scheduling and interconnection based on the computing power requests, performs intelligent business calculations based on the scheduled and interconnected computing power, obtains intelligent business results, and sends the intelligent business results directly to the intelligent network function of the intelligent plane through the point-to-point interface.

[0109] Method 2: Service-based interface interaction.

[0110] For example, the intelligent network function of the intelligent plane uses a service-oriented interface to send computing power requests to the intelligent network function of the computing plane. The intelligent network function of the computing plane then performs computing power scheduling and interconnection based on the computing power requests, performs intelligent service calculations based on the scheduled and interconnected computing power, obtains intelligent service results, and sends the intelligent service results back to the intelligent network function of the intelligent plane through the service-oriented interface.

[0111] Method 3: Interaction is based on anchor point functions, with different anchor point functions set for the data plane, intelligence plane, and computation plane. Different planes have different anchor point functions for external interaction, and different intelligent network functions within the same plane need to use the anchor point functions of that plane to interact with externally.

[0112] For example, the intelligent network function of the intelligent plane uses the anchor point function of the intelligent plane to send computing power requests to the intelligent network function of the computing plane. The intelligent network function of the computing plane then performs computing power scheduling and interconnection based on the computing power requests, and performs intelligent service calculations based on the scheduled and interconnected computing power to obtain intelligent service results, and sends the intelligent service results to the intelligent network function of the intelligent plane through the anchor point function of the computing plane.

[0113] Based on the communication network architecture shown in Figure 3, Figure 4 is a schematic diagram of intelligent network functions included in different planes according to an embodiment of this disclosure. The difference from Figure 2 is that, in Figure 4, intelligent network functions for the computing plane are added, and these intelligent network functions include at least one of the following functions:

[0114] The Computing Power Control Function (CPCF) is used to parse the computing power demand information corresponding to the computing power request, perform computing power orchestration based on the computing power demand information, obtain computing power policy information and computing power routing information, and send the computing power policy information and computing power routing information to the Computing Interconnect Function (CPIF). Computing power orchestration, for example, involves orchestrating at least one Computing Execution Function (CPEF).

[0115] The Computing Power Interconnection Function (CPIF) is used to schedule and interconnect at least one Computing Execution Function (CPEF) based on computing power policy information and computing power routing information, thereby opening up computing power channels.

[0116] The Computing Power Execution Function (CPEF) is used for scheduling and interconnection based on the Computing Power Interconnect Function (CPIF) to perform intelligent business computation and obtain intelligent business results.

[0117] In Figure 4, the intelligent model training function AIMTF of the intelligent surface is also used to send a computing power request to the intelligent network function (computation control function CPCF) of the computing surface if the computing power is insufficient, so that the computing surface can provide computing power to realize intelligent business computing.

[0118] Figure 5 is a schematic diagram of another communication network architecture provided by an embodiment of the present disclosure. The difference from Figure 1 is that in this embodiment, the core network also includes the intelligent network function of the orchestration management plane. The orchestration management plane summarizes the orchestration management functions of each plane in the communication network architecture shown in Figure 4 into the orchestration management plane.

[0119] In Figure 5, the radio access network is used to send intelligent service requests and / or computing power requests to the intelligent network function of the orchestration management plane. The intelligent network function of the orchestration management plane is used to parse the intelligent service information carried in the intelligent service request and send the intelligent service information to the intelligent network function of the intelligent plane; and / or, to parse the computing power demand information corresponding to the computing power request, perform computing power orchestration based on the computing power demand information to obtain computing power policy information and computing power routing information, and send the computing power policy information and computing power routing information to the intelligent network function of the computing plane.

[0120] Based on the communication network architecture shown in Figure 5, Figure 6 is a schematic diagram of intelligent network functions included in different aspects according to an embodiment of this disclosure. The difference from Figure 4 is that, in Figure 6, an intelligent network function for the orchestration management surface is added. The intelligent network function for the orchestration management surface includes at least one of the following functions:

[0121] The functions include Computational Control Function (CPCF), Data Orchestration Function (DOF), Service Requirement Adaptation Function (SRAF), and Intelligent Orchestration Management Function (AIMOF).

[0122] Specifically, the Computational Control Function (CPCF) is moved from the Computational Plane to the Orchestration Management Plane, the Data Orchestration Function (DOF) is moved from the Data Plane to the Orchestration Management Plane, and the Intelligent Orchestration Management Function (AIMOF) is moved from the Intelligent Plane to the Orchestration Management Plane. This consolidates the orchestration management functions of each plane in the communication network architecture shown in Figure 4 into the Orchestration Management Plane. Simultaneously, the Computational Control Function (CPCF) in the Computational Plane, the Data Orchestration Function (DOF) in the Data Plane, and the Intelligent Orchestration Management Function (AIMOF) in the Intelligent Plane are removed to avoid functional duplication.

[0123] The Service Request Adaptation (SRAF) function parses the intelligent service information carried in intelligent service requests sent by the radio access network and transmits this information to the intelligent network functions of the intelligent plane. The SRAF can receive intent requests from the service layer, which can originate from the terminal, within the 6G communication network, or from a third party. The SRAF parses these intent requests, converting them into network-understandable requirements, and then orchestrates and adapts the services based on factors such as connectivity, computing power, data, and algorithms.

[0124] Figure 7 shows a communication network architecture based on the service-oriented nature of the radio access network control plane provided in an embodiment of this disclosure. The difference between this architecture and the communication network architecture shown in Figure 4 is that, in Figure 7, the radio access network control plane is service-oriented into two network functions: Radio Connection Management Function (RCCF) and Radio Transmission Management Function (RSMF).

[0125] Radio Connection Management Function (RCCF): Radio connection management function, carrying / supporting / including one or more of the following functions: establishing / modifying / releasing / deleting / reconfiguring access control, radio resource control (RRC), paging, and mobility signaling between the control terminal and the radio access network; controlling the establishment / modification / releasing / deleting of access, paging, and mobility signaling within the radio access network between the radio access network and the core network.

[0126] Radio Transmission Management Function (RSMF): Used to carry / support / include one or more of the following functions: control the establishment / modification / release / deletion of the channel (first channel) between the control terminal and the radio access network for transmitting UE user plane data, and control the establishment / modification / release / deletion of the channel (second channel) between the radio access network and the core network for transmitting UE user plane data in the radio access network.

[0127] Figure 8 illustrates a communication network architecture based on a 5G communication network that integrates radio access network and core network functions, according to an embodiment of this disclosure. In Figure 8, the control plane functions of the 5G communication network's radio access network are functionally reconfigured with the access and mobility management function (AMF) and session management function (SMF) of the core network, respectively. Similarly, the user plane functions of the 5G communication network's radio access network are functionally reconfigured with the user plane functions of the core network. The functional reconfiguration results in four network functions: Public Unit (PU), Access Connection Management Function (ACMF), Session Bearer Management Function (SBMF), and User Plane Service Function (UPSF).

[0128] The common unit (PU) includes the physical layer, media access control layer, radio link control layer, and packet compression protocol (PDCP) layer.

[0129] The Access Connection Management Function (ACMF) includes the connection control function of Radio Resource Control (RRC), mobility management function, handover function, paging function, etc., as well as the access and intelligent access and mobility management functions of the core network (iAMF).

[0130] The Session Bearer Management Function (SBMF) includes functions other than those of Radio Resource Control (RRC) such as connectionless control, mobility management, handover, and paging, as well as the functions of the core network's Intelligent Session Management Function (iSMF).

[0131] The User Plane Service Function (UPSF) includes user data packet processing and forwarding functions of the Service Data Adaptation Protocol (SDAP) of the Radio Access Network, as well as the User Plane Function (UPF) of the core network. The UPS function can also be dynamically and flexibly reconfigured based on the location between the Public Unit (PU) and the Data Network (DN), the UPS capacity, and other scenarios; these are called ordinary UPS and anchor UPS.

[0132] Example 1

[0133] Figure 9 is a flowchart illustrating the AI ​​model training process based on the communication network architecture shown in Figure 4. In Figure 9, the AI ​​model training includes the following steps 1 to 13:

[0134] 1. The terminal needs to retrain the AI ​​model so that when optimizing the AI ​​model, it sends an AI model training request to the Radio Access Network (RAN, which includes CU-CP, CU-UP, and DU). The service type is intelligent service. The AI ​​model training request carries the training parameters required to train the AI ​​model, such as the model type of the AI ​​model to be trained and the data required for training.

[0135] 2. After receiving the AI ​​model training request, the Radio Access Network (RAN) parses the AI ​​model training request, determines that the service type is intelligent service, and selects the Intelligent Orchestration Management Function (AIMOF).

[0136] 3. The Radio Access Network (RAN) sends the AI ​​model training request to the Intelligent Orchestration Management Function (AIMOF). The AI ​​model training request carries the training parameters required to train the AI ​​model.

[0137] 4. The AIMOF intelligent orchestration management function parses AI model training requests, determines that intelligent business is AI model training, and selects the intelligent model training function AIMTF.

[0138] 5. The AIMOF intelligent orchestration management function sends an AI model training request to the AIMTF intelligent model training function. The AI ​​model training request carries the training parameters required to train the AI ​​model.

[0139] 6. The Intelligent Model Training Function (AIMTF) sends a data request to the Data Agent (DA) on the data plane based on the training parameters. The Data Agent (DA) then sends a data response to the AIMTF, which includes the training data required to train the AI ​​model and the AI ​​model's identifier.

[0140] 7. The AIMTF intelligent model training function sends an AI model query request to the AIMRF intelligent model registration function based on the AI ​​model's identifier. The AI ​​model query request carries the AI ​​model's identifier. The AIMRF intelligent model registration function returns an AI model query response, which also carries the AI ​​model's identifier.

[0141] 8. The AIMTF (Intelligent Model Training Function) sends an AI model retrieval request to the AIMSF (Intelligent Model Storage Function), which includes the AI ​​model's identifier. The AIMSF then sends an AI model retrieval response to the AIMTF (Intelligent Model Training Function), which includes the AI ​​model in the response.

[0142] 9. Intelligent Model Training Function: AIMTF retrains the AI ​​model based on the training data.

[0143] 10. The Intelligent Model Training Function (AIMTF) sends the trained AI model to the Intelligent Model Validation Function (AIMAF) for validation. Specifically, the AIMTF sends an AI model validation request to the AIMAF, which includes the trained AI model. The AIMAF then returns an AI model validation response.

[0144] 11. The AIMTF (AI Model Training Function) sends the verified AI model to the AIMRF (AI Model Registration Function) for registration. Specifically, the AIMTF sends an AI model registration request to the AIMRF, which includes the verified AI model. The AIMRF then returns an AI model registration response.

[0145] 12. The Intelligent Model Training Function (AIMTF) sends the registered AI model to the Intelligent Model Storage Function (AIMSF) for storage. Specifically, the AIMTF sends an AI model storage request to the AIMSF, which includes the registered AI model. The AIMSF then returns an AI model storage response.

[0146] 13. The AIMTF intelligent model training function sends an AI model training response to the terminal through the intelligent orchestration management function AIMOF and the radio access network RAN, indicating that the AI ​​model training has been completed.

[0147] Example 2

[0148] Figure 10 is a flowchart illustrating the AI ​​model training process based on the communication network architecture shown in Figure 6. In Figure 10, the AI ​​model training includes the following steps 1 through 13:

[0149] 1. When a terminal needs to retrain its AI model to optimize it, it sends an AI model training request to the Radio Access Network (RAN, which includes CU-CP, CU-UP, and DU). The service type is intelligent service. The AI ​​model training request carries the training parameters required to train the AI ​​model, such as the model type and the data required for training. The RAN then forwards the AI ​​model training request to the Service Requirements Adaptation Function (SRAF) in the orchestration management plane for parsing.

[0150] 2. Business Requirement Adaptation Function: After receiving an AI model training request, SRAF parses the AI ​​model training request, determines that the business type is intelligent business, and selects the intelligent orchestration management function AIMOF.

[0151] 3. Business requirement adaptation function: SRAF sends the AI ​​model training request to the intelligent orchestration management function AIMOF. The AI ​​model training request carries the training parameters required to train the AI ​​model.

[0152] 4. The AIMOF intelligent orchestration management function parses AI model training requests, determines that intelligent business is AI model training, and selects the intelligent model training function AIMTF.

[0153] 5. The AIMOF intelligent orchestration management function sends an AI model training request to the AIMTF intelligent model training function. The AI ​​model training request carries the training parameters required to train the AI ​​model.

[0154] 6. The Intelligent Model Training Function (AIMTF) sends a data request to the Data Agent (DA) on the data plane based on the training parameters. The Data Agent (DA) then sends a data response to the AIMTF, which includes the training data required to train the AI ​​model and the AI ​​model's identifier.

[0155] 7. The AIMTF intelligent model training function sends an AI model query request to the AIMRF intelligent model registration function based on the AI ​​model's identifier. The AI ​​model query request carries the AI ​​model's identifier. The AIMRF intelligent model registration function returns an AI model query response, which also carries the AI ​​model's identifier.

[0156] 8. The AIMTF (Intelligent Model Training Function) sends an AI model retrieval request to the AIMSF (Intelligent Model Storage Function), which includes the AI ​​model's identifier. The AIMSF then sends an AI model retrieval response to the AIMTF (Intelligent Model Training Function), which includes the AI ​​model in the response.

[0157] 9. Intelligent Model Training Function: AIMTF retrains the AI ​​model based on the training data.

[0158] 10. The Intelligent Model Training Function (AIMTF) sends the trained AI model to the Intelligent Model Validation Function (AIMAF) for validation. Specifically, the AIMTF sends an AI model validation request to the AIMAF, which includes the trained AI model. The AIMAF then returns an AI model validation response.

[0159] 11. The AIMTF (AI Model Training Function) sends the verified AI model to the AIMRF (AI Model Registration Function) for registration. Specifically, the AIMTF sends an AI model registration request to the AIMRF, which includes the verified AI model. The AIMRF then returns an AI model registration response.

[0160] 12. The Intelligent Model Training Function (AIMTF) sends the registered AI model to the Intelligent Model Storage Function (AIMSF) for storage. Specifically, the AIMTF sends an AI model storage request to the AIMSF, which includes the registered AI model. The AIMSF then returns an AI model storage response.

[0161] 13. The AIMTF intelligent model training function sends an AI model training response to the terminal through the intelligent orchestration management function AIMOF and the radio access network RAN, indicating that the AI ​​model training has been completed.

[0162] Example 3

[0163] Figure 11 is a schematic diagram of the AI ​​model training process based on the communication network architecture shown in Figure 7. In Figure 11, the AI ​​model training includes the following steps 1 to 13:

[0164] 1. When the terminal needs to retrain the AI ​​model to optimize the AI ​​model, it sends an AI model training request to the Radio Access Network (RAN, which includes RCCF, RSMF, CU-UP, and DU). The service type is intelligent service. The AI ​​model training request carries the training parameters required to train the AI ​​model, such as the model type of the AI ​​model to be trained and the data required for training.

[0165] 2. After receiving the AI ​​model training request, the Radio Access Network (RAN) parses the AI ​​model training request, determines that the service type is intelligent service, and selects the Intelligent Orchestration Management Function (AIMOF).

[0166] 3. The Radio Access Network (RAN) sends the AI ​​model training request to the Intelligent Orchestration Management Function (AIMOF). The AI ​​model training request carries the training parameters required to train the AI ​​model.

[0167] 4. The AIMOF intelligent orchestration management function parses AI model training requests, determines that intelligent business is AI model training, and selects the intelligent model training function AIMTF.

[0168] 5. The AIMOF intelligent orchestration management function sends an AI model training request to the AIMTF intelligent model training function. The AI ​​model training request carries the training parameters required to train the AI ​​model.

[0169] 6. The Intelligent Model Training Function (AIMTF) sends a data request to the Data Agent (DA) on the data plane based on the training parameters. The Data Agent (DA) then sends a data response to the AIMTF, which includes the training data required to train the AI ​​model and the AI ​​model's identifier.

[0170] 7. The AIMTF intelligent model training function sends an AI model query request to the AIMRF intelligent model registration function based on the AI ​​model's identifier. The AI ​​model query request carries the AI ​​model's identifier. The AIMRF intelligent model registration function returns an AI model query response, which also carries the AI ​​model's identifier.

[0171] 8. The AIMTF (Intelligent Model Training Function) sends an AI model retrieval request to the AIMSF (Intelligent Model Storage Function), which includes the AI ​​model's identifier. The AIMSF then sends an AI model retrieval response to the AIMTF (Intelligent Model Training Function), which includes the AI ​​model in the response.

[0172] 9. Intelligent Model Training Function: AIMTF retrains the AI ​​model based on the training data.

[0173] 10. The Intelligent Model Training Function (AIMTF) sends the trained AI model to the Intelligent Model Validation Function (AIMAF) for validation. Specifically, the AIMTF sends an AI model validation request to the AIMAF, which includes the trained AI model. The AIMAF then returns an AI model validation response.

[0174] 11. The AIMTF (AI Model Training Function) sends the verified AI model to the AIMRF (AI Model Registration Function) for registration. Specifically, the AIMTF sends an AI model registration request to the AIMRF, which includes the verified AI model. The AIMRF then returns an AI model registration response.

[0175] 12. The Intelligent Model Training Function (AIMTF) sends the registered AI model to the Intelligent Model Storage Function (AIMSF) for storage. Specifically, the AIMTF sends an AI model storage request to the AIMSF, which includes the registered AI model. The AIMSF then returns an AI model storage response.

[0176] 13. The AIMTF intelligent model training function sends an AI model training response to the terminal through the intelligent orchestration management function AIMOF and the radio access network RAN, indicating that the AI ​​model training has been completed.

[0177] Example 4

[0178] Figure 12 is a flowchart illustrating the AI ​​model training process based on the communication network architecture shown in Figure 8. In Figure 12, the AI ​​model training includes the following steps 1 to 13:

[0179] 1. When the terminal needs to retrain the AI ​​model to optimize the AI ​​model, it sends an AI model training request to the Radio Access Network (RAN, which includes the Public Unit, PU). The service type is intelligent service. The AI ​​model training request carries the training parameters required to train the AI ​​model, such as the model type of the AI ​​model to be trained and the data required for training.

[0180] 2. After receiving the AI ​​model training request, the Public Unit (PU) sends the AI ​​model training request to the Intelligent Bearer Management Function (SBMF). The Intelligent Bearer Management Function (SBMF) parses the AI ​​model training request, determines that the business type is intelligent business, and selects the Intelligent Orchestration Management Function (AIMOF).

[0181] 3. The Intelligent Host Management Function (SBMF) sends the AI ​​model training request to the Intelligent Orchestration Management Function (AIMOF). The AI ​​model training request carries the training parameters required to train the AI ​​model.

[0182] 4. The AIMOF intelligent orchestration management function parses AI model training requests, determines that intelligent business is AI model training, and selects the intelligent model training function AIMTF.

[0183] 5. The AIMOF intelligent orchestration management function sends an AI model training request to the AIMTF intelligent model training function. The AI ​​model training request carries the training parameters required to train the AI ​​model.

[0184] 6. The Intelligent Model Training Function (AIMTF) sends a data request to the Data Agent (DA) on the data plane based on the training parameters. The Data Agent (DA) then sends a data response to the AIMTF, which includes the training data required to train the AI ​​model and the AI ​​model's identifier.

[0185] 7. The AIMTF intelligent model training function sends an AI model query request to the AIMRF intelligent model registration function based on the AI ​​model's identifier. The AI ​​model query request carries the AI ​​model's identifier. The AIMRF intelligent model registration function returns an AI model query response, which also carries the AI ​​model's identifier.

[0186] 8. The AIMTF (Intelligent Model Training Function) sends an AI model retrieval request to the AIMSF (Intelligent Model Storage Function), which includes the AI ​​model's identifier. The AIMSF then sends an AI model retrieval response to the AIMTF (Intelligent Model Training Function), which includes the AI ​​model in the response.

[0187] 9. Intelligent Model Training Function: AIMTF retrains the AI ​​model based on the training data.

[0188] 10. The Intelligent Model Training Function (AIMTF) sends the trained AI model to the Intelligent Model Validation Function (AIMAF) for validation. Specifically, the AIMTF sends an AI model validation request to the AIMAF, which includes the trained AI model. The AIMAF then returns an AI model validation response.

[0189] 11. The AIMTF (AI Model Training Function) sends the verified AI model to the AIMRF (AI Model Registration Function) for registration. Specifically, the AIMTF sends an AI model registration request to the AIMRF, which includes the verified AI model. The AIMRF then returns an AI model registration response.

[0190] 12. The Intelligent Model Training Function (AIMTF) sends the registered AI model to the Intelligent Model Storage Function (AIMSF) for storage. Specifically, the AIMTF sends an AI model storage request to the AIMSF, which includes the registered AI model. The AIMSF then returns an AI model storage response.

[0191] 13. The AIMTF intelligent model training function sends an AI model training response to the terminal through the intelligent orchestration management function AIMOF and the radio access network RAN, indicating that the AI ​​model training has been completed.

[0192] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0193] Those skilled in the art will understand that although some embodiments described herein include certain features included in other embodiments but not others, combinations of features from different embodiments are meant to be within the scope of this disclosure and form different embodiments.

[0194] Those skilled in the art will understand that the descriptions of the various embodiments have different focuses, and for parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0195] Although embodiments of the present disclosure have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present disclosure, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A communication network architecture, the communication network architecture comprising a wireless access network and a core network, the core network comprising intelligent network functions of the data plane and intelligent network functions of the intelligent plane; The wireless access network is used to receive intelligent service requests and send the intelligent service information carried in the intelligent service requests to the intelligent network function of the intelligent plane. The intelligent network function of the intelligent plane is used to obtain the data required for the intelligent service information from the intelligent network function of the data plane, determine the intelligent service result based on the data required for the intelligent service information, and send the intelligent service response through the wireless access network.

2. The communication network architecture according to claim 1, wherein, The interaction between the intelligent network function of the data plane and the intelligent network function of the intelligent plane includes at least one of the following: Point-to-point interface interaction; Service-oriented interface interaction; Interaction is based on anchor point functionality, and different anchor point functions are set for the data plane and the smart plane respectively.

3. The communication network architecture according to claim 1, wherein, The core network also includes intelligent network functions for the computing plane; The intelligent network function of the intelligent plane is also used to send a computing power request to the intelligent network function of the computing plane when determining the intelligent business result based on the data required by the intelligent business information, if the computing power is insufficient. The intelligent network function of the computing plane is used to perform computing power scheduling and interconnection based on the computing power request, and to perform intelligent business calculations based on the scheduled and interconnected computing power to obtain intelligent business results.

4. The communication network architecture according to any one of claims 1 to 3, wherein, The core network also includes intelligent network functions for the orchestration and management plane; The wireless access network is used to send intelligent service requests and / or computing power requests to the intelligent network function of the orchestration management plane; The intelligent network function of the orchestration management plane is used to parse the intelligent service information carried in the intelligent service request and send the intelligent service information to the intelligent network function of the intelligent plane. And / or, used to parse the computing power demand information corresponding to the computing power request, perform computing power orchestration based on the computing power demand information, obtain computing power strategy information and computing power routing information, and send the computing power strategy information and the computing power routing information to the intelligent network function of the computing plane.

5. The communication network architecture according to claim 1, wherein, The intelligent network function of the intelligent surface includes at least one of the following functions: The Intelligent Orchestration and Management Function (AIMOF) is used to decompose AI business into at least one AI task based on the intelligent business information, and map the at least one AI task to the corresponding demand information. The Intelligent Model Training Function (AIMTF) is used to obtain the training data required for training an AI model from the intelligent network function of the data plane, and to train the AI ​​model based on the training data. The Intelligent Model Validation Function (AIMAF) is used to validate the AI ​​model trained by the Intelligent Model Training Function (AIMTF). The Intelligent Model Registration Function (AIMRF) is used to register AI models that have passed the Intelligent Model Validation Function (AIMAF) to the network. The Intelligent Model Storage Function (AIMSF) is used to store AI models registered by the Intelligent Model Registration Function (AIMRF).

6. The communication network architecture according to claim 5, wherein, The intelligent model training function AIMTF is also used to send a computing power request to the intelligent network function of the computing plane if the computing power is insufficient.

7. The communication network architecture according to claim 1 or 5, wherein, The intelligent network function of the data plane includes at least one of the following functions: Data Management Function (DRF) is used to manage terminal data and network data; The Data Orchestration Function (DOF) is used to receive data service requests, convert the data service requests into corresponding data pipeline construction requests, and send the data pipeline construction requests to the Data Control Function (DCF). Data Control Function (DCF) is used to construct a data pipeline based on the data pipeline construction request. The Data Agent (DA) is used to provide data service capabilities for the data in the data pipeline constructed by the Data Control Function (DCF).

8. The communication network architecture according to claim 7, wherein, The data agent DA is also used to report the data service capabilities to the data orchestration function DOF and the data control function DCF; The data orchestration function DOF is also used to select a data proxy DA to provide data service capabilities based on the data service request and the data service capabilities reported by at least one of the data proxy DAs; The Data Control Function (DCF) is also used to register or deregister the Data Agent DA based on the data service capabilities reported by the Data Agent DA.

9. The communication network architecture according to claim 3, wherein, The intelligent network functionality of the computational plane includes at least one of the following functions: The Computation Control Function (CPCF) is used to parse the computing power demand information corresponding to the computing power request, perform computing power orchestration based on the computing power demand information, obtain computing power strategy information and computing power routing information, and send the computing power strategy information and computing power routing information to the Computation Interconnect Function (CPIF). The Compute Interconnect Function (CPIF) is used to schedule and interconnect at least one Compute Execution Function (CPEF) based on the computing power policy information and the computing power routing information. The Computation Execution Function (CPEF) is used to perform intelligent service computation based on the scheduling and interconnection of the Computation Interconnection Function (CPIF) to obtain intelligent service results.

10. The communication network architecture according to claim 4, wherein, The intelligent network function of the orchestration management plane includes at least one of the following functions: Computational Control Function (CPCF); Data orchestration function DOF; The Service Requirement Adaptation Function (SRAF) is used to parse the intelligent service information carried in the intelligent service request sent by the radio access network and send the intelligent service information to the intelligent network function of the intelligent plane. AIMOF (Intelligent Orchestration and Management Function) The functions that are identical to those in the intelligent network functions of the orchestration management plane among the intelligent network functions of the data plane, the intelligent network functions of the intelligent plane, and the intelligent network functions of the computing plane are deleted.

11. The communication network architecture according to claim 5, wherein, The AI ​​task is an AI model training task; The intelligent orchestration management function AIMOF sends an AI model training request to the intelligent model training function AIMTF, and the AI ​​model training request carries the training parameters required to train the AI ​​model. The AIMTF intelligent model training function obtains the training data and AI model identifier required for training the AI ​​model from the intelligent network function of the data plane based on the training parameters. The AIMTF (AI Model Training Function) sends an AI model retrieval request to the AIMSF (AI Model Storage Function) based on the identifier of the AI ​​model. The AIMSF then sends an AI model retrieval response to the AIMTF, which includes the AI ​​model. The AIMTF intelligent model training function retrains the AI ​​model based on the training data; The AIMTF intelligent model training function sends the trained AI model to the AIMAF intelligent model verification function for verification. The AIMTF intelligent model training function will send the verified AI model to the AIMRF intelligent model registration function for registration. The AIMTF intelligent model training function sends the registered AI model to the AIMSF intelligent model storage function for storage.

Images