Artificial intelligence model management method and related equipment

Abstract

The invention provides an artificial intelligence model management method and related equipment, and relates to the technical field of communication, and the method comprises the steps: receiving an artificial intelligence AI model transaction list sent by a cloud node, determining the online transaction budget of an AI model based on a wireless access network intelligent layer, completing the online transaction of the AI model with the cloud node, and obtaining the right to use the AI model, an online AI service list is established through a wireless access network intelligent layer, first information is sent, AI service is provided for a user node based on the first information, and the first information is used for sending the online AI service list to the user node. By means of the mode, the life cycle management process of the AI model is provided and completed, AI model scheduling and transmission across network domains are supported, it is ensured that the AI model is converted from a plug-in type to an endogenous type, the wireless access network intelligent layer supports upward completion of AI model online transaction with a cloud service provider and online management of the AI model, and the AI model management efficiency is improved. And the online AI service can be provided for the user node in a downward manner, and the AI service can be managed in an online manner.

Description

Background Art

[0002] In the existing research on wireless network intelligence, most artificial intelligence (AI) models are "plug-in" and deployed into the radio access network to solve use-case level application tasks, such as beam management and positioning enhancement use cases in 3GPP. When other problems need to be solved, a model needs to be retrained and redeployed.

[0003] It should be noted that the information disclosed in the above background art is only used to enhance the understanding of the background of the present disclosure, and thus may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0004] The present disclosure provides an artificial intelligence model management method and related devices, which at least to a certain extent change the management and use methods of AI models from "plug-in" to "endogenous", and provide a complete AI model management process.

[0005] Other features and advantages of the present disclosure will become apparent through the following detailed description, or will be learned in part through the practice of the present disclosure.

[0006] In a first aspect, an embodiment in the present disclosure provides an artificial intelligence model management method, which is applied to a network node. The method includes:

[0007] Receiving an artificial intelligence (AI) model transaction list sent by a cloud node;

[0008] Determining an online transaction budget for the AI model based on the intelligent layer of the radio access network;

[0009] Completing an online transaction of the AI model with the cloud node according to the AI model transaction list and the online transaction budget for the AI model, so that the cloud node issues the right to use the AI model to the network node;

[0010] Based on the received right to use the AI model, establishing an online AI service list through the intelligent layer of the radio access network;

[0011] Sending a first message, and providing an AI service for a user node based on the first message; the first message is used to send the online AI service list to the user node.

[0012] In a second aspect, an embodiment in the present disclosure provides an artificial intelligence model management method, including:

[0013] Establishing an online transaction mechanism for the AI model;

[0014] Sending an AI model transaction list to a network node; the AI model transaction list includes: AI models available at the cloud node;

[0015] Complete an online transaction of an AI model with a network node;

[0016] Send the right to use the AI model to the network node.

[0017] In a third aspect, an embodiment in the present disclosure provides an artificial intelligence model management method, including:

[0018] Receive the first information sent by the network node; the first information is used to send an online AI service list to the user node; the online AI service list is established by the network node completing an online transaction of an AI model with the cloud node according to the AI model transaction list and the online transaction budget of the AI model, so that the cloud node issues the right to use the AI model to the network node;

[0019] Send the third information to the user node; the third information includes: the online AI service list.

[0020] In a fourth aspect, an embodiment in the present disclosure provides an artificial intelligence model management method, including:

[0021] Receive the third information; the third information includes: the online AI service list; the online AI service list is established by the network node completing an online transaction of an AI model with the cloud node according to the AI model transaction list and the online transaction budget of the AI model, so that the cloud node issues the right to use the AI model to the network node.

[0022] In a fifth aspect, an embodiment in the present disclosure provides a network node, including:

[0023] A first communication module, configured to receive the artificial intelligence AI model transaction list sent by the cloud node;

[0024] A first model management module, configured to determine the online transaction budget of the AI model based on the intelligent layer of the radio access network;

[0025] A first model management module, configured to complete an online transaction of an AI model with the cloud node according to the AI model transaction list and the online transaction budget of the AI model, so that the cloud node issues the right to use the AI model to the network node;

[0026] A first model management module, configured to establish an online AI service list through the intelligent layer of the radio access network based on the right to use the AI model;

[0027] A first communication module, configured to send the first information to provide an AI service for the user node based on the first information; the first information is used to send an online AI service list to the user node.

[0028] In a sixth aspect, an embodiment in the present disclosure provides a cloud node, including:

[0029] The second model management module is used to establish an online trading mechanism for AI models;

[0030] The second communication module is used to send an AI model trading list to a network node; the AI model trading list includes: the AI models available on the cloud node;

[0031] The second model management module is also used to complete the online trading of AI models with the network node;

[0032] The second communication module is also used to issue the right to use the AI model to the network node.

[0033] In a seventh aspect, an embodiment in the present disclosure provides a transmission node, including:

[0034] The third communication module is used to receive the first information sent by the network node; the first information is used to send an online AI service list to the user node; the online AI service list is established by the network node completing the online trading of AI models with the cloud node according to the AI model trading list and the online trading budget of the AI models, so that the cloud node issues the right to use the AI model to the network node;

[0035] The third communication module is also used to send the third information to the user node; the third information includes: the online AI service list.

[0036] In an eighth aspect, an embodiment in the present disclosure provides a user node, including:

[0037] The fourth communication module is used to receive the third information; the third information includes: the online AI service list; the online AI service list is established by the network node completing the online trading of AI models with the cloud node according to the AI model trading list and the online trading budget of the AI models, so that the cloud node issues the right to use the AI model to the network node.

[0038] In a ninth aspect, an embodiment in the present disclosure provides a communication system, including: the network node as in the fifth aspect, the cloud node as in the sixth aspect, the transmission node as in the seventh aspect, and the user node as in the eighth aspect.

[0039] In a tenth aspect, an embodiment of the present disclosure provides an electronic device, including: a processor; and a memory for storing executable instructions of the processor; wherein, the processor is configured to execute the method in the first aspect above by executing the executable instructions.

[0040] In an eleventh aspect, an embodiment of the present disclosure provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the method in the first aspect above is implemented.

[0041] In a twelfth aspect, according to another aspect of the present disclosure, there is also provided a computer program product or a computer program. The computer program product or the computer program includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method of any one of the above.

[0042] An artificial intelligence model management method and related devices provided by an embodiment of the present disclosure relate to the field of communication technologies. The method includes: receiving an artificial intelligence (AI) model transaction list sent by a cloud node, determining an online transaction budget for the AI model based on an intelligent layer of a radio access network, completing an online transaction of the AI model with the cloud node according to the AI model transaction list and the online transaction budget for the AI model, so that the cloud node issues the right to use the AI model to a network node, and based on the right to use the AI model, establishing an online AI service list through the intelligent layer of the radio access network and sending a first message; the first message is used to send the online AI service list to a user node. By the above method, a complete AI service process is provided, which can complete the life cycle management process of the AI model, and supports the scheduling and transmission of the AI model across network domains, ensuring the transformation of the AI model from plug-in type to endogenous type. The intelligent layer of the radio access network supports "upward" to complete the online transaction of the AI model with a cloud service provider and online management of the AI model, and also supports "downward" to provide an online AI service for a user node and online management of the AI service.

[0043] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings herein are incorporated into the specification and form a part of the specification, showing embodiments consistent with the present disclosure, and are used together with the specification to explain the principles of the present disclosure. Obviously, the accompanying drawings in the following description are only some embodiments of the present disclosure, and those of ordinary skill in the art can obtain other drawings without creative efforts based on these drawings.

[0045] Figure 1 A schematic structural diagram of a communication system in an embodiment of the present disclosure is shown;

[0046] Figure 2 An interaction diagram of an artificial intelligence model management method in an embodiment of the present disclosure is shown;

[0047] Figure 3 An interaction diagram of an artificial intelligence model management method in an embodiment of the present disclosure is shown;

[0048] Figure 4One of the flowcharts showing an artificial intelligence model management method in an embodiment of the present disclosure;

[0049] Figure 5 Two of the flowcharts showing an artificial intelligence model management method in an embodiment of the present disclosure;

[0050] Figure 6 Three of the flowcharts showing an artificial intelligence model management method in an embodiment of the present disclosure;

[0051] Figure 7 Four of the flowcharts showing an artificial intelligence model management method in an embodiment of the present disclosure;

[0052] Figure 8 Schematic diagram of the structure of a network node in an embodiment of the present disclosure;

[0053] Figure 9 Schematic diagram of the structure of a cloud node in an embodiment of the present disclosure;

[0054] Figure 10 Schematic diagram of the structure of a transmission node in an embodiment of the present disclosure;

[0055] Figure 11 Schematic diagram of the structure of a user node in an embodiment of the present disclosure;

[0056] Figure 12 Schematic diagram of the structure of an electronic device in an embodiment of the present disclosure. Detailed implementation manners

[0057] Example embodiments will now be described more fully with reference to the accompanying drawings. However, the example embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be more complete and comprehensive, and will fully convey the concept of the example embodiments to those skilled in the art. The features, structures, or characteristics described may be combined in any suitable manner in one or more embodiments.

[0058] In addition, the accompanying drawings are only schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0059] Explanation of the terms in the present disclosure:

[0060] AI: Artificial Intelligence, Artificial Intelligence.

[0061] 6G: 6th Generation Mobile Communication Technology, 6th Generation Mobile Communication Technology.

[0062] RAN: Radio Access Network, Radio Access Network.

[0063] UE: User Equipment, User Equipment.

[0064] LCM: Life Cycle Management, Life Cycle Management.

[0065] OAM: Orchestration and Management, Orchestration and Management.

[0066] CU: Centralized Unit, Centralized Unit.

[0067] DU: Distributed Unit, Distributed Unit.

[0068] RU: Radio Unit, Radio Unit.

[0069] NWDAF: Network Data Analytics Function, Network Data Analytics Function.

[0070] OTT Server: Over-the-top Server, Over-the-top Server.

[0071] SIB: System Information Broadcast, System Information Broadcast.

[0072] UPF: User Plane Function, User Plane Function.

[0073] QoS: Quality of Service, Quality of Service.

[0074] RRC: Radio Resource Control, Radio Resource Control.

[0075] DCI: Downlink Control Information, Downlink Control Information.

[0076] UCI: Uplink Control Information, Uplink Control Information.

[0077] MAC CE: MAC Control Element, the media access control layer control unit.

[0078] OAM: Orchestration and Management, the process of arranging and managing.

[0079] The radio access network is a key component of modern telecommunication systems, typically consisting of multiple base stations that use radio frequency to manage wireless communication with devices within a specified geographical area. Traditional radio access networks are managed using static configurations, and engineers manually optimize network performance based on predefined parameters and traffic patterns. However, with the explosive growth of mobile data traffic and the proliferation of connected devices brought about by 5G technology, the complexity of managing radio access networks has increased significantly, giving rise to the need for intelligent systems that can automate and optimize RAN operations in real time, and spawning the concept of wireless network intelligence, which uses artificial intelligence (AI) technology to improve the efficiency, performance, and adaptability of radio access networks.

[0080] Wireless network intelligence generally deploys AI models on edge nodes close to user terminals. On the one hand, it can collect data from various network nodes (including base stations, user equipment, and core networks), and apply AI models to process and analyze this data in real time to predict network conditions, optimize resource allocation, reduce the latency of data transmission to the cloud, and improve the quality of service for users. On the other hand, wireless network intelligence supports the deployment of a variety of emerging applications, such as virtual reality, augmented reality, and autonomous driving technologies, which can bring stronger scalability and adaptability to the network. Currently, wireless network intelligence is widely used in fields such as smart homes, autonomous driving, and industrial automation, and can achieve functions such as real-time monitoring, anomaly detection, and predictive maintenance.

[0081] In the research on 6G-oriented technologies, such as industrial Internet, autonomous driving, and intelligent healthcare, existing work has increasingly focused on online scenarios, where data typically arrives in the form of data streams. At each time node, it is difficult to obtain a globally optimal solution because future changes cannot be perceived. In addition, with the increase in the number of user terminals at the network edge and the growing demand for AI services, wireless network intelligence faces multiple challenges such as limited computing and communication resources and unguaranteed data security. How to support online AI model management and orchestration strategies in radio access networks has become a key issue.

[0082] In the existing research on wireless network intelligence, most artificial intelligence (AI) models are "plug-in" and deployed into the radio access network to solve use-case-level application tasks, such as beam management and positioning enhancement use cases in 3GPP. When other problems need to be solved, a model needs to be retrained and deployed again.

[0083] Based on the inventors' thinking on this, in the embodiments of the present disclosure, an artificial intelligence model management method and related devices are provided, which relate to the field of communication technologies. The method includes: receiving an artificial intelligence (AI) model transaction list sent by a cloud node, determining the online transaction budget of the AI model based on the intelligent layer of the radio access network, completing the online transaction of the AI model with the cloud node, obtaining the right to use the AI model, establishing an online AI service list through the intelligent layer of the radio access network, and sending a first message for sending the online AI service list to a user node. Through the above method, the life cycle management process of the AI model is provided and completed, and cross-network-domain AI model scheduling and transmission are supported, ensuring the transformation of the AI model from plug-in to endogenous. Under the cloud-edge-end collaborative architecture, the intelligent layer of the radio access network supports "upward" online transactions of the AI model with cloud service providers and online management of the AI model, and also supports "downward" providing online AI services to user nodes and online management of AI services.

[0084] Figure 1 The structural schematic diagram of a communication system in the embodiments of the present disclosure is shown, as Figure 1 shown, the communication system 100 may include: a cloud node 101, a network node 102, a transmission node 103, and a user node 104.

[0085] The cloud node 101 may communicate with the network node 102 through a private interface, the network node 102 may communicate with the transmission node 103 through an NG interface, and the transmission node 103 may communicate with the user node 104 through a wireless air interface. Among them, there may be multiple user nodes 104. Taking a UE as an example, it may include: UE1, UE2, and UE3.

[0086] The user node 104 may be a user equipment node, including enterprise users, individual users, etc.

[0087] The transmission node 103 may be a base station equipment node located at the network edge, including CU / DU / RU, and the CU includes CU-UP (user plane) and CU-CP (control plane).

[0088] The network node 102 may be a node deployed by an operator on the radio access network side and may include at least one of the following: OAM, OTT Server, NWDAF, etc., which are network functions or entities for network management and act as communication service providers.

[0089] The cloud node 101 can be a node deployed by a cloud service provider on the core network cloud server, serving as an AI model provider and a cloud computing service provider.

[0090] Among them, the Radio Access Network AI layer (RAN AI layer) can be deployed on network nodes, which can be at least one of network virtual functions or physical entities, and can be independently deployed for the online unified management and allocation of AI models on the radio access network side.

[0091] It includes at least one of the following functions: online trading of AI models, online management of AI models, online management of AI services, online provision of AI services, and online collection of data.

[0092] Online trading of AI models: The Radio Access Network AI layer formulates an online trading budget for AI models by analyzing network load conditions, environmental information, profitability, AI service information, and the AI service requirements of user nodes.

[0093] Online management of AI models: The Radio Access Network AI layer performs model registration for all AI models / model groups that have completed online transactions and records their basic information; the Radio Access Network AI layer can perform model fine-tuning for AI models / model groups based on the model monitoring information feedback by user nodes.

[0094] Online management of AI services: The Radio Access Network AI layer performs model identification for the registered AI models / model groups, identifies the AI services that the AI models can provide, establishes an online AI service list, is responsible for the identification, configuration, pricing of AI services, and the real-time update of the information of the AI services subscribed by user nodes.

[0095] Online provision of AI services: The Radio Access Network AI layer provides online AI services for user nodes based on the subscription results of user nodes, and deploys and infers AI models.

[0096] Online collection of data: The Radio Access Network AI layer online collects the network data of network nodes themselves, the AI service requirement data, AI service request data, and model monitoring data of user nodes for performing corresponding optimizations.

[0097] The following will explain this exemplary embodiment in detail with reference to the accompanying drawings and embodiments.

[0098] First, Figure 2 One of the interaction diagrams of an artificial intelligence model management method in an embodiment of the present disclosure is shown. As Figure 2 shown, applied to a communication system, the method provided in the embodiment of the present disclosure includes the following steps:

[0099] S202: The cloud node establishes an online trading mechanism for AI models.

[0100] S204: The cloud node sends an AI model transaction list to the network node. The AI model transaction list includes: the AI models available at the cloud node.

[0101] In a possible embodiment, the cloud node and multiple network nodes form an online AI model trading market. The cloud node establishes an online AI model trading mechanism, creates an AI model transaction list based on the AI models available at the cloud node, and announces the trading prices of the AI models in the list.

[0102] S206: The network node determines the online AI model trading budget based on the intelligent layer of the radio access network.

[0103] In a possible embodiment, there may be multiple network nodes. The number of network nodes is not limited in this disclosure. Each network node formulates an online AI model trading budget based on the intelligent layer of the radio access network for purchasing AI models in the AI model trading market.

[0104] S208: The cloud node and the network node complete the online AI model transaction.

[0105] S210: The cloud node issues the right of use of the AI model to the network node.

[0106] In a possible embodiment, the network node can enter the online AI model trading market by means of registration information and obtain the right of use of one or more AI models through online model trading. It can also be one or more AI model groups. At the same time, the network node can also withdraw from the model trading market at any time.

[0107] In a possible embodiment, the network node can initiate a registration process to the cloud node, report the registration information of the network node, and obtain the qualification for online AI model trading.

[0108] S212: The network node establishes an online AI service list based on the right of use of the AI model through the intelligent layer of the radio access network.

[0109] In a possible embodiment, each network node completes model registration for the purchased AI model / model group based on the intelligent layer of the radio access network. Each network node can establish its own online AI service list.

[0110] S214: The network node sends the first information to the transmission node, and provides AI services for the user node based on the first information. The first information is used to send the online AI service list to the user node.

[0111] S216: The transmission node sends the third information to the user node. The third information is used to send the online AI service list to the user node.

[0112] In a possible embodiment, the network node provides AI services to all user nodes within its coverage area based on the online AI service list, and distributes the online AI service list to all user nodes through the transmission node.

[0113] In a possible embodiment, the network node may receive the second information sent by the transmission node, where the second information is used to subscribe to a target AI model, and provide the AI service of the target AI model for the user node through the intelligent layer of the radio access network.

[0114] In a possible embodiment, for any user node's request for an AI service, it is executed based on a workflow, and corresponding AI services are provided according to the AI service requirements of different user nodes.

[0115] For example, user node A requires high-throughput performance guarantee for AI services, while user node B needs to obtain AI services with low latency and high reliability. In the actual execution process, the workload is embodied in the form of one or more workflows, and each workflow has different requirements for AI services.

[0116] In a possible embodiment, the AI services include: concurrent services and sequential services; the concurrent services provide AI services for the workflows of multiple user nodes or multiple workflows of the same user node simultaneously; the sequential services provide AI services for the workflows of multiple user nodes or multiple workflows of the same user node in sequence.

[0117] In a possible embodiment, the network node may receive the model monitoring feedback information sent by the transmission node through the intelligent layer of the radio access network, and update the online AI service list in real time according to the model monitoring feedback information. The model monitoring feedback information is the information sent by the user node after performing model monitoring to the network node.

[0118] In a possible embodiment, the network node may receive the model monitoring feedback information through the intelligent layer of the radio access network. If the network node has the model fine-tuning permission, the network node may adjust the target AI model according to the model monitoring feedback information.

[0119] In a possible embodiment, the usage rights of the AI model include at least one of the following: model parameter usage rights, model usage instruction acquisition rights, API call rights, log acquisition rights, AI model reference information acquisition rights, model fine-tuning rights.

[0120] In a possible embodiment, the online trading mechanism of the AI model includes at least one of the following: the admission mechanism of the network node, the exit mechanism, the AI model trading method, the AI model trading time limit, the quantity limit, and the participation limit.

[0121] In a possible embodiment, the transmission node may receive fourth information sent by the user node, where the fourth information is used to subscribe to a target AI model, and send second information to the network node, where the second information is used to subscribe to the target AI model.

[0122] In a possible embodiment, the transmission node receives model monitoring feedback information and forwards it to the network node.

[0123] In a possible embodiment, after the user node receives the third information, it may determine the AI service budget, and based on the AI service budget, AI service requirements, and auxiliary information, select a list of AI service targets from the online AI service list, and send fourth information according to the list of AI service targets; the fourth information is used to subscribe to the target AI model.

[0124] In a possible embodiment, the user node may perform model monitoring and send model monitoring feedback information.

[0125] Figure 3 FIG. 2 shows a second interaction schematic diagram of an artificial intelligence model management method in an embodiment of the present disclosure. In the following embodiments, it is described by taking the need for the network node to send the model inference result to the user node as an example. As Figure 3 shown, the following steps are included:

[0126] S302: The cloud node establishes an online trading mechanism for AI models.

[0127] S304: The cloud node sends an AI model trading list to the network node. The AI model trading list includes: the AI models available at the cloud node.

[0128] In a possible embodiment, the online trading mechanism for AI models includes an admission and exit mechanism for network nodes, an AI model trading method, time limits for AI model trading, quantity limits, participation limits, etc.

[0129] In a possible embodiment, the online trading of AI models indicates that many elements of this trading market, including the types of AI models, trading prices, trading participants, etc., are dynamically changing. These elements may change with the environment, policies, and their own needs. Therefore, the model trading is in an online form.

[0130] In a possible embodiment, the transaction price of the AI model changes in real time and can be determined based on at least one of the time complexity of the AI model, the space complexity, the storage space of the AI model, the processing power of the AI model (e.g., token length), the number of storage bits of the AI model, the generation ability of the AI model (e.g., generation speed, maximum length, etc.), the quality of the database on which the AI model depends, and the training cost of the AI model. For example, the transaction price of the AI model is charged C yuan for generating B tokens every A seconds (in the database token mode).

[0131] In a possible embodiment, there can be multiple metrics for measuring the transaction price of the AI model, which are not limited in this disclosure and may include: energy consumption, equivalent services, cash value, etc.

[0132] In a possible embodiment, the cloud node is responsible for determining the AI model transaction method, including at least one of the following:

[0133] (1) Competitive transaction method: Each network node participates in the AI model transaction in the form of an auction, etc. The cloud node announces the initial price of a certain AI model, and each network node proposes a new bidding price for this price. The cloud node coordinates uniformly and continuously adjusts the current price of the AI model. Eventually, the network node with the highest bid can obtain the right to use the AI model. If no network node bids, the cloud node decides the ownership of the right to use the model;

[0134] (2) Non-competitive transaction method: Each network node participates in the AI model transaction in a non-competitive manner. For example, if the bid of the network node meets the requirements of the cloud node, it can obtain the right to use the model.

[0135] In a possible embodiment, the AI model transaction list includes at least one of the following: AI model ID, AI model usage period, AI model real-time transaction price, and AI model function information.

[0136] The AI model ID is a universal ID that can be uniquely identified within the entire network and is used for identification and use between different nodes.

[0137] The AI model function information includes at least one of the following: model structure of the AI model, model input and output formats, data requirements and data processing methods, model performance indicators, and other information.

[0138] In a possible embodiment, the AI model transaction list is transmitted from the cloud node to all network nodes through an interface customized by the cloud service provider.

[0139] S306: The network node determines the online transaction budget of the AI model based on the intelligent layer of the radio access network.

[0140] In a possible embodiment, the network node(s) can be one or more network nodes.

[0141] In a possible embodiment, the method for formulating the online transaction budget of the AI model for any network node A includes the following steps: The user node reports the AI service requirements of the user node to network node A through the transmission node. The radio access network intelligent layer of network node A collects network data within a preset time period. The radio access network intelligent layer of network node A estimates the obtainable revenue based on the aggregated network data information and in combination with the model monitoring feedback information of the user node, and formulates the AI model transaction budget according to the preset budget rules.

[0142] In a possible embodiment, the network data includes at least one of the following: the network revenue growth situation (profitability) within a preset time period, the network load status, network management information, network service requirements, updates of AI service-related information, updates of AI model deployment-related information, and other historical information.

[0143] In a possible embodiment, the preset budget rules may include at least one of the following: price budget, quantity budget, function budget, and time budget.

[0144] The price budget may include: the total price of purchasing the AI model, the upper limit of the unit price of each AI model, etc.

[0145] The quantity budget may include: the quantity of purchasing the AI model, the maximum quantity, etc.

[0146] The function budget may include: the functions of purchasing the AI model, the preferentially considered functions, the expected effects, etc.

[0147] The time budget may include: the effective time of purchasing the AI model, the maximum usage time, etc.

[0148] In a possible embodiment, the online transaction budget of the AI model formulated by the network node is time-sensitive, representing the budget of the network node for the AI model within the current time period. In the next time period, the network node will re-formulate the budget based on the radio access network intelligent layer.

[0149] S308: The network node initiates a registration process to the cloud node and sends registration information.

[0150] In a possible embodiment, the registration information of the network node includes at least one of the following: its operator ID, location information, permission information, etc. The above registration information can be reported to the cloud node through a custom interface.

[0151] S310: The cloud node and the network node complete the online transaction of the AI model.

[0152] S312: The cloud node issues the right to use the AI model to the network node.

[0153] In a possible embodiment, the network node selects one or more AI models according to the online trading budget of the AI model in the current time period to form a target AI model / model group, and performs model online trading.

[0154] In a possible embodiment, the network node first filters out all AI models that meet the requirements of the network node in the AI model function information according to the AI models in the AI model trading list, then selects K models that meet the quantity requirements, and then determines the target AI model / model group according to the K models.

[0155] In a possible embodiment, the target AI model can be determined according to the K models in various ways. For example, the principle of the lowest cost (that is, preferentially select AI models with lower prices), the principle of the highest performance, etc.

[0156] In a possible embodiment, for S310, the process of online trading may include the following steps: The network node pays the corresponding amount to the cloud node according to the trading price of the AI model corresponding to the target AI model / model group in the AI model trading list, and reports the AI model ID of the target AI model. The cloud node confirms the receipt of the payment, sends a confirmation message ACK to the network node, the cloud node retrieves the target AI model / model group in the AI model trading list, and opens the right to use the corresponding AI model / model group to the network node; the network node confirms that it has obtained the right to use the AI model / model group.

[0157] In a possible embodiment, the payment methods for model online trading include offline payment and online payment methods, which can be determined by negotiation between the network node and the cloud node, and the transaction is completed through an online payment platform or offline channels.

[0158] Among them, the payment process can be encrypted through blockchain technology and other means to ensure transaction security.

[0159] In a possible embodiment, the right to use the AI model means that the corresponding right to use the AI model can be obtained within the specified AI model usage period.

[0160] The right to use the AI model may include at least one of the following: the right to use model parameters, the right to obtain model usage instructions, API call rights, log acquisition rights, the right to obtain AI model reference information, and model fine-tuning rights.

[0161] In a possible embodiment, the AI model usage period is set by the cloud node and informed to the corresponding network node before model online trading.

[0162] In a possible embodiment, if a network node needs to withdraw from the model market, it can perform the following steps: send a withdrawal request signaling to the cloud node, receive the confirmation information feedback from the cloud node, after the cloud node feeds back the confirmation information, the network node executes the withdrawal mechanism and no longer participates in model transactions, and the cloud node deletes the registration information of the corresponding network node.

[0163] In a possible embodiment, the AI model registration process includes the intelligent layer of the radio access network recording the AI model information and identifying the AI services that the AI model can provide for the AI models / model groups purchased in the current time period.

[0164] In a possible embodiment, the AI model information includes at least one of the following: purchase time, cost information, energy consumption information, performance information.

[0165] S314: Based on the AI model usage right, the network node establishes an online AI service list through the intelligent layer of the radio access network.

[0166] In a possible embodiment, the intelligent layer of the radio access network of the network node can identify the AI services that each AI model can provide, and one AI model can provide one or more AI services. The intelligent layer of the radio access network integrates the identified AI service information and establishes an online AI service list.

[0167] In a possible embodiment, the online AI service list needs to provide the basic information of the AI service, including at least one of the following: AI service index, AI service nature, AI service function, AI service value, performance reference information, AI model deployment method, AI service format requirement, privacy protection requirement, value-added service.

[0168] The AI service index represents the index value of the AI service, which is used to distinguish different AI services within the coverage area of the network node.

[0169] The AI service nature represents the basic characteristics of the AI service, such as classification, regression, clustering, text generation, etc.

[0170] The AI service function represents the functions that the AI service can provide, which is a finer-grained function division corresponding to a certain AI service nature, such as beam management, high-precision positioning, coverage optimization, cell handover decision, fault detection, root cause analysis, sentiment analysis, etc.

[0171] The AI service value represents the price that the user node needs to pay for this AI service, which is related to the capabilities of the AI model used by the AI service and the resource consumption of the network node, and is also related to the specific AI service function. The same AI model may provide multiple AI services with different AI service values.

[0172] The performance reference information represents the expected performance reference information of the AI service, which is based on specific AI service functions and is associated with obtaining the AI model reference information sent by the cloud node.

[0173] The AI model deployment method indicates whether the AI model needs to be deployed on the user node or the network node when providing the AI service, or whether both bilateral models are jointly deployed. It depends on the actual deployment method of the network node for the AI model and is also associated with the actual application requirements of the user node.

[0174] The AI service format requirement represents the format requirements that the local user data needs to meet when the user node executes the AI service. For example, the adjustments that need to be made in terms of input and output data dimensions, multimodality, data volume, prompt words, etc.

[0175] The privacy protection requirement indicates whether the AI service provides user privacy protection capabilities, whether the input or output data of the AI model is confidential. If there are privacy protection issues, the user needs to be informed in advance and use this AI service with caution.

[0176] The value-added service represents additional AI services derived from the AI service in terms of AI service functions, etc. on the basis of this AI service. Therefore, the corresponding AI service value will also change.

[0177] The AI service value is formulated by the intelligent layer of the radio access network, and the pricing rule is configured by the network node itself.

[0178] In a possible embodiment, the online AI service list can be divided according to service nature, target audience (such as government and enterprise users, commercial users, ordinary users), service scale (such as large models, small models), etc.

[0179] S316: The network node sends the first information to the transport node, and the first information is used to send the online AI service list to the user node.

[0180] In a possible embodiment, in the method of the present disclosure, the online AI service list can be sent to all user nodes in the form of control plane broadcast, and the transport node can carry the online AI service list information through signaling such as SIB1. The first information can be: signaling such as SIB1.

[0181] In a possible embodiment, in the method of the present disclosure, the online AI service list can also be sent to all user nodes through the user plane.

[0182] In a possible embodiment, the online AI service list signaling transmission can include at least one of the following in the table: As shown in Table 1:

[0183] Table 1

[0184]

[0185] S318: The transmission node sends the third information to the user node, and the third information is used to send the online AI service list to the user node.

[0186] S320: The user node determines the AI service budget. Based on the AI service budget, AI service requirements, and auxiliary information, the user node selects a list of AI service targets from the online AI service list, according to the list of AI service targets.

[0187] In a possible embodiment, the process of the user node generating the list of AI service targets mainly includes the following steps: The user node formulates the AI service budget within the current time period. The user node filters out the AI services whose AI service functions meet the AI service requirements of the user node based on the AI services in the online AI service list. The user node sorts the AI services in ascending order according to the value of the AI services and the degree of adaptation of other auxiliary information to the requirements of the user node among the filtered AI services. The user node selects the maximum N AI services that meet the budget requirements from the sorted AI services to generate a list of AI service targets, where the value of N is determined by the user node or configured by the network node.

[0188] In a possible embodiment, the AI service budget is related to at least one of the following: the computing power of the user node, data processing and storage capabilities, network status, and AI service requirements. The rules of the AI service budget are the same as the rules of the online transaction budget of the AI model.

[0189] In a possible embodiment, the auxiliary information may include at least one of the following: performance, user privacy protection.

[0190] In a possible embodiment, the AI service budget changes dynamically with the user's own situation and the external environment, and the user node formulates the AI service budget for the current time period according to the information obtained in real time.

[0191] The granularity of the time period can be a frame, a half-frame, a time slot, etc. The finer the granularity of the time period, the more the user node will obtain the AI service budget information in real time, so as to make real-time AI service decisions.

[0192] In a possible embodiment, the list of AI service targets includes at least one of the following: AI service index, whether the AI service requires user privacy protection capabilities, etc.

[0193] S322: The user node sends the fourth information to the transmission node according to the list of AI service targets, and the fourth information is used to subscribe to the target AI model.

[0194] In a possible embodiment, the online AI service list is dynamically changing, and the user node's own requirements are also dynamically changing. Therefore, within each time period, the user node can dynamically adjust the AI service target list and the subscription result of the AI service, that is, the target AI model to be subscribed, to meet its own performance requirements.

[0195] The fourth information includes at least one of the following: AI service index, AI model deployment method selection, subscription time, UE configuration information, performance monitoring method, performance feedback period, privacy protection requirement, value-added service indication. The AI service index represents the index value of the AI service subscribed by the user node. The AI model deployment method selection represents the AI model deployment method selected by the user node for executing the AI service, including at least the following: deployment on the user node, deployment on the network node, or bilateral deployment method.

[0196] In a possible embodiment, according to the different deployment locations of the AI model, the ways for the network node to provide the AI service include at least one of the following forms, as shown below:

[0197] (1) Unilateral model deployment (AI model is deployed on the network node): The user node first performs data collection, and then uploads the data to the intelligent layer of the radio access network of the network node to complete AI model inference. The intelligent layer of the radio access network sends the inference result to the user node, and the user node performs model performance monitoring. A possible form is that the user node, based on the network large model, first inputs a prompt word to the large model deployed in the intelligent layer of the radio access network, and then the intelligent layer of the radio access network sends the output result of the large model to the user node.

[0198] (2) Unilateral model deployment (AI model is deployed on the user node): The intelligent layer of the radio access network of the network node first deploys the AI model on the user node through the downlink data channel, and then the user node performs data collection and performs model inference based on the deployed AI model to obtain the model inference result and execute the model performance monitoring mechanism. A possible form is that the user node, through the AI model on the intelligent wearable device, based on the data collected during its own movement, monitors the movement situation in real time and provides intelligent analysis results.

[0199] (3) Bilateral model deployment (AI model is deployed on both the network node and the user node): The intelligent layer of the radio access network of the network node first deploys the model on the AI side on the user node through the downlink data channel, and then the user node performs data collection and performs model inference based on the deployed model on the AI side to obtain the intermediate model inference result, and uploads the intermediate model inference result to the intelligent layer of the radio access network. The intelligent layer of the radio access network uses the network-side AI model to complete the subsequent model inference based on the intermediate model inference result and sends the final inference result to the user node, and the user node performs model performance monitoring.

[0200] The subscription time indicates the time when the user node subscribes to the AI service. During this time, the user node has the right to use this AI service, and the unit can be seconds, minutes, hours, days, weeks, etc.

[0201] UE configuration information represents the configuration operations that the user node needs to complete on the local client to execute this AI service. It can include at least one of the following: user data format configuration, workload configuration, transmission resource configuration, computing resource configuration, storage resource configuration, etc.

[0202] The performance monitoring method indicates that when the user node executes the AI service, it monitors the performance metrics of the AI service locally, which can include at least one of the following: AI metrics (e.g., accuracy, recall, etc.), communication metrics (e.g., throughput, spectral efficiency, data rate, latency, jitter, packet loss rate, etc.), service metrics (e.g., quality of service, service rate, service density, coverage, connection, etc.), energy consumption metrics (e.g., power consumption), computing power metrics (e.g., CPU / GPU utilization), user experience (QoE, etc.), security metrics (ability to resist network attacks).

[0203] The performance feedback period indicates that the user node feeds back the model monitoring performance of the AI service to the network node according to this period, and feeds back according to the performance metrics indicated by the performance monitoring method.

[0204] The privacy protection requirement represents the user node's demand for data privacy protection. If there is a privacy protection requirement, the network node needs to ensure that the user's local data, the data generated by the AI model, and other relevant configuration information are not shared or uploaded to other nodes; the value-added service indication represents whether the user node subscribes to a certain value-added service. The fourth information can be called the online AI service subscription request information, which is specific to the AI service. If the user node subscribes to M AI services, it needs to report the corresponding M service subscription request information, and the arrangement of the information can be in the form of a list: {online AI subscription request information of AI service 1,..., online AI subscription request information of AI service M}.

[0205] S324: The transmission node sends the second information to the network node, and the second information is used to subscribe to the target AI model.

[0206] In a possible embodiment, the transmission node can transmit the fourth information through at least one of the following information: uplink DCI, MAC CE, or RRC Reconfiguration control signaling. If the fourth information is configured to be transmitted by DCI, as shown in Table 2:

[0207] Table 2

[0208]

[0209] In a possible embodiment, the user node needs to pay the subscription fee for the AI service. The AI service subscription fee includes the usage fee of the AI service (corresponding to the value of the AI service), and also includes related fees (for example, whether to subscribe to additional AI services, information collection fees). If the AI model deployment method in the online AI service subscription request information is selected to be deployed at the user node or bilaterally deployed, the AI service subscription fee also includes the overhead of the network node transmitting the model to the user node.

[0210] In a possible embodiment, the AI service subscription fee can be transmitted by the user node to the network node through the user plane dedicated signaling, or can be transmitted to the network node through the control plane signaling. For example, uplink DCI, MAC CE or RRC signaling can be attached after the service subscription request information for unified transmission, or can be generated separately for signaling transmission.

[0211] S326: The network node provides the AI service of the target AI model for the user node through the intelligent layer of the radio access network.

[0212] S328: The network node sends the model inference result to the transmission node.

[0213] S330: The transmission node sends the model inference result to the user node.

[0214] S332: The user node performs model monitoring.

[0215] S334: The user node sends the model monitoring feedback information to the transmission node.

[0216] S336: The transmission node sends the model monitoring feedback information to the network node.

[0217] In a possible embodiment, the intelligent layer of the radio access network of the network node updates the basic information of the AI service recorded in the corresponding online AI service list based on the model monitoring feedback information of each user node.

[0218] For example, if user node 1 feedbacks that the performance of AI service 1 cannot reach the reference value given in the performance reference information, the network node will modify the performance reference information corresponding to AI service 1 in the online AI service list accordingly. The modified value can be the average value of the feedback information of all user nodes for AI service 1 or the value after comprehensive processing.

[0219] In a possible embodiment, if the cloud node grants the model fine-tuning permission in the usage right of the AI model in the network node, the network node can perform model fine-tuning on the AI model / model group corresponding to the AI service subscribed by the user node based on the model monitoring feedback information of the user node, and the fine-tuning method can be selected by the network node itself or configured by the cloud node.

[0220] The above Figure 2 and Figure 3 The steps in are independently performed in each time period and continuously iterated, and are executed in an online manner.

[0221] Figure 4 shows one of the flowcharts of an artificial intelligence model management method in an embodiment of the present disclosure. As Figure 4 shown, applied to a network node, it includes the following steps:

[0222] S402: Receive the artificial intelligence (AI) model transaction list sent by the cloud node.

[0223] S404: Determine the online transaction budget of the AI model based on the intelligent layer of the radio access network.

[0224] S406: Complete the online transaction of the AI model with the cloud node according to the AI model transaction list and the online transaction budget of the AI model, so that the cloud node issues the usage right of the AI model to the network node.

[0225] S408: Establish an online AI service list through the intelligent layer of the radio access network based on the received usage right of the AI model.

[0226] S410: Send the first information, and provide the AI service for the user node based on the first information. The first information is used to send the online AI service list to the user node.

[0227] Figure 5 shows another flowchart of an artificial intelligence model management method in an embodiment of the present disclosure. As Figure 5 shown, applied to the cloud node, it includes the following steps:

[0228] S502: Establish an online transaction mechanism for the AI model.

[0229] S504: Send the AI model transaction list to the network node. The AI model transaction list includes: the AI models available in the cloud node.

[0230] S506: Complete the online transaction of the AI model with the network node.

[0231] S508: Issue the usage right of the AI model to the network node.

[0232] Figure 6FIG. 3 shows a flowchart of an artificial intelligence model management method according to an embodiment of the present disclosure. As shown in Figure 6 shown, it is applied to a transmission node and includes the following steps:

[0233] S602: Receive first information sent by a network node. The first information is used to send an online AI service list to a user node. The online AI service list is established by the network node completing an online AI model transaction with a cloud node according to an AI model transaction list and an online AI model transaction budget, so that the cloud node issues the right to use the AI model to the network node.

[0234] S604: Send third information to the user node. The third information includes: the online AI service list.

[0235] Figure 7 FIG. 4 shows a flowchart of an artificial intelligence model management method according to an embodiment of the present disclosure. As shown in Figure 7 shown, it is applied to a user node and includes the following steps:

[0236] S702: Receive third information. The third information includes: the online AI service list. The online AI service list is established by the network node completing an online AI model transaction with a cloud node according to an AI model transaction list and an online AI model transaction budget, so that the cloud node issues the right to use the AI model to the network node.

[0237] Through the artificial intelligence model management method in the present disclosure, a complete AI service process is provided, which can complete the life cycle management process of the AI model, and supports cross-network domain AI model scheduling and transmission, ensuring the transformation of the AI model from plug-in to endogenous. Under the cloud-edge-end collaborative architecture, the intelligent layer of the radio access network supports "upward" to complete the online AI model transaction with the cloud service provider and online management of the AI model, and also supports "downward" to provide online AI services to the user node and online management of the AI services.

[0238] Configure the intelligent layer of the radio access network and deploy it on the network node, and give specific function definitions. Within each time period, the intelligent layer of the radio access network is responsible for managing processes such as transaction, deployment, inference, and fine-tuning of the AI model of this node, improving the ability of the network node to uniformly manage multiple elements such as data, model, and service. Provide a bottom-up online AI model transaction mechanism, provide a top-down online AI service provision mechanism, provide a customized model fine-tuning service for edge devices, provide an online AI service subscription request signaling that complies with 3GPP standards, and give the corresponding implementation process.

[0239] Based on the same inventive concept, embodiments of the present disclosure also provide a network node, a cloud node, a transmission node, and a user node. Since the principle of problem-solving in this device embodiment is similar to that in the above method embodiment, the implementation of this embodiment can refer to the implementation of the above method embodiment, and the repeated parts will not be elaborated.

[0240] Figure 8 FIG. shows a schematic structural diagram of a network node in an embodiment of the present disclosure, as Figure 8 shown, the network node 80 includes: a first communication module 801 and a first model management module 802.

[0241] The first communication module 801 is configured to receive an AI model transaction list sent by the cloud node; the first model management module 802 is configured to determine the online transaction budget of the AI model based on the radio access network intelligent layer; the first model management module 802 is configured to complete the online transaction of the AI model with the cloud node according to the AI model transaction list and the online transaction budget of the AI model, so that the cloud node issues the usage right of the AI model to the network node; the first model management module 802 is configured to establish an online AI service list through the radio access network intelligent layer based on the usage right of the AI model; the first communication module 801 is configured to send a first message to provide an AI service for the user node based on the first message; the first message is used to send the online AI service list to the user node.

[0242] Figure 9 FIG. shows a schematic structural diagram of a cloud node in an embodiment of the present disclosure, as Figure 9 shown, the cloud node 90 includes: a second communication module 901 and a second model management module 902.

[0243] The second model management module 902 is configured to establish an online transaction mechanism for the AI model; the second communication module 901 is configured to send the AI model transaction list to the network node; the AI model transaction list includes: the AI models available in the cloud node; the second model management module 902 is further configured to complete the online transaction of the AI model with the network node; the second communication module 901 is further configured to issue the usage right of the AI model to the network node.

[0244] Figure 10 FIG. shows a schematic structural diagram of a transmission node in an embodiment of the present disclosure, as Figure 10 shown, the transmission node 100 includes: a third communication module 1001.

[0245] The third communication module 1001 is configured to receive the first information sent by the network node; the first information is used to send an online AI service list to the user node; the online AI service list is established by the network node completing an online AI model transaction with the cloud node according to the AI model transaction list and the online transaction budget of the AI model, so that the cloud node issues the usage right of the AI model to the network node; the third communication module 1001 is further configured to send the third information to the user node; the third information includes: the online AI service list.

[0246] Figure 11 The following shows a schematic structural diagram of a user node in an embodiment of the present disclosure, as Figure 11 shown, the user node 110 includes: a fourth communication module 1101.

[0247] The fourth communication module 1101 is configured to receive the third information; the third information includes: the online AI service list; the online AI service list is established by the network node completing an online AI model transaction with the cloud node according to the AI model transaction list and the online transaction budget of the AI model, so that the cloud node issues the usage right of the AI model to the network node.

[0248] Those skilled in the art can understand that various aspects of the present disclosure can be implemented as a system, a method, or a program product. Therefore, various aspects of the present disclosure can be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or an implementation combining hardware and software aspects, which can be collectively referred to as "circuitry", "module", or "system" here.

[0249] Next, refer to Figure 12 to describe the electronic device 1200 according to this embodiment of the present disclosure. Figure 12 The electronic device 1200 shown is only an example and should not impose any limitation on the functions and usage scope of the embodiments of the present disclosure.

[0250] As Figure 12 shown, the electronic device 1200 is presented in the form of a general-purpose computing device. The components of the electronic device 1200 may include but are not limited to: at least one of the above-mentioned processing units 1210, at least one of the above-mentioned storage units 1220, and a bus 1230 connecting different system components (including the storage unit 1220 and the processing unit 1210).

[0251] Among them, the storage unit stores program code, and the program code can be executed by the processing unit 1210, so that the processing unit 1210 executes the steps according to various exemplary embodiments of the present disclosure described in the above "Exemplary Method" section of this specification. For example, the processing unit 1210 can execute the steps of any one of the above method embodiments.

[0252] The storage unit 1220 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 12201 and / or a cache storage unit 12202, and may further include a read-only storage unit (ROM) 12203.

[0253] The storage unit 1220 may also include a program / utilities 12204 having a set (at least one) of program modules 12205. Such program modules 12205 include, but are not limited to: an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment.

[0254] The bus 1230 may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus structures.

[0255] The electronic device 1200 may also communicate with one or more external devices 1240 (such as a keyboard, a pointing device, a Bluetooth device, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 1200, and / or may communicate with any device that enables the electronic device 1200 to communicate with one or more other computing devices (such as a router, a modem, etc.). Such communication may be through an input / output (I / O) interface 1250. Also, the electronic device 1200 may communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through a network adapter 1260. As shown in the figure, the network adapter 1260 communicates with other modules of the electronic device 1200 through the bus 1230. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 1200, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0256] Through the description of the above embodiments, those skilled in the art can easily understand that the example embodiments described herein can be implemented by software, or can be implemented by a combination of software and necessary hardware. Therefore, the technical solution according to the embodiments of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the embodiments of the present disclosure.

[0257] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method in the above embodiments.

[0258] In an exemplary embodiment of the present disclosure, there is also provided a computer-readable storage medium, which may be a readable signal medium or a readable storage medium. A program product capable of implementing the above method of the present disclosure is stored thereon. In some possible implementation manners, various aspects of the present disclosure may also be implemented in the form of a program product, which includes program code. When the program product runs on a terminal device, the program code is used to cause the terminal device to execute the steps according to various exemplary embodiments of the present disclosure described in the "Exemplary Method" section of this specification.

[0259] More specific examples of the computer-readable storage medium in the present disclosure may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0260] In the present disclosure, the computer-readable storage medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries the readable program code. Such a propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The readable signal medium may also be any readable medium other than the readable storage medium, which can send, propagate, or transmit a program used by or in conjunction with an instruction execution system, apparatus, or device.

[0261] Optionally, the program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination of the above.

[0262] In specific implementation, program code for performing the operations of the present disclosure can be written in any combination of one or more programming languages. The programming languages include object-oriented programming languages such as Java, C++, etc., and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, executed as an independent software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user's computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (e.g., by connecting through the Internet using an Internet service provider).

[0263] It should be noted that although several modules or units of the devices for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more of the above-described modules or units can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0264] In addition, although the steps of the methods in the present disclosure are described in a specific order in the drawings, this does not require or imply that the steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additionally or alternatively, some steps can be omitted, multiple steps can be combined into one step for execution, and / or one step can be decomposed into multiple steps for execution, etc.

[0265] Through the description of the above embodiments, those skilled in the art can easily understand that the example embodiments described herein can be implemented by software or by a combination of software and necessary hardware. Therefore, the technical solutions according to the embodiments of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a mobile terminal, or a network device, etc.) to execute the methods according to the embodiments of the present disclosure.

[0266] Other embodiments of the present disclosure will be readily apparent to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. The present disclosure is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include known common general knowledge or conventional technical means in the technical field not disclosed in the present disclosure. The specification and examples are only to be considered as exemplary, and the true scope and spirit of the present disclosure are pointed out by the appended claims.

Claims

1. An artificial intelligence model management method, characterized in that, Applied to a network node, the method includes: Receiving an artificial intelligence (AI) model transaction list sent by a cloud node; Determining an online transaction budget for the AI model based on the intelligent layer of the radio access network; Completing an online transaction of the AI model with the cloud node according to the AI model transaction list and the online transaction budget for the AI model, so that the cloud node issues the right to use the AI model to the network node; Based on the received right to use the AI model, establishing an online AI service list through the intelligent layer of the radio access network; Sending first information to provide an AI service for a user node based on the first information; the first information is used to send the online AI service list to the user node.

2. The method according to claim 1, wherein The step of sending first information to provide an AI service for a user node based on the first information includes: Sending first information, where the first information is used to send the online AI service list to the user node; Receiving second information; the second information is used to subscribe to a target AI model; Providing the AI service of the target AI model for the user node through the intelligent layer of the radio access network.

3. The method according to claim 1, wherein The AI service includes: concurrent services and sequential services; The concurrent service provides AI services for the workflows of multiple user nodes or multiple workflows of the same user node simultaneously; the sequential service provides AI services for the workflows of multiple user nodes or multiple workflows of the same user node sequentially.

4. The method according to claim 1, wherein The method further includes: Receiving model monitoring feedback information through the intelligent layer of the radio access network; Updating the online AI service list in real time according to the model monitoring feedback information.

5. The method according to claim 2, characterized in that, The method further includes: Receiving model monitoring feedback information through the intelligent layer of the radio access network; If the network node has the permission to fine-tune the model, adjusting the target AI model according to the model monitoring feedback information.

6. The method according to claim 1, wherein The method further includes: Initiating a registration process to the cloud node and reporting the registration information of the network node; Obtaining the qualification for online transactions of AI models.

7. The method according to claim 1, wherein The right to use the AI model includes at least one of the following: the right to use model parameters, the right to obtain model usage instructions, the right to call an application programming interface (API), the right to obtain logs, the right to obtain reference information of the AI model, and the right to fine-tune the model.

8. An artificial intelligence model management method, characterized in that, Applied to a cloud node, the method includes: Establishing an online transaction mechanism for AI models; Sending an AI model transaction list to a network node; the AI model transaction list includes: the AI models available on the cloud node; Completing an online transaction of the AI model with the network node; Issuing the right to use the AI model to the network node.

9. The method according to claim 8, characterized in that, The online transaction mechanism for AI models includes at least one of the following: an access mechanism for network nodes, an exit mechanism, an AI model transaction method, a time limit for AI model transactions, a quantity limit, and a participation limit.

10. The method according to claim 8, wherein The method further includes: Receiving the registration information sent by the network node; Allowing the network node to conduct online transactions of AI models.

11. An artificial intelligence model management method, characterized in that, Applied to a transmission node, the method includes: Receive the first information sent by the network node; the first information is used to send the online AI service list to the user node; the online AI service list is established by the network node completing an online AI model transaction with the cloud node according to the AI model transaction list and the online transaction budget of the AI model, so that the cloud node issues the right to use the AI model to the network node. Send the third information to the user node; the third information includes: the online AI service list.

12. The method according to claim 11, wherein The method further includes: Receive the fourth information sent by the user node, where the fourth information is used to subscribe to a target AI model. Send the second information to the network node, where the second information is used to subscribe to the target AI model.

13. The method according to claim 11, wherein The method further includes: Receive the model monitoring feedback information and forward it to the network node.

14. An artificial intelligence model management method, characterized in that, Applied to the user node, the method includes: Receive the third information; the third information includes: the online AI service list; the online AI service list is established by the network node completing an online AI model transaction with the cloud node according to the AI model transaction list and the online transaction budget of the AI model, so that the cloud node issues the right to use the AI model to the network node.

15. The method according to claim 14, wherein The method further includes: Determine the AI service budget. Based on the AI service budget, AI service requirements, and auxiliary information, select a list of AI service targets from the online AI service list. According to the list of AI service targets, send the fourth information; the fourth information is used to subscribe to a target AI model.

16. The method according to claim 14, wherein The method further includes: Perform model monitoring. Send model monitoring feedback information.

17. A network node, characterized in that, Includes: The first communication module is used to receive the artificial intelligence AI model transaction list sent by the cloud node. The first model management module is used to determine the online transaction budget of the AI model based on the intelligent layer of the radio access network. The first model management module is used to complete an online AI model transaction with the cloud node according to the AI model transaction list and the online transaction budget of the AI model, so that the cloud node issues the right to use the AI model to the network node. The first model management module is used to establish an online AI service list through the intelligent layer of the radio access network based on the right to use the AI model. The first communication module is used to send the first information, and provide AI services for the user node based on the first information; the first information is used to send the online AI service list to the user node.

18. A cloud node, characterized in that, Includes: The second model management module is used to establish an online transaction mechanism for the AI model. The second communication module is used to send the AI model transaction list to the network node; the AI model transaction list includes: the AI models available at the cloud node. The second model management module is further used to complete an online AI model transaction with the network node. The second communication module is further used to issue the right to use the AI model to the network node.

19. A transmission node, characterized in that, Includes: The third communication module is used to receive the first information sent by the network node. The first information is used to send the online AI service list to the user node; the online AI service list is established by the network node completing an online AI model transaction with the cloud node according to the AI model transaction list and the online transaction budget of the AI model, so that the cloud node issues the right to use the AI model to the network node. The third communication module is further configured to send third information to the user node; the third information includes: the online AI service list.

20. A user node, characterized in that, Including: The fourth communication module is configured to receive the third information. The third information includes: the online AI service list. The online AI service list is established by the network node completing an online AI model transaction with the cloud node according to the AI model transaction list and the online transaction budget of the AI model, so that the cloud node issues the right to use the AI model to the network node.

21. A communication system, characterized in that, Including: The network node according to claim 17, the cloud node according to claim 18, the transmission node according to claim 19, and the user node according to claim 20.

Images