Artificial intelligence (AI) model deployment method and apparatus

Abstract

The present disclosure relates to an artificial intelligence (AI) model deployment method and apparatus. The method comprises: a first node determining that the operating state of the first node meets a preset condition; and when the operating state of the first node meets the preset condition, the first node executing an AI deployment operation, wherein the AI deployment operation is one of the operations for altering an AI model operating in the first node, and the first node is any one of a terminal, an access network device, and a network element of a core network.

Description

A method and apparatus for deploying an artificial intelligence (AI) model

[0001] Cross-references to related applications

[0002] This disclosure claims priority to Chinese Patent Application No. 202411611242.4, filed on November 12, 2024, entitled "A Method and Apparatus for Deploying an Artificial Intelligence (AI) Model", the entire contents of which are incorporated herein by reference. Technical Field

[0003] This disclosure relates to the field of communications, and more particularly to a method and apparatus for deploying an artificial intelligence (AI) model. Background Technology

[0004] Currently, with the development of artificial intelligence (AI) technology, mobile communication systems are beginning to explore the use of AI to improve network operating efficiency and service quality.

[0005] For example, in 5G mobile communication technology, an AI model runs within the Network Data Analytic Function (NWDAF). When nodes in the network require intelligent analysis, they can send relevant requests to the NWDAF along with corresponding parameters. Upon receiving the request, the NWDAF can obtain the intelligent analysis results based on the parameters and the AI ​​model, and then feed the results back to the node.

[0006] As can be seen, in the above-mentioned related technologies, when a node in the network needs intelligent analysis, the node first needs to send a request carrying parameters to NWDAF, and then NWDAF uses the AI ​​model to obtain the intelligent analysis results and then feeds the intelligent analysis results back to the node. The whole process is quite complicated. Summary of the Invention

[0007] To address the aforementioned technical issues, this disclosure provides a method and apparatus for deploying an artificial intelligence (AI) model.

[0008] Firstly, a method for deploying an artificial intelligence (AI) model is provided. This method is applied to a first node in a mobile communication system. The method includes: the first node determining that its operating state meets preset conditions; and the first node performing an AI deployment operation when its operating state meets the preset conditions. The AI ​​deployment operation is one of the operations that changes the AI ​​model running in the first node. The first node can be any one of a terminal, an access network device, or a core network element.

[0009] In some implementations, the first node determines that its operating state meets preset conditions, including: the first node detects that its operating state meets preset conditions. When the first node's operating state meets the preset conditions, the first node performs an AI deployment operation, including: the first node performs an AI deployment operation when it detects that its operating state meets the preset conditions.

[0010] In some implementations, AI deployment operations include one or more of the following: activating a first AI model, deactivating a second AI model, and switching a running AI model to a third AI model. The first, second, and third AI models are AI models pre-deployed in the first node.

[0011] In some implementations, when the first node detects that its operating status meets preset conditions, it performs an AI deployment operation, including: the first node sending a request message to the second node; wherein the request message instructs the second node to send an AI model to the first node; the first node receiving a first instruction message sent by the second node according to the request message; the first instruction message including a fourth AI model; the first node performing the AI ​​deployment operation according to the first instruction message; the AI ​​deployment operation includes: activating the fourth AI model, or switching the currently running AI model to the fourth AI model.

[0012] In some implementations, the request information includes the running status of the first node.

[0013] In some implementations, the first node determines that its operating state meets preset conditions, including: the first node receives second indication information from the second node; the second indication information is used to instruct the second node to detect that the first node's operating state meets preset conditions; and the first node performs an AI deployment operation when the first node's operating state meets preset conditions, including: the first node performs an AI deployment operation after receiving the second indication information.

[0014] In some implementations, AI deployment operations include one or more of the following: activating the fifth AI model, deactivating the sixth AI model, and switching the currently running AI model to the seventh AI model. The fifth, sixth, and seventh AI models are pre-deployed in the first node.

[0015] In some implementations, the second instruction information carries an eighth AI model; after receiving the second instruction information, the first node performs AI deployment operations, including: after receiving the second instruction information, the first node activates the eighth AI model carried in the second instruction information in the first node, or switches the currently running AI model in the first node to the eighth AI model carried in the second instruction information.

[0016] In some implementations, the first node is a terminal; preset conditions include one or more of the following: the first node is moved to an area where AI usage is restricted, the first node is in power-saving mode, the operating system of the first node is changed, the utilization rate of the storage resources of the first node is changed, the utilization rate of the computing power of the first node is changed, and the accuracy of the AI ​​model running in the first node is lower than the accuracy threshold.

[0017] In some implementations, the first node is a network element of the core network; preset conditions include one or more of the following: the first node is deployed to an area where the use of AI is restricted, the load state of the first node changes, and the accuracy of the AI ​​model running in the first node is lower than an accuracy threshold.

[0018] In some implementations, the first node is an access network device; preset conditions include one or more of the following: changes in the utilization rate of the air interface resources of the first node, changes in the load status of the first node, changes in the utilization rate of the computing power of the first node, and the accuracy of the AI ​​model running in the first node being lower than the accuracy threshold.

[0019] Secondly, a method for deploying an artificial intelligence (AI) model is provided. This method is applied to a mobile communication system, which includes a first node and a second node. The method includes: the second node acquiring the operating status of the first node; the first node is any one of a terminal, an access network device, or a network element of the core network. After the second node acquires that the operating status of the first node meets preset conditions, it sends a first instruction message to the first node; the first instruction message is used to instruct the first node to perform an AI deployment operation. The AI ​​deployment operation includes changing the AI ​​model running in the first node.

[0020] In some implementations, the first instruction information is specifically used to instruct the first node to perform one or more of the following: activate the first AI model, deactivate the second AI model, and switch the running AI model to the third AI model; wherein the first AI model, the second AI model, and the third AI model are AI models pre-deployed in the first node.

[0021] In some implementations, the first instruction information includes a fourth AI model; specifically, the first instruction information is used to instruct the first node to activate the fourth AI model in the first node, or to switch the running AI model in the first node to the fourth AI model.

[0022] In some implementations, the second node obtains the running status of the first node, including: the second node receiving status information from the first node; the status information includes the running status of the first node.

[0023] In some implementations, the second node obtains the running status of the first node, including: the second node detects the running status of the first node.

[0024] In some implementations, when the first node is a terminal, preset conditions include one or more of the following: the first node moves to an area where AI usage is restricted; the first node is in power-saving mode; the operating system of the first node changes; the utilization rate of the first node's storage resources changes; the utilization rate of the first node's computing power changes; and the accuracy of the AI ​​model running in the first node is lower than an accuracy threshold. Alternatively, when the first node is a network element in the core network, preset conditions include one or more of the following: the first node is deployed to an area where AI usage is restricted; the load state of the first node changes; and the accuracy of the AI ​​model running in the first node is lower than an accuracy threshold. Alternatively, when the first node is an access network device, preset conditions include one or more of the following: the utilization rate of the first node's air interface resources changes; the load state of the first node changes; the utilization rate of the first node's computing power changes; and the accuracy of the AI ​​model running in the first node is lower than an accuracy threshold.

[0025] Thirdly, a method for deploying an artificial intelligence (AI) model is provided. This method is applied to a mobile communication system, which includes a first node and a second node. The method includes: the second node receiving request information from the first node; the request information instructing the first node to send an AI model to the first node; the first node is any one of a terminal, an access network device, or a network element of the core network. The second node, based on the request information, sends first instruction information to the first node; the first instruction information includes a first AI model; the first instruction information instructs the first node to activate a second AI model in the first node, or to switch the currently running AI model in the first node to the second AI model.

[0026] In some implementations, the request information includes the running status of the first node; the second node sends a first instruction to the first node based on the request information, including: after the second node determines that the running status of the first node meets the preset conditions based on the request information, it sends the first instruction to the first node.

[0027] In some implementations, the method further includes: the second node detecting the running status of the first node; the second node sending first indication information to the first node according to the request information, including: after determining that the running status of the first node meets the preset conditions, the second node sending first indication information to the first node according to the request information.

[0028] Fourthly, a communication device is provided, applied to a first node in a mobile communication system. The communication device includes: a memory, a transceiver, and a processor. The memory is used to store computer programs; the transceiver is used to send and receive data under the control of the processor; the processor is used to read the computer program in the memory and perform the following operations: determining that the operating state of the first node meets preset conditions; and, if the operating state of the first node meets the preset conditions, executing an AI deployment operation; wherein, the AI ​​deployment operation is one of the operations that changes the AI ​​model running in the first node; the first node is any one of a terminal, an access network device, or a network element of the core network.

[0029] Fifthly, a communication device is provided for a second node in a mobile communication system, the mobile communication system including a first node and a second node. The communication device includes: a memory, a transceiver, and a processor; the memory is used to store a computer program; the transceiver is used to send and receive data under the control of the processor; the processor is used to read the computer program in the memory and perform the following operations: obtaining the operating status of the first node; the first node is any one of a terminal, an access network device, or a network element of the core network; after obtaining that the operating status of the first node meets preset conditions, sending first instruction information to the first node; the first instruction information is used to instruct the first node to perform an AI deployment operation; wherein, the AI ​​deployment operation includes the operation of changing the AI ​​model running in the first node.

[0030] In a sixth aspect, a communication device is provided for a second node in a mobile communication system, the mobile communication system including a first node and a second node. The communication device includes: a memory, a transceiver, and a processor; the memory is used to store a computer program; the transceiver is used to send and receive data under the control of the processor; the processor is used to read the computer program in the memory and perform the following operations: receiving request information from the first node; the request information is used to instruct the first node to send an AI model; the first node is any one of a terminal, an access network device, or a network element of a core network; according to the request information, sending first instruction information to the first node; the first instruction information includes a first AI model; the first instruction information is used to instruct the first node to activate a second AI model in the first node, or to switch the currently running AI model in the first node to the second AI model.

[0031] In a seventh aspect, a communication device is provided, which is applied to a first node in a mobile communication system. The communication device includes:

[0032] The determining unit is used to determine whether the operating status of the first node meets preset conditions;

[0033] The processing unit is used to execute AI deployment operations when the running status of the first node meets preset conditions;

[0034] Among them, AI deployment operation is one of the operations that changes the AI ​​model running in the first node; the first node is any one of the terminal, access network equipment or core network element.

[0035] Eighthly, a communication device is provided, wherein the communication device is applied to a second node in a mobile communication system, the mobile communication system including a first node and a second node, and the communication device includes:

[0036] The processing unit is used to obtain the operating status of the first node; the first node can be any one of a terminal, an access network device, or a core network element.

[0037] The communication unit is used to send a first instruction message to the first node after obtaining the running status of the first node and meeting the preset conditions; the first instruction message is used to instruct the first node to perform AI deployment operations.

[0038] AI deployment operations include operations that modify the AI ​​model running in the first node.

[0039] A ninth aspect provides a communication device applied to a second node in a mobile communication system, the mobile communication system including a first node and a second node, the communication device comprising: a receiving unit for receiving request information from a first node; the request information instructing the first node to send an AI model; the first node being any one of a terminal, an access network device, or a network element of a core network; and a sending unit for sending first instruction information to the first node according to the request information; the first instruction information including a first AI model; the first instruction information instructing the first node to activate a second AI model in the first node, or to switch a running AI model in the first node to the second AI model.

[0040] A tenth aspect provides a communication system including a first node and a second node; wherein the first node is configured to perform the method provided by the first aspect or any implementation thereof; the second node is configured to perform the method provided by the second aspect or any two implementations thereof, or the second node is configured to perform the method provided by the third aspect or any implementation thereof.

[0041] Eleventhly, a processor-readable storage medium is provided, the processor-readable storage medium storing a program for causing the processor to perform the method provided by the first aspect or any implementation thereof, or for causing the processor to perform the method provided by the second aspect or any implementation thereof, or for causing the processor to perform the method provided by the third aspect or any implementation thereof.

[0042] In a twelfth aspect, a computer program product is provided, the computer program product including instructions that, when executed on a processor, implement a method as described in the first aspect or any implementation thereof, or implement a method as described in the second aspect or any implementation thereof, or implement a method as described in the third aspect or any implementation thereof.

[0043] In the method provided by this disclosure, on the one hand, compared to the related technologies that deploy AI models in NWDAF network elements specifically responsible for network intelligence, this disclosure can deploy AI models in terminals, access network devices, or core network elements (i.e., the first node). On the other hand, this disclosure can perform AI deployment operations when the operating state of the first node meets preset conditions, thereby achieving the effect of changing the AI ​​model running in the first node according to the operating state of the first node. This achieves the goal of flexibly deploying AI models in the first node. Attached Figure Description

[0044] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0045] To more clearly illustrate the technical solutions in the embodiments or related technologies of this disclosure, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0046] Figure 1 is a schematic diagram of the structure of a mobile communication system provided in an embodiment of this disclosure;

[0047] Figure 2 is a flowchart illustrating one of the methods for deploying an AI model according to an embodiment of this disclosure;

[0048] Figure 3 is a second schematic flowchart of an AI model deployment method provided in an embodiment of this disclosure;

[0049] Figure 4 is a third flowchart illustrating an AI model deployment method provided in this embodiment of the present disclosure;

[0050] Figure 5 is a flowchart illustrating a method for deploying an AI model according to an embodiment of this disclosure;

[0051] Figure 6 is a fifth flowchart illustrating an AI model deployment method provided in an embodiment of this disclosure;

[0052] Figure 7 is a flowchart of a method for deploying an AI model according to an embodiment of this disclosure;

[0053] Figure 8 is a flowchart of a method for deploying an AI model according to an embodiment of this disclosure (the seventh one).

[0054] Figure 9 is a flowchart illustrating an AI model deployment method according to an embodiment of this disclosure (the eighth one).

[0055] Figure 10 is a flowchart of a method for deploying an AI model according to an embodiment of this disclosure;

[0056] Figure 11 is a flowchart of an AI model deployment method provided in an embodiment of this disclosure;

[0057] Figure 12 is a flowchart of an AI model deployment method provided in an embodiment of this disclosure;

[0058] Figure 13 is a flowchart of an AI model deployment method provided in an embodiment of this disclosure, number 12.

[0059] Figure 14 is a flowchart of an AI model deployment method provided in an embodiment of this disclosure, number thirteen.

[0060] Figure 15 is a flowchart of an AI model deployment method provided in an embodiment of this disclosure, number fourteen.

[0061] Figure 16 is a flowchart of an AI model deployment method provided in an embodiment of this disclosure, number fifteen.

[0062] Figure 17 is a flowchart of an AI model deployment method provided in an embodiment of this disclosure, number sixteen.

[0063] Figure 18 is a flowchart of an AI model deployment method provided in an embodiment of this disclosure, number seventeen.

[0064] Figure 19 is a flowchart of an AI model deployment method provided in an embodiment of this disclosure, number eighteen.

[0065] Figure 20 is a flowchart of an AI model deployment method provided in an embodiment of this disclosure, number nineteen.

[0066] Figure 21 is a flowchart of an AI model deployment method provided in an embodiment of this disclosure;

[0067] Figure 22 is a flowchart of an AI model deployment method according to an embodiment of this disclosure, number twenty-one.

[0068] Figure 23 is a schematic diagram of one of the structures of a communication device provided in an embodiment of this disclosure;

[0069] Figure 24 is a second schematic diagram of the structure of a communication device provided in an embodiment of this disclosure;

[0070] Figure 25 is a third schematic diagram of the structure of a communication device provided in an embodiment of this disclosure;

[0071] Figure 26 is a fourth schematic diagram of a communication device provided in an embodiment of this disclosure. Detailed Implementation

[0072] To better understand the above-mentioned objectives, features, and advantages of this disclosure, the solutions disclosed herein will be further described below. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0073] Numerous specific details are set forth in the following description in order to provide a full understanding of this disclosure, but this disclosure may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only some, and not all, of the embodiments of this disclosure.

[0074] First, the technical terms involved in the embodiments of this disclosure will be introduced:

[0075] Artificial Intelligence (AI) model.

[0076] An AI model refers to a system trained using machine learning and other technologies, through computer algorithms and data, capable of performing specific tasks. When this system is running, it can simulate human intelligent behavior to complete specific tasks including image recognition, speech recognition, natural language processing, and machine translation.

[0077] The core of an AI model is its algorithm and data. Through continuous iterative learning and optimization, AI models can continuously improve their accuracy and efficiency.

[0078] AI models typically consist of model architecture, parameters, and training methods. The model architecture defines the data processing flow, the parameters are the objects the model learns and optimizes, and the training methods guide how the model learns and improves.

[0079] The training process for AI models typically includes steps such as data preprocessing, model selection and optimization, and model training.

[0080] Data preprocessing includes cleaning and formatting the raw data. Preprocessing removes noise and outliers, ensuring data quality and accuracy.

[0081] Model selection and optimization include: selecting appropriate model structures and algorithms based on the characteristics and requirements of the task, and adjusting and optimizing parameters to obtain better model performance.

[0082] Model training includes: training the model using the training set, optimizing the model's weights and features through repeated iterations and learning, and improving the model's performance.

[0083] In related technologies, AI models can be applied to mobile communication systems to optimize network performance. For example, in 5G (5th Generation Mobile Communication Technology), AI models can run within the Network Data Analytic Function (NWDAF). When nodes in the network require intelligent analysis, they can send relevant requests to the NWDAF along with corresponding parameters. Upon receiving the request, the NWDAF can obtain intelligent analysis results based on the parameters and the AI ​​model, and then feed these results back to the nodes.

[0084] The application scenarios of the embodiments of this disclosure are described below:

[0085] The technical solutions provided in this disclosure are applicable to a variety of mobile communication systems. For example, applicable mobile communication systems may include Long Term Evolution (LTE) systems, LTE Frequency Division Duplex (FDD) systems, LTE Time Division Duplex (TDD) systems, Long Term Evolution Advanced (LTE-A) systems, Universal Mobile Telecommunication System (UMTS), Worldwide Interoperability for Microwave Access (WiMAX) systems, 5G New Radio (NR) systems and their evolved communication systems, and 6G (sixth generation mobile communication technology) systems. These systems may include terminals and network equipment. The systems may also include a core network component, such as the Evolved Packet Core (EPC) and the 5G Core Network (5GC).

[0086] For example, Figure 1 is a schematic diagram of a network architecture applying the technical solution provided in the embodiments of this disclosure. The network may include: a terminal, a radio access network (RAN) or an access network (AN) (RAN and AN are collectively referred to as (R)AN), and a core network (CN).

[0087] A terminal can be a device with wireless transceiver capabilities. It can have various names, such as user equipment (UE), access device, terminal unit, terminal station, mobile station, mobile station, remote station, remote terminal, mobile device, wireless communication device, terminal agent, or terminal apparatus. Terminals can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; they can also be deployed on water (such as on ships); and they can be deployed in the air (such as on airplanes, balloons, and satellites). Terminals include handheld devices, vehicle-mounted devices, wearable devices, or computing devices with wireless communication capabilities. For example, a terminal can be a mobile phone, tablet computer, or computer with wireless transceiver capabilities. Terminals can also be virtual reality (VR) devices, augmented reality (AR) devices, wireless terminals for industrial control, wireless terminals in autonomous driving, wireless terminals in telemedicine, wireless terminals in smart grids, wireless terminals in smart cities, and wireless terminals in smart homes. In this embodiment of the disclosure, the device for implementing the terminal's functions can be the terminal itself, or it can be any device capable of supporting the terminal in implementing those functions, such as a chip system. In this disclosure, the chip system can consist of chips, or it can include chips and other discrete components.

[0088] In (R)AN, the main components include access network equipment. Access network equipment can also be called base stations. Base stations can include various forms of base stations, such as macro base stations, micro base stations (also known as small stations), relay stations, and access points. Specifically, they can be: access points (APs) in Wireless Local Area Networks (WLANs), base stations (BTSs) in Global System for Mobile Communications (GSM) or Code Division Multiple Access (CDMA), base stations (NodeBs, NBs) in Wideband Code Division Multiple Access (WCDMA), evolved Node Bs (eNBs or eNodeBs) in LTE, relay stations or access points, or next-generation Node Bs (gNBs) in vehicle-mounted equipment, wearable devices, and 5G networks, or base stations in future Public Land Mobile Networks (PLMNs), etc.

[0089] The core network comprises multiple core network elements (or network function elements). Taking a 5G system as an example, as shown in Figure 1, the core network in a 5G system includes: access and mobility management services (AMF) elements, session management function (SMF) elements, policy control function (PCF) elements, user plane function (UPF) elements, application function (AF) elements, authentication server function (AUSF) elements, unified data management (UDM) elements, and network data analytics function (NWDAF) elements.

[0090] In addition, the core network may also include some network elements not shown in Figure 1, such as: security anchor function (SEAF) network elements, authentication credential repository and processing function (ARPF), which will not be described in detail in this embodiment.

[0091] Among them, NWDAF is a network element in 5G networks responsible for collecting, analyzing and processing data in the network, and making corresponding control decisions based on the analysis results.

[0092] NWDAF includes a set of rules and strategies to guide NWDAF in data analysis and decision control. These rules and strategies are collectively referred to as "NWDAF control rules".

[0093] By formulating reasonable NWDAF control rules, it can be ensured that NWDAF can quickly and accurately analyze and judge the network status based on real-time data in the network, and take corresponding control measures to optimize network performance. These NWDAF control rules include a variety of conditional judgments and control strategies, which can be customized and adjusted according to different network needs and scenarios.

[0094] The specific content of NWDAF control rules includes: data acquisition rules, data analysis rules, and decision control rules. Data acquisition rules specify how NWDAF collects data from the network. These rules include the frequency of data acquisition, the data type collected, and the data source. Data analysis rules specify how NWDAF analyzes and processes the collected data. These rules include data analysis algorithms, analysis metrics, and thresholds. Decision control rules specify how NWDAF makes corresponding decisions and controls based on the data analysis results. These rules include the conditions for decision-making, the decision-making strategy, and the decision-making actions.

[0095] The technical solutions provided in the embodiments of this disclosure are described below with reference to examples:

[0096] As mentioned above, in related technologies, mobile communication systems typically employ external AI models, which run the AI ​​model within the NWDAF. When nodes in the network require intelligent analysis, the nodes must first send a request carrying parameters to the NWDAF. Then, the NWDAF uses the AI ​​model to obtain the intelligent analysis results and feeds them back to the nodes. The entire process is quite complex.

[0097] To address the aforementioned issues, this embodiment considers the following: External AI models can be converted into internal AI models, i.e., the AI ​​model can be deployed on one or more nodes in a terminal, access network device, or core network element of a mobile communication system (hereinafter, the node deploying the AI ​​model in the terminal, access network device, or core network element is referred to as the first node). This allows the first node deploying the AI ​​model to complete the training and inference of the AI ​​model according to its own conditions.

[0098] Specifically, in the first node, the relationship between AI functions and AI models is shown in Table 1 below:

[0099] Table 1

[0100] In other words, in this embodiment of the disclosure, one or more AI models can run in the first node. These AI models, as described in Table 1 above, are used to implement one or more AI functions. The current problem to be solved is how to deploy the AI ​​models in the first node.

[0101] Based on the above considerations, this disclosure provides a method for deploying an AI model. In this method, a first node in a mobile communication system can perform an AI deployment operation based on its operating state. The first node can be any one of a terminal, access network equipment, or a core network element in the mobile communication system, and the AI ​​deployment operation can specifically be one of the operations that changes the AI ​​model running in the first node.

[0102] Specifically, as shown in Figure 2, after the first node determines that the running state of the first node meets the preset conditions (i.e., S101), the first node can perform AI deployment operations (i.e., S102) when the running state of the first node meets the preset conditions.

[0103] In the method provided by this disclosure, on the one hand, compared to the related technologies that deploy AI models in NWDAF network elements specifically responsible for network intelligence, this disclosure can deploy AI models in terminals, access network devices, or core network elements (i.e., the first node). On the other hand, this disclosure can perform AI deployment operations when the operating state of the first node meets preset conditions, thereby achieving the effect of changing the AI ​​model running in the first node according to the operating state of the first node. This achieves the goal of flexibly deploying AI models in the first node.

[0104] The following describes the specific implementation process of the method provided in this disclosure when applied to three types of nodes: terminals, access network devices, and core network elements.

[0105] In the first implementation, the method provided in this disclosure can be applied to a terminal. For example, the first node in Figure 2 above can be a terminal. Specifically, when the method provided in this disclosure is applied to a terminal (for ease of description, the terminal implementing the method provided in this disclosure will be referred to as terminal E1 below), the method may include:

[0106] S201. Terminal E1 determines that the operating state of terminal E1 meets the preset conditions (hereinafter referred to as preset condition F1 for ease of description).

[0107] S202. When the operating state of terminal E1 meets the preset condition F1, the AI ​​deployment operation is executed.

[0108] Among them, AI deployment operation can be one of the operations that changes the AI ​​model running in terminal E1.

[0109] Specifically, the above AI deployment operations may include: activating the AI ​​model, deactivating the AI ​​model, or switching the AI ​​model.

[0110] The following six possible designs will be used to describe the contents included in the above-mentioned preset condition F1:

[0111] In the first possible design, the preset condition F1 may include: the terminal E1 moves to an area where the use of AI is restricted.

[0112] The above design allows terminal E1 to trigger an AI deployment operation when it moves to an area where AI usage is restricted. For example, this AI deployment operation could be either deactivating the AI ​​model running on terminal E1 or switching the AI ​​model running on terminal E1. This ensures that terminal E1 can function normally in the currently restricted AI usage area.

[0113] In the second possible design, the preset condition F1 may include: terminal E1 is in power saving mode.

[0114] The above design allows the AI ​​deployment operation to be triggered when the terminal E1 is in power-saving mode. For example, this AI deployment operation could be either deactivating the AI ​​model running on the terminal E1 or switching the AI ​​model running on the terminal E1. This reduces the power consumption of the terminal E1 when it is in power-saving mode, thus achieving power saving.

[0115] In the third possible design, the preset condition F1 may include: a change in the operating system of terminal E1.

[0116] The above design allows for the execution of AI deployment operations (e.g., deactivating, activating, or switching the AI ​​model running on terminal E1) when the operating system of terminal E1 changes. This ensures that an AI model compatible with the operating system can run on terminal E1 even when its operating system changes.

[0117] In the fourth possible design, the preset condition F1 may include: a change in the utilization rate of the storage resources of terminal E1.

[0118] For example, the change in the utilization rate of the storage resources of terminal E1 may specifically include: the utilization rate of the storage resources of terminal E1 is higher than the utilization rate threshold (hereinafter referred to as threshold v1), and / or, the utilization rate of the storage resources of terminal E1 is lower than the utilization rate threshold (hereinafter referred to as threshold v2).

[0119] In other words, the preset condition F1 can specifically include: the utilization rate of the storage resources of terminal E1 is higher than the threshold v1, and / or the utilization rate of the storage resources of terminal E1 is lower than the threshold v2.

[0120] The values ​​of threshold v1 and threshold v2 can be set according to the actual application, and there are no restrictions in this embodiment.

[0121] Through the above design, an AI deployment operation can be triggered when the storage resource utilization rate of terminal E1 changes, ensuring that the AI ​​model running on terminal E1 matches the storage resource utilization rate of terminal E1. For example, when the storage resource utilization rate of terminal E1 exceeds the threshold v1, the above design can trigger an AI deployment operation to switch the AI ​​model running on terminal E1 to an AI model that consumes less storage resources, or deactivate the AI ​​model running on terminal E1, thereby reducing the storage resource utilization rate of terminal E1.

[0122] For example, when the storage resource utilization rate of terminal E1 is lower than the threshold v2, the above design can trigger the execution of AI deployment operations, switch the AI ​​model running in terminal E1 to a more complex and powerful AI model, or activate more AI models in terminal E1, so that the AI ​​model running in terminal E1 matches the storage resource utilization rate of terminal E1.

[0123] In the fifth possible design, the preset condition F1 may include: a change in the utilization rate of the computing power of terminal E1.

[0124] For example, the utilization rate of computing power of terminal E1 may change, specifically including: the utilization rate of computing power of terminal E1 is higher than the utilization rate threshold (hereinafter referred to as threshold v3), and / or, the utilization rate of computing power of terminal E1 is lower than the utilization rate threshold (hereinafter referred to as threshold v4).

[0125] In other words, the preset condition F1 can specifically include: the utilization rate of the computing power of terminal E1 is higher than the threshold v3, and / or the utilization rate of the computing power of terminal E1 is lower than the threshold v4.

[0126] The values ​​of threshold v3 and threshold v4 can be set according to actual applications, and there are no restrictions in this embodiment.

[0127] Through the above design, an AI deployment operation can be triggered when the computing power utilization rate of terminal E1 changes, so that the AI ​​model running on terminal E1 matches the computing power utilization rate of terminal E1. For example, when the computing power utilization rate of terminal E1 is higher than the threshold v3, the above implementation method can trigger the execution of an AI deployment operation (e.g., switching the AI ​​model running on terminal E1 to an AI model that consumes less computing power, or deactivating the AI ​​model running on terminal E1), thereby matching the AI ​​model running on terminal E1 with the computing power utilization rate of terminal E1.

[0128] In the sixth possible design, the preset condition F1 may include: the accuracy of the AI ​​model running in terminal E1 is lower than the accuracy threshold.

[0129] With the above design, when the accuracy of the AI ​​model running in terminal E1 is lower than the accuracy threshold, an AI deployment operation can be triggered (e.g., deactivating the AI ​​model running in terminal E1 with an accuracy lower than the accuracy threshold, activating other AI models in terminal E1, or switching the AI ​​model running in terminal E1), so as to avoid affecting the performance of terminal E1 due to the accuracy of the running AI model being lower than the accuracy threshold.

[0130] It is understood that in practical applications, the content of the preset condition F1 can be one or more of the above six designs, and this disclosure embodiment does not impose any restrictions on this.

[0131] In some implementations, considering that terminal E1 can detect its own operating state to determine whether the operating state of terminal E1 meets preset conditions, as shown in Figure 3, the above S201 may specifically include:

[0132] S201a, Terminal E1 detects that the operating status of Terminal E1 meets the preset condition F1.

[0133] For example, terminal E1 can determine its operating status by monitoring the current network performance and various performance parameters of terminal E1 itself, so as to determine whether the operating status of terminal E1 meets the preset condition F1.

[0134] Specifically, S202 mentioned above may include:

[0135] S202a. When terminal E1 detects that its operating status meets the preset condition F1, it performs an AI deployment operation.

[0136] In the above implementation, terminal E1 can detect its operating status. If the operating status of terminal E1 meets the preset condition F1, an AI deployment operation can be triggered to change the AI ​​model running on terminal E1 based on its operating status. This achieves the effect of flexibly deploying AI models on terminal E1.

[0137] In some possible designs, it is considered that one or more AI models can be pre-deployed in terminal E1. This way, when the operating state of terminal E1 meets the preset condition F1, terminal E1 can directly utilize the pre-deployed AI models to perform AI deployment operations. Therefore, the AI ​​deployment operation performed in S202a above may specifically include one or more of the following:

[0138] Activate the AI ​​model (hereinafter referred to as AI model AI_1), deactivate the AI ​​model (hereinafter referred to as AI model AI_2), and switch the running AI model to another AI model (hereinafter referred to as AI model AI_3).

[0139] Among them, AI model AI_1, AI model AI_2 and AI model AI_3 are AI models pre-deployed in terminal E1.

[0140] In some possible designs, as shown in Figure 4, the method may also include:

[0141] S203, Terminal E1 sends a notification message (hereinafter referred to as notification message M1) to network node N1.

[0142] Network node N1 can be a node that performs intelligent management in the mobile communication system. For example, network node N1 can be one of the following in the core network: NWDAF, AI Control Function (AICF), or SMF.

[0143] The notification information M1 may include model information of the AI ​​model running on terminal E1. For example, the model information may include all or part of the following: model ID, model version number, model attributes, model parameter set, model capabilities, effective time of model use, and scope of model application.

[0144] S204. Based on the notification information M1, network node N1 saves and updates the model information of the AI ​​model running in terminal E1.

[0145] For example, a model information table can be maintained in network node N1. This model information table records the model information of AI models running on multiple devices (e.g., multiple terminals). Then, when network node N1 receives notification information M1, it saves and updates the model information of the AI ​​model running on terminal E1 recorded in the model information table.

[0146] S205. After saving and updating the model information of the AI ​​model running in terminal E1, network node N1 sends a notification confirmation message M2 to terminal E1.

[0147] Among them, the notification confirmation information M2 is used to indicate that network node N1 has saved and updated the model information of the AI ​​model running in terminal E1.

[0148] With the above design, after performing AI deployment operations on terminal E1, the model information of the currently running AI model on terminal E1 can be sent to network node N1, so that network node N1 can save and update the model information of the AI ​​model running on terminal E1. In this way, during subsequent equipment maintenance and other work, the model information of the AI ​​model running on terminal E1 can be conveniently obtained from network node N1.

[0149] In some possible designs, considering that when the operating state of terminal E1 meets preset condition F1, if the AI ​​model to be run is not pre-deployed in terminal E1, then in this embodiment of the disclosure, terminal E1 can also obtain the AI ​​model to be run by requesting the AI ​​model to be distributed from network node N1. Specifically, as shown in Figure 5, the above S202a may further include:

[0150] S202a1. When terminal E1 detects that the operating status of terminal E1 meets the preset condition F1, it sends a request message (hereinafter referred to as request message M3) to network node N1.

[0151] Among them, the request information M3 is used to instruct network node N1 to send the AI ​​model to terminal E1.

[0152] Specifically, network node N1 can be a node that performs intelligent management in a mobile communication system. For example, network node N1 can be one of NWDAF, AICF, or SMF in the core network.

[0153] S202a2, Network node N1 sends instruction information M4 to terminal E1 based on request information M3.

[0154] The instruction information M4 includes the AI ​​model (hereinafter referred to as AI model AI_4). Specifically, the instruction information M4 includes the executable file of AI model AI_4 and the attribute information of AI model AI_4.

[0155] S202a3 and terminal E1 execute AI deployment operations according to instruction information M4.

[0156] The AI ​​deployment operations include: activating the AI ​​model AI_4, or switching the currently running AI model to the AI ​​model AI_4.

[0157] Through the above design, when the operating state of terminal E1 meets the preset condition F1, terminal E1 can request the AI ​​model to be deployed from network node N1, thereby enabling terminal E1 to obtain the required AI model (i.e., the aforementioned AI model AI_4). This allows for the modification of the AI ​​model running on terminal E1 based on its operating state, even when the required AI model is not pre-deployed. This achieves the effect of flexibly deploying AI models on terminal E1.

[0158] In some possible implementations, the above request information M3 may include: the operating status of terminal E1.

[0159] For example, the above request information M3 may include some or all of the following information: the location of terminal E1, whether terminal E1 is in power saving mode, the operating system running on terminal E1, the utilization rate of terminal E1's storage resources, the utilization rate of terminal E1's computing power, and the accuracy of the AI ​​model currently running on terminal E1.

[0160] In this way, after receiving the request information M3, network node N1 can determine whether the operating status of terminal E1 meets the preset conditions (such as the preset condition F1 mentioned above) based on the operating status of terminal E1 carried in the request information M3. After determining that the operating status of terminal E1 meets the preset conditions, it then sends the AI ​​model to terminal E1 by executing the content of S202a2.

[0161] Furthermore, as shown in Figure 6, S202a2 may specifically include:

[0162] S202a21. After determining that the operating status of terminal E1 meets the preset conditions based on the request information M3, network node N1 sends instruction information M4 to terminal E1.

[0163] In some implementations, it is considered that the operating status of terminal E1 can be detected by a network node. Then, when the operating status of terminal E1 is detected to meet preset conditions, the network node can send an indication message to terminal E1, enabling terminal E1 to determine that its operating status meets the preset conditions. Therefore, as shown in Figure 7, the above S201 may specifically include:

[0164] S201b, Terminal E1 receives indication information (hereinafter referred to as indication information M5) from network node N1.

[0165] Among them, the indication information M5 is used to indicate that network node N1 has detected that the operating status of terminal E1 meets the preset conditions.

[0166] Specifically, network node N1 can be a node that performs intelligent management in a mobile communication system. For example, network node N1 can be one of NWDAF, AICF, or SMF in the core network.

[0167] For example, network node N1 can subscribe to network performance analysis, congestion analysis, and UE-related analysis results from network elements in the network, and then determine whether the operating status of terminal E1 meets preset conditions based on the subscribed analysis results. After determining that the operating status of terminal E1 meets the preset conditions, network node N1 sends indication information M5 to terminal E1.

[0168] In addition, the above-mentioned S202 may specifically include:

[0169] After receiving instruction information M5, S202b and terminal E1 execute AI deployment operations.

[0170] In some possible designs, the instruction information M5 may include: content related to AI deployment operations.

[0171] For example, instruction information M5 may include one or more of the following: activation, deactivation, or switching. In this way, upon receiving instruction information M5, terminal E1 can directly determine the AI ​​deployment operation to be performed based on the content of instruction information M5.

[0172] Additionally, after terminal E1 receives instruction information M5 and performs the AI ​​deployment operation, as shown in Figure 7, the method may further include:

[0173] S206, Terminal E1 sends notification information M6 to network node N1.

[0174] The notification message M6 is used to indicate that the AI ​​deployment operation has been completed in the terminal E1.

[0175] In the above implementation, network node N1 can detect the operating status of terminal E1. When network node N1 detects that the operating status of terminal E1 meets the preset condition F1, it sends an instruction message M5 to terminal E1, causing terminal E1 to trigger the execution of AI deployment operations. This achieves the goal of changing the AI ​​model running on terminal E1 based on its operating status, thereby realizing the effect of flexibly deploying AI models on terminal E1.

[0176] In some possible designs, the AI ​​deployment operations performed in S202b above may specifically include one or more of the following:

[0177] Activate the AI ​​model (hereinafter referred to as AI model AI_5), deactivate the AI ​​model (hereinafter referred to as AI model AI_6), and switch the running AI model to another AI model (hereinafter referred to as AI model AI_7).

[0178] Among them, AI model AI_5, AI model AI_6 and AI model AI_7 are AI models pre-deployed in terminal E1.

[0179] In the above design, one or more AI models can be pre-deployed in terminal E1. When terminal E1 receives instruction information M5 from network node N1, it can use the pre-deployed AI models to perform AI deployment operations, thereby changing the AI ​​model running in terminal E1 according to its operating status. This achieves the effect of flexibly deploying AI models in terminal E1.

[0180] In some possible implementations, the aforementioned instruction information M5 may include: the identifier of the AI ​​model corresponding to the AI ​​deployment operation.

[0181] For example, the aforementioned instruction information M5 may include the identifiers (e.g., model IDs) of all or some of the AI ​​models in the aforementioned AI models AI_5, AI_6, and AI_7.

[0182] In this way, after receiving the instruction information M5, the terminal E1 can determine which AI models need to be deployed based on the instruction information M5.

[0183] In some possible designs, the aforementioned instruction information M5 carries an AI model (hereinafter referred to as AI model AI_8).

[0184] The instruction information M5 carries the AI ​​model AI_8, which can be understood as: the instruction information M5 includes the executable file of the AI ​​model AI_8 and the attribute information of the AI ​​model AI_8.

[0185] In the above design, it is considered that when the running state of terminal E1 meets the preset condition F1, if the AI ​​model to be run at this time is not pre-deployed in terminal E1, then in this embodiment of the disclosure, the AI ​​model can be carried in the instruction information M5, so that the network node N1 sends the AI ​​model to terminal E1, so that terminal E1 can obtain the AI ​​model to be run.

[0186] Furthermore, as shown in Figure 8, the above S202b may specifically include:

[0187] S202b1 After receiving the instruction information M5, terminal E1 activates the AI ​​model AI_8 carried in the instruction information M5 in terminal E1, or switches the running AI model to the AI ​​model AI_8 carried in the instruction information M5 in terminal E1.

[0188] Specifically, if the instruction information M5 carries the AI ​​model AI_8, the terminal E1 can complete the AI ​​deployment operation based on the executable file of the AI ​​model AI_4 and the attribute information of the AI ​​model AI_8 included in the instruction information M5. The AI ​​deployment operation may include: activating the AI ​​model AI_8 carried in the instruction information M5, or switching the currently running AI model in the terminal E1 to the AI ​​model AI_8 carried in the instruction information M5.

[0189] In the second implementation, the method provided in this disclosure can be applied to network elements of the core network. For example, the first node in Figure 2 above can be a network element of the core network. Specifically, when the method provided in this disclosure is applied to a network element of the core network (for ease of description, the network element of the core network implementing the method provided in this disclosure will be referred to as network element E2 below), the method may include:

[0190] S301. Network element E2 determines that the operating state of network element E2 meets the preset conditions (hereinafter referred to as preset condition F2 for ease of description).

[0191] For example, network element E2 can be one of the following in the core network: SMF, AMF, UPF, PCF, UDM, AF, and AUSF.

[0192] S202 and network element E2 perform AI deployment operations when the operating status of network element E2 meets the preset condition F2.

[0193] Among them, AI deployment operation can be one of the operations that changes the AI ​​model running in the network element E2.

[0194] Specifically, the above AI deployment operations may include any one of the following: activating the AI ​​model, deactivating the AI ​​model, or switching the AI ​​model.

[0195] The following describes the contents of the above-mentioned preset condition F2 in three possible designs:

[0196] In the first possible design, the preset condition F2 may include: the network element E2 is deployed in an area where the use of AI is restricted.

[0197] With the above design, when network element E2 is deployed in an area where AI usage is restricted, an AI deployment operation can be triggered. For example, this AI deployment operation could be either deactivating the AI ​​model running in network element E2 or switching the AI ​​model running in network element E2. This allows network element E2 to operate normally in areas where AI usage is restricted.

[0198] In the second possible design, the preset condition F2 may include: a change in the load state of network element E2.

[0199] For example, the load status of network element E2 may change, specifically including: the load of network element E2 is higher than the load threshold v5, and / or the load of network element E2 is lower than the load threshold v6.

[0200] In other words, the preset condition F2 can specifically include: the load of network element E2 is higher than the load threshold v5, and / or the load of network element E2 is lower than the load threshold v6.

[0201] The values ​​of thresholds v5 and v6 can be set according to actual applications, and are not restricted in this embodiment.

[0202] Through the above design, AI deployment operations can be triggered when the load state of network element E2 changes, ensuring that the AI ​​models running in network element E2 match its load state. For example, when the load of network element E2 is higher than threshold v5, the above design can trigger AI deployment operations to switch the AI ​​models running in network element E2 to AI models with lower resource consumption, or to deactivate the AI ​​models running in network element E2, thereby matching the AI ​​models running in network element E2 with its load state. As another example, when the load of network element E2 is lower than threshold v6, the above design can trigger AI deployment operations to switch the AI ​​models running in network element E2 to AI models with more complex parameters and more powerful functions, or to activate more AI models in network element E2, thereby matching the AI ​​models running in network element E2 with its load state.

[0203] In the third possible design, the preset condition F2 may include: the accuracy of the AI ​​model running in network element E2 is lower than the accuracy threshold.

[0204] With the above design, when the accuracy of the AI ​​model running in network element E2 is lower than the accuracy threshold, an AI deployment operation can be triggered (e.g., deactivating the AI ​​model running in network element E2 with an accuracy lower than the accuracy threshold, activating other AI models in network element E2, or switching the AI ​​model running in network element E2), so as to avoid affecting the performance of network element E2 due to the accuracy of the running AI model being lower than the accuracy threshold.

[0205] It is understood that in practical applications, the content of the preset condition F2 can be one or more of the above three designs, and this disclosure embodiment does not impose any restrictions on this.

[0206] In some implementations, considering that network element E2 can detect its own operating status to determine whether the operating status of network element E2 meets preset conditions, as shown in Figure 9, the above S301 may specifically include:

[0207] S301a, Network element E2 detects that the operating status of network element E2 meets the preset condition F2.

[0208] For example, network element E2 can determine its operating status by monitoring the current network performance and its own performance, so as to determine whether the operating status of network element E2 meets the preset condition F2.

[0209] Specifically, S302 mentioned above may include:

[0210] S302a and network element E2 execute AI deployment operations when the network element E2's operating status meets the preset condition F2.

[0211] In the above implementation, network element E2 can detect its operating status. If the operating status of network element E2 meets the preset condition F2, an AI deployment operation can be triggered. This allows the AI ​​model running in network element E2 to be changed according to its operating status, thus achieving the effect of flexibly deploying AI models in network element E2.

[0212] In some possible designs, it is considered that one or more AI models can be pre-deployed in network element E2. This way, when the operating state of network element E2 meets the preset condition F2, network element E2 can directly utilize the pre-deployed AI models to perform AI deployment operations. Therefore, the AI ​​deployment operation performed in S302a above may specifically include one or more of the following:

[0213] Activate the AI ​​model (hereinafter referred to as AI model AI_9), deactivate the AI ​​model (hereinafter referred to as AI model AI_10), and switch the running AI model to another AI model (hereinafter referred to as AI model AI_11).

[0214] Among them, AI model AI_9, AI model AI_10 and AI model AI_11 are AI models pre-deployed in network element E2.

[0215] In some possible designs, as shown in Figure 10, the method may also include:

[0216] S303, network element E2 sends a notification message (hereinafter referred to as notification message M7) to network node N2.

[0217] Network node N2 can be a node that performs intelligent management in the mobile communication system. For example, network node N2 can be one of the following in the core network: NWDAF, AI Control Function (AICF), or SMF.

[0218] The notification information M7 may include model information of the AI ​​model running in the network element E2. For example, the model information may include all or part of the following: model ID, model version number, model attributes, model parameter set, model capabilities, effective time of model use, and scope of model application.

[0219] S304. Based on notification information M7, network node N2 saves and updates the model information of the AI ​​model running in network element E2.

[0220] For example, a model information table can be maintained in network node N2. This model information table records the model information of AI models running in multiple devices (e.g., multiple core network elements). Then, when network node N2 receives notification information M7, it saves and updates the model information of the AI ​​model running in network element E2 recorded in the model information table.

[0221] S305. After saving and updating the model information of the AI ​​model running in network element E2, network node N2 sends a notification confirmation message M8 to network element E2.

[0222] Among them, the notification confirmation message M8 is used to indicate that network node N2 has saved and updated the model information of the AI ​​model running in network element E2.

[0223] Through the above design, after performing AI deployment operations in network element E2, the model information of the currently running AI model in network element E2 can be sent to network node N2, so that network node N2 can save and update the model information of the AI ​​model running in network element E2. In this way, during subsequent equipment maintenance and other work, the model information of the AI ​​model running in network element E2 can be conveniently obtained from network node N2.

[0224] In some possible designs, considering that when the operating state of network element E2 meets the preset condition F2, if the AI ​​model to be run is not pre-deployed in network element E2, then in this embodiment of the disclosure, network element E2 can also obtain the AI ​​model to be run by requesting the AI ​​model from network node N2. Specifically, as shown in Figure 11, the above-mentioned S302a may further include:

[0225] S302a1, when network element E2 detects that its operating status meets the preset condition F2, it sends a request message (hereinafter referred to as request message M9) to network node N2.

[0226] Among them, the request information M9 is used to instruct network node N2 to send the AI ​​model to network element E2.

[0227] Specifically, network node N2 can be a node that performs intelligent management in a mobile communication system. For example, network node N2 can be one of NWDAF, AICF, or SMF in the core network.

[0228] S302a2, Network node N2 sends instruction information M10 to network element E2 according to request information M9.

[0229] The instruction information M10 includes the AI ​​model (hereinafter referred to as AI model AI_12). Specifically, the instruction information M10 includes the executable file of AI model AI_12 and the attribute information of AI model AI_12.

[0230] S302a3 and network element E2 execute AI deployment operations according to instruction information M10.

[0231] The AI ​​deployment operations include: activating the AI ​​model AI_12, or switching the currently running AI model to the AI ​​model AI_12.

[0232] Through the above design, when the operating state of network element E2 meets the preset condition F2, network element E2 can request the AI ​​model to be deployed from network node N2, thereby enabling network element E2 to obtain the required AI model (i.e., the aforementioned AI model AI_12). This allows for the modification of the AI ​​model running in network element E2 based on its operating state, even when the required AI model is not pre-deployed in network element E2. This achieves the effect of flexible deployment of AI models in network element E2.

[0233] In some possible implementations, the above request information M9 may include the operating status of network element E2.

[0234] For example, the above request information M9 may include some or all of the information such as the region where the network element E2 is deployed, the load status of the network element E2, and the accuracy of the AI ​​model currently running in the network element E2.

[0235] In this way, after receiving the request information M9, network node N2 can determine whether the operating status of network element E2 meets the preset conditions (such as the preset condition F2 mentioned above) based on the operating status of network element E2 carried in the request information M9. After determining that the operating status of network element E2 meets the preset conditions, it sends the AI ​​model to network element E2 by executing the content of S302a2.

[0236] Furthermore, as shown in Figure 12, S302a2 may specifically include:

[0237] S302a21. After determining that the operating status of network element E2 meets the preset conditions based on the request information M9, network node N2 sends instruction information M10 to network element E2.

[0238] In some implementations, it is considered that the operating status of network element E2 can be detected by network node N2. Furthermore, when the operating status of network element E2 is detected to meet preset conditions, network node N2 can send an indication message to network element E2, thereby enabling network element E2 to determine that its operating status meets the preset conditions. Therefore, as shown in Figure 13, the above-mentioned S301 may specifically include:

[0239] S301b, network element E2 receives indication information (hereinafter referred to as indication information M11) from network node N2.

[0240] Among them, the indication information M11 is used to indicate that the network node N2 has detected that the operating status of the network element E2 meets the preset conditions.

[0241] Specifically, network node N2 can be a node that performs intelligent management in a mobile communication system. For example, network node N2 can be one of NWDAF, AICF, or SMF in the core network.

[0242] For example, network node N2 can subscribe to network performance analysis and congestion analysis results from network elements in the network, and then determine whether the operating status of network element E2 meets preset conditions based on the subscribed analysis results. Then, after determining that the operating status of network element E2 meets the preset conditions, network node N2 sends indication information M11 to network element E2.

[0243] In addition, the above-mentioned S302 may specifically include:

[0244] After receiving instruction information M11, S302b and network element E2 execute AI deployment operations.

[0245] In some possible designs, the instruction information M11 may include the content of AI deployment operations.

[0246] For example, instruction information M11 may include one or more of the following: activation, deactivation, or switching. In this way, when network element E2 receives instruction information M11, it can directly determine the AI ​​deployment operation to be performed based on the content of instruction information M11.

[0247] In addition, after network element E2 receives instruction information M11 and performs AI deployment operation, as shown in Figure 13, the method may further include:

[0248] S306, network element E2 sends notification information M12 to network node N2.

[0249] Among them, notification information M12 is used to indicate that the AI ​​deployment operation has been completed in network element E2.

[0250] In the above implementation, network node N2 can detect the operating status of network element E2. When network node N2 detects that the operating status of network element E2 meets the preset condition F2, it sends an instruction message M11 to network element E2, causing network element E2 to trigger the execution of an AI deployment operation. This achieves the goal of changing the AI ​​model running in network element E2 based on its operating status, thereby realizing the effect of flexibly deploying AI models in network element E2.

[0251] In some possible designs, the AI ​​deployment operations performed in S302b above may specifically include one or more of the following:

[0252] Activate the AI ​​model (hereinafter referred to as AI model AI_13), deactivate the AI ​​model (hereinafter referred to as AI model AI_14), and switch the running AI model to another AI model (hereinafter referred to as AI model AI_15).

[0253] Among them, AI model AI_13, AI model AI_14 and AI model AI_15 are AI models pre-deployed in network element E2.

[0254] In the above design, one or more AI models can be pre-deployed in network element E2. When network element E2 receives instruction information M11 from network node N2, it can utilize the pre-deployed AI models to perform AI deployment operations, thereby achieving the goal of changing the AI ​​models running in network element E2 according to its operational status. This enables flexible deployment of AI models in network element E2.

[0255] In some possible implementations, the aforementioned instruction information M11 may include: the identifier of the AI ​​model corresponding to the AI ​​deployment operation.

[0256] For example, the aforementioned instruction information M11 may include the identifiers (e.g., model IDs) of all or some of the AI ​​models in the aforementioned AI models AI_13, AI_14, and AI_15.

[0257] In this way, after receiving the instruction information M11, the network element E2 can determine which AI models need to be deployed based on the instruction information M11.

[0258] In some possible designs, the aforementioned instruction information M11 carries an AI model (hereinafter referred to as AI model AI_16).

[0259] The instruction information M11 carries the AI ​​model AI_16, which can be understood as: the instruction information M11 includes the executable file of the AI ​​model AI_16 and the attribute information of the AI ​​model AI_16.

[0260] In the above design, it is considered that when the operating state of network element E2 meets the preset condition F2, if the AI ​​model to be run at this time is not pre-deployed in network element E2, then in this embodiment of the disclosure, the AI ​​model can be carried in the instruction information M11, so that network node N2 sends the AI ​​model to network element E2, so that network element E2 can obtain the AI ​​model to be run.

[0261] Furthermore, as shown in Figure 14, the above-mentioned S302b may specifically include:

[0262] After receiving the instruction information M11, S302b1 and network element E2 activate the AI ​​model AI_16 carried in the instruction information M11 in network element E2, or switch the running AI model in network element E2 to the AI ​​model AI_16 carried in the instruction information M11.

[0263] Specifically, when the instruction information M11 carries the AI ​​model AI_16, network element E2 can complete the AI ​​deployment operation based on the executable file and attribute information of the AI ​​model AI_16 included in the instruction information M11. The AI ​​deployment operation may include: activating the AI ​​model AI_16 carried in the instruction information M11, or switching the currently running AI model in network element E2 to the AI ​​model AI_16 carried in the instruction information M11.

[0264] In the third implementation, the method provided in this disclosure can be applied to an access network device. For example, the first node in Figure 2 above can be an access network device. Specifically, when the method provided in this disclosure is applied to an access network device (for ease of description, the access network device implementing the method provided in this disclosure will be referred to as access network device E3 below), the method may include:

[0265] S401. Access network device E3 determines that the operating status of access network device E3 meets the preset conditions (hereinafter referred to as preset condition F3 for ease of description).

[0266] S402. When the operating status of access network device E3 meets the preset condition F3, the AI ​​deployment operation is performed.

[0267] Among them, AI deployment operation can be one of the operations that changes the AI ​​model running in the access network device E3.

[0268] Specifically, the above AI deployment operations may include: activating the AI ​​model, deactivating the AI ​​model, or switching the AI ​​model.

[0269] The following describes the contents of the aforementioned preset condition F3 in four possible designs:

[0270] In the first possible design, the preset condition F3 may include: a change in the utilization rate of air interface resources of the access network device E3.

[0271] For example, a change in the utilization rate of air interface resources of access network device E3 may specifically include: the utilization rate of air interface resources of access network device E3 being higher than the threshold v7, and / or the utilization rate of air interface resources of access network device E3 being lower than the threshold v8.

[0272] In other words, the preset condition F3 can specifically include: the utilization rate of air interface resources of access network device E3 is higher than the threshold v7, and / or the utilization rate of air interface resources of access network device E3 is lower than the threshold v8.

[0273] Through the above design, AI deployment operations can be triggered when the air interface resources of access network device E3 change, ensuring that the AI ​​models running in access network device E3 match the utilization rate of its air interface resources. For example, when the utilization rate of access network device E3's air interface resources is higher than threshold v7, the above design can trigger AI deployment operations to switch the AI ​​models running in access network device E3 to AI models with lower resource consumption, or deactivate the AI ​​models running in access network device E3, thereby matching the AI ​​models running in access network device E3 with the utilization rate of its air interface resources. As another example, when the load of access network device E3 is lower than threshold v8, the above design can trigger AI deployment operations to switch the AI ​​models running in access network device E3 to AI models with more complex parameters and more powerful functions, or activate more AI models in access network device E3, thereby matching the AI ​​models running in access network device E3 with the utilization rate of its air interface resources.

[0274] In the second possible design, the preset condition F3 may include: a change in the load state of the access network device E3.

[0275] For example, a change in the load state of access network device E3 may specifically include: the load of access network device E3 being higher than the load threshold v9, and / or the load of access network device E3 being lower than the load threshold v10.

[0276] In other words, the preset condition F2 can specifically include: the load of access network device E3 is higher than the load threshold v9, and / or the load of access network device E3 is lower than the load threshold v10.

[0277] The values ​​of thresholds v9 and v10 can be set according to actual applications, and are not restricted in this embodiment.

[0278] Through the above design, AI deployment operations can be triggered when the load state of access network device E3 changes, ensuring that the AI ​​models running in access network device E3 match its load state. For example, when the load of access network device E3 is higher than threshold v9, the above design can trigger AI deployment operations to switch the AI ​​models running in access network device E3 to AI models with lower resource consumption, or to deactivate the AI ​​models running in access network device E3, thereby matching the AI ​​models running in access network device E3 with its load state. As another example, when the load of access network device E3 is lower than threshold v10, the above design can trigger AI deployment operations to switch the AI ​​models running in access network device E3 to AI models with more complex parameters and more powerful functions, or to activate more AI models in access network device E3, thereby matching the AI ​​models running in access network device E3 with the utilization rate of its storage resources.

[0279] In the third possible design, the preset condition F3 may include: a change in the utilization rate of the computing power of the access network device E3.

[0280] For example, the utilization rate of the computing power of access network device E3 may change, specifically including: the utilization rate of the computing power of access network device E3 is higher than the utilization rate threshold (hereinafter referred to as threshold v11), and / or, the utilization rate of the computing power of access network device E3 is lower than the utilization rate threshold (hereinafter referred to as threshold v12).

[0281] In other words, the preset condition F3 may specifically include: the utilization rate of the computing power of the access network device E3 is higher than the utilization rate threshold (hereinafter referred to as threshold v11), and / or, the utilization rate of the computing power of the access network device E3 is lower than the utilization rate threshold (hereinafter referred to as threshold v12).

[0282] The values ​​of threshold v11 and threshold v12 can be set according to the actual application, and there are no restrictions in this embodiment.

[0283] Through the above design, when the utilization rate of the computing power of access network device E3 changes, an AI deployment operation can be triggered to match the AI ​​model running in access network device E3 with its computing power utilization rate. For example, when the utilization rate of the computing power of access network device E3 is higher than the threshold v11, the above implementation method can trigger an AI deployment operation (e.g., switching the AI ​​model running in access network device E3 to an AI model that consumes less computing power, or deactivating the AI ​​model running in access network device E3), thereby matching the AI ​​model running in access network device E3 with its computing power utilization rate. As another example, when the utilization rate of the computing power of access network device E3 is lower than the threshold v12, the above implementation method can trigger an AI deployment operation (e.g., switching the AI ​​model running in access network device E3 to an AI model with more complex parameters and more powerful functions, or activating more AI models in access network device E3), thereby matching the AI ​​model running in access network device E3 with its computing power utilization rate.

[0284] In the fourth possible design, the preset condition F3 may include: the accuracy of the AI ​​model running in the access network device E3 is lower than the accuracy threshold.

[0285] With the above design, when the accuracy of the AI ​​model running in the access network device E3 is lower than the accuracy threshold, an AI deployment operation can be triggered (e.g., deactivating the AI ​​model running in the access network device E3 whose accuracy is lower than the accuracy threshold, activating other AI models in the access network device E3, or switching the AI ​​model running in the access network device E3), so as to avoid affecting the performance of the access network device E3 due to the accuracy of the running AI model being lower than the accuracy threshold.

[0286] It is understood that in practical applications, the content of the preset condition F3 can be one or more of the above four designs, and this disclosure embodiment does not impose any restrictions on this.

[0287] In some implementations, it is considered that: access network device E3 can detect its own operating status to determine whether the operating status of access network device E3 meets preset conditions. Therefore, as shown in Figure 15, the above S401 may specifically include:

[0288] S401a, Access network device E3 detects that the operating status of access network device E3 meets the preset condition F3.

[0289] For example, access network device E3 can determine its operating status by monitoring the current network performance and various performance parameters of access network device E3 itself, so as to determine whether the operating status of access network device E3 meets the preset condition F3.

[0290] Specifically, S402 mentioned above may include:

[0291] S402a. When the access network device E3 detects that its operating status meets the preset condition F3, it performs an AI deployment operation.

[0292] In the above implementation, the access network device E3 can detect its operating status. If the operating status of access network device E3 meets the preset condition F3, an AI deployment operation can be triggered. This allows the AI ​​model running in access network device E3 to be modified according to its operating status, thus achieving the effect of flexibly deploying AI models in access network device E3.

[0293] In some possible designs, it is considered that one or more AI models can be pre-deployed in the access network device E3. This way, when the operating state of the access network device E3 meets the preset condition F3, the access network device E3 can directly utilize the pre-deployed AI models to perform AI deployment operations. Therefore, the AI ​​deployment operation performed in S402a above may specifically include one or more of the following:

[0294] Activate the AI ​​model (hereinafter referred to as AI model AI_17), deactivate the AI ​​model (hereinafter referred to as AI model AI_18), and switch the running AI model to another AI model (hereinafter referred to as AI model AI_19).

[0295] Among them, AI model AI_17, AI model AI_18 and AI model AI_19 are AI models pre-deployed in access network device E3.

[0296] In some possible designs, as shown in Figure 16, the method may also include:

[0297] S403, Access network device E3 sends a notification message (hereinafter referred to as notification message M13) to network node N3.

[0298] Among them, network node N3 can be a node that manages access network nodes in a mobile communication system.

[0299] For example, network node N3 can be one of the following: Operation Administration and Maintenance (OAM), Service Management and Orchestration (SMO), and AICF.

[0300] The notification information M13 may include model information of the AI ​​model running in the access network device E3. For example, the model information may include all or part of the following: model ID, model version number, model attributes, model parameter set, model capabilities, effective time of model use, and scope of model application.

[0301] S404, Network node N3 saves and updates the model information of the AI ​​model running in access network device E3 according to notification information M13.

[0302] For example, a model information table can be maintained in network node N3. This model information table records the model information of AI models running in multiple devices (e.g., multiple access network devices). Then, when network node N3 receives notification information M13, network node N2 saves and updates the model information of the AI ​​model running in access network device E3 recorded in the model information table.

[0303] S405 After saving and updating the model information of the AI ​​model running in the access network device E3, the network node N3 sends a notification confirmation message M14 to the access network device E3.

[0304] Among them, the notification confirmation message M14 is used to indicate that network node N3 has saved and updated the model information of the AI ​​model running in access network device E3.

[0305] Through the above design, after performing AI deployment operations in access network device E3, the model information of the currently running AI model in access network device E3 can be sent to network node N3, so that network node N3 can save and update the model information of the AI ​​model running in access network device E3. This allows for convenient retrieval of the model information of the AI ​​model running in access network device E3 from network node N3 during subsequent equipment maintenance and other tasks.

[0306] In some possible designs, considering that when the operating state of access network device E3 meets preset condition F3, if the AI ​​model to be run is not pre-deployed in access network device E3, then in this embodiment of the disclosure, access network device E3 can also obtain the AI ​​model to be run by requesting the AI ​​model from network node N3. Specifically, as shown in Figure 17, the above-mentioned S402a may further include:

[0307] S402a1. When the access network device E3 detects that the operating status of the access network device E3 meets the preset condition F3, it sends a request message (hereinafter referred to as request message M15) to the network node N3.

[0308] Among them, request information M15 is used to instruct network node N3 to send the AI ​​model to access network device E3.

[0309] For example, network node N3 can be one of the following: Operation Administration and Maintenance (OAM), Service Management and Orchestration (SMO), and AICF.

[0310] S402a2, Network Node N3 sends Instruction Information M16 to Access Network Device E3 based on Request Information M15.

[0311] The instruction information M16 includes the AI ​​model (hereinafter referred to as AI model AI_20). Specifically, the instruction information M16 includes the executable file of AI model AI_20 and the attribute information of AI model AI_20.

[0312] S402a3 and access network device E3 perform AI deployment operations according to instruction information M16.

[0313] The AI ​​deployment operations include: activating the AI ​​model AI_20, or switching the currently running AI model to the AI ​​model AI_20.

[0314] Through the above design, when the operating state of access network device E3 meets the preset condition F3, access network device E3 can request the AI ​​model to be deployed from network node N3, thereby enabling access network device E3 to obtain the required AI model (i.e., the aforementioned AI model AI_20). This allows for the modification of the AI ​​model running in access network device E3 based on its operating state, even when the required AI model is not pre-deployed in access network device E3. This achieves the effect of flexible deployment of AI models in access network device E3.

[0315] In some possible implementations, the above request information M15 may include: the operating status of access network device E3.

[0316] For example, the above request information M15 may include some or all of the following information: the utilization rate of air interface resources of access network device E3, the load status of access network device E3, the utilization rate of computing power of access network device E3, and the accuracy of AI models running in access network device E3.

[0317] In this way, after receiving the request information M15, network node N3 can determine whether the operating status of access network device E3 meets the preset conditions (such as the preset condition F3 mentioned above) based on the operating status of access network device E3 carried in the request information M15. After determining that the operating status of access network device E3 meets the preset conditions, it then sends the AI ​​model to access network device E3 by executing the content of S402a2.

[0318] Furthermore, as shown in Figure 18, S402a2 may specifically include:

[0319] S402a21. After determining that the operating status of access network device E3 meets the preset conditions based on request information M15, network node N3 sends instruction information M16 to access network device E3.

[0320] In some implementations, it is considered that the network node N3 can detect the operating status of the access network device E3. Then, when the operating status of the access network device E3 is detected to meet preset conditions, the network node N3 can send an indication message to the access network device E3, thereby enabling the access network device E3 to determine that its operating status meets the preset conditions. Therefore, as shown in Figure 19, the above S401 may specifically include:

[0321] S401b, Access network device E3 receives indication information (hereinafter referred to as indication information M17) from network node N3.

[0322] Among them, the indication information M17 is used to indicate that the network node N3 has detected that the operating status of the access network device E3 meets the preset conditions.

[0323] For example, network node N3 can subscribe to network performance analysis and congestion analysis results from network elements in the network, and then determine whether the operating status of access network device E3 meets preset conditions based on the subscribed analysis results. Then, after determining that the operating status of access network device E3 meets the preset conditions, network node N3 sends indication information M17 to access network device E3.

[0324] In addition, the above-mentioned S402 may specifically include:

[0325] After receiving instruction information M17, S402b and access network device E3 execute AI deployment operations.

[0326] In some possible designs, the instruction information M17 may include: content related to AI deployment operations.

[0327] For example, instruction information M17 may include one or more of the following: activation, deactivation, or switching. In this way, when access network device E3 receives instruction information M17, it can directly determine the AI ​​deployment operation to be performed based on the content of instruction information M17.

[0328] Additionally, after the access network device E3 receives the instruction information M17 and performs the AI ​​deployment operation, as shown in Figure 19, the method may further include:

[0329] S406. Access network device E3 sends notification information M18 to network node N3.

[0330] The notification message M18 is used to indicate that the AI ​​deployment operation has been completed in the access network device E3.

[0331] In the above implementation, network node N3 can detect the operating status of access network device E3. When network node N3 detects that the operating status of access network device E3 meets preset condition F3, it sends indication information M17 to access network device E3, causing access network device E3 to trigger the execution of AI deployment operations. This achieves the goal of changing the AI ​​model running in access network device E3 based on its operating status, thereby realizing the effect of flexibly deploying AI models in access network device E3.

[0332] In some possible designs, the AI ​​deployment operations performed in S402b above may specifically include one or more of the following:

[0333] Activate the AI ​​model (hereinafter referred to as AI model AI_21), deactivate the AI ​​model (hereinafter referred to as AI model AI_22), and switch the running AI model to another AI model (hereinafter referred to as AI model AI_23).

[0334] Among them, AI model AI_21, AI model AI_22 and AI model AI_23 are AI models pre-deployed in network element E2.

[0335] In the above design, one or more AI models can be pre-deployed in the access network device E3. When access network device E3 receives instruction information M17 from network node N3, it can utilize the pre-deployed AI models to perform AI deployment operations. This allows the AI ​​models running in access network device E3 to be changed according to its operating status, thus achieving the effect of flexible deployment of AI models in access network device E3.

[0336] In some possible implementations, the aforementioned instruction information M17 may include: the identifier of the AI ​​model corresponding to the AI ​​deployment operation.

[0337] For example, the aforementioned instruction information M17 may include the identifiers (e.g., model IDs) of all or some of the AI ​​models in the aforementioned AI models AI_21, AI_22, and AI_23.

[0338] In this way, after receiving the instruction information M17, the access network device E3 can determine which AI models need to be deployed based on the instruction information M17.

[0339] In some possible designs, the aforementioned instruction information M17 carries an AI model (hereinafter referred to as AI model AI_24).

[0340] The instruction information M17 contains the AI ​​model AI_24, which can be understood as: the instruction information M17 includes the executable file of the AI ​​model AI_24 and the attribute information of the AI ​​model AI_24.

[0341] In the above design, it is considered that when the operating state of the access network device E3 meets the preset condition F3, if the AI ​​model to be run at this time is not pre-deployed in the access network device E3, then in this embodiment of the disclosure, the AI ​​model can be carried in the instruction information M17, so that the network node N3 sends the AI ​​model to the access network device E3, so that the access network device E3 can obtain the AI ​​model to be run.

[0342] Furthermore, as shown in Figure 20, the above S402b may specifically include:

[0343] S402b1. After receiving the instruction information M17, the access network device E3 activates the AI ​​model AI_24 carried in the instruction information M17, or switches the currently running AI model to the AI ​​model AI_24 carried in the instruction information M11.

[0344] Specifically, if the instruction information M17 carries the AI ​​model AI_24, the access network device E3 can complete the AI ​​deployment operation based on the executable file and attribute information of the AI ​​model AI_24 included in the instruction information M17. The AI ​​deployment operation may include: activating the AI ​​model AI_24 carried in the instruction information M17, or switching the currently running AI model in the access network device E3 to the AI ​​model AI_24 carried in the instruction information M17.

[0345] Figures 2-20 above mainly describe the method provided in this embodiment from the perspective of the first node deploying the AI ​​model (wherein the first node can be terminal E1, core network element E2, or access network device E3).

[0346] The method provided in this disclosure embodiment will now be described from the perspective of the node that manages the first node (hereinafter referred to as the second node). As shown in Figure 21, the method may include:

[0347] S501, The second node obtains the running status of the first node.

[0348] The first node can be any one of the following: a terminal, an access network device, or a core network element. For example, the first node can be any one of the following in Figures 3-20: terminal E1, core network element E2, or access network device E3.

[0349] The second node can be any one of network node N1, network node N2, or network node N3 in Figures 3-20 above.

[0350] The specific implementation process of S501 can be seen in the descriptions of Figures 7, 8, 13, 14, 19 and 20, which describe the specific implementation process of network node N1 detecting the operating status of terminal E1, network node N2 detecting the operating status of network element E2, or network node N3 detecting the operating status of access network device E3.

[0351] Alternatively, the specific implementation process of S501 can be described with reference to Figures 6, 12 and 18, which describe the specific implementation process of network node N1 determining the operating status of terminal E1 based on request information M3, network node N2 determining the operating status of network element E2 based on request information M9, and network node N3 determining the operating status of access network device E3 based on request information M15.

[0352] S502. After the second node obtains that the running status of the first node meets the preset conditions, it sends the first instruction information to the first node.

[0353] The first instruction information is used to instruct the first node to perform AI deployment operations.

[0354] AI deployment operations include operations that modify the AI ​​model running in the first node.

[0355] For example, the first instruction information can specifically be instruction information M5 as described in Figures 7 and 8 above. Upon receiving instruction information M5, terminal E1 performs an AI deployment operation; that is, instruction information M5 instructs terminal E1 to perform the AI ​​deployment operation.

[0356] For example, the first instruction information can also be the instruction information M11 described in Figures 13 and 14 above. Upon receiving the instruction information M11, network element E2 performs the AI ​​deployment operation; that is, the instruction information M11 instructs network element E2 to perform the AI ​​deployment operation.

[0357] For example, the first instruction information can also be the instruction information M17 described in Figures 19 and 20 above. Upon receiving the instruction information M17, the access network device E3 performs the AI ​​deployment operation; that is, the instruction information M17 instructs the access network device E3 to perform the AI ​​deployment operation.

[0358] For example, the first instruction information can also be the request information M3 in the description corresponding to Figure 6 above.

[0359] For example, the first instruction information can also be the request information M9 in the description corresponding to Figure 12 above.

[0360] For example, the first instruction information can also be the request information M15 in the description corresponding to Figure 18 above.

[0361] The specific implementation process of S502 can be referred to the specific implementation processes of S201b, S301b, and S401b in Figures 7, 8, 13, 14, 19, and 20. Alternatively, the specific implementation process of S502 can be referred to the specific implementation processes of S202a21, S302a21, and S402a21 in Figures 6, 12, and 18.

[0362] In some implementations, the first instruction information is specifically used to instruct the first node to perform one or more of the following: activate the first AI model, deactivate the second AI model, or switch the running AI model to the third AI model.

[0363] Among them, the first AI model, the second AI model, and the third AI model are AI models pre-deployed in the first node.

[0364] For example, the first AI model, the second AI model, and the third AI model mentioned above can be AI model AI_5, AI model AI_6, and AI model AI_7 in S202b, or AI model AI_13, AI model AI_14, and AI model AI_15 in S302b, or AI model AI_21, AI model AI_22, and AI model AI_23 in S402b.

[0365] In some implementations, the first instruction information includes a fourth AI model.

[0366] Specifically, the first instruction information is used to instruct the first node to activate the fourth AI model in the first node, or to switch the running AI model in the first node to the fourth AI model.

[0367] For example, the fourth AI model mentioned above can be: AI model AI_8 in the description corresponding to Figure 8, or AI model AI_16 in the description corresponding to Figure 14, or AI model AI_24 in the description corresponding to Figure 20.

[0368] In some implementations, S501 specifically includes:

[0369] S501a, The second node receives status information from the first node.

[0370] The status information includes the running status of the first node.

[0371] For example, the aforementioned status information could be the request information M3 for the operating status of the bearer terminal E1 mentioned above, or the request information M9 for the operating status of the bearer network element E2, or the request information M15 for the operating status of the bearer access network device E3.

[0372] The specific implementation process of S501a can be referred to the process in S202a1, S302a1 and S402a1 above, in which network node N1, network node N2 and network node N3 respectively obtain request information M3, request information M9 and request information M15.

[0373] In some implementations, S501 specifically includes:

[0374] S501b: The second node detects the operating status of the first node.

[0375] The specific implementation process of S501b can be seen in Figures 7, 8, 13, 14, 19 and 20, which describe the process by which network node N1, network node N2 and network node N3 respectively detect the operating status of terminal E1, network element E2 and access network device E3.

[0376] In some implementations, when the first node is a terminal, preset conditions include: the first node is moved to an area where AI usage is restricted, the first node is in power-saving mode, the operating system of the first node is changed, the utilization rate of the storage resources of the first node is changed, the utilization rate of the computing power of the first node is changed, and the accuracy of the AI ​​model running in the first node is lower than an accuracy threshold, or one or more of these conditions.

[0377] Alternatively, in some implementations, when the first node is a network element of the core network, preset conditions are included, such as: the first node is deployed to an area where the use of AI is restricted, the load state of the first node changes, and the accuracy of the AI ​​model running in the first node is lower than an accuracy threshold, or one or more of these conditions.

[0378] Alternatively, in some implementations, when the first node is an access network device, preset conditions are included, such as: changes in the utilization rate of the first node's air interface resources, changes in the first node's load status, changes in the utilization rate of the first node's computing power, and one or more of the following: the accuracy of the AI ​​model running in the first node is lower than the accuracy threshold.

[0379] Additionally, as shown in Figure 22, the method may also include:

[0380] S601, The second node receives the request information from the first node.

[0381] The request information is used to instruct the first node to send the AI ​​model.

[0382] The first node can be any one of the following: a terminal, an access network device, or a core network element.

[0383] The specific implementation process of S601 can be referred to the specific implementation processes of S202a1, S302a1 and S402a1 mentioned above.

[0384] S602. The second node sends the first instruction information to the first node based on the request information.

[0385] The first instruction information includes the first AI model.

[0386] The first instruction information is used to instruct the first node to activate the second AI model in the first node, or to switch the running AI model in the first node to the second AI model.

[0387] The specific implementation process of S602 can be referred to the specific implementation processes of S202a2, S302a2 and S402a2 mentioned above.

[0388] In some implementations, S602 specifically includes:

[0389] S602a. After the second node determines that the operating status of the first node meets the preset conditions based on the request information, it sends the first instruction information to the first node.

[0390] The request information includes the running status of the first node.

[0391] The specific implementation process of S602a can be referred to the specific implementation processes of S202a21, S302a21 and S402a21 mentioned above.

[0392] In some implementations, the method may also include: the second node detecting the running status of the first node.

[0393] S602 may specifically include:

[0394] S602b: After determining that the operating status of the first node meets the preset conditions, the second node sends the first instruction information to the first node according to the request information.

[0395] For example, when the second node is network node N1 mentioned above, network node N1 can detect the operating status of terminal E1. Then, after network node N1 receives request information M3 through S202a1 in Figure 5, network node N1 can determine whether the operating status of the first node meets the preset condition F1 based on the detection result. And when it is determined that the operating status of network node N1 meets the preset condition, it sends instruction information M4 (i.e., first instruction information) to terminal E1 according to request information M3.

[0396] Figure 23 is a schematic diagram of the structure of a communication device provided in an embodiment of this disclosure.

[0397] On the one hand, the communication device can be used to perform some or all of the steps of the first node in Figure 2-22 above (wherein the first node can be the terminal E1, network element E2 or access network device E3 above).

[0398] On the other hand, the communication device can be used to perform some or all of the steps of the second node in Figure 2-22 (wherein the second node can be the network node N1, network node N2 or network node N3).

[0399] The communication device may include some or all of the components in the memory 701, transceiver 702, processor 703, and bus interface 704.

[0400] The transceiver 702 is used to receive and send data under the control of the processor 703.

[0401] In Figure 23, the bus architecture may include any number of interconnected buses and bridges, specifically linking various circuits of one or more processors represented by processor 703 and memory represented by memory 701. The bus architecture may also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein.

[0402] In the bus architecture, the bus interface 704 is used to provide the interface.

[0403] The transceiver 702 may be a combination of multiple components, including a transmitter and a receiver, providing a unit for communicating with various other devices over a transmission medium, including wireless channels, wired channels, optical fibers, and other transmission media.

[0404] The processor 703 is responsible for managing the bus architecture and general processing, while the memory 701 can store the data used by the processor 703 when performing operations.

[0405] The processor 703 can be a central processing unit (CPU), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a complex programmable logic device (CPLD). The processor can also adopt a multi-core architecture.

[0406] The processor 703 executes any of the methods provided in the embodiments of this disclosure according to the obtained executable instructions by calling a program stored in memory. The processor 703 and the memory 701 may also be physically separated.

[0407] In the first aspect, when the communication device 70 is used to perform all or part of the steps of the first node described above, the memory 701 is used to store a computer program. The transceiver 702 is used to transmit and receive data under the control of the processor 703. The processor 703 is used to read the computer program in the memory 701 and perform the following operations:

[0408] Determine that the operating status of the first node meets the preset conditions.

[0409] If the operating status of the first node meets preset conditions, an AI deployment operation is executed. The AI ​​deployment operation is one of the operations that changes the AI ​​model running in the first node. The first node can be any one of a terminal, access network device, or core network element.

[0410] In some implementations, determining that the running state of the first node meets preset conditions includes: detecting that the running state of the first node meets preset conditions.

[0411] If the running status of the first node meets the preset conditions, perform the AI ​​deployment operation, including: if the running status of the first node is detected to meet the preset conditions, perform the AI ​​deployment operation.

[0412] In some implementations, AI deployment operations include one or more of the following: activating a first AI model, deactivating a second AI model, and switching a running AI model to a third AI model. The first, second, and third AI models are AI models pre-deployed in the first node.

[0413] In some implementations, AI deployment operations are executed when the running status of the first node meets preset conditions, including:

[0414] If the operating status of the first node meets the preset conditions, a request message is sent to the second node; the request message is used to instruct the second node to send the AI ​​model to the first node.

[0415] Receive the first instruction information sent by the second node according to the request information; the first instruction information includes the fourth AI model.

[0416] Based on the first instruction, perform the AI ​​deployment operation; the AI ​​deployment operation includes: activating the fourth AI model, or switching the currently running AI model to the fourth AI model.

[0417] In some implementations, the request information includes the running status of the first node.

[0418] In some implementations, the running state of the first node is determined to meet preset conditions, including:

[0419] The system receives a second indication from a second node; this second indication indicates that the second node has detected that the operating status of the first node meets preset conditions. If the operating status of the first node meets the preset conditions, the system performs an AI deployment operation, including: performing the AI ​​deployment operation after receiving the second indication.

[0420] In some implementations, AI deployment operations include one or more of the following: activating the fifth AI model, deactivating the sixth AI model, and switching the currently running AI model to the seventh AI model. The fifth, sixth, and seventh AI models are pre-deployed in the first node.

[0421] In some implementations, the second instruction information carries the eighth AI model.

[0422] Upon receiving the second instruction information, perform AI deployment operations, including: activating the eighth AI model carried in the second instruction information in the first node, or switching the running AI model in the first node to the eighth AI model carried in the second instruction information.

[0423] In some implementations, the first node is a terminal; preset conditions include one or more of the following: the first node is moved to an area where AI usage is restricted, the first node is in power-saving mode, the operating system of the first node is changed, the utilization rate of the storage resources of the first node is changed, the utilization rate of the computing power of the first node is changed, and the accuracy of the AI ​​model running in the first node is lower than the accuracy threshold.

[0424] In some implementations, the first node is a network element of the core network; preset conditions include one or more of the following: the first node is deployed to an area where the use of AI is restricted, the load state of the first node changes, and the accuracy of the AI ​​model running in the first node is lower than an accuracy threshold.

[0425] In some implementations, the first node is an access network device; preset conditions include one or more of the following: changes in the utilization rate of the air interface resources of the first node, changes in the load status of the first node, changes in the utilization rate of the computing power of the first node, and the accuracy of the AI ​​model running in the first node being lower than the accuracy threshold.

[0426] In the second aspect, when the communication device 70 is used to execute all or part of the steps of the second node described above, the memory 701 is used to store a computer program. The transceiver 702 is used to send and receive data under the control of the processor 703. The processor 703 is used to read the computer program in the memory 701 and perform the following operations:

[0427] Obtain the operating status of the first node. The first node can be any one of the following: a terminal, an access network device, or a core network element.

[0428] After obtaining the running status of the first node and finding that it meets the preset conditions, a first instruction message is sent to the first node; the first instruction message is used to instruct the first node to perform AI deployment operations; wherein, the AI ​​deployment operations include operations that change the AI ​​model running in the first node.

[0429] In some implementations, the first instruction information is specifically used to instruct the first node to perform one or more of the following: activate the first AI model, deactivate the second AI model, and switch the running AI model to the third AI model. The first, second, and third AI models are AI models pre-deployed in the first node.

[0430] In some implementations, the first instruction information includes a fourth AI model. Specifically, the first instruction information is used to instruct the first node to activate the fourth AI model within the first node, or to switch the currently running AI model to the fourth AI model within the first node.

[0431] In some implementations, the running status of the first node is obtained, including:

[0432] Receive status information from the first node; the status information includes the running status of the first node.

[0433] In some implementations, the running status of the first node is obtained, including:

[0434] Detect the running status of the first node.

[0435] In some implementations, when the first node is a terminal, preset conditions include: the first node is moved to an area where AI usage is restricted, the first node is in power-saving mode, the operating system of the first node is changed, the utilization rate of the storage resources of the first node is changed, the utilization rate of the computing power of the first node is changed, and the accuracy of the AI ​​model running in the first node is lower than an accuracy threshold, or one or more of these conditions.

[0436] Alternatively, if the first node is a network element in the core network, preset conditions may be set, including: the first node is deployed to an area where AI usage is restricted, the load status of the first node changes, and the accuracy of the AI ​​model running in the first node is lower than an accuracy threshold, or one or more of these conditions may be met.

[0437] Alternatively, if the first node is an access network device, preset conditions may be set, including: changes in the utilization rate of the first node's air interface resources, changes in the first node's load status, changes in the utilization rate of the first node's computing power, and one or more of the following: the accuracy of the AI ​​model running in the first node is lower than the accuracy threshold.

[0438] In the third aspect, when the communication device 70 is used to execute all or part of the steps of the second node described above, the memory 701 is used to store the computer program. The transceiver 702 is used to send and receive data under the control of the processor 703. The processor 703 is used to read the computer program in the memory 701 and perform the following operations:

[0439] Receive a request message from the first node. The request message instructs the first node to send the AI ​​model; the first node can be any one of a terminal, access network device, or core network element.

[0440] Based on the request information, a first instruction message is sent to the first node; the first instruction message includes a first AI model; the first instruction message is used to instruct the first node to activate a second AI model in the first node, or to switch the running AI model in the first node to the second AI model.

[0441] In some implementations, the request information includes the running status of the first node; according to the request information, sending the first instruction information to the first node includes: after determining that the running status of the first node meets the preset conditions according to the request information, sending the first instruction information to the first node.

[0442] In some implementations, the processor 703 is also used to detect the operating status of the first node. Based on the request information, it sends first indication information to the first node, including:

[0443] After determining that the operating status of the first node meets the preset conditions, the first instruction information is sent to the first node according to the request information.

[0444] Based on the same technical concept, this disclosure also provides a communication device. This communication device can implement the function of the first node in the aforementioned embodiments. Referring to FIG24, it is a structural schematic diagram of a communication device provided in an embodiment of this disclosure.

[0445] As shown in the figure, the communication device 80 may include: a determining unit 801 and a processing unit 802. Wherein:

[0446] The determining unit 801 is used to determine whether the operating status of the first node meets the preset conditions.

[0447] The processing unit 802 is used to perform AI deployment operations when the running status of the first node meets preset conditions;

[0448] Among them, AI deployment operation is one of the operations that changes the AI ​​model running in the first node; the first node is any one of the terminal, access network equipment or core network element.

[0449] In some implementations, the determining unit 801 is used to determine whether the operating state of the first node meets preset conditions, including:

[0450] The determining unit 801 is used to detect that the operating status of the first node meets preset conditions;

[0451] Processing unit 802 is used to execute AI deployment operations when the running status of the first node meets preset conditions, including:

[0452] The processing unit 802 is used to perform AI deployment operations when the operating status of the first node meets preset conditions.

[0453] In some implementations, AI deployment operations include one or more of the following: activating a first AI model, deactivating a second AI model, or switching a running AI model to a third AI model.

[0454] Among them, the first AI model, the second AI model, and the third AI model are AI models pre-deployed in the first node.

[0455] In some implementations, the processing unit 802 is used to perform AI deployment operations when the running status of the first node meets preset conditions, including:

[0456] The processing unit 802 is used to send a request message to the second node when the operating status of the first node meets the preset conditions; wherein the request message is used to instruct the second node to send an AI model to the first node.

[0457] Processing unit 802 is used to receive first instruction information sent by the second node according to the request information; the first instruction information includes a fourth AI model;

[0458] The processing unit 802 is used to perform AI deployment operations according to the first instruction information; the AI ​​deployment operations include: activating the fourth AI model, or switching the running AI model to the fourth AI model.

[0459] In some implementations, the request information includes the running status of the first node.

[0460] In some implementations, the determining unit 801 is used to determine whether the operating state of the first node meets preset conditions, including:

[0461] The determining unit 801 is used to receive second indication information from the second node; the second indication information is used to indicate that the second node has detected that the operating status of the first node meets preset conditions;

[0462] Processing unit 802 is used to execute AI deployment operations when the running status of the first node meets preset conditions, including:

[0463] The processing unit 802 is used to perform AI deployment operations after receiving the second instruction information.

[0464] In some implementations, AI deployment operations include one or more of the following: activating the fifth AI model, deactivating the sixth AI model, or switching the running AI model to the seventh AI model.

[0465] Among them, the fifth AI model, the sixth AI model, and the seventh AI model are AI models that are pre-deployed in the first node.

[0466] In some implementations, the second instruction information carries an eighth AI model. The processing unit 802, upon receiving the second instruction information, performs an AI deployment operation, including:

[0467] The processing unit 802 is configured to, upon receiving the second instruction information, activate the eighth AI model carried in the second instruction information in the first node, or switch the running AI model in the first node to the eighth AI model carried in the second instruction information.

[0468] In some implementations, the first node is a terminal; preset conditions include one or more of the following: the first node is moved to an area where AI usage is restricted, the first node is in power-saving mode, the operating system of the first node is changed, the utilization rate of the storage resources of the first node is changed, the utilization rate of the computing power of the first node is changed, and the accuracy of the AI ​​model running in the first node is lower than the accuracy threshold.

[0469] In some implementations, the first node is a network element of the core network; preset conditions include one or more of the following: the first node is deployed to an area where the use of AI is restricted, the load state of the first node changes, and the accuracy of the AI ​​model running in the first node is lower than an accuracy threshold.

[0470] In some implementations, the first node is an access network device; preset conditions include one or more of the following: changes in the utilization rate of the air interface resources of the first node, changes in the load status of the first node, changes in the utilization rate of the computing power of the first node, and the accuracy of the AI ​​model running in the first node being lower than the accuracy threshold.

[0471] Based on the same technical concept, this disclosure also provides a communication device. This communication device can implement the function of the second node in the aforementioned embodiments. Referring to FIG25, a schematic diagram of the structure of a communication device provided in an embodiment of this disclosure is shown. As shown, the communication device 90 may include: a processing unit 901 and a communication unit 902. Wherein:

[0472] The processing unit 901 is used to obtain the operating status of the first node; the first node is any one of a terminal, an access network device, or a core network element.

[0473] The communication unit 902 is used to send a first instruction message to the first node after obtaining that the running status of the first node meets the preset conditions; the first instruction message is used to instruct the first node to perform AI deployment operations.

[0474] AI deployment operations include operations that modify the AI ​​model running in the first node.

[0475] In some implementations, the first instruction information is specifically used to instruct the first node to perform one or more of the following: activate the first AI model, deactivate the second AI model, and switch the running AI model to the third AI model; wherein the first AI model, the second AI model, and the third AI model are AI models pre-deployed in the first node.

[0476] In some implementations, the first instruction information includes a fourth AI model; specifically, the first instruction information is used to instruct the first node to activate the fourth AI model in the first node, or to switch the running AI model in the first node to the fourth AI model.

[0477] In some implementations, processing unit 901 is used to obtain the running status of the first node, including:

[0478] The processing unit 901 is used to receive status information from the first node; the status information includes the operating status of the first node.

[0479] In some implementations, processing unit 901 is used to obtain the running status of the first node, including:

[0480] Processing unit 901 is used to detect the operating status of the first node.

[0481] In some implementations, when the first node is a terminal, preset conditions include one or more of the following: the first node moves to an area where AI usage is restricted; the first node is in power-saving mode; the operating system of the first node changes; the utilization rate of the first node's storage resources changes; the utilization rate of the first node's computing power changes; and the accuracy of the AI ​​model running in the first node is lower than an accuracy threshold. Alternatively, when the first node is a network element in the core network, preset conditions include one or more of the following: the first node is deployed to an area where AI usage is restricted; the load state of the first node changes; and the accuracy of the AI ​​model running in the first node is lower than an accuracy threshold. Alternatively, when the first node is an access network device, preset conditions include one or more of the following: the utilization rate of the first node's air interface resources changes; the load state of the first node changes; the utilization rate of the first node's computing power changes; and the accuracy of the AI ​​model running in the first node is lower than an accuracy threshold.

[0482] Based on the same technical concept, this disclosure also provides a communication device. This communication device can implement the function of the second node in the aforementioned embodiments. Referring to FIG26, a schematic diagram of the structure of a communication device provided in an embodiment of this disclosure is shown. As shown, the communication device 100 may include: a receiving unit 1001 and a transmitting unit 1002. Wherein:

[0483] The receiving unit 1001 is used to receive request information from the first node; the request information is used to instruct the first node to send an AI model; the first node is any one of a terminal, an access network device, or a network element of the core network.

[0484] The sending unit 1002 is used to send first instruction information to the first node according to the request information; the first instruction information includes a first AI model; the first instruction information is used to instruct the first node to activate a second AI model in the first node, or to switch the running AI model in the first node to the second AI model.

[0485] In some implementations, the request information includes the running status of the first node.

[0486] Sending unit 1002 is configured to send first indication information to the first node according to the request information, including:

[0487] The sending unit 1002 is used to send first indication information to the first node after determining that the operating status of the first node meets the preset conditions based on the request information.

[0488] In some implementations, the communication device 100 further includes:

[0489] The processing unit 1003 is used to detect the running status of the first node.

[0490] Sending unit 1002 is configured to send first indication information to the first node according to the request information, including:

[0491] The sending unit 1002 is used to send first instruction information to the first node according to the request information after determining that the operating status of the first node meets the preset conditions.

[0492] In some implementations, when the first node is a terminal, preset conditions include one or more of the following: the first node moves to an area where AI usage is restricted; the first node is in power-saving mode; the operating system of the first node changes; the utilization rate of the first node's storage resources changes; the utilization rate of the first node's computing power changes; and the accuracy of the AI ​​model running in the first node is lower than an accuracy threshold. Alternatively, when the first node is a network element in the core network, preset conditions include one or more of the following: the first node is deployed to an area where AI usage is restricted; the load state of the first node changes; and the accuracy of the AI ​​model running in the first node is lower than an accuracy threshold. Alternatively, when the first node is an access network device, preset conditions include one or more of the following: the utilization rate of the first node's air interface resources changes; the load state of the first node changes; the utilization rate of the first node's computing power changes; and the accuracy of the AI ​​model running in the first node is lower than an accuracy threshold.

[0493] It should be noted that the communication devices 70, 80, 90 and 100 provided in the embodiments of this disclosure can implement all or part of the method steps implemented in the above method embodiments and can achieve the same technical effect. Here, the parts and beneficial effects that are the same as those in the method embodiments will not be described in detail.

[0494] It should be noted that the division of units in the embodiments of this disclosure is illustrative and only represents one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional units in the various embodiments of this disclosure can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units described above can be implemented in hardware or as software functional units.

[0495] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a processor-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the related technology, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this disclosure.

[0496] This disclosure also provides a computer-readable storage medium storing computer instructions; when the computer-readable storage medium is run on a processor, the processor performs any of the methods provided in this disclosure.

[0497] This disclosure also provides a computer program product containing computer instructions that, when run on a processor, enable the processor to perform any of the methods provided in this disclosure.

[0498] This disclosure provides a chip including a processor that, when executing instructions, causes the chip to perform any of the methods provided in this disclosure. The instructions may originate from internal memory or external memory. In some embodiments, the chip further includes input / output circuitry serving as a communication interface.

[0499] It should be noted that, in this document, relational terms such as "first" and "second" are used merely 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. Furthermore, 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. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0500] The above description is merely a specific embodiment of this disclosure, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method of deploying an artificial intelligence (AI) model, wherein, The method is applied to a first node in a mobile communication system, and the method comprises: The first node determines that a running state of the first node meets a preset condition; The first node performs an AI deployment operation in a case where the running state of the first node meets the preset condition; The AI deployment operation is one of operations for changing an AI model running in the first node; and the first node is any one of a terminal, an access network device, or a network element of a core network.

2. The method of claim 1, wherein The first node determines that a running state of the first node meets a preset condition, comprising: The first node detects that the running state of the first node meets the preset condition; The first node performs an AI deployment operation in a case where the running state of the first node meets the preset condition, comprising: The first node performs the AI deployment operation in a case where it is detected that the running state of the first node meets the preset condition.

3. The method of claim 2, wherein, The AI deployment operation comprises one or more of the following: activating a first AI model, deactivating a second AI model, and switching a running AI model to a third AI model; The first AI model, the second AI model, and the third AI model are AI models pre-deployed in the first node.

4. The method of claim 2, wherein, The first node performs an AI deployment operation in a case where it is detected that the running state of the first node meets the preset condition, comprising: The first node sends request information to a second node in a case where it is detected that the running state of the first node meets the preset condition; wherein the request information is used to instruct the second node to send an AI model to the first node; The first node receives first indication information sent by the second node according to the request information; the fourth AI model is included in the first indication information; The first node performs an AI deployment operation according to the first indication information; the AI deployment operation comprises activating the fourth AI model or switching a running AI model to the fourth AI model.

5. The method of claim 4, wherein, The request information includes the running state of the first node.

6. The method of claim 1, wherein The first node determines that a running state of the first node meets a preset condition, comprising: The first node receives second indication information from a second node; the second indication information is used to instruct the second node to detect that the running state of the first node meets a preset condition; The first node performs an AI deployment operation in a case where the running state of the first meets the preset condition, comprising: The first node performs the AI deployment operation after receiving the second indication information.

7. The method of claim 6, wherein, The AI deployment operation comprises one or more of the following: activating a fifth AI model, deactivating a sixth AI model, and switching a running AI model to a seventh AI model; The fifth AI model, the sixth AI model, and the seventh AI model are AI models pre-deployed in the first node.

8. The method of claim 6, wherein, The second indication information carries an eighth AI model; after receiving the second indication information, the first node performs an AI deployment operation, including: After receiving the second indication information, the first node activates the eighth AI model carried in the second indication information in the first node, or switches a running AI model in the first node to the eighth AI model carried in the second indication information.

9. The method of any one of claims 1-8, wherein, The first node is a terminal. The preset condition includes one or more of the following: the first node moves to an area where AI is limited, the first node is in a power saving mode, an operating system of the first node changes, a usage rate of a storage resource of the first node changes, a usage rate of computing power of the first node changes, and an accuracy of an AI model running in the first node is lower than an accuracy threshold.

10. The method of any one of claims 1-8, wherein, The first node is a network element of a core network. The preset condition includes one or more of the following: the first node is deployed to an area where AI is limited, a load state of the first node changes, and an accuracy of an AI model running in the first node is lower than an accuracy threshold.

11. The method according to any one of claims 1-8, wherein, The first node is an access network device. The preset condition includes one or more of the following: a usage rate of an air interface resource of the first node changes, a load state of the first node changes, a usage rate of computing power of the first node changes, and an accuracy of an AI model running in the first node is lower than an accuracy threshold. 12.A method of deploying an artificial intelligence (AI) model, wherein, The method is applied to a mobile communication system including a first node and a second node, and the method includes: The second node obtains a running state of the first node; the first node is any one of a terminal, an access network device, or a network element of a core network. After the second node obtains that the running state of the first node meets a preset condition, the second node sends first indication information to the first node; the first indication information is used to instruct the first node to perform an AI deployment operation. The AI deployment operation includes an operation of changing an AI model running in the first node.

13. The method of claim 12, wherein, The first indication information is used to instruct the first node to perform one or more of the following: activate a first AI model, deactivate a second AI model, and switch a running AI model to a third AI model. The first AI model, the second AI model, and the third AI model are AI models pre-deployed in the first node.

14. The method of claim 12, wherein, The first indication information includes a fourth AI model; the first indication information instructs the first node to activate the fourth AI model in the first node or switch a running AI model in the first node to the fourth AI model.

15. The method according to any one of claims 12-14, wherein, The second node obtains a running state of the first node, including: The second node receives state information from the first node; the state information includes the running state of the first node.

16. The method of any one of claims 12-14, wherein, The second node obtains a running state of the first node, including: The second node detects the running state of the first node.

17. The method of any one of claims 12-14, wherein, In a case where the first node is a terminal, the preset condition includes one or more of the following: the first node moving to an area where the use of AI is limited, the first node being in a power saving mode, an operating system of the first node changing, a usage rate of a storage resource of the first node changing, a usage rate of computing power of the first node changing, and an accuracy of an AI model running in the first node being lower than an accuracy threshold; or, in a case where the first node is a network element of a core network, the preset condition includes one or more of the following: the first node being deployed to an area where the use of AI is limited, a load state of the first node changing, and an accuracy of an AI model running in the first node being lower than an accuracy threshold; or, in a case where the first node is an access network device, the preset condition includes one or more of the following: a usage rate of an air interface resource of the first node changing, a load state of the first node changing, a usage rate of computing power of the first node changing, and an accuracy of an AI model running in the first node being lower than an accuracy threshold. 18.A method of deploying an artificial intelligence (AI) model, wherein, The method is applied to a mobile communication system including a first node and a second node, and includes: The second node receives request information from the first node; the request information is used to indicate that an AI model is sent to the first node; the first node is any one of a terminal, an access network device, or a network element of a core network; The second node sends first indication information to the first node according to the request information; the first indication information includes a first AI model; the first indication information is used to instruct the first node to activate a second AI model in the first node or to switch an AI model running in the first node to the second AI model.

19. The method of claim 18, wherein, The request information includes a running state of the first node; The second node sends first indication information to the first node according to the request information, including: The second node sends first indication information to the first node after determining that the running state of the first node meets a preset condition.

20. The method of claim 18, wherein, The method further includes that the second node detects the running state of the first node; The second node sends first indication information to the first node according to the request information, including: The second node sends first indication information to the first node according to the request information after determining that the running state of the first node meets a preset condition.

21. A communications device, wherein, A first node applied to a mobile communication system, the communication device includes a memory, a transceiver, and a processor; The memory is used to store a computer program; the transceiver is used to transceive data under the control of the processor; and the processor is used to read the computer program in the memory and perform the following operations: determining that a running state of the first node meets a preset condition; in a case where the running state of the first node meets the preset condition, performing an AI deployment operation; The AI deployment operation is one of operations for changing an AI model running in the first node.

22. The communication apparatus according to claim 21, wherein The determining that the running state of the first node meets the preset condition comprises: detecting that the running state of the first node meets the preset condition; The executing the AI deployment operation in the case where the running state of the first node meets the preset condition comprises: The AI deployment operation comprises one or more of the following: activating a first AI model, deactivating a second AI model, and switching a running AI model to a third AI model.

23. The communication apparatus according to claim 22, wherein, The first AI model, the second AI model, and the third AI model are AI models pre-deployed in the first node. The executing the AI deployment operation in the case where the running state of the first node meets the preset condition comprises:

24. The communication apparatus according to claim 22, wherein, The AI deployment operation comprises one or more of the following: activating a first AI model, deactivating a second AI model, and switching a running AI model to a third AI model. The first AI model, the second AI model, and the third AI model are AI models pre-deployed in the first node. The executing the AI deployment operation in the case where the running state of the first node meets the preset condition comprises: The AI deployment operation comprises one or more of the following: activating a first AI model, deactivating a second AI model, and switching a running AI model to a third AI model.

25. The communication apparatus according to claim 24, wherein, The first AI model, the second AI model, and the third AI model are AI models pre-deployed in the first node. The executing the AI deployment operation in the case where the running state of the first node meets the preset condition comprises: The AI deployment operation comprises one or more of the following: activating a first AI model, deactivating a second AI model, and switching a running AI model to a third AI model. The first AI model, the second AI model, and the third AI model are AI models pre-deployed in the first node. The executing the AI deployment operation in the case where the running state of the first node meets the preset condition comprises: The AI deployment operation comprises one or more of the following: activating a first AI model, deactivating a second AI model, and switching a running AI model to a third AI model.

27. The communication apparatus according to claim 26, wherein, The first AI model, the second AI model, and the third AI model are AI models pre-deployed in the first node. The second indication information carries an eighth AI model, and the executing the AI deployment operation after receiving the second indication information comprises:

28. The communication apparatus according to claim 26, wherein, The first node is a terminal. ​ 29. The communication apparatus according to any of claims 21-28, wherein, ​ The preset condition comprises one or more of the following: the first node moving to an area where the use of AI is limited, the first node being in a power saving mode, an operating system of the first node changing, a usage rate of a storage resource of the first node changing, a usage rate of computing power of the first node changing, and an accuracy of an AI model running in the first node being lower than an accuracy threshold.

30. The communications apparatus of any of claims 21-28, wherein, The first node is a network element of a core network. The preset condition comprises one or more of the following: the first node being deployed to an area where the use of AI is limited, a load state of the first node changing, and an accuracy of an AI model running in the first node being lower than an accuracy threshold.

31. The communication apparatus according to any of claims 21-28, wherein, The first node is an access network device. The preset condition comprises one or more of the following: a usage rate of an air interface resource of the first node changing, a load state of the first node changing, a usage rate of computing power of the first node changing, and an accuracy of an AI model running in the first node being lower than an accuracy threshold.

32. A communications device, comprising: A second node applied to a mobile communication system, the mobile communication system comprising a first node and the second node, the communication device comprising: a memory, a transceiver, and a processor; The memory is configured to store a computer program; the transceiver is configured to transceive data under control of the processor; and the processor is configured to read the computer program in the memory and perform the following operations: obtaining a running state of the first node; the first node being any one of a terminal, an access network device, or a network element of a core network; after obtaining that the running state of the first node meets a preset condition, sending first indication information to the first node; the first indication information being used to instruct the first node to perform an AI deployment operation; The AI deployment operation comprises an operation of changing an AI model running in the first node.

33. A communications device, wherein, A second node applied to a mobile communication system, the mobile communication system comprising a first node and the second node, the communication device comprising: a memory, a transceiver, and a processor; The memory is configured to store a computer program; the transceiver is configured to transceive data under control of the processor; and the processor is configured to read the computer program in the memory and perform the following operations: receiving request information from the first node; the request information being used to instruct sending an AI model to the first node; the first node being any one of a terminal, an access network device, or a network element of a core network. According to the request information, sending first indication information to the first node; the first indication information comprising a first AI model; the first indication information being used to instruct the first node to activate a second AI model in the first node, or to switch an AI model running in the first node to the second AI model.

34. A communication system, wherein, comprising a first node and a second node; wherein the first node is configured to perform the method of any one of claims 1-11; and wherein the second node is configured to perform the method of any one of claims 12-17, or the second node is configured to perform the method of any one of claims 18-20.

35. A processor-readable storage medium, wherein, The processor readable storage medium stores a program for causing the processor to perform the method of any one of claims 1-11, or the program for causing the processor to perform the method of any one of claims 12-17, or the program for causing the processor to perform the method of any one of claims 18-20.

Images