Communication method and communication apparatus

Abstract

The present application discloses a communication method and a communication apparatus. The method is applied to the technical field of communications. The communication method comprises: receiving a service request from a terminal device, the service request comprising an identifier of a first application or an identifier of a first AI model. Determining at least one of model configuration information and task orchestration information according to communication state information between a terminal device and a first access network device, the first access network device being an access network device of a serving cell in which the terminal device is located, and the first AI model being at least deployed on the first access network device. Adopting the present application allows for better AI model configuration and task orchestration.

Description

Communication methods and communication devices

[0001] This application claims priority to Chinese Patent Application No. 202411455856.8, filed on October 17, 2024, entitled "Communication Method and Communication Device", the entire contents of which are incorporated herein by reference. Technical Field

[0002] This application relates to the field of communication technology, and in particular to a communication method and communication device. Background Technology

[0003] With the rapid development of artificial intelligence (AI) technology and machine learning algorithms, AI models, represented by neural networks and Transformers, have demonstrated powerful capabilities in numerous fields, giving rise to intelligent applications across various industries. In the communications field, on the one hand, AI models can replace traditional communication modules to achieve more intelligent network functions, i.e., AI for Network (AI4Net). For example, AI models can be used to achieve beam management, channel prediction, and resource allocation. On the other hand, the computing, transmission capabilities, and sensed information within the network can also support third-party AI applications on terminal devices. This can be achieved by offloading the computation of third-party AI applications to computing nodes within the network, or by using network-sensed environmental information and user behavior to provide better personalized services for third-party AI applications. Given these two potential scenarios and driving forces, the deep integration of AI models and communications has become one of the important visions for future communication systems.

[0004] As AI models and communication converge, future wireless networks will offer services beyond mere communication, such as computing and sensing, providing more real-time, seamless, and personalized AI services. In this scenario of AI model-communication convergence, achieving adaptive and flexible configuration of AI models has become a pressing technical challenge. Summary of the Invention

[0005] This application provides a communication method and a communication device that considers the communication status information of the air interface between the terminal device and the first access network device when determining the model configuration information and / or task orchestration information of the AI ​​model. It can adaptively and flexibly adjust the configuration and task orchestration of the AI ​​model based on the communication status information of the air interface, and can achieve better AI model configuration and task orchestration.

[0006] In a first aspect, embodiments of this application provide a communication method, wherein the method can be executed by a model management network element or by a component of the model management network element (e.g., a processor, a chip, or a chip system). The communication method may include:

[0007] Receive a service request from a terminal device, the service request including the identifier of the first application;

[0008] Based on the communication status information between the terminal device and the first access network device, determine at least one of the model configuration information and task orchestration information;

[0009] The model configuration information includes the identifier of the first AI model determined for the first application and the configuration information of the first AI model; the task orchestration information is the task orchestration information of the first AI model;

[0010] The first access network device is the access network device of the serving cell where the terminal device is located; the first AI model is deployed at least on the first access network device.

[0011] Secondly, embodiments of this application provide a communication method, wherein the method can be executed by a model management network element or by a component of the model management network element (e.g., a processor, chip, or chip system). The communication method may include:

[0012] The system receives a service request from a terminal device, the service request including an identifier of a first AI model. For example, the first AI model may be determined by the terminal device for a first application.

[0013] Based on the communication status information between the terminal device and the first access network device, at least one of the following is determined: model configuration information of the first AI model and task orchestration information of the first AI model.

[0014] The first access network device is the access network device of the serving cell where the terminal device is located; the first AI model is deployed at least on the first access network device.

[0015] By implementing the method described in the first or second aspect, when determining the model configuration information and / or task orchestration information, the communication status information between the terminal device and the first access network device is taken into account, that is, the communication status information of the air interface between the terminal device and the first access network device, which fluctuates greatly, is taken into account. Based on the communication status information of the air interface, the AI ​​model configuration and / or task orchestration are adaptively adjusted, which can achieve better AI model configuration and task orchestration.

[0016] Based on the first or second aspect, in one possible implementation, the communication status information includes current wireless communication information between the terminal device and the first access network device, historical wireless communication information between the terminal device and the first access network device, or predicted wireless communication information between the terminal device and the first access network device.

[0017] By implementing this approach, since the communication status information of the air interface changes significantly, AI model configuration and / or task orchestration can be achieved based on historical, current, or predicted air interface communication status information, thus enabling better AI model configuration and task orchestration.

[0018] Based on the first or second aspect, in one possible implementation, at least one of the following is determined according to the communication status information between the terminal device and the first access network device: model configuration information and task orchestration information, including:

[0019] Based on the communication status information and the first information, determine the model configuration information; and / or,

[0020] Based on the communication status information and the second information, the task scheduling information is determined;

[0021] The first information includes at least one of the following: computing status information of the first access network device, model deployment information of the first access network device, location movement information of the terminal device, and usage information of the terminal device on the first application;

[0022] The second information includes at least one of the following: the computing status information of the first access network device, and the model deployment information of the first access network device;

[0023] The computing status information of the first access network device includes at least one of the following: computing resource information of the first access network device, and computing resource occupancy information of the first access network device;

[0024] The model deployment information of the first access network device includes information about the AI ​​models that have been deployed on the first access network device.

[0025] Implementing this approach allows for further consideration of the computing status information of the first access network device during AI model configuration or task orchestration. This ensures that the first access network device has sufficient computing resources to complete the AI ​​model computation. Furthermore, the model deployment information of the first access network device can be considered during AI model configuration or task orchestration, thereby prioritizing the selection of AI models already deployed on the first access network device and improving AI model computation efficiency. Finally, the location and movement information of the terminal device, as well as its usage information of the first application, can be considered during AI model configuration to improve the accuracy of AI model configuration.

[0026] Based on the first or second aspect, in one possible implementation, the method further includes:

[0027] Receive at least one of the following from the first access network device: computing status information of the first access network device, model deployment information of the first access network device, location movement information of the terminal device, and usage information of the terminal device for the first application.

[0028] When this approach is implemented, if the model management network element is not the first access network device, it can obtain relevant information from the first access network device, thereby satisfying the AI ​​model configuration and task orchestration needs when the model management network element and the first access network device are different devices.

[0029] Based on the first or second aspect, in one possible implementation, the method further includes:

[0030] Send at least one of the model configuration information and the task orchestration information to the first access network device.

[0031] Implementing this approach can meet the needs of AI model configuration and task orchestration when the model management network element and the first access network device are different devices.

[0032] Based on the first or second aspect, in one possible implementation, the first AI model is also deployed on the terminal device;

[0033] The first information and / or the second information also include at least one of the following: computing status information of the terminal device, and model deployment information of the terminal device;

[0034] The computing status information of the terminal device includes at least one of the following: computing resource information of the terminal device, and computing resource occupancy information of the terminal device;

[0035] The model deployment information for terminal devices includes information about the AI ​​models that have already been deployed on the terminal devices.

[0036] In implementing this approach, the first AI model is also deployed on the terminal device. That is, the first AI model can be deployed on the first access network device and the terminal device. The first access network device and the terminal device can jointly train or infer the first AI model. In this scenario, when configuring the AI ​​model and orchestrating tasks, it is necessary to consider the computing status information and model deployment information of the terminal device to ensure that there are sufficient resources for AI model computing and to improve the computing efficiency of the AI ​​model.

[0037] Based on the first or second aspect, in one possible implementation, the method further includes:

[0038] Receive at least one of the following from the terminal device: computing status information and model deployment information from the terminal device.

[0039] Implementing this method can obtain relatively accurate computing status information and model deployment information of terminal devices.

[0040] Based on the first or second aspect, in one possible implementation, the method further includes:

[0041] Send at least one of the following to the terminal device: model configuration information and task orchestration information.

[0042] Implementing this method can meet the needs of AI model configuration and task orchestration in scenarios where the first access network device and terminal device jointly train or infer the first AI model.

[0043] Based on the first or second aspect, in one possible implementation, the first AI model is also deployed in the second access network device;

[0044] The first information and / or the second information also include at least one of the following: computing status information of the second access network device, and model deployment information of the second access network device;

[0045] The computing status information of the second access network device includes at least one of the following: computing resource information of the second access network device, and computing resource occupancy information of the second access network device;

[0046] The model deployment information for the second access network device includes information about the AI ​​models that have already been deployed on the second access network device.

[0047] By implementing this approach, the first AI model can be deployed on at least two access network devices, enabling more flexible AI model deployment and reducing the load on the first access network device.

[0048] Based on the first or second aspect, in one possible implementation, the method further includes:

[0049] Receive at least one of the following from the second access network device: the computing status information of the second access network device and the model deployment information of the second access network device.

[0050] By implementing this method, the computing status information and model deployment information of the second access network device can be obtained, which makes it easier for the first access network device to determine whether it is necessary to work with the second access network device to train and infer the AI ​​model.

[0051] Based on the first or second aspect, in one possible implementation, the method further includes:

[0052] Send at least one of the following to the second access network device: model configuration information and task orchestration information.

[0053] In one possible implementation, the first AI model is also deployed on a cloud server;

[0054] The first and / or second information also includes cloud server model deployment information;

[0055] The cloud server's model deployment information includes information about at least one sub-model of the first AI model that is deployed on the cloud server.

[0056] By implementing this method, the first AI model can also be deployed on a cloud server. In other words, the cloud server participates in the training and inference of the first AI model, which can improve the overall efficiency of the system.

[0057] Based on the first or second aspect, in one possible implementation, the method further includes:

[0058] Receive model deployment information from the cloud server.

[0059] By implementing this method, the cloud server can provide model deployment information, that is, information indicating at least one sub-model deployed on the cloud server, which can accurately determine the information of the sub-model deployed on the cloud server.

[0060] Based on the first or second aspect, in one possible implementation, the method further includes:

[0061] Receive the contract information from the first application on the cloud server;

[0062] Based on the communication status information between the terminal device and the first access network device, determine at least one of the following: model configuration information and task orchestration information, including:

[0063] Based on the communication status information and contract information, determine at least one of the following: model configuration information and task orchestration information.

[0064] This approach takes into account the first application's contract information when determining model configuration and task orchestration information, thereby providing better model configuration and task orchestration.

[0065] Based on the first or second aspect, in one possible implementation, the contract information includes at least one of the following:

[0066] Information on at least one alternative model supported by the first application, model compilation method, Quality of Service (QoS) parameters of the first application, runtime environment parameters of the first application, Service Level Agreement (SLA) parameters of the first application, and model requirements information of the cloud server.

[0067] The first AI model is determined from at least one candidate model;

[0068] The cloud server's model requirement information includes indication information, which indicates whether the AI ​​model associated with the first application is deployed on the cloud server.

[0069] Implementing this approach provides various information, including the contract information for the first application, thereby enabling better model configuration and task orchestration.

[0070] Thirdly, embodiments of this application provide a communication method, wherein the method can be executed by a cloud server or by a component of the cloud server (e.g., a processor, chip, or chip system). The communication method may include:

[0071] Receive a registration request from the terminal device, the registration request including the identifier of the first application;

[0072] In response to the registration request, the first application's contract information is sent. The contract information is used to determine at least one of the model configuration information and task orchestration information of the AI ​​model associated with the first application.

[0073] The third approach involves the cloud server providing contract information upon receiving a registration request. This contract information can be taken into account in the configuration of AI models and task orchestration, resulting in better model configuration and task orchestration.

[0074] Based on the third aspect, in one possible implementation, the contract information includes at least one of the following:

[0075] Information on at least one alternative model supported by the first application, model compilation method, QoS parameters of the first application, runtime environment parameters of the first application, SLA parameters of the first application, and model requirements information of the cloud server.

[0076] The model configuration information associated with the first application includes the identifier of a first AI model determined for the first application, wherein the first AI model is determined from at least one candidate model.

[0077] The cloud server's model requirement information includes indication information, which indicates whether the AI ​​model associated with the first application is deployed on the cloud server.

[0078] Implementing this approach provides various information, including contract details, to enable better model configuration and task orchestration.

[0079] Based on the third aspect, in one possible implementation, if the AI ​​model associated with the first application is deployed on a cloud server, the method also includes:

[0080] Receive service requests from terminal devices, the service requests including the identifier of the first application or the identifier of the first AI model;

[0081] In response to a service request, send the model deployment information of the cloud server;

[0082] The cloud server's model deployment information includes information about at least one sub-model deployed on the cloud server within the AI ​​model associated with the first application.

[0083] Implementing this approach provides a way to deploy AI models on cloud servers, improving the efficiency of AI model training and inference.

[0084] Fourthly, embodiments of this application provide a communication method, wherein the method can be executed by a first access network device or by a component of the first access network device (e.g., a processor, chip, or chip system). The communication method may include:

[0085] Receive service requests from terminal devices, the service requests including the identifier of the first application or the identifier of the first artificial intelligence (AI) model;

[0086] Send a service request to the model management network element;

[0087] Receive at least one of the following from the model management network element: model configuration information and task orchestration information.

[0088] In one possible implementation, at least one of the following is sent to the model management network element: the computing status information of the first access network device, the model deployment information of the first access network device, the location movement information of the terminal device, and the usage information of the terminal device on the first application.

[0089] The computing status information of the first access network device includes at least one of the following: computing resource information of the first access network device, and computing resource occupancy information of the first access network device;

[0090] The model deployment information of the first access network device includes information about the AI ​​models that have been deployed on the first access network device.

[0091] Fifthly, embodiments of this application provide a communication method, wherein the method is executed by a first access network device, or may be executed by a component of the first access network device (e.g., a processor, chip, or chip system). The communication method may include:

[0092] Receive a service request from a terminal device, the service request including the identifier of the first application;

[0093] Based on the communication status information between the terminal device and the first access network device, determine at least one of the model configuration information and task orchestration information;

[0094] The model configuration information includes the identifier of the first AI model determined for the first application and the configuration information of the first AI model; the task orchestration information is the task orchestration information of the first AI model;

[0095] The first access network device is the access network device of the serving cell where the terminal device is located; the first AI model is deployed at least on the first access network device.

[0096] Sixthly, embodiments of this application provide a communication method, wherein the method is executed by a first access network device, or may be executed by a component of the first access network device (e.g., a processor, chip, or chip system). The communication method may include:

[0097] The system receives a service request from a terminal device, the service request including an identifier of a first AI model. For example, the first AI model may be determined by the terminal device for a first application.

[0098] Based on the communication status information between the terminal device and the first access network device, at least one of the following is determined: model configuration information of the first AI model and task orchestration information of the first AI model.

[0099] The first access network device is the access network device of the serving cell where the terminal device is located; the first AI model is deployed at least on the first access network device.

[0100] Based on the fifth or sixth aspect, in one possible implementation, the communication status information includes current wireless communication information between the terminal device and the first access network device, historical wireless communication information between the terminal device and the first access network device, or predicted wireless communication information between the terminal device and the first access network device.

[0101] Based on the fifth or sixth aspect, in one possible implementation, at least one of the following is determined according to the communication status information between the terminal device and the first access network device: model configuration information and task orchestration information of the first AI model, including:

[0102] Based on the communication status information and the first information, determine the model configuration information; and / or,

[0103] Based on the communication status information and the second information, the task scheduling information is determined;

[0104] The first information includes at least one of the following: computing status information of the first access network device, model deployment information of the first access network device, location movement information of the terminal device, and usage information of the terminal device on the first application;

[0105] The second information includes at least one of the following: the computing status information of the first access network device, and the model deployment information of the first access network device;

[0106] The computing status information of the first access network device includes at least one of the following: computing resource information of the first access network device, and computing resource occupancy information of the first access network device;

[0107] The model deployment information of the first access network device includes information about the AI ​​models that have been deployed on the first access network device.

[0108] Based on the fifth or sixth aspect, in one possible implementation, the first AI model is also deployed on the terminal device;

[0109] The first information and / or the second information also include at least one of the following: computing status information of the terminal device, and model deployment information of the terminal device;

[0110] The computing status information of the terminal device includes at least one of the following: computing resource information of the terminal device, and computing resource occupancy information of the terminal device;

[0111] The model deployment information for terminal devices includes information about the AI ​​models that have already been deployed on the terminal devices.

[0112] Based on the fifth or sixth aspect, in one possible implementation, the method further includes:

[0113] Receive at least one of the following from the terminal device: computing status information and model deployment information from the terminal device.

[0114] Based on the fifth or sixth aspect, in one possible implementation, the method further includes:

[0115] Send at least one of the following to the terminal device: model configuration information and task orchestration information.

[0116] Based on the fifth or sixth aspect, in one possible implementation, the first AI model is also deployed in the second access network device;

[0117] The first information and / or the second information also include at least one of the following: computing status information of the second access network device, and model deployment information of the second access network device;

[0118] The computing status information of the second access network device includes at least one of the following: computing resource information of the second access network device, and computing resource occupancy information of the second access network device;

[0119] The model deployment information for the second access network device includes information about the AI ​​models that have already been deployed on the second access network device.

[0120] Based on the fifth or sixth aspect, in one possible implementation, the method further includes:

[0121] Receive at least one of the following from the second access network device: the computing status information of the second access network device and the model deployment information of the second access network device.

[0122] In one possible implementation, the method also includes:

[0123] Send at least one of the following to the second access network device: model configuration information of the first AI model and task orchestration information of the first AI model.

[0124] Based on the fifth or sixth aspect, in one possible implementation, the first AI model is also deployed on a cloud server;

[0125] The first and / or second information also includes cloud server model deployment information;

[0126] The cloud server's model deployment information includes information about at least one sub-model of the first AI model that is deployed on the cloud server.

[0127] Based on the fifth or sixth aspect, in one possible implementation, the method further includes:

[0128] Receive model deployment information from the cloud server.

[0129] Based on the fifth or sixth aspect, in one possible implementation, the method further includes:

[0130] Receive the contract information from the first application on the cloud server;

[0131] Based on the communication status information between the terminal device and the first access network device, determine at least one of the following: model configuration information and task orchestration information, including:

[0132] Based on the communication status information and contract information, determine at least one of the following: model configuration information and task orchestration information.

[0133] In one possible implementation, the contract information includes at least one of the following:

[0134] Information on at least one alternative model supported by the first application, model compilation method, QoS parameters of the first application, runtime environment parameters of the first application, SLA parameters of the first application, and model requirements information of the cloud server.

[0135] The first AI model is determined from at least one candidate model;

[0136] The cloud server's model requirement information includes indication information, which indicates whether the AI ​​model associated with the first application is deployed on the cloud server.

[0137] In a seventh aspect, embodiments of this application provide a communication method, wherein the method can be executed by a second access network device or by a component of the second access network device (e.g., a processor, chip, or chip system). The communication method may include:

[0138] Send at least one of the following to the first access network device: computing status information of the second access network device and model deployment information of the second access network device; the computing status information of the second access network device includes at least one of the following: computing resource information of the second access network device and computing resource occupancy information of the second access network device; the model deployment information of the second access network device includes information on the AI ​​models already deployed by the second access network device.

[0139] Based on the seventh aspect, in one possible implementation, the method further includes:

[0140] Receive at least one of the model configuration information and the task orchestration information from the first access network device.

[0141] Eighthly, embodiments of this application provide a communication apparatus, including a unit for performing the method in the first aspect or any possible implementation, or a unit for performing the method in the second aspect or any possible implementation, or a unit for performing the method in the third aspect or any possible implementation, or a unit for performing the method in the fourth aspect or any possible implementation, or a unit for performing the method in the fifth aspect or any possible implementation, or a unit for performing the method in the sixth aspect or any possible implementation, or a unit for performing the method in the seventh aspect or any possible implementation.

[0142] Ninthly, embodiments of this application provide a communication device including a processor for executing the methods in any possible implementations of the first to seventh aspects described above. Alternatively, the processor is configured to execute a program stored in a memory, wherein when the program is executed, the methods in any possible implementations of the first to seventh aspects described above are executed.

[0143] In one possible implementation, the memory is located outside the aforementioned communication device.

[0144] In one possible implementation, the memory is located within the aforementioned communication device.

[0145] In this embodiment of the application, the processor and memory can also be integrated into a single device, that is, the processor and memory can be integrated together.

[0146] In one possible implementation, the communication device further includes a transceiver for receiving or transmitting signals.

[0147] In a tenth aspect, embodiments of this application provide a communication device, which includes a logic circuit and an interface, wherein the logic circuit and the interface are coupled; the interface is used for inputting and / or outputting information, and the logic circuit is used for performing processing operations.

[0148] Eleventhly, embodiments of this application provide a computer-readable storage medium for storing a computer program that, when run on a computer, causes the methods shown in any possible implementation of the first to seventh aspects to be executed.

[0149] In a twelfth aspect, embodiments of this application provide a computer program product comprising a computer program or computer code that, when run on a computer, causes the methods shown in any possible implementation of the first to seventh aspects to be executed.

[0150] In a thirteenth aspect, embodiments of this application provide a computer program that, when run on a computer, executes the methods shown in any possible implementation of the first to seventh aspects described above.

[0151] In a fourteenth aspect, embodiments of this application provide a communication system, which includes a terminal device and a model management network element; or, the communication system includes a terminal device and a first access network device.

[0152] In one possible implementation, the communication system also includes a cloud server.

[0153] The technical effects achieved by the second to fourteenth aspects mentioned above can be referred to the technical effects of the first aspect or the beneficial effects in the method embodiments shown below, and will not be repeated here. Attached Figure Description

[0154] Figure 1 is a schematic diagram of a communication system provided in an embodiment of this application;

[0155] Figure 2 is a schematic diagram of the unified configuration and task orchestration of AI models by a cloud server provided in an embodiment of this application;

[0156] Figure 3 is a schematic diagram of the deployment location of the MC&TO functional module provided in the embodiment of this application;

[0157] Figure 4 is a flowchart illustrating a communication method provided in an embodiment of this application;

[0158] Figure 5 is an example of a communication method provided in an embodiment of this application;

[0159] Figure 6 is another example of the communication method provided in the embodiments of this application;

[0160] Figure 7 is an example of a communication method provided in an embodiment of this application;

[0161] Figure 8 is another example of the communication method provided in the embodiments of this application;

[0162] Figure 9 is a schematic diagram of the structure of a communication device provided in an embodiment of this application;

[0163] Figure 10 is a schematic diagram of another communication device provided in an embodiment of this application;

[0164] Figure 11 is a schematic diagram of the structure of another communication device provided in an embodiment of this application. Detailed Implementation

[0165] Figure 1 is a schematic diagram of a communication system 1000 provided in an embodiment of this application. As shown in Figure 1, the communication system 1000 includes a wireless access network 100 and a core network 200. Optionally, the communication system 1000 may also include an Internet 300. The Internet 300 may include a cloud server.

[0166] The wireless access network 100 may include at least one access network device (as shown in Figure 1, 110a and 110b) and at least one terminal device (as shown in Figure 1, 120a-120j). The terminal devices are connected wirelessly to the access network devices, and the access network devices are connected wirelessly or via a wired connection to the core network. The core network devices and access network devices may be independent physical devices, or the functions of the core network devices and the logical functions of the access network devices may be integrated on the same physical device, or a single physical device may integrate some of the functions of the core network devices and some of the functions of the access network devices. Terminal devices and access network devices may be interconnected via wired or wireless connections. Figure 1 is only a schematic diagram; the communication system 1000 may also include other network devices, such as wireless relay devices and wireless backhaul devices, which are not shown in Figure 1.

[0167] Access network equipment can be a base station, an evolved NodeB (eNodeB), a transmission reception point (TRP), a next-generation NodeB (gNB) in a 5G mobile communication system, access network equipment in an open radio access network (O-RAN), a next-generation base station in a 6G mobile communication system, a base station in a future mobile communication system, or an access node in a wireless fidelity (WiFi) system; or it can be a module or unit that performs some of the functions of a base station, such as a central unit (CU), a distributed unit (DU), a CU control plane (CU-CP) module, or a CU user plane (CU-UP) module. Access network equipment can be a macro base station (as shown in Figure 1, 110a), a micro base station or an indoor station (as shown in Figure 1, 110b), or a relay node, etc. The embodiments of this application do not limit the specific technology or device form used in the access network equipment.

[0168] In this application embodiment, the apparatus for implementing the functions of the access network device can be the access network device itself; it can also be an apparatus capable of supporting the access network device in implementing the functions, such as a chip system, hardware circuit, software module, or hardware circuit plus software module. This apparatus can be installed in the access network device or can be used in conjunction with the access network device. In this application embodiment, the chip system can be composed of chips or can include chips and other discrete devices. For ease of description, the following uses the example of an access network device as the apparatus for implementing the functions of the access network device to describe the technical solutions provided in this application embodiment.

[0169] Terminal devices can also be referred to as terminals, user equipment (UE), mobile stations, mobile terminals, etc. Terminal devices can be widely used in communication scenarios across various environments, including but not limited to one or more of the following: device-to-device (D2D), vehicle-to-everything (V2X), machine-type communication (MTC), Internet of Things (IoT), virtual reality, augmented reality, industrial control, autonomous driving, telemedicine, smart grids, smart furniture, smart offices, smart wearables, smart transportation, or smart cities, etc. Terminal devices can be mobile phones, tablets, computers with wireless transceiver capabilities, wearable devices, vehicles, drones, helicopters, airplanes, ships, robots, robotic arms, or smart home devices, etc. This application does not limit the specific technologies or device forms used in the terminal devices.

[0170] In the embodiments of this application, the apparatus for implementing the functions of the terminal device can be the terminal device itself; it can also be an apparatus capable of supporting the terminal device in implementing the functions, such as a chip system, hardware circuit, software module, or hardware circuit plus software module. This apparatus can be installed in the terminal device or can be used in conjunction with the terminal device. For ease of description, the following description uses the example of a terminal device as the apparatus for implementing the functions of the terminal device to describe the technical solutions provided in the embodiments of this application.

[0171] Access network equipment and terminal equipment can be fixed or mobile. Access network equipment and / or terminal equipment can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; they can also be deployed on water; and they can also be deployed in the air on aircraft, balloons, and satellites.

[0172] This application does not limit the application scenarios of access network devices and terminal devices. Access network devices and terminal devices can be deployed in the same or different scenarios. For example, access network devices and terminal devices can be deployed on land at the same time; or, access network devices can be deployed on land and terminal devices can be deployed on water, etc., and so on.

[0173] The roles of access network devices and terminal devices can be relative. For example, the helicopter or drone 120i in Figure 1 can be configured as a mobile access network device. For terminal devices 120j that access the wireless access network 100 via 120i, terminal device 120i is an access network device; however, for access network device 110a, 120i is a terminal device. That is, 110a and 120i communicate via a wireless air interface protocol. Alternatively, 110a and 120i can communicate via an interface protocol between access network devices. In this case, relative to 110a, 120i is also an access network device. Therefore, both access network devices and terminal devices can be collectively referred to as communication devices. 110a and 110b in Figure 1 can be called communication devices with access network device functions, and 120a-120j in Figure 1 can be called communication devices with terminal device functions.

[0174] In the embodiments of this application, the functions of the access network device can be executed by modules (such as chips) within the access network device, or by a control subsystem that includes the functions of the access network device. This control subsystem, including the functions of the access network device, can be a control center in the aforementioned application scenarios such as smart grids, industrial control, intelligent transportation, and smart cities. Similarly, the functions of the terminal device can be executed by modules (such as chips or modems) within the terminal device, or by a device that includes the functions of the terminal device.

[0175] The core network 200 includes one or more core network devices. These core network devices include, but are not limited to, one or more of the following network elements: authentication server function (AUSF) network element, unified data management (UDM) network element, unified data repository (UDR) network element, network repository function (NRF) network element, network exposure function (NEF) network element, application function (AF) network element, policy control function (PCF) network element, access and mobility management function (AMF) network element, session management function (SMF) network element, user plane function (UPF) network element, binding support function (BSF) network element, and network data analytics function (NWDAF) network element.

[0176] In some implementations, the access network device in this application embodiment may include one or more central units (CUs), one or more distributed units (DUs), and one or more radio units (RUs). The CU is used to connect to the core network and one or more DUs. For example, the CU may have some of the core network's functions. The CU may include a CU-control plane (CP) and a CU-user plane (UP).

[0177] In different system architectures, CU (or CU-CP and CU-UP), DU, or RU may have different names, but those skilled in the art will understand their meaning. For example, in an open RAN (ORAN) system architecture, CU can also be called an open CU (O-CU), DU can also be called an O-DU, CU-CP can also be called an O-CU-CP, CU-UP can also be called an O-CU-UP, and RU can also be called an O-RU. For ease of description, this application uses CU, CU-CP, CU-UP, DU, and RU as examples. Any of the units among CU (or CU-CP, CU-UP), DU, and RU in this application can be implemented through software modules, hardware modules, or a combination of software modules and hardware modules. It is understood that in an ORAN system, the access network equipment may also include a near real-time access network intelligent controller (RIC).

[0178] In the first implementation, an independent network element, referred to as a model management network element, can be introduced into the communication system shown in Figure 1. This model management network element determines the configuration information and / or task orchestration information of the AI ​​model and sends this information to other network elements or devices. The model management network element can be directly connected to the access network devices in the communication system, or indirectly connected through other network elements. These other network elements can be core network elements such as authentication management functions (AMF) or user plane functions (UPF). In some implementations, if the model management network element is deployed in the access network, it can be an independent RAN node (also called an anchor point). This model management network element can uniformly manage multiple RANs, i.e., it can perform model configuration and task orchestration for AI models on multiple RANs. In some implementations, the model management network element can also be deployed in the core network or the Internet; this application does not impose any limitations on this.

[0179] The second implementation involves embedding a model configuration and task orchestration (MC&TO) module within the first network element of the communication system shown in Figure 1. This MC&TO module determines the configuration information and / or task orchestration information of the AI ​​model and sends this information to other network elements or devices. This first network element can be an access network device, a core network device, or a network management system, such as Operation Administration and Maintenance (OAM). In some implementations, the MC&TO module can also be deployed on a cloud server, with the first network element being the cloud server. The MC&TO module can be a newly added program function on the first network element, or an independent module on the first network element, which can be implemented through software and / or hardware.

[0180] The following examples illustrate how to integrate the MC&TO function module into an access network device. For example, the MC&TO function can be integrated into the CU of the access network device, such as the CU-CP and / or CU-UP. For example, the MC&TO function can be integrated into the DU of the access network device. For example, the MC&TO function can be integrated into the RIC of the access network device.

[0181] Before detailing the method of this application, let me first briefly introduce some of the terms used in this application.

[0182] 1. AI Model

[0183] The AI ​​model in this application embodiment can be composed of various deep neural networks. Neural networks are a specific implementation of machine learning. Neural networks can be used to perform classification tasks, prediction tasks, and can also be used to establish conditional probability distributions between variables. Common neural networks include deep neural networks (DNNs) and generative neural networks (GNNs).

[0184] AI models can be associated with applications, meaning that AI services for associated applications can be achieved through the training or inference of AI models.

[0185] An AI model can be deployed on one or more devices.

[0186] If an AI model is deployed on a single device, the computing resources (computing power / model / data) of that single device can perform all the functions of the AI ​​model. This AI model can also be called a one-sided model.

[0187] If an AI model is deployed on multiple devices, it can be understood as dividing the AI ​​model into multiple parts, with each part deployed on a different device. The computing resources (computing power / model / data) of the multiple devices are used together for training or inference to complete all the functions of the AI ​​model.

[0188] For example, if an AI model is deployed on two devices, the AI ​​model is divided into two parts: one part is deployed on one device, and the other part is deployed on the other device. The computing resources of the two devices are jointly used for training or inference to complete the function of the AI ​​model. This AI model can also be called a two-sided model.

[0189] For example, if an AI model is deployed on three or more devices, the AI ​​model is divided into at least three parts: one part is deployed on one device, and different parts are deployed on different devices. The computing resources of these at least three devices are jointly used for training or inference to complete the function of the AI ​​model. This AI model can also be called a multilateral model.

[0190] In some embodiments, dividing an AI model into multiple parts can be understood as dividing an AI model into multiple sub-models, and dividing multiple sub-models into multiple parts, each part may include one or more sub-models.

[0191] If an AI model is deployed on both terminal and network devices, it can be called end-to-end network collaboration. The network device can be an access network device, a core network device, or an OAM (Operational Access Management) device. The computing resources of the terminal and network devices are jointly used for training or inference to complete the function of the AI ​​model.

[0192] If an AI model is deployed on terminal devices, network devices, and cloud servers, it can be called end-to-end cloud collaboration. The computing resources of terminal devices, network devices, and cloud servers are jointly used for training or inference to complete the function of the AI ​​model.

[0193] 2. Model configuration information

[0194] Model configuration information may include at least one of the following: model selection information, model quantization and pruning, model monitoring information, transmission method, compression method, and AI model parameters.

[0195] Model selection information indicates the AI ​​model selected or determined from the model pool corresponding to the application. This model pool may include multiple candidate models corresponding to the application. In this embodiment, determining a first AI model for a first application is used as an example. The first AI model can be an AI model selected or determined from the model pool corresponding to the first application. The model selection information may include the identifier of the first AI model.

[0196] Model monitoring information may include monitoring time intervals, which means that the model execution status is fed back once every monitoring time interval, such as whether the model prediction is accurate every monitoring time interval.

[0197] The transmission method can indicate the transmission method for intermediate data between multiple devices when multiple devices are jointly conducting training or inference.

[0198] Compression method can indicate the compression method used when multiple devices are jointly training or inference, for transmitting intermediate data between the devices.

[0199] The parameters of an AI model can include one or more of the following:

[0200] (1) Model architecture information, also known as neural network structure-related information, such as the type of neural network (e.g., deep neural network or generative neural network), the number of layers of the neural network, the number of neurons, etc.

[0201] (2) Model parameters, also known as the connection method of neurons or network layers;

[0202] (3) Neuron parameters can be understood as the parameters of each neuron in a neural network, such as weights, biases and activation functions.

[0203] 3. Task scheduling information

[0204] When an AI model is deployed across multiple devices, it needs to be divided into multiple parts, each part including one or more sub-models, and each part including one or more sub-models deployed on one device. The task orchestration information in this embodiment can indicate the division and deployment of the AI ​​model. Specifically, the task orchestration information may include at least one of the following: the device on which the AI ​​model is deployed, the sub-models deployed on each device, and dynamic reconfiguration information.

[0205] The dynamic reconstruction information indicates that the workload of at least two devices jointly performing training or inference can be dynamically adjusted, meaning the tasks can be dynamically and flexibly divided. For example, the amount of data processed between at least two devices jointly performing training or inference can be flexibly adjusted, or the sub-models deployed on at least two devices jointly performing training or inference can be flexibly adjusted. For instance, if the first AI model is deployed on a terminal device and an access network device, meaning the terminal device and access network device jointly perform training or inference, and the terminal device fails to complete the computation within a preset time when executing the task assigned to it, the number of sub-models deployed on the terminal device can be reduced, or the amount of data processed by the sub-models deployed on the terminal device can be reduced. Conversely, the number of sub-models deployed on the access network device can be increased, or the amount of data processed by the sub-models deployed on the access network device can be increased, thereby achieving dynamic task adjustment.

[0206] In the AI4Net scenario, there exists a network-inherent RAN AI model for enhancing communication functions, including one-sided and two-sided models. This AI model implements beam management, channel prediction, and resource allocation, and can replace traditional communication modules to achieve more intelligent network functions. The AI ​​model in the AI4Net scenario has a clearly defined task and is primarily controlled by the network side. Its basic model configuration process is as follows: the terminal device reports computing resources, the network side configures the model, and sends the configured AI model to the terminal device. Once configured, the AI ​​model remains unchanged. The fixed AI model configuration method in the AI4Net scenario cannot meet the diverse and personalized needs of third-party applications, lacking flexible model configuration.

[0207] To meet the needs of third-party applications, in the Net4AI scenario, AI model configuration and task orchestration can be uniformly performed through a cloud server. Figure 2 illustrates this unified AI model configuration and task orchestration using a cloud server. In Figure 2, terminal device 10 is connected to access network device 11 via a wireless communication link, access network device 11 is connected to core network device 12, and core network device 12 is connected to cloud server 13. Terminal device 10, access network device 11, and cloud server 13 have computing resources for AI model computation, and AI models can be deployed on terminal device 10, access network device 11, and cloud server 13. Terminal device 10 sends a service request to cloud server 13, and cloud server 13 can configure and orchestrate AI models based on the computing resources of access network device 11 and terminal device 10 in the mobile communication network, and send model configuration information and task orchestration information to access network device 11 and terminal device 10.

[0208] In this approach, the cloud server can only configure and orchestrate AI models based on the computing resources (such as computing power) of the devices deploying AI models. The cloud server cannot configure models and orchestrate tasks based on the fluctuating communication status information of the air interface. Furthermore, the cloud server cannot perceive the multi-user situation on the air interface side. For example, the cloud server cannot perceive other terminal devices that share channel resources or access network equipment with terminal device 10, making it difficult to achieve optimal model configuration and task orchestration.

[0209] To address the issues of not considering fluctuating air interface communication status information and the lack of flexible AI model configuration when determining model configuration and task orchestration, this embodiment of the application, upon receiving a service request from a terminal device, can determine at least one of model configuration information and task orchestration information based on the air interface communication status information between the terminal device and the first access network device. Furthermore, it can also determine at least one of model configuration information and task orchestration information based on relevant information about the computing resources of the device deploying the AI ​​model, thereby achieving better model configuration and task orchestration. Moreover, since this embodiment of the application can adaptively determine model configuration information and task orchestration information based on communication status information, flexible model configuration and task orchestration can also be achieved.

[0210] For example, an independent model management network element may be introduced in the access network, core network, or Internet, which can determine at least one of model configuration information and task orchestration information.

[0211] For example, an MC&TO function module can be added to the first network element. This MC&TO function module can determine at least one of the model configuration information and task orchestration information. As shown in Figure 3, which is a schematic diagram of the deployment location of the MC&TO function module provided in this embodiment, the MC&TO function module can be added to the access network device (i.e., the first network element is the access network device), or it can be added to the core network device or OAM (i.e., the first network element is the OAM), or it can be added to the cloud server (i.e., the first network element is the cloud server). The MC&TO function module can be a newly added program function in the access network device, core network device, OAM, or cloud server, or it can be a newly added independent module. For example, the MC&TO function module can be a chip on the first network element. The MC&TO function module determines at least one of the model configuration information and task orchestration information based on the communication status information of the air interface between the terminal device and the first access network device.

[0212] The AI ​​model in this embodiment is deployed at least on network elements within a mobile communication network. For example, the AI ​​model is deployed at least on access network equipment, or at least on core network equipment. The AI ​​model can also be deployed on terminal devices and / or cloud servers.

[0213] The following examples illustrate embodiments of the communication method of this application. It should be noted that the various technical solutions (or embodiments) of this application can be implemented independently or in combination based on certain inherent connections. This application does not impose limitations. Furthermore, various terms and definitions between the embodiments can be referenced mutually. In each embodiment of this application, different implementation methods can also be implemented in combination or independently.

[0214] Please refer to Figure 4, which is a flowchart illustrating a communication method provided in an embodiment of this application. Figure 1 can be a system architecture diagram to which this communication method is applicable. The execution order of the various steps of the communication method shown in Figure 4 is not limited in this embodiment. As shown in Figure 4, the communication method of this embodiment includes, but is not limited to, the following steps. It is understood that in some scenarios, some but not all of the following steps may be included, and this application does not limit this. In the following description, the communication device can be an introduced independent model management network element. The relevant description of the model management network element can be referred to the description of the foregoing embodiments. Alternatively, the communication device can be a first network element, which includes an MC&TO functional module. The first network element can be one of the following: an access network device, a core network device, an OAM, or a cloud server. The relevant introduction of the MC&TO functional module can be referred to the description of the foregoing embodiments.

[0215] 401, the terminal device sends a service request. Correspondingly, the communication device receives the service request from the terminal device.

[0216] The service request may include the identifier of the first application or the identifier of the first AI model. In this embodiment, the identifier of the first application may also be replaced with the identifier of a function, which may be model configuration and task orchestration to achieve a function.

[0217] If the service request includes an identifier for a first AI model, this first AI model can be determined by the terminal device. For example, the first AI model can be associated with a first application; the association of the first application with the first AI model can be understood as the first AI model being an AI model that implements the AI ​​service of the first application. The first AI model can be identified by the identifier of the first application. For instance, the first AI model can be an alternative model selected by the terminal device from at least one alternative model supported by the first application. The service request can be sent by the first application of the terminal device.

[0218] In some implementations, the terminal device can send service requests to the communication device through other network elements or devices. For example, if the communication device is a model management network element, the terminal device can send service requests to the model management network element through a first access network device, which can be the access network device of the serving cell where the terminal device is located.

[0219] The first AI model in this application embodiment can be deployed at least on a first access network device. Exemplarily, the first AI model can also be deployed on a terminal device. Exemplarily, the first AI model can also be deployed on a cloud server. Exemplarily, the first AI model can also be deployed on a core network device. When the first AI model is deployed on at least two devices, it can be understood that the at least two devices jointly train or infer the first AI model. At least one sub-model of the first AI model can be deployed on each device, and the sub-models deployed on different devices are different. Exemplarily, the first AI model can be deployed on a first access network device and a terminal device, or the first AI model can be deployed on a first access network device, a terminal device, and a cloud server, or the first AI model can be deployed on at least two access network devices and a terminal device, where the at least two access network devices include the first access network device and its adjacent access network devices, or the first AI model can be deployed on at least two access network devices, a terminal device, and a cloud server.

[0220] 402. The communication device determines at least one of the following: model configuration information and task orchestration information, based on the communication status information between the terminal device and the first access network device.

[0221] In one implementation, the service request includes an identifier of a first application, and the model configuration information may include an identifier of a first AI model determined or selected for the first application. The first AI model may be selected from at least one candidate model included in a preset model pool corresponding to the first application. The model configuration information may also include configuration information of the first AI model, which may include at least one of the following: model quantization pruning, model monitoring information, transmission method, compression method, etc. In this implementation, the task orchestration information is the task orchestration information of the selected or determined first AI model.

[0222] In another implementation, the service request includes the identifier of the first AI model, and the model configuration information is the configuration information of the first AI model. The configuration information of the first AI model may include at least one of the following: model quantization pruning, model monitoring information, transmission method, compression method, etc. In this implementation, the task orchestration information is the task orchestration information of the first AI model requested by the terminal device.

[0223] The communication status information between the terminal device and the first access network device can be understood as the communication status information of the air interface between the terminal device and the first access network device. The communication status information may include at least one of the following: current wireless communication information, historical wireless communication information, or predicted wireless communication information. The wireless communication information may include, but is not limited to, at least one of the following: time-frequency resource allocation information, the number of terminal devices accessing the first access network device, the maximum transmission rate of the terminal device, the guaranteed transmission rate of the terminal device, the communication latency, packet loss rate, and throughput of the terminal device.

[0224] If the communication device is a first access network device, the communication status information between the terminal device and the first access network device can be obtained by the communication device itself. If the communication device is another network element or device, such as a model management network element, the communication status information between the terminal device and the first access network device can be sent by the first access network device to the communication device.

[0225] For example, a communication device can determine model configuration information based on communication status information and the computational status information of at least one device deploying an AI model. The device's computational status information may include the device's computational resource information and / or computational resource occupancy information. The model configuration information may include the identifier of the selected AI model and related model configuration information, such as the quantization accuracy, pruning method, transmission quantization, or transmission sparsity of model parameters. This ensures that the transmission and computational requirements of the AI ​​model determined by the model configuration information can meet the user's required accuracy and latency, such as meeting the communication latency required by the communication status information or meeting the guaranteed transmission rate and throughput required by the communication status information. If multiple AI models meet these requirements, the AI ​​model with the lowest system overhead can be selected, or an AI model can be selected according to preset rules.

[0226] For example, a communication device can determine task orchestration information based on communication status information. For instance, if the number of terminal devices accessing the first access network device is greater than a threshold, fewer tasks can be assigned to the first access network device. This could mean assigning fewer sub-models to the first access network device or processing less data with the sub-models deployed on the first access network device, thereby preventing the first access network device from failing to complete calculations in a timely manner.

[0227] In some embodiments, the communication device may determine the model configuration information based on communication status information and first information, wherein the first information may include at least one of the following 1-4:

[0228] 1. Computational status information of the first access network device

[0229] The first AI model is deployed on the first access network device, and the first information may include the computing status information of the first access network device. The computing status information of the first access network device includes at least one of the following: computing resource information of the first access network device, and computing resource occupancy information of the first access network device. The computing resource information may include, but is not limited to, at least one of the following: computing power information, memory information, central processing unit (CPU) performance information, etc.

[0230] Information on computing resource usage can include the remaining free computing resources or the computing resource utilization rate, etc. Computing resources can be at least one of the following: computing power resources, memory resources, etc.

[0231] 2. Model deployment information of the first access network device

[0232] The first AI model is deployed on the first access network device, and the first information may include the model deployment information of the first access network device. The model deployment information of the first access network device includes information about the AI ​​models already deployed on the first access network device. For example, the model deployment information may include the identifier, function, etc., of the AI ​​models already deployed on the first access network device. The AI ​​models already deployed on the first access network device may be AI models deployed for other terminal devices or other applications. Considering the information of the AI ​​models already deployed on the first access network device during model configuration can fully utilize the AI ​​models already deployed on the first access network device and improve computational efficiency.

[0233] 3. Location and movement information of terminal devices

[0234] The location movement information of the terminal device may include at least one of the following: the current location of the terminal device, the historical movement trajectory of the terminal device, and the predicted movement trajectory of the terminal device.

[0235] In this embodiment, the model configuration information can be determined based on the location and movement information of the terminal device. For example, if the first application or function requested by the terminal device is related to location, such as positioning, navigation, scene recognition, or other related functions or applications, the appropriate AI model can be selected based on the location or movement speed of the terminal device to provide a better service experience.

[0236] For example, considering the fluctuations or degradation in transmission performance caused by user mobility, model configuration information with smaller transmission requirements can be selected in the model configuration, such as using transmission settings with lower quantization bits or higher transmission sparsity.

[0237] 4. Terminal device usage information for the first application

[0238] The usage information of the terminal device on the first application can be used to create a user profile for that terminal device. A user profile can be derived from the historical usage information of the terminal device's user on the first application. For example, if the first application is a social media application, the user profile can be derived from the exercise information posted by the terminal device's user on the social media application, identifying them as a sports enthusiast.

[0239] In this embodiment, model configuration information can be determined based on the usage information of the first application on the terminal device. For example, a finely tuned personalized model can be selected to provide services based on the user profile or preferences of the terminal device. The wireless network can classify users based on their communication behavior or determine their mobility preferences.

[0240] In some embodiments, the communication device may determine the task orchestration information of the first AI model based on communication status information and second information. The second information may include at least one of the following: the computing status information of the first access network device and the model deployment information of the first access network device. For a detailed description of the computing status information and model deployment information of the first access network device, please refer to points 1 and 2 of the foregoing embodiments, and will not be repeated here.

[0241] By considering the computational status information of the first access network device when determining the task orchestration information of the first AI model, it can be ensured that the first access network device has sufficient computing resources to perform the computation of the AI ​​model. Furthermore, by considering the AI ​​models already deployed on the first access network device when determining the task orchestration information of the first AI model, tasks of AI models already deployed on the first access network device can be orchestrated for the first access network device, thereby improving the computational efficiency of the first access network device.

[0242] After determining the model configuration information and / or task orchestration information, the communication device can send the determined model configuration information and / or task orchestration information to the device or network element deploying the first AI model. For example, the communication device can send the model configuration information and / or task orchestration information to the first access network device deploying the first AI model. The first access network device configures the sub-models deployed on the first access network device according to the model configuration information. The first access network device determines the sub-models deployed on the first access network device according to the task orchestration information. The sub-models deployed on the first access network device are the sub-models included in the first AI model.

[0243] The first AI model in this application embodiment can be deployed not only on the first access network device but also on a terminal device. The first access network device and the terminal device jointly train or infer the first AI model. For example, a portion of the sub-models in the first AI model are deployed on the first access network device, and another portion of the sub-models are deployed on the terminal device.

[0244] When determining model configuration information and / or task orchestration information, at least one of the following must also be considered: the terminal device's computing status information and the terminal device's model deployment information. That is, the aforementioned first information also includes at least one of the terminal device's computing status information and the terminal device's model deployment information. The aforementioned second information also includes at least one of the terminal device's computing status information and the terminal device's model deployment information. The terminal device's computing status information can be sent by the terminal device to the communication device, and the terminal device's model deployment information can also be sent by the terminal device to the communication device. For example, the terminal device's computing status information and / or model deployment information can be carried in a service request sent by the terminal device.

[0245] The computing status information of a terminal device includes at least one of the following: computing resource information of the terminal device, and computing resource occupancy information of the terminal device. Computing resource information may include, but is not limited to, at least one of the following: computing power information, memory information, central processing unit (CPU) performance information, etc.

[0246] Information on computing resource usage can include the remaining free computing resources or the computing resource utilization rate, etc. Computing resources can be at least one of the following: computing power resources, memory resources, etc.

[0247] The model deployment information for terminal devices includes information about the AI ​​models already deployed on the terminal device. For example, model deployment information may include the identifier, functionality, etc., of the AI ​​models already deployed on the terminal device. These AI models may be deployed for other applications. When configuring models or orchestrating tasks, considering the information about the AI ​​models already deployed on the terminal device allows for full utilization of these models, eliminating the need to send models to the terminal device and improving computational efficiency.

[0248] After determining the model configuration information and / or task orchestration information, the communication device can send the determined model configuration information and / or task orchestration information to the device or network element deploying the first AI model. For example, the communication device can send the model configuration information and / or task orchestration information to the terminal device deploying the first AI model. The terminal device configures the sub-model deployed on the terminal device according to the model configuration information. The terminal device determines the sub-model deployed on the terminal device according to the task orchestration information. In some embodiments, to reduce the information transmitted between the communication device and the terminal device, the communication device can send the model configuration information and / or task orchestration information related to the terminal device to the terminal device. For example, the communication device can send the model configuration information and task orchestration information of the sub-model deployed on the terminal device to the terminal device.

[0249] The first AI model in this application embodiment can be deployed not only in the first access network device but also in the second access network device. For example, the first AI model can also be deployed in a terminal device. The second access network device can be an adjacent access network device of the first access network device. In this application embodiment, the computing resources of adjacent access network devices can be used to jointly train or infer the first AI model, thus fully utilizing the computing resources of adjacent access network devices and improving computing efficiency.

[0250] When determining model configuration information and / or task orchestration information, at least one of the following must also be considered: the computational status information of the second access network device and the model deployment information of the second access network device. That is, the aforementioned first information also includes at least one of the computational status information and model deployment information of the second access network device. The aforementioned second information also includes at least one of the computational status information and model deployment information of the second access network device. The computational status information of the second access network device can be sent from the second access network device to the communication device, and the model deployment information of the second access network device can also be sent from the second access network device to the communication device. The communication device and the second access network device can periodically exchange computational status information and / or model deployment information, or the communication device can obtain at least one of the computational status information and model deployment information of the second access network device from the second access network device after receiving a service request from the terminal device.

[0251] The computing status information of the second access network device includes at least one of the following: computing resource information of the second access network device, and computing resource occupancy information of the second access network device. The computing resource information may include, but is not limited to, at least one of the following: computing power information, memory information, CPU performance information, etc.

[0252] Information on computing resource usage can include the remaining free computing resources or the computing resource utilization rate, etc. Computing resources can be at least one of the following: computing power resources, memory resources, etc.

[0253] The model deployment information of the second access network device includes information about the AI ​​models already deployed on the second access network device. For example, the model deployment information may include the identifier, function, etc., of the AI ​​models already deployed on the second access network device. These AI models may be deployed for other applications or other terminal devices. Considering the information about the AI ​​models already deployed on the second access network device during model configuration or task orchestration can fully utilize these models and improve computational efficiency.

[0254] After determining the model configuration information and / or task orchestration information, the communication device can send the determined model configuration information and / or task orchestration information to the device or network element deploying the first AI model. For example, the communication device can send the model configuration information and / or task orchestration information to the second access network device deploying the first AI model. The second access network device configures the sub-model deployed on the second access network device according to the model configuration information. The second access network device determines the sub-model deployed on the second access network device according to the task orchestration information. In some embodiments, to reduce the information transmitted between the communication device and the second access network device, the communication device can send the model configuration information and / or task orchestration information related to the second access network device to the second access network device. For example, the communication device can send the model configuration information and task orchestration information of the sub-model deployed on the second access network device to the second access network device.

[0255] The first AI model in this application embodiment can also be deployed on a cloud server. For example, at least one sub-model in the first AI model can be deployed on a cloud server. For example, the at least one sub-model deployed on the cloud server can be a relatively core part. The cloud server can send model deployment information to the communication device, and the model deployment information includes information about the at least one sub-model deployed on the cloud server. Accordingly, the communication device can consider the information of the at least one sub-model deployed on the cloud server when determining model configuration information and / or task orchestration information, and the aforementioned first information and / or second information also includes the model deployment information of the cloud server.

[0256] Before step 401, the terminal device may send a registration request to the cloud server. The registration request may include the identifier of the first application. In response to receiving the registration request, the cloud server may send subscription information to the terminal device. For example, the cloud server may also send subscription information to the communication device. The subscription information may be used by the communication device to determine at least one of the model configuration information and task orchestration information of the AI ​​model associated with the first application.

[0257] The contract information includes at least one of the following:

[0258] 1. Information on at least one alternative model supported by the first application.

[0259] The first AI model can be determined from the at least one candidate model. The at least one candidate model can be a model supported by a first application, determined by the cloud server from a preset model pool. The preset model pool corresponds to the first application.

[0260] 2. Model compilation method

[0261] The model compilation method can be determined by the cloud server according to the model intellectual property protection rules. The model compilation method can include public format or private compilation format.

[0262] 3. QoS parameters for the first application

[0263] 4. Operating environment parameters of the first application

[0264] The runtime environment parameters of the first application can be the AI ​​framework of the first application and the relevant settings parameters of the AI ​​service execution environment of the first application.

[0265] 5. SLA parameters for the first application

[0266] 6. Cloud server model requirements information

[0267] The model requirement information for the cloud server includes indication information indicating whether the AI ​​model associated with the first application is deployed on the cloud server. If the indication information indicates that the AI ​​model associated with the first application is deployed on the cloud server, the model requirement information for the cloud server may also include dynamic adjustment information. For example, the dynamic adjustment information may indicate whether at least one sub-model deployed on the cloud server can be dynamically adjusted; for example, it may indicate whether a sub-model deployed on the cloud server can be added, or a sub-model deployed on the cloud server can be reduced. For example, the dynamic adjustment information may indicate whether the amount of data processed by at least one sub-model deployed on the cloud server can be dynamically adjusted.

[0268] The following examples, using Figures 5 and 6, illustrate communication methods when the communication device is a newly introduced model management network element in the network, enabling centralized model configuration and task orchestration through the model management network element.

[0269] Please refer to Figure 5, which illustrates an example of a communication method provided in this application. Figure 5 shows an example where a first AI model is deployed on a terminal device and a first access network device, without the need for a cloud server; that is, the first AI model is not deployed on a cloud server. The terminal device and the first access network device jointly train or infer the first AI model. This scenario can also be referred to as a closed-loop AI service within the network. It is understood that the execution order of the steps in the embodiment shown in Figure 5 is not limited. In some application scenarios, some steps of the embodiment shown in Figure 5 may also be included. The steps in Figure 5 are described below:

[0270] 501, the terminal device sends a registration request to the cloud server.

[0271] 502, the cloud server sent the contract information.

[0272] The registration request includes the identifier of a first application, which may be sent by the first application installed on the terminal device. This first application is an AI application. The registration request is used to request a subscription registration. The subscription registration process can be initiated by the terminal device or by the cloud server. In response to the registration request, the cloud server sends subscription information, which can also be called registration information. Specifically, the cloud server can send subscription information to the terminal device. The cloud server can also send subscription information to the model management network element, allowing the model management network element to determine the model configuration information and / or task orchestration information of the AI ​​model associated with the first application based on this subscription information. The cloud server can also send subscription information to the first access network device, which is the access network device of the serving cell where the terminal device is located, and this first access network device serves the terminal device.

[0273] The contract information may include at least one of the following: information on at least one alternative model supported by the first application, the model compilation method, the QoS parameters of the first application, the SLA parameters of the first application, the AI ​​framework parameters, etc.

[0274] 503, the terminal device sends a service request to the first access network device.

[0275] A service request can also be called a session request. For example, a service request may include computing state information of the terminal device. For a description of the computing state information of the terminal device, please refer to the description in the foregoing embodiments, which will not be repeated here.

[0276] For example, the service request may include model deployment information of the terminal device. For a description of the model deployment information of the terminal device, please refer to the description of the foregoing embodiments, which will not be repeated here.

[0277] The first access network device can also send the service request to the cloud server.

[0278] 504, the first access network device sends a service request to the model management network element.

[0279] 505, the first access network device sends at least one of the following to the model management network element: communication status information, computing status information, model deployment information, location and movement information of the terminal device, and usage information of the terminal device for the first application.

[0280] Steps 504 and 505 can be sent via the same message or different messages.

[0281] The first access network device can send to the model management network element at least one of the following: communication status information between the first access network device and the terminal device over the air interface; computing status information of the first access network device; computing status information of the terminal device; model deployment information of the first access network device; model deployment information of the terminal device; location movement information of the terminal device; and terminal device usage information for the first application.

[0282] Depending on the deployment location of the model management network element, the first access network device can send information to the model management network element through different interfaces. For example, if the model management network element is deployed in the access network, such as as an anchor base station, the Xn interface is used to send information to the model management network element. As another example, if the model management network element is deployed in the core network, the Nx interface is used to send information to the model management network element. Yet another example, if the model management network element is deployed in the interconnection network, information is sent to the model management network element through an Application Programming Interface (API).

[0283] 506, The model management network element determines at least one of the following: model configuration information and task orchestration information.

[0284] The model management network element can determine at least one of the following: communication status information, computing status information of the first access network device, computing status information of the terminal device, model deployment information of the first access network device, model deployment information of the terminal device, location and movement information of the terminal device, terminal device usage information for the first application, and subscription information. The model configuration information may include the identifier of the first AI model determined by the model management network element. The first AI model may be determined from at least one alternative model indicated in the subscription information. The task orchestration information is the task orchestration information for the determined first AI model, such as which devices or network elements the first AI model is deployed on, and the sub-models that need to be deployed on each device or network element.

[0285] In some embodiments, the model management network element may also determine QoS based on at least one of the following: subscription information, communication status, computing status information of the first access network device, and computing status information of the terminal device.

[0286] 507. The model management network element sends at least one of the following to the first access network device and the terminal device: model configuration information and task orchestration information.

[0287] For example, the model management network element can send at least one of the determined model configuration information and task orchestration information to the first access network device to enable the first access network device to perform model configuration and task orchestration. In some implementations, the model management network element can send at least one of the model configuration information and task orchestration information related to the first access network device to the first access network device. For example, the task orchestration information related to the first access network device may include information related to the sub-models deployed on the first access network device. The model configuration information related to the first access network device may include the configuration information of the sub-models deployed on the first access network device.

[0288] For example, the model management network element can send at least one of the determined model configuration information and task orchestration information to the terminal device to enable the terminal device to perform model configuration and task orchestration. In some implementations, the model management network element can send at least one of the model configuration information and task orchestration information related to the terminal device to the terminal device. For example, the task orchestration information related to the terminal device may include information about the sub-models deployed on the terminal device. The model configuration information related to the terminal device may include the configuration information of the sub-models deployed on the terminal device.

[0289] 508. The model management network element sends at least one of the following to the cloud server: model configuration information or task orchestration information.

[0290] The model management network element can send at least one of the determined model configuration information and task orchestration information to the cloud server so that the cloud server can record the model configuration information and task orchestration information.

[0291] 509. The terminal device configures the model according to the model configuration information, and / or determines the sub-models deployed on the terminal device according to the task orchestration information.

[0292] The terminal device can determine at least one sub-model deployed on the terminal device through task orchestration information. It is understood that at least one sub-model deployed on the terminal device is included in the first AI model. Further details regarding task orchestration information can be found in the terminology explanation of the foregoing embodiments.

[0293] The terminal device can configure at least one sub-model deployed on the terminal device according to the model configuration information.

[0294] 510. The first access network device configures the model according to the model configuration information, and / or determines the sub-models deployed on the first access network device according to the task orchestration information.

[0295] The first access network device can determine at least one sub-model deployed on the first access network device through task orchestration information. It can be understood that at least one sub-model deployed on the first access network device is included in the first AI model. Further description of task orchestration information can be found in the terminology explanation of the foregoing embodiments.

[0296] The first access network device can configure at least one sub-model deployed on the first access network device according to the model configuration information.

[0297] 511. The cloud server records at least one of the following: model configuration information and task orchestration information.

[0298] The cloud server can record at least one of the received model configuration information and task orchestration information to determine the deployment status of the first AI model.

[0299] This application provides a centralized model management network element for model configuration and task orchestration. The network-closed loop enables the configuration and task orchestration of AI models. By introducing relevant information from terminal devices, wireless networks, and application servers, it ensures that the model configuration and task orchestration can meet the service requirements of the application, thereby improving the overall system efficiency while ensuring business experience.

[0300] Furthermore, centralized task orchestration allows for better management of multiple access network devices and the AI ​​services they serve, facilitating unified management, expansion, and updates.

[0301] Please refer to Figure 6 for another example of the communication method provided in this application embodiment. Figure 6 illustrates an example where a first AI model is deployed on a terminal device, a first access network device, and a cloud server. This application embodiment provides an AI service with end-to-end cloud collaboration, where the terminal device, the first access network device, and the cloud server jointly train or infer the first AI model. For reasons such as service performance and privacy protection, some sub-models of the first application need to be executed by the cloud server. It is understood that the execution order of each step in the embodiment shown in Figure 6 is not limited. In some application scenarios, some steps of the embodiment shown in Figure 6 may also be included. The steps in Figure 6 are described below:

[0302] 601, The terminal device sends a registration request to the cloud server.

[0303] 602, the cloud server sends the contract information.

[0304] For details regarding steps 601 and 602, please refer to steps 501 and 502 of the embodiment in Figure 5.

[0305] The difference from step 502 is that the contract information in step 602 may also include cloud server model requirement information.

[0306] In this embodiment, the model requirement information of the cloud server may include indication information indicating that the AI ​​model associated with the first application is also deployed on the cloud server. In some embodiments, the model requirement information of the cloud server may further include dynamic adjustment information. For example, the dynamic adjustment information may indicate whether at least one sub-model deployed on the cloud server can be dynamically adjusted, such as indicating whether a sub-model deployed on the cloud server can be added or removed. For example, the dynamic adjustment information may indicate whether the amount of data processed by at least one sub-model deployed on the cloud server can be dynamically adjusted.

[0307] 603, The terminal device sends a service request to the first access network device.

[0308] The first access network device can also send service requests to the cloud server.

[0309] 604, The first access network device sends a service request to the model management network element.

[0310] 605. The first access network device sends at least one of the following to the model management network element: communication status information, computing status information, model deployment information, location and movement information of the terminal device, and usage information of the terminal device for the first application.

[0311] Steps 604 and 605 can be sent via the same message or different messages.

[0312] Steps 603-605 in this embodiment can be referred to the relevant description of steps 503-505 in embodiment 5, and will not be repeated here.

[0313] 606, The cloud server sends the model deployment information of the cloud server to the model management network element.

[0314] When a cloud server receives a service request, it can send its model deployment information to the model management network element. This deployment information may include information about at least one sub-model deployed on the cloud server. For example, it may include at least one of the following: the level corresponding to the sub-model deployed on the cloud server, dynamic adjustment information, and data interaction methods. Dynamic adjustment information may indicate whether at least one sub-model deployed on the cloud server can be dynamically adjusted. Data interaction methods may indicate the compression and quantization methods of intermediate data used in model computation.

[0315] 607. The model management network element determines at least one of the following: model configuration information and task orchestration information.

[0316] The model management network element can determine at least one of the following: communication status information, computing status information of the first access network device, computing status information of the terminal device, model deployment information of the first access network device, model deployment information of the terminal device, location and movement information of the terminal device, terminal device usage information for the first application, and subscription information. Furthermore, the model management network element can also consider the model deployment information of the cloud server when determining the model configuration information and task orchestration information; that is, it determines at least one of the model configuration information and task orchestration information based on the model deployment information of the cloud server.

[0317] In some embodiments, the model management network element can also determine QoS parameters.

[0318] In some embodiments, steps 606 and 607 may also involve the cloud server determining the task orchestration information and instructing it to the model management network element.

[0319] 608. The model management network element sends at least one of the following to the first access network device and the terminal device: model configuration information and task orchestration information.

[0320] 609. The model management network element sends at least one of the following to the cloud server: model configuration information and task orchestration information.

[0321] 610. The terminal device configures the model according to the model configuration information, and / or determines the sub-models deployed on the terminal device according to the task orchestration information.

[0322] 611. The first access network device configures the model according to the model configuration information, and / or determines the sub-models deployed on the first access network device according to the task orchestration information.

[0323] Steps 608-611 in this embodiment are referred to steps 507-510 in embodiment 5, and will not be repeated here.

[0324] 612. The cloud server configures the sub-models deployed on the cloud server and records at least one of the following: model configuration information and task orchestration information.

[0325] The cloud server can record at least one of the received model configuration information and task orchestration information to determine the deployment status of the first AI model. Since some sub-models in the first AI model are deployed on the cloud server, the cloud server will configure the sub-models deployed on the cloud server.

[0326] The difference between this embodiment and the embodiment in Figure 5 is that the first AI model associated with the first application is also deployed on a cloud server. The cloud server participates in the training or inference of the first AI model. When configuring the model and arranging tasks for the first AI model, it can comprehensively consider the computing capabilities of the cloud server, access network equipment, and terminal equipment, thereby providing a more reasonable task arrangement and model configuration. Since the cloud server also participates in the training or inference of the first AI model, it can also improve the overall efficiency of the system.

[0327] The following examples, using Figures 7 and 8, illustrate a communication method when the communication device is a first access network device serving a terminal device. The first access network device may include an MC&TO function module, which can determine at least one of the model configuration information and task orchestration information of the first AI model. This MC&TO function module can be a chip on the first access network device. The first AI model can be deployed on multiple access network devices, including the first access network device, thereby fully utilizing the distributed computing resources on the wireless network side. Optionally, each access network device may include an MC&TO function module. Figures 7 and 8 illustrate this with an example of the first AI model deployed on two access network devices (the first access network device and the second access network device, respectively). The second access network device can be an adjacent access network device to the first access network device, and the first access network device is the access network device of the serving cell where the terminal device is located.

[0328] Please refer to Figure 7, which illustrates an example of a communication method provided in this application. Figure 7 shows an example where a first AI model is deployed on a terminal device, a first access network device, and a second access network device, without the need for a cloud server; that is, the first AI model is not deployed on a cloud server. The terminal device, the first access network device, and the second access network device jointly train or infer the first AI model. This scenario can also be referred to as a closed-loop AI service within the network. It is understood that the execution order of the steps in the embodiment shown in Figure 7 is not limited. In some application scenarios, some steps of the embodiment shown in Figure 7 may also be included. The steps in Figure 7 are described below:

[0329] 701, The terminal device sends a registration request to the cloud server.

[0330] 702, the cloud server sends the contract information.

[0331] For details regarding steps 701 and 702, please refer to steps 501 and 502 of the embodiment in Figure 5.

[0332] 703, The first access network device and the second access network device interact to calculate at least one of the following: status information and model deployment information.

[0333] The first access network device and the second access network device can periodically exchange at least one of the following: computation status information and model deployment information. Specifically, the second access network device sends its own computation status information and model deployment information to the first access network device. This allows the first access network device to determine whether the AI ​​model needs to be deployed on the second access network device. Conversely, the first access network device sends its own computation status information and model deployment information to the second access network device, enabling the second access network device to determine whether the AI ​​model needs to be deployed on the first access network device.

[0334] In some embodiments, the first access network device and the second access network device can also exchange communication status information. Specifically, the first access network device sends its air interface communication status information to the second access network device, and the second access network device sends its air interface communication status information to the first access network device.

[0335] In some embodiments, step 703 may be executed after the first access network device receives the service request.

[0336] 704, The terminal device sends a service request to the first access network device.

[0337] The first access network device can also send service requests to the cloud server.

[0338] Please refer to step 503 of embodiment 5 in this application for step 704, which will not be repeated here.

[0339] 705, The first access network device determines at least one of the following: model configuration information and task orchestration information.

[0340] The first access network device can determine at least one of the following: communication status information between the first access network device and the terminal device, computing status information of the first access network device, computing status information of the terminal device, model deployment information of the first access network device, model deployment information of the terminal device, location and movement information of the terminal device, usage information of the terminal device for the first application, and subscription information.

[0341] In some embodiments, when determining model configuration information and / or task orchestration information, the first access network device may also consider at least one of the computing status information and model deployment information of the second access network device. For example, if the first access network device does not have sufficient computing resources to complete the computation, it can utilize the computing resources of the second access network device. Or, for example, if the model already deployed by the second access network device has better performance, the first access network device can utilize the model resources of the second access network device.

[0342] Step 705 can be broken down into two steps: the first access network device determines the model configuration information of the first AI model, and the first access network device determines the task orchestration information of the first AI model.

[0343] 706. The first access network device sends at least one of the following to the terminal device: model configuration information and task orchestration information.

[0344] For example, the first access network device may send at least one of the determined model configuration information and task orchestration information to the terminal device to enable the terminal device to perform model configuration and task orchestration. In some implementations, the first access network device may send at least one of the model configuration information and task orchestration information related to the terminal device to the terminal device. For example, the task orchestration information related to the terminal device may include information related to the sub-models deployed on the terminal device. The model configuration information related to the terminal device may include the configuration information of the sub-models deployed on the terminal device.

[0345] 707. The first access network device sends at least one of the following to the second access network device: model configuration information and task orchestration information.

[0346] For example, the first access network device may send at least one of the determined model configuration information and task orchestration information to the second access network device to enable the second access network device to perform model configuration and task orchestration. In some implementations, the first access network device may send at least one of the model configuration information and task orchestration information related to the second access network device to the second access network device. For example, the task orchestration information related to the second access network device may include information related to the sub-models deployed on the second access network device. The model configuration information related to the second access network device may include the configuration information of the sub-models deployed on the second access network device.

[0347] 708. The first access network device sends at least one of the following to the cloud server: model configuration information and task orchestration information.

[0348] The first access network device can send at least one of the determined model configuration information and task orchestration information to the cloud server so that the cloud server can record the model configuration information and task orchestration information.

[0349] 709. The terminal device configures the model according to the model configuration information, and / or determines the sub-models deployed on the terminal device according to the task orchestration information.

[0350] 710. The first access network device configures the model according to the model configuration information, and / or determines the sub-models deployed on the first access network device according to the task orchestration information.

[0351] Steps 709-710 in this embodiment can be referred to steps 509-510 in embodiment 5, and will not be repeated here.

[0352] 711. The second access network device configures the model according to the model configuration information, and / or determines the sub-models deployed on the second access network device according to the task orchestration information.

[0353] The second access network device can determine at least one sub-model deployed on the second access network device through task orchestration information. It can be understood that at least one sub-model deployed on the second access network device is included in the first AI model. For further details on task orchestration information, please refer to the terminology explanation in the foregoing embodiments.

[0354] The second access network device can configure at least one sub-model deployed on the second access network device according to the model configuration information.

[0355] 712, The cloud server records at least one of the following: model configuration information and task orchestration information.

[0356] Step 712 of this application embodiment can be referred to step 511 of the embodiment in Figure 5, and will not be repeated here.

[0357] This application embodiment deploys the MC&TO functional module independently for each access network device, providing greater flexibility in network topology. It eliminates the need for additional model management network elements, thus saving deployment costs.

[0358] Please refer to Figure 8 for another example of the communication method provided in this application embodiment. Figure 8 shows an example of a first AI model deployed on a terminal device, a first access network device, a second access network device, and a cloud server. This application embodiment provides an AI service with end-to-end cloud collaboration, where the terminal device, the first access network device, the second access network device, and the cloud server jointly train or infer the first AI model. For reasons such as service performance and privacy protection, some sub-models of the first application need to be executed by the cloud server. It is understood that the execution order of each step in the embodiment shown in Figure 8 is not limited. In some application scenarios, some steps of the embodiment shown in Figure 8 may also be included. The steps in Figure 8 are described below:

[0359] 801, The terminal device sends a registration request to the cloud server.

[0360] 802, the cloud server sends the contract information.

[0361] For details regarding steps 801 and 802, please refer to steps 501 and 502 of the embodiment in Figure 5.

[0362] The difference between step 502 and step 802 is that the contract information in step 802 may also include cloud server model requirement information. For details regarding cloud server model requirement information, please refer to the description of step 602 in embodiment 6, which will not be repeated here.

[0363] 803, the first access network device and the second access network device interact to calculate at least one of the following: status information and model deployment information.

[0364] Step 803 in this embodiment can be referred to the relevant description of step 703 in embodiment 7, and will not be repeated here.

[0365] 804, the terminal device sends a service request to the first access network device.

[0366] Please refer to step 503 of embodiment 5 in this application for step 704, which will not be repeated here.

[0367] 805, the first access network device sends a service request to the cloud server.

[0368] 806, the cloud server sends the cloud server model deployment information to the first access network device.

[0369] When a cloud server receives a service request, it sends its model deployment information to the first access network device. This model deployment information may include information about at least one sub-model deployed on the cloud server. For example, it may include at least one of the following: the level corresponding to the sub-model deployed on the cloud server, dynamic adjustment information, and data interaction methods. Dynamic adjustment information may indicate whether at least one sub-model deployed on the cloud server can be dynamically adjusted. Data interaction methods may indicate the compression method, quantization method, etc., of the intermediate data used in model computation.

[0370] 807, the first access network device determines at least one of the model configuration information and task orchestration information.

[0371] When determining model configuration information and task orchestration information, the first access network device may consider at least one of the following: communication status information between the first access network device and the terminal device, computing status information of the first access network device, computing status information of the terminal device, model deployment information of the first access network device, model deployment information of the terminal device, location and movement information of the terminal device, usage information of the terminal device for the first application, subscription information, computing status information of the second access network device, and model deployment information of the second access network device. In addition, it may also consider the model deployment information of the cloud server, thereby achieving better model configuration and task orchestration.

[0372] 808, The first access network device sends at least one of the following to the terminal device: model configuration information and task orchestration information.

[0373] 809. The first access network device sends at least one of the following to the second access network device: model configuration information and task orchestration information.

[0374] 810. The first access network device sends at least one of the following to the cloud server: model configuration information and task orchestration information.

[0375] Please refer to steps 706-708 of embodiment 7 for steps 808-810 of this application embodiment.

[0376] 811. The terminal device configures the model according to the model configuration information, and / or determines the sub-models deployed on the terminal device according to the task orchestration information.

[0377] 812, the first access network device configures the model according to the model configuration information, and / or determines the sub-models deployed on the first access network device according to the task orchestration information.

[0378] Steps 709-710 in this embodiment can be referred to steps 509-510 in embodiment 5, and will not be repeated here.

[0379] 813. The second access network device configures the model according to the model configuration information, and / or determines the sub-models deployed on the second access network device according to the task orchestration information.

[0380] Steps 811-813 in this embodiment are referred to steps 709-711 in embodiment 7, and will not be repeated here.

[0381] 814. The cloud server configures the sub-models deployed on the cloud server and records at least one of the following: model configuration information and task orchestration information.

[0382] The cloud server can record at least one of the received model configuration information and task orchestration information to determine the deployment status of the first AI model. Since some sub-models in the first AI model are deployed on the cloud server, the cloud server will configure the sub-models deployed on the cloud server.

[0383] The difference between this embodiment and the embodiment in Figure 7 is that the first AI model associated with the first application is also deployed on a cloud server. The cloud server participates in the training or inference of the first AI model. When configuring the first AI model and arranging tasks, it can comprehensively consider the computing capabilities of the cloud server, access network equipment, and terminal equipment, thereby providing a more reasonable task arrangement and model configuration. Since the cloud server also participates in the training or inference of the first AI model, it can also improve the overall efficiency of the system.

[0384] The following describes the communication device provided in the embodiments of this application.

[0385] This application divides the communication device into functional modules according to the above-described method embodiments. For example, each function can be divided into its own functional modules, or two or more functions can be integrated into one processing module. The integrated modules can be implemented in hardware or as software functional modules. It should be noted that the module division in this application is illustrative and represents only one logical functional division; in actual implementation, other division methods may be used. The communication device of this application embodiment will be described in detail below with reference to Figures 9 to 11.

[0386] Figure 9 is a schematic diagram of a communication device provided in an embodiment of this application. As shown in Figure 9, the communication device 1000 can correspondingly implement the functions or steps implemented by the communication device in the above-described method embodiments, or the communication device 1000 can correspondingly implement the functions or steps implemented by the cloud server in the above-described method embodiments, or the communication device 1000 can correspondingly implement the functions or steps implemented by the second access network device, or the communication device can correspondingly implement the functions or steps implemented by the first access network device. The communication device 1000 includes: a transceiver unit 1100 and a processing unit 1200. The transceiver unit 1100 is used to perform receiving or sending operations, and the processing unit 1200 is used to perform processing operations.

[0387] In some possible implementations, the communication device 1000 can correspondingly implement the behavior and functions of the communication device in the above method embodiments. For example, the communication device 1000 can be a communication device or a component (e.g., a chip or circuit) applied in the communication device. The transceiver unit 1100 can, for example, be used to perform all the receiving or transmitting operations performed by the communication device in the above method embodiments. The processing unit 1200 is used to perform all operations performed by the communication device other than the receiving and transmitting operations.

[0388] The transceiver unit 1100 is used to receive service requests from terminal devices, the service requests including the identifier of a first application or the identifier of a first artificial intelligence (AI) model;

[0389] The processing unit 1200 is configured to determine at least one of model configuration information and task orchestration information based on the communication status information between the terminal device and the first access network device; the first access network device is the access network device of the serving cell where the terminal device is located;

[0390] When the service request includes the identifier of the first application, the model configuration information includes the identifier of the first AI model determined for the first application and the configuration information of the first AI model; when the service request includes the identifier of the first AI model, the model configuration information is the configuration information of the first AI model.

[0391] The task orchestration information is the task orchestration information of the first AI model;

[0392] The first access network device is the access network device of the serving cell where the terminal device is located; the first AI model is deployed at least on the first access network device.

[0393] In one possible implementation, the communication status information includes at least one of the following: current wireless communication information between the terminal device and the first access network device, historical wireless communication information between the terminal device and the first access network device, or predicted wireless communication information between the terminal device and the first access network device.

[0394] In one possible implementation, the processing unit 1200 is specifically used for:

[0395] Based on the communication status information and the first information, determine the model configuration information; and / or,

[0396] The task orchestration information is determined based on the communication status information and the second information;

[0397] The first information includes at least one of the following: computing status information of the first access network device, model deployment information of the first access network device, location movement information of the terminal device, and usage information of the terminal device for the first application;

[0398] The second information includes at least one of the following: the computing status information of the first access network device, and the model deployment information of the first access network device;

[0399] The computing status information of the first access network device includes at least one of the following: computing resource information of the first access network device, and computing resource occupancy information of the first access network device;

[0400] The model deployment information of the first access network device includes information about the AI ​​models that have been deployed on the first access network device.

[0401] In one possible implementation, the transceiver unit 1100 is further configured to receive at least one of the following from the first access network device: the computing status information of the first access network device, the model deployment information of the first access network device, the location movement information of the terminal device, and the usage information of the terminal device on the first application.

[0402] In one possible implementation, the transceiver unit 1100 is further configured to send at least one of the model configuration information and the task orchestration information to the first access network device.

[0403] In one possible implementation, the first AI model is also deployed on the terminal device;

[0404] The first information and / or the second information further include at least one of the following: the computing status information of the terminal device, and the model deployment information of the terminal device;

[0405] The computing status information of the terminal device includes at least one of the following: computing resource information of the terminal device, and computing resource occupancy information of the terminal device;

[0406] The model deployment information of the terminal device includes information about the AI ​​models that have been deployed on the terminal device.

[0407] In one possible implementation, the transceiver unit 1100 is further configured to receive at least one of the terminal device's computing status information and the terminal device's model deployment information from the terminal device.

[0408] In one possible implementation, the transceiver unit 1100 is further configured to send at least one of the model configuration information of the first AI model and the task orchestration information of the first AI model to the terminal device.

[0409] In one possible implementation, the first AI model is also deployed in a second access network device;

[0410] The first information and / or the second information further include at least one of the following: the computing status information of the second access network device, and the model deployment information of the second access network device;

[0411] The computing status information of the second access network device includes at least one of the following: computing resource information of the second access network device, and computing resource occupancy information of the second access network device;

[0412] The model deployment information of the second access network device includes information about the AI ​​models that have been deployed on the second access network device.

[0413] In one possible implementation, the transceiver unit 1100 is further configured to receive at least one of the following: the computing status information of the second access network device and the model deployment information of the second access network device.

[0414] In one possible implementation, the transceiver unit 1100 is further configured to send at least one of model configuration information and task orchestration information to the second access network device.

[0415] In one possible implementation, the first AI model is also deployed on a cloud server;

[0416] The first information and / or the second information also include the model deployment information of the cloud server;

[0417] The model deployment information of the cloud server includes information on at least one sub-model deployed on the cloud server in the first AI model.

[0418] In one possible implementation, the transceiver unit 1100 is further configured to receive model deployment information of the cloud server from the cloud server.

[0419] In one possible implementation, the transceiver unit 1100 is further configured to receive subscription information of the first application from the cloud server;

[0420] The processing unit 1200 is specifically used to: determine at least one of the model configuration information and task orchestration information based on the communication status information and the subscription information.

[0421] In one possible implementation, the contract information includes at least one of the following:

[0422] Information on at least one alternative model supported by the first application, model compilation method, QoS parameters of the first application, runtime environment parameters of the first application, SLA parameters of the first application, and model requirement information of the cloud server;

[0423] The first AI model is determined from the at least one candidate model;

[0424] The model requirement information of the cloud server includes indication information, which indicates whether the AI ​​model associated with the first application is deployed on the cloud server.

[0425] The detailed description and beneficial effects of the device embodiment shown in Figure 9 can be found in the description of the foregoing method embodiment, and will not be repeated here.

[0426] Reusing Figure 9, the communication device 1000 can correspondingly implement the behavior and functions of the cloud server in the above method embodiments. For example, the communication device 1000 can be a cloud server, or it can be a component (e.g., a chip or circuit) applied in the cloud server. The transceiver unit 1100 can, for example, be used to perform all the receiving or sending operations performed by the cloud server in the above method embodiments. The processing unit 1200 is used to perform all operations performed by the cloud server except for the receiving and sending operations.

[0427] In one possible design, transceiver unit 1100 is used to receive a registration request from a terminal device, the registration request including the identifier of a first application;

[0428] The processing unit 1200 is used to respond to the registration request and send the contract information of the first application through the transceiver unit 1100. The contract information is used to determine at least one of the model configuration information and task orchestration information associated with the first application.

[0429] In one possible design, the contract information includes at least one of the following:

[0430] Information on at least one alternative model supported by the first application, model compilation method, QoS parameters of the first application, runtime environment parameters of the first application, SLA parameters of the first application, and model requirement information of the cloud server;

[0431] The model configuration information associated with the first application includes the identifier of a first AI model determined for the first application, wherein the first AI model is determined from the at least one candidate model.

[0432] The model requirement information of the cloud server includes indication information, which indicates whether the AI ​​model associated with the first application is deployed on the cloud server.

[0433] In one possible design, the transceiver unit 1100 is further configured to receive a service request from the terminal device, the service request including the identifier of the first application;

[0434] If the AI ​​model associated with the first application is deployed on the cloud server, the processing unit 1200 is also used to respond to the service request by sending the model deployment information of the cloud server through the transceiver unit 1100.

[0435] The model deployment information of the cloud server includes information on at least one sub-model deployed on the cloud server in the first AI model.

[0436] The detailed description and beneficial effects of the device embodiment shown in Figure 9 can be found in the description of the foregoing method embodiment, and will not be repeated here.

[0437] The communication device and cloud server of this application embodiment have been described above. The following describes possible product forms of the communication device and cloud server. It should be understood that any product with the functions of the communication device described in FIG. 9 above, or any product with the functions of the cloud server described in FIG. 9 above, falls within the protection scope of this application embodiment. It should also be understood that the following description is merely illustrative and does not limit the product forms of the communication device and cloud server of this application embodiment to these forms.

[0438] In one possible implementation, in the communication device shown in FIG9, the processing unit 1200 may be one or more processors, and the transceiver unit 1100 may be a transceiver. Alternatively, the transceiver unit 1100 may also be a transmitting unit and a receiving unit, where the transmitting unit may be a transmitter and the receiving unit may be a receiver. The transmitting unit and the receiving unit are integrated into a single device, such as a transceiver. In the embodiments of this application, the processor and the transceiver may be coupled, etc. The connection method between the processor and the transceiver is not limited in the embodiments of this application.

[0439] Figure 10 is a schematic diagram of another communication device 2000 provided in an embodiment of this application. The communication device in Figure 10 may be the communication equipment described above, or it may be the cloud server described above, or it may be the second access network device described above, or it may be the first access network device described above.

[0440] As shown in Figure 10, the communication device 2000 includes one or more processors 2200 and transceivers 2100. The transceiver 2100 can implement the functions of the transceiver unit 1100, and the processor 2200 can implement the functions of the processing unit 1200.

[0441] In various implementations of the communication device shown in Figure 10, the transceiver may include a receiver for performing a receiving function (or operation) and a transmitter for performing a transmitting function (or operation). The transceiver is also used to communicate with other devices / appliances via a transmission medium.

[0442] Optionally, the communication device 2000 may further include one or more memories 2300 for storing program instructions and / or data. The memory 2300 is coupled to the processor 2200. The coupling in this embodiment is an indirect coupling or communication connection between devices, units, or modules, and can be electrical, mechanical, or other forms, used for information exchange between devices, units, or modules. The processor 2200 may operate in conjunction with the memory 2300. The processor 2200 can execute program instructions stored in the memory 2300.

[0443] This embodiment does not limit the specific connection medium between the transceiver 2100, processor 2200, and memory 2300. In Figure 10, the transceiver 2100, processor 2200, and memory 2300 are connected via a bus 2400, indicated by a thick line. The connection methods between other components are merely illustrative and not intended to be limiting. The bus can be categorized as an address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used in Figure 10, but this does not imply that there is only one bus or one type of bus.

[0444] In the embodiments of this application, the processor may be a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., and can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly manifested as being executed by a hardware processor, or being executed by a combination of hardware and software modules within the processor.

[0445] In this application embodiment, the memory may include, but is not limited to, non-volatile memory such as hard disk drive (HDD) or solid-state drive (SSD), random access memory (RAM), erasable programmable read-only memory (EPROM), read-only memory (ROM), or compact disc read-only memory (CD-ROM), etc. Memory is any storage medium capable of carrying or storing program code having instruction or data structure forms, and capable of being read and / or written by a computer (such as the communication device shown in this application), but is not limited to this. The memory in this application embodiment may also be a circuit or any other device capable of implementing storage functions, used to store program instructions and / or data.

[0446] The processor 2200 is primarily used for processing communication protocols and data, controlling the entire communication device, executing software programs, and processing software program data. The memory 2300 is primarily used for storing software programs and data. The transceiver 2100 may include control circuitry and an antenna. The control circuitry is primarily used for converting baseband signals to radio frequency signals and processing radio frequency signals. The antenna is primarily used for transmitting and receiving radio frequency signals in the form of electromagnetic waves. Input / output devices, such as touchscreens, displays, and keyboards, are primarily used for receiving user input data and outputting data to the user.

[0447] When the communication device is powered on, the processor 2200 can read the software program in the memory 2300, interpret and execute the instructions of the software program, and process the data of the software program. When data needs to be transmitted wirelessly, the processor 2200 performs baseband processing on the data to be transmitted and outputs the baseband signal to the radio frequency (RF) circuit. The RF circuit processes the baseband signal and transmits the RF signal outward in the form of electromagnetic waves through the antenna. When data is sent to the communication device, the RF circuit receives the RF signal through the antenna, converts the RF signal into a baseband signal, and outputs the baseband signal to the processor 2200. The processor 2200 converts the baseband signal into data and processes the data.

[0448] In another implementation, the radio frequency circuitry and antenna can be set up independently of the processor performing baseband processing. For example, in a distributed scenario, the radio frequency circuitry and antenna can be arranged remotely, independent of the communication device.

[0449] It is understood that the communication device shown in the embodiments of this application may have more components than those in Figure 10, and the embodiments of this application do not limit this. The methods executed by the processor and transceiver shown above are only examples, and the specific steps executed by the processor and transceiver can be referred to the methods described above.

[0450] In another possible implementation, in the communication device shown in FIG9, the processing unit 1200 can be one or more logic circuits, and the transceiver unit 1100 can be an input / output interface, or a communication interface, or an interface circuit, or an interface, etc. Alternatively, the transceiver unit 1100 can also be a transmitting unit and a receiving unit. The transmitting unit can be an output interface, and the receiving unit can be an input interface. The transmitting unit and the receiving unit are integrated into one unit, such as an input / output interface. As shown in FIG11, the communication device shown in FIG11 includes a logic circuit 3001 and an interface 3002. That is, the above-mentioned processing unit 1200 can be implemented using the logic circuit 3001, and the transceiver unit 1100 can be implemented using the interface 3002. Among them, the logic circuit 3001 can be a chip, a processing circuit, an integrated circuit, or a system on chip (SoC) chip, etc., and the interface 3002 can be a communication interface, an input / output interface, a pin, etc. For example, FIG11 uses the above-mentioned communication device as a chip, which includes a logic circuit 3001 and an interface 3002.

[0451] In this embodiment, the logic circuit and the interface can also be coupled to each other. The specific connection method between the logic circuit and the interface is not limited in this embodiment.

[0452] It is understood that the communication device shown in the embodiments of this application can implement the method provided in the embodiments of this application in hardware form or in software form, etc., and the embodiments of this application do not limit it in this way.

[0453] This application also provides a wireless communication system, which includes a communication device and a terminal device, and the communication device and the terminal device can be used to perform the methods in any of the foregoing embodiments.

[0454] Optionally, the communication system may also include a cloud server.

[0455] Optionally, the communication system may also include a second access network device.

[0456] Furthermore, this application also provides a computer-readable storage medium storing computer code. When the computer code is executed on a computer, it causes the computer to perform the operations and / or processes performed by the communication device in the method provided in this application, or causes the computer to perform the operations and / or processes performed by the cloud server in the method provided in this application, or causes the computer to perform the steps and / or processes performed by the second access network device in the method provided in this application, or causes the computer to perform the steps and / or processes performed by the first access network device in the method provided in this application, or causes the computer to perform the steps and / or processes performed by the model management network element in the method provided in this application.

[0457] This application also provides a computer program product, which includes computer code or a computer program that, when run on a computer, causes the operations and / or processes performed by the communication device or cloud server or second access network device or first access network device or model management network element in the method provided in this application to be executed.

[0458] In the embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or units, or it may be an electrical, mechanical, or other form of connection.

[0459] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected according to actual needs to achieve the technical effects of the solutions provided in the embodiments of this application.

[0460] Furthermore, the functional units in the various embodiments of this application 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 unit can be implemented in hardware or as a software functional unit.

[0461] 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 computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes, 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 readable storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned readable storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0462] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A communication method characterized by comprising: include: Receive a service request from a terminal device, the service request including the identifier of a first application or the identifier of a first artificial intelligence (AI) model; Based on the communication status information between the terminal device and the first access network device, determine at least one of the model configuration information and task orchestration information; If the service request includes the identifier of the first application, the model configuration information includes the identifier of the first AI model determined for the first application and the configuration information of the first AI model. If the service request includes the identifier of the first AI model, the model configuration information is the configuration information of the first AI model; The task orchestration information is the task orchestration information of the first AI model; The first access network device is the access network device of the serving cell where the terminal device is located; the first AI model is deployed at least on the first access network device.

2. The method of claim 1, wherein, The communication status information includes at least one of the following: current wireless communication information between the terminal device and the first access network device, historical wireless communication information between the terminal device and the first access network device, or predicted wireless communication information between the terminal device and the first access network device.

3. The method of claim 1 or 2, wherein, The step of determining at least one of the following based on the communication status information between the terminal device and the first access network device, namely model configuration information and task orchestration information, includes: Based on the communication status information and the first information, determine the model configuration information; and / or, Based on the communication status information and the second information, the task scheduling information is determined; The first information includes at least one of the following: computing status information of the first access network device, model deployment information of the first access network device, location movement information of the terminal device, and usage information of the terminal device for the first application; The second information includes at least one of the following: the computing status information of the first access network device, and the model deployment information of the first access network device; The computing status information of the first access network device includes at least one of the following: computing resource information of the first access network device, and computing resource occupancy information of the first access network device; The model deployment information of the first access network device includes information about the AI ​​models that have been deployed on the first access network device.

4. The method of claim 3, wherein, The method further includes: The device receives at least one of the following: the computing status information of the first access network device, the model deployment information of the first access network device, the location movement information of the terminal device, and the usage information of the terminal device for the first application.

5. The method of claim 3 or 4, wherein, The method further includes: Send at least one of the model configuration information and the task orchestration information to the first access network device.

6. The method according to any one of claims 3 to 5, wherein, The first AI model is also deployed on the terminal device; The first information and / or the second information further include at least one of the following: the computing status information of the terminal device, and the model deployment information of the terminal device; The computing status information of the terminal device includes at least one of the following: computing resource information of the terminal device, and computing resource occupancy information of the terminal device; The model deployment information of the terminal device includes information about the AI ​​models that have been deployed on the terminal device.

7. The method of claim 6, wherein, The method further includes: Receive at least one of the following from the terminal device: the terminal device's computing status information and the terminal device's model deployment information.

8. The method of claim 6 or 7, wherein, The method further includes: Send at least one of the model configuration information and the task orchestration information to the terminal device.

9. The method according to any one of claims 3 to 8, wherein, The first AI model is also deployed in the second access network device; The first information and / or the second information further include at least one of the following: the computing status information of the second access network device, and the model deployment information of the second access network device; The computing status information of the second access network device includes at least one of the following: computing resource information of the second access network device, and computing resource occupancy information of the second access network device; The model deployment information of the second access network device includes information about the AI ​​models that have been deployed on the second access network device.

10. The method of claim 9, wherein, The method further includes: Receive at least one of the following from the second access network device: the computing status information of the second access network device and the model deployment information of the second access network device.

11. The method of claim 9 or 10, wherein, The method further includes: Send at least one of the model configuration information and the task orchestration information to the second access network device.

12. The method according to any one of claims 4 to 11, wherein, The first AI model is also deployed on a cloud server; The first information and / or the second information also include the model deployment information of the cloud server; The model deployment information of the cloud server includes information on at least one sub-model deployed on the cloud server in the first AI model.

13. The method of claim 12, wherein, The method further includes: Receive model deployment information of the cloud server from the cloud server.

14. The method of any one of claims 1-13, wherein, The method further includes: Receive the contract information of the first application from the cloud server; The step of determining at least one of the following based on the communication status information between the terminal device and the first access network device, namely model configuration information and task orchestration information, includes: Based on the communication status information and the contract information, determine at least one of the model configuration information and the task orchestration information.

15. The method of claim 14, wherein, The contract information includes at least one of the following: Information on at least one alternative model supported by the first application, model compilation method, QoS parameters of the first application, runtime environment parameters of the first application, SLA parameters of the first application, and model requirement information of the cloud server; The first AI model is determined from the at least one candidate model; The model requirement information of the cloud server includes indication information, which indicates whether the AI ​​model associated with the first application is deployed on the cloud server.

16. A method of communication, comprising: include: Receive a registration request from a terminal device, the registration request including the identifier of the first application; In response to the registration request, the contract information of the first application is sent, wherein the contract information is used to determine at least one of the model configuration information and task orchestration information associated with the first application.

17. The method of claim 16, wherein, The contract information includes at least one of the following: Information on at least one alternative model supported by the first application, model compilation method, QoS parameters of the first application, runtime environment parameters of the first application, SLA parameters of the first application, and model requirement information of the cloud server; The model configuration information associated with the first application includes the identifier of a first AI model determined for the first application, wherein the first AI model is determined from the at least one candidate model. The model requirement information of the cloud server includes indication information, which indicates whether the AI ​​model associated with the first application is deployed on the cloud server.

18. The method of claim 16 or 17, wherein, The method further includes: Receive a service request from the terminal device, the service request including the identifier of the first application or the identifier of the first AI model; If the AI ​​model associated with the first application is deployed on the cloud server, in response to the service request, the model deployment information of the cloud server is sent. The model deployment information of the cloud server includes information on at least one sub-model deployed on the cloud server in the AI ​​model associated with the first application.

19. A communications device, characterized by The communication device includes a unit for performing the method as described in any one of claims 1-15; or includes a unit for performing the method as described in any one of claims 16-18.

20. A communications device, characterized by Including the processor; The processor is coupled to the memory; The memory is used to store instructions; The processor is configured to execute the instructions to cause the method described in any one of claims 1-15 to be performed; or... The processor is configured to execute the instructions to cause the method described in any one of claims 16-18 to be performed.

21. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program that, when executed, performs the method according to any one of claims 1-15; or, performs the method according to any one of claims 16-18.

22. A computer program product, characterised in that, The computer program product includes a computer program that, when run on a computer, executes the method according to any one of claims 1-15; or, executes the method according to any one of claims 16-18.

Images