Artificial intelligence model deployment method and device

Abstract

The invention discloses an AI model deployment method and device. The method comprises the following steps: a first access network device receives a first request from a terminal, and the first request comprises an identifier of a first AI model; the first access network device determines a first AI model; wherein the first AI model is one of at least one AI model cached by the first access network device or the second access network device, and the second access network device is adjacent to the first access network device; and the first access network device sends the first AI model to the terminal. According to the embodiment of the invention, the AI model is cached in the access network device, so that the transmission of the AI model can be realized independent of a PDU session mechanism, and the effects of protecting user privacy, reducing network resource occupation, improving model deployment efficiency and the like can be considered.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to an artificial intelligence (AI) model deployment method and device. Background Art

[0002] With the development of AI technology, more and more companies are using AI models to provide services to terminals. Currently, there are two main modes for terminals to use AI models:

[0003] (1) Online mode: The terminal uses the protocol data unit (PDU) session mechanism to call the application program interface (API) of the AI ​​model in the data network (DN) to access the AI ​​model. However, the online mode has the risk of leaking user privacy and is unavailable to users offline, resulting in a poor experience.

[0004] (2) Offline mode: The terminal uses the PDU session mechanism to download the AI ​​model in the DN to the local for deployment, and directly accesses the AI ​​model locally. Although the offline mode can ensure that users can use the AI ​​model offline and reduce the risk of privacy leakage, the terminal needs to occupy more network resources when downloading the AI ​​model because the amount of data of the AI ​​model is very large.

[0005] Moreover, both of the above methods rely on the PDU session mechanism to transmit AI model data. Therefore, the operator's network only provides access and connection services for terminals and AI model providers, that is, the AI ​​model is transparently transmitted, and the data transmission efficiency of the AI ​​model is low.

[0006] Therefore, how to deploy AI models to balance protecting user privacy, reducing network resource usage and improving the deployment efficiency of AI models is a technical problem to be solved in this application. Summary of the invention

[0007] The present application provides an AI model deployment method and device, which can achieve the technical effects of protecting user privacy, reducing network resource usage and improving the deployment efficiency of AI models.

[0008] In a first aspect, a method for deploying an AI model is provided, which can be applied to an access network device. The access network device can be an access network device or a chip in the access network device, and the method includes: a first access network device receives a first request from a terminal, and the first request includes an identifier of a first AI model; the first access network device determines the first AI model; wherein the first AI model is one of at least one AI model cached by the first access network device or the second access network device, and the second access network device is adjacent to the first access network device; the first access network device sends the first AI model to the terminal.

[0009] In an embodiment of the present application, an access network device (such as a first access network device and a second access network device) has the function of caching an AI model. After receiving the first request from the terminal, the first access network device can send the AI ​​model cached by itself or an adjacent access network device (such as a second access network device) to the terminal. In an embodiment of the present application, the transmission of the AI ​​model can be implemented independently of the PDU session mechanism. The terminal does not need to access the API of the AI ​​model in the DN through the PDU session mechanism, but downloads the AI ​​model to local use, thereby protecting user privacy; it does not need to download the AI ​​model from the DN through the PDU session mechanism, thereby reducing the occupation of network resources (such as network load, bandwidth, etc.); and the AI ​​model can be cached in the access network device, and the transmission path of the AI ​​model is short, which can improve the transmission efficiency of the AI ​​model, thereby improving the model deployment efficiency. It can be seen that the embodiment of the present application can take into account the protection of user privacy, the reduction of network resource occupation and the improvement of the deployment efficiency of the AI ​​model.

[0010] In one possible design, the source of the first AI model cached by the first access network device may be other adjacent access network devices or core network elements, without limitation. For example, the first access network device may obtain the first AI model from the second access network device and save the first AI model to the cache of the first access network device; or, the first access network device may receive the first AI model from the first data processing function network element and save the first AI model to the cache of the first access network device.

[0011] In one possible design, the first request may be a request sent to the first access network device. For example, the first request may be carried in a signaling radio bearer (SRB). Of course, the signaling radio bearer is only one possible implementation method, and the actual method of the terminal sending the first request to the first access network device is not limited to this.

[0012] In this way, the terminal can request the first AI model from the access network.

[0013] In one possible design, the first access network device determines the first AI model, including: the first access network device queries whether there is the first AI model in the cache of the first access network device according to the first request. If the first AI model is in the cache of the first access network device, the first access network device sends the first AI model cached by itself to the terminal.

[0014] In this way, when there is a first AI model in the cache of the first access network device, the first access network device preferentially returns the first AI model cached by itself to the terminal, shortening the transmission path as much as possible and improving the model deployment efficiency.

[0015] In one possible design, if the first access network device does not have the first AI model in the cache, the first access network device may query the second access network device for the first AI model, and the first access network device receives the first AI model sent by the second access network device.

[0016] In this way, when the adjacent access network device has the first AI model cached, the first access network device can obtain the first AI model from the adjacent access network device and return it to the terminal without obtaining the first AI model from the core network, thereby shortening the transmission path and improving the model deployment efficiency.

[0017] In one possible design, after the first access network device receives the first AI model sent by the second access network device, the first access network device may also save the first AI model to the cache of the first access network device.

[0018] In this way, the next time the first access network device receives a request for the first AI model, it can send the first AI model cached by itself to the terminal, thereby improving the efficiency of model deployment.

[0019] In one possible design, the first request may be a request sent to an application function. For example, the first request is carried in a data radio bearer (DRB). Accordingly, after receiving the first request, the first access network device may forward the first request to the application function network element. Afterwards, the first access network device may receive the first AI model from the first data processing function network element, and may also save the first AI model to a cache of the first access network device.

[0020] In this way, the first access network device can receive the first AI model from the core network (such as a data processing function network element). This method can be independent of the PDU session mechanism and can reduce the occupation of network resources. At the same time, after receiving the first AI model, the first access network device can cache the first AI model. Then, the next time the first access network device receives a request for the first AI model, it can send the cached first AI model to the terminal, thereby improving the efficiency of model deployment.

[0021] In one possible design, after the first access network device receives the first request from the terminal, the first access network device may also send a first duration to the terminal, where the first duration is the duration that the terminal waits for the first AI model to start transmitting.

[0022] Since the data volume of the AI ​​model is generally large and requires a certain amount of transmission time, the first access network device can instruct the terminal on the waiting time. On the one hand, this can avoid the problem of low model deployment efficiency caused by the terminal waiting for a long time when the network does not have the first AI model; on the other hand, it can avoid the problem of the terminal mistakenly thinking that the model deployment has failed and frequently sending requests to the network when the network has the first AI model.

[0023] In one possible design, the first access network device can also receive update information from the first data processing function network element, and the update information is used to update the first AI model.

[0024] In this way, the AI ​​model cached in the access network device can be updated to improve the user experience.

[0025] In a second aspect, a method for deploying an AI model is provided, which can be applied to a terminal, which may be a terminal device or a chip or device in the terminal device, comprising: the terminal sends a first request to a first access network device, the first request including an identifier of a first AI model; the terminal receives the first AI model from the first access network device; wherein the first AI model is one of at least one AI model cached by the first access network device or the second access network device, and the second access network device is adjacent to the first access network device.

[0026] In the embodiment of the present application, the terminal can obtain the AI ​​model from the access network device (such as the first access network device and the second access network device), so that the transmission of the AI ​​model can be realized independently of the PDU session mechanism. The terminal does not need to access the API of the AI ​​model in the DN through the PDU session mechanism, but downloads the AI ​​model to the local for use, thereby protecting user privacy, and does not need to download the AI ​​model from the DN through the PDU session mechanism, thereby reducing the occupation of network resources (such as network load, bandwidth, etc.). In addition, the AI ​​model can be cached in the access network device, and the transmission path of the AI ​​model is short, which can improve the transmission efficiency of the AI ​​model, thereby improving the efficiency of model deployment.

[0027] In one possible design, the first request may be a request sent to the first access network device. For example, the first request may be carried in a signaling radio bearer. Of course, the signaling radio bearer is only one possible implementation method, and the actual method of the terminal sending the first request to the first access network device is not limited thereto.

[0028] In this way, it is possible to obtain the first AI model from the access network.

[0029] In a possible design, the first request may be a request sent to the AF. For example, the first request is carried in a data radio bearer.

[0030] In this way, it is possible to obtain the first AI model from the core network.

[0031] In one possible design, after the terminal sends a first request to the first access network device and before the terminal receives the first AI model from the first access network device, it also includes: if the terminal waits for more than a preset time, or the terminal receives information from the first access network device indicating that the deployment of the first AI model has failed, the terminal sends a second request to the application function network element through the first access network device, and the second request includes an identifier of the first AI model, and the second request is carried in the data wireless bearer.

[0032] In this way, the terminal first tries to obtain the AI ​​model from the access network, and then obtains the AI ​​model from the core network if it fails. This can ensure that the terminal obtains the AI ​​model and improve the reliability of the solution.

[0033] In one possible design, after the terminal sends a first request to the first access network device and before the terminal receives the first AI model from the first access network device, it also includes: the terminal receives a first time length from the first access network device; and the terminal waits for the first AI model to start transmitting within the first time length.

[0034] In this way, the terminal can wait for the first AI model within the time length indicated by the first access network device. On the one hand, it can avoid the problem of low model deployment efficiency caused by the terminal waiting for the first AI model when the network does not have the first AI model. On the other hand, it can avoid the problem of the terminal frequently sending requests to the network when the network has the first AI model.

[0035] In one possible design, the preset duration is the second duration, and the second duration is the duration of a timer configured for the terminal.

[0036] In this way, the terminal waits for the first AI model within the duration of the timer. On the one hand, it can avoid the problem of low model deployment efficiency caused by the terminal waiting for the first AI model when the network does not have the first AI model. On the other hand, it can avoid the problem of the terminal frequently sending requests to the network when the network has the first AI model.

[0037] According to a third aspect, an AI model deployment method is provided, which can be applied to a data processing function network element or a chip in a data processing function network element. Taking the application of the method to a first data processing function network element as an example, the method includes: the first data processing function network element receives first configuration information from a control function network element, and the first configuration information includes an identifier of a first access network device; the first data processing function network element receives a first AI model from a first data storage function network element; and the first data processing function network element sends the first AI model to the first access network device according to the identifier of the first access network device.

[0038] In an embodiment of the present application, the first data processing function network element can transmit the first AI model to the first access network device according to the configuration of the control function network element, so that the first access network device caches and sends the AI ​​model to the terminal. This solution can realize AI model deployment independently of the PDU session mechanism, and can take into account the effects of protecting user privacy and reducing network resource usage.

[0039] In one possible design, the first configuration information also includes a transmission protocol. The transmission protocol is related to the first data pipe, which is a data pipe between the first data storage functional network element and the first data processing functional network element; or, the transmission protocol is related to the first data pipe identifier, which is used to indicate the data pipe between the first data storage functional network element and the first data processing functional network element. Accordingly, the first data processing functional network element receives the first AI model from the first data storage functional network element, including: the first data processing functional network element receives the first AI model from the first data storage functional network element based on the transmission protocol.

[0040] In this way, the control function network element can flexibly configure the transmission protocol between the first data storage function network element and the first data processing function network element according to the type of data to be transmitted between the first data storage function network element and the first data processing function network element, thereby improving the data (such as AI model) transmission efficiency between the first data storage function network element and the first data processing function network element.

[0041] In one possible design, after the first data processing function network element sends the first AI model to the first access network device, it can also send the identifier of the first access network device and the identifier of the first AI model to the control function network element.

[0042] In this way, it is convenient for the control function network element to update the model cache information of the first access network device, so as to better control and manage the AI ​​model cached by the first access network device.

[0043] In one possible design, the first data processing function network element can also receive update information from the first data storage function network element, and the update information is used to update the first AI model; the first data processing function network element sends the update information to the first access network device.

[0044] In this way, the first data processing functional network element can assist the first data storage functional network element in sending update information to the first access network device, thereby updating the AI ​​model cached by the access network device and improving user experience.

[0045] In a fourth aspect, a method for deploying an AI model is provided. The method can be applied to a data storage functional network element or a chip in a data storage functional network element. Taking the method applied to the first data storage functional network element as an example, the method includes: the first data storage functional network element receives second configuration information from the control functional network element, the second configuration information includes an identifier of the first data processing functional network element and an identifier of the first AI model; the first data storage functional network element sends the first AI model to the first data processing functional network element.

[0046] In an embodiment of the present application, the first data storage functional network element can transmit the first AI model to the first data processing functional network element according to the configuration of the control functional network element, so that the data processing functional network element sends the first AI model to the access network device. This solution can realize AI model deployment independently of the PDU session mechanism, and can take into account the effects of protecting user privacy and reducing network resource usage.

[0047] In one possible design, the second configuration information also includes a transmission protocol. The transmission protocol is related to the first data pipeline, and the first data pipeline is a data pipeline between the first data storage functional network element and the first data processing functional network element; or, the transmission protocol is related to the first data pipeline identifier, and the first data pipeline identifier is used to indicate the data pipeline between the first data storage functional network element and the first data processing functional network element. Accordingly, the first data storage functional network element sends the first AI model to the first data processing functional network element, including: the first data storage functional network element sends the first AI model to the first data processing functional network element based on the transmission protocol.

[0048] In this way, the data transmission efficiency of the first AI model between the first data storage functional network element and the first data processing functional network element can be improved.

[0049] In one possible design, the first data storage functional network element can also receive subscription information from the control functional network element, and the subscription information includes an identifier of the first AI model; when the first AI model stored in the first data storage functional network element is updated, the first data storage functional network element sends notification information to the control functional network element according to the subscription information, and the notification information is used to indicate that the first AI model has been updated.

[0050] In this way, when the first AI model stored in the first data storage functional network element is updated, the control functional network element can be notified in time so that the control functional network element can control and update the first AI model cached by the access network device in time.

[0051] In one possible design, the first data storage functional network element can also send update information to the first data processing functional network element, and the update information is used to update the first AI model.

[0052] In this way, the AI ​​model cached in the access network device can be updated to improve the user experience.

[0053] In a fifth aspect, a method for deploying an AI model is provided, which can be applied to a control function network element or a chip in a control function network element. Taking the application of the method to a control function network element as an example, the method includes: the control function network element receives a third request from an application function network element or a user plane network element, and the third request includes an identifier of a first AI model; the control function network element determines a first data storage function network element and a first data processing function network element according to the third request; the control function network element sends first configuration information to the first data processing function network element and sends second configuration information to the first data storage function network element; wherein the first configuration information includes an identifier of a first access network device, and the second configuration information includes an identifier of the first data processing function network element and an identifier of the first AI model.

[0054] In an embodiment of the present application, the control function network element can configure the first data processing function network element and the first data storage function network element to transmit the first AI model to the first access network device. This solution can implement AI model deployment independently of the PDU session mechanism, and can take into account the effects of protecting user privacy and reducing network resource usage.

[0055] In one possible design, the first configuration information and the second configuration information also include a transmission protocol. The transmission protocol is related to the first data pipeline, which is a data pipeline between the first data storage function network element and the first data processing function network element; or the transmission protocol is related to the first data pipeline identifier, which is used to indicate the data pipeline between the first data storage function network element and the first data processing function network element.

[0056] In this way, the data transmission efficiency of the first AI model between the first data storage functional network element and the first data processing functional network element can be improved.

[0057] In one possible design, the data packet containing the second request carries the address information of the terminal; the control function network element can also query the identifier of the access network device corresponding to the terminal from the user plane function network element based on the address information of the terminal to obtain the identifier of the first access network device.

[0058] In this way, the first AI model can be sent to the access network device (ie, the first access network device) that provides services to the terminal, thereby ensuring the reliability of the AI ​​model transmission.

[0059] In one possible design, the control function network element can also receive the identifier of the first access network device and the identifier of the first AI model from the first data processing function network element; update the model cache information according to the identifier of the first access network device and the identifier of the first AI model, and the model cache information includes information of the AI ​​model cached on at least one access network device.

[0060] In this way, the control function network element can update the model cache information of the first access network device to better control and manage the AI ​​model cached by the first access network device.

[0061] In one possible design, the control function network element may also send subscription information to the first data storage function network element, where the subscription information includes an identifier of the first AI model; the control function network element receives notification information from the first data storage function network element, where the notification information is used to indicate that the first AI model has been updated; the control function network element configures the first data storage function network element and the first data processing function network element to send update information to the first access network device according to the model cache information, where the update information is used to update the first AI model.

[0062] In this way, the control function network element can timely access the first AI model cached by the network device to update it when the first AI model is updated, thereby improving user experience.

[0063] In a sixth aspect, a method for deploying an AI model is provided, including: a control function network element receives a third request from an application function network element or a user plane network element, the third request including an identifier of a first AI model; the control function network element determines a first data storage function network element and a first data processing function network element according to the third request; the control function network element sends first configuration information to the first data processing function network element and sends second configuration information to the first data storage function network element; wherein the first configuration information includes an identifier of a first access network device, and the second configuration information includes an identifier of the first data processing function network element and an identifier of the first AI model; the first data storage function network element receives the second configuration information from the control function network element, and sends the first AI model to the first data processing function network element; the first data processing function network element receives the first configuration information from the control function network element, and receives the first AI model from the first data storage function network element; and sends the first AI model to the first access network device according to the identifier of the first access network device.

[0064] In a seventh aspect, a communication device is provided, which includes a module or unit or technical means for executing the method described in the first aspect or any possible design of the first aspect.

[0065] For example, the device may include:

[0066] A transceiver module, configured to receive a first request from a terminal, where the first request includes an identifier of a first AI model;

[0067] A processing module, configured to determine a first AI model; wherein the first AI model is one of at least one AI model cached by the first access network device or the second access network device, and the second access network device is adjacent to the first access network device;

[0068] The transceiver module is also used to send the first AI model to the terminal.

[0069] In an eighth aspect, a communication device is provided, which includes a module or unit or technical means for executing the method described in the second aspect or any possible design of the second aspect.

[0070] For example, the device may include:

[0071] A transceiver module is used to send a first request to a first access network device, wherein the first request includes an identifier of a first AI model; and receive the first AI model from the first access network device; wherein the first AI model is one of at least one AI model cached by the first access network device or the second access network device, and the second access network device is adjacent to the first access network device.

[0072] In a ninth aspect, a communication device is provided, which includes a module or unit or technical means for executing the method described in the third aspect or any possible design of the third aspect.

[0073] For example, the device may include:

[0074] A transceiver module is configured to receive first configuration information from a control function network element, wherein the first configuration information includes an identifier of a first access network device; and receive a first AI model from a first data storage function network element;

[0075] A processing module is used to send the first AI model to the first access network device according to the identifier of the first access network device.

[0076] In a tenth aspect, a communication device is provided, which includes a module or unit or technical means for executing the method described in the fourth aspect or any possible design of the fourth aspect.

[0077] For example, the device may include:

[0078] The transceiver module is used to receive second configuration information from the control function network element, where the second configuration information includes an identifier of the first data processing function network element and an identifier of the first AI model; and send the first AI model to the first data processing function network element.

[0079] In the eleventh aspect, a communication device is provided, which includes a module or unit or technical means for executing the method described in the fifth aspect or any possible design of the fifth aspect.

[0080] For example, the device may include:

[0081] A transceiver module, configured to receive a third request from an application function network element or a user plane network element, where the third request includes an identifier of the first AI model;

[0082] A processing module, configured to determine a first data storage function network element and a first data processing function network element according to a third request;

[0083] The transceiver module is also used to send first configuration information to the first data processing function network element and to send second configuration information to the first data storage function network element; wherein the first configuration information includes the identifier of the first access network device, and the second configuration information includes the identifier of the first data processing function network element and the identifier of the first AI model.

[0084] In the twelfth aspect, a communication device is provided, comprising a processor and an interface circuit, wherein the interface circuit is used to receive signals from other communication devices outside the communication device and transmit them to the processor, or to send signals from the processor to other communication devices outside the communication device, and the processor is used to implement the method described in the first aspect or any possible design of the first aspect or the second aspect or any possible design of the second aspect or the third aspect or any possible design of the third aspect or the fourth aspect or any possible design of the fourth aspect or the fifth aspect or any possible design of the fifth aspect through logic circuits or execution instructions.

[0085] In a possible design, the communication device further includes a memory for storing the instructions. Optionally, the memory and the processor are integrated together.

[0086] In the thirteenth aspect, a computer-readable storage medium is provided, wherein a computer program or instruction is stored in the storage medium. When the computer program or instruction is executed by a processor, the method described in the first aspect or any possible design of the first aspect, the second aspect or any possible design of the second aspect, the third aspect or any possible design of the third aspect, the fourth aspect or any possible design of the fourth aspect, or the fifth aspect or any possible design of the fifth aspect is implemented.

[0087] In the fourteenth aspect, a computer program product is provided, comprising a computer program or instructions, which, when executed by a processor, enables the method described in the first aspect or any possible design of the first aspect, the second aspect or any possible design of the second aspect, the third aspect or any possible design of the third aspect, the fourth aspect or any possible design of the fourth aspect, or the fifth aspect or any possible design of the fifth aspect to be implemented.

[0088] In a fifteenth aspect, a communication system is provided, comprising a communication device as described in aspects seven to eleven, or comprising a communication device as described in aspect twelfth.

[0089] The beneficial effects of the sixth to fifteenth aspects mentioned above can refer to the beneficial effects of the corresponding designs in the first to fifth aspects, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] Figure 1 A schematic diagram of a possible communication system provided in an embodiment of the present application;

[0091] Figure 2 A schematic diagram of another possible communication system provided in an embodiment of the present application;

[0092] Figure 3 A flowchart of an AI model deployment method provided in an embodiment of the present application;

[0093] Figure 4 A flow chart of obtaining an identifier of a first AI model for a terminal;

[0094] Figure 5 A flow chart for a first access network device to obtain a first AI model from a second access network device;

[0095] Figure 6 A flowchart for obtaining a first AI model from a core network for a terminal;

[0096] Figure 7 A flow chart for obtaining a first AI model from a core network for a first access network device;

[0097] Figure 8 A flowchart for obtaining a first AI model from a core network for a terminal;

[0098] Fig. 9 A flowchart of a method for updating an AI model is provided for an embodiment of the present application;

[0099] Fig.10 A schematic diagram of a communication device 1000 is provided for an embodiment of the present application;

[0100] Fig.11A schematic diagram of another communication device 1100 is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0101] The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0102] The technical solution of the embodiment of the present application can be applied to various communication systems, such as: long term evolution (LTE) system, evolved LTE (LTE-advanced, LTE-A) system, universal mobile telecommunications system (UMTS), and fifth generation (5G) mobile communication system, beyond 5G (B5G) mobile communication system, or sixth generation (6G) and other communication systems evolved after 5G, etc. The communication system can also be a device-to-device (D2D) network, a WiFi network, a machine-to-machine (M2M) network, an Internet of Things (IoT) network or other networks.

[0103] See also Figure 1 , is a schematic diagram of a possible communication system provided in an embodiment of the present application. The network functions and entities included in the system mainly include: user equipment (UE), radio access network ((radio) access network, (R) AN), user plane function (UPF), data network (DN), access and mobility management function (AMF), session management function (SMF), application function (AF), etc.

[0104] Among them, the user equipment may also be called terminal equipment, terminal, mobile station, mobile terminal, etc. The following description takes the terminal as an example.

[0105] The terminals can be widely used in various scenarios, for example, device-to-device (D2D), vehicle to everything (V2X) communication, machine-type communication (MTC), Internet of things (IoT), virtual reality, augmented reality, industrial control, autonomous driving, telemedicine, smart grid, smart furniture, smart office, smart wearable, smart transportation, smart city, etc. The terminal can be a mobile phone, a tablet computer, a laptop computer, a PDA, a mobile internet device (MID), a wearable device (such as a smart watch, a smart bracelet, a pedometer, smart glasses, etc.), a vehicle-mounted device (such as a car, a bicycle, an electric car, an airplane, a ship, a train, a high-speed train, etc.), a satellite terminal, a virtual reality (VR) device, an augmented reality (AR) device, a smart point of sale (POS) machine, a customer-premises equipment (CPE), a wireless terminal in industrial control, a smart home device (such as a refrigerator, a television, an air conditioner, an electric meter, etc.), an intelligent robot, a robotic arm, workshop equipment, a wireless terminal in unmanned driving, a wireless terminal in telemedicine, a wireless terminal in a smart grid, a wireless terminal in transportation safety, a wireless terminal in a smart city, or a wireless terminal in a smart home, a flying device (such as an intelligent robot, a hot air balloon, a drone, an airplane), etc. The terminal device may also be a vehicle device, such as a vehicle device, a vehicle-mounted module, a vehicle-mounted chip, an onboard unit (OBU) or a telematics box (T-BOX), etc. The terminal device may also be other devices with terminal functions, for example, the terminal device may also be a device that functions as a terminal in D2D communication. The embodiments of the present application do not limit the specific technology and specific device form used by the terminal.

[0106] (R)AN may also be referred to as a wireless access network device, an access network device, an access network equipment, a wireless access network equipment, an access network node, an access network element, etc. The following description will be given using an access network device as an example.

[0107] The access network device is used to help the terminal achieve wireless access. Multiple access network devices in a communication system can be nodes of the same type or different types. In some scenarios, the roles of the access network device and the terminal are relative. For example, a helicopter or drone is configured as a mobile base station. For those terminals that access the network through the helicopter or drone, the helicopter or drone is an access network device; but for the base station that the helicopter or drone accesses, the helicopter or drone is a terminal. Access network devices and terminals are sometimes referred to as communication devices.

[0108] In a possible scenario, the access network device may be a base station, an evolved NodeB (eNodeB), a transmitting and receiving point (TRP), a transmitting point (TP), a next generation NodeB (gNB), a next generation base station in a 6th generation (6G) mobile communication system, a base station in a future mobile communication system, a satellite, or an access point (AP) in a WiFi system, an integrated access and backhaul (IAB) node, an access network device in a mobile switching center non-terrestrial network (NTN) communication system, that is, it may be deployed on a high altitude platform or satellite, etc. The access network device may be a macro base station, a micro base station or an indoor station, a relay node or a donor node, or a wireless controller in a CRAN scenario. The access network device may also be a device that functions as a base station in device to device (D2D) communication, vehicle networking communication, drone communication, and machine communication. Optionally, the access network device may also be a server, a wearable device, a vehicle or an onboard device, etc. For example, the access network device in the vehicle to everything (V2X) technology may be a road side unit (RSU).

[0109] In another possible scenario, multiple access network devices collaborate to assist the terminal in achieving wireless access, and different access network devices respectively implement part of the functions of the base station. For example, the access network device may be a centralized unit (CU), a distributed unit (DU), a CU-control plane (CP), a CU-user plane (UP), or a radio unit (RU). The CU and DU may be separately configured, or may be included in the same network element, such as a baseband unit (BBU). The RU may be included in a radio frequency device or a radio frequency unit, such as a remote radio unit (RRU), an active antenna unit (AAU), or a remote radio head (RRH). It is understood that the access network device may be a CU node, a DU node, or a device including a CU node and a DU node. In addition, the CU may be divided into an access network device in the access network RAN, or the CU may be divided into an access network device in the core network (CN), without limitation here.

[0110] In different systems, CU (or CU-CP and CU-UP), DU or RU may also have different names, but those skilled in the art can understand their meanings. For example, in an open RAN (ORAN) system, CU may also be called O-CU (open CU), DU may also be called O-DU, CU-CP may also be called O-CU-CP, CU-UP may also be called O-CU-UP, and RU may also be called O-RU. For the convenience of description, CU, CU-CP, CU-UP, DU and RU are described as examples in this application. Any unit of CU (or CU-CP, CU-UP), DU and RU in this application may be implemented by a software module, a hardware module, or a combination of a software module and a hardware module.

[0111] In the embodiment of the present application, the access network device has a cache function, and the cache (RANcache) of the access network device can save the AI ​​model received from other network elements (such as data storage function network elements) or devices (such as another access network device). In addition, the access network devices can interact with each other and transfer AI models. In addition, the access network device also supports UE to request model deployment and transmits its own cached AI model to the UE.

[0112] DN: A data network that provides business services to users. Generally, the client is located in the terminal and the server is located in the data network. The data network can be a private network, such as a local area network, or an external network that is not controlled by the operator, such as the Internet. It can also be a proprietary network jointly deployed by operators, such as a network that provides Internet protocol (IP) multimedia core network subsystem (IMS) services. The provider of the AI ​​model can provide the AI ​​model to the UE through the DN. For example, the over the top (OTT) server is deployed in the DN, and the UE can request the AI ​​model from the OTT server in the DN.

[0113] SMF: used for session management, terminal IP address allocation and management, selection of endpoints for manageable user equipment plane functions, policy control, or charging function interfaces, and downlink data notification.

[0114] AMF: used for mobility management and access management, etc. For example, it can be the mobility management entity (MME) function in a 4G communication network or the AMF network element in a 5G network.

[0115] AF: Used for data routing affected by applications, access to network open functions, or interaction with the policy framework for policy control.

[0116] UPF: used for packet routing and forwarding, or quality of service (QoS) processing of user plane data.

[0117] See also Figure 2 , is a schematic diagram of another possible communication system provided in an embodiment of the present application. Figure 1 Compared to the system shown, Figure 2 The system shown also includes the following network elements:

[0118] Data storage function (e.g., data storage function, DSF) network element: responsible for data storage functions, such as storing AI models. In specific implementations, DSF network elements can be deployed independently on a physical entity, or integrated with other network elements (such as unified data repository (UDR) network elements) on the same physical entity, which is not limited in the embodiments of this application.

[0119] Data processing function (e.g., data processing function, DPF) network element: responsible for data processing functions. For example, in the implementation of this application, the DPF network element can transmit the AI ​​model to the RAN, and the AI ​​model can be opened and stored on the RAN side. In a specific implementation, the DPF network element can be independently deployed on a physical entity, or it can be integrated with other network elements (such as UPF network elements) on the same physical entity, which is not limited by the embodiments of this application. In a possible example, the DSF network element may be a UDR network element.

[0120] Control function network element: responsible for the control function of AI services, capable of data orchestration for AI data-related network elements (such as data collection coordination function (DCCF) network elements, network data analysis function (NWDAF) network elements, data analysis storage function (ADRF) network elements, DSF network elements, DPF network elements, etc.). Among them, the orchestration includes the selection of network elements, the operation instructions of network elements, the establishment of data channels (selection of transmission protocols), etc. For example, the control function network element is specifically an artificial intelligence service control function (for example: artificial intelligence service control function, AISCF) network element, which can be responsible for orchestrating the selection of DSF and DPF and the data transmission between DSF and DPF. In a specific implementation, the AISCF network element can be a type of service control function (SCF) network element. The AISCF network element can be independently deployed on a physical entity, and the AISCF network element can also be integrated with other network elements on the same physical entity, which is not limited in the embodiments of the present application.

[0121] It should be understood that the names of the above network elements are only examples, and this application does not exclude the situation where the network elements are named differently in the future, or the functions of the network elements are merged. With the evolution of technology, any device or network element that can realize the functions of the above network elements is within the protection scope of this application.

[0122] In addition, in actual applications, the above network architecture may also include other network elements, which is not limited in this application.

[0123] See also Figure 3 , is an AI model deployment method provided in an embodiment of the present application, which can be applied to Figure 1 and Figure 2 The communication system shown in the method includes S301 to S303:

[0124] S301. A terminal sends a first request, and a first access network device receives the first request.

[0125] The first request is used to request the first AI model, and the first request includes an identifier of the first AI model.

[0126] The AI ​​model described in the embodiments of the present application includes but is not limited to a machine learning (ML) model, a deep learning (DL) model, etc. In a specific example, the AI ​​model is a large language model (LLM), that is, a deep learning model trained with a large amount of text data, which can generate natural language text or understand the meaning of language text. It can be understood that in this article, the AI ​​model can be replaced by an ML model, a DL model, or an LLM, etc.

[0127] In a possible implementation, the terminal also obtains an identifier of the first AI model before sending the first request.

[0128] For example, see Figure 4 As shown, before S301, the following is also executed:

[0129] S300a. The terminal sends a model information request to the AF (it can be understood that the model information request can also be replaced by other names) to query which AI models are in the network (specifically, for example, DSF); the AF receives the model information request.

[0130] S300b. AF returns a model information response to the terminal (it can be understood that the model information response can also be replaced by other names), which carries information about the AI ​​models existing in the network, such as the identifier of at least one AI model; the terminal receives the model information response.

[0131] Optionally, the model information request may carry type information to indicate the type of at least one AI model. Accordingly, the AF may return the identifier of at least one AI model of the corresponding type according to the type information.

[0132] S300c: The terminal determines an identifier of an AI model, such as an identifier of the first AI model, from the identifiers of at least one AI model according to the received model information response, and then executes step S301.

[0133] In this way, the terminal can send the first request based on the AI ​​model existing in the network, avoiding requesting the AI ​​model that is not in the network, thereby improving the reliability of the solution.

[0134] Understandably, Figure 4 Optional steps are marked with dashed lines.

[0135] S302. The first access network device determines a first AI model.

[0136] The first AI model is one of at least one AI model cached by the first access network device; or, the first AI model is one of at least one AI model cached by the second access network device, and the second access network device is adjacent to the first access network device.

[0137] It is understandable that the first access network device may have one or more adjacent access network devices. When the first access network device has multiple adjacent access network devices, the second access network device is one of the multiple adjacent access network devices. The adjacent access network device of the first access network device refers to an access network device that can communicate directly with the first access network device, or an access network device that has only one hop to communicate with the first access network device, that is, the next hop of the first access network device is its adjacent access network device.

[0138] In some embodiments, the first access network device determines the first AI model, which can also be replaced by describing as: the first access network device obtains the first AI model, or the first access network device searches for the first AI model, and so on.

[0139] S303. The first access network device sends a first AI model, and the terminal receives the first AI model.

[0140] Specifically, the first access network device may transmit the first AI model based on a wireless bearer between the first access network device and the terminal.

[0141] In a possible implementation, after the first access network device determines the first AI model and before sending the first AI model, the first access network device also establishes a wireless bearer with the terminal. Alternatively, the first access network device may also use a wireless bearer that has been previously established with the terminal (e.g., a wireless bearer used in the PDU session phase) to transmit the first AI model.

[0142] In this way, the first AI model can be transmitted based on wireless bearer to improve the reliability of transmission.

[0143] In a possible implementation, after receiving the first request and before sending the first AI model (or establishing a wireless bearer), the first access network device may also return a first response to the terminal, where the first response is used to indicate that the first access network device has received the first request. Optionally, the first response also indicates that the first access network device has the first AI model cached or has not cached the first AI model.

[0144] In this way, the terminal can determine that the first request is sent successfully.

[0145] In a possible implementation, after determining the first AI model and before sending the first AI model (or establishing a wireless bearer), the first access network device may also send a first duration to the terminal, where the first duration is used to indicate the duration for the terminal to wait for the first AI model to start transmitting. For example, the terminal starts timing after receiving the first duration. If the terminal still does not receive data from the first AI model after the timing exceeds the first duration, the terminal no longer waits for the transmission of the first AI model and determines that the deployment of the first AI model has failed (or the request for the first AI model has failed).

[0146] Optionally, the duration may be carried in the first response.

[0147] Optionally, the first duration returned when the first access network device caches the first AI model may be different from the first duration returned when the first access network device does not cache the first AI model. For example, when the first access network device caches the first AI model, the first duration is a first value, and when the first access network device does not cache the first AI model, the first duration is a second value, and the first value is less than the second value. In this way, the possibility of obtaining the first AI model from other access network devices or core network elements when the first access network device does not cache the first AI model is taken into account, which can improve the success rate of the deployment of the first AI model.

[0148] Alternatively, the first duration returned when the first access network device has the first AI model cached is the same as the first duration returned when the first access network device does not have the first AI model cached. In this way, the implementation complexity can be reduced.

[0149] In a possible implementation, the terminal is configured with a timer. After sending the first request (or receiving the first response), the terminal starts the timer. If the terminal has not received the data of the first AI model after the timer times out (such as exceeding the second time period), the terminal no longer waits for the transmission of the first AI model, and determines that the deployment of the first AI model has failed (or the request for the first AI model has failed).

[0150] Optionally, in a specific implementation, the first duration and the timer configuration may also exist at the same time, and the terminal may determine whether the first AI model deployment fails based on either one of them.

[0151] As an example, see Figure 4 As shown, after S302 and before S303, the following is also executed:

[0152] S304a. The first access network device sends a first response to the terminal, and the terminal receives the first response.

[0153] Optionally, the first response carries the first duration.

[0154] S304b. The first access network device sends a radio bearer establishment request to the terminal (it can be understood that the radio bearer establishment request can also be replaced by other names), and the terminal receives the radio bearer establishment request.

[0155] S304c. The terminal sends a radio bearer establishment response to the first access network device (it can be understood that the radio bearer establishment response can also be replaced by other names), and the first access network device receives the radio bearer establishment response.

[0156] After the radio bearer is established, the first AI model can be transmitted based on the radio bearer (ie, executing S303 ).

[0157] It can be understood that the embodiment of the present application takes the case where the terminal requests one AI model (i.e., the first AI model) at a time as an example. However, in actual applications, the terminal can also request multiple AI models at a time. For example, the first request in S301 may carry not only the identifier of the first AI model but also the identifiers of other AI models. In S303, the first access network device may send not only the first AI model but also other AI models. For the acquisition method and transmission method of other AI models, reference may be made to the acquisition method and transmission method for the first AI model, and no further description will be given.

[0158] In an embodiment of the present application, an access network device (such as a first access network device, a second access network device, etc.) has the function of caching an AI model. After receiving the first request from the terminal, the first access network device can send the cached AI model of itself or an adjacent access network device (such as a second access network device) to the terminal. It can be seen that in the embodiment of the present application, the deployment of the AI ​​model can be implemented independently of the PDU session mechanism. The terminal does not need to access the API of the AI ​​model in the DN through the PDU session mechanism, but downloads the AI ​​model to local use, so user privacy can be protected; the terminal does not need to download the AI ​​model from the DN through the PDU session mechanism, so the occupancy of network resources (such as network load, bandwidth, etc.) can be reduced; and the AI ​​model can be cached in the access network device, and the transmission path of the AI ​​model is short, which can improve the transmission efficiency of the AI ​​model. It can be seen that the embodiment of the present application can take into account the protection of user privacy, the reduction of network resource occupancy and the improvement of the deployment efficiency of the AI ​​model.

[0159] In one possible design, the first request is a request sent to the first access network device, or in other words, the terminal requests the first AI model from the first access network device.

[0160] For example, the first request may be carried in a signaling radio bearer SRB. It is understood that after the SRB carrying the first request is transmitted to the first access network device, the first access network device may not forward the SRB to the core network (such as AMF, etc.).

[0161] Of course, SRB is only one possible implementation method, and the actual method in which the terminal sends the first request to the first access network device is not limited thereto.

[0162] Optionally, when the first request is a request sent to the first access network device, the first access network device may first query whether there is a first AI model in the cache of the first access network device based on the first request; if the first access network device has the first AI model cached, the first access network device sends the first AI model to the first terminal.

[0163] Further optionally, if the first access network device does not cache the first AI model, the first access network device queries the second access network device for the first AI model. If the second access network device caches the first AI model, the first access network device may receive the first AI model from the second access network device and send the first AI model to the terminal.

[0164] Further optionally, if the second access network device does not cache the first AI model, the first access network device can also obtain the first AI model from a core network network element (such as a data processing function network element) and then send the first AI model to the terminal.

[0165] Here are some possible examples:

[0166] Example 1: There is a first AI model in the cache of the first access network device. The first access network device determines the first AI model from its own cache and sends it to the terminal. The first AI model cached by the first access network device may be obtained from other access network devices and cached locally before receiving the first request (such as after receiving the request carrying the identifier of the first AI model sent by the terminal last time) (the acquisition method can refer to Example 2), or obtained from a core network element (such as a data processing function element) and cached locally (the acquisition method can refer to Example 3). The embodiment of the present application does not limit the source of the first AI model cached by the first access network device.

[0167] For the specific implementation process of this example, please refer to Figure 4 In the process shown, step S302 may specifically include the first access network device reading the first AI model from the local cache according to the identifier of the first AI model.

[0168] In Example 1, the terminal can obtain the AI ​​model from the first access network device without interacting with the core network, and the AI ​​model can be obtained quickly.

[0169] Example 2: There is no first AI model in the cache of the first access network device. The first access network device queries (or requests) the first AI model from an adjacent access network device (such as the second access network device), and the first access network device receives the first AI model from the second access network device.

[0170] It can be understood that the first access network device may have one or more adjacent access network devices.

[0171] When the first access network device has only one adjacent access network device (ie, the second access network device), the first access network device requests the first AI model from the adjacent access network device.

[0172] When the first access network device has multiple adjacent access network devices, the first access network device may request the first AI model from each adjacent access network device. After the first access network device receives the first AI model sent by any adjacent access network device, the first access network device may send information to other adjacent access network devices (such as the third access network device and the fourth access network device) to instruct to stop querying or transmitting the first AI model; or,

[0173] The first access network device stores model cache information of each adjacent access network device (for example, the first access network device locally stores a mapping relationship between the access network device identifier and the AI ​​model identifier). The first access network device determines the adjacent access network device (such as the second access network device) that caches the first AI model based on the model cache information of each adjacent access network device, and requests the first AI model from the adjacent access network device.

[0174] For example, see Figure 5 The process of the first access network device successfully acquiring the first AI model from the second access network device may include:

[0175] S302a: After receiving the first request, the first access network device determines that there is no first AI model in its cache.

[0176] S302b. The first access network device sends a model cache request to the second access network device (it can be understood that the model cache request can also be replaced by other names). The request carries the identifier of the first AI model, and the second access network device receives the model cache request.

[0177] S302c. The second access network device sends a first model cache response (it can be understood that the model cache response can also be replaced by other names) and a first AI model to the first access network device, and the first access network device receives the first model cache response and the first AI model.

[0178] It can be understood that the second access network device may first send a first model cache response (the response may indicate that the second access network device has cached the first AI model), and then send the first AI model; the second access network device may also send the first model cache response and the first AI model at the same time, for example, carrying the first AI model in the first model cache response; or, using the first AI model as the first model cache response, that is, only sending the first AI model, etc., which is not limited in the embodiments of the present application.

[0179] S302d. The first access network device caches the first AI model.

[0180] Afterwards, the first access network device can send the first AI model to the terminal.

[0181] It can be understood that since the data volume of the first AI model is large and may be transmitted in multiple times, the first access network device can cache the data of the first AI model while sending the cached data of the first AI model to the terminal.

[0182] Alternatively, the first access network device may not cache the first AI model, but only forward the received first AI model to the terminal.

[0183] It is understood that the present application does not limit the order between S304a-S304c and S302a-S302d. Figure 5 The order shown is only an example.

[0184] It can be understood that S302 may include S302a, S302b, S302c, and S302d.

[0185] In Example 2, the terminal can obtain the AI ​​model from the adjacent access network device of the first access network device without interacting with the core network, and the AI ​​model can be obtained quickly.

[0186] Example 3: There is no first AI model in the cache of the first access network device. The first access network device further queries the adjacent access network device (such as the second access network device) for the first AI model. Since the adjacent access network device does not have the first AI model either, the first access network device further obtains the first AI model from a core network element (such as a data processing function network element).

[0187] In one possible implementation, after the first access network device fails to obtain the first AI model from the second access network device, it may send information indicating the failure of the deployment of the first AI model to the terminal (hereinafter referred to as failure information for the sake of description), so that after receiving the failure information, the terminal sends a second request to the application function network element, so that the terminal requests the first AI model from the core network element (such as the data processing function network element). Alternatively, if the terminal does not receive the first AI model after waiting for more than a preset time (such as the first time or the second time described above), the terminal sends a second request to the application function network element, so that the terminal requests the first AI model from the core network element (such as the data processing function network element). Among them, the terminal sends the second request to the application function network element, which may be sending the second request to the application function network element via the first access network device, etc.

[0188] For example, see Figure 5 The process in which the first access network device fails to obtain the first AI model from the second access network device may include the following steps:

[0189] S302a: The first access network device determines that there is no first AI model in its cache.

[0190] S302b. The first access network device sends a model cache request to the second access network device, where the request carries an identifier of the first AI model, and the second access network device receives the model cache request.

[0191] S302e. The second access network device sends a second model cache response to the first access network device (the response may indicate that the second access network device does not cache the first AI model), and the first access network device receives the second model cache response.

[0192] Alternatively, the second access network device may not send the second model cache response to the first access network device. After waiting for a certain period of time, the first access network device does not receive any response (or the first AI model), and determines that the second access network device does not cache the first AI model.

[0193] S302f. The first access network device sends a failure message to the terminal. The terminal receives the failure message and determines that the deployment of the first AI model has failed.

[0194] Alternatively, the first access network device does not send a failure message, and the terminal does not receive the first AI model after a preset time period (such as the first time period or the second time period mentioned above), and determines that the deployment of the first AI model has failed.

[0195] It can be understood that S302 may include S302a, S302b, S302e, and S302f.

[0196] See also Figure 6After the terminal receives the failure information (i.e., S302e), or after the terminal waits for more than a preset time, the process of the terminal requesting the first AI model from the core network element may include the following steps:

[0197] S401. The terminal sends a second request to an application function network element via a first access network device, etc., and the application function network element receives the second request.

[0198] The second request carries the identifier of the first AI model, and the second request can be carried in a data radio bearer (DRB).

[0199] S402. The application function network element sends a third request to the control function network element based on the second request. The control function network element receives the third request, where the third request includes an identifier of the first AI model.

[0200] Optionally, the application function network element may directly forward the second request, that is, the third request is the second request; or, the application function network element may process the second request before forwarding it, that is, the third request is different from the second request, and this embodiment of the present application does not limit this.

[0201] Optionally, after receiving the third request, the control function network element may also send a third response to the application function network element, and the application function network element (via the first access network device, etc.) sends a third response to the terminal (the third response and the second response may be the same or different), and the second response and the third response may carry a duration, such as a third duration, which is used to indicate the duration that the terminal waits for the first AI model to start transmitting. The third duration may be the same or different from the first duration or the second duration mentioned above, and this embodiment of the application does not limit this.

[0202] After receiving the third request, the control function network element orchestrates the data storage function network element and the data processing function network element to send the first AI model to the first access network device, for example, S403 to S405.

[0203] S403: The control function network element determines a first data processing function network element and a first data storage function network element.

[0204] It can be understood that one or more data storage functional network elements and one or more data processing functional network elements can be deployed in the network. When the control functional network element arranges the data storage functional network element, it can select the data storage functional network element that stores the first AI model. When the control functional network element arranges the data processing functional network element, it can select the data processing functional network element according to the transmission quality (such as latency, bandwidth, or jitter, etc.). The selected data processing functional network element is the first data processing functional network element, and the selected data storage functional network element is the first data storage functional network element.

[0205] It can be understood that the number of data processing function network elements selected by the control function network element can also be multiple. When there are multiple data processing function network elements, the control function network element sends configuration information to each data processing function network element, so that each data processing function network element can receive the first AI model from the upstream node (the upstream node can be the first data storage function network element or another data processing function network element or other forwarding node) and send the first AI model to the downstream node (the downstream node can be the first access network device or another data processing function network element or other forwarding node). For example, the first data processing function network element receives the first AI model from the first data storage function network element, the first data processing function network element sends the first AI model to the second data storage function network element, and the second data processing function network element sends the first AI model to the first access network device. For ease of description, the embodiment of the present application takes one data processing function network element as an example.

[0206] S404: The control function network element sends first configuration information to the first data processing function network element, and the first data processing function network element receives the first configuration information.

[0207] In a possible implementation, the first configuration information includes an identifier of the first access network device.

[0208] It can be understood that when the terminal sends the second request, it will carry the address information of the terminal (such as IP address) in the packet header of the data packet containing the second request. When the application function network element forwards the second request, it will also forward the address information to the control function network element. The control function network element can query the identifier of the access network device corresponding to the terminal from the user plane function network element based on the address information of the terminal, and obtain the identifier of the first access network device, thereby carrying the identifier of the first access network device in the first configuration information.

[0209] In a possible implementation, the first configuration information may not include the identifier of the first access network device, and the first data processing function network element queries the identifier of the first access network device from the user plane function network element according to the address information of the terminal.

[0210] S405. The control function network element sends second configuration information to the first data storage function network element, and the first data storage function network element receives the second configuration information.

[0211] Among them, the second configuration information includes the identifier of the first data processing function network element and the identifier of the first AI model.

[0212] It can be understood that the embodiment of the present application does not limit the sequence of S404 and S405.

[0213] S406. After receiving the second configuration information, the first data storage functional network element sends the first AI model to the first data processing functional network element.

[0214] For example, the first AI model is determined from the AI ​​models stored in itself according to the identifier of the first AI model, and the first AI model is sent to the first data processing function network element according to the identifier of the first data processing function network element.

[0215] In a possible implementation, the first configuration information and the second configuration information may also include a transmission protocol. The transmission protocol is related to the first data pipe, which is a data pipe between the first data storage function network element and the first data processing function network element; or, the transmission protocol is related to the first data pipe identifier, which is used to indicate the data pipe between the first data storage function network element and the first data processing function network element. Accordingly, the first data storage function network element sends the first AI model based on the transmission protocol, and the first data processing function network element receives the first AI model based on the transmission protocol.

[0216] Among them, the transmission protocol includes but is not limited to one or more of quick user datagram protocol (UDP) network connections (quick UDP internet connections, QUIC), loss tolerant transmission protocol (loss tolerant transmission protocol, LTP), transmission control protocol (transmission control protocol, TCP), hypertext transfer protocol (hypertext transfer protocol, HTTP), etc.

[0217] In an embodiment of the present application, the transmission protocol between the data processing functional network element and the data storage functional network element is configurable, and the control functional network element can flexibly configure the transmission protocol according to the type of data to be transmitted between the data processing functional network element and the data storage functional network element. For example, considering that the data transmission volume of the AI ​​model is large, the QUIC or LTP protocol can be configured for the first data processing functional network element and the first data storage functional network element, so that the first data processing functional network element and the first data storage functional network element transmit the AI ​​model based on the QUIC or LTP protocol, thereby improving the transmission efficiency of the AI ​​model between the first data processing functional network element and the first data storage functional network element, and reducing or avoiding network congestion.

[0218] S407. After receiving the first configuration information and the first AI model, the first data processing function network element sends the first AI model to the first access network device according to the identifier of the first access network device, and the first access network device receives the first AI model.

[0219] S408. The first access network device caches the first AI model.

[0220] Afterwards, the first access network device may send the first AI model to the terminal (ie, execute S303).

[0221] It can be understood that S302 may also include S407 and S408.

[0222] In another possible implementation, after the first access network device fails to obtain the first AI model from the second access network device, the first access network device may not return failure information to the terminal, but directly request the first AI model from the core network, so that the first access network device obtains the first AI model from the core network network element (such as a data processing function network element).

[0223] For example, see Figure 7 , the process of the first access network device obtaining the first AI model from the core network may include the following steps:

[0224] S501. The first access network device sends a fourth request to the user plane function network element, where the fourth request includes an identifier of the first AI model, and the user plane function network element receives the fourth request.

[0225] In a specific implementation, the first access network device may send a fourth request to the user plane functional network element based on the data path between the first access network device and the user plane functional network element. Exemplarily, the first access network device may deploy a session proxy (SP) functional module, the SP functional module may trigger the establishment of a data path between the first access network device and the user plane functional network element, and the session management functional network element may configure session information for the user plane functional network element and the first access network device (i.e., there is no need to establish a DRB from the terminal to the first access network device), so that the first access network device can directly interact with the user plane functional network element.

[0226] S502. The user plane function network element sends a third request to the control function network element, and the control function network element receives the third request, where the third request includes an identifier of the first AI model.

[0227] The third request and the fourth request may be the same or different without limitation.

[0228] It can be understood that here the example is that the first access network device requests the first AI model from the control function network element through the user plane function network element. In actual applications, the first access network device can also request the first AI model from the control function network element through other network elements or directly request the first AI model from the control function network element.

[0229] S503: The control function network element determines a first data processing function network element and a first data storage function network element.

[0230] S504: The control function network element sends first configuration information to the first data processing function network element, and the first data processing function network element receives the first configuration information.

[0231] S505: The control function network element sends second configuration information to the first data storage function network element, and the first data storage function network element receives the second configuration information.

[0232] S506. After receiving the second configuration information, the first data storage functional network element sends the first AI model to the first data processing functional network element, and the first data processing functional network element receives the first AI model.

[0233] S507. After receiving the first configuration information and the first AI model, the first data processing function network element sends the first AI model to the first access network device according to the identifier of the first access network device, and the first access network device receives the first AI model.

[0234] S508. The first access network device caches the first AI model.

[0235] The specific implementation of the above S502 to S508 can refer to the above S402 to S408 and will not be described in detail.

[0236] Afterwards, the first access network device may send the first AI model to the terminal (ie, execute S303).

[0237] Figure 7 The scheme shown and Figure 6 The differences between the schemes shown are: Figure 6 The terminal sends a third request to the control function network element via the application function network element, Figure 7 The first access network device sends the third request to the control function network element through the user plane function network element.

[0238] In Example 3, the terminal first attempts to obtain the AI ​​model from the access network (such as the first access network device and its adjacent access network devices), and then obtains the AI ​​model from the core network if it fails. This can ensure that the terminal eventually obtains the AI ​​model, thereby improving the reliability of the solution.

[0239] Through the above design, the terminal can request the first AI model from the access network (such as the first access network device or the adjacent access network device of the first access network device). If the access network has the first AI model, there is no need to interact with the core network, which can minimize the occupation of network resources and improve the deployment efficiency of the AI ​​model; if the access network does not have the first AI model, the first AI model can also be obtained from the core network, which improves the reliability of the deployed AI model.

[0240] In another possible design, the first request is a request sent to a core network element (such as an application function network element or a control function network element, etc.), or the terminal requests the first AI model from the core network element. For example, the first request can be carried in a signaling radio bearer (SRB).

[0241] For example, see Figure 8 , the process of the terminal requesting the first AI model from the core network element is as follows:

[0242] S601. The terminal sends a first request to an application function network element through a first access network device. The application function network element receives the first request, where the first request includes an identifier of a first AI model.

[0243] Specifically, the terminal sends a first request to the first access network device, and the first access network device forwards the first request to the application function network element after receiving the first request. It can be understood that S601 includes S301.

[0244] S602. The application function network element sends a third request to the control function network element based on the first request. The control function network element receives the third request, where the third request includes an identifier of the first AI model.

[0245] Optionally, the application function network element may directly forward the first request, that is, the third request is the first request; or, the application function network element may process the first request before forwarding it, that is, the third request is different from the first request, and this embodiment of the present application does not limit this.

[0246] S603: The control function network element determines a first data processing function network element and a first data storage function network element.

[0247] S604: The control function network element sends first configuration information to the first data processing function network element, and the first data processing function network element receives the first configuration information.

[0248] S605: The control function network element sends second configuration information to the first data storage function network element, and the first data storage function network element receives the second configuration information.

[0249] S606. After receiving the second configuration information, the first data storage functional network element sends the first AI model to the first data processing functional network element, and the first data processing functional network element receives the first AI model.

[0250] S607. After receiving the first configuration information and the first AI model, the first data processing function network element sends the first AI model to the first access network device according to the identifier of the first access network device, and the first access network device receives the first AI model.

[0251] S608. The first access network device caches the first AI model.

[0252] It can be understood that S302 may include S602 to S608.

[0253] Afterwards, the first access network device may send the first AI model to the terminal (ie, execute S303).

[0254] The specific implementation of the above S602 to S608 can refer to the above S402 to S408, which will not be repeated here.

[0255] Figure 8 The scheme shown and Figure 6 The differences between the schemes shown are: Figure 6 After the terminal attempts to obtain the first AI model from the first access network device and its adjacent access network devices and fails, the terminal requests the first AI model from the core network element again; Figure 8 The terminal directly requests the first AI model from the core network element.

[0256] It can be understood that after the terminal obtains the first AI model, the next time the terminal needs to obtain the first AI model, it can directly initiate a request to the first access network device (refer to Figures 4 to 7 scheme shown).

[0257] Through the above design, the terminal can directly obtain the first AI model from the core network, and this process does not need to rely on the PDU session mechanism, which can take into account the protection of user privacy, reduce network resource usage and improve AI model transmission efficiency.

[0258] In one possible design, the control function network element may record model cache information, where the model cache information includes information about AI models cached on at least one access network device. Exemplarily, the control function network element stores a mapping relationship between an identifier of the access network device and an identifier of the AI ​​model, where the mapping relationship indicates which AI models are cached by each access network device in at least one access network device. For example, Table 1 is an example of model cache information recorded by the control function network element:

[0259] Table 1 Model cache information

[0260]

[0261] It can be seen from Table 1 that the first access network device caches the first AI model and the second AI model, and the second access network device caches the second AI model and the third AI model.

[0262] It can be understood that Table 1 is only an example, and the format of the actual model cache information is not limited thereto.

[0263] In one possible implementation, after the first data processing function network element sends the first AI model to the first access network device, the first data processing function network element sends the identifier of the first access network device and the identifier of the first AI model to the control function network element, and then the control function network element can update the model cache information of the first access network device based on the received identifier of the first access network device and the identifier of the first AI model.

[0264] In another possible implementation, after the first access network device (for example, from the second access network device or the first data processing function network element) receives and caches the first AI model, the first access network device (for example, through the user plane function network element) sends the identifier of the first access network device and the identifier of the first AI model to the control function network element, and the control function network element can update the model cache information of the first access network device based on the received identifier of the first access network device and the identifier of the first AI model.

[0265] It can be understood that if the control function network element originally does not have the model cache information of the first access network device, the control function network element may update the model cache information by creating the model cache information of the first access network device, such as creating a table entry corresponding to the first access network device; if the control function network element originally records the model cache information of the first access network device, the control function network element may update the model cache information by adding new information content based on the original model cache information of the first access network device, such as filling in the identifier of the first AI model in the table entry corresponding to the first access network device. In addition, the first access network device or the first data processing function network element may also report other information content (such as deletion indication, replacement indication, etc.), so that the control function network element may also update the model cache information by other operations such as replacement and deletion, which is not limited in the embodiments of the present application.

[0266] Through the above design, the control function network element can obtain information about the AI ​​models cached by each access network device to better control and manage the AI ​​models cached by each access network device.

[0267] In one possible design, the control function network element can also subscribe to the model update event from the first data storage function network element, so that when the AI ​​model stored on the data storage function network element changes, the control function network element can promptly learn of the change in the AI ​​model and update the AI ​​model cached in the access network device.

[0268] For example, see Fig. 9 , is an example of updating an AI model, including:

[0269] S701. The control function network element sends subscription information to the first data storage function network element, and the first data storage function network element receives the subscription information, where the subscription information includes an identifier of the first AI model.

[0270] S702. When the first AI model stored in the first data storage functional network element is updated, the first data storage functional network element sends notification information to the control functional network element according to the subscription information, and the control functional network element receives the notification information, where the notification information is used to indicate that the first AI model is updated.

[0271] S703. The control function network element determines which access network devices need to update the first AI model based on the model cache information stored in the control function network element.

[0272] For example, if only the first access network device stores the first AI model, then only the first access network device needs to update the first AI model.

[0273] S704: The control function network element selects a data processing function network element and a data storage function network element according to the determined access network device.

[0274] For example, a first data processing functional network element and a first data storage functional network element are determined.

[0275] S705. The control function network element sends third configuration information to the first data processing function network element, and the first data processing function network element receives the third configuration information.

[0276] S706. The control function network element sends fourth configuration information to the first data storage function network element, and the first data storage function network element receives the fourth configuration information.

[0277] It is understandable that S705 and S706 are not in any particular order.

[0278] The content of the third configuration information may refer to the first configuration information above, and the content of the fourth configuration information may refer to the second configuration information above.

[0279] S707: The first data storage functional network element sends update information to the first data processing functional network element, and the first data processing functional network element receives the update information.

[0280] The update information may be the changed data part in the first AI model, or may be all the data parts of the first AI model, which is not limited in the embodiment of the present application.

[0281] S708. The first data processing function network element sends update information to the first access network device, and the first access network device receives the update information.

[0282] S709. The first access network device updates the cached first AI model based on the update information.

[0283] Understandably, Fig. 9Taking the example of the control function network element subscribing to model update events for one AI model (i.e., the first AI model) at a time, in actual applications, the control function network element can also subscribe to model update events for multiple AI models at a time, such as subscribing to model update events for all models saved by the data storage function network element, which is not limited in the embodiments of the present application.

[0284] certainly, Fig. 9 The example of AISCF managing and controlling the update of AI models is given. In actual applications, other network elements can also manage or control the update of AI models, or DSF itself can manage or control the update of AI models. For example, DSF can save model cache information, and when a model is updated, the update information is directly pushed to the access network device through DPF.

[0285] Through the above design, when the AI ​​model saved by the data storage functional network element is updated, the AI ​​model cached by the access network device can be updated in time, thereby ensuring the timeliness of the update of the AI ​​model and improving the user experience.

[0286] It can be understood that the above-mentioned embodiments of the present application can be implemented separately or in combination with each other, and the embodiments of the present application are not limited.

[0287] The method provided by the embodiment of the present application is introduced above in combination with the accompanying drawings, and the device provided by the embodiment of the present application is introduced below in combination with the accompanying drawings.

[0288] Based on the same technical concept, the embodiment of the present application provides a communication device, which includes a module / unit / means for executing the method executed by the device in the above method embodiment. The module / unit / means can be implemented by software, or by hardware, or the corresponding software can be implemented by hardware.

[0289] For example, see Fig.10 , is a schematic diagram of a communication device 1000 provided in an embodiment of the present application, wherein the device 1000 includes a transceiver module 101 and a processing module 102.

[0290] When the device 1000 is a first access network device or is located in a first access network device, the functions of each module of the device 1000 are as follows:

[0291] The transceiver module 101 is configured to receive a first request from a terminal, where the first request includes an identifier of a first AI model;

[0292] The processing module 102 is configured to determine a first AI model; wherein the first AI model is one of at least one AI model cached by the first access network device or the second access network device, and the second access network device is adjacent to the first access network device;

[0293] The transceiver module 101 is also used to send the first AI model to the terminal.

[0294] When the device 1000 is a terminal or is located at a terminal, the functions of the modules of the device 1000 are as follows:

[0295] The transceiver module 101 is used to send a first request to a first access network device, wherein the first request includes an identifier of a first AI model; and receive the first AI model from the first access network device; wherein the first AI model is one of at least one AI model cached by the first access network device or the second access network device, and the second access network device is adjacent to the first access network device.

[0296] When the device 1000 is a first data processing function network element or is located in the first data processing function network element, the functions of the modules of the device 1000 are as follows:

[0297] The transceiver module 101 is configured to receive first configuration information from a control function network element, wherein the first configuration information includes an identifier of a first access network device; and receive a first AI model from a first data storage function network element;

[0298] The processing module 102 is used to send the first AI model to the first access network device according to the identifier of the first access network device.

[0299] When the device 1000 is a first data storage functional network element or is located in a first data storage functional network element, the functions of each module of the device 1000 are as follows:

[0300] The transceiver module 101 is used to receive second configuration information from the control function network element, where the second configuration information includes an identifier of the first data processing function network element and an identifier of the first AI model; and send the first AI model to the first data processing function network element.

[0301] When the device 1000 is a control function network element or is located in a control function network element, the functions of each module of the device 1000 are as follows:

[0302] The transceiver module 101 is configured to receive a third request from an application function network element or a user plane network element, where the third request includes an identifier of the first AI model;

[0303] The processing module 102 is used to determine a first data storage function network element and a first data processing function network element according to the third request;

[0304] The transceiver module 101 is also used to send first configuration information to the first data processing function network element and to send second configuration information to the first data storage function network element; wherein the first configuration information includes the identifier of the first access network device, and the second configuration information includes the identifier of the first data processing function network element and the identifier of the first AI model.

[0305] In specific implementation, the above-mentioned device 1000 can have a variety of product forms. Several possible product forms are introduced below.

[0306] like Fig.11 As shown, an embodiment of the present application provides a communication device 1100, which includes a processor 1110 and an interface circuit 1120. The interface circuit 1120 is used to receive signals from other communication devices outside the communication device and transmit them to the processor 1110, or send signals from the processor 1110 to other communication devices outside the communication device. The processor 1110 is used to implement the method performed by any device or network element in the above method embodiment through a logic circuit or execution instruction.

[0307] The processor 1110 and the interface circuit 1120 are coupled to each other. It is understood that the interface circuit 1120 may be a transceiver or an input / output interface. Optionally, the communication device 1100 may further include a memory 1130 for storing instructions executed by the processor 1110 or storing input data required by the processor 1110 to execute instructions or storing data generated after the processor 1110 executes instructions.

[0308] When the above communication device is a chip applied to a terminal, the chip implements the functions of the terminal in the above method embodiment. The chip receives information from other modules in the terminal (such as a radio frequency module or an antenna), and the information is sent by the base station to the terminal; or the terminal chip sends information to other modules in the terminal (such as a radio frequency module or an antenna), and the information is sent by the terminal to the base station.

[0309] When the above-mentioned communication device is a module applied to an access network device, the access network device module implements the functions of the access network device in the above-mentioned method embodiment. The access network device module receives information from other modules in the access network device (such as a radio frequency module or an antenna), and the information is sent by the terminal to the access network device; or, the access network device module sends information to other modules in the access network device (such as a radio frequency module or an antenna), and the information is sent by the access network device to the terminal. The access network device module here can be a baseband chip of the access network device, or it can be a DU or other module. The DU here can be a DU under the open radio access network (O-RAN) architecture.

[0310] It should be understood that the processor mentioned in the embodiments of the present application can be implemented by hardware or by software. When implemented by hardware, the processor can be a logic circuit, an integrated circuit, etc. When implemented by software, the processor can be a general-purpose processor implemented by reading software code stored in a memory.

[0311] Exemplarily, the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0312] It should be understood that the memory mentioned in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct memory bus random access memory (DirectRambus RAM, DR RAM).

[0313] It should be noted that when the processor is a general-purpose processor, DSP, ASIC, FPGA or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, the memory (storage module) can be integrated into the processor.

[0314] It should be noted that the memory described herein is intended to include but is not limited to these and any other suitable types of memory. Based on the same technical concept, an embodiment of the present application also provides a computer-readable storage medium, in which a computer program or instruction is stored. When the computer program or instruction is executed by a processor, the method performed by any device or network element in the above method embodiment is implemented.

[0315] Based on the same technical concept, an embodiment of the present application also provides a computer program product, including a computer program or instructions. When the computer program or instructions are executed by a processor, the method executed by any device or network element in the above method embodiment is implemented.

[0316] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0317] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0318] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0319] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

Claims

1. An artificial intelligence (AI) model deployment method, characterized in that: include: The first access network device receives a first request from the terminal, where the first request includes an identifier of the first AI model; The first access network device determines the first AI model; wherein the first AI model is one of at least one AI model cached by the first access network device or a second access network device, and the second access network device is adjacent to the first access network device; The first access network device sends the first AI model to the terminal.

2. The method according to claim 1, characterized in that The method further comprises: The first access network device obtains the first AI model from the second access network device, and saves the first AI model to a cache of the first access network device; or, The first access network device receives the first AI model from the first data processing function network element, and saves the first AI model to a cache of the first access network device.

3. The method according to claim 1 or 2, characterized in that The first request is carried in a signaling radio bearer.

4. The method according to any one of claims 1 to 3, characterized in that: The first access network device determines a first AI model, including: The first access network device queries, based on the first request, whether the first AI model is in a cache of the first access network device.

5. The method according to claim 4, characterized in that The first access network device does not have the first AI model in its cache; and the method further includes: The first access network device queries the second access network device for the first AI model; The first access network device receives the first AI model sent by the second access network device.

6. The method according to claim 5, characterized in that The method further comprises: The first access network device saves the first AI model to a cache of the first access network device.

7. The method according to claim 1, characterized in that The first request is carried in a data radio bearer; The method further comprises: The first access network device forwards the first request to an application function network element; The first access network device receives the first AI model from the first data processing function network element, and saves the first AI model to a cache of the first access network device.

8. The method according to any one of claims 1 to 7, characterized in that: After the first access network device receives the first request from the terminal, the method further includes: The first access network device sends a first duration to the terminal, where the first duration is the duration that the terminal waits for the first AI model to start transmitting.

9. The method according to any one of claims 1 to 11, characterized in that: The method further comprises: The first access network device receives update information from a first data processing function network element, and the update information is used to update the first AI model.

10. An artificial intelligence (AI) model deployment method, characterized in that: include: The terminal sends a first request to the first access network device, where the first request includes an identifier of the first AI model; The terminal receives the first AI model from the first access network device; wherein the first AI model is one of at least one AI model cached by the first access network device or a second access network device, and the second access network device is adjacent to the first access network device.

11. The method according to claim 10, characterized in that The first request is carried in a signaling radio bearer.

12. The method according to claim 10, characterized in that The first request is carried in a data radio bearer.

13. The method according to claim 11, characterized in that After the terminal sends the first request to the first access network device and before the terminal receives the first AI model from the first access network device, the method further includes: If the terminal waits for more than a preset time, or the terminal receives information from the first access network device indicating that the deployment of the first AI model has failed, the terminal sends a second request to the application function network element through the first access network device, where the second request includes an identifier of the first AI model, and the second request is carried in a data wireless bearer.

14. The method according to claim 13, characterized in that The preset duration is a first duration; After the terminal sends the first request to the first access network device and before the terminal receives the first AI model from the first access network device, the method further includes: The terminal receives a first duration from the first access network device; The terminal waits for the first AI model to start transmitting within the first time period.

15. The method according to claim 13, characterized in that The preset duration is a second duration, and the second duration is the duration of a timer configured by the terminal.

16. An artificial intelligence (AI) model deployment method, characterized in that: include: The first data processing function network element receives first configuration information from the control function network element, where the first configuration information includes an identifier of the first access network device; The first data processing functional network element receives a first AI model from a first data storage functional network element; The first data processing function network element sends the first AI model to the first access network device according to the identifier of the first access network device.

17. The method according to claim 16, characterized in that The first configuration information also includes a transmission protocol, where the transmission protocol is related to a first data pipeline, where the first data pipeline is a data pipeline between the first data storage function network element and the first data processing function network element; The first data processing functional network element receives the first AI model from the first data storage functional network element, including: The first data processing functional network element receives the first AI model from the first data storage functional network element based on the transmission protocol.

18. The method according to claim 16 or 17, characterized in that After the first data processing function network element sends the first AI model to the first access network device, the method further includes: The first data processing function network element sends the identifier of the first access network device and the identifier of the first AI model to the control function network element.

19. The method according to any one of claims 16 to 18, characterized in that: The method further comprises: The first data processing functional network element receives update information from the first data storage functional network element, where the update information is used to update the first AI model; The first data processing function network element sends the update information to the first access network device.

20. An artificial intelligence (AI) model deployment method, characterized in that: include: The first data storage function network element receives second configuration information from the control function network element, where the second configuration information includes an identifier of the first data processing function network element and an identifier of the first AI model; The first data storage function network element sends the first AI model to the first data processing function network element.

21. The method according to claim 20, characterized in that The second configuration information also includes a transmission protocol, where the transmission protocol is related to a first data pipeline, where the first data pipeline is a data pipeline between the first data storage function network element and the first data processing function network element; The first data storage function network element sends the first AI model to the first data processing function network element, including: The first data storage function network element sends the first AI model to the first data processing function network element based on the transmission protocol.

22. The method according to claim 20 or 21, characterized in that The method further comprises: The first data storage function network element receives subscription information from the control function network element, where the subscription information includes an identifier of the first AI model; When the first AI model stored in the first data storage functional network element is updated, the first data storage functional network element sends notification information to the control functional network element according to the subscription information, where the notification information is used to indicate that the first AI model is updated.

23. The method of claim 22, wherein: The method further comprises: The first data storage function network element sends update information to the first data processing function network element, where the update information is used to update the first AI model.

24. An artificial intelligence (AI) model deployment method, characterized in that: include: The control function network element receives a third request from the application function network element or the user plane network element, where the third request includes an identifier of the first AI model; The control function network element determines a first data storage function network element and a first data processing function network element according to the third request; The control function network element sends first configuration information to the first data processing function network element and sends second configuration information to the first data storage function network element; wherein the first configuration information includes an identifier of the first access network device, and the second configuration information includes an identifier of the first data processing function network element and an identifier of the first AI model.

25. The method of claim 24, wherein: The first configuration information and the second configuration information also include a transmission protocol, and the transmission protocol is related to a first data pipeline, and the first data pipeline is a data pipeline between the first data storage function network element and the first data processing function network element.

26. The method according to claim 24 or 25, characterized in that The data packet containing the second request carries the address information of the terminal; and the method further includes: The control function network element queries the user plane function network element for the identifier of the access network device corresponding to the terminal based on the address information of the terminal, and obtains the identifier of the first access network device.

27. The method according to any one of claims 24 to 26, characterized in that: The method further comprises: The control function network element receives the identifier of the first access network device and the identifier of the first AI model from the first data processing function network element; The control function network element updates the model cache information according to the identifier of the first access network device and the identifier of the first AI model, where the model cache information includes information of the AI ​​model cached on at least one access network device.

28. The method of claim 27, wherein: The method further comprises: The control function network element sends subscription information to the first data storage function network element, where the subscription information includes an identifier of the first AI model; The control function network element receives notification information from the first data storage function network element, where the notification information is used to indicate that the first AI model is updated; The control function network element configures the first data storage function network element and the first data processing function network element according to the model cache information to send update information to the first access network device, and the update information is used to update the first AI model.

29. An artificial intelligence (AI) model deployment method, characterized in that: include: The control function network element receives a third request from the application function network element or the user plane network element, where the third request includes an identifier of the first AI model; The control function network element determines a first data storage function network element and a first data processing function network element according to the third request; The control function network element sends first configuration information to the first data processing function network element and sends second configuration information to the first data storage function network element; wherein the first configuration information includes an identifier of the first access network device, and the second configuration information includes an identifier of the first data processing function network element and an identifier of the first AI model; The first data storage function network element receives the second configuration information from the control function network element, and sends the first AI model to the first data processing function network element; The first data processing function network element receives the first configuration information from the control function network element, and receives the first AI model from the first data storage function network element; and sends the first AI model to the first access network device according to the identifier of the first access network device.

30. A communication device, characterized in that: The device includes a module for executing the method as described in any one of claims 1 to 9, or includes a module for executing the method as described in any one of claims 10 to 15, or includes a module for executing the method as described in any one of claims 16 to 19, or includes a module for executing the method as described in any one of claims 20 to 23, or includes a module for executing the method as described in any one of claims 24 to 28.

31. A communication device, characterized in that: It includes a processor and an interface circuit, wherein the interface circuit is used to receive signals from other communication devices outside the communication device and transmit them to the processor, or send signals from the processor to other communication devices outside the communication device, and the processor is used to implement the method as described in any one of claims 1 to 9, or implement the method as described in any one of claims 10 to 15, or implement the method as described in any one of claims 16 to 19, or implement the method as described in any one of claims 20 to 23, or implement the method as described in any one of claims 24 to 28 through a logic circuit or execution instruction.

32. A computer-readable storage medium, characterized in that: The storage medium stores a computer program or instruction. When the computer program or instruction is executed by the processor, the method as described in any one of claims 1 to 9 is implemented, or the method as described in any one of claims 10 to 15 is implemented, or the method as described in any one of claims 16 to 19 is implemented, or the method as described in any one of claims 20 to 23 is implemented, or the method as described in any one of claims 24 to 28 is implemented.

33. A computer program product, characterized in that Contains a computer program or an instruction, which, when executed by a processor, enables the method according to any one of claims 1 to 9 to be implemented, or enables the method according to any one of claims 10 to 15 to be implemented, or enables the method according to any one of claims 16 to 19 to be implemented, or enables the method according to any one of claims 20 to 23 to be implemented, or enables the method according to any one of claims 24 to 28 to be implemented.

Images