Data processing method, device and network equipment

Abstract

The present application provides a data processing method, device and network equipment. Since the base station of the present application has AI service capability, after the first base station that is determined to be able to provide the requested target AI native service to the user terminal receives a PDU session request message, it will determine the user terminal's request to establish a PDU session to achieve the target AI native service based on the PDU session attribute identifier carried by it, configure the first base station to establish a target DRB for the PDU session with the user terminal, and specifically use it to transmit the corresponding PDU session data. The first base station responds to the PDU session data to achieve the target AI native service without having to parse it through the application layer and then divert it, thereby improving the execution efficiency of the target AI native service and greatly reducing the edge AI application latency. There is no need to additionally deploy dedicated UPF and MEC computing power devices, thereby greatly improving the computing power utilization of the base station, and no need for external transmission, thereby reducing the transmission cost and improving the data processing security.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a data processing method, device and network equipment. Background Art

[0002] With the integrated application and development of 5G (5th Generation Mobile Communication Technology) mobile communication technology and artificial intelligence (AI) technology, it has been widely used in application scenarios such as the Internet of Things (IoT), edge computing, smart manufacturing, self-driving cars, telemedicine, and augmented reality / virtual reality, bringing far-reaching impacts to all walks of life.

[0003] When AI services need to be implemented in 5G networks, a logical data transmission channel needs to be established between the user equipment (UE) and the data network (DN), such as Figure 1 The PDU (Packet Data Unit) session shown in the figure is used to meet the communication requirements between the AI ​​application client in the UE and the AI ​​server in the DN network.

[0004] However, in actual communication, Figure 1 As shown in the figure, the PDU session established by the 5G core network (5G Core, 5GC) is composed of two sub-logical channel links: DRB (Data Radio Bearer) between the UE and the base station, and N3 Tunnel between the base station and the UPF (User Plane Function) of the 5G core network. As a result, the communication between the AI ​​application client in the UE and the AI ​​server in the DN network is limited by the network performance parameters such as the latency and bandwidth of the N3 Tunnel and the link between the UPF and the DN network, which reduces the performance of edge AI applications. Summary of the invention

[0005] In view of the above problems, this application provides the following solutions:

[0006] A first aspect of the present application provides a data processing method, the data processing method comprising:

[0007] The first base station receives a packet data unit PDU session request message; the first base station is a base station that can provide the requested target artificial intelligence AI native service for the user terminal, and the PDU session request message carries a PDU session attribute identifier;

[0008] Determining, based on the PDU session attribute identifier, that the user terminal requests to establish a PDU session to implement a target AI native service, configuring a target data radio bearer DRB for establishing a PDU session between the first base station and the user terminal; the target DRB supports the implementation of the target AI native service;

[0009] Receiving the PDU session data transmitted by the target DRB from the user terminal;

[0010] In response to the PDU session data, the target AI native service is implemented.

[0011] In a possible implementation, the implementing the target AI native service in response to the PDU session data includes any one of the following:

[0012] Sending the PDU session data to an AI server deployed by the first base station, so that the AI ​​server completes the target AI native service;

[0013] Determine at least one second base station that supports the target AI native service, offload the PDU session data to AI servers deployed by the first base station and the at least one second base station, and complete the target AI native service through collaboration of the AI ​​servers;

[0014] Determine at least one second base station that supports the target AI native service, offload the PDU session data to AI servers deployed by each of the at least one second base station, and complete the target AI native service through collaboration of each of the AI ​​servers;

[0015] Among them, the second base station and the first base station belong to a target base station cluster that can provide the target AI native service, the AI ​​servers in the target base station cluster communicate through a data network, and the service addresses of the base stations in the target base station cluster for the target AI native service are the same.

[0016] In a possible implementation, implementing the target AI native service in response to the PDU session data further includes:

[0017] Determine idle service resources currently owned by the AI ​​server deployed by the first base station, and expected service resources required to complete the target AI native service;

[0018] Based on the idle service resources and the expected service resources, determining whether the target base station cluster needs to collaborate to complete the target AI native service;

[0019] If not, executing the step of sending the PDU session data to the AI ​​server deployed by the first base station, and the AI ​​server completes the target AI native service;

[0020] If necessary, execute the step of determining at least one second base station that supports the target AI native service.

[0021] A second aspect of the present application provides a data processing method, the data processing method comprising:

[0022] Receiving a packet data unit (PDU) session establishment request message sent by a user terminal; the PDU session establishment request message carries a service identifier of a target artificial intelligence (AI) native service requested by the user terminal;

[0023] Determine the service resource demand indicated by the service identifier and the service resources currently possessed by the first base station accessed by the user terminal;

[0024] If the service resources currently possessed by the first base station match the service resource requirements, send a PDU session request message to the first base station; the PDU session request message carries a PDU session attribute identifier, and the PDU session attribute identifier is the service identifier, or an address identifier obtained by processing based on the attribute information in the PDU session establishment request message;

[0025] Receive a PDU session reply message fed back by the first base station; the PDU session reply message is sent by the first base station after successfully configuring the target data radio bearer DRB for establishing a PDU session between the first base station and the user terminal based on the PDU session attribute identifier; the target DRB is used to transmit the PDU session data sent by the user terminal for the target AI native service.

[0026] A third aspect of the present application provides a data processing device, wherein the data processing device is deployed in a base station, and the base station is configured as a first base station capable of providing a requested target artificial intelligence AI native service for a user terminal, and the data processing device includes:

[0027] The first receiving module is used to receive a packet data unit PDU session request message; the PDU session request message carries a PDU session attribute identifier;

[0028] A first configuration module is used to determine, based on the PDU session attribute identifier, that the user terminal requests to establish a PDU session to implement a target AI native service, and configure a target data radio bearer DRB for establishing a PDU session between the first base station and the user terminal; the target DRB supports the implementation of the target AI native service;

[0029] A second receiving module, configured to receive the PDU session data transmitted by the target DRB from the user terminal;

[0030] An execution module is used to implement the target AI native service in response to the PDU session data.

[0031] A fourth aspect of the present application provides a data processing device, which is deployed in a core network, and includes:

[0032] A first receiving module is configured to receive a packet data unit (PDU) session establishment request message sent by a user terminal; the PDU session establishment request message carries a service identifier of a target artificial intelligence (AI) native service requested by the user terminal;

[0033] A first determining module, configured to determine a service resource requirement indicated by the service identifier and a service resource currently possessed by the first base station accessed by the user terminal;

[0034] A first sending module is configured to send a PDU session request message to the first base station if the service resources currently possessed by the first base station match the service resource requirements; the PDU session request message carries a PDU session attribute identifier, and the PDU session attribute identifier is the service identifier, or an address identifier obtained by processing based on the attribute information in the PDU session establishment request message;

[0035] The second receiving module is used to receive the PDU session reply message fed back by the first base station; the PDU session reply message is sent by the first base station after successfully configuring the target data radio bearer DRB for establishing a PDU session between the first base station and the user terminal based on the PDU session attribute identifier; the target DRB is used to transmit the PDU session data sent by the user terminal for the target AI native service.

[0036] A fifth aspect of the present application provides a network device, the network device comprising: at least one communication element, at least one memory and at least one processor, wherein:

[0037] The communication element is used to realize the communication connection between the network device and the user terminal;

[0038] The processor is used to load and execute the computer program stored in the memory to implement the various steps of the data processing method as described above.

[0039] It can be seen that in the data processing method proposed in this application, since this application deploys the AI ​​server in the base station, the base station has the AI ​​service capability. In this way, after the first base station that can provide the requested target AI native service to the user terminal receives the PDU session request message, it will determine the user terminal's request to establish a PDU session to achieve the target AI native service based on the PDU session attribute identifier carried by it, and configure the first base station to establish a target data radio bearer DRB with the user terminal for the PDU session, that is, to achieve the target AI native service dedicated DRB. After the first base station receives the PDU session data transmitted from the user terminal by the target DRB, it directly responds to the PDU session data to achieve the target AI native service, without having to parse through the application layer and then divert, which improves the execution efficiency of the target AI native service and greatly reduces the edge AI application latency. And there is no need to deploy additional dedicated UPF and MEC computing power devices, and the target AI native service is directly achieved on the base station side, which greatly improves the computing power utilization of the base station, is not subject to external network interference, meets the user data localization processing requirements, and improves data processing security. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and the originals and elements are not necessarily drawn to scale.

[0041] Figure 1 This is a schematic diagram of the data transmission channel structure of a traditional PDU session;

[0042] Figure 2 A schematic diagram of a data transmission channel structure of a PDU session applicable to the data processing method proposed in this application;

[0043] Figure 3 A flowchart of a data processing method proposed in Example 1 of the present application;

[0044] Figure 4 A flowchart of a data processing method proposed in Embodiment 2 of the present application;

[0045] Figure 5 A flowchart of a data processing method proposed in Embodiment 3 of the present application;

[0046] Figure 6 A schematic diagram of a data transmission channel structure of another PDU session applicable to the data processing method proposed in this application;

[0047] Figure 7 A flowchart of a data processing method proposed in Embodiment 4 of the present application;

[0048] Figure 8 A schematic diagram of the structure of a base station applicable to the data processing method proposed in this application;

[0049] Fig. 9 A flowchart of a data processing method proposed in Embodiment 5 of the present application;

[0050] Fig.10 A signaling flow diagram of a data processing method proposed in Embodiment 6 of the present application;

[0051] Fig.11 A schematic diagram of the structure of a data processing device proposed in Embodiment 1 of the present application;

[0052] Fig.12 A schematic diagram of the structure of a data processing device proposed in Embodiment 2 of the present application;

[0053] Fig.13 The figure is a schematic diagram of the hardware structure of a network device suitable for the data processing method proposed in this application. DETAILED DESCRIPTION

[0054] The embodiments of the present application are described below in conjunction with the drawings in the embodiments of the present application. The terms used in the implementation mode of the present application are only used to explain the specific embodiments of the present application and are not intended to limit the present application. The embodiments of the present application will be described below in conjunction with the drawings. It is known to those of ordinary skill in the art that with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0055] Moreover, the terms "first", "second" etc. in the full text description content and the accompanying drawings of the present application are used to distinguish similar objects, and need not be used to describe a specific order or sequential order. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, and this is only to describe the distinction mode adopted by the objects of the same attributes when describing in the embodiments of the present application. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, so that the process, method, system, product or equipment comprising a series of units need not be limited to those units, but may include other units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0056] In order to solve the technical problems described in the background technology section, it is necessary to deploy AI applications as close to base stations as physically possible, reduce the data transmission distance between AI application terminals and AI application servers, and thus reduce the number of network element hops generated during the communication between the two. Based on this, in order to reduce latency and increase data processing speed, it is proposed to deploy AI applications on edge devices that are closer to the data source. In one solution, the following can be used: Figure 1In the process of deploying UPF (User Plane Function) network elements near the base station, and deploying MEC (Multi-access Edge Computing) servers at the same time, or even deploying the base station and UPF network elements on the same server. However, this requires the deployment of UPF functions on the base station to analyze user data in real time, which increases the message cache time and analysis and calculation time, and significantly increases the transmission delay from the base station to the application server. In addition, since a large amount of computing resources need to be allocated, this is a huge challenge for base stations with limited computing power, and they can only serve this base station.

[0057] Another solution is to add dedicated data diversion hardware, and generate corresponding data diversion strategies by monitoring the control protocol between the base station and the core network to achieve the diversion of data to be transmitted on the user side. However, this method of configuring UPF dedicated computing resources or independent computing cards has high purchase costs, and requires the reservation of relevant resources on the base station in advance, and the equipment utilization rate is very inefficient. Because it runs independently outside the 5G network, user data processing is not controlled by the 5G network, and it is difficult to deploy and implement commercially as an application service. If it is necessary to support UE (User Equipment) mobile scenarios, resources need to be reserved in the wireless range, which is even more inefficient.

[0058] In order to further improve the above technical problems, this application proposes a data processing method, which deploys artificial intelligence AI native services in the base station to use the idle network capacity of the base station itself to provide AI native services. In this way, when the UE initiates an AI native service request, it will change Figure 1 The structure of the 5G network PDU (Packet Data Unit) session of the 3GPP standard shown in FIG. 1 is transformed into the following Figure 2 The PDU session shown only includes DRB (Data Radio Bearer) but not N3 Tunnel, so that after the base station receives the packet data unit PUD session request message, it configures the DRB for the AI ​​native service requested by the UE, and the base station receives the PDU session data for the AI ​​native service sent by the UE from the DRB, and responds to the PDU session through the network idle capacity of the base station to implement the AI ​​native service requested by the UE. It can be seen that the AI ​​native service is implemented directly in the base station without external transmission, and the processing process does not need to be resolved, which can protect data privacy and avoid the transmission of a large amount of data generated during the implementation of the AI ​​native service in the 5G network, saving transmission costs. It is no longer limited by network performance parameters and improves the execution time of the AI ​​native service. The data processing method of the embodiment of the present application is described in detail below with reference to the accompanying drawings.

[0059] Reference Figure 3, which is a flow chart of a data processing method proposed in the first embodiment of the present application. The method can be applied to a base station where an AI server is deployed, that is, a base station that can provide AI native services, referring to Figure 2 The data transmission channel structure diagram of the PDU session is shown in FIG. Figure 3 As shown, the data processing method proposed in this embodiment may include but is not limited to:

[0060] Step S31, the first base station receives a PDU session request message; the first base station is a base station that can provide the requested target AI native service for the user terminal, and the PDU session request message carries a PDU session attribute identifier;

[0061] In an embodiment of the present application, in a scenario where a user terminal UE requests a certain AI native service (such as an intelligent and personalized service provided by AI technology based on a cloud native architecture, which can be recorded as a target AI native service), such as application scenarios such as intelligent content creation, intelligent office, and intelligent customer service implemented based on an AI model, a corresponding PDU session establishment request message can be initiated, so that the core network of the wireless communication network analyzes it and determines that there is a first base station in the wireless communication network that can provide the requested target AI native service for the UE, and a decision can be made to establish such a session. Figure 2 The PDU session structure shown sends a corresponding PDU session request message to the first base station, ensuring that the PDU session request message carries a PDU session attribute identifier for the target PDU native service.

[0062] Among them, the communication process between the UE and the core network can be implemented according to the 3rd Generation Partnership Project 3GPP standard, and the embodiments of this application will not be described in detail here. The above-mentioned PDU session attribute identifier can be used to indicate that the UE requests to establish a PDU session to implement the AI ​​native service. This application does not limit its content. It is necessary to define the PDU session attribute identifiers corresponding to various AI native services in the core network of the wireless communication network in advance to ensure that it is distinguished from the identifiers of other non-base station AI native services.

[0063] Optionally, the above-mentioned PDU session attribute identifier of the present application can be one or more combinations of DNN (Data Network Naming, data network naming / identification, such as) and S-NSSAI (Single Network Slice Selection Assistance Information, single network slice identifier), etc. The DNN can be set to: "DNN= AI-Native"; the S-NSSAI can use the custom ID number allowed in the current communication network standard (such as 3GPP standard), such as "SST=100, SD=1", etc., to distinguish the DNN and S-NSSAI of non-base station AI native services. Of course, the above-mentioned PDU session attribute identifier can also be an identifier obtained by specializing some IEs in the PDU session to distinguish ordinary DU session services, such as setting the TNL address to a loopback address, etc. This application does not limit the content of the above-mentioned PDU session attribute identifier and can be flexibly configured according to actual conditions.

[0064] It should be noted that the PDU session request message received by the first base station, in addition to the PDU session attribute identifier described above, may also include other content, such as the terminal identifier of the UE that initiates the target AI native service, etc., which can be determined according to the corresponding communication standard requirements, and this application will not elaborate on this. In addition, in order to ensure that the core network can accurately detect whether the base station accessed by the UE can provide the target AI native service, each base station can report its own service resources to the core network in advance. After the service resources of the base station itself change, the updated service resources will be reported to the core network in a timely manner, so that the core network can accurately know the current AI service capabilities of each base station.

[0065] Of course, for each base station belonging to the same base station cluster, their respective service resources (which may include pre-configured and dynamically updated service resources, etc.) can be published to the data network of the base station cluster, so that each base station can know the service resources currently possessed by other base stations. In the process of each base station responding to the AI ​​native service request initiated by its own access UE, when multiple base stations need to collaborate to implement the AI ​​native service, they can accurately identify other base stations that can participate in the implementation of the AI ​​native service, thereby improving the execution efficiency and reliability of the AI ​​native service.

[0066] Step S32, determining that the user terminal requests to establish a PDU session to implement the target AI native service based on the PDU session attribute identifier, and configuring the first base station to establish a target DRB for the PDU session with the user terminal;

[0067] Combined with the above description of the PDU session attribute identifier, after the first base station parses the PDU session request message to obtain the PDU session attribute identifier, it can determine that the user terminal requests to establish a PDU session to implement the target AI native service through the pre-defined meaning of the PDU session attribute identifier, that is, the user terminal requests to establish a PDU session for implementing the target AI native service. According to the above description of the technical concept of the method provided in this application, the first base station will decide to establish a DRB dedicated to the target AI native service, that is, configure the target DRB in the PDU session established between the first base station and the UE, so that the target DRB supports the implementation of the target AI native service. This application does not limit the implementation method of how the first base station configures the target DRB, and it can be implemented in combination with the PDU session establishment method under the wireless communication network standard. The implementation process is not described in detail in this embodiment.

[0068] In actual applications, after the first base station configures the target DRB, it can inform the corresponding user terminal UE that the PDU session establishment request requested by the UE has been accepted. In this process, the first base station can carry the DRB information of the target DRB through the RRCReconfiguration (Radio Resource Control Reconfiguration, ref. 3GPP TS38.331) message in the RRC (Radio Resource Control) protocol to complete the creation of the target DRB, and at the same time notify the UE that the PDU session establishment request for the target AI native service requested to be implemented has been accepted. This application does not elaborate on how to complete the creation process of the target DRB between the first base station and the UE through RRCReconfiguration. It should be understood that in wireless communication networks applicable to non-3GPP standards, the configuration of the target DRB can be adaptively adjusted according to the communication principles of the corresponding network protocol.

[0069] Step S33, receiving the PDU session data transmitted by the target DRB from the user terminal;

[0070] Step S34, in response to the PDU session data, implement the target AI native service.

[0071] After the first base station configures the target DRB, it can feedback the DRB configuration result, that is, the response result to the PDU session request message, to the user terminal and core network and other equipment in accordance with the wireless communication network standard, and notify the UE that the PDU session establishment request requested by the UE to implement the target AI native service has been accepted. The UE can send user data through the PDU session, that is, the configured target DRB, such as the PDU session data for implementing the target AI native service. The core network can also monitor and manage the PDU session established this time based on the response result feedback from the first base station. The implementation method is not described in detail in this application.

[0072] In this way, when the first base station receives user data sent by the user terminal, it is determined that it is PDU session data transmitted through the target DRB, indicating that it is PDU session data for implementing the target AI native service requested by the user. The first base station can respond to the PDU session data in accordance with the data processing decision configured for the target AI native service (such as PDU session data from the target DRB transmission is directly sent to the AI ​​server deployed by the first base station for processing, and the service results obtained from the processing are fed back to the corresponding UE through the target DRB), and implement the target AI native service, such as implementing image recognition AI model training or reasoning services, and implementing the data upload and download required for the service through the target DRB transmission.

[0073] It can be seen that the target AI native service requested by the user terminal is directly implemented by the AI ​​service capability of the base station itself, without the need to deploy dedicated UPF and MEC servers, which greatly improves the computing power utilization of the base station. In addition, since the target AI native service is implemented at the network layer within the base station, it does not need to be transmitted externally and is not subject to external network interference and is no longer restricted by external networks. Figure 1 The network performance parameters such as latency and bandwidth of the N3 Tunnel and the link between the UPF and DN networks shown in the figure improve data processing efficiency and security while saving transmission costs, and reliably meet the performance requirements of edge AI applications.

[0074] Reference Figure 4 , which is a flow chart of a data processing method proposed in Embodiment 2 of the present application. This embodiment can describe an optional detailed implementation method of implementing a target AI native service by a first base station responding to PDU session data transmitted from a target DRB, such as Figure 4 As shown, the data processing method proposed in this embodiment may include:

[0075] Step S41, the first base station receives a PDU session request message; the first base station is a base station that can provide the requested target AI native service for the user terminal, and the PDU session request message carries a PDU session attribute identifier;

[0076] Step S42, determining that the user terminal requests to establish a PDU session to implement a target AI native service based on the PDU session attribute identifier, and configuring a target DRB for the first base station to establish a PDU session with the user terminal;

[0077] Step S43, receiving the PDU session data transmitted by the target DRB from the user terminal;

[0078] Regarding the implementation process of step S41 to step S43, reference may be made to the description of the corresponding parts of the above embodiment, such as the description of the corresponding parts of step S31 to step S33, which will not be repeated in this embodiment.

[0079] Step S44, determining at least one second base station that supports the target AI native service;

[0080] Step S45: The received PDU session data is diverted to the AI ​​servers deployed by the first base station and at least one second base station, respectively, and the target AI native service is completed through the collaboration of each AI server.

[0081] In an embodiment of the present application, for each base station deployed in a wireless communication network, multiple base stations that can provide the same AI native service can be grouped into a base station cluster, and the AI ​​servers deployed at each base station in the base station cluster can communicate through the data network DN, so that the multiple base stations in the base station cluster can collaborate to complete an AI native service, thereby improving the reliability of the implementation of the AI ​​native service.

[0082] Based on this, the first base station determines that the PDU session data transmitted by the target DRB is received, that is, the user data of the target AI native service requested by the user terminal. In order to improve service efficiency and reliability, the first base station can first determine whether there is a base station cluster for the target AI native service, that is, whether there is at least one second base station (which is different from the first base station) that supports the target AI native service. If so, it means that there is currently a base station cluster that implements the target AI native service. The first base station can directly implement the diversion of the PDU session data at the network layer, and divert the PDU session data transmitted by the target DRB to the DN between the first base station and the at least one second base station, and complete the target AI native service through the collaboration of multiple AI servers within the DN (that is, the AI ​​servers deployed by each of the multiple base stations).

[0083] For example, for large models with billions or even tens of billions of parameters that need to be deployed, the required resources are often relatively large, and a single base station cannot independently carry and run the AI ​​service implemented by the large model. The large model can be divided into different modules and then deployed to multiple base stations respectively. These multiple base stations collaborate to process the same AI native service implemented based on the large model.

[0084] It can be seen that in the implementation process of step S44 and step S45, the first base station can obtain the target model deployment requirements for implementing the target AI native service; the target model deployment requirements indicate the segmentation results of the target model to be deployed, and the target model can be a machine learning model for implementing the target AI native service, such as a large model LM, a large language model LLM or other models. In this way, the first base station can determine at least one second base station that matches the segmentation result based on the target model deployment requirements, that is, the base station that needs to deploy each module of the target model (that is, the segmentation result), and these multiple base stations can constitute a base station cluster that provides the target AI native service.

[0085] Among them, since the modules divided by the target model deploy different data or execute different subtasks of the target AI native service, the first base station can, according to the segmentation result, divert the PDU session data transmitted by the target DRB to the AI ​​server deployed by the first base station itself and the AI ​​servers deployed by the determined second base stations, to ensure that the AI ​​servers of each base station that deploys the modules after the target model segmentation can receive the corresponding part of the PDU session data and collaborate to complete the target AI native service, such as completing the deployment of the target model or realizing AI applications based on the target model.

[0086] It should be noted that in the process of a first base station and at least one second base station collaborating to implement a target AI native service, how the first base station diverts the PDU session data transmitted by the target DRB, which part of the PDU session data is transmitted to which second base station through the DN, and which part of the data is retained in the first base station for execution. As in the example above, it can be determined based on the deployment results of the target modeler for implementing the target AI native service, or it can be determined based on the service resources (such as various hardware resources and software resources, etc.) actually possessed by each base station that can provide the target AI native service, so as to ensure that the PDU session data diverted to the base station can be successfully executed and collaborate to complete the requested target AI native service.

[0087] In addition, for the target AI native service requested by the user terminal, if the first base station has sufficient resources to support its implementation, after the first base station receives the PDU session data transmitted by the target DRB, it does not need to be diverted to the second base station. The AI ​​server deployed by the first base station itself responds to the PDU session data to implement the target AI native service. The implementation process can be determined according to the service type. This application does not elaborate on the implementation process of how the AI ​​server implements the target AI native service.

[0088] In some other embodiments proposed in the present application, the first base station may also directly divert all PDU session data transmitted by the target DRB to at least one second base station. The first base station, as an access gateway for the PDU session data, does not need to execute the target AI native service. Instead, the second base stations to which the PDU session data is diverted respond to implement the target AI native service requested by the user terminal.

[0089] Based on the above analysis, refer to Figure 5 The flowchart of a data processing method proposed in the third embodiment of the present application is shown. This embodiment can describe the data processing flow performed by the first base station after the target DRB between the first base station and the user terminal is established. Regarding the establishment process of the target DRB dedicated to the target AI native service, refer to the description of the corresponding part of the context embodiment. This embodiment will not be described in detail here. Figure 5As shown, the data processing method proposed in this embodiment may include but is not limited to the following steps:

[0090] Step S51, the first base station receives PDU session data transmitted by a target DRB from a user terminal; the target DRB supports the implementation of a target AI native service requested by the user terminal;

[0091] Step S52, determining the idle service resources currently possessed by the AI ​​server deployed by the first base station, and the expected service resources required to complete the target AI native service;

[0092] Step S53: Based on the idle service resources and the expected service resources, determine whether the target base station cluster needs to collaborate to complete the target AI native service; if not, proceed to step S54; if yes, execute step S55;

[0093] Combined with the above, the composition of the PDU session for implementing the AI ​​native service established in this application is as follows: Figure 2 As shown, after the first base station is configured to implement the target DRB for the target AI native service, the PDU session data required by the user terminal to implement the target AI native service can be transmitted to the first base station through the target DRB. After the first base station receives the PDU session data, it does not need to transmit it to the outside (such as the core network), and the target AI native service requested by the user terminal can be completed directly within the local wireless network of the base station.

[0094] Based on this, before the first base station responds to the PDU session data transmitted by the target DRB, in order to determine whether the PDU session data needs to be diverted, the idle service resources currently possessed by the first base station can be obtained first, that is, the idle service resources currently possessed by the AI ​​server deployed by the first base station to implement the AI ​​native service, and the expected service resources required to complete the target AI native service requested by the UE can be determined. After comparison, it is determined whether the idle service resources of the first base station can support the completion of the target AI native service, so as to avoid the situation where the first base station’s idle service resources are insufficient and it still responds to the PDU session data by itself, resulting in the failure of the execution of the target AI native service requested by the UE, or requiring the UE to wait for a long time for the first base station to have sufficient idle service resources (which can change dynamically according to the changes in the tasks performed by the first base station) before execution, thereby reducing the execution reliability and efficiency of the AI ​​native service.

[0095] Therefore, after comparing the above service resources, it is determined that the idle service resources of the first base station are greater than the above expected service resources, indicating that the first base station itself can complete the target AI native service, and step S54 can be directly executed; on the contrary, when the idle service resources of the first base station are less than or equal to the above expected service resources, it means that the first base station itself cannot complete the target AI native service, and the target base station cluster (that is, a set of multiple base stations that can provide the target AI native service) for the target AI native service needs to collaborate to complete the target AI native service, and step S55 can be executed, thereby ensuring that the base station side that can provide AI native services proposed in this application can reliably and efficiently complete the target AI native service requested by the UE.

[0096] It should be noted that, as described in the above embodiment, when the idle service resources of the first base station are sufficient, the first base station can still divert the received PDU session data to the second base station of the target base station cluster. It is not limited to the processing method described in step S54, which can be determined based on pre-configured data processing rules.

[0097] In addition, in some other embodiments, in the implementation method of step S53, in order to reserve certain service resources for the first base station to perform other tasks, the spatial service resources of the first base station can reserve a certain margin. At this time, if the idle service resources of the first base station are greater than the above-mentioned expected service resources, and the difference between the idle service resources and the expected service resources is greater than the reserved service resource amount (which can be determined based on the historical data of the first base station, such as the amount of service resources required to execute a certain type of task that is frequently performed by the first base station, etc., the present application does not impose any restrictions on its numerical value, and it can be flexibly adjusted according to actual conditions), step S54 can be executed; otherwise, step S55 can be executed.

[0098] It can be seen that the implementation method of step S53 can be determined according to the spatial service resource configuration requirements of the first base station in the actual scenario, including but not limited to the comparison implementation method described above.

[0099] Step S54, sending the PDU session data to the AI ​​server deployed by the first base station, and the AI ​​server completes the target AI native service;

[0100] Following the above analysis, when the first base station determines that it implements the target AI native service requested by the UE, the first base station can transmit the received PDU session data to the AI ​​server deployed by itself, and the AI ​​server responds to the PDU session data to complete the target AI native service, such as

[0101] Step S55, determining at least one second base station from the target base station cluster;

[0102] Step S56: The PDU session data is diverted to the AI ​​servers deployed by the at least one second base station, and the target AI native service is completed through the collaboration of each AI server.

[0103] After the above-mentioned comparison of the idle service resources of the first base station and the expected service resources for completing the target AI native service, it is determined that the idle service resources of the first base station are insufficient and the target base station cluster collaboration that supports the target AI native service is needed. At least one second base station that participates in implementing the target AI native service can be selected from the target base station cluster. The selection of the second base station can be implemented based on the idle service resources currently possessed by each base station in the target base station cluster and the collaboration requirements of the first base station for the target AI native service (such as dividing the target AI native service into multiple subtasks with respective service resources required), but is not limited to the implementation method of step S55 described in this embodiment.

[0104] Afterwards, if Figure 6 The PDU session data processing schematic diagram shown in FIG. Figure 6 The base station 1 shown in the figure can use its communication with the second base station (such as Figure 6 The data network between the base stations N) shown in the figure diverts the PDU session data transmitted by the target DRB to the AI ​​servers deployed in each second base station. In this way, each base station responds to the diverted PDU session data to implement the subtasks assigned to the corresponding base station in the target AI native service.

[0105] As described in the above example, the image recognition AI service based on the large model is implemented. Since this application divides the large model (or other AI models) into different modules in advance (this application does not limit the method of dividing the AI ​​model, which may be determined according to the situation), and deploys them to different base stations, these base stations jointly process the tasks implemented by calling the large model. Therefore, in the process of UE calling the large model to implement the image recognition task, the first base station receives the PDU session data transmitted by the target DRB, such as the image data to be recognized and the prompt words used for image recognition. The first base station can divert the PDU session data to the second base station where some modules of the large model are deployed according to the deployment result of the large model on the base station side. The multiple base stations that deploy the large model collaborate to complete the image recognition task. The implementation process is not described in detail in this application.

[0106] Based on the above analysis, the AI ​​server deployed in the base station can be a server designed for the AI ​​application that the base station needs to provide, so as to provide corresponding AI native services. It can adopt a heterogeneous hardware architecture and can be equipped with acceleration chips such as a graphics processing unit (GPU), a field-programmable gate array (FPGA) or an application-specific integrated circuit (ASIC). It uses a combination of a central processing unit (CPU) and an acceleration chip to meet the needs of high-throughput interconnection, and provide powerful computing power support for AI application scenarios such as image processing, video surveillance, intelligent customer service, and marketing recommendations, supporting deep learning training of large models and intelligent application reasoning processes.

[0107] Reference Figure 7 , which is a flow chart of a data processing method proposed in Embodiment 4 of the present application. Based on the data processing method described above, the receiving and processing process of PDU session data can be further processed.

[0108] Line detailed description, such as Figure 7 As shown, the data processing method proposed in this embodiment may include:

[0109] Step S71: The first base station receives a PDU session request message; the PDU session request message carries a PDU session attribute identifier;

[0110] Step S72, determining that the user terminal requests to establish a PDU session to implement a target AI native service based on the PDU session attribute identifier, and configuring a target DRB for the first base station to establish a PDU session with the user terminal;

[0111] The implementation process of step S71 and step S72 can refer to the description of the corresponding part of the context embodiment, and this embodiment will not be described in detail here. Figure 1 The composition of the 5G network PDU session in the 3GPP standard shown in FIG. Figure 2 and Figure 6 The composition of the PDU session shown makes the PDU session established between the first base station and the user terminal UE include the target DRB configured for the target AI native service, and no longer include the above-mentioned N3 Tunnel logical link, thereby reliably avoiding the adverse effects of N3Tunnel's transmission delay, bandwidth and other network performance on the realization of the target AI native service performance.

[0112] Step S73, processing the data packet sent by the user terminal through the wireless air interface protocol stack to obtain PDU session data from the user terminal;

[0113] In the embodiment of the present application, the data transmission between the base station and the user terminal can be implemented through a wireless air interface protocol stack, and the wireless air interface protocol stack may include but is not limited to a physical layer (such as a PHYSICAL Layer, referred to as a PHY layer), a MAC (Medium Access Control) layer, an RLC (Radio Link Control) layer, a PDCP (Packet Data Convergence Protocol) layer, a SDAP (Service Data Adaptation Protocol) layer, and an RRC (Radio Resource Control) layer, etc. The present application does not limit the protocol stack structure of the wireless communication network, and can be reused. Figure 1 The protocol stack implementation in the PDU session communication scenario shown can also be adaptively adjusted according to communication needs to meet the requirements of the corresponding network communication protocol. This application does not give detailed examples one by one.

[0114] Based on the above analysis, the data packet sent by the user terminal can be decapsulated through the wireless air interface protocol stack to obtain the PDU session data from the user terminal. The decapsulation implementation process of the data packet can be determined based on the communication principle of the wireless air interface protocol stack, which is not described in detail in this application.

[0115] Step S74, determining that the PDU session data comes from the target DRB by numbering the entity of the wireless air interface protocol stack;

[0116] Data radio bearer DRB is a general term for different layers of protocol entities and configurations assigned by the base station to the UE, including the various layers of protocol entities included in the wireless air interface protocol stack listed above. It serves as a channel connecting the base station and the UE, and provides the same data packet forwarding processing for the data packets from the UE. Since the PDU session between the UE and the first base station may be configured with multiple DRBs, in order to distinguish these multiple DRBs, a one-to-one unique identifier, such as a DRB number, can be configured to identify different data streams from the same UE or different UEs. The identifier can be designed based on the wireless air interface protocol to ensure the base station's management and scheduling of different data streams and achieve efficient and stable data transmission.

[0117] It can be seen that the entity number in step S74 can be an identifier such as a DRB number, thereby determining whether the DRB transmitting the PDU session data from the UE is the target DRB, that is, whether the PDU session data from the UE received by the first base station is the improved PUD session data of this application, or the PDU session data that complies with the third generation partnership project 3GPP standard, that is, whether the PDU session data received by the first base station from the UE comes from the target DRB specially configured for the target AI native service. It should be noted that the identifier used to distinguish different DRBs includes but is not limited to the entity number proposed in this embodiment.

[0118] In actual applications, the first base station determines in the above manner that the received PDU session data does not come from the target DRB, indicating that the PDU session data is PDU session data that complies with the 3GPP standard and implements non-AI native services. The first base station can upload the PDU session data to the core network, such as Figure 1 The UPF network element shown can be subsequently processed according to the 3GPP standard process to meet the communication needs of ordinary PUD sessions. The implementation process will not be described in detail in this application.

[0119] Among them, for the transmission data between the first base station and the core network UPF network element, that is, the PDU session data received by the first base station, the PDU session data can be processed according to the transmission protocol between the first base station and the UPF network element, such as IP (Internet Protocol), Ethernet (Ethernet Protocol), GTP-U (GPRS Tunneling Protocol in user plane, a user plane protocol), SCTP (Stream Control Transmission Protocol), etc., such as encapsulation, and the processed data packet is transmitted to the UPF network element through the N3 Tunnel, and the core network responds to the PDU session data to implement the corresponding service (non-AI native service).

[0120] Step S75: Send the PDU session data to the data network where the AI ​​server deployed by the first base station is located, and at least one AI server in the data network implements the target AI native service.

[0121] When the first base station determines that the received PDU session data implements the target AI native service, the first base station may implement the target AI native service alone according to the method described in the corresponding part above, or may cooperate with at least one second base station supporting the target AI native service to implement the target AI native service, such as Figure 6As shown, the AI ​​servers deployed by the first base station and each second base station can communicate through the data network DN. In this way, the AI ​​client of the UE can communicate with the AI ​​server in the DN through the target DRB, and the first base station accessed by the UE can divert the PDU session data transmitted by the target DRB to the DN, and the AI ​​server in the DN implements the target AI native service, such as training or reasoning of the AI ​​model.

[0122] If the service result obtained by the first base station and / or at least one determined second base station (i.e., the base station that collaborates to implement the target AI native service) to implement the target AI native service needs to be fed back to the UE, the service result can be fed back to the UE through the target DRB of the wireless air interface. As in the example above, the service result can be an image recognition result, a question and answer result in an intelligent customer service scenario, a product recommendation result in a marketing recommendation scenario, etc., which can be determined based on the PDU session data content and its corresponding target AI native service type. The implementation process is not described in detail in this application.

[0123] Based on the above description of the data processing method at the base station side, Figure 8 As shown, a base station supporting AI native services (such as the first base station mentioned above) may include three parts: a wireless protocol module 81, a transmission protocol module 82 and an AI service module 83, wherein:

[0124] The wireless protocol module 81 is responsible for the wireless air interface protocol between the base station and the user terminal, and realizes data processing through the wireless air interface protocol stack, such as the above steps S73 and S74. It is also responsible for the network protocol (such as 5G protocol) processing between the base station and the core network (such as 5GC), such as the network protocol followed by the N2 interface and the N3 interface to realize the data transmission of the N2 / N3 tunnel, such as the transmission of PDU session request message / reply message, reporting of service resources, etc. In addition, it is also responsible for configuring the data processing strategy of PDU session data based on the AI ​​native service request of the user terminal to realize data diversion processing.

[0125] The transmission protocol module 82 is responsible for the transmission protocol processing between the base station and the external network element entity, as well as the encapsulation and decapsulation of related data packets. For example, the data transmission of the N3 Tunnel between the first base station and the UPF network element mentioned above (such as the PDU session data of the non-target DRB transmission received by the wireless protocol module 81, sent to the transmission protocol module 82, and then encapsulated by the transmission protocol module 82 with GTP-U and sent to the UPF network element for processing), and data transmission between base station clusters (such as the diverted PUD session data, service resources published by each base station, etc.).

[0126] The AI ​​service module 83 is responsible for the data processing of the AI ​​native service provided by the base station. For example, the PDU session data transmitted by the target DRB received by the wireless protocol module 81 is sent to the AI ​​service module 83, and the AI ​​service module 83 determines whether other base stations need to cooperate in processing, such as the processing described in the above steps S52-S55. If collaborative processing is required, the AI ​​service module 83 can send the PDU session data that needs to be diverted to the transmission protocol module 82, and the transmission protocol module 82 sends it to the collaborative base station (such as the second base station mentioned above) for processing in the form of IP or Ethernet messages. The service results obtained can be sent to the wireless protocol module 81, and the wireless protocol module 81 sends the service results to the user terminal through the target DRB.

[0127] The above introduces the data processing method provided by the embodiment of the present application from the base station side, and the following will introduce the data processing method provided by the embodiment of the present application from the core network side.

[0128] Reference Fig. 9 , is a flow chart of a data processing method proposed in Embodiment 5 of the present application. The method can be applied to the core network of a wireless communication network, such as the control plane of 5GC, and can include network elements such as AMF, SMF, and UDM. The present application does not elaborate on the types of network elements constituting the core network and their communication principles. Only 5GC is used as an example for explanation. Fig. 9 As shown, the data processing method proposed in this embodiment may include:

[0129] Step S91, receiving a PDU session establishment request message sent by a user terminal; the PDU session establishment request message carries a service identifier of a target AI native service requested by the user terminal;

[0130] Step S92, determining the service resource demand indicated by the service identifier and the service resources currently possessed by the first base station accessed by the user terminal;

[0131] Step S93: If the service resources currently possessed by the first base station match the above service resource requirements, a PDU session request message is sent to the first base station;

[0132] When the UE requests the target AI native service, it can send a PDU session establishment request message carrying the corresponding service identifier. As analyzed above, the service identifier can be DNN or S-NSSAI or a combination of the two, such as "DNN= AI-Native", "SST=100, SD=1" or a combination of two. Since the service identifier has been predefined on the core network side of the wireless communication network, after obtaining the service identifier, it can quickly and accurately know that the UE requests to establish a PDU session to implement the target AI native service based on its predefined content.

[0133] In this way, according to the data processing method proposed in the present application, the service resource requirements indicated by the service identifier can be analyzed, such as the expected service resources required to implement the target AI native service, and the service resources currently possessed by the first base station accessed by the UE (i.e., software and hardware information that characterizes the current AI service capabilities of the first base station, etc.) can be queried, such as the current idle service resources dynamically reported by the first base station through the N2 interface (such as when its idle service resources increase, decrease, or update, etc.), etc. The core network can analyze and determine whether the first base station can provide relevant AI services to the UE, i.e., the target AI native service requested by the UE.

[0134] Combined with the above description of the target AI native service, since the service resource types and / or sizes required to implement different AI native services are often different, the core network can determine whether the first base station can provide the target AI native service requested by the UE by judging whether the service resources of the first base station match the service resource requirements of the PDU session establishment request from the UE. If it is determined that the first base station can provide the target AI native service or part of its services, the core network decides to establish a PDU session for the UE, and the PDU session does not require the participation of the UPF network element. It only needs to configure a dedicated DRB to implement the target AI native service and transmit the PDU session data sent by the UE to the first base station.

[0135] Afterwards, the core network may send a PDU session request message to the first base station, and the PDU session request message carries a PDU session attribute identifier, and the PDU session attribute identifier is a service identifier in a PDU session establishment request message from the UE, such as S-NSSAI, etc. If the service identifier cannot be carried, some IEs in the PDU session may be processed specifically to obtain a PDU session attribute identifier. At this time, an address identifier may be obtained based on the attribute information in the PDU session establishment request message, such as setting a special N3 Tunnel Info auxiliary identifier. The implementation method is not limited in this application and may be determined as appropriate.

[0136] Exemplarily, the core network can set the address value in the attribute "UL NG-U UP TNL" (ref.3GPP TS38.413) in the PDU session request message to the local loopback address (such as "GTP tunnel .Endpoint IP Address =127.0.0.1"), and set the unique value of GTP.TEID to identify the user service flow (i.e., the PDU session service flow of the target AI native service) and distinguish the ordinary PDU session service flow (i.e., the PDU session service flow of the non-AI native service), but is not limited to the address identification content.

[0137] It should be noted that for the service identifier of the target AI native service, in addition to reusing the service identifier of the wireless communication network protocol (such as 5G), the present application can also pre-define a new service identifier for the target AI native service in the core network, such as S-NSSAI ID, etc. However, this requires processing the protocol of the wireless communication network to make it compatible with the service identifier and implement the data processing method proposed in this application.

[0138] In some further embodiments, the PDU session request message may also carry a service resource requirement policy or identification information of the PDU session requested to be established by the UE, thereby instructing the first base station to establish a PDU session between it and the UE. In this way, the first base station can directly configure a DRB dedicated to the target AI native service requested by the UE.

[0139] Step S94: receiving a PDU session reply message fed back by the first base station.

[0140] Combined with the data processing method described above from the base station side, after the first base station successfully configures the target DRB for establishing a PDU session between the first base station and the user terminal UE based on the PDU session attribute identifier carried by the PDU session request message (i.e., a dedicated DRB for transmitting the PDU session data sent by the user terminal for the target AI native service), it can send a PDU session reply message to the core network to notify the core network that the wireless resource allocation for the requested PDU session has been completed, so that the PDU session is in an "active" state / running state, and the PDU session can be monitored and managed to achieve rational use of wireless resources.

[0141] Based on this, in the process of monitoring and managing the PDU session, the status information of the PDU session established between the first base station and the user terminal can be obtained, such as the PDU session end request message sent by the user terminal, or the detection that the user terminal has not synchronized the update message with the core network for a long time, or the message response timeout, etc. The core network can respond to the status information and release the target DRB configured by the first base station for the PUD session according to the content of the status information, or add at least one DRB for implementing other AI native services in the PDU session to instruct the first base station to perform corresponding resource processing according to the corresponding message. Among them, regarding the implementation process of modifying the PDU session, the PDU session establishment process can be referred to, and this application will not be described in detail.

[0142] To summarize, in an embodiment of the present application, after the core network receives the PDU session establishment request message sent by the UE, it determines that the UE requests to implement the target AI native service based on the service identifier carried by it. Since the base station of the present application has AI service capabilities, it can determine whether the service resources of the first base station accessed by the UE match the service resource requirements indicated by the service identifier. If they match, it will decide to establish a PDU session containing only DRBs without the participation of UPF network elements, that is, to establish an AI native service PDU session that is different from an ordinary PDU session. At this time, the core network directly sends a PDU session request message carrying a PDU session attribute identifier to the first base station, so that the first base station can establish a PDU session containing a target DRB for the target AI native service, and transmit PDU session data for implementing the target AI native service, and implement the target AI native service on the base station side. There is no need to transmit the PDU session data externally, which improves data processing security and efficiency, saves transmission costs, and since there is no need to parse the PDU session data at the application layer, the execution time of the base station for the AI ​​native service is greatly improved.

[0143] During the data processing method implemented on the core network side, if it is determined that the service resources currently possessed by the first base station do not match the service resource requirements, the core network may refuse to respond to the PDU session establishment request of the user terminal, and feedback a PDU session establishment refusal message to the user terminal. In another possible implementation, if it is further determined that the service resource requirements match the service resources currently possessed by the second base station (i.e., at least one other base station in the target base station cluster except the first base station), a PDU session request message may still be sent to the first base station. At this time, the PDU session request message may carry the base station identifier of the second base station, so that the first base station and the second base station collaborate to realize the target AI native service. In combination with the above base station side embodiment, the first base station, as the access gateway of the user terminal, may divert the PDU session data from the user terminal transmitted by the target DRB to the second base station, and the AI ​​server deployed by the second base station responds to the PDU session data to realize the target AI native service requested by the user terminal.

[0144] Combined with the data processing methods described above from the base station side and the core network side, the following will take the AI ​​model training / inference scenario under the 5G network as an example to illustrate the implementation process of the data processing method. Fig.10 The signaling flow chart of the data processing method proposed in Embodiment 6 of the present application is shown, and the data processing method may include:

[0145] Step S101, the user terminal sends a PDU session establishment request message to the 5G core network;

[0146] Among them, the PDU session establishment request message carries the service identifier of the target AI native service requested by the user terminal.

[0147] Step S102, the 5G core network determines that the service resources currently possessed by the first base station match the service resource requirements indicated by the service identifier;

[0148] Step S103, the 5G core network sends a PDU session request message to the first base station;

[0149] Among them, the PDU session request message carries a PDU session attribute identifier, which is the service identifier in the PDU session establishment request message from the UE, or the address identifier obtained by processing the attribute information in the PDU session establishment request message, etc.

[0150] Step S104: The first base station determines, based on the PDU session attribute identifier, that the user terminal requests to establish a PDU session to implement a target AI native service, and configures a data processing strategy for a target DRB established to implement the target AI native service;

[0151] Step S105, the first base station configures a target DRB for establishing a PDU session with the user terminal, and notifies the user terminal that the PDU session establishment request requested by the user terminal has been accepted;

[0152] Step S106, the first base station sends a PDU session reply message to the 5G core network;

[0153] Regarding the implementation process of step S101-step S106, please refer to the description of the corresponding part of the above embodiment, which will not be repeated in this embodiment. Among them, the first base station is the base station accessed by the user terminal. When deciding to establish a DRB dedicated to the target AI native service, namely the target DRB, the data processing strategy is configured, that is, the PDU session data received from the target DRB is directly sent to the AI ​​server in the first base station for processing, and the data processed by the AI ​​server is also sent to the user terminal through the target DRB. This application does not limit the representation method of the data processing decision.

[0154] During the process of configuring the target DRB by the first base station, the configuration information of the target DRB (i.e., the above-mentioned DRB information) can be sent to the user terminal through the communication method supported by the RRC protocol to complete the creation of the target DRB, so that when the user terminal sends PDU session data to realize the target AI native service, it is sent to the first base station through the target DRB. The first base station can process the PDU session data transmitted by the target DRB according to the configured data processing strategy.

[0155] Step S107: The user terminal uploads PDU session data for implementing the target AI native service to the first base station through the target DRB;

[0156] Step S108: The first base station and / or the second base station responds to the PDU session data according to the data processing strategy, implements the target AI native service, and obtains the corresponding service result.

[0157] Step S109: The first base station sends the service result to the user terminal through the target DRB.

[0158] In the case where the target AI native service is AI model training / inference, the PDU session data sent by the user terminal may include data required to implement AI model training / inference, such as task samples / data to be processed, task prompts, etc. After receiving the PDU session data, the first base station may determine, based on the actual deployment results of the AI ​​model on the base station side, whether the first base station completes the AI ​​model training / inference alone, or shunts the PDU session data to the second base station (i.e., at least one base station that supports the target AI native service) through the data network, and the first base station and the second base station collaborate to complete the AI ​​model training / inference. Such as federated learning, etc., depending on the situation.

[0159] It should be noted that when multiple base stations need to collaborate to complete the target AI native service, these multiple base stations can communicate with the user terminal by sharing an AI service virtual IP address (i.e., service address). In this way, the user terminal's AI native service request can also use the service address as the service destination address, ensuring that the corresponding PDU session data is reliably sent to the base station through the target DRB, thereby improving the execution reliability of the target AI native service.

[0160] The above describes the data processing method provided in the embodiments of the present application. The following describes a device for executing the above data processing method.

[0161] Reference Fig.11 , is a structural diagram of a data processing device proposed in Embodiment 1 of the present application, and the data processing device can be applied to the above-mentioned base station, that is, a base station capable of providing a requested target AI native service to a user terminal, such as Fig.11 As shown, the data processing device may include:

[0162] The first receiving module 111 is used to receive a packet data unit PDU session request message; the PDU session request message carries a PDU session attribute identifier;

[0163] The first configuration module 112 is used to determine, based on the PDU session attribute identifier, that the user terminal requests to establish a PDU session to implement a target AI native service, and configure a target data radio bearer DRB for establishing a PDU session between the first base station and the user terminal; the target DRB supports the implementation of the target AI native service;

[0164] A second receiving module 113 is used to receive the PDU session data transmitted by the target DRB from the user terminal;

[0165] The execution module 114 is used to implement the target AI native service in response to the PDU session data.

[0166] In some embodiments, the execution module 114 may include any of the following processing units:

[0167] A first processing unit, configured to send the PDU session data to an AI server deployed by the first base station, so that the AI ​​server completes the target AI native service;

[0168] A second processing unit is configured to determine at least one second base station that supports the target AI native service, and to offload the PDU session data to AI servers respectively deployed by the first base station and the at least one second base station, so that the target AI native service is completed through collaboration of the AI ​​servers;

[0169] A third processing unit is configured to determine at least one second base station that supports the target AI native service, and to offload the PDU session data to the AI ​​servers respectively deployed by the at least one second base station, so that the target AI native service is completed through the collaboration of the AI ​​servers;

[0170] Among them, the second base station and the first base station belong to a target base station cluster that can provide the target AI native service, the AI ​​servers in the target base station cluster communicate through a data network, and the service addresses of the base stations in the target base station cluster for the target AI native service are the same.

[0171] In some further embodiments, the execution module 114 may further include:

[0172] A first determining unit, configured to determine idle service resources currently possessed by the AI ​​server deployed by the first base station, and expected service resources required to complete the target AI native service;

[0173] The second determination unit is used to determine whether the target base station cluster needs to collaborate to complete the target AI native service based on the idle service resources and the expected service resources; if not, trigger the first processing unit to send the PDU session data to the AI ​​server deployed by the first base station, and the AI ​​server completes the target AI native service; if necessary, trigger the second processing unit or the third processing unit to determine at least one second base station that supports the target AI native service.

[0174] In a possible implementation, the second processing unit may include:

[0175] A first acquisition unit is used to acquire a target model deployment requirement for implementing the target AI native service; the target model deployment requirement indicates a segmentation result of a target model to be deployed, and the target model is a machine learning model for implementing the target AI native service;

[0176] A third determining unit, configured to determine at least one second base station matching the segmentation result according to the target model deployment requirement;

[0177] A diversion unit is used to divert the PDU session data to the AI ​​servers deployed by the first base station and the at least one second base station respectively according to the segmentation result, so as to complete the deployment of the target model.

[0178] In some embodiments, the second receiving module 113 may include:

[0179] a fourth processing unit, configured to process the data packet sent by the user terminal through a wireless air interface protocol stack to obtain PDU session data from the user terminal;

[0180] The fourth determining unit is used to determine that the PDU session data comes from the target DRB by numbering the entity of the wireless air interface protocol stack.

[0181] Based on the above analysis, the data processing device may further include at least one of the following modules:

[0182] A first transmission module is used to determine that the PDU session data does not come from the target DRB by numbering the entity of the wireless air interface protocol stack, and transmit the PDU session data to the core network to respond to the PDU session data according to the third generation partnership project 3GPP standard to implement corresponding services;

[0183] A second transmission module is used to obtain a service result of implementing the target AI native service, and feed back the service result to the user terminal through the target DRB of the wireless air interface;

[0184] A service resource reporting module is used for each base station in the target base station cluster to report its own service resources to the core network; the service resources represent the AI ​​service capabilities of the corresponding base station;

[0185] The second configuration module is used to determine that when receiving the PDU session data transmitted by the target DRB from the user terminal, it configures a data processing strategy for the target AI native service to trigger the execution module 114 to implement the target AI native service in accordance with the data processing strategy in response to the PDU session data.

[0186] Reference Fig.12 , is a structural diagram of a data processing device provided in Embodiment 2 of the present application, the data processing device is deployed in a core network, such as Fig.12 As shown, the data processing device may include:

[0187] The first receiving module 121 is used to receive a packet data unit PDU session establishment request message sent by a user terminal; the PDU session establishment request message carries a service identifier of a target artificial intelligence AI native service requested by the user terminal;

[0188] A first determining module 122, configured to determine a service resource requirement indicated by the service identifier and a service resource currently possessed by the first base station to which the user terminal accesses;

[0189] A first sending module 123 is configured to send a PDU session request message to the first base station if the service resources currently possessed by the first base station match the service resource requirements; the PDU session request message carries a PDU session attribute identifier, and the PDU session attribute identifier is the service identifier, or an address identifier obtained by processing the attribute information in the PDU session establishment request message;

[0190] The second receiving module 124 is used to receive a PDU session reply message fed back by the first base station; the PDU session reply message is sent by the first base station after successfully configuring the target data radio bearer DRB for establishing a PDU session between the first base station and the user terminal based on the PDU session attribute identifier; the target DRB is used to transmit the PDU session data sent by the user terminal for the target AI native service.

[0191] In some embodiments, the data processing device may further include at least one of the following modules:

[0192] A second sending module is configured to feed back a PDU session establishment rejection message to the user terminal if the service resources currently possessed by the first base station do not match the service resource requirements;

[0193] A second determination module is configured to determine, if the service resources currently possessed by the first base station do not match the service resource demand, that the service resource demand matches the service resources currently possessed by the second base station, and trigger the first sending module 123 to send a PDU session request message to the first base station; the PDU session request message carries a base station identifier of the second base station, so that the first base station collaborates with the second base station to implement the target AI native service;

[0194] The PDU session management module obtains the status information of the PDU session established between the first base station and the user terminal, and in response to the status information, releases the target DRB configured for the PDU session, or adds at least one DRB for implementing other AI native services in the PDU session.

[0195] An embodiment of the present application also provides a computer program product, including computer-readable instructions. When the computer-readable instructions are executed on a network device (such as a base station or a core network deployment device), the network device implements any data processing method provided in the corresponding embodiment of the present application.

[0196] Wherein, when loading and executing computer-readable instructions on a computer, all or part of the process or function described in the embodiment of the present application is generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer-readable instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer-readable instructions may be transmitted from a website site, a computer, a training device, or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, training device, or data center, to meet the data / instruction transmission requirements in the execution process of the data processing method proposed in the embodiment of the present application.

[0197] Reference Fig.13 , is a schematic diagram of the hardware structure of a network device applicable to the data processing method proposed in this application, and the network device may be a base station, such as Fig.13 As shown, the network device may include: at least one communication element 131, at least one memory 132 and at least one processor 133, wherein:

[0198] At least one communication element 131, at least one memory 132 and at least one processor 133 can communicate with each other via a bus. The bus can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig.13 Only one bidirectional line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0199] The communication element 131 can be used to realize the communication connection between the network device and the user terminal, and can also be used to realize the communication connection between the network device and the core network deployment device (such as the network element entity included in the 5GC, etc.) as needed, so as to realize the transmission of the corresponding message during the execution of the data processing method. It can be seen that the communication element 131 may include communication elements corresponding to wireless communication methods such as wifi, 5G / 6G and other mobile communications. In order to meet the data transmission between the internal components of the network device, it may also include one or more interfaces that support wired communication methods, such as general-purpose input / output (GPIO) interface, USB interface, universal asynchronous receiver / transmitter (UART) interface, etc. This application does not limit the composition structure of the communication element 131 to realize this function and its corresponding communication transmission mechanism, which can be determined according to the circumstances.

[0200] The memory 132 can be used to store a computer program (multiple computer-readable instructions) that implements the data processing method proposed in the embodiments of the present application; the processor 133 can load and execute the computer program stored in the memory 132 to implement the various steps of the data processing method proposed in the embodiments of the present application. The implementation process can refer to the description of the corresponding part of the method embodiment executed on the corresponding device side above.

[0201] In the embodiment of the present application, the memory 132 may include various storage media. The processor 133 may include any one or more of a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), a digital signal processor (DSP), an application specific integrated circuit (ASIC), or a field-programmable gate array (FPGA).

[0202] It should be understood that Fig.13 The structure of the network device shown does not constitute a limitation on the network device in the embodiment of the present application. In actual applications, the network device may include Fig.13 This application does not give detailed examples of more or fewer components shown, or combinations of certain components, such as basic components of a base station, etc.

[0203] A computer-readable storage medium is also provided in an embodiment of the present application. The storage medium carries one or more computer programs. When the one or more computer programs are executed by a network device, the network device can implement any data processing method provided in the corresponding embodiment of the present application.

[0204] The computer-readable storage medium may be any available medium that can be stored by a computer or a data storage device such as a training device or a data center that includes one or more available media. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid state drive (SSD)).

[0205] It should also be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed over multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the drawings of the device embodiments provided by the present application, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines.

[0206] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. In this specification, each embodiment is described in a progressive or parallel manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the embodiments can be referred to each other. For the device and network equipment disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

Claims

1. A data processing method, characterized in that: The data processing method comprises: The first base station receives a packet data unit PDU session request message; the first base station is a base station that can provide the requested target artificial intelligence AI native service for the user terminal, and the PDU session request message carries a PDU session attribute identifier; Determining, based on the PDU session attribute identifier, that the user terminal requests to establish a PDU session to implement a target AI native service, configuring a target data radio bearer DRB for establishing a PDU session between the first base station and the user terminal; the target DRB supports the implementation of the target AI native service; Receiving the PDU session data transmitted by the target DRB from the user terminal; In response to the PDU session data, implementing the target AI native service; The implementing the target AI native service in response to the PDU session data includes any one of the following: Sending the PDU session data to an AI server deployed by the first base station, so that the AI ​​server completes the target AI native service; Determine at least one second base station that supports the target AI native service, offload the PDU session data to AI servers deployed by the first base station and the at least one second base station, and complete the target AI native service through collaboration of the AI ​​servers; Determine at least one second base station that supports the target AI native service, offload the PDU session data to AI servers deployed by each of the at least one second base station, and complete the target AI native service through collaboration of each of the AI ​​servers; Among them, the second base station and the first base station belong to a target base station cluster that can provide the target AI native service, the AI ​​servers in the target base station cluster communicate through a data network, and the service addresses of the base stations in the target base station cluster for the target AI native service are the same.

2. The data processing method according to claim 1, characterized in that: The implementing the target AI native service in response to the PDU session data further includes: Determine idle service resources currently owned by the AI ​​server deployed by the first base station, and expected service resources required to complete the target AI native service; Based on the idle service resources and the expected service resources, determining whether the target base station cluster needs to collaborate to complete the target AI native service; If not, executing the step of sending the PDU session data to the AI ​​server deployed by the first base station, and the AI ​​server completes the target AI native service; If necessary, execute the step of determining at least one second base station that supports the target AI native service.

3. The data processing method according to claim 1, characterized in that: The determining at least one second base station that supports the target AI native service, and offloading the PDU session data to AI servers deployed by the first base station and the at least one second base station respectively, includes: Obtaining a target model deployment requirement for implementing the target AI native service; the target model deployment requirement indicates a segmentation result of a target model to be deployed, the target model being a machine learning model for implementing the target AI native service; Determining at least one second base station matching the segmentation result according to the target model deployment requirement; According to the segmentation result, the PDU session data is diverted to the AI ​​servers deployed by the first base station and the at least one second base station respectively to complete the deployment of the target model.

4. The data processing method according to any one of claims 1 to 3, characterized in that: The receiving the PDU session data transmitted by the target DRB from the user terminal includes: Processing the data packet sent by the user terminal through the wireless air interface protocol stack to obtain PDU session data from the user terminal; By numbering the entity of the wireless air interface protocol stack, it is determined that the PDU session data comes from the target DRB, and the step of executing the step is responded to the PDU session data to realize the target AI native service.

5. The data processing method according to claim 4, characterized in that: The data processing method further includes at least one of the following: By numbering the entity of the wireless air interface protocol stack, determining that the PDU session data is not from the target DRB, transmitting the PDU session data to the core network, so as to respond to the PDU session data according to the third generation partnership project 3GPP standard and implement corresponding services; Obtain a service result for implementing the target AI native service, and feed the service result back to the user terminal through the target DRB of the wireless air interface; Each base station in the target base station cluster reports its own service resources to the core network; the service resources represent the AI ​​service capabilities of the corresponding base station; When it is determined that the target DRB transmits the PDU session data from the user terminal, a data processing strategy for the target AI native service is configured to implement the target AI native service in response to the PDU session data in accordance with the data processing strategy.

6. A data processing method, characterized in that: The data processing method comprises: Receiving a packet data unit (PDU) session establishment request message sent by a user terminal; the PDU session establishment request message carries a service identifier of a target artificial intelligence (AI) native service requested by the user terminal; Determine the service resource demand indicated by the service identifier and the service resources currently possessed by the first base station accessed by the user terminal; If the service resources currently possessed by the first base station match the service resource requirements, send a PDU session request message to the first base station; the PDU session request message carries a PDU session attribute identifier, and the PDU session attribute identifier is the service identifier, or an address identifier obtained by processing based on the attribute information in the PDU session establishment request message; receiving a PDU session reply message fed back by the first base station; the PDU session reply message is sent by the first base station after successfully configuring a target data radio bearer DRB for establishing a PDU session between the first base station and the user terminal based on the PDU session attribute identifier; the target DRB is used to transmit the PDU session data for the target AI native service sent by the user terminal; The target AI native service is implemented by responding to the PDU session data, and the implementation process of the target AI native service includes any one of the following: The first base station sends the PDU session data to an AI server deployed by the first base station, and the AI ​​server completes the target AI native service; The first base station determines at least one second base station that supports the target AI native service, distributes the PDU session data to AI servers deployed by the first base station and the at least one second base station, and completes the target AI native service through collaboration of the AI ​​servers; The first base station determines at least one second base station that supports the target AI native service, and distributes the PDU session data to AI servers respectively deployed by the at least one second base station, so that the target AI native service is completed through collaboration of the AI ​​servers; Among them, the second base station and the first base station belong to a target base station cluster that can provide the target AI native service, the AI ​​servers in the target base station cluster communicate through a data network, and the service addresses of the base stations in the target base station cluster for the target AI native service are the same.

7. A data processing device, characterized in that: The data processing device is deployed in a base station, and the base station is configured as a first base station capable of providing a requested target artificial intelligence AI native service for a user terminal. The data processing device includes: The first receiving module is used to receive a packet data unit PDU session request message; the PDU session request message carries a PDU session attribute identifier; A first configuration module is configured to determine, based on the PDU session attribute identifier, that the user terminal requests to establish a PDU session to implement a target AI native service, and configure a target data radio bearer DRB for establishing a PDU session between the first base station and the user terminal; the target DRB supports the implementation of the target AI native service; A second receiving module, configured to receive the PDU session data transmitted by the target DRB from the user terminal; An execution module, configured to implement the target AI native service in response to the PDU session data; The execution module implements the target AI native service in response to the PDU session data, including any one of the following: The execution module sends the PDU session data to the AI ​​server deployed by the first base station, and the AI ​​server completes the target AI native service; The execution module determines at least one second base station that supports the target AI native service, distributes the PDU session data to AI servers deployed by the first base station and the at least one second base station, and completes the target AI native service through the collaboration of the AI ​​servers; The execution module determines at least one second base station that supports the target AI native service, and distributes the PDU session data to the AI ​​servers respectively deployed by the at least one second base station, so that the target AI native service is completed through the collaboration of each of the AI ​​servers; Among them, the second base station and the first base station belong to a target base station cluster that can provide the target AI native service, the AI ​​servers in the target base station cluster communicate through a data network, and the service addresses of the base stations in the target base station cluster for the target AI native service are the same.

8. A data processing device, characterized in that: The data processing device is deployed in the core network, and the data processing device includes: A first receiving module is configured to receive a packet data unit (PDU) session establishment request message sent by a user terminal; the PDU session establishment request message carries a service identifier of a target artificial intelligence (AI) native service requested by the user terminal; A first determining module, configured to determine a service resource requirement indicated by the service identifier and a service resource currently possessed by the first base station accessed by the user terminal; A first sending module is configured to send a PDU session request message to the first base station if the service resources currently possessed by the first base station match the service resource requirements; the PDU session request message carries a PDU session attribute identifier, and the PDU session attribute identifier is the service identifier, or an address identifier obtained by processing based on the attribute information in the PDU session establishment request message; A second receiving module is configured to receive a PDU session reply message fed back by the first base station; the PDU session reply message is sent by the first base station after successfully configuring a target data radio bearer DRB for establishing a PDU session between the first base station and the user terminal based on the PDU session attribute identifier; the target DRB is used to transmit the PDU session data sent by the user terminal for the target AI native service; The target AI native service is implemented by responding to the PDU session data, and the implementation process of the target AI native service includes any one of the following: The first base station sends the PDU session data to an AI server deployed by the first base station, and the AI ​​server completes the target AI native service; The first base station determines at least one second base station that supports the target AI native service, distributes the PDU session data to AI servers deployed by the first base station and the at least one second base station, and completes the target AI native service through collaboration of the AI ​​servers; The first base station determines at least one second base station that supports the target AI native service, and distributes the PDU session data to AI servers respectively deployed by the at least one second base station, so that the target AI native service is completed through collaboration of the AI ​​servers; Among them, the second base station and the first base station belong to a target base station cluster that can provide the target AI native service, the AI ​​servers in the target base station cluster communicate through a data network, and the service addresses of the base stations in the target base station cluster for the target AI native service are the same.

9. A network device, characterized in that: The network device comprises: at least one communication element, at least one memory and at least one processor, wherein: The communication element is used to realize the communication connection between the network device and the user terminal; The processor is used to load and execute the computer program stored in the memory to implement the various steps of the data processing method described in any one of claims 1 to 5.

Images