AI model scheduling method, system and device

Abstract

The invention relates to the technical field of communication, and provides an AI model scheduling method, system and device, and the method comprises the steps: receiving position information, an intelligent calculation service identifier and an intelligent calculation model identifier through a visiting place AI capability management module, and determining to schedule a same-region model through a stored intelligent calculation model registration form, the service transmission channel interaction module interacts with a deployment place AI capability application module and a visit place AI capability application module, and respectively performs control plane signaling interaction with a visit place service transmission channel user plane network element and a deployment place service transmission channel user plane network element through a service transmission channel control plane network element; and enabling a visit place AI capability application module, a visit place service transmission channel user plane network element, a deployment place service transmission channel user plane network element and a deployment place AI capability application module to obtain core network tunnel information of adjacent network elements, and establishing a user plane data transmission channel to transmit an intelligent calculation model copy. The copy is transmitted from a deployment place to a visit place AI capability application module, and support is provided for intelligent calculation business.

Description

Technical Field

[0001] The present disclosure belongs to the field of communication technology, and in particular relates to an AI model scheduling method, system, and device. Background Art

[0002] On the one hand, with the rapid development of artificial intelligence (AI), generative AI, general-purpose large models, and industry-specific large models are gradually maturing and becoming part of people's daily lives. On the other hand, with the advancement of mobile communications, mobile communications have become an integral part of production and daily life. The high bandwidth, low latency, and wide connectivity of 5G and 5G-A naturally suit the application needs of AI technology. Combining these two approaches to further enhance social productivity and gradually empower a wide range of industries has become a key area of ​​exploration.

[0003] The current network architectures for edge computing in mobile communication networks can be divided into the following four types: Type 1: Edge intelligent computing power is provided by mobile user terminals, and edge inference models are deployed locally on mobile terminals. Type 2: Edge intelligent computing power is set up in an edge data center (such as a city computer room or a campus computer room). Users access the corresponding DN (Data Network) through the mobile communication network to obtain edge intelligent computing services. The third type: combining intelligent computing power with mobile communication network gateways (such as UPF (User Plane Function)), such as MEC (Mobile Edge Computing) technology; The fourth type: intelligent computing power in mobile communication networks, such as NWDAF (Network Data Analytics Function).

[0004] However, all of the above methods have problems: For the first architecture, it uses the mobile terminals of mobile users. The computing power and device power of mobile terminals are limited and cannot support high computing power, long time and diversified intelligent computing needs.

[0005] For the second architecture, its intelligent computing power is deployed in the edge data center. The mobile communication network needs to first establish a terminal-to-DN bearer for the user, and then call the intelligent computing service, which has a latency problem; generally, the user side needs to manually initiate a specific SAAS service trigger, which has poor versatility.

[0006] For the third architecture, 5G MEC can integrate artificial intelligence and machine learning algorithms at the edge of the network. Although it is closer to the user edge, it is logically decoupled from the mobile communication network and relies on the deployment location of the core network gateway (such as UPF).

[0007] In the fourth architecture, NWDAF is a core network element, but it analyzes network data to improve network capabilities and cannot provide users with business-related AI capabilities.

[0008] From the above analysis, we can see that in the face of the growing demand for AI computing, existing mobile communication networks use AI capabilities as one of the means to improve network capabilities. When faced with the demand for low-latency and highly versatile intelligent computing services, the intelligent computing service implementation solutions under the above four architectures are not a complete solution.

[0009] In view of this, it is urgent to propose an intelligent computing network architecture based on which intelligent computing models can be scheduled to support intelligent computing services. Summary of the Invention

[0010] To solve the above problems, the present disclosure provides an AI model scheduling method, system and device to support intelligent computing services.

[0011] In a first aspect, an AI model scheduling method is provided, comprising: The AI ​​capability management module of the visited location receives the location information, intelligent computing service identifier, and intelligent computing model identifier; The AI ​​capability management module at the visited location uses the stored intelligent computing model registration table to determine whether there is an AI capability application module that matches both the intelligent computing service identifier and the intelligent computing model identifier, and if the area information corresponding to the matching AI capability application module does not match the received location information, the module performs the following same-area scheduling: The visited-location AI capability management module initiates an AI model scheduling service request to the visited-location AI capability application module, requesting allocation of core network tunnel information for the visited-location AI capability application module. The intelligent computing model registration table stores a correspondence between the intelligent computing service identifier, intelligent computing model identifier, regional information, and AI capability application module information. The visited-location AI capability application module returns the first core network tunnel information to the visited-location AI capability management module. The visited-site AI capability management module initiates an AI model scheduling service request to the deployment-site AI capability application module pointed to by the matching AI capability application module information, requesting the allocation of the deployment-site AI capability application module-side core network tunnel information. The request carries the intelligent computing service identifier and intelligent computing model identifier. The deployment-site AI capability application module returns the fourth core network tunnel information to the visited-site AI capability management module. The visited location AI capability management module initiates an AI model scheduling service request to the visited location service transmission channel control plane network element. The request carries the first core network tunnel information, the fourth core network tunnel information, the location information, and the regional information of the deployed location AI capability application module. The visited service transmission channel control plane network element uses the location information carried in the AI ​​model scheduling service request to address the visited service transmission channel user plane network element, and uses the area information to address the deployed service transmission channel user plane network element. When it is determined that the visited service transmission channel user plane network element and the deployed service transmission channel user plane network element are the same network element, the user plane network element sends a user plane tunnel establishment request carrying the first core network tunnel information and the fourth core network tunnel information to the visited service transmission channel user plane network element. The user plane network element of the visited service transmission channel sends the second core network tunnel information to the control plane network element of the visited service transmission channel; The visited-site service transmission channel control plane network element sends the second core network tunnel information to the visited-site AI capability application module and the deployed-site AI capability application module via the visited-site AI capability management module. At this point, the visited-site AI capability application module and the deployed-site AI capability application module establish a user plane intelligent computing service model transmission channel. The AI ​​capability application module at the deployed site uses the intelligent computing service identifier and the intelligent computing model identifier to select an intelligent computing model, and passes a copy of the selected intelligent computing model to the AI ​​capability application module at the visited site.

[0012] Furthermore, the method further comprises: The visited service transmission channel control plane network element uses the AI ​​model to schedule the location information and area information carried in the service request. When it is determined that the visited service transmission channel user plane network element and the deployed service transmission channel user plane network element are different network elements, it sends a user plane tunnel establishment request carrying the first core network tunnel information to the visited service transmission channel user plane network element, and sends a user plane tunnel establishment request carrying the fourth core network tunnel information to the deployed service transmission channel user plane network element. The user plane network element of the visited service transmission channel sends the second core network tunnel information to the control plane network element of the visited service transmission channel; The user plane network element of the service transmission channel at the deployment site sends the third core network tunnel information to the control plane network element of the service transmission channel at the visited site; The visited-site service transmission channel control plane network element sends the second core network tunnel information to the visited-site AI capability application module via the visited-site AI capability management module, and sends the third core network tunnel information to the deployment-site AI capability application module; The control plane network element of the visited site service transmission channel sends the second core network tunnel information to the user plane network element of the deployed site service transmission channel, and sends the third core network tunnel information to the user plane network element of the visited site service transmission channel. At this point, the visited site AI capability application module and the deployed site AI capability application module establish a user plane intelligent computing service model transmission channel. The AI ​​capability application module at the deployed site uses the intelligent computing service identifier and the intelligent computing model identifier to select an intelligent computing model, and passes a copy of the selected intelligent computing model to the AI ​​capability application module at the visited site.

[0013] In a second aspect, an AI model scheduling method is provided, the method comprising: The AI ​​capability management module of the visited location receives the location information, intelligent computing service identifier, and intelligent computing model identifier; The AI ​​capability management module at the visited location uses the stored intelligent computing model registration table to determine if there is an AI capability application module that matches both the intelligent computing service ID and the intelligent computing model ID. It then performs the following cross-region scheduling: The visited-location AI capability management module initiates an AI model scheduling service request to the visited-location AI capability application module, requesting allocation of core network tunnel information on the visited-location AI capability application module side. The intelligent computing model registration table stores the correspondence between the intelligent computing service identifier, intelligent computing model identifier, regional information, and AI capability application module information. The visited-location AI capability application module returns the first core network tunnel information to the visited-location AI capability management module. The visited-site AI capability management module initiates an AI model scheduling service request to the visited-site service transmission channel control plane network element, which carries the first core network tunnel information and location information. The visited-site service transmission channel control plane network element addresses the visited-site service transmission channel user plane network element based on the location information and sends a user plane tunnel establishment request, which carries the first core network tunnel information. The visited-site service transmission channel control plane network element receives the second core network tunnel information fed back by the visited-site service transmission channel user plane network element and transmits it to the visited-site AI capability application module via the visited-site AI capability management module. The AI ​​capability management module at the visited location addresses the AI ​​capability management module at the deployed location based on the intelligent computing service identifier and the intelligent computing model identifier, and sends an AI model scheduling service request to the AI ​​capability management module at the deployed location. The request carries the intelligent computing service identifier, the intelligent computing model identifier, and the second core network tunnel information. The AI ​​capability management module at the deployment site determines the AI ​​capability application module at the deployment site based on the intelligent computing service identifier and the intelligent computing model identifier, and initiates the AI ​​model scheduling service process to the module, requesting the AI ​​capability application module at the deployment site to allocate core network tunnel information. The AI ​​capability application module at the deployment site returns the fourth core network tunnel information to the AI ​​capability management module at the deployment site; The AI ​​capability management module at the deployment site initiates an AI model scheduling service request to the service transport channel control plane network element at the deployment site. The request carries the fourth core network tunnel information, the second core network tunnel information, and the area information responsible for the AI ​​capability application module at the deployment site. The service transport channel control plane network element at the deployment site and the service transport channel control plane network element at the visited site may be the same network element or different network elements. The service transmission channel control plane network element at the deployment site addresses the service transmission channel user plane network element at the deployment site according to the area information responsible for the AI ​​capability application module at the deployment site, and initiates a user plane tunnel establishment request to the service transmission channel user plane network element at the deployment site. The request carries the fourth core network tunnel information and the second core network tunnel information. The user plane network element of the service transmission channel at the deployment site returns the third core network tunnel information to the control plane network element of the service transmission channel at the deployment site; The service transmission channel control plane network element at the deployment site returns the third core network tunnel information to the AI ​​capability management module at the deployment site; The AI ​​capability management module at the deployment site returns the AI ​​model scheduling service establishment request result to the AI ​​capability management module at the visited site. The result includes: the third core network tunnel information; The visited location AI capability management module returns the third core network tunnel information to the visited location service transmission channel control plane network element. The visited location service transmission channel control plane network element returns the third core network tunnel information to the visited location service transmission channel user plane network element. At this point, the intelligent computing model scheduling service transmission channel is established. The AI ​​capability application module at the deployed site uses the intelligent computing service identifier and the intelligent computing model identifier to select an intelligent computing model, and passes a copy of the selected intelligent computing model to the AI ​​capability application module at the visited site.

[0014] Furthermore, the AI ​​capability management module at the visited location sends an AI model scheduling service request to the AI ​​capability management module at the deployed location, including: The AI ​​capability management module at the visited location sends an AI model scheduling service request to the AI ​​capability management module at the deployed location via the service transmission channel control plane network element at the visited location and the service transmission channel control plane network element at the deployed location; and / or The AI ​​capability management module at the deployment site returns the AI ​​model scheduling service establishment request result to the AI ​​capability management module at the visiting site, including: The AI ​​capability management module at the deployment site returns the AI ​​model scheduling service request establishment result to the AI ​​capability management module at the visited site via the service transmission channel control plane network element at the visited site and the service transmission channel control plane network element at the deployment site.

[0015] Furthermore, after the AI ​​capability application module at the deployed location selects an intelligent computing model using the intelligent computing service identifier and the intelligent computing model identifier, and transmits a copy of the selected intelligent computing model to the AI ​​capability application module at the visited location, the method further includes: The AI ​​capability application module at the visited location returns the intelligent computing model transfer completion information to the AI ​​capability management module at the visited location, and the AI ​​capability management module registers the model.

[0016] Furthermore, the AI ​​capability management module of the visited location performs addressing in one of the following two ways: Method 1: Configure the global intelligent computing model registration information table locally and notify the global and other AI capability management modules when the registration information changes. Method 2: The AI ​​capability management module completes registration with the network resource database when the intelligent computing model registration information changes. The registration information includes: AI capability management module identifier, AI capability management module address, intelligent computing service identifier, intelligent computing model identifier, AI capability application module identifier, and regional information; the AI ​​capability management module performs AI capability management module service discovery on the network resource database. The request message includes the intelligent computing service identifier and the intelligent computing model identifier, and selects the AI ​​capability management module based on the AI ​​capability management module service discovery result returned by the network resource database.

[0017] Thirdly, an AI model scheduling method is provided, which is applied to the AI ​​capability management module at the visited location, including: Receive location information, intelligent computing service identifier, and intelligent computing model identifier; Using the stored intelligent computing model registration table, it is determined that there is an AI capability application module that matches both the intelligent computing service identifier and the intelligent computing model identifier, and the area information corresponding to the matching AI capability application module does not match the received location information. In this case, intra-area scheduling is performed, and an AI model scheduling service request is initiated to the visited location AI capability application module, requesting allocation of the core network tunnel information on the visited location AI capability application module side. The intelligent computing model registration table stores the correspondence between the intelligent computing service identifier, the intelligent computing model identifier, the area information, and the AI ​​capability application module information. Receive the first core network tunnel information returned by the visited AI capability application module; Initiate an AI model scheduling service request to the AI ​​capability application module at the deployment location pointed to by the matching AI capability application module information, requesting the allocation of the core network tunnel information on the AI ​​capability application module side at the deployment location. The request carries the intelligent computing service identifier and intelligent computing model identifier. Receive the fourth core network tunnel information returned by the AI ​​capability application module at the deployment site; Initiate an AI model scheduling service request to the control plane network element of the visited service transmission channel. The request carries the first core network tunnel information, the fourth core network tunnel information, the location information, and the regional information of the deployed AI capability application module. When it is determined that the user plane network element of the visited service transmission channel and the user plane network element of the deployed service transmission channel are the same network element, the second core network tunnel information sent by the control plane network element of the visited service transmission channel is forwarded to the visited AI capability application module and the deployed AI capability application module, wherein the second core network tunnel information is the core network tunnel information of the user plane network element end of the visited service transmission channel.

[0018] Furthermore, when it is determined that the user plane network element of the visited service transmission channel and the user plane network element of the deployed service transmission channel are different network elements, the second core network tunnel information sent by the control plane network element of the visited service transmission channel is forwarded to the visited AI capability application module, and the third core network tunnel information is forwarded to the AI ​​capability application module of the deployed site, wherein the third core network tunnel information is the core network tunnel information of the user plane network element end of the deployed site service transmission channel.

[0019] Fourthly, an AI model scheduling method is provided, which is applied to the AI ​​capability management module at the visited location, including: Receive location information, intelligent computing service identifier, and intelligent computing model identifier; Using the stored intelligent computing model registration table, it is determined that there is no AI capability application module that matches both the intelligent computing service identifier and the intelligent computing model identifier. An AI model scheduling service request is initiated to the visited location AI capability application module, requesting allocation of the core network tunnel information on the visited location AI capability application module side. The intelligent computing model registration table stores the correspondence between the intelligent computing service identifier, intelligent computing model identifier, regional information, and AI capability application module information. Receive the first core network tunnel information returned by the visited AI capability application module; Initiate an AI model scheduling service request to the control plane network element of the visited service transmission channel, carrying the first core network tunnel information and location information; Forwarding the second core network tunnel information sent by the visited service transmission channel control plane network element to the visited AI capability application module; The AI ​​capability management module at the deployment location is addressed based on the intelligent computing service identifier and intelligent computing model identifier, and an AI model scheduling service request is sent. The request carries the intelligent computing service identifier, intelligent computing model identifier, and second core network tunnel information. Receive the AI ​​model scheduling service request establishment result returned by the AI ​​capability management module at the deployment site, where the result includes: third core network tunnel information, where the third core network tunnel information is the core network tunnel information of the user plane network element end of the deployment site service transmission channel; Return the third core network tunnel information to the visited service transmission channel control plane network element.

[0020] Furthermore, it also includes: Receive the intelligent computing model transfer completion information returned by the AI ​​capability application module at the visited location and register the intelligent computing model.

[0021] Furthermore, addressing is performed using either of the following two methods: Method 1: Configure the global intelligent computing model registration information table locally and notify the global and other AI capability management modules when the registration information changes. Method 2: When the registration information of the intelligent computing model changes, registration with the network resource database is completed. The registration information includes: AI capability management module identifier, AI capability management module address, intelligent computing service identifier, intelligent computing model identifier, AI capability application module identifier, and regional information; the AI ​​capability management module performs AI capability management module service discovery on the network resource database. The request message includes the intelligent computing service identifier and the intelligent computing model identifier, and selects the AI ​​capability management module based on the AI ​​capability management module service discovery result returned by the network resource database.

[0022] In a fifth aspect, an AI model scheduling method is provided, which is applied to the AI ​​capability application module at the visited location, including: Receive AI model scheduling service requests from the AI ​​capability management module at the visited location; Allocate the first core network tunnel information and return it to the visited location AI capability management module; Receive the second core network tunnel information sent by the visited AI capability management module; wherein the second core network tunnel information is the core network tunnel information of the user plane network element end of the visited service transmission channel; Receive the intelligent computing model copy sent by the AI ​​capability application module at the deployment site.

[0023] In a sixth aspect, an AI model scheduling method is provided, which is applied to a control plane network element of a visited service transmission channel. When performing same-area scheduling, the method includes: Receive an AI model scheduling service request initiated by the visited AI capability management module, the request carrying the first core network tunnel information, the fourth core network tunnel information, the location information, and the regional information of the deployed AI capability application module. The first core network tunnel information refers to the core network tunnel information of the visited AI application module, and the fourth core network tunnel information refers to the core network tunnel information of the deployed AI application module. Use the location information carried in the AI ​​model scheduling service request to address the user plane network element of the visited service transmission channel, use the area information to address the user plane network element of the deployed service transmission channel, and when it is determined that the user plane network element of the visited service transmission channel and the user plane network element of the deployed service transmission channel are the same network element, send a user plane tunnel establishment request carrying the first core network tunnel information and the fourth core network tunnel information to the user plane network element of the visited service transmission channel; receiving the second core network tunnel information sent by the user plane network element of the visited service transmission channel; The second core network tunnel information is sent to the visited site AI capability application module and the deployed site AI capability application module respectively via the visited site AI capability management module.

[0024] Furthermore, the method further comprises: Using the location information and area information carried in the AI ​​model scheduling service request, when it is determined that the user plane network element of the visited service transmission channel and the user plane network element of the deployed service transmission channel are different network elements, a user plane tunnel establishment request carrying the first core network tunnel information is sent to the user plane network element of the visited service transmission channel, and a user plane tunnel establishment request carrying the fourth core network tunnel information is sent to the user plane network element of the deployed service transmission channel; receiving the second core network tunnel information sent by the user plane network element of the visited service transmission channel; Receiving the third core network tunnel information sent by the user plane network element of the service transmission channel at the deployment site; Sending the second core network tunnel information to the visited location AI capability application module via the visited location AI capability management module, and sending the third core network tunnel information to the deployed location AI capability application module; The second core network tunnel information is sent to the user plane network element of the service transmission channel at the deployment site, and the third core network tunnel information is sent to the user plane network element of the service transmission channel at the visited site.

[0025] In a seventh aspect, an AI model scheduling method is provided, which is applied to a control plane network element of a visited service transmission channel. When performing cross-region scheduling, the method includes: Receive an AI model scheduling service request initiated by the visited location AI capability management module, which carries the first core network tunnel information and location information; Addressing the user plane network element of the visited service transmission channel according to the location information, sending a user plane tunnel establishment request, and carrying the first core network tunnel information in the request; Receive the second core network tunnel information fed back by the user plane network element of the user's visited service transmission channel, and transmit it to the visited AI capability application module via the visited AI capability management module; Receive the third core network tunnel information returned by the visited AI capability management module, where the third core network tunnel information is the core network tunnel information of the user plane network element end of the deployed service transmission channel; Return the third core network tunnel information to the user plane network element of the visited service transmission channel.

[0026] In an eighth aspect, an AI model scheduling method is provided, which is applied to the AI ​​capability application module at the deployment site and includes: Receive an AI model scheduling service request initiated by the visited AI capability management module to allocate core network tunnel information for the deployed AI capability application module. The request carries the intelligent computing service identifier and intelligent computing model identifier. Allocating fourth core network tunnel information; Return the fourth core network tunnel information to the visited AI capability management module; When the user plane network element of the visited site service transmission channel and the user plane network element of the deployed site service transmission channel are the same network element, receiving the second core network tunnel information sent by the visited site AI capability management module, and interacting with the visited site service transmission channel user plane network element to facilitate the visited site AI capability application module and the deployed site AI capability application module to establish a user plane intelligent computing service model transmission channel; wherein the second core network tunnel information is the core network tunnel information of the user plane network element end of the visited site service transmission channel; Use the intelligent computing service identifier and intelligent computing model identifier to select the intelligent computing model, and pass the selected intelligent computing model copy to the visited AI capability application module through the established user-side intelligent computing service model transmission channel.

[0027] Furthermore, when it is determined that the user plane network element of the visited service transmission channel and the user plane network element of the deployed service transmission channel are not the same network element, the third core network tunnel information sent by the visited AI capability management module is received, wherein the third core network tunnel information is the core network tunnel information of the user plane network element end of the deployed service transmission channel.

[0028] In a ninth aspect, an AI model scheduling method is provided, which is applied to the AI ​​capability application module at the deployment site and includes: Receive the AI ​​model scheduling service process initiated by the AI ​​capability management module at the deployment site to request the AI ​​capability application module at the deployment site to allocate core network tunnel information; Allocating fourth core network tunnel information; Return the fourth core network tunnel information to the AI ​​capability management module at the deployment site, so that the AI ​​capability management module at the deployment site carries the fourth core network tunnel information, the second core network tunnel information, and the regional information in the AI ​​model scheduling service request and sends it to the service transmission channel control plane network element at the deployment site, where the second core network tunnel information refers to the core network tunnel information at the user plane network element end of the visited service transmission channel; Use the intelligent computing service identifier and intelligent computing model identifier to select the intelligent computing model, and pass the selected intelligent computing model copy to the visited AI capability application module through the established user-side intelligent computing service model transmission channel.

[0029] In a tenth aspect, an AI model scheduling method is provided, which is applied to a user-plane network element of a visited service transmission channel, and when performing same-area scheduling, includes: Receive a user plane tunnel establishment request sent by a visited service transport channel control plane network element, where the request carries first core network tunnel information and fourth core network tunnel information, where the first core network tunnel information refers to the core network tunnel information of the visited AI application module, and the fourth core network tunnel information refers to the core network tunnel information of the deployed AI application module; Allocating second core network tunnel information; Sending the second core network tunnel information to the visited service transmission channel control plane network element; Forward the intelligent computing model copy sent by the AI ​​capability module at the deployment site to the AI ​​capability application module at the visited site.

[0030] In an eleventh aspect, an AI model scheduling method is provided, which is applied to a user-plane network element of a visited service transmission channel, and when performing cross-region scheduling, includes: receiving a user plane tunnel establishment request sent by a visited service transport channel control plane network element, where the request carries first core network tunnel information, wherein the first core network tunnel information refers to the core network tunnel information of the visited AI application module; Allocating second core network tunnel information; Sending the second core network tunnel information to the visited service transmission channel control plane network element; Receiving the third core network tunnel information sent by the control plane network element of the visited service transmission channel, wherein the third core network tunnel information refers to the core network tunnel information of the user plane network element end of the deployed service transmission channel; Forward the intelligent computing model copy sent by the user-plane network element of the service transmission channel at the deployment site to the AI ​​capability application module at the visited site.

[0031] In a twelfth aspect, an AI model scheduling method is provided, which is applied to a user-plane network element of a service transmission channel at a deployment location, and when performing intra-area scheduling, includes: Receive a user plane tunnel establishment request from a control plane network element of a visited service transmission channel, where the request carries information about a fourth core network tunnel. The fourth core network tunnel information refers to the core network tunnel information of the AI ​​capability application module at the deployment site. Allocate third core network tunnel information; Sending the third core network tunnel information to the visited service transmission channel control plane network element; receiving second core network tunnel information sent by a control plane network element of a visited service transmission channel, wherein the second core network tunnel information refers to core network tunnel information at a user plane network element end of the visited service transmission channel; Forward the intelligent computing model copy sent by the AI ​​capability application module at the deployment site to the AI ​​capability application module at the visited site.

[0032] In a thirteenth aspect, an AI model scheduling method is provided, which is applied to a user-plane network element of a service transmission channel at a deployment location, and when performing cross-region scheduling, includes: Receive a user plane tunnel establishment request sent by a control plane network element of a service transport channel at a deployment site, where the request carries information about a fourth core network tunnel and information about a second core network tunnel. The second core network tunnel information is information about a core network tunnel at the user plane network element end of the visited service transport channel, and the fourth core network tunnel information is information about a core network tunnel at the AI ​​capability application module end of the deployment site. Allocate third core network tunnel information; Return the third core network tunnel information to the service transmission channel control plane network element at the deployment site; Forward the intelligent computing model copy sent by the AI ​​capability application module at the deployed location to the control plane network element of the service transmission channel at the visited location.

[0033] In a fourteenth aspect, an AI model scheduling method is provided, which is applied to a service transmission channel control plane network element at a deployment site, and when performing cross-region scheduling, includes: Receive an AI model scheduling service request sent by the AI ​​capability management module at the deployment site. The request carries the fourth core network tunnel information, the second core network tunnel information, and the area information responsible for the AI ​​capability application module at the deployment site. The second core network tunnel information is the core network tunnel information of the user plane network element end of the visited service transmission channel, and the fourth core network tunnel information is the core network tunnel information of the AI ​​capability application module end at the deployment site. According to the area information of the deployed AI capability application module, the module addresses the user plane network element of the deployed service transmission channel and initiates a user plane tunnel establishment request to the user plane network element of the deployed service transmission channel. The request carries the fourth core network tunnel information and the second core network tunnel information. Receiving the third core network tunnel information returned by the user plane network element of the service transmission channel at the deployment site; Return the third core network tunnel information to the AI ​​capability management module at the deployment site.

[0034] In a fifteenth aspect, an AI model scheduling method is provided, which is applied to an AI capability management module at a deployment site, including: Receive an AI model scheduling service request sent by the visited location AI capability management module, where the request carries the second core network tunnel information, where the second core network tunnel information is the core network tunnel information of the user plane network element end of the visited location service transmission channel; Initiate an AI model scheduling service request to the AI ​​capability application module at the deployment site to allocate the core network tunnel information of the AI ​​capability application module at the deployment site. The request carries the intelligent computing service identifier and the intelligent computing model identifier. Receiving fourth core network tunnel information returned by the deployed AI capability application module, wherein the fourth core network tunnel information is core network tunnel information of the deployed AI capability application module end; Send an AI model scheduling service request to the service transmission channel control plane network element at the deployment site. The request carries the fourth core network tunnel information, the second core network tunnel information, and the area information responsible for the deployment site's AI capability application module. receiving third core network tunnel information sent by a control plane network element of a deployed service transmission channel, wherein the third core network tunnel information is core network tunnel information of a user plane network element end of the deployed service transmission channel; Return the AI ​​model scheduling service establishment request result to the visited AI capability management module, which includes: the third core network tunnel information.

[0035] In a sixteenth aspect, a visit location AI capability management module includes: A first receiving module is used to receive location information, an intelligent computing service identifier, and an intelligent computing model identifier; A first sending module is configured to use the stored intelligent computing model registration table to perform intra-area scheduling when it is determined that there is an AI capability application module that matches both the intelligent computing service identifier and the intelligent computing model identifier and the area information corresponding to the matching AI capability application module does not match the received location information, and to initiate an AI model scheduling service request to the visited location AI capability application module to request allocation of core network tunnel information on the visited location AI capability application module side, wherein the intelligent computing model registration table stores a correspondence between the intelligent computing service identifier, the intelligent computing model identifier, the area information, and the AI ​​capability application module information; A second receiving module is configured to receive the first core network tunnel information returned by the visited location AI capability application module; The second sending module is used to initiate an AI model scheduling service request to the AI ​​capability application module at the deployment site pointed to by the matching AI capability application module information, requesting the allocation of the core network tunnel information on the AI ​​capability application module side at the deployment site, and the request carries the intelligent computing service identifier and the intelligent computing model identifier; A third receiving module is configured to receive fourth core network tunnel information returned by the deployed AI capability application module; The third sending module is configured to initiate an AI model scheduling service request to the control plane network element of the visited service transmission channel, where the request carries the first core network tunnel information, the fourth core network tunnel information, the location information, and the regional information of the deployed AI capability application module; The forwarding module forwards the second core network tunnel information sent by the control plane network element of the visited service transmission channel to the visited AI capability application module and the deployed AI capability application module when it is determined that the user plane network element of the visited service transmission channel and the user plane network element of the deployed service transmission channel are the same network element, wherein the second core network tunnel information is the core network tunnel information at the user plane network element end of the visited service transmission channel.

[0036] In a seventeenth aspect, a visiting location AI capability application module is provided, including: A first receiving module is configured to receive an AI model scheduling service request initiated by a visited location AI capability management module; an allocation module, configured to allocate first core network tunnel information; A sending module is used to return the information to the AI ​​capability management module at the visited location; The second receiving module is used to receive the second core network tunnel information sent by the visited AI capability management module; wherein the second core network tunnel information is the core network tunnel information of the user plane network element end of the visited service transmission channel The third receiving module is used to receive a copy of the intelligent computing model sent by the AI ​​capability application module at the deployment site.

[0037] In an eighteenth aspect, a control plane network element for a visited service transmission channel is provided, which, when scheduling in the same area, includes: A first receiving module is configured to receive, during intra-region scheduling, an AI model scheduling service request initiated by the visited-site AI capability management module, the request carrying first core network tunnel information, fourth core network tunnel information, location information, and regional information of the deployed-site AI capability application module, wherein the first core network tunnel information refers to the core network tunnel information of the visited-site AI application module, and the fourth core network tunnel information refers to the core network tunnel information of the deployed-site AI application module; An addressing module is used to address the user plane network element of the service transmission channel at the visited location using the location information carried in the AI ​​model scheduling service request, and to address the user plane network element of the service transmission channel at the deployed location using the area information; A first sending module is configured to send a user plane tunnel establishment request carrying the first core network tunnel information and the fourth core network tunnel information to the user plane network element of the visited service transmission channel when it is determined that the user plane network element of the visited service transmission channel and the user plane network element of the deployed service transmission channel are the same network element; The second receiving module is configured to receive the second core network tunnel information sent by the user plane network element of the visited service transmission channel; The second sending module is configured to send the second core network tunnel information to the visited site AI capability application module and the deployed site AI capability application module via the visited site AI capability management module.

[0038] In a nineteenth aspect, a visited service transmission channel control plane network element is provided, which includes: a third receiving module, configured to receive an AI model scheduling service request initiated by the visited-location AI capability management module, the request carrying first core network tunnel information and location information; wherein the first core network tunnel information refers to the core network tunnel information of the visited-location AI application module; An addressing module, used for addressing a user plane network element of a visited service transmission channel according to location information; A third sending module is configured to send a user plane tunnel establishment request to a user plane network element of a visited service transmission channel, wherein the request carries the first core network tunnel information; A fourth receiving module is configured to receive the second core network tunnel information fed back by the user plane network element of the user's visited site service transmission channel, and transmit the information to the visited site AI capability application module via the visited site AI capability management module; A fifth receiving module is configured to receive the third core network tunnel information returned by the visited AI capability management module, wherein the third core network tunnel information is the core network tunnel information of the user plane network element end of the deployed service transmission channel; The fourth sending module is configured to return the third core network tunnel information to the user plane network element of the visited service transmission channel.

[0039] In a twentieth aspect, a deployment-site AI capability application module is provided, which, when performing intra-area scheduling, includes: A first receiving module is configured to receive an AI model scheduling service request initiated by the visited AI capability management module to allocate core network tunnel information of the deployed AI capability application module, wherein the request carries an intelligent computing service identifier and an intelligent computing model identifier; A first allocation module, configured to allocate fourth core network tunnel information; A first sending module, configured to return the fourth core network tunnel information to the visited location AI capability management module; A second receiving module is configured to receive, when the visited site service transmission channel control plane network element determines that the visited site service transmission channel user plane network element and the deployed site service transmission channel user plane network element are the same network element, the second core network tunnel information sent by the visited site AI capability management module, wherein the second core network tunnel information is the core network tunnel information of the visited site service transmission channel user plane network element end; A selection module is used to select an intelligent computing model using an intelligent computing service identifier and an intelligent computing model identifier; The second sending module transmits the selected intelligent computing model copy to the visited AI capability application module.

[0040] In a twenty-first aspect, a deployment-site AI capability application module is provided, which includes: A third receiving module is configured to receive an AI model scheduling service process initiated by the AI ​​capability management module at the deployment site to request the AI ​​capability application module at the deployment site to allocate core network tunnel information; A second allocation module, configured to allocate fourth core network tunnel information; A third sending module is configured to return the fourth core network tunnel information to the AI ​​capability management module at the deployment site, so that the AI ​​capability management module at the deployment site carries the fourth core network tunnel information, the second core network tunnel information, and the regional information in the AI ​​model scheduling service request and sends it to the service transmission channel control plane network element at the deployment site, wherein the second core network tunnel information refers to the core network tunnel information of the user plane network element end of the visited service transmission channel; The fourth sending module is used to transmit the selected intelligent computing model copy to the visited AI capability application module through the established user-side intelligent computing business model transmission channel.

[0041] In a twenty-second aspect, a user-plane network element for a visited service transmission channel, when performing intra-area scheduling, includes: a first receiving module, configured to receive a user plane tunnel establishment request sent by a visited-site service transport channel control plane network element, the request carrying first core network tunnel information and fourth core network tunnel information, wherein the first core network tunnel information refers to the core network tunnel information of the visited-site AI application module, and the fourth core network tunnel information refers to the core network tunnel information of the deployed-site AI application module; A first allocation module, configured to allocate second core network tunnel information; The first sending module is configured to send the second core network tunnel information to a control plane network element of a visited service transmission channel.

[0042] In a twenty-third aspect, a user-plane network element for a visited service transmission channel, when performing cross-region scheduling, includes: a second receiving module, configured to receive a user plane tunnel establishment request sent by a visited service transport channel control plane network element, the request carrying first core network tunnel information, wherein the first core network tunnel information refers to the core network tunnel information of the visited AI application module; A second allocation module, configured to allocate second core network tunnel information; A second sending module sends the second core network tunnel information to the visited service transmission channel control plane network element; A third receiving module is configured to receive third core network tunnel information sent by a control plane network element of a visited service transmission channel, wherein the third core network tunnel information refers to core network tunnel information of a user plane network element end of a deployed service transmission channel; The third sending module forwards the intelligent computing model copy sent by the user-plane network element of the deployed service transmission channel to the AI ​​capability application module at the visited location.

[0043] In a twenty-fourth aspect, a user-plane network element for a deployed service transmission channel is provided, which, when performing intra-area scheduling, includes: A first receiving module is configured to receive a user plane tunnel establishment request sent by a visited service transmission channel control plane network element, the request carrying fourth core network tunnel information, wherein the fourth core network tunnel information refers to the core network tunnel information of the AI ​​capability application module end at the deployment site; A distribution module, configured to distribute third core network tunnel information; A first sending module, configured to send the third core network tunnel information to a visited service transmission channel control plane network element; A second receiving module is configured to receive second core network tunnel information sent by a control plane network element of a visited service transmission channel, wherein the second core network tunnel information refers to core network tunnel information at a user plane network element end of the visited service transmission channel; The second sending module is used to forward the intelligent computing model copy sent by the AI ​​capability application module at the deployment site to the AI ​​capability application module at the visited site.

[0044] In a twenty-fifth aspect, a service transmission channel control plane network element is provided, which, when performing cross-region scheduling, includes: A first receiving module is configured to receive an AI model scheduling service request sent by an AI capability management module at a deployment site, the request carrying information about a fourth core network tunnel, information about a second core network tunnel, and information about an area under the responsibility of an AI capability application module at a deployment site, wherein the second core network tunnel information is information about a core network tunnel at a user plane network element end of a visited site service transmission channel, and the fourth core network tunnel information is information about a core network tunnel at a deployment site AI capability application module end; An addressing module is used to address the user plane network element of the deployed service transmission channel according to the area information of the deployed AI capability application module; The first sending module is configured to initiate a user plane tunnel establishment request to a user plane network element of a deployed service transmission channel, wherein the request carries information about the fourth core network tunnel and information about the second core network tunnel; The second receiving module is used to receive the third core network tunnel information returned by the user plane network element of the service transmission channel at the deployment site; The second sending module is used to return the third core network tunnel information to the AI ​​capability management module at the deployment site.

[0045] In aspect 26, a deployment AI capability management module is provided, including: a first receiving module, configured to receive an AI model scheduling service request sent by a visited location AI capability management module, the request carrying second core network tunnel information, wherein the second core network tunnel information is core network tunnel information of a user plane network element end of a visited location service transmission channel; The first sending module is used to initiate an AI model scheduling service request for allocating core network tunnel information of the AI ​​capability application module at the deployment site to the AI ​​capability application module at the deployment site, wherein the request carries an intelligent computing service identifier and an intelligent computing model identifier; a second receiving module, configured to receive fourth core network tunnel information returned by the deployed AI capability application module, wherein the fourth core network tunnel information is core network tunnel information of the deployed AI capability application module; The second sending module is used to send an AI model scheduling service request to the service transmission channel control plane network element of the deployment site, where the request carries the fourth core network tunnel information, the second core network tunnel information, and the area information responsible for the AI ​​capability application module of the deployment site; A third receiving module is configured to receive third core network tunnel information sent by a control plane network element of a deployed service transmission channel, wherein the third core network tunnel information is core network tunnel information of a user plane network element end of the deployed service transmission channel; The third sending module is used to return the AI ​​model scheduling service request establishment result to the visited AI capability management module, where the result includes: third core network tunnel information.

[0046] In aspect 27, an AI model scheduling system is provided, which, when performing same-area scheduling, includes: the visited site AI capability management module described in aspect 16, the visited site AI capability application module described in aspect 17, the visited site service transmission channel control plane network element described in aspect 18, the deployed site AI capability application module described in aspect 20, the visited site service transmission channel user plane network element described in aspect 22, and the deployed site service transmission channel user plane network element described in aspect 24; During cross-region scheduling, it includes: the visited AI capability management module described in the sixteenth aspect, the visited AI capability application module described in the seventeenth aspect, the visited service transmission channel control plane network element described in the nineteenth aspect, the visited service transmission channel user plane network element described in the twenty-third aspect, the deployed service transmission channel control plane network element described in the twenty-fifth aspect, the deployed AI capability application module described in the twenty-first aspect, and the deployed AI capability management module described in the twenty-sixth aspect.

[0047] In a twenty-eighth aspect, an electronic device is provided, comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus; Memory for storing computer programs; The processor is used to implement the above method when executing the program stored in the memory.

[0048] In aspect 29, a computer storage medium is provided, wherein a computer program is stored in the computer storage medium, and the computer program implements the above method when executed by a processor.

[0049] Compared with the prior art, the present disclosure has the following advantages: This disclosure provides AI capabilities to mobile terminal users in the form of services, integrating AI service capabilities into mobile communication networks, and delivering AI computing power to users from the core network. Compared to deploying AI in the data network (DN), computing power is closer to users, significantly shortening service processes. Compared to mobile terminals carrying AI computing power, moving intelligent computing power to the edge network eliminates terminal computing, reduces terminal power consumption, and enables light-load operation. Compared to MEC carrying AI computing power, moving intelligent computing power to the core network allows the core network to implement functions such as mobility management and service expansion. This solves problems such as high computing latency, poor versatility, and difficult mobility management.

[0050] This disclosure proposes a mobile communications core network that provides AI services for mobile terminal users. It also provides a method for establishing a service transmission channel and a model scheduling method. The service transmission channel establishment method establishes a service transmission channel, which can then be used to provide AI services to user terminals. The model scheduling method transfers copies of intelligent computing models from the deployment site to the AI ​​capability application module at the visited site, providing support for intelligent computing services.

[0051] Other features and advantages of the present disclosure will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present disclosure. The purposes and other advantages of the present disclosure can be realized and obtained by the structures indicated in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0053] Figure 1 A schematic diagram of an edge intelligent computing network architecture for mobile communication network users according to an embodiment of the present disclosure is shown; Figure 2 A schematic diagram of a business process for a user to use an intelligent computing service according to an embodiment of the present disclosure is shown; Figure 3 A schematic diagram of a network topology of an AI model scheduling business process according to an embodiment of the present disclosure is shown; Figure 4A flowchart of the AI ​​model same-zone scheduling service according to an embodiment of the present disclosure is shown; Figure 5 A diagram illustrating interactions between network elements for a same-zone scheduling service using an AI model according to an embodiment of the present disclosure is shown; Figure 6 A diagram illustrating interactions between network elements for cross-region scheduling services using an AI model according to an embodiment of the present disclosure is shown; Figure 7 Another diagram illustrating interactions between network elements for cross-region scheduling services using an AI model according to an embodiment of the present disclosure is shown; Figure 8 A flowchart of the AI ​​model cross-region scheduling business process according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present disclosure more clear, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present disclosure without making any creative efforts shall fall within the scope of protection of the present disclosure.

[0055] Figure 1 FIG2 shows a schematic diagram of an edge computing network architecture for mobile communication network users according to an embodiment of the present disclosure. Figure 1 As shown, an edge intelligent computing network architecture for mobile communication network users in an embodiment of the present disclosure includes: an AI capability application module, an AI capability management module, a service transmission channel control plane network element, a service transmission channel user plane network element, a user data network element, a user PCC (Policy Control and Charging) policy control network element, and an access and mobility management network element. The service transmission channel user plane network element is connected to an external data network (DN), which has an application function (AF).

[0056] The edge intelligent computing network architecture disclosed in this disclosure can be based on the existing 5G core network (5GC) or 4G core network (EPC) network architecture, adding an AI capability application module and an AI capability management module, and updating the functions of the service transmission channel control plane network element and the service transmission channel user plane network element. This disclosure provides a process for introducing an intelligent computing model into a mobile communication network, a service / signaling process for establishing an intelligent computing service data transmission channel for mobile communication user UEs, and a service / signaling process for completing intelligent computing model scheduling. These processes involve modifications to the interfaces between the aforementioned network elements. Figure 1The dashed lines represent signaling interactions between control plane channels, and the solid lines represent service interactions on user plane channels. The purple solid line represents the established user plane transport channel for intelligent computing services, while the purple dashed line represents signaling interactions between the AI ​​capability application module and the AI ​​capability management module, as well as between the AI ​​capability management module and the control plane network elements of the service transport channel.

[0057] The service transmission channel control plane network element disclosed in the present invention may be an SMF (Session Management Function) in a 5G core network; and may be an SGW-C / SGW (Serving Gateway-Control / Serving Gateway) in a 4G core network.

[0058] The user plane network element of the service transmission channel disclosed in the present invention may be a UPF in a 5G core network; in a 4G core network, it may be an SGW-U / SGW (Serving Gateway-User / Serving Gateway, serving gateway-control plane / serving gateway).

[0059] The user data network element disclosed in the present invention may be a UDM (Unified Data Management) in a 5G core network; and may be an HSS (Home Subscriber Server) in a 4G core network.

[0060] The user PCC policy control network element disclosed in the present invention may be a PCF (Policy Control Function) in a 5G core network and a PCRF (Policy and Charging Control Function) in a 4G core network.

[0061] in: (1) The AI ​​capability application module is used to carry intelligent computing power, provide users with a variety of intelligent computing models, and create core network tunnels based on the request of the AI ​​capability management module during the establishment of intelligent computing service transmission channels, which serve as the starting point or end point of the tunnel to complete the corresponding business process.

[0062] Among them, multiple intelligent calculation models are provided to users, including general intelligent calculation models and / or intelligent calculation models for user-specific contracts. The intelligent calculation models for user-specific contracts can be divided into user-initiated contracts for enterprise consumers (ToC) and user-initiated contracts for enterprise businesses (ToB).

[0063] AI capability application modules can be deployed in a hierarchical or distributed manner; they can be logically integrated with existing network elements (such as UPF, wireless access network elements, and NWDAF); one AI capability application module can correspond to one or more wireless access network elements, and one AI capability application module can correspond to multiple service transmission channel user plane elements.

[0064] (2) The AI ​​capability management module is used to complete the management and scheduling of intelligent computing models and computing power. When a new intelligent computing model is added to the mobile network, the intelligent computing model is deployed to the AI ​​capability application module by interacting with the service transmission channel control plane network element, the service transmission channel user plane network element, etc.; after the user accesses the mobile network, the AI ​​capability application module is addressed to provide intelligent computing services. When the AI ​​capability management module does not deploy the corresponding intelligent computing model locally, it completes the model scheduling and updating by interacting with the service transmission channel control plane network element and the AI ​​capability application module; cleans up the intelligent computing models that have not been used for a long time in the AI ​​capability application module; and periodically checks and updates the AI ​​capability management information. It can be logically combined with existing network elements (such as the service transmission channel control plane network element).

[0065] In the present disclosure, one AI capability management module may correspond to one or more AI capability application modules, and one or more AI capability management modules may correspond to one service transmission channel control plane network element.

[0066] The AI ​​capability management module includes an intelligent computing service identifier, an intelligent computing model identifier, and information about the AI ​​capability application module that deploys the intelligent computing model represented by the intelligent computing model identifier. The AI ​​capability application module information includes: the AI ​​capability application module identifier and / or the AI ​​capability application module address, as well as information about the region served by the AI ​​capability application module. The information included in the AI ​​capability management module can be specifically described by the intelligent computing model registration table as shown in Table (1) below.

[0067] Table (1) Intelligent Computing Model Registration Form (Example)

[0068] The above-mentioned intelligent computing service identifier can be S-NSSAI+DNN used in 5G networks and APN used in 4G networks, which is used to distinguish intelligent computing services from other service types (such as data services and voice services).

[0069] The intelligent computing model identifier is used to uniquely identify intelligent computing models under the same intelligent computing service identifier. Intelligent computing models are divided into general intelligent computing models and user-specific intelligent computing models. The same intelligent computing model can be stored in different AI capability application modules, and the same AI capability application module can store different intelligent computing models.

[0070] (3) The service transmission channel control plane network element is used to select the AI ​​capability management module based on the intelligent computing service identifier. It cooperates with the AI ​​capability management module, the service transmission channel user plane network element, and the user data network element to complete the intelligent computing model scheduling business process.

[0071] (4) The service transmission channel user plane network element is used to establish the service transmission channel for the AI ​​model scheduling service together with the AI ​​capability application module based on the request of the service transmission channel control plane network element in the AI ​​model scheduling service process.

[0072] (5) User data network element, used for signing contracts and storing smart computing service contract data. Smart computing service contract data includes: smart computing service identifier and smart computing model identifier.

[0073] (6) User PCC policy control network element, used for signing contracts and storing smart computing related policy control information. The relevant policy control information is used to provide users with a differentiated smart computing service experience.

[0074] The following describes the business process of this disclosure: 1. Introducing intelligent computing models into business processes 1) User contracts for intelligent computing services When a user signs up for the intelligent computing service at the business hall, the user data network element signs the user's intelligent computing service identifier and intelligent computing model identifier. At the same time, the user PCC policy data network element signs the user's intelligent computing service-related policy control information.

[0075] 2) Local deployment of the contract model If the intelligent computing model you subscribe to is not in the local AI capability application module, you can deploy it in the following ways: 1. Locally upload the corresponding intelligent computing model to the AI ​​capability application module; 2. If the intelligent computing model is in an AI capability application module in another region, the management plane can be used to control the user plane to transfer a copy of the intelligent computing model to the AI ​​capability application module in the user's contracted location. The AI ​​capability application module initiates the AI ​​model scheduling service process. For details, see the cross-region intelligent computing model scheduling in the AI ​​model scheduling service process described in Process 3. The user plane model transfer process is as follows: AI capability application module in another region - user plane network element of the service transmission channel in another region - user plane network element of the service transmission channel in the contracted location - AI capability application module in the contracted location.

[0076] 3) Model update When the intelligent computing model is iteratively updated, the AI ​​capability application module cooperates to complete the global intelligent computing model update.

[0077] 4) Model registration After the intelligent computing model is introduced, scheduled, or changed, the AI ​​capability management module registers the intelligent computing model related information.

[0078] 2. Business process for establishing user intelligent computing business channels The establishment of a user intelligent computing service session first requires interaction with the network. Specifically, the user-side intelligent computing service transmission channel is established through core network control plane signaling interaction. Subsequent users use the established intelligent computing service channel to transmit intelligent computing service data.

[0079] The essence of establishing an intelligent computing service channel is that the core network selects the AI ​​capability application module used for this service based on user-related information and establishes a core network user-plane service transmission channel. The establishment of a service transmission channel includes: the service transmission channel control plane network element allocates an IP address to the UE, the AI ​​capability application module allocates core network tunnel information for the user's intelligent computing session, and the wireless access network network element allocates access network tunnel information for the user's intelligent computing session. The UE needs to know the IP address assigned to it, the wireless access network network element needs to know the core network tunnel information, and the AI ​​capability application module needs to know the access network tunnel information. At the same time, the core network also needs to complete related processes such as authentication and authorization of the user's intelligent computing service, association of intelligent computing session management policies, and registration of user data network element sessions.

[0080] The following steps are based on the essence of the above and describe the business process of establishing the intelligent computing service channel in combination with the user's intelligent computing service contract, such as Figure 2 As shown, the following steps are included: Step 0: The user equipment (UE) accesses the network and completes registration.

[0081] Specifically, the UE accesses the network, and the access and mobility management network element in the network interacts with related network elements through signaling to complete mobility management such as two-way authentication and location update between the network and the UE, and the UE enters the registered state.

[0082] During the UE registration process, the access and mobility management network element obtains the user identity, user location information and user policy control network element identification information.

[0083] The user identifier may be a SUPI / IMSI (Subscription Permanent Identifier / International Mobile Subscriber Identity) or a GPSI / MSISDN (Generic Public Subscription Identifier / Mobile Subscriber ISDN Number). In the 5G core network, the user policy control element is the PCF, so the user policy control element identifier is the PCF ID. In the 4G core network, the user policy control element is the PCRF, so the user policy control element identifier is the PCRF ID.

[0084] The above process may specifically follow the existing 5G standards, 4G standards and the standards of the new generation of mobile communications.

[0085] Step 1: The UE sends an intelligent computing service channel establishment request to the access and mobility management network element via the wireless access network network element; the access and mobility management network element determines the service transmission channel control plane network element based on the intelligent computing service identifier, and sends the user identifier, user location information, user PCC policy control network element identifier, intelligent computing service identifier and intelligent computing model identifier to the determined service transmission channel control plane network element.

[0086] In this step 1, when the user needs to use the intelligent computing service, the user sends an intelligent computing service channel establishment request to the network through the UE.

[0087] The intelligent computing service channel establishment request carries an intelligent computing service identifier and an intelligent computing model identifier.

[0088] Specifically, the intelligent computing service identifier here can be represented by S-NSSAI+DNN in the 5G network and by APN in the 4G network.

[0089] The access and mobility management network element may determine the service transmission channel control plane network element in any of the following three ways based on the intelligent computing service identifier: Method 1: When the access and mobility management network element locally configures a matching list of the intelligent computing service identifier to the identifier and address of the service transmission channel control plane network element, the access and mobility management network element can search the matching list for the identifier and address of the service transmission channel control plane network element that matches the intelligent computing service identifier carried in the intelligent computing service channel establishment request; Method 2: During or before the UE registration process in step S1, the service transmission channel control plane network element sends a registration message to the network resource database (e.g., NRF) carrying the intelligent computing service identifier for which it is responsible and its own address information, completing registration in the network resource database. At this point, the access and mobility management network element executes a service discovery request for the service transmission channel control plane network element to the network resource database, the request containing the intelligent computing service identifier. The network resource database uses the intelligent computing service identifier contained in the discovery request to find a matching transmission channel control plane network element identifier and address from the stored service transmission channel control plane network element registration information. The network resource database returns a service discovery result carrying a matching transmission channel control plane network element identifier and address to the access and mobility management network element. The access and mobility management network element can then select the service transmission channel control plane network element according to the transmission channel control plane network element identifier and address in the service discovery result.

[0090] Method 3: A matching list of the intelligent computing service identifier and the identifier and address of the service transmission channel control plane network element is pre-configured in the network resource database (for example, DNS (Domain Name System)); the access and mobility management network element initiates a query request to the network resource database, and the query request includes the intelligent computing service identifier; the network resource database uses the intelligent computing service identifier in the query request to search for the matching identifier and address of the transmission channel control plane network element in the configured matching list, and returns it to the access and mobility management network element in the query result; the access and mobility management network element then selects the service transmission channel control plane network element based on the query result returned by the network resource database.

[0091] Step 2: The service transport channel control plane network element interacts with the user PCC policy data network element based on the user identifier to obtain the user session management subscription data. Using this user session management subscription data, the network element determines whether to accept or reject the UE's intelligent service channel establishment request. If accepted, the process proceeds to step 3. If rejected, the process ends.

[0092] Specifically, in this step 2, the service transmission channel control plane network element determines whether the user session management subscription data corresponding to the user identifier has been stored locally based on the user identifier; if so, it accepts or rejects the UE's intelligent computing service channel establishment request based on the UE's user session management subscription data; if not, it accesses the user data network element (such as UDM) to obtain the user session management subscription data therefrom, and accepts or rejects the UE's intelligent computing service channel establishment request based on the UE's user session management subscription data. The service transmission channel control plane network element accepts the UE's intelligent computing service notification establishment request when the user represented by the user identifier has signed up for an intelligent computing session; otherwise, it rejects the UE's intelligent computing service channel establishment request.

[0093] It should be noted that if there is a scenario where the AI ​​capability application module is selected based on policy information, then in this step 2, the service transmission channel control plane network element also determines whether it is necessary to select the AI ​​capability application module based on the policy information based on the user session management contract data; if necessary, the user PCC policy control network element identifier is used to access the user PCC policy control network element, and policy information containing QoS (Quality of Service) information is obtained from the user PCC policy control network element. The policy information is used to establish an intelligent computing session management policy association, and the policy is associated with the intelligent computing session management. After completing the AI ​​capability application module selection and user IP address allocation, the intelligent computing session management policy association modification in step 4 is executed, and the IP address allocated to the user is also sent to the user PCC policy control network element; if not required, step 3 is executed.

[0094] Here, the PCC policy control network element needs to record the user IP address and other information (such as IP Domain) to distinguish unique sessions. If subsequent services trigger policies, the PCC policy control network element will select the corresponding session and issue a policy based on the saved user IP address and other information.

[0095] The above-mentioned user PCC policy control network element stores intelligent computing session-related policy information (including QoS information), intelligent computing policy control rule information, and intelligent computing policy control request trigger information.

[0096] Step 3: The service transmission channel control plane network element allocates an IP address to the UE and selects an intelligent computing capability management module; transmits the intelligent computing service identifier, intelligent computing model identifier, policy information (if the policy network element is accessed in step 2), user session management subscription data and user location information to the selected AI capability management module, and requests the AI ​​capability management module to select an AI capability application module; the AI ​​capability management module selects an AI capability application module based on the intelligent computing service identifier, intelligent computing model identifier, user session management subscription data, policy information and user location information, and sends the selected AI capability application module information to the service transmission channel control plane network element.

[0097] In this step 3, the service transmission channel control plane network element can be co-located with the AI ​​capability management module, or can have the ability to address the AI ​​capability management module.

[0098] When the service transmission channel control plane network element and the AI ​​capability management module are jointly provided, in this step 3, the service transmission channel control plane network element transparently transmits the intelligent computing service identifier, intelligent computing model identifier, user session management subscription data, and user location information to the AI ​​capability management module; When the service transmission channel control plane network element and the AI ​​capability management module are not installed together, the service transmission channel control plane network element addresses the AI ​​capability management module according to the intelligent computing service identifier, and transmits the intelligent computing service identifier, intelligent computing model identifier, user session management contract data and user location information to the AI ​​capability management module.

[0099] The AI ​​capability management module selects the AI ​​capability application module according to the intelligent computing service identifier, intelligent computing model identifier, user session management contract data, policy information and user location information, specifically including: the AI ​​capability management module filters out the AI ​​capability application module information that matches the intelligent computing service identifier, intelligent computing model identifier and user location information transmitted by the service transmission channel control plane network element from the stored intelligent computing model registration table, and then further filters out the AI ​​capability application module based on the service load and capacity of the user session management contract data and the filtered AI capability application module information, as well as the policy information, wherein the intelligent computing model registration table stores the correspondence between the intelligent computing service identifier, intelligent computing model identifier, regional information and AI capability application module information, wherein the AI ​​capability application module information includes: AI application module identifier and / or AI application module address.

[0100] For example, if a user is located in a residential complex in Chaoyang District, Beijing, and has a core network location ID of TAI1, and requests an intelligent computing service with an ID of 1 and an intelligent computing model ID of 1, the AI ​​capability management module selects AI capability application modules 1, 2, 3, 4, and 5 that meet the request. The module also selects the optimal AI capability application module based on the user's user session management subscription data and the service load and capacity of each AI capability application module. The module then returns the identifier or address of the AI ​​capability application module.

[0101] In addition, if the AI ​​capability management module finds no corresponding AI capability application module after the above screening, it will initiate the AI ​​model scheduling service process (see the AI ​​model scheduling service process in process 3 for details) and complete the model management registration. The AI ​​capability management module will then send the selected AI capability application module information to the service transmission channel control plane network element.

[0102] Step 4: If the user PCC policy data NE was not accessed to obtain policy information in step 2, then in this step 4, the service transmission channel control plane NE needs to execute the PCC process of intelligent computing session management policy control. That is, using the user PCC policy control NE identifier, access the user PCC policy control NE, obtain policy information, establish an intelligent computing session management policy association, associate the policy with intelligent computing session management, and send the policy information to the AI ​​capability management module, where the policy information includes QoS information. If the policy data NE was accessed and the policy information was obtained in step 2, then in this step 4, the service transmission channel control plane NE initiates a session management modification to the user PCC policy data NE.

[0103] Step 5: The service transmission channel control plane network element sends user location information, intelligent computing service identifier, and intelligent computing model identifier to the AI ​​capability application module through the AI ​​capability management network element, requesting the AI ​​capability application module to allocate core network tunnel information; the AI ​​capability application module allocates core network tunnel information and reports the core network tunnel information to the service transmission channel control plane network element through the AI ​​capability management module.

[0104] Core network tunnel information contains the address of the interface tunnel used for a specific intelligent computing session, which consists of a TEID (Tunnel Endpoint Identifier) ​​and an IP address.

[0105] Step 6: The service transmission channel control plane network element returns the session identifier, intelligent computing service identifier, IP address allocated to the user, core network tunnel information of the AI ​​capability application module, session establishment acceptance information, and QoS information to the access and mobility management network element. The access and mobility management network element, the wireless access network, and the UE complete the information exchange related to the establishment of the intelligent computing session. The wireless access network network element allocates an access network tunnel for the intelligent computing session; the user terminal sends the accepted or rejected QoS list to the wireless access network network element, and the wireless access network element sends the access network tunnel information and QoS list to the service transmission channel control plane network element via the access and mobility management network element; The access and mobility management network element stores the received information in the associated intelligent computing session context of the UE.

[0106] Specifically, the information interaction related to the establishment of the intelligent computing session includes: the access and mobility management network element forwards the intelligent computing session identifier, core network tunnel information, QoS information, and session establishment acceptance information to the wireless access network network element, and uses the intelligent computing service identifier, IP address allocated to the user, QoS information, and session establishment acceptance information as information in the intelligent computing service channel establishment response, and transmits it to the UE through the wireless access network network element.

[0107] The wireless access network element allocates an access network tunnel for the intelligent computing session Step 7: The service transmission channel control plane network element releases the corresponding QoS profile according to the QoS list rejected by the user terminal, and sends the access network tunnel information to the AI ​​capability application module through the AI ​​capability management module.

[0108] Step 8: The service transmission channel control plane network element initiates a registration process to the user-owned user data network element, and the user-owned user data network element saves the intelligent computing service transmission channel status information.

[0109] Among them, the registration message in the registration process includes: user identification, intelligent computing service identification, session identification and service transmission channel control plane network element identification.

[0110] The user data network element to which the user belongs saves the status information of the intelligent computing service transmission channel, including: user ID, established intelligent computing session ID and its intelligent computing service ID, and service service transmission channel control plane network element ID; and saves it in the UE context in the contract data corresponding to the user ID and the service transmission channel control plane network element data.

[0111] Step 9: The service transmission channel control plane network element, access and mobility management network element, radio access network element, and UE complete the subsequent processes of establishing service transmission.

[0112] Subsequent business processes include: IPv6 address allocation, intelligent computing service transmission channel control plane network elements subscribing to UE mobility events from access and mobility management network elements, etc.

[0113] The above steps 6 to 9 are similar to the existing service transmission channel establishment process for data services. The difference is that the service transmission channel user plane network element is replaced by the AI ​​capability application module in this proposal.

[0114] At this point, the control plane completes the establishment of the service transmission channel. The user plane can transmit intelligent computing service data through the user plane routing (UE-radio access network element-AI capability application module).

[0115] The solution disclosed in this disclosure adds edge computing AI capabilities to the mobile communication network architecture, making AI capabilities part of the network capabilities, just like SMS services, phone services, and Internet services. Mobile users can use intelligent computing models on demand.

[0116] 3. AI Model Scheduling Process In step 0, after completing network registration, the user initiates the intelligent computing service process. The AI ​​capability management module at the visited location fails to match the corresponding AI capability application module address based on the intelligent computing service identifier (for example, S-NSSAI+DNN (Single Network Slice Selection Assistance Information+Data Network Name) used in 5G networks and APN (Access Point Name) used in 4G networks), intelligent computing model identifier, and user location information (such as TAI).

[0117] Case 1: Scheduling intelligent computing models within the same region. The AI ​​capability management module at the visited location can simultaneously match the intelligent computing service identifier (for example, S-NSSAI+DNN used in 5G networks, APN used in 4G networks) and the intelligent computing model identifier. This requires scheduling intelligent computing models across different AI application modules within the same AI capability management module. This means that the intelligent computing model requested by the user is in the AI ​​capability application module managed by the visited location AI capability management module, but not in the AI ​​capability application module at the user's visited location.

[0118] For example, suppose a user has signed up for Intelligent Computing Model 1 for Intelligent Computing Service 1. This Intelligent Computing Model 1 is deployed in the Haidian District AI Capability Application Module, but not in the Chaoyang District AI Capability Application Module. Both the Haidian District AI Capability Application Module and the Chaoyang District AI Capability Application Module are managed by the Beijing AI Capability Management Module. When a user accesses the mobile communication network in Chaoyang District, Beijing, Chaoyang District is the user's location. When the user applies to use Intelligent Computing Model 1 for the signed Intelligent Computing Service, the Chaoyang District AI Capability Application Module does not deploy this Intelligent Computing Model 1. Therefore, the Beijing AI Capability Management Module initiates the model scheduling process and schedules Intelligent Computing Model 1 for Intelligent Computing Service 1 from the Haidian District AI Capability Application Module to the Chaoyang District AI Capability Application Module. Afterwards, the Chaoyang District AI capability application module deployed the intelligent computing model 1 of the intelligent computing service 1. The next time the user accesses the mobile communication network in Chaoyang District, Beijing and uses the intelligent computing model 1, if the model has not been cleared, it can be directly matched to the intelligent computing model 1 of the intelligent computing service 1 in the Chaoyang District AI capability application module, and there is no need to perform AI scheduling services.

[0119] Case 2: Cross-regional intelligent computing model scheduling. If the AI ​​capability management module at the visited location does not simultaneously match the intelligent computing service identifier (for example, S-NSSAI+DNN used in 5G networks, APN used in 4G networks) and the intelligent computing model identifier, scheduling of the intelligent computing model across both AI capability management modules and AI application modules is required. This means that the intelligent computing model requested by the user is in the AI ​​capability application module managed by the AI ​​capability management module in a different region, but not in the AI ​​capability application module at the user's visited location.

[0120] For example: A user accesses the mobile communication network in Chaoyang District, Beijing, and applies to use the intelligent computing model 1 of the contracted intelligent computing service. However, the Chaoyang District AI capability application module has not deployed the intelligent computing model 1. The model is deployed in the Zhengzhou AI capability application module managed by the Henan AI capability management module. Therefore, the Beijing AI capability management module initiates the model scheduling process and schedules the intelligent computing model 1 in the Zhengzhou AI capability application module to the Chaoyang District AI capability application module.

[0121] For case 1, execute steps 1, 2.a, 3.a, 4.a, 5, and 6. Figure 4 and Figure 5 As shown; For case 2, execute steps 1, 2.b, 3.b, 4.b, 5, and 6. Figure 6 、 Figure 7 and Figure 8 shown.

[0122] Figure 5 The example shows the interaction diagram between network elements of the AI ​​model scheduling service in the same area; Figure 6 The example in the figure shows an interaction diagram when the service transmission channel control plane network element in the user's visited area and the service transmission channel control plane network element in the area where the model is located are different network elements; Figure 7 The example in the figure shows an interaction diagram when the service transmission channel control plane network element in the user's visited area and the service transmission channel control plane network element in the area where the model is located are the same network element.

[0123] It should be noted that in the solution disclosed herein, the initiation of the AI ​​model scheduling process is not limited to the triggering of the above-mentioned step 0. For example, when the user makes an appointment to visit a place in advance or knows the user's visiting place in advance based on the relevant information provided by the user, when it is determined that there is no contracted intelligent computing model at the visiting place, the AI ​​model can be triggered to be scheduled from the area where the model is located to the visiting area; based on the relevant statistical data of the intelligent computing model, when it is necessary to promote a certain intelligent computing model that meets the conditions from its location to the visiting place, the AI ​​model can also be triggered to be scheduled from the area where the model is located to the visiting area.

[0124] Scheduling requires the AI ​​capability management module at the location to receive location information, intelligent computing service identifiers, and intelligent computing model identifiers. Using the stored intelligent computing model registration table, if it determines that no AI capability application module information matches the intelligent computing service identifier, intelligent computing model identifier, and location information, it can initiate the corresponding AI model scheduling process based on the specific matching situation. The following steps are explained based on the assumption that the received location information is user location information.

[0125] Step 1: The visited-site AI capability management module initiates an AI model scheduling service process to the user's visited-site AI capability application module and requests the visited-site AI capability application module to allocate core network tunnel information (including the visited-site AI capability application module's address information). The visited-site AI capability application module then returns the first core network tunnel information to the visited-site AI capability management module, i.e., the core network tunnel information allocated by the visited-site AI capability application module. In step 2.a, the visited-site AI capability management module initiates an AI model scheduling service process to the AI ​​capability application module that deploys the intelligent computing model, i.e., the deployment-site AI capability application module, and requests the deployment-site AI capability application module to allocate core network tunnel information (including the address information of the deployment-site AI capability application module), i.e., the fourth core network tunnel information. The deployment-site AI capability application module returns the fourth core network tunnel information, i.e., the core network tunnel information of the deployment-site AI capability application module, to the visited-site AI capability management module. The visited-site AI capability management module initiates an AI model scheduling service request to the control plane network element of the visited-site service transport channel to which it belongs. The request carries the core network tunnel information of the user's visited-site AI capability application module (including the address information of the AI ​​capability application module), the core network tunnel information of the AI ​​capability application module that deploys the intelligent computing model (including the address information of the AI ​​capability application module), the area information (e.g., TAI List) of the AI ​​capability application module that deploys the intelligent computing model, and the user's location information (e.g., TAI (Tracking Area Identity)). It should be noted that the user's visited location is the visited location, the intelligent computing deployment location is the location of the intelligent computing model, the user-plane network element of the service transmission channel of the user's visited location is the user-plane network element of the service transmission channel of the visited location, and the names of other network elements are similar and will not be repeated here.

[0126] In step 3.a, the visited-site service transport channel control plane network element addresses the visited-site service transport channel user plane network element based on the user's location information and sends a user plane tunnel establishment request. The request carries the core network tunnel information of the visited-site AI capability application module (including the address information of the AI ​​capability application module). The visited-site service transport channel user plane network element returns the local core network tunnel information (including the address information of the service transport channel user plane network element), which is also the second core network tunnel information. The visited-site service transport channel control plane network element addresses the intelligent computing deployment site service transport channel user plane network element based on the regional information (e.g., TAI List) where the intelligent computing model's AI capability application module is deployed and sends a user plane tunnel establishment request. The information contains the core network tunnel information of the AI ​​capability application module deployed (including the address information of the AI ​​capability application module), which is also the fourth core network information. The intelligent computing deployment site service transport channel user plane network element returns the local core network tunnel information (including the address information of the service transport channel user plane network element), which is also the third core network tunnel information.

[0127] Figure 3 The same-area scheduling process is shown in area A. When the user plane network element of the service transmission channel at the user's visit location is the same as the user plane network element of the service transmission channel at the intelligent computing deployment location, the established intelligent computing model schedules the user plane routing as follows: Figure 3 As shown in process ①; when the user plane network element of the service transmission channel at the user's visit location is different from the user plane network element of the service transmission channel at the intelligent computing deployment location, the established intelligent computing model schedules the user plane routing as follows Figure 3 As shown in process ②.

[0128] In addition, in a specific implementation, the visited service transmission channel control plane network element may also use the location information carried in the AI ​​model scheduling service request to address the visited service transmission channel user plane network element, and use the area information to address the deployment service transmission channel user plane network element. When it is determined that the visited service transmission channel user plane network element and the deployment service transmission channel user plane network element are the same network element, a user plane tunnel establishment request carrying the first core network tunnel information and the fourth core network tunnel information is sent to the visited service transmission channel user plane network element; the visited service transmission channel user plane network element sends the second core network tunnel information to the visited service transmission channel control plane network element; When it is determined that the user plane network element of the visited service transmission channel and the user plane network element of the deployed service transmission channel are different network elements, a user plane tunnel establishment request carrying the first core network tunnel information is sent to the user plane network element of the visited service transmission channel, and a user plane tunnel establishment request carrying the fourth core network tunnel information is sent to the user plane network element of the deployed service transmission channel; the user plane network element of the visited service transmission channel sends the second core network tunnel information to the control plane network element of the visited service transmission channel.

[0129] In step 4.a, the control plane network element of the visited site service transmission channel sends the core network tunnel information of the user plane network element of the visited site service transmission channel (including the address information of the user plane network element of the service transmission channel) to the AI ​​capability application module of the visited site through the AI ​​capability management network element of the visited site, and sends the core network tunnel information of the user plane network element of the visited site service transmission channel (including the address information of the user plane network element of the service transmission channel) to the user plane network element of the service transmission channel at the intelligent computing deployment site. The control plane network element of the visited site service transmission channel sends the core network tunnel information of the user plane network element of the intelligent computing deployment site (including the address information of the user plane network element of the service transmission channel) to the AI ​​capability application module where the intelligent computing model is deployed through the AI ​​capability management network element, and sends the core network tunnel information of the user plane network element of the intelligent computing deployment site (including the address information of the user plane network element of the service transmission channel) to the user plane network element of the visited site service transmission channel.

[0130] If the user plane network element of the service transmission channel at the user's visited location is the same as the user plane network element of the service transmission channel at the intelligent computing deployment location, this step does not require the completion of information interaction between the user plane network element of the service transmission channel at the user's visited location and the user plane network element of the service transmission channel at the intelligent computing deployment location.

[0131] User plane routing① (such as Figure 3 Process ①): AI capability application module at the user's visit location - user-plane network element of the service transmission channel at the user's visit location (user-plane network element of the service transmission channel at the intelligent computing deployment location) - AI capability application module at the intelligent computing deployment location.

[0132] User plane routing② (such as Figure 3 Process ②): AI capability application module at the user's visit location - user-plane network element of the service transmission channel at the user's visit location - user-plane network element of the service transmission channel at the intelligent computing deployment location - AI capability application module at the intelligent computing deployment location.

[0133] In step 2.b, the visited-site AI capability management module initiates an AI model scheduling service request to the visited-site service transport channel control plane network element. This message includes the core network tunnel information of the visited-site AI capability application module (including the address information of the AI ​​capability application module) and the user's location information (e.g., TAI). The visited-site service transport channel control plane network element, based on the user's location information, addresses the visited-site service transport channel user plane network element and sends a user plane tunnel establishment request. This request includes the core network tunnel information of the visited-site AI capability application module (including the address information of the AI ​​capability application module). The visited-site service transport channel control plane network element receives the core network tunnel information (including the address information of the service transport channel user plane network element) from the visited-site service transport channel user plane network element and synchronizes this information with the visited-site AI capability management module. The visited-site AI capability management module then sends the core network tunnel information (including the address information of the service transport channel user plane network element) from the visited-site service transport channel user plane network element to the visited-site AI capability application module.

[0134] In step 3.b, the AI ​​capability management module at the visited site addresses the AI ​​capability management module in another region based on the intelligent computing service identifier (e.g., S-NSSAI+DNN used in 5G networks, APN used in 4G networks) and the intelligent computing model identifier (see Process 4. AI Capability Management Module Addressing Service Process). The module then sends an AI model scheduling service request to the addressed AI capability management module at the deployment site. The request includes: the intelligent computing service identifier (e.g., S-NSSAI+DNN used in 5G networks, APN used in 4G networks), the intelligent computing model identifier, and the core network tunnel information of the user plane network element of the service transmission channel at the visited site (including the address information of the user plane network element of the service transmission channel). Specifically, sending an AI model scheduling service request to the addressed deployment site AI capability management module includes: the visited site AI capability management module sends the AI ​​model scheduling service request to the deployment site AI capability management module via the visited site service transmission channel control plane network element and the deployment site service transmission channel control plane network element in sequence.

[0135] In step 4.b, after receiving the AI ​​model scheduling service request, the AI ​​capability management module at the intelligent computing deployment site initiates the AI ​​model scheduling service process to the AI ​​capability application module deployed in the intelligent computing model based on the intelligent computing service identifier (for example, S-NSSAI+DNN used in 5G networks, APN used in 4G networks) and the intelligent computing model identifier, and requests the AI ​​capability application module to allocate core network tunnel information (including the address information of the AI ​​capability application module). The AI ​​capability application module returns the allocated core network tunnel information (including the address information of the AI ​​capability application module). The AI ​​capability management module at the intelligent computing deployment site initiates an AI model scheduling service request to the control plane network element of the service transmission channel at the intelligent computing deployment site. The message includes: the core network tunnel information of the AI ​​capability application module at the deployment site (including the address information of the AI ​​capability application module), the regional information of the AI ​​capability application module at the deployment site (for example, TAIList), and the core network tunnel information of the user plane network element of the service transmission channel at the user visit site. The service transmission channel control plane network element at the deployment site of the intelligent computing service transmission channel addresses the user plane network element at the deployment site based on the regional information of the AI ​​capability application module at the deployment site (such as TAI List), and initiates a user plane tunnel establishment request to the user plane network element at the deployment site. The request includes: the core network tunnel information of the AI ​​capability application module at the deployment site (including the address information of the AI ​​capability application module) and the core network tunnel information of the user plane network element of the service transmission channel at the user's visit site. The user plane network element at the deployment site returns the core network tunnel information (including the address information of the AI ​​capability application module) of the user plane network element of the deployment site; the service transmission channel control plane network element at the deployment site returns the core network tunnel information (including the address information of the AI ​​capability application module) of the user plane network element of the deployment site to the AI ​​capability management module at the intelligent computing deployment site. The AI ​​capability management module at the intelligent computing deployment site returns the AI ​​model scheduling service request establishment result (including: core network tunnel information of the user plane network element of the deployment site's service transmission channel) to the AI ​​capability management module at the visited site; the AI ​​capability management module at the visited site returns the core network tunnel information of the user plane network element of the deployment site's service transmission channel to the control plane network element of the visited site's service transmission channel; the control plane network element of the visited site's service transmission channel returns the core network tunnel information of the user plane network element of the deployment site's service transmission channel to the user plane network element of the visited site's service transmission channel.

[0136] Specifically, the AI ​​capability management module at the intelligent computing deployment site returns the AI ​​model scheduling service establishment request result to the AI ​​capability management module at the visited site, including: The AI ​​capability management module at the intelligent computing deployment site returns the AI ​​model scheduling service request establishment result to the AI ​​capability management module at the visited site via the service transmission channel control plane network element at the visited site and the service transmission channel control plane network element at the deployment site.

[0137] Figure 3 The figure shows the inter-region scheduling route between area A (visited site) and area B (deployment site), also known as user plane route ③: visited site AI capability application module - visited site service transmission channel user plane network element - deployment site service transmission channel user plane network element - deployment site AI capability application module.

[0138] In step 5, the AI ​​capability application module at the intelligent computing deployment site selects the corresponding intelligent computing model based on the intelligent computing model identifier and intelligent computing service identifier (for example, S-NSSAI+DNN used in 5G networks and APN used in 4G networks), and sends a copy of the model to the AI ​​capability application module at the user's visited site through the user plane service transmission channel; the AI ​​capability application module at the user's visited site sends a model transfer completion message to the AI ​​capability management module at the user's visited site (intelligent computing service identifier (for example, S-NSSAI+DNN used in 5G networks and APN used in 4G networks), intelligent computing model identifier, regional information (such as TAI List), AI capability application module identifier and / or address information).

[0139] In step 6, the AI ​​capability management module at the user's visit location completes model information registration (intelligent computing service identifier (for example, S-NSSAI+DNN used in 5G networks, APN used in 4G networks), intelligent computing model identifier, regional information (such as TAI List), AI capability application module identifier, and / or address information).

[0140] In the above-mentioned AI model scheduling service, the essence is to use the AI ​​capability management module and the transmission channel control plane network element to establish the user plane channel between the visited AI capability application module, the visited service transmission channel user plane network element, the deployed service transmission channel user plane network element and the deployed AI capability application module. Each network element allocates the core network tunnel information of the local end for model scheduling, and also needs to obtain the core network tunnel information of the adjacent network elements through interaction with other network elements, as shown in the following table (2).

[0141] Table (2) Tunnel information

[0142] 4. AI Capability Management Module Addressing Business Processes There are several ways to do this: Method 1: Locally configure the global intelligent computing model registration information table When the intelligent computing model registration information changes, the AI ​​capability management module notifies other global AI capability management modules of the local intelligent computing model registration information (or the change information). Each AI capability management module maintains a global intelligent computing model registration information table, which includes: AI capability management module identifier, AI capability management module address, intelligent computing service identifier (for example, S-NSSAI+DNN used in 5G networks, APN used in 4G networks), intelligent computing model identifier, AI capability application module identifier, and regional information (such as TAI List).

[0143] Method 2: Service Registration and Service Discovery When the intelligent computing model registration information changes, the AI ​​capability management module completes registration with the network resource database (such as NRF). The registration information includes: AI capability management module identifier, AI capability management module address, intelligent computing service identifier (for example, S-NSSAI+DNN used in 5G networks, APN used in 4G networks), intelligent computing model identifier, AI capability application module identifier, and regional information (such as TAI List). The AI ​​capability management module performs AI capability management module service discovery on the network resource database (such as NRF). The request message includes the intelligent computing service identifier (for example, S-NSSAI+DNN used in 5G networks, APN used in 4G networks), the intelligent computing model identifier, and selects the AI ​​capability management module based on the AI ​​capability management module service discovery result returned by the network resource database (such as NRF).

[0144] In summary, the main innovations of the present disclosure are: 1) This paper proposes an edge intelligent computing network architecture for mobile communication network users, along with methods for establishing service transmission channels and scheduling intelligent computing services. Based on the existing 5G core network (5GC) and 4G core network (EPC) network architectures, it adds AI capability application modules and AI capability management modules, updates the functions of the user service transmission channel control network element and the user plane network element of the user service transmission channel, and develops a service / signaling process for introducing intelligent computing models into mobile communication networks, establishing intelligent computing service data transmission channels for mobile communication user UEs, and completing intelligent computing model scheduling. This process also involves modifications to the interfaces between these network elements. 2) A new AI capability application module. This module carries intelligent computing power and provides users with a variety of intelligent computing models, such as a general intelligent computing model and an intelligent computing model for personalized user contracts (which can be divided into ToC user active contracting and ToB user active contracting). During the establishment of the intelligent computing service transmission channel, based on the request of the AI ​​capability management module, a core network tunnel is created as the starting point or end point of the tunnel to complete the corresponding service process. 3) A new AI capability management module is added. This module manages and schedules models and computing power. When a new intelligent computing model is added to the mobile network, it interacts with the service transmission channel control network element and the service transmission channel user plane network element to deploy the intelligent computing model to the AI ​​capability application module. After the user accesses the mobile network, it addresses the AI ​​capability application module and provides intelligent computing services. When the corresponding intelligent computing model is not deployed locally, it interacts with the service transmission channel control network element and the AI ​​capability application module to complete model scheduling and updates. AI capability management information is periodically checked and updated.

[0145] 4) The service transmission channel control network element selects the AI ​​capability management module based on the intelligent computing service identifier, and cooperates with the AI ​​capability management module, the service transmission channel user plane network element, and the user data network element to complete the intelligent computing model scheduling service process; 5) In the AI ​​model scheduling service process, the service transmission channel user-plane network element establishes the AI ​​model scheduling service transmission channel together with the AI ​​capability application module based on the service transmission channel control network element request; 6) New interfaces: interface between the AI ​​capability application module and the radio access network element, interface between the AI ​​capability application module and the service transmission channel user plane network element, interface between the AI ​​capability application module and the AI ​​capability management module, interface between the AI ​​capability management module and the service transmission channel control network element, and interface between cross-region AI capability management modules.

[0146] Based on the above-mentioned intelligent computing service channel establishment method, the embodiment of the present disclosure also provides an intelligent computing service channel establishment system corresponding to the above-mentioned intelligent computing service channel establishment method. Figure 1 As shown, it includes: access and mobility management network element, service transmission channel control plane network element, AI capability management module, AI capability application module and wireless access network element; among which: The access and mobility management network element is used to receive an intelligent computing service channel establishment request carrying an intelligent computing service identifier and an intelligent computing model identifier sent by the user terminal UE via the wireless access network network element after the user terminal UE accesses the network and completes registration, determine the service transmission channel control plane network element according to the intelligent computing service identifier, and send the user location information, intelligent computing service identifier and intelligent computing model identifier to the determined service transmission channel control plane network element; and forward the intelligent computing session identifier, core network tunnel information, and session establishment acceptance information to the wireless access network network element, and use the intelligent computing service identifier, the IP address assigned to the user, and the session establishment acceptance information as the information carried in the intelligent computing service channel establishment response, and transparently transmit it to the UE through the wireless access network network element; The service transmission channel control plane network element is used to send user location information, intelligent computing service identifier and intelligent computing model identifier to the AI ​​capability management module and allocate an IP address to the UE; and send user location information, intelligent computing service identifier and intelligent computing model identifier to the AI ​​capability application module pointed to by the AI ​​capability application module information, requesting the AI ​​capability application module to allocate core network tunnel information; and return the intelligent computing session identifier, intelligent computing service identifier, IP address allocated to the user and core network tunnel information to the access and mobility management network element; The AI ​​capability management module is used to determine the AI ​​capability application module information based on the intelligent computing service identifier, intelligent computing model identifier, and user location information, and feed it back to the service transmission channel control plane network element; The AI ​​capability application module is used to distribute core network tunnel information and feed it back to the service transmission channel control plane network element; The wireless access network element is used to forward the intelligent computing service channel establishment request sent by the user equipment UE to the access and mobility management network element, and allocate access network tunnel information for the intelligent computing session; forward the intelligent computing service channel establishment response sent by the access and mobility management network element to the UE, and send the access network tunnel information to the AI ​​capability application module via the access and mobility management network element, the service transmission channel control plane network element and the AI ​​capability management module.

[0147] Furthermore, the access and mobility management network element is specifically used to determine the service transmission channel control plane network element in any one of the following three ways: Way 1: The access and mobility management network element searches for the identifier and address of the service transmission channel control plane network element that matches the intelligent computing service identifier from the locally configured matching list, and uses the network element pointed to by the identifier and address of the found service transmission channel control plane network element as the service transmission channel control plane network element, wherein the matching list contains the matching relationship between the intelligent computing service identifier and the identifier and address of the service transmission channel control plane network element; Way 2: The access and mobility management network element executes a service transmission channel control plane network element service discovery request to the network resource database, wherein the request contains the intelligent computing service identifier; and receives the service discovery result returned by the network resource database, wherein the service discovery The result includes the transmission channel control plane network element identifier and address that matches the intelligent computing service identifier; the network element pointed to by the discovery result is used as the service transmission channel control plane network element; wherein, the service transmission channel control plane network element registers the intelligent computing service identifier for itself and its own address information in the network resource database; method 3: the access and mobility management network element initiates a query request to the network resource database, wherein the query request includes the intelligent computing service identifier; receives the query response returned by the network resource database, wherein the query response includes the transmission channel control plane network element identifier and address that matches the intelligent computing service identifier; the network element pointed to by the query response is used as the service transmission channel control plane network element; wherein, a matching list of the intelligent computing service identifier to the identifier and address of the service transmission channel control plane network element is pre-configured in the network resource database.

[0148] Furthermore, the AI ​​capability management module is specifically used to filter out AI capability application module information that matches the intelligent computing service identifier, intelligent computing model identifier and user location information transmitted by the service transmission channel control plane network element from the stored intelligent computing model registration table; and further filter out AI capability application modules based on the user session management contract data and the service load and capacity of the filtered AI capability application module information, wherein the intelligent computing model registration table stores the correspondence between the intelligent computing service identifier, intelligent computing model identifier, regional information and AI capability application module information, wherein the AI ​​capability application module information includes: AI application module identifier and / or AI application module address.

[0149] Furthermore, the access and mobility management network element is further configured to send the user identifier to the determined service transmission channel control plane network element; The service transmission channel control plane network element is also used to use the user identifier to obtain user session management contract data before sending user location information, intelligent computing service identifier and intelligent computing model identifier to the AI ​​capability management module and allocating an IP address to the UE, determine that the user represented by the user identifier has signed an intelligent computing session, and accept the UE's intelligent computing service notification establishment request.

[0150] Furthermore, the access and mobility management network element is further configured to send a user PCC policy control network element identifier to the determined service transmission channel control plane network element; The service transmission channel control plane network element is also used to determine the need to select the AI ​​capability application module based on the policy information according to the user session management contract data before sending the intelligent user location information, intelligent computing service identifier and intelligent computing model identifier to the AI ​​capability management module and allocating the IP address of the intelligent computing service to the UE; use the user PCC policy control network element identifier to access the user PCC policy control network element, obtain the user's policy information, and use the policy information to establish an intelligent computing session management policy association; and after receiving the AI ​​capability application module information, perform the intelligent computing session management policy association modification and send the IP address allocated to the user to the user PCC policy control network element.

[0151] Furthermore, the access and mobility management network element is further configured to send a user PCC policy control network element identifier to the determined service transmission channel control plane network element; The service transmission channel control plane network element is also used to perform intelligent computing session management policy association modification after receiving AI capability application module information, and send the IP address allocated to the user to the user PCC policy control network element.

[0152] Furthermore, the service transmission channel control plane network element is specifically used to, when it is jointly installed with the AI ​​capability management module, transmit the intelligent computing service identifier, intelligent computing model identifier, user session management contract data and user location information to the AI ​​capability management module; when it is not jointly installed with the AI ​​capability management module, address the AI ​​capability management module according to the intelligent computing service identifier, and transmit the intelligent computing service identifier, intelligent computing model identifier, user session management contract data and user location information to the AI ​​capability management module.

[0153] Furthermore, the policy information includes Qos information; The service transmission channel control plane network element is also used to return QoS information to the access and mobility management network elements; The access and mobility management network element is also used to transparently transmit QoS information to the UE through the radio access network element; The wireless access network element is further configured to receive an accepted or rejected QoS list sent by the user terminal after allocating access network tunnel information for the intelligent computing session; and send the accepted or rejected QoS list and access network tunnel information to the service transmission channel control plane element via the access and mobility management network element; The service transmission channel control plane network element is also used to release the corresponding QoS configuration file according to the QoS list rejected by the user terminal, and send access network tunnel information to the AI ​​capability application module through the AI ​​capability management module.

[0154] The present disclosure also provides a scheduling method based on the above-mentioned AI model. The embodiments of the present disclosure also provide a device corresponding to the above-mentioned AI model scheduling method, which is specifically as follows: 1) A visit location AI capability management module, including: A first receiving module is used to receive location information, an intelligent computing service identifier, and an intelligent computing model identifier; A first sending module is configured to use the stored intelligent computing model registration table to perform intra-area scheduling when it is determined that there is an AI capability application module that matches both the intelligent computing service identifier and the intelligent computing model identifier and the area information corresponding to the matching AI capability application module does not match the received location information, and to initiate an AI model scheduling service request to the visited location AI capability application module to request allocation of core network tunnel information on the visited location AI capability application module side, wherein the intelligent computing model registration table stores a correspondence between the intelligent computing service identifier, the intelligent computing model identifier, the area information, and the AI ​​capability application module information; A second receiving module is configured to receive the first core network tunnel information returned by the visited location AI capability application module; The second sending module is used to initiate an AI model scheduling service request to the AI ​​capability application module at the deployment site pointed to by the matching AI capability application module information, requesting the allocation of the core network tunnel information on the AI ​​capability application module side at the deployment site, and the request carries the intelligent computing service identifier and the intelligent computing model identifier; A third receiving module is configured to receive fourth core network tunnel information returned by the deployed AI capability application module; The third sending module is configured to initiate an AI model scheduling service request to the control plane network element of the visited service transmission channel, where the request carries the first core network tunnel information, the fourth core network tunnel information, the location information, and the regional information of the deployed AI capability application module; The forwarding module forwards the second core network tunnel information sent by the control plane network element of the visited service transmission channel to the visited AI capability application module and the deployed AI capability application module when it is determined that the user plane network element of the visited service transmission channel and the user plane network element of the deployed service transmission channel are the same network element, wherein the second core network tunnel information is the core network tunnel information at the user plane network element end of the visited service transmission channel.

[0155] 2) Provide a visitor location AI capability application module, including: A first receiving module is configured to receive an AI model scheduling service request initiated by a visited location AI capability management module; an allocation module, configured to allocate first core network tunnel information; A sending module is used to return the information to the AI ​​capability management module at the visited location; The second receiving module is used to receive the second core network tunnel information sent by the visited AI capability management module; wherein the second core network tunnel information is the core network tunnel information of the user plane network element end of the visited service transmission channel The third receiving module is used to receive a copy of the intelligent computing model sent by the AI ​​capability application module at the deployment site.

[0156] 3) Provide a control plane network element for the visited service transmission channel, which includes: A first receiving module is configured to receive an AI model scheduling service request initiated by the visited AI capability management module, wherein the request carries first core network tunnel information, fourth core network tunnel information, location information, and regional information of the deployed AI capability application module, wherein the first core network tunnel information refers to the core network tunnel information of the visited AI application module end, and the fourth core network tunnel information refers to the core network tunnel information of the deployed AI application module end; An addressing module is used to address the user plane network element of the service transmission channel at the visited location using the location information carried in the AI ​​model scheduling service request, and to address the user plane network element of the service transmission channel at the deployed location using the area information; A first sending module is configured to send a user plane tunnel establishment request carrying the first core network tunnel information and the fourth core network tunnel information to the user plane network element of the visited service transmission channel when it is determined that the user plane network element of the visited service transmission channel and the user plane network element of the deployed service transmission channel are the same network element; The second receiving module is configured to receive the second core network tunnel information sent by the user plane network element of the visited service transmission channel; The second sending module is configured to send the second core network tunnel information to the visited site AI capability application module and the deployed site AI capability application module via the visited site AI capability management module.

[0157] 4) A control plane network element for a visited service transmission channel, when performing cross-region scheduling, includes: a third receiving module, configured to receive an AI model scheduling service request initiated by the visited-location AI capability management module, the request carrying first core network tunnel information and location information; wherein the first core network tunnel information refers to the core network tunnel information of the visited-location AI application module; An addressing module, used for addressing a user plane network element of a visited service transmission channel according to location information; A third sending module is configured to send a user plane tunnel establishment request to a user plane network element of a visited service transmission channel, wherein the request carries the first core network tunnel information; A fourth receiving module is configured to receive the second core network tunnel information fed back by the user plane network element of the user's visited site service transmission channel, and transmit the information to the visited site AI capability application module via the visited site AI capability management module; A fifth receiving module is configured to receive the third core network tunnel information returned by the visited AI capability management module, wherein the third core network tunnel information is the core network tunnel information of the user plane network element end of the deployed service transmission channel; The fourth sending module is configured to return the third core network tunnel information to the user plane network element of the visited service transmission channel.

[0158] 5) Provide an AI capability application module for deployment, which includes: A first receiving module is configured to receive an AI model scheduling service request initiated by the visited AI capability management module to allocate core network tunnel information of the deployed AI capability application module, wherein the request carries an intelligent computing service identifier and an intelligent computing model identifier; A first allocation module, configured to allocate fourth core network tunnel information; A first sending module, configured to return the fourth core network tunnel information to the visited location AI capability management module; A second receiving module is configured to receive, when the visited site service transmission channel control plane network element determines that the visited site service transmission channel user plane network element and the deployed site service transmission channel user plane network element are the same network element, the second core network tunnel information sent by the visited site AI capability management module, wherein the second core network tunnel information is the core network tunnel information of the visited site service transmission channel user plane network element end; A selection module is used to select an intelligent computing model using an intelligent computing service identifier and an intelligent computing model identifier; The second sending module transmits the selected intelligent computing model copy to the visited AI capability application module.

[0159] 6) Provide an AI capability application module for deployment, which includes: A third receiving module is configured to receive an AI model scheduling service process initiated by the AI ​​capability management module at the deployment site to request the AI ​​capability application module at the deployment site to allocate core network tunnel information; A second allocation module, configured to allocate fourth core network tunnel information; A third sending module is configured to return the fourth core network tunnel information to the AI ​​capability management module at the deployment site; The fourth sending module is used to transmit the selected intelligent computing model copy to the visited AI capability application module through the established user-side intelligent computing business model transmission channel.

[0160] 7) A user plane network element for a visited service transmission channel, when performing intra-area scheduling, includes: a first receiving module, configured to receive a user plane tunnel establishment request sent by a visited-site service transport channel control plane network element, the request carrying first core network tunnel information and fourth core network tunnel information, wherein the first core network tunnel information refers to the core network tunnel information of the visited-site AI application module, and the fourth core network tunnel information refers to the core network tunnel information of the deployed-site AI application module; A first allocation module, configured to allocate second core network tunnel information; The first sending module is configured to send the second core network tunnel information to a control plane network element of a visited service transmission channel.

[0161] 8) A user plane network element for a visited service transmission channel, when performing cross-region scheduling, includes: a second receiving module, configured to receive a user plane tunnel establishment request sent by a visited service transport channel control plane network element, the request carrying first core network tunnel information, wherein the first core network tunnel information refers to the core network tunnel information of the visited AI application module; A second allocation module, configured to allocate second core network tunnel information; A second sending module sends the second core network tunnel information to the visited service transmission channel control plane network element; A third receiving module is configured to receive third core network tunnel information sent by a control plane network element of a visited service transmission channel, wherein the third core network tunnel information refers to core network tunnel information of a user plane network element end of a deployed service transmission channel; The third sending module forwards the intelligent computing model copy sent by the user-plane network element of the deployed service transmission channel to the AI ​​capability application module at the visited location.

[0162] 9) Provide a user-plane network element for a deployed service transmission channel, which includes: A first receiving module is configured to receive a user plane tunnel establishment request sent by a visited service transmission channel control plane network element, the request carrying fourth core network tunnel information, wherein the fourth core network tunnel information refers to the core network tunnel information of the AI ​​capability application module end at the deployment site; A distribution module, configured to distribute third core network tunnel information; A first sending module, configured to send the third core network tunnel information to a visited service transmission channel control plane network element; A second receiving module is configured to receive second core network tunnel information sent by a control plane network element of a visited service transmission channel, wherein the second core network tunnel information refers to core network tunnel information at a user plane network element end of the visited service transmission channel; The second sending module is used to forward the intelligent computing model copy sent by the AI ​​capability application module at the deployment site to the AI ​​capability application module at the visited site.

[0163] 10) Provide a service transmission channel control plane network element for deployment, which includes: A first receiving module is configured to receive an AI model scheduling service request sent by an AI capability management module at a deployment site, the request carrying information about a fourth core network tunnel, information about a second core network tunnel, and information about an area under the responsibility of an AI capability application module at a deployment site, wherein the second core network tunnel information is information about a core network tunnel at a user plane network element end of a visited site service transmission channel, and the fourth core network tunnel information is information about a core network tunnel at a deployment site AI capability application module end; An addressing module is used to address the user plane network element of the deployed service transmission channel according to the area information of the deployed AI capability application module; The first sending module is configured to initiate a user plane tunnel establishment request to a user plane network element of a deployed service transmission channel, wherein the request carries information about the fourth core network tunnel and information about the second core network tunnel; The second receiving module is used to receive the third core network tunnel information returned by the user plane network element of the service transmission channel at the deployment site; The second sending module is used to return the third core network tunnel information to the AI ​​capability management module at the deployment site.

[0164] 11) Provide a deployment-based AI capability management module, including: a first receiving module, configured to receive an AI model scheduling service request sent by a visited location AI capability management module, the request carrying second core network tunnel information, wherein the second core network tunnel information is core network tunnel information of a user plane network element end of a visited location service transmission channel; The first sending module is used to initiate an AI model scheduling service request for allocating core network tunnel information of the AI ​​capability application module at the deployment site to the AI ​​capability application module at the deployment site, wherein the request carries an intelligent computing service identifier and an intelligent computing model identifier; a second receiving module, configured to receive fourth core network tunnel information returned by the deployed AI capability application module, wherein the fourth core network tunnel information is core network tunnel information of the deployed AI capability application module; The second sending module is used to send an AI model scheduling service request to the service transmission channel control plane network element of the deployment site, where the request carries the fourth core network tunnel information, the second core network tunnel information, and the area information responsible for the AI ​​capability application module of the deployment site; A third receiving module is configured to receive third core network tunnel information sent by a control plane network element of a deployed service transmission channel, wherein the third core network tunnel information is core network tunnel information of a user plane network element end of the deployed service transmission channel; The third sending module is used to return the AI ​​model scheduling service request establishment result to the visited AI capability management module, where the result includes: third core network tunnel information.

[0165] The present disclosure also provides an AI model scheduling system. When scheduling in the same area, see Figure 4 , including: a visited site AI capability management module, a visited site AI capability application module, a visited site service transmission channel control plane network element for intra-area scheduling, a deployment site AI capability application module for intra-area scheduling, a visited site service transmission channel user plane network element for intra-area scheduling, and a deployment site service transmission channel user plane network element for intra-area scheduling; In case of cross-region scheduling, see Figure 8 , including: visited site AI capability management module, visited site AI capability application module, visited site service transmission channel control plane network element for user cross-region scheduling, visited site service transmission channel user plane network element for user cross-region scheduling, deployed site service transmission channel control plane network element for user cross-region scheduling, deployed site AI capability application module for user cross-region scheduling, deployed site AI capability management module for user cross-region scheduling.

[0166] Based on the same inventive concept as the above disclosure, the present disclosure also provides an electronic device. The electronic device of the present disclosure embodiment includes at least one processor and at least one memory electrically connected to each other, the memory being electrically connected to the processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method described above.

[0167] It should be noted that the electrical connection between the above-mentioned units does not necessarily mean the connection between lines. An indirect connection method can be applied to the embodiments of the present disclosure as long as the purpose of the present disclosure is achieved.

[0168] Based on the same inventive concept, the present disclosure further provides a computer storage medium, wherein the computer storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0169] Although the present disclosure has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure.

Claims

1. An AI model scheduling method, characterized in that: include: The AI ​​capability management module of the visited location receives the location information, intelligent computing service identifier, and intelligent computing model identifier; The AI ​​capability management module at the visited location uses the stored intelligent computing model registration table to determine whether there is an AI capability application module that matches both the intelligent computing service identifier and the intelligent computing model identifier, and if the area information corresponding to the matching AI capability application module does not match the received location information, the module performs the following same-area scheduling: The visited-location AI capability management module initiates an AI model scheduling service request to the visited-location AI capability application module, requesting allocation of core network tunnel information for the visited-location AI capability application module. The intelligent computing model registration table stores a correspondence between the intelligent computing service identifier, intelligent computing model identifier, regional information, and AI capability application module information. The visited-location AI capability application module returns the first core network tunnel information to the visited-location AI capability management module. The visited-site AI capability management module initiates an AI model scheduling service request to the deployment-site AI capability application module pointed to by the matching AI capability application module information, requesting the allocation of the deployment-site AI capability application module-side core network tunnel information. The request carries the intelligent computing service identifier and intelligent computing model identifier. The deployment-site AI capability application module returns the fourth core network tunnel information to the visited-site AI capability management module. The visited location AI capability management module initiates an AI model scheduling service request to the visited location service transmission channel control plane network element. The request carries the first core network tunnel information, the fourth core network tunnel information, the location information, and the regional information of the deployed location AI capability application module. The visited service transmission channel control plane network element uses the location information carried in the AI ​​model scheduling service request to address the visited service transmission channel user plane network element, and uses the area information to address the deployed service transmission channel user plane network element. When it is determined that the visited service transmission channel user plane network element and the deployed service transmission channel user plane network element are the same network element, the user plane network element sends a user plane tunnel establishment request carrying the first core network tunnel information and the fourth core network tunnel information to the visited service transmission channel user plane network element. The user plane network element of the visited service transmission channel sends the second core network tunnel information to the control plane network element of the visited service transmission channel; The visited-site service transmission channel control plane network element sends the second core network tunnel information to the visited-site AI capability application module and the deployed-site AI capability application module via the visited-site AI capability management module. At this point, the visited-site AI capability application module and the deployed-site AI capability application module establish a user plane intelligent computing service model transmission channel. The AI ​​capability application module at the deployed site uses the intelligent computing service identifier and the intelligent computing model identifier to select an intelligent computing model, and passes a copy of the selected intelligent computing model to the AI ​​capability application module at the visited site.

2. The method according to claim 1, characterized in that The method further comprises: The visited service transmission channel control plane network element uses the AI ​​model to schedule the location information and area information carried in the service request. When it is determined that the visited service transmission channel user plane network element and the deployed service transmission channel user plane network element are different network elements, it sends a user plane tunnel establishment request carrying the first core network tunnel information to the visited service transmission channel user plane network element, and sends a user plane tunnel establishment request carrying the fourth core network tunnel information to the deployed service transmission channel user plane network element. The user plane network element of the visited service transmission channel sends the second core network tunnel information to the control plane network element of the visited service transmission channel; The user plane network element of the service transmission channel at the deployment site sends the third core network tunnel information to the control plane network element of the service transmission channel at the visited site; The visited-site service transmission channel control plane network element sends the second core network tunnel information to the visited-site AI capability application module via the visited-site AI capability management module, and sends the third core network tunnel information to the deployment-site AI capability application module; The control plane network element of the visited site service transmission channel sends the second core network tunnel information to the user plane network element of the deployed site service transmission channel, and sends the third core network tunnel information to the user plane network element of the visited site service transmission channel. At this point, the visited site AI capability application module and the deployed site AI capability application module establish a user plane intelligent computing service model transmission channel. The AI ​​capability application module at the deployed site uses the intelligent computing service identifier and the intelligent computing model identifier to select an intelligent computing model, and passes a copy of the selected intelligent computing model to the AI ​​capability application module at the visited site.

3. An AI model scheduling method, characterized in that: The method comprises: The AI ​​capability management module of the visited location receives the location information, intelligent computing service identifier, and intelligent computing model identifier; The AI ​​capability management module at the visited location uses the stored intelligent computing model registration table to determine if there is an AI capability application module that matches both the intelligent computing service ID and the intelligent computing model ID. It then performs the following cross-region scheduling: The visited-location AI capability management module initiates an AI model scheduling service request to the visited-location AI capability application module, requesting allocation of core network tunnel information on the visited-location AI capability application module side. The intelligent computing model registration table stores the correspondence between the intelligent computing service identifier, intelligent computing model identifier, regional information, and AI capability application module information. The visited-location AI capability application module returns the first core network tunnel information to the visited-location AI capability management module. The visited-site AI capability management module initiates an AI model scheduling service request to the visited-site service transmission channel control plane network element, which carries the first core network tunnel information and location information. The visited-site service transmission channel control plane network element addresses the visited-site service transmission channel user plane network element based on the location information and sends a user plane tunnel establishment request, which carries the first core network tunnel information. The visited-site service transmission channel control plane network element receives the second core network tunnel information fed back by the visited-site service transmission channel user plane network element and transmits it to the visited-site AI capability application module via the visited-site AI capability management module. The AI ​​capability management module at the visited location addresses the AI ​​capability management module at the deployed location based on the intelligent computing service identifier and the intelligent computing model identifier, and sends an AI model scheduling service request to the AI ​​capability management module at the deployed location. The request carries the intelligent computing service identifier, the intelligent computing model identifier, and the second core network tunnel information. The AI ​​capability management module at the deployment site determines the AI ​​capability application module at the deployment site based on the intelligent computing service identifier and the intelligent computing model identifier, and initiates the AI ​​model scheduling service process to the module, requesting the AI ​​capability application module at the deployment site to allocate core network tunnel information. The AI ​​capability application module at the deployment site returns the fourth core network tunnel information to the AI ​​capability management module at the deployment site; The AI ​​capability management module at the deployment site initiates an AI model scheduling service request to the service transport channel control plane network element at the deployment site. The request carries the fourth core network tunnel information, the second core network tunnel information, and the area information responsible for the AI ​​capability application module at the deployment site. The service transport channel control plane network element at the deployment site and the service transport channel control plane network element at the visited site may be the same network element or different network elements. The service transmission channel control plane network element at the deployment site addresses the service transmission channel user plane network element at the deployment site according to the area information responsible for the AI ​​capability application module at the deployment site, and initiates a user plane tunnel establishment request to the service transmission channel user plane network element at the deployment site. The request carries the fourth core network tunnel information and the second core network tunnel information. The user plane network element of the service transmission channel at the deployment site returns the third core network tunnel information to the control plane network element of the service transmission channel at the deployment site; The service transmission channel control plane network element at the deployment site returns the third core network tunnel information to the AI ​​capability management module at the deployment site; The AI ​​capability management module at the deployment site returns the AI ​​model scheduling service establishment request result to the AI ​​capability management module at the visited site. The result includes: the third core network tunnel information; The visited location AI capability management module returns the third core network tunnel information to the visited location service transmission channel control plane network element. The visited location service transmission channel control plane network element returns the third core network tunnel information to the visited location service transmission channel user plane network element. At this point, the intelligent computing model scheduling service transmission channel is established. The AI ​​capability application module at the deployed site uses the intelligent computing service identifier and the intelligent computing model identifier to select an intelligent computing model, and passes a copy of the selected intelligent computing model to the AI ​​capability application module at the visited site.

4. The method according to claim 3, characterized in that The AI ​​capability management module at the visited location sends an AI model scheduling service request to the AI ​​capability management module at the deployed location, including: The AI ​​capability management module at the visited location sends an AI model scheduling service request to the AI ​​capability management module at the deployed location via the service transmission channel control plane network element at the visited location and the service transmission channel control plane network element at the deployed location; and / or The AI ​​capability management module at the deployment site returns the AI ​​model scheduling service establishment request result to the AI ​​capability management module at the visiting site, including: The AI ​​capability management module at the deployment site returns the AI ​​model scheduling service request establishment result to the AI ​​capability management module at the visited site via the service transmission channel control plane network element at the visited site and the service transmission channel control plane network element at the deployment site.

5. The method according to claim 3 or 4, characterized in that After the AI ​​capability application module at the deployed site selects an intelligent computing model using the intelligent computing service identifier and the intelligent computing model identifier, and transmits a copy of the selected intelligent computing model to the AI ​​capability application module at the visited site, the method further includes: The AI ​​capability application module at the visited location returns the intelligent computing model transfer completion information to the AI ​​capability management module at the visited location, and the AI ​​capability management module registers the model.

6. The method according to claim 3 or 4, characterized in that The AI ​​capability management module for the visited location uses one of the following two methods for addressing: Method 1: Configure the global intelligent computing model registration information table locally and notify the global and other AI capability management modules when the registration information changes. Method 2: The AI ​​capability management module completes registration with the network resource database when the intelligent computing model registration information changes. The registration information includes: AI capability management module identifier, AI capability management module address, intelligent computing service identifier, intelligent computing model identifier, AI capability application module identifier, and regional information; the AI ​​capability management module performs AI capability management module service discovery on the network resource database. The request message includes the intelligent computing service identifier and the intelligent computing model identifier, and selects the AI ​​capability management module based on the AI ​​capability management module service discovery result returned by the network resource database.

7. An AI model scheduling method, characterized in that: Applied in the AI ​​capability management module for visited locations, including: Receive location information, intelligent computing service identifier, and intelligent computing model identifier; Using the stored intelligent computing model registration table, it is determined that there is an AI capability application module that matches both the intelligent computing service identifier and the intelligent computing model identifier, and the area information corresponding to the matching AI capability application module does not match the received location information. In this case, intra-area scheduling is performed, and an AI model scheduling service request is initiated to the visited location AI capability application module, requesting allocation of the core network tunnel information on the visited location AI capability application module side. The intelligent computing model registration table stores the correspondence between the intelligent computing service identifier, the intelligent computing model identifier, the area information, and the AI ​​capability application module information. Receive the first core network tunnel information returned by the visited AI capability application module; Initiate an AI model scheduling service request to the AI ​​capability application module at the deployment location pointed to by the matching AI capability application module information, requesting the allocation of the core network tunnel information on the AI ​​capability application module side at the deployment location. The request carries the intelligent computing service identifier and intelligent computing model identifier. Receive the fourth core network tunnel information returned by the AI ​​capability application module at the deployment site; Initiate an AI model scheduling service request to the control plane network element of the visited service transmission channel. The request carries the first core network tunnel information, the fourth core network tunnel information, the location information, and the regional information of the deployed AI capability application module. When it is determined that the user plane network element of the visited service transmission channel and the user plane network element of the deployed service transmission channel are the same network element, the second core network tunnel information sent by the control plane network element of the visited service transmission channel is forwarded to the visited AI capability application module and the deployed AI capability application module, wherein the second core network tunnel information is the core network tunnel information of the user plane network element end of the visited service transmission channel.

8. The method according to claim 7, characterized in that When it is determined that the user plane network element of the visited service transmission channel and the user plane network element of the deployed service transmission channel are different network elements, the second core network tunnel information sent by the control plane network element of the visited service transmission channel is forwarded to the visited AI capability application module, and the third core network tunnel information is forwarded to the AI ​​capability application module of the deployed site, wherein the third core network tunnel information is the core network tunnel information of the user plane network element end of the deployed site service transmission channel.

9. An AI model scheduling method, characterized in that: Applied in the AI ​​capability management module for visited locations, including: Receive location information, intelligent computing service identifier, and intelligent computing model identifier; Using the stored intelligent computing model registration table, it is determined that there is no AI capability application module that matches both the intelligent computing service identifier and the intelligent computing model identifier. An AI model scheduling service request is initiated to the visited location AI capability application module, requesting allocation of the core network tunnel information on the visited location AI capability application module side. The intelligent computing model registration table stores the correspondence between the intelligent computing service identifier, intelligent computing model identifier, regional information, and AI capability application module information. Receive the first core network tunnel information returned by the visited AI capability application module; Initiate an AI model scheduling service request to the control plane network element of the visited service transmission channel, carrying the first core network tunnel information and location information; Forwarding the second core network tunnel information sent by the visited service transmission channel control plane network element to the visited AI capability application module; The AI ​​capability management module at the deployment location is addressed based on the intelligent computing service identifier and intelligent computing model identifier, and an AI model scheduling service request is sent. The request carries the intelligent computing service identifier, intelligent computing model identifier, and second core network tunnel information. Receive the AI ​​model scheduling service request establishment result returned by the AI ​​capability management module at the deployment site, where the result includes: third core network tunnel information, where the third core network tunnel information is the core network tunnel information of the user plane network element end of the deployment site service transmission channel; Return the third core network tunnel information to the visited service transmission channel control plane network element.

10. The method according to claim 9, characterized in that Also includes: Receive the intelligent computing model transfer completion information returned by the AI ​​capability application module at the visited location and register the intelligent computing model.

11. The method according to claim 9 or 10, characterized in that Addressing is done in one of two ways: Method 1: Configure the global intelligent computing model registration information table locally and notify the global and other AI capability management modules when the registration information changes. Method 2: When the registration information of the intelligent computing model changes, registration with the network resource database is completed. The registration information includes: AI capability management module identifier, AI capability management module address, intelligent computing service identifier, intelligent computing model identifier, AI capability application module identifier, and regional information; the AI ​​capability management module performs AI capability management module service discovery on the network resource database. The request message includes the intelligent computing service identifier and the intelligent computing model identifier, and selects the AI ​​capability management module based on the AI ​​capability management module service discovery result returned by the network resource database.

12. An AI model scheduling method, characterized in that: AI capability application modules applied in the visited locations include: Receive AI model scheduling service requests from the AI ​​capability management module at the visited location; Allocate the first core network tunnel information and return it to the visited location AI capability management module; Receive the second core network tunnel information sent by the visited AI capability management module; wherein the second core network tunnel information is the core network tunnel information of the user plane network element end of the visited service transmission channel; Receive the intelligent computing model copy sent by the AI ​​capability application module at the deployment site.

13. An AI model scheduling method, characterized in that: Applied to a control plane network element of a visited service transmission channel, when performing intra-area scheduling, the method includes: Receive an AI model scheduling service request initiated by the visited AI capability management module, the request carrying the first core network tunnel information, the fourth core network tunnel information, the location information, and the regional information of the deployed AI capability application module. The first core network tunnel information refers to the core network tunnel information of the visited AI application module, and the fourth core network tunnel information refers to the core network tunnel information of the deployed AI application module. Use the location information carried in the AI ​​model scheduling service request to address the user plane network element of the visited service transmission channel, use the area information to address the user plane network element of the deployed service transmission channel, and when it is determined that the user plane network element of the visited service transmission channel and the user plane network element of the deployed service transmission channel are the same network element, send a user plane tunnel establishment request carrying the first core network tunnel information and the fourth core network tunnel information to the user plane network element of the visited service transmission channel; receiving the second core network tunnel information sent by the user plane network element of the visited service transmission channel; The second core network tunnel information is sent to the visited site AI capability application module and the deployed site AI capability application module respectively via the visited site AI capability management module.

14. The method according to claim 13, characterized in that The method further comprises: Using the location information and area information carried in the AI ​​model scheduling service request, when it is determined that the user plane network element of the visited service transmission channel and the user plane network element of the deployed service transmission channel are different network elements, a user plane tunnel establishment request carrying the first core network tunnel information is sent to the user plane network element of the visited service transmission channel, and a user plane tunnel establishment request carrying the fourth core network tunnel information is sent to the user plane network element of the deployed service transmission channel; receiving the second core network tunnel information sent by the user plane network element of the visited service transmission channel; Receiving the third core network tunnel information sent by the user plane network element of the service transmission channel at the deployment site; Sending the second core network tunnel information to the visited location AI capability application module via the visited location AI capability management module, and sending the third core network tunnel information to the deployed location AI capability application module; The second core network tunnel information is sent to the user plane network element of the service transmission channel at the deployment site, and the third core network tunnel information is sent to the user plane network element of the service transmission channel at the visited site.

15. An AI model scheduling method, characterized in that: Applied to a control plane network element of a visited service transmission channel, when performing cross-region scheduling, the method includes: Receive an AI model scheduling service request initiated by the visited location AI capability management module, which carries the first core network tunnel information and location information; Addressing the user plane network element of the visited service transmission channel according to the location information, sending a user plane tunnel establishment request, and carrying the first core network tunnel information in the request; Receive the second core network tunnel information fed back by the user plane network element of the user's visited service transmission channel, and transmit it to the visited AI capability application module via the visited AI capability management module; Receive the third core network tunnel information returned by the visited AI capability management module, where the third core network tunnel information is the core network tunnel information of the user plane network element end of the deployed service transmission channel; Return the third core network tunnel information to the user plane network element of the visited service transmission channel.

16. An AI model scheduling method, characterized in that: The AI ​​capability application module applied in the deployment area includes: Receive an AI model scheduling service request initiated by the visited AI capability management module to allocate core network tunnel information for the deployed AI capability application module. The request carries the intelligent computing service identifier and intelligent computing model identifier. Allocating fourth core network tunnel information; Return the fourth core network tunnel information to the visited AI capability management module; When the user plane network element of the visited site service transmission channel and the user plane network element of the deployed site service transmission channel are the same network element, receiving the second core network tunnel information sent by the visited site AI capability management module, and interacting with the visited site service transmission channel user plane network element to facilitate the visited site AI capability application module and the deployed site AI capability application module to establish a user plane intelligent computing service model transmission channel; wherein the second core network tunnel information is the core network tunnel information of the user plane network element end of the visited site service transmission channel; Use the intelligent computing service identifier and intelligent computing model identifier to select the intelligent computing model, and pass the selected intelligent computing model copy to the visited AI capability application module through the established user-side intelligent computing service model transmission channel.

17. The method according to claim 16, characterized in that When it is determined that the user plane network element of the visited service transmission channel and the user plane network element of the deployed service transmission channel are not the same network element, third core network tunnel information sent by the visited AI capability management module is received, wherein the third core network tunnel information is core network tunnel information of the user plane network element end of the deployed service transmission channel.

18. An AI model scheduling method, characterized in that: The AI ​​capability application module applied in the deployment area, when performing cross-region scheduling, includes: Receive the AI ​​model scheduling service process initiated by the AI ​​capability management module at the deployment site to request the AI ​​capability application module at the deployment site to allocate core network tunnel information; Allocating fourth core network tunnel information; Return the fourth core network tunnel information to the AI ​​capability management module at the deployment site, so that the AI ​​capability management module at the deployment site carries the fourth core network tunnel information, the second core network tunnel information, and the regional information in the AI ​​model scheduling service request and sends it to the service transmission channel control plane network element at the deployment site, where the second core network tunnel information refers to the core network tunnel information at the user plane network element end of the visited service transmission channel; Use the intelligent computing service identifier and intelligent computing model identifier to select the intelligent computing model, and pass the selected intelligent computing model copy to the visited AI capability application module through the established user-side intelligent computing service model transmission channel.

19. An AI model scheduling method, characterized in that: Applied to the user-plane network element of the visited service transmission channel, during intra-area scheduling, including: Receive a user plane tunnel establishment request sent by a visited service transport channel control plane network element, where the request carries first core network tunnel information and fourth core network tunnel information, where the first core network tunnel information refers to the core network tunnel information of the visited AI application module, and the fourth core network tunnel information refers to the core network tunnel information of the deployed AI application module; Allocating second core network tunnel information; Sending the second core network tunnel information to the visited service transmission channel control plane network element; Forward the intelligent computing model copy sent by the AI ​​capability module at the deployment site to the AI ​​capability application module at the visited site.

20. An AI model scheduling method, characterized in that: Applied to the user-plane network element of the visited service transmission channel, during cross-region scheduling, including: receiving a user plane tunnel establishment request sent by a visited service transport channel control plane network element, where the request carries first core network tunnel information, wherein the first core network tunnel information refers to the core network tunnel information of the visited AI application module; Allocating second core network tunnel information; Sending the second core network tunnel information to the visited service transmission channel control plane network element; Receiving the third core network tunnel information sent by the control plane network element of the visited service transmission channel, wherein the third core network tunnel information refers to the core network tunnel information of the user plane network element end of the deployed service transmission channel; Forward the intelligent computing model copy sent by the user-plane network element of the service transmission channel at the deployment site to the AI ​​capability application module at the visited site.

21. An AI model scheduling method, characterized in that: Applied to the user-plane network elements of the service transmission channel at the deployment site, during intra-area scheduling, including: Receive a user plane tunnel establishment request from a control plane network element of a visited service transmission channel, where the request carries information about a fourth core network tunnel. The fourth core network tunnel information refers to the core network tunnel information of the AI ​​capability application module at the deployment site. Allocate third core network tunnel information; Sending the third core network tunnel information to the visited service transmission channel control plane network element; receiving second core network tunnel information sent by a control plane network element of a visited service transmission channel, wherein the second core network tunnel information refers to core network tunnel information at a user plane network element end of the visited service transmission channel; Forward the intelligent computing model copy sent by the AI ​​capability application module at the deployment site to the AI ​​capability application module at the visited site.

22. An AI model scheduling method, characterized in that: Applied to the user-plane network elements of the service transmission channel at the deployment site, during cross-area scheduling, including: Receive a user plane tunnel establishment request sent by a control plane network element of a service transport channel at a deployment site, where the request carries information about a fourth core network tunnel and information about a second core network tunnel. The second core network tunnel information is information about a core network tunnel at the user plane network element end of the visited service transport channel, and the fourth core network tunnel information is information about a core network tunnel at the AI ​​capability application module end of the deployment site. Allocate third core network tunnel information; Return the third core network tunnel information to the service transmission channel control plane network element at the deployment site; Forward the intelligent computing model copy sent by the AI ​​capability application module at the deployed location to the control plane network element of the service transmission channel at the visited location.

23. An AI model scheduling method, characterized in that: Applied to the control plane network element of the service transmission channel at the deployment site, when performing cross-area scheduling, including: Receive an AI model scheduling service request sent by the AI ​​capability management module at the deployment site. The request carries the fourth core network tunnel information, the second core network tunnel information, and the area information responsible for the AI ​​capability application module at the deployment site. The second core network tunnel information is the core network tunnel information of the user plane network element end of the visited service transmission channel, and the fourth core network tunnel information is the core network tunnel information of the AI ​​capability application module end at the deployment site. According to the area information of the deployed AI capability application module, the module addresses the user plane network element of the deployed service transmission channel and initiates a user plane tunnel establishment request to the user plane network element of the deployed service transmission channel. The request carries the fourth core network tunnel information and the second core network tunnel information. Receiving the third core network tunnel information returned by the user plane network element of the service transmission channel at the deployment site; Return the third core network tunnel information to the AI ​​capability management module at the deployment site.

24. An AI model scheduling method, characterized in that: The AI ​​capability management module applied in the deployment site includes: Receive an AI model scheduling service request sent by the visited location AI capability management module, where the request carries the second core network tunnel information, where the second core network tunnel information is the core network tunnel information of the user plane network element end of the visited location service transmission channel; Initiate an AI model scheduling service request to the AI ​​capability application module at the deployment site to allocate the core network tunnel information of the AI ​​capability application module at the deployment site. The request carries the intelligent computing service identifier and the intelligent computing model identifier. Receiving fourth core network tunnel information returned by the deployed AI capability application module, wherein the fourth core network tunnel information is core network tunnel information of the deployed AI capability application module end; Send an AI model scheduling service request to the service transmission channel control plane network element at the deployment site. The request carries the fourth core network tunnel information, the second core network tunnel information, and the area information responsible for the deployment site's AI capability application module. receiving third core network tunnel information sent by a control plane network element of a deployed service transmission channel, wherein the third core network tunnel information is core network tunnel information of a user plane network element end of the deployed service transmission channel; Return the AI ​​model scheduling service establishment request result to the visited AI capability management module, which includes: the third core network tunnel information.

25. A visiting location AI capability management module, characterized in that: include: A first receiving module is used to receive location information, an intelligent computing service identifier, and an intelligent computing model identifier; A first sending module is configured to use the stored intelligent computing model registration table to perform intra-area scheduling when it is determined that there is an AI capability application module that matches both the intelligent computing service identifier and the intelligent computing model identifier and the area information corresponding to the matching AI capability application module does not match the received location information, and to initiate an AI model scheduling service request to the visited location AI capability application module to request allocation of core network tunnel information on the visited location AI capability application module side, wherein the intelligent computing model registration table stores a correspondence between the intelligent computing service identifier, the intelligent computing model identifier, the area information, and the AI ​​capability application module information; A second receiving module is configured to receive the first core network tunnel information returned by the visited location AI capability application module; The second sending module is used to initiate an AI model scheduling service request to the AI ​​capability application module at the deployment site pointed to by the matching AI capability application module information, requesting the allocation of the core network tunnel information on the AI ​​capability application module side at the deployment site, and the request carries the intelligent computing service identifier and the intelligent computing model identifier; A third receiving module is configured to receive fourth core network tunnel information returned by the deployed AI capability application module; The third sending module is configured to initiate an AI model scheduling service request to the control plane network element of the visited service transmission channel, where the request carries the first core network tunnel information, the fourth core network tunnel information, the location information, and the regional information of the deployed AI capability application module; The forwarding module forwards the second core network tunnel information sent by the control plane network element of the visited service transmission channel to the visited AI capability application module and the deployed AI capability application module when it is determined that the user plane network element of the visited service transmission channel and the user plane network element of the deployed service transmission channel are the same network element, wherein the second core network tunnel information is the core network tunnel information at the user plane network element end of the visited service transmission channel.

26. A visiting location AI capability application module, characterized in that: include: A first receiving module is configured to receive an AI model scheduling service request initiated by a visited location AI capability management module; an allocation module, configured to allocate first core network tunnel information; A sending module is used to return the information to the AI ​​capability management module at the visited location; A second receiving module is configured to receive the second core network tunnel information sent by the access AI capability management module; The second core network tunnel information is the core network tunnel information of the user plane network element end of the visited service transmission channel; The third receiving module is used to receive a copy of the intelligent computing model sent by the AI ​​capability application module at the deployment site.

27. A visited service transmission channel control plane network element, characterized in that: When dispatching in the same area, it includes: a first receiving module, configured to receive an AI model scheduling service request initiated by a visited-site AI capability management module, the request carrying first core network tunnel information, fourth core network tunnel information, location information, and regional information of a deployed-site AI capability application module, wherein the first core network tunnel information refers to core network tunnel information of the visited-site AI application module, and the fourth core network tunnel information refers to core network tunnel information of the deployed-site AI application module; An addressing module is used to address the user plane network element of the service transmission channel at the visited location using the location information carried in the AI ​​model scheduling service request, and to address the user plane network element of the service transmission channel at the deployed location using the area information; A first sending module is configured to send a user plane tunnel establishment request carrying the first core network tunnel information and the fourth core network tunnel information to the user plane network element of the visited service transmission channel when it is determined that the user plane network element of the visited service transmission channel and the user plane network element of the deployed service transmission channel are the same network element; The second receiving module is configured to receive the second core network tunnel information sent by the user plane network element of the visited service transmission channel; The second sending module is configured to send the second core network tunnel information to the visited site AI capability application module and the deployed site AI capability application module via the visited site AI capability management module.

28. A visited service transmission channel control plane network element, characterized in that: In cross-region dispatching, it includes: a third receiving module, configured to receive an AI model scheduling service request initiated by the visited-location AI capability management module, the request carrying first core network tunnel information and location information; wherein the first core network tunnel information refers to the core network tunnel information of the visited-location AI application module; An addressing module, used for addressing a user plane network element of a visited service transmission channel according to location information; A third sending module is configured to send a user plane tunnel establishment request to a user plane network element of a visited service transmission channel, wherein the request carries the first core network tunnel information; A fourth receiving module is configured to receive the second core network tunnel information fed back by the user plane network element of the user's visited site service transmission channel, and transmit the information to the visited site AI capability application module via the visited site AI capability management module; A fifth receiving module is configured to receive the third core network tunnel information returned by the visited AI capability management module, wherein the third core network tunnel information is the core network tunnel information of the user plane network element end of the deployed service transmission channel; The fourth sending module is configured to return the third core network tunnel information to the user plane network element of the visited service transmission channel.

29. A deployed AI capability application module, characterized in that: When performing intra-area scheduling, it includes: A first receiving module is configured to receive an AI model scheduling service request initiated by the visited AI capability management module to allocate core network tunnel information of the deployed AI capability application module, wherein the request carries an intelligent computing service identifier and an intelligent computing model identifier; A first allocation module, configured to allocate fourth core network tunnel information; A first sending module, configured to return the fourth core network tunnel information to the visited location AI capability management module; A second receiving module is configured to receive, when the visited site service transmission channel control plane network element determines that the visited site service transmission channel user plane network element and the deployed site service transmission channel user plane network element are the same network element, the second core network tunnel information sent by the visited site AI capability management module, wherein the second core network tunnel information is the core network tunnel information of the visited site service transmission channel user plane network element end; A selection module is used to select an intelligent computing model using an intelligent computing service identifier and an intelligent computing model identifier; The second sending module transmits the selected intelligent computing model copy to the visited AI capability application module.

30. A deployed AI capability application module, characterized in that: When conducting cross-region dispatch, it includes: A third receiving module is configured to receive an AI model scheduling service process initiated by the AI ​​capability management module at the deployment site to request the AI ​​capability application module at the deployment site to allocate core network tunnel information; A second allocation module, configured to allocate fourth core network tunnel information; A third sending module is configured to return the fourth core network tunnel information to the AI ​​capability management module at the deployment site, so that the AI ​​capability management module at the deployment site carries the fourth core network tunnel information, the second core network tunnel information, and the regional information in the AI ​​model scheduling service request and sends it to the service transmission channel control plane network element at the deployment site, wherein the second core network tunnel information refers to the core network tunnel information of the user plane network element end of the visited service transmission channel; The fourth sending module is used to transmit the selected intelligent computing model copy to the visited AI capability application module through the established user-side intelligent computing business model transmission channel.

31. A user plane network element for a visited service transmission channel, characterized in that: When performing intra-area scheduling, it includes: a first receiving module, configured to receive a user plane tunnel establishment request sent by a visited-site service transport channel control plane network element, the request carrying first core network tunnel information and fourth core network tunnel information, wherein the first core network tunnel information refers to the core network tunnel information of the visited-site AI application module, and the fourth core network tunnel information refers to the core network tunnel information of the deployed-site AI application module; A first allocation module, configured to allocate second core network tunnel information; The first sending module is configured to send the second core network tunnel information to a control plane network element of a visited service transmission channel.

32. A user plane network element for a visited service transmission channel, characterized in that: When conducting cross-region dispatch, it includes: a second receiving module, configured to receive a user plane tunnel establishment request sent by a visited service transport channel control plane network element, the request carrying first core network tunnel information, wherein the first core network tunnel information refers to the core network tunnel information of the visited AI application module; A second allocation module, configured to allocate second core network tunnel information; A second sending module sends the second core network tunnel information to the visited service transmission channel control plane network element; A third receiving module is configured to receive third core network tunnel information sent by a control plane network element of a visited service transmission channel, wherein the third core network tunnel information refers to core network tunnel information of a user plane network element end of a deployed service transmission channel; The third sending module forwards the intelligent computing model copy sent by the user-plane network element of the deployed service transmission channel to the AI ​​capability application module at the visited location.

33. A deployed service transmission channel user plane network element, characterized in that: When performing intra-area scheduling, it includes: A first receiving module is configured to receive a user plane tunnel establishment request sent by a visited service transmission channel control plane network element, the request carrying fourth core network tunnel information, wherein the fourth core network tunnel information refers to the core network tunnel information of the AI ​​capability application module end at the deployment site; A distribution module, configured to distribute third core network tunnel information; A first sending module, configured to send the third core network tunnel information to a visited service transmission channel control plane network element; A second receiving module is configured to receive second core network tunnel information sent by a control plane network element of a visited service transmission channel, wherein the second core network tunnel information refers to core network tunnel information at a user plane network element end of the visited service transmission channel; The second sending module is used to forward the intelligent computing model copy sent by the AI ​​capability application module at the deployment site to the AI ​​capability application module at the visited site.

34. A deployed service transmission channel control plane network element, characterized in that: When conducting cross-region dispatch, it includes: A first receiving module is configured to receive an AI model scheduling service request sent by an AI capability management module at a deployment site, the request carrying information about a fourth core network tunnel, information about a second core network tunnel, and information about an area under the responsibility of an AI capability application module at a deployment site, wherein the second core network tunnel information is information about a core network tunnel at a user plane network element end of a visited site service transmission channel, and the fourth core network tunnel information is information about a core network tunnel at a deployment site AI capability application module end; An addressing module is used to address the user plane network element of the deployed service transmission channel according to the area information of the deployed AI capability application module; The first sending module is configured to initiate a user plane tunnel establishment request to a user plane network element of a deployed service transmission channel, wherein the request carries information about the fourth core network tunnel and information about the second core network tunnel; The second receiving module is used to receive the third core network tunnel information returned by the user plane network element of the service transmission channel at the deployment site; The second sending module is used to return the third core network tunnel information to the AI ​​capability management module at the deployment site.

35. A deployed AI capability management module, characterized in that: include: a first receiving module, configured to receive an AI model scheduling service request sent by a visited location AI capability management module, the request carrying second core network tunnel information, wherein the second core network tunnel information is core network tunnel information of a user plane network element end of a visited location service transmission channel; The first sending module is used to initiate an AI model scheduling service request for allocating core network tunnel information of the AI ​​capability application module at the deployment site to the AI ​​capability application module at the deployment site, wherein the request carries an intelligent computing service identifier and an intelligent computing model identifier; a second receiving module, configured to receive fourth core network tunnel information returned by the deployed AI capability application module, wherein the fourth core network tunnel information is core network tunnel information of the deployed AI capability application module; The second sending module is used to send an AI model scheduling service request to the service transmission channel control plane network element of the deployment site, where the request carries the fourth core network tunnel information, the second core network tunnel information, and the area information responsible for the AI ​​capability application module of the deployment site; A third receiving module is configured to receive third core network tunnel information sent by a control plane network element of a deployed service transmission channel, wherein the third core network tunnel information is core network tunnel information of a user plane network element end of the deployed service transmission channel; The third sending module is used to return the AI ​​model scheduling service request establishment result to the visited AI capability management module, where the result includes: third core network tunnel information.

36. An AI model scheduling system, characterized in that: In the case of same-area scheduling, the device comprises: the visited-site AI capability management module according to claim 25, the visited-site AI capability application module according to claim 26, the visited-site service transmission channel control plane network element according to claim 27, the deployed-site AI capability application module according to claim 29, the visited-site service transmission channel user plane network element according to claim 31, and the deployed-site service transmission channel user plane network element according to claim 33; During cross-region scheduling, the system includes: the visited site AI capability management module according to claim 25, the visited site AI capability application module according to claim 26, the visited site service transmission channel control plane network element according to claim 28, the visited site service transmission channel user plane network element according to claim 32, the deployment site service transmission channel control plane network element according to claim 34, the deployment site AI capability application module according to claim 30, and the deployment site AI capability management module according to claim 35.

37. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; Memory for storing computer programs; A processor, configured to implement the method described in any one of claims 7 to 24 when executing a program stored in a memory.

38. A computer storage medium, characterized in that The computer storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 7 to 24 is implemented.

Images