Cross-domain model training method, device, network device, readable storage medium and program product

Abstract

The present application relates to a cross-domain model training method, device, network equipment, readable storage medium and program product, and relates to the field of wireless communication technology. The present application can improve the utilization rate of network resources. The method includes: obtaining a first model training request sent by a training request end, matching a model training server and each model training client in a network element of each network domain according to the model training requirement information in the first model training request, the model training server and each model training client belong to different network domains, sending a second model training request to the model training server according to the model training requirement information and each model training client, the model training server is used to trigger each model training client to perform local training of the target model according to the model training requirement information, obtaining the model training result of the target model according to the model training requirement information and the local model information sent by each model training client, and sending the model training result to the training request end.

Description

Technical Field

[0001] The present application relates to the field of wireless communication technology, and in particular to a cross-domain model training method, apparatus, network equipment, computer-readable storage medium, and computer program product. Background Art

[0002] Mobile communication systems are mainly based on the standard architecture developed by 3GPP (3rd Generation Partnership Project), and most of the implementation process relies on dedicated hardware devices. This traditional architecture has significant advantages in performance and stability, but the dedicated and closed nature of hardware resources limits the flexibility and scalability of the system.

[0003] With the rise of artificial intelligence (AI) technology, many studies have been conducted on using AI and other technologies to optimize communication systems.

[0004] The current artificial intelligence in mobile communication systems is one-to-one oriented towards application scenarios, and each working group independently carries out artificial intelligence design, which results in the superposition of artificial intelligence functions in the network, and artificial intelligence capabilities such as computing power and models cannot be coordinated across domains (across network domains).

[0005] With current technology, network domains such as wireless access networks, core networks, network management systems, and the cloud will deploy computing resources with different capabilities. However, based on current technology, these computing powers are difficult to coordinate across domains, and there is a technical problem of low network resource utilization. Summary of the invention

[0006] Based on this, it is necessary to provide a cross-domain model training method, device, network equipment, computer-readable storage medium and computer program product to address the above-mentioned technical problems.

[0007] In a first aspect, the present application provides a cross-domain model training method, including:

[0008] Obtaining a first model training request sent by a training request end; the first model training request includes model training requirement information;

[0009] According to the model training requirement information, a model training server and each model training client are matched in the network elements of each network domain; the model training server and the model training client belong to different network domains;

[0010] According to the model training requirement information and each of the model training clients, sending a corresponding second model training request to the model training server;

[0011] Among them, the model training server is used to trigger each of the model training clients to perform local training of the target model according to the model training requirement information, obtain the model training result of the target model according to the model training requirement information and the local model information sent by each of the model training clients, and send the model training result to the training request end.

[0012] In one of the embodiments, according to the model training requirement information, a model training server and each model training client are matched in the network elements of each network domain, including: according to the model training requirement information, a network element having model training capability information satisfying the model training requirement information is matched in the network element maintenance information as the model training server and each model training client; wherein the network element maintenance information includes the model training capability information of the network elements of each of the network domains maintained.

[0013] In one of the embodiments, before matching the network element having the model training capability information that meets the model training requirement information in the network element maintenance information as the model training server and each of the model training clients according to the model training requirement information, the method further includes: sending a capability information acquisition request to the network element with artificial intelligence capability in each of the network domains; receiving the model training capability information returned by the network element with artificial intelligence capability in each of the network domains; and obtaining the network element maintenance information according to the model training capability information.

[0014] In one of the embodiments, according to the model training requirement information, a model training server and each model training client are matched in the network elements of each network domain, including: if the model training requirement information includes network element designation information of a first target network element used as the model training server and each second target network element used as the each model training client, and the model training capability information of the first target network element and each second target network element meets the model training requirement information, then the model training server and each model training client are obtained.

[0015] In one of the embodiments, the model training server and the model training clients are matched in the network elements of each network domain according to the model training requirement information, including: if the model training requirement information includes network element designation information of a first target network element used as the model training server and second target network elements used as the model training clients, and the model training capability information of one or more network elements among the first target network element and the second target network elements does not meet the model training requirement information, then the model training server and / or the model training client are re-matched in the network elements of each of the network domains.

[0016] In one of the embodiments, according to the model training requirement information, the model training server and the model training clients are matched in the network elements of each network domain, including: if the model training requirement information contains a target network domain combination identifier, the target network domain combination is determined from multiple network domain combinations according to the target network domain combination identifier; the target network domain combination includes a first network domain where the model training server is located and a second network domain where the model training clients are located; according to the model training requirement information, the model training server is matched in the network elements of the first network domain; according to the model training requirement information, the model training clients are matched in the network elements of the second network domain.

[0017] In one of the embodiments, the model training requirement information includes data source requirement information; matching each model training client in the network element of the second network domain according to the model training requirement information includes: matching each model training client in the network element of the second network domain according to the data source requirement information and the model training capability information of the network element of the second network domain; the data source requirement information is used to indicate the data source required for local training of the target model.

[0018] In a second aspect, the present application also provides a cross-domain model training method, including:

[0019] Receive a second model training request sent by the intelligent orchestration and management function; the second model training request is obtained by the intelligent orchestration and management function according to model training requirement information and each model training client; the model training server and the model training client belong to different network domains; the model training server and each model training client are matched in network elements of each network domain by the intelligent orchestration and management function according to the model training requirement information; the model training requirement information is obtained by the intelligent orchestration and management function according to the first model training request sent by the training request end; trigger each model training client to perform local training of the target model according to the model training requirement information; obtain the model training result of the target model according to the model training requirement information and the local model information sent by each model training client; send the model training result to the training request end.

[0020] In a third aspect, the present application also provides a cross-domain model training device, including:

[0021] A request acquisition module, used to acquire a first model training request sent by a training request end; the first model training request includes model training requirement information;

[0022] A training matching module, used to match a model training server and each model training client in a network element of each network domain according to the model training requirement information; the model training server and the model training client belong to different network domains;

[0023] A request sending module, used for sending a corresponding second model training request to the model training server according to the model training requirement information and each of the model training clients;

[0024] Among them, the model training server is used to trigger each of the model training clients to perform local training of the target model according to the model training requirement information, obtain the model training result of the target model according to the model training requirement information and the local model information sent by each of the model training clients, and send the model training result to the training request end.

[0025] In a fourth aspect, the present application also provides a cross-domain model training device, including:

[0026] A request receiving module, used to receive a second model training request sent by the intelligent orchestration and management function; the second model training request is obtained by the intelligent orchestration and management function according to the model training requirement information and each model training client; the model training server and the model training client belong to different network domains; the model training server and each model training client are obtained by the intelligent orchestration and management function by matching in the network elements of each network domain according to the model training requirement information; the model training requirement information is obtained by the intelligent orchestration and management function according to the first model training request sent by the training request end;

[0027] A training trigger module, used to trigger each of the model training clients to perform local training of the target model according to the model training requirement information;

[0028] A result obtaining module, used for obtaining the model training result of the target model according to the model training requirement information and the local model information sent by each of the model training clients;

[0029] The result sending module is used to send the model training result to the training request end.

[0030] In a fifth aspect, the present application further provides a network device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0031] Obtain a first model training request sent by a training request end; the first model training request includes model training requirement information; according to the model training requirement information, match a model training server and each model training client in the network elements of each network domain; the model training server and the model training client belong to different network domains; according to the model training requirement information and each model training client, send a corresponding second model training request to the model training server; wherein the model training server is used to trigger each model training client to perform local training of a target model according to the model training requirement information, obtain a model training result of the target model according to the model training requirement information and the local model information sent by each model training client, and send the model training result to the training request end.

[0032] In a sixth aspect, the present application further provides a network device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0033] Receive a second model training request sent by the intelligent orchestration and management function; the second model training request is obtained by the intelligent orchestration and management function according to model training requirement information and each model training client; the model training server and the model training client belong to different network domains; the model training server and each model training client are matched in network elements of each network domain by the intelligent orchestration and management function according to the model training requirement information; the model training requirement information is obtained by the intelligent orchestration and management function according to the first model training request sent by the training request end; trigger each model training client to perform local training of the target model according to the model training requirement information; obtain the model training result of the target model according to the model training requirement information and the local model information sent by each model training client; send the model training result to the training request end.

[0034] In a seventh aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0035] Obtain a first model training request sent by a training request end; the first model training request includes model training requirement information; according to the model training requirement information, match a model training server and each model training client in the network elements of each network domain; the model training server and the model training client belong to different network domains; according to the model training requirement information and each model training client, send a corresponding second model training request to the model training server; wherein the model training server is used to trigger each model training client to perform local training of a target model according to the model training requirement information, obtain a model training result of the target model according to the model training requirement information and the local model information sent by each model training client, and send the model training result to the training request end.

[0036] In an eighth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0037] Receive a second model training request sent by the intelligent orchestration and management function; the second model training request is obtained by the intelligent orchestration and management function according to model training requirement information and each model training client; the model training server and the model training client belong to different network domains; the model training server and each model training client are matched in network elements of each network domain by the intelligent orchestration and management function according to the model training requirement information; the model training requirement information is obtained by the intelligent orchestration and management function according to the first model training request sent by the training request end; trigger each model training client to perform local training of the target model according to the model training requirement information; obtain the model training result of the target model according to the model training requirement information and the local model information sent by each model training client; send the model training result to the training request end.

[0038] In a ninth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0039] Obtain a first model training request sent by a training request end; the first model training request includes model training requirement information; according to the model training requirement information, match a model training server and each model training client in the network elements of each network domain; the model training server and the model training client belong to different network domains; according to the model training requirement information and each model training client, send a corresponding second model training request to the model training server; wherein the model training server is used to trigger each model training client to perform local training of a target model according to the model training requirement information, obtain a model training result of the target model according to the model training requirement information and the local model information sent by each model training client, and send the model training result to the training request end.

[0040] In a tenth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0041] Receive a second model training request sent by the intelligent orchestration and management function; the second model training request is obtained by the intelligent orchestration and management function according to model training requirement information and each model training client; the model training server and the model training client belong to different network domains; the model training server and each model training client are matched in network elements of each network domain by the intelligent orchestration and management function according to the model training requirement information; the model training requirement information is obtained by the intelligent orchestration and management function according to the first model training request sent by the training request end; trigger each model training client to perform local training of the target model according to the model training requirement information; obtain the model training result of the target model according to the model training requirement information and the local model information sent by each model training client; send the model training result to the training request end.

[0042] The above cross-domain model training method, device, network equipment, computer-readable storage medium and computer program product obtain the first model training request sent by the training request end, and match the model training server and each model training client in the network element of each network domain according to the model training requirement information in the first model training request. The model training server and the model training client belong to different network domains. According to the model training requirement information and each model training client, the corresponding second model training request is sent to the model training server. The model training server is used to trigger each model training client to perform local training of the target model according to the model training requirement information, and obtain the model training result of the target model according to the model training requirement information and the local model information sent by each model training client, and send the model training result to the training request end, thereby realizing cross-network domain model training, which can fully utilize and coordinate the computing resources in different network domains, improve the utilization rate of network resources, improve the network intelligent service capability and the flexibility of computing power call, and can also ensure that the local original data does not leave the network domain, and ensure data privacy and security. In addition, the method of the present application is based on the current network architecture evolution and has good compatibility with the current technical architecture. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 An application environment diagram of a cross-domain model training method in one embodiment;

[0045] Figure 2 A schematic diagram of a process of a cross-domain model training method in one embodiment;

[0046] Figure 3 A flowchart of steps for obtaining network element maintenance information in one embodiment;

[0047] Figure 4 A flowchart of steps for matching a server and a client in one embodiment;

[0048] Figure 5 A schematic diagram of a flow chart of a cross-domain model training method in another embodiment;

[0049] Figure 6 A schematic diagram of an interactive process of a cross-domain model training method in one embodiment;

[0050] Figure 7It is a structural block diagram of a cross-domain model training device in one embodiment;

[0051] Figure 8 It is a structural block diagram of a cross-domain model training device in another embodiment;

[0052] Fig. 9 FIG. 4 is a diagram showing the internal structure of a network device in one embodiment. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0054] Some explanations of terms related to this application:

[0055] End-to-end AI orchestration and management function (E2EAMO): can be deployed on the network management system, logically connecting the artificial intelligence functions of network domains such as wireless access network, core network and cloud, and is responsible for the unified orchestration and management of network intelligence.

[0056] Radio Access Network: Radio Access Network, RAN.

[0057] Core Network: Core Network, CN.

[0058] Network Data Analytics Function (NWDAF): The network element in 5GC (5G core network) that is responsible for collecting, analyzing and processing data generated in the network.

[0059] Open Radio Access Network Alliance: Open Radio Access Network Alliance, O-RAN.

[0060] The integration of network and intelligence is one of the important directions of 6G network evolution. As network computing power deployment becomes more extensive and decentralized, how to coordinate computing resources in different domains in the network has become an important challenge in 6G network design. There are significant gaps in deployment location, functional architecture, hardware standards, etc. among wireless access networks, core networks, and network management systems. Realizing cross-domain AI collaboration from the perspective of AI management and orchestration will be a requirement for 6G system architecture design, and it is also the key to improving the overall intelligence level of the network, optimizing resource utilization, and improving user experience.

[0061] In the current network, wireless access networks, core networks, network management systems and the cloud all deploy computing resources of different capabilities. However, based on the current network architecture, these computing powers are difficult to collaborate across network domains, resulting in poor availability of computing resources, weak end-to-end intelligent service capabilities, and difficulty in ensuring data privacy and security.

[0062] The cross-domain model training method of the present application can connect the intelligent capabilities of distributed deployment in the network through end-to-end intelligent orchestration and management functions. The end-to-end intelligent orchestration and management functions belong to the network management system and can be deployed in one or more locations in the regional cloud and the central cloud. Through the cross-domain model training method, it can fully utilize and coordinate the computing resources in different network domains, improve the utilization rate of network resources, improve the network's intelligent service capabilities and the flexibility of computing power calls, and ensure that local original data does not leave the network domain, ensuring data privacy and security. In addition, the method of the present application is based on the current network architecture evolution and has good compatibility with the current technical architecture.

[0063] The cross-domain model training method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. The application environment is a communication system, which may include intelligent orchestration and management functions, as well as various network elements in network domains such as the core network domain, wireless access network domain, terminal domain and cloud domain. Among them, the cross-domain model training method of the present application can be applied to the cross-domain intelligent collaboration scenario of 6G intelligent endogenous network, and may involve resource collaboration and management orchestration across network domains such as wireless access network, core network and network management system. In the cross-domain model training method of the present application, the main network elements / functions involved may include but are not limited to traditional 3GPP base stations, OAM (Operation Administration and Maintenance), network data analysis function NWDAF, and may also include RAN intelligent network elements for 6G networks, intelligent network elements of network management systems, and end-to-end intelligent orchestration and management functions. As Figure 1 As shown, the end-to-end intelligent orchestration and management functions can be connected to the intelligent functions of different network domains through service interfaces or dedicated interfaces, and can provide unified network intelligent management and orchestration functions for 6G networks. The intelligent orchestration and management functions can have the intelligent ability of distributed deployment in the connected network. Through the cross-domain model training method of this application, it can fully utilize and coordinate the computing resources in different network domains, improve the utilization rate of network resources, improve the network intelligent service capabilities and the flexibility of computing power call, and ensure that local original data does not leave the network domain, ensuring data privacy and security. In addition, the method of this application is based on the current network architecture evolution and has good compatibility with the current technical architecture.

[0064] The cross-domain model training method of the present application is described below with various embodiments and corresponding drawings.

[0065] In an exemplary embodiment, Figure 2 As shown, the present application provides a cross-domain model training method, which can be applied to Figure 1 The intelligent orchestration and management function in the network management system can be deployed in one or more locations in the regional cloud and the central cloud. The method can include the following steps:

[0066] Step S201, obtaining a first model training request sent by a training request end.

[0067] The training request end may be a network device that initiates model training across network domains. The training request end may send a model training request to the intelligent orchestration and management function, and the model training request may be recorded as a first model training request. The intelligent orchestration and management function may receive the first model training request sent by the training request end. The first model training request may include model training requirement information and other information. The model training requirement information is the requirement information for model training, which may include but is not limited to the accuracy threshold of model training, the maximum training cycle of the model, the maximum training time of the model, and model parameter information.

[0068] Step S202: According to the model training requirement information, a model training server and each model training client are matched in the network elements of each network domain.

[0069] Among them, the intelligent orchestration and management function can match the network elements that can meet the model training demand information in the network elements of each network domain as the model training server and each model training client according to the model training demand information contained in the first model training request. Among them, the model training server can be a network element of the first network domain in each network domain, the model training client can be a network element of the second network domain in each network domain, the number of model training clients can be multiple, and the model training server can be one. Among them, the model training server and the model training client belong to different network domains, such as the model training server belongs to the core network domain, and each model training client belongs to the wireless access network domain.

[0070] Step S203: Send a corresponding second model training request to the model training server according to the model training requirement information and each model training client.

[0071] Among them, the intelligent orchestration and management function can generate a corresponding second model training request according to the model training requirement information and each model training client. The second model training request may include model training requirement information and relevant information for indicating each model training client. The intelligent orchestration and management function can send the second model training request to the model training server.

[0072] Among them, the model training server can be used to trigger each model training client to perform local training of the target model based on the second model training request and the model training requirement information after receiving the second model training request. Among them, the target model refers to the model that the training request end needs to train, which can be specified by the training request end in the first model training request, or determined by the intelligent orchestration and management function, or a model pre-specified by the training request end in the intelligent orchestration and management function. The determination form of the target model is not limited here. The target model can be transmitted to the model training server by the intelligent orchestration and management function in the second model training request. The model training server can initialize the target model first and then transmit it to each model training client, and each model training client trains the target model locally to obtain local model information. The local model information can be used to indicate the results of one or more rounds of training of the target model by each model training client locally. The local model information may include but is not limited to the loss function value. Each model training client can send the local model information to the model training server, and the model training server obtains the model training result of the target model according to the model training requirement information and the local model information sent by each model training client. As an example, the model training result may be training failure or training success. The model training server may send the model training result as a model training response to the training request end. As an example, the model training server may send training failure information to the training request end, or may send training success information including the trained target model to the training request end.

[0073] The cross-domain model training method of this embodiment obtains the first model training request sent by the training request end, and matches the model training server and each model training client in the network element of each network domain according to the model training requirement information in the first model training request. The model training server and the model training client belong to different network domains. According to the model training requirement information and each model training client, the corresponding second model training request is sent to the model training server. The model training server is used to trigger each model training client to perform local training of the target model according to the model training requirement information, and obtain the model training result of the target model according to the model training requirement information and the local model information sent by each model training client, and send the model training result to the training request end, thereby realizing cross-network domain model training, which can fully utilize and coordinate the computing resources in different network domains, improve the utilization rate of network resources, improve the network intelligent service capability and the flexibility of computing power call, and can also ensure that the local original data does not leave the network domain, and ensure data privacy and security. In addition, the method of this embodiment is based on the current network architecture evolution and has good compatibility with the current technical architecture.

[0074] In an exemplary embodiment, step S202 of matching the model training server and each model training client in the network elements of each network domain according to the model training requirement information may include:

[0075] According to the model training requirement information, network elements with model training capability information that meets the model training requirement information are matched in the network element maintenance information as model training servers and model training clients.

[0076] In this embodiment, the intelligent orchestration and management function can maintain network element maintenance information locally. The network element maintenance information is the network element information maintained locally by the intelligent orchestration and management function. The network element maintenance information may include model training capability information of each network element in each network domain maintained locally by the intelligent orchestration and management function. Among them, the model training capability information refers to the relevant information of the model training capability possessed by the network element, such as available computing resources, storage resource information, etc. Among them, the intelligent orchestration and management function can maintain the model training capability of each network element in the network element maintenance information through the identifier (ID) of each network element. Therefore, when the intelligent orchestration and management function needs to match the model training server and each model training client for the training request end, it can match the network element with the model training capability information that meets the model training requirement information in the network element maintenance information as the model training server and each model training client, thereby improving the matching efficiency of the model training server and each model training client, thereby improving the efficiency of model training across network domains.

[0077] In an exemplary embodiment, Figure 3As shown, before matching the network elements with model training capability information that meets the model training requirement information in the network element maintenance information as the model training server and each model training client according to the model training requirement information, the following steps may also be included:

[0078] Step S301: Send a capability information acquisition request to a network element with artificial intelligence capability in each network domain.

[0079] In this step, the intelligent orchestration and management function may send a capability information acquisition request to the network elements with artificial intelligence capabilities in each network domain, and the capability information acquisition request is used to request the network elements with artificial intelligence capabilities in each network domain to return their own model training capability information. The model training capability information may include but is not limited to available computing resources and storage resource information.

[0080] Step S302: receiving model training capability information returned by network elements with artificial intelligence capabilities in each network domain.

[0081] Step S303: Obtain network element maintenance information according to the model training capability information.

[0082] In step S302 and step S303, the intelligent orchestration and management function receives the model training capability information returned by each network element with artificial intelligence capability, and can obtain the network element maintenance information based on the model training capability information and the identification of each network element.

[0083] The solution of this embodiment can be that the intelligent orchestration and management function requests the model training capability information of the network elements with artificial intelligence capabilities in each network domain in advance, so as to obtain the network element maintenance information in the local maintenance of the intelligent orchestration and management function, and provide data support for the rapid matching of the model training server and each model training client.

[0084] In an exemplary embodiment, step S202 of matching the model training server and each model training client in the network elements of each network domain according to the model training requirement information may include:

[0085] If the model training requirement information includes network element designation information of the first target network element used as the model training server and each second target network element used as each model training client, and the model training capability information of the first target network element and each second target network element meets the model training requirement information, then the model training server and each model training client are obtained.

[0086] In this embodiment, the training request end may specify the model training server and each model training client so that the cross-domain model training has a certain flexibility. Among them, the model training requirement information of the first model training request sent by the training request end may include the network element designation information of the first target network element used as the model training server and each second target network element used as each model training client, and the network element designation information may be used to represent the model training server and each model training client specified by the training request end. Among them, the first target network element is the model training server specified by the training request end, that is, the first target network element used as the model training server. The second target network element is the model training client specified by the training request end, that is, the second target network element used as the model training client. Therefore, if the model training requirement information includes the network element designation information of the first target network element used as the model training server and each second target network element used as each model training client, and the model training capability information of the first target network element and each second target network element meets the model training requirement information, then the intelligent orchestration and management function can match the model training server and each model training client for it. Among them, the intelligent orchestration and management function can obtain the model training capability information of the first target network element and each second target network element based on the network element maintenance information maintained locally, thereby determining whether they meet the model training requirement information.

[0087] In an exemplary embodiment, step S202 of matching the model training server and each model training client in the network element of each network domain according to the model training requirement information may also include:

[0088] If the model training requirement information includes network element designation information of the first target network element used as the model training server and the second target network elements used as the model training clients, and the model training capability information of one or more network elements among the first target network element and the second target network elements does not meet the model training requirement information, the model training server and / or model training client are re-matched in the network elements of each network domain.

[0089] In this embodiment, the training request end may specify the model training server and each model training client so that the cross-domain model training has a certain flexibility, and when the model training capability information of one or more networks in the model training server and each model training client specified by the training request end does not meet the model training requirement information, the model training server and / or model training client may be actively adapted for the training request end to improve the reliability of the matching. Wherein, if the model training requirement information includes the network element designation information of the first target network element used as the model training server and each second target network element used as each model training client, and the model training capability information of one or more network elements in the first target network element and each second target network element does not meet the model training requirement information, the intelligent orchestration and management function may re-match the model training server in the network elements of each network domain, or re-match one or more model training clients, or re-match the model training server and one or more model training clients.

[0090] In an exemplary embodiment, Figure 4 As shown, step S202 of matching the model training server and each model training client in the network elements of each network domain according to the model training requirement information may include:

[0091] Step S401: If the model training requirement information includes a target network domain combination identifier, the target network domain combination is determined from multiple network domain combinations according to the target network domain combination identifier.

[0092] In this embodiment, the training request end can specify a combination of the target network domains where the model training server and each model training client are located (referred to as a target network domain combination), which can also make cross-domain model training have a certain degree of flexibility.

[0093] The model training requirement information of the first model training request sent by the training request end may include a target network domain combination identifier, and the target network domain combination identifier may be used to represent the above-mentioned target network domain combination specified by the training request end. The target network domain combination identifier may be represented by a combination identifier (such as combination 1 to combination 4 represent multiple combinations). The target network domain combination may include a first network domain where the model training server is located and a second network domain where each model training client is located.

[0094] Therefore, if the model training requirement information includes a target network domain combination identifier, the intelligent orchestration and management function can determine the target network domain combination from multiple network domain combinations according to the target network domain combination identifier. Among them, the intelligent orchestration and management function can also maintain multiple network domain combinations in advance, and use different combination identifiers to distinguish different network domain combinations. The following Table 1 illustrates some network domain combinations:

[0095] Table 1 Network domain combinations

[0096]

[0097] Among them, ‌gNB‌ is a base station node in the 5G network, and its full name is "‌Next Generation Node B", which is the next generation node B. It is the core component of the 5G network and is responsible for tasks such as allocation and scheduling of wireless resources and maintenance of communication connections.

[0098] Step S402: According to the model training requirement information, a model training server is matched in the network elements of the first network domain; according to the model training requirement information, each model training client is matched in the network elements of the second network domain.

[0099] Thus, the intelligent orchestration and management function can match the model training server in the network elements of the first network domain according to the model training requirement information, and the intelligent orchestration and management function can obtain the model training capability information of the network elements of the first network domain according to the network element maintenance information maintained locally, thereby obtaining the model training server in the network elements of the first network domain. The intelligent orchestration and management function can also match the model training clients in the network elements of the second network domain according to the model training requirement information, and can also obtain the model training capability information of the network elements in the second network domain according to the network element maintenance information maintained locally, thereby obtaining the model training clients in the network elements of the second network domain.

[0100] In an exemplary embodiment, the model training requirement information may further include data source requirement information; in step S402, matching each model training client in the network element of the second network domain according to the model training requirement information may include:

[0101] According to the data source requirement information and the model training capability information of the network elements in the second network domain, each model training client is matched in the network elements in the second network domain. The data source requirement information is used to indicate the data source required for local training of the target model.

[0102] In this embodiment, the training request end may specify the data source required for local training of the target model so that the cross-domain model training can adapt to the data source requirements. The model training requirement information of the first model training request sent by the training request end may also include data source requirement information, which is used to indicate the data source required for local training of the target model.

[0103] Therefore, when matching model training clients, the intelligent orchestration and management function can combine the data source demand information and the model training capability information of the network elements in the second network domain, and match the network elements in the second network domain that have model training capability information that meets the data source demand information and can obtain the corresponding data source data as the model training client.

[0104] In an exemplary embodiment, Figure 5 As shown, a cross-domain model training method is also provided, which can be executed by a model training server, and the model training server can be a network element such as a base station. The method may include the following steps:

[0105] Step S501: receiving a second model training request sent by the intelligent orchestration and management function.

[0106] Among them, the second model training request is obtained by the intelligent orchestration and management function according to the model training requirement information and each model training client; the model training server and the model training client belong to different network domains; the model training server and each model training client are matched in the network elements of each network domain by the intelligent orchestration and management function according to the model training requirement information; the model training requirement information is obtained by the intelligent orchestration and management function according to the first model training request sent by the training request end.

[0107] Step S502: trigger each model training client to perform local training of the target model according to the model training requirement information.

[0108] Step S503, obtaining the model training result of the target model according to the model training requirement information and the local model information sent by each model training client.

[0109] Step S504: Send the model training result to the training request end.

[0110] In this embodiment, the model training server can receive the second model training request sent by the intelligent orchestration and management function, and trigger each model training client to perform local training of the target model according to the model training requirement information in the second model training request. The model training server can first initialize the target model and then transmit it to each model training client, and each model training client trains the target model locally to obtain local model information. The local model information can be used to indicate the results of one or more rounds of local training of the target model by each model training client. The local model information may include but is not limited to the loss function value. Each model training client can send the local model information to the model training server, and the model training server obtains the model training result of the target model according to the model training requirement information and the local model information sent by each model training client. The model training server can send the model training result to the training request end.

[0111] The solution of this embodiment can realize cross-network domain model training under the coordination of intelligent orchestration and management functions, fully utilize and coordinate computing resources in different network domains, improve network resource utilization, enhance network intelligent service capabilities and the flexibility of computing power call, and ensure that local original data does not leave the network domain, thereby protecting data privacy and security.

[0112] In an exemplary embodiment, Figure 6 As shown, a cross-domain model training method is also provided. In this embodiment, the method may include the following steps:

[0113] Steps in the network element information maintenance phase:

[0114] Step S11: Send a capability information acquisition request to the network elements with artificial intelligence capabilities in each network domain.

[0115] Step S12: Receive model training capability information returned by network elements with artificial intelligence capabilities in each network domain; obtain network element maintenance information based on the model training capability information.

[0116] During the network element information maintenance phase, the intelligent orchestration and management function can send capability information acquisition requests to network elements with artificial intelligence capabilities in each network domain to request model training capability information returned by each network element, such as node computing resources, storage resources, and other information. Based on this, the network element maintenance information can be obtained in combination with the network element identifier.

[0117] Steps in the cross-domain model training processing phase:

[0118] Step S21: The training request end sends a first model training request to the intelligent orchestration and management function.

[0119] Step S22: The intelligent orchestration and management function matches the model training server and each model training client for the training request end according to the locally maintained network element maintenance information and model training requirement information. Among them, the model training server can be a management system intelligent network element, and the model training client can be multiple base stations. For example, a network element with sufficient computing resources can be selected as a model training server, etc., and the corresponding network element can be selected as a model training client according to the data source requirement information and model training capability information, etc.

[0120] Step S23: The intelligent orchestration and management function may send model training confirmation information to the model training server to confirm whether the model training server can complete the model training task.

[0121] Step S24: The intelligent orchestration and management function may send model training confirmation information to each model training client to confirm whether each model training client can complete the model training task.

[0122] Step S25: The model training server may return confirmation information of whether the model training task can be completed to the intelligent orchestration and management function. If the model training server cannot complete the model training task, the intelligent orchestration and management function may return to step S22 to re-match the model training server.

[0123] Step S26: Each model training client may return confirmation information of whether the model training task can be completed to the intelligent orchestration and management function. If the model training client (one or more) cannot complete the model training task, the intelligent orchestration and management function may return to step S22 to re-match the model training client (one or more).

[0124] Step S27: The intelligent orchestration and management function sends a corresponding second model training request to the model training server according to the model training requirement information and each model training client to trigger the model training process.

[0125] Step S31: the model training server triggers local training of the target model to each model training client, and transmits model training information such as the initialized target model.

[0126] Step S32: Each model training client collects model training data from the data source respectively.

[0127] Step S33: Each model training client performs training of the target model locally.

[0128] Step S34: Each model training client reports local model information to the model training server.

[0129] Step S35: The model training server may obtain the model training result of the target model according to the local model information sent by each model training client. The model training server may aggregate the local model information into a global model to obtain the model training result of the target model.

[0130] Step S36: The model training server can determine whether the current target model meets the model training requirement information of the training request end. The model training requirement information may include threshold information corresponding to model accuracy requirements, model training cycle, model training duration, etc. For example, if the model accuracy meets the requirements, or the model training cycle or training duration exceeds the given threshold information in the request, the training of the target model can be stopped; if not, step S37 can be executed.

[0131] Step S37: The model training server synchronizes the global model information of the target model to each model training client. The global model information can be the model information obtained by integrating the local model information of the target model, such as the gradient and loss value of the model.

[0132] Step S38: Each model training client updates local model parameters.

[0133] Repeat steps S32 to S38 until the model accuracy meets the requirements, or the model training cycle or training duration exceeds the given threshold information in the request.

[0134] Step S39: The model training server may return the trained target model or output to the training requester.

[0135] In this regard, the current 3GPP network usually only supports the features of federated learning in the core network, and the artificial intelligence capabilities of different network domains cannot be coordinated and interoperable. For example, the computing resources in different network domains can only be used for intelligent scenarios in the domain, and artificial intelligence capabilities cannot be coordinated across domains. At the same time, due to the uneven deployment of network computing resources and the distribution of intelligent demand, the current network cannot achieve cross-domain artificial intelligence collaboration, which may lead to low network resource utilization and weak end-to-end service capabilities.

[0136] The solution of this embodiment can be based on the current network architecture. Through end-to-end intelligent orchestration and management, it can connect the artificial intelligence functions of network domains such as wireless access networks, core networks, and network management systems, and realize unified management and orchestration of the entire network intelligence, which can provide key technical support for the vision of endogenous intelligence in 6G networks. Based on end-to-end intelligent orchestration and management, the cross-domain model training method can be applied to cross-domain federated learning, and the servers and clients participating in federated learning can be flexibly selected to realize cross-domain distributed training of models. When the computing power within the domain cannot support federated learning, the end-to-end intelligent orchestration and management can be used to flexibly select the servers and clients of federated learning, or the training request end can specify the servers and clients of federated learning. On the one hand, this can ensure that local data does not leave the domain and protect the privacy and security of the original data. On the other hand, it can make full use of artificial intelligence resources in different domains through the coordination of models and computing power, realize distributed model training, and further improve the utilization of network resources. To a certain extent, it solves the current problem of difficult coordination of artificial intelligence between network domains, as well as the contradiction between centralized training and data privacy, and promotes the improvement of the capabilities of end-to-end intelligent services. It can further improve the utilization of network resources through cross-domain artificial intelligence, and at the same time provide a feasible path for the endogenous evolution of network intelligence, and provide a solution for building a new generation of intelligent wireless access networks. The solution of this embodiment can be based on the evolution of the current network architecture, has high compatibility and feasibility, and supports computing power coordination across network management systems, wireless access networks, and core networks, and can be applied to next-generation communication systems such as O-RAN.

[0137] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0138] Based on the same inventive concept, the embodiment of the present application also provides a cross-domain model training device for implementing the cross-domain model training method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more cross-domain model training device embodiments provided below can refer to the above limitations on the cross-domain model training method, which will not be repeated here.

[0139] In an exemplary embodiment, Figure 7 As shown, a cross-domain model training device is provided, and the device 700 may include:

[0140] The request acquisition module 701 is used to acquire a first model training request sent by a training request end; the first model training request includes model training requirement information;

[0141] A training matching module 702 is used to match a model training server and each model training client in the network elements of each network domain according to the model training requirement information; the model training server and the model training client belong to different network domains;

[0142] A request sending module 703 is used to send a corresponding second model training request to the model training server according to the model training requirement information and each of the model training clients;

[0143] Among them, the model training server is used to trigger each of the model training clients to perform local training of the target model according to the model training requirement information, obtain the model training result of the target model according to the model training requirement information and the local model information sent by each of the model training clients, and send the model training result to the training request end.

[0144] In an exemplary embodiment, the training matching module 702 is used to match network elements having model training capability information that meets the model training requirement information in the network element maintenance information as the model training server and each of the model training clients according to the model training requirement information; wherein the network element maintenance information includes the model training capability information of the network elements of each of the network domains maintained.

[0145] In an exemplary embodiment, the training matching module 702 is also used to send a capability information acquisition request to the network elements with artificial intelligence capabilities in each of the network domains; receive the model training capability information returned by the network elements with artificial intelligence capabilities in each of the network domains; and obtain the network element maintenance information based on the model training capability information.

[0146] In an exemplary embodiment, the training matching module 702 is used to obtain the model training server and each of the model training clients if the model training requirement information includes network element designation information of a first target network element used as the model training server and each of the second target network elements used as the each model training client, and the model training capability information of the first target network element and each of the second target network elements meets the model training requirement information.

[0147] In an exemplary embodiment, the training matching module 702 is used to re-match the model training server and / or the model training client in the network elements of each of the network domains if the model training requirement information includes network element designation information of a first target network element serving as the model training server and second target network elements serving as the model training clients, and the model training capability information of one or more of the first target network element and the second target network elements does not meet the model training requirement information.

[0148] In an exemplary embodiment, the training matching module 702 is used to determine the target network domain combination from multiple network domain combinations according to the target network domain combination identifier if the model training requirement information includes a target network domain combination identifier; the target network domain combination includes a first network domain where the model training server is located and a second network domain where each of the model training clients is located; according to the model training requirement information, the model training server is matched in the network element of the first network domain; according to the model training requirement information, each of the model training clients is matched in the network element of the second network domain.

[0149] In an exemplary embodiment, the model training requirement information includes data source requirement information; the training matching module 702 is used to match each of the model training clients in the network elements of the second network domain according to the data source requirement information and the model training capability information of the network elements of the second network domain; the data source requirement information is used to indicate the data source required for local training of the target model.

[0150] In an exemplary embodiment, Figure 8 As shown, a cross-domain model training device is provided, and the device 800 may include:

[0151] The request receiving module 801 is used to receive a second model training request sent by the intelligent orchestration and management function; the second model training request is obtained by the intelligent orchestration and management function according to the model training requirement information and each model training client; the model training server and the model training client belong to different network domains; the model training server and each model training client are obtained by the intelligent orchestration and management function according to the model training requirement information in the network elements of each network domain; the model training requirement information is obtained by the intelligent orchestration and management function according to the first model training request sent by the training request end;

[0152] A training trigger module 802 is used to trigger each of the model training clients to perform local training of the target model according to the model training requirement information;

[0153] A result obtaining module 803 is used to obtain the model training result of the target model according to the model training requirement information and the local model information sent by each of the model training clients;

[0154] The result sending module 804 is used to send the model training result to the training request end.

[0155] Each module in the above cross-domain model training device can be implemented in whole or in part by software, hardware, and a combination thereof. Each of the above modules can be embedded in or independent of a processor in a network device in the form of hardware, or can be stored in a memory in a network device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0156] In an exemplary embodiment, a network device is provided. The network device may be a base station, an intelligent network element, etc., and its internal structure diagram may be as follows: Fig. 9 As shown. The network device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the network device is used to provide computing and control capabilities. The memory of the network device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the network device can be used to store data such as network element maintenance information and model training capability information. The input / output interface of the network device is used to exchange information between the processor and an external device. The communication interface of the network device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a cross-domain model training method is implemented.

[0157] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the network device to which the scheme of the present application is applied. The specific network device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0158] In one embodiment, a network device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0159] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0160] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0161] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0162] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0163] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0164] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A cross-domain model training method, characterized in that: Applied to intelligent orchestration and management functions, the method includes: Obtaining a first model training request sent by a training request end; the first model training request includes model training requirement information; If the model training requirement information includes a target network domain combination identifier, the target network domain combination is determined from multiple network domain combinations according to the target network domain combination identifier; the target network domain combination includes a first network domain where a model training server is located and a second network domain where each model training client is located; according to the model training requirement information, the model training server is matched in a network element of the first network domain; according to the model training requirement information, each model training client is matched in a network element of the second network domain; the model training server and the model training client belong to different network domains; each model training client belongs to the same network domain; According to the model training requirement information and each of the model training clients, sending a corresponding second model training request to the model training server; Among them, the model training server is used to trigger each of the model training clients to perform local training of the target model according to the model training requirement information, obtain the model training result of the target model according to the model training requirement information and the local model information sent by each of the model training clients, and send the model training result to the training request end.

2. The method according to claim 1, characterized in that The method further comprises: According to the model training requirement information, network elements having model training capability information that meets the model training requirement information are matched in the network element maintenance information as the model training server and each of the model training clients; wherein the network element maintenance information includes the model training capability information of the network elements of each of the network domains maintained.

3. The method according to claim 2, characterized in that Before matching a network element having model training capability information that meets the model training requirement information in the network element maintenance information as the model training server and each of the model training clients according to the model training requirement information, the method further includes: Sending a capability information acquisition request to a network element with artificial intelligence capability in each of the network domains; Receiving the model training capability information returned by the network element with artificial intelligence capability in each of the network domains; The network element maintenance information is obtained according to the model training capability information.

4. The method according to claim 1, characterized in that: The method further comprises: If the model training requirement information includes network element designation information of the first target network element used as the model training server and the second target network elements used as the model training clients, and the model training capability information of the first target network element and the second target network elements meets the model training requirement information, then the model training server and the model training clients are obtained.

5. The method according to claim 1, characterized in that The method further comprises: If the model training requirement information includes network element designation information of the first target network element used as the model training server and the second target network elements used as the model training clients, and the model training capability information of one or more network elements among the first target network element and the second target network elements does not meet the model training requirement information, the model training server and / or the model training client are re-matched in the network elements in each of the network domains.

6. The method according to claim 1, characterized in that The model training requirement information includes data source requirement information; According to the model training requirement information, each of the model training clients is matched in the network element of the second network domain, including: According to the data source requirement information and the model training capability information of the network elements of the second network domain, each of the model training clients is matched in the network elements of the second network domain; The data source requirement information is used to indicate the data source required for local training of the target model.

7. A cross-domain model training method, characterized in that: Applied to a model training server, the method comprises: Receive a second model training request sent by the intelligent orchestration and management function; the second model training request is obtained by the intelligent orchestration and management function according to the model training requirement information and each model training client; the model training server and the model training client belong to different network domains; each of the model training clients belongs to the same network domain; if the model training requirement information contains a target network domain combination identifier, the intelligent orchestration and management function determines the target network domain combination from multiple network domain combinations according to the target network domain combination identifier; the target network domain combination includes the first network domain where the model training server is located and the second network domain where each of the model training clients is located; according to the model training requirement information, the model training server is matched in the network element of the first network domain; according to the model training requirement information, each of the model training clients is matched in the network element of the second network domain; the model training requirement information is obtained by the intelligent orchestration and management function according to the first model training request sent by the training request end; Triggering each of the model training clients to perform local training of the target model according to the model training requirement information; Obtaining a model training result of the target model according to the model training requirement information and the local model information sent by each of the model training clients; The model training result is sent to the training request end.

8. A cross-domain model training device, characterized in that: Applied to intelligent arrangement and management functions, the device comprises: A request acquisition module, used to acquire a first model training request sent by a training request end; the first model training request includes model training requirement information; A training matching module, used for determining a target network domain combination from multiple network domain combinations according to the target network domain combination identifier if the model training requirement information includes a target network domain combination identifier; the target network domain combination includes a first network domain where a model training server is located and a second network domain where each model training client is located; according to the model training requirement information, matching the model training server in a network element of the first network domain; according to the model training requirement information, matching each model training client in a network element of the second network domain; the model training server and the model training client belong to different network domains; each model training client belongs to the same network domain; A request sending module, used for sending a corresponding second model training request to the model training server according to the model training requirement information and each of the model training clients; Among them, the model training server is used to trigger each of the model training clients to perform local training of the target model according to the model training requirement information, obtain the model training result of the target model according to the model training requirement information and the local model information sent by each of the model training clients, and send the model training result to the training request end.

9. A cross-domain model training device, characterized in that: Applied to a model training server, the device comprises: A request receiving module, used to receive a second model training request sent by an intelligent orchestration and management function; the second model training request is obtained by the intelligent orchestration and management function according to model training requirement information and each model training client; the model training server and the model training client belong to different network domains; each of the model training clients belongs to the same network domain; if the model training requirement information contains a target network domain combination identifier, the intelligent orchestration and management function determines the target network domain combination from multiple network domain combinations according to the target network domain combination identifier; the target network domain combination includes a first network domain where the model training server is located and a second network domain where each of the model training clients is located; according to the model training requirement information, the model training server is matched in a network element of the first network domain; according to the model training requirement information, each of the model training clients is matched in a network element of the second network domain; the model training requirement information is obtained by the intelligent orchestration and management function according to the first model training request sent by the training request end; A training trigger module, used to trigger each of the model training clients to perform local training of the target model according to the model training requirement information; A result obtaining module, used for obtaining the model training result of the target model according to the model training requirement information and the local model information sent by each of the model training clients; The result sending module is used to send the model training result to the training request end.

10. A network device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 or claim 7 are implemented.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 or claim 7 are implemented.

12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 or claim 7 are implemented.

Images