An integrated control method, control device, and control system for heterogeneous networks.

Abstract

An integrated control method, control device, and control system for heterogeneous networks, relating to the fields of data communication and network technology, is disclosed. The method includes: real-time acquisition of a first network demand Rr from network users, inputting it into a Transformer model, and outputting a second network entity command Iv; dividing the second network entity command Iv into network domain commands and network subdomain commands, and sequentially issuing them to the physical network; wherein the Transformer model includes three layers: the upper Transformer maps the diverse first network demand Rr to an integrated representation of a second network demand Ru; the middle Transformer maps the second network demand Ru to an integrated representation of a first network entity command Iu; and the lower Transformer maps the first network entity command Iu to diverse second network entity commands Iv. This application enables integrated control of heterogeneous networks through the Transformer model.

Description

Technical Field

[0001] This application relates to the fields of data communication and network technology, specifically to an integrated control method, control device and control system for heterogeneous networks. Background Technology

[0002] Multimodal networks aim to separate network technology architecture from its supporting environment, maximizing the innovative vitality of emerging application-driven network technology architectures and significantly reducing the barriers to application deployment and service provision of new network architectures. Multimodal networks propose a fully definable network baseline technology concept, constructing a new generation of information and communication network infrastructure based on a multimodal network environment, adapting to various network architectures and vertical industry application scenarios with programmable "plug-and-play" interfaces.

[0003] One widely accepted definition of network modality is that any physical network entity defined according to network technology, industry applications supported by the network, network operation and maintenance system, and network deployment can be abstracted into a network modality that can run in a multimodal network environment through standardized software and hardware interfaces.

[0004] Of the four aspects of network modes mentioned above, multimodal network design and implementation based on network technology architecture is fundamental and is also the focus of current mainstream research and design. Network technology architecture involves addressing methods, routing methods, forwarding behavior, and other aspects. Network modes classified according to this aspect can be called network subdomains.

[0005] From the perspective of technological development, multimodal networks can be seen as a product of the in-depth development of Software Defined Networking (SDN). Unlike SDN, which mainly focuses on defining network routing (flow tables) by the application layer, multimodal networks allow addressing methods, forwarding behavior, network routing, and application scenarios to be defined by the upper layer.

[0006] The goal of multimodal networks is full-dimensional definability, meaning that all features of the network subdomains can be modified by the control surface. However, integrated control of multiple modalities is quite challenging due to the discreteness and diversity of features.

[0007] In addition to the diversity brought about by multimodal networks, other heterogeneous networks also exhibit diverse characteristics. For example, the configurations and instructions of heterogeneous networks are also diverse, which brings considerable difficulties to the integrated control of heterogeneous networks. Furthermore, the large number of network users and their diverse needs make it difficult for heterogeneous networks to collaborate in meeting these diverse needs. Moreover, the independent control of existing heterogeneous networks makes it difficult to manage and schedule resources in a unified manner, making it very difficult to allocate and utilize network resources rationally and fully realize their value. Summary of the Invention

[0008] This application provides an integrated control method, control device, and control system for heterogeneous networks, which can solve the technical problem that it is difficult to achieve integrated control of heterogeneous networks in the prior art.

[0009] In a first aspect, embodiments of this application provide an integrated control method for heterogeneous networks, the method comprising:

[0010] The first network demand Rr of network users is obtained in real time and input into the Transformer model, and the second network entity instruction Iv is output.

[0011] The second network entity instruction Iv is divided into network domain instructions and network subdomain instructions and sent to the physical network in sequence.

[0012] The Transformer model includes three layers: the upper layer maps the diverse first network requirement Rr to the unified representation of the second network requirement Ru; the middle layer maps the second network requirement Ru to the unified representation of the first network entity instruction Iu; and the lower layer maps the first network entity instruction Iu to the diverse second network entity instruction Iv.

[0013] In conjunction with the first aspect, in one implementation, the Transformer model is obtained through training, and the training set includes a network requirement set and a second network entity instruction set.

[0014] Collect the original network requirements of each network user and describe them according to a preset format to obtain the corresponding first network requirement Rr, forming the network requirement set;

[0015] Collect the network domain instructions and network subdomain instructions corresponding to each original network requirement to form the second network entity instruction set.

[0016] In conjunction with the first aspect, in one implementation, the training set further includes a network resource set, which includes the physical topology of the network and network subdomains.

[0017] In conjunction with the first aspect, in one implementation, before the second network requirement Ru enters the middle-layer Transformer, it further includes:

[0018] Analyze the importance and priority of the demand characteristics of the second network demand Ru;

[0019] When there is only one important requirement feature, pool that requirement feature.

[0020] When there are multiple important demand features, a weighted pool is performed on these multiple important demand features.

[0021] In conjunction with the first aspect, in one implementation, the formula for pooling the demand feature is:

[0022] Pc = Max(Pn normalized to 1) < <n,

[0023] Where Pc represents the demand features after pooling, Pn represents the set of demand features before pooling, and n represents the index of the most important demand feature.

[0024] The formula for weighted pooling is:

[0025] Pc = Σ(Pn*Wn) normalized to one.

[0026] Where Wn represents the weight.

[0027] In conjunction with the first aspect, in one implementation, the first network entity instruction Iu adopts a custom structure description.

[0028] In conjunction with the first aspect, in one implementation, before the first network entity instruction Iu enters the lower-level Transformer, it further includes:

[0029] Analyze the importance and priority of the instruction characteristics of the first network entity instruction Iu;

[0030] Pooling instruction features includes:

[0031] For instruction characteristics of operation domain, object domain, and parameter domain, joint pooling or independent pooling is performed;

[0032] In addition, non-repeating instruction features are selected as pooling features.

[0033] In conjunction with the first aspect, in one implementation, selecting non-repeating instruction features as pooling features includes:

[0034] Repeating factors are selected sequentially in ascending order. If the instruction feature corresponding to the selected repeating factor is repeated, the next repeating factor is selected for iteration until instruction features without repetition are obtained.

[0035] A second aspect provides a control device for an integrated control method of heterogeneous networks, the control device comprising:

[0036] The acquisition module is used to obtain the first network demand Rr of network users in real time;

[0037] The Transformer model takes a first network requirement Rr as input and outputs a second network entity instruction Iv. The Transformer model includes three Transformer layers: the upper Transformer maps the diverse first network requirements Rr to a unified representation of the second network requirements Ru; the middle Transformer maps the second network requirements Ru to a unified representation of the first network entity instruction Iu; and the lower Transformer maps the first network entity instruction Iu to diverse second network entity instructions Iv.

[0038] The issuing module is used to divide the network entity instruction Iv into network domain instructions and network subdomain instructions, and issue them to the physical network in sequence.

[0039] Thirdly, a control system is provided, comprising:

[0040] The aforementioned control device is used to achieve integrated control of heterogeneous networks;

[0041] The interface components include a management interface component and a network interface component. The management interface component is used to provide a control command interface and an operation interface for the control device, and the network interface component is used to provide network domain and network subdomain interfaces.

[0042] A database component is used to store the training dataset of the control device and the Transformer dataset, the Transformer dataset including the Transformer model algorithm and the Transformer model parameters.

[0043] The beneficial effects of the technical solutions provided in this application include:

[0044] The Transformer model automatically maps the first network demand Rr of network users to the second network entity instruction Iv, enabling the network to respond to user needs in real time and quickly meet them, providing appropriate QoS (Quality of Service). The unified representation of the second network demand Ru makes the unified control adaptive to changes in network user needs; the unified representation of the first network entity instruction Iu makes the unified control adaptive to changes in the network, network domain, and network subdomains; the Transformer model enables unified control of heterogeneous networks.

[0045] The Transformer model comprises three Transformer layers, simplifying the overall design. The diverse primary network requirements (Rr) of network users are not directly coupled with the diverse instructions of network domains and subdomains, facilitating independent implementation of each. When a type of input changes, only the corresponding Transformer layer is affected, without fundamentally impacting the entire Transformer model, significantly improving the adaptability and scalability of integrated control. Attached Figure Description

[0046] Figure 1 This is a flowchart illustrating the integrated control method for heterogeneous networks according to an embodiment of this application.

[0047] Figure 2 This is a flowchart illustrating the Transformer model in an embodiment of this application;

[0048] Figure 3 This is a schematic diagram illustrating the integrated characterization and feature parameterization process of network requirements in an embodiment of this application;

[0049] Figure 4 This is a schematic diagram illustrating the demand feature pooling process of a network user group in an embodiment of this application.

[0050] Figure 5 This is a schematic diagram illustrating the integrated representation and parameterization of the first network entity instruction Iu in an embodiment of the present invention;

[0051] Figure 6 This is the first network entity instruction Iu set feature analysis and pooling process in this embodiment of the invention;

[0052] Figure 7 This is a schematic diagram illustrating the integrated control of heterogeneous networks and network subdomains in an embodiment of this application;

[0053] Figure 8 This is a schematic diagram of the logic structure of the control system according to an embodiment of the present invention. Detailed Implementation

[0054] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0055] First, some of the technical terms used in this application will be explained to help those skilled in the art understand this application.

[0056] Transformer is a sequence-to-sequence model based on attention mechanisms. Its core idea is to use self-attention to capture the contextual relationships between different positions in the input sequence.

[0057] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0058] In a first aspect, embodiments of this application provide an integrated control method for heterogeneous networks, the control method comprising:

[0059] The first network demand Rr of network users is obtained in real time and input into the Transformer model. The Transformer model outputs the second network entity instruction Iv.

[0060] The second network entity instruction Iv is divided into network domain instructions and network subdomain instructions, which are then issued to the physical network in sequence.

[0061] like Figure 2 As shown, the above Transformer model includes three Transformer layers. The upper Transformer is used to map the diverse first network requirements Rr to the unified representation of the second network requirements Ru, which can be expressed as:

[0062] Ru = Transformer1(Rr),

[0063] Here, Transformer1 represents the upper-layer Transformer. The upper-layer Transformer implements the process of taking the first network requirement Rr as an integrated representation and Transformer feature parameters. The input and rules of the upper-layer Transformer can be customized according to the actual situation, but the output of the upper-layer Transformer is set.

[0064] The middle-layer Transformer maps the second network requirement Ru to the first network entity instruction Iu of the unified representation, which can be expressed as:

[0065] Iu = Transformer2(Ru),

[0066] Here, Transformer2 represents the middle-layer Transformer. The first network entity instruction can be an instruction of a network, a network domain, or a network subdomain. The structure of the unified representation is predefined, for example:

[0067] Instruction code: I1

[0068] Instruction operation domain: C1

[0069] Instruction data field: D1

[0070] D1 Data Field Object 1:O1

[0071] D1 Data Field Operation 1: A1

[0072] D1 Data Field Parameter 1: P1

[0073] D1 Data Field Object 2: O2

[0074] D1 Data Domain Operation 2: A2

[0075] D1 Data Field Parameter 2: P2

[0076] ......}

[0077] The lower-level Transformer maps the first network entity instruction Iu to the diversified second network entity instruction Iv. This can be represented as:

[0078] Iv = Transformer3(Iu),

[0079] Here, Transformer3 represents the lower-level Transformer. The lower-level Transformer is equivalent to transforming the unified representation of the first network entity instruction Iu into an actual usable entity instruction. The output and rules of the lower-level Transformer can be customized according to the actual situation, but the input of the lower-level Transformer is predefined.

[0080] In this embodiment, the integrated control of heterogeneous networks requires intelligent mapping of diverse needs to diverse instructions or configurations of the heterogeneous networks. Considering the diverse network needs of discrete network users, and the diverse instructions of heterogeneous networks, network domains, and network subdomains, direct mapping is complex, sacrifices scalability, and reduces efficiency. Therefore, this embodiment uses a three-layer Transformer model, where each layer is relatively independent. By integrating the second network requirement Ru and the first network entity instruction Iu, direct mapping is simple and enables integrated intelligent control. If changes caused by the application only involve one layer of Transformer, other layers can remain unchanged, improving scalability and efficiency.

[0081] The Transformer model described above is obtained through training, and the training set includes the network requirement set and the second network entity instruction set.

[0082] By collecting the original network requirements of each network user and describing them according to a preset format, the corresponding first network requirement Rr is obtained, forming the network requirement set.

[0083] The second network entity instruction set is formed by collecting network domain instructions and network subdomain instructions corresponding to each original network requirement.

[0084] In one embodiment, reference is made to Figure 1 , Figure 1 This is a schematic flowchart illustrating an embodiment of the integrated control method for heterogeneous networks proposed in this application. Figure 1 As shown, in this embodiment, in addition to the steps of the above embodiments, the integrated control method also includes the process of collecting the Transformer model training set. This embodiment includes the following steps:

[0085] S01. Physical topology awareness and network subdomain awareness are both forms of network self-awareness. Physical topology awareness obtains the network topology, such as which network devices are present. Network subdomain awareness addresses the question of which forwarding subdomain a specific service flow is carried in. The collected physical topology and network subdomains form a network resource set.

[0086] S02. Network Demand Set Awareness: Various user demand collection methods can be employed to collect the original network demands of network users. Original network demands refer to the initial demands directly from the users. These original network demands are analyzed and described as first network demands Rr according to a preset format. All first network demands Rr form a network demand set. The preset format can be customized according to actual circumstances, for example:

[0087] {Network User: User A}

[0088] Target server: S1

[0089] User requirement: No perceptible buffering or latency when watching videos.

[0090] S03, Network Domain Instruction Set Awareness and Network Subdomain Instruction Set Awareness: The network domain is part of the physical network, and it collects network domain instructions corresponding to various original network requirements. Network subdomain instruction set awareness is equivalent to collecting network subdomain instructions corresponding to various original network requirements. The network domain instruction set and the network subdomain instruction set form the second network entity instruction set.

[0091] S4. Train the Transformer model using the training set formed by the network demand set and the second network entity instruction set to obtain the trained Transformer model.

[0092] S5. Input the first network requirement Rr of network users in real time into the Transformer model, and the Transformer model outputs the second network entity instruction Iv.

[0093] S6. Divide the second network entity instruction Iv into a network domain instruction and a network sub-domain instruction, and sequentially issue them to the physical network in two steps.

[0094] In this embodiment, the above steps S01 - S03 belong to the pre - preparation. Actually, in other embodiments, the Transformer model can be pre - trained in advance, and then the trained Transformer model can be directly used. Steps S5 and S6 can be cycled.

[0095] Further, in one embodiment, the data stream of the Transformer model's perception of network requirements is as Figure 3 shown, representing the change of the data stream from the network user's appeal to the demand characterization of the network user group. It specifically includes four stages. The data in the first stage is the natural representation of the network user's appeal, the data in the second stage is the representation of the demand for network resources corresponding to the network user's appeal (i.e., the network requirement), the data in the third stage is the representation of the network user's network requirement at the QoS level, and the data in the fourth stage is the description of the network requirement using meta - mathematical parameters, that is, feature parameterization, to form the second network requirement Ru.

[0096] Further, as Figure 2 shown, in one embodiment, before the second network requirement Ru enters the middle - layer Transformer, there is also a process of pooling the second network requirement Ru.

[0097] First, conduct a demand feature analysis of the second network requirement Ru to analyze the importance and priority of the demand features. Among them, the importance and priority can be adjusted at any time according to the needs of different users, and the demand feature analysis can determine whether the important demand feature is one or more.

[0098] Then pool the demand features, mainly using two methods:

[0099] One is based on the principle of selecting the most important demand feature. When there is only one important demand feature, pool this demand feature. The formula is:

[0100] Pc = Max(Pn normalized) << n, where Pc represents the demand feature after pooling, Pn represents the set of demand features before pooling, and n represents the serial number of the most important demand feature.

[0101] The other is to comprehensively consider all features. When there are multiple important demand features, perform weighted pooling on multiple important demand features. The formula is:

[0102] Pc = Σ(Pn*Wn) normalized, where Wn represents the weight.

[0103] In this embodiment, because the number of network users is enormous, and the network needs of each user are discrete and independent, the number of demand features is vast. Demand feature pooling can reduce the number of features that need to be processed, and the pooled features are representative. For example... Figure 4 As shown, after pooling the network demand feature matrix of the network user group, it can be clearly seen that the feature matrix becomes smaller.

[0104] Furthermore, in one embodiment, the network domain instruction set awareness in step S03 includes: intelligent analysis of network domain service flows, awareness of network subdomains, awareness of network domain service flow instruction sets, and parameterization of the instruction sets. Specifically, intelligent analysis of network domain service flows is oriented towards a single network user's service, based on the awareness and analysis of a specific service of a particular network user. Awareness of network subdomains and awareness of network domain service flow instruction sets also target a specific service of a particular network user. Parameterization of the instruction sets is a subsequent processing step after awareness is achieved.

[0105] Because heterogeneous networks have a large number of network entity types—the number of network subdomains Nm may reach hundreds, the number of network types Nt may reach more than ten, and the number of network domains Nd may reach dozens—the number of network entities Nn = Nm * Nt * Nd is considerable. Furthermore, with the increasing sophistication of network programmability, the instruction sets for each type of network entity are likely to be very large and complex. Therefore, feature analysis and feature pooling of network entity instruction characteristics are necessary.

[0106] like Figure 2 As shown, in one embodiment, before the first network entity instruction Iu enters the lower-level Transformer, the process of pooling the first network entity instruction Iu is also included.

[0107] First, the instruction characteristics of the first network entity instruction Iu are analyzed to determine their importance and priority. The importance and priority can be adjusted as needed by different users.

[0108] Then, instruction features are pooled to reduce the number of instruction features that need to be processed, and to ensure that the pooled instruction features are representative. Instruction feature pooling includes joint pooling or independent pooling for instruction features in the operation domain, object domain, and parameter domain; it also includes representative domain pooling, which selects instruction features without repetition as pooling features.

[0109] Selecting unique instruction features as pooling features can be done using a hierarchical iterative approach. This includes selecting repetition factors sequentially from smallest to largest. If the selected repetition factor corresponds to a repetitive instruction feature, the next repetition factor is selected for iteration, and so on, until unique instruction features are obtained. For example, the instruction feature with the smallest repetition factor can be selected; if the instruction feature with the smallest repetition factor is repetitive, the instruction feature with the second smallest repetition factor can be selected for iteration, and so on.

[0110] In one embodiment, the Transformer model's perception of changes in network domain instructions and modal instructions includes two stages. The first stage is the service flow requirement stage, where the data consists of instructions that need to be executed to implement a specific service for a particular user. The second stage is the service flow standard instruction stage, where the data is an integrated representation of the aforementioned instructions. As in the above embodiment, the integrated representation and parameterization of the first network entity instruction Iu are as follows: Figure 5 As shown. Figure 6 The diagram illustrates the instruction feature pooling process for the first network entity instruction Iu (including network domain instructions and network subdomain instructions). Before pooling, the network domain and network subdomain instruction features consist of a three-dimensional dataset comprising the operation domain, object domain, and parameter domain. After pooling, based on the needs of the network user group, the instruction features are a selection and simplification of the three-dimensional dataset comprising the operation domain, object domain, and parameter domain.

[0111] Furthermore, the pooling process described above applies to all network requests or all network entity instructions. The first network entity instruction Iu and the second network request Ru represent multiple data of a class, not just one.

[0112] In this integrated control method, intelligent perception, demand feature analysis, and demand feature pooling of network demands enable real-time understanding of network user demands, accurate grasp of user demand characteristics, and minimization of the overhead of processing user demands. Intelligent perception, instruction feature analysis, and instruction feature pooling of network entity instruction sets enable real-time perception of network entity instructions corresponding to service flows, real-time instruction feature analysis, and reduced overhead of processing network instructions.

[0113] Secondly, embodiments of this application also provide a control device for an integrated control method based on heterogeneous networks. For example... Figure 7 As shown, the control device in this embodiment can achieve integrated control of heterogeneous networks and network subdomains. The heterogeneous network includes 1 to x network domains, which are relatively independent of each other. Each network domain can support 1 to y network subdomains, which are also logically relatively independent of each other. Network users provide network requirements to the control device, and the control device provides network commands or configurations. Figure 7Modality 1, Modality 2, and Modality y, which are abstracted from the network domain, are all network subdomains, where x and y both represent ordinal numbers.

[0114] In one embodiment, the control device includes an acquisition module, a Transformer model, and a distribution module.

[0115] The acquisition module is used to obtain the first network demand Rr of network users in real time.

[0116] The Transformer model takes a first network requirement Rr as input and outputs a second network entity instruction Iv. The Transformer model consists of three Transformer layers: the upper Transformer maps the diverse first network requirements Rr to a unified representation of the second network requirement Ru; the middle Transformer maps the second network requirement Ru to a unified representation of the first network entity instruction Iu; and the lower Transformer maps the first network entity instruction Iu to the diverse second network entity instruction Iv.

[0117] The distribution module is used to divide network entity commands (Iv) into network domain commands and network subdomain commands, and then distribute them to the physical network in sequence.

[0118] Furthermore, the control device also includes a demand feature pooling module and an instruction feature pooling module.

[0119] The requirement feature pooling module is located between the upper-layer Transformer and the middle-layer Transformer. It is used to analyze the importance and priority of the requirement features of the second network requirement Ru. When there is only one important requirement feature, the requirement feature is pooled. When there are multiple important requirement features, the multiple important requirement features are weighted and pooled.

[0120] The instruction feature pooling module is located between the middle-layer Transformer and the lower-layer Transformer. It is used to analyze the importance and priority of the instruction features of the first network entity instruction Iu and to pool the instruction features, including: joint pooling or independent pooling for instruction features of operation domain, object domain, and parameter domain; and selecting instruction features without repetition as pooling features.

[0121] The functions of each module in the control device correspond to the steps in the control method embodiments described above, and their functions and implementation processes will not be elaborated here.

[0122] Thirdly, such as Figure 8As shown, this application also provides an embodiment of a control system, including the control device described in the above embodiment, which is used to realize integrated control of heterogeneous networks. Furthermore, the control system also includes an interface component and a database component.

[0123] Furthermore, in one embodiment, the interface component includes a management interface component and a network interface component. The management interface component is used to provide the control device with an integrated control command interface and an operation interface; the network interface component is used to provide network domain and network subdomain interfaces.

[0124] The database component stores the control unit's training dataset and the Transformer dataset. The training dataset includes network requirements sets, network resource sets, network domain instruction sets, and network subdomain instruction sets. The Transformer dataset provides the model algorithms and parameters for the Transformer model.

[0125] The control unit includes a model training component that trains the Transformer model through feedback and iteration. It also includes an instruction translation component for maintaining the Transformer model and translating network control commands.

[0126] The control system in this embodiment can realize the network user's network demand perception, maintain the Transformer model, and achieve integrated control of heterogeneous networks by combining a three-layer Transformer with an integrated representation of network demand and network entity instructions.

[0127] It should be noted that the sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0128] The terms "comprising" and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. The terms "first," "second," and "third," etc., are used to distinguish different objects, etc., and do not indicate a sequence, nor do they limit "first," "second," and "third" to different types.

[0129] In the description of the embodiments of this application, terms such as "exemplary," "for example," or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary," "for example," or "for instance" is intended to present the relevant concepts in a concrete manner.

[0130] In the description of the embodiments of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more.

[0131] In some processes described in the embodiments of this application, multiple operations or steps are included in a specific order. However, it should be understood that these operations or steps may not be executed in the order they appear in the embodiments of this application, or they may be executed in parallel. The sequence number of the operation is only used to distinguish different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.

[0132] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device to execute the methods described in the various embodiments of this application.

[0133] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. An integrated control method for heterogeneous networks, characterized in that, The method includes: The first network demand Rr of network users is obtained in real time and input into the Transformer model, and the second network entity instruction Iv is output. The second network entity instruction Iv is divided into network domain instructions and network subdomain instructions and sent to the physical network in sequence. The Transformer model includes three layers: the upper layer maps the diverse first network requirement Rr to the unified representation of the second network requirement Ru; the middle layer maps the second network requirement Ru to the unified representation of the first network entity instruction Iu; and the lower layer maps the first network entity instruction Iu to the diverse second network entity instruction Iv.

2. The integrated control method for heterogeneous networks as described in claim 1, characterized in that, The Transformer model is obtained through training, and the training set includes a network requirement set and a second network entity instruction set. Collect the original network requirements of each network user and describe them according to a preset format to obtain the corresponding first network requirement Rr, forming the network requirement set; Collect the network domain instructions and network subdomain instructions corresponding to each original network requirement to form the second network entity instruction set.

3. The integrated control method for heterogeneous networks as described in claim 2, characterized in that, The training set also includes a network resource set, which includes the network's physical topology and network subdomains.

4. The integrated control method for heterogeneous networks as described in claim 2, characterized in that: Before the second network requirement Ru enters the middle-layer Transformer, it also includes: Analyze the importance and priority of the demand characteristics of the second network demand Ru; When there is only one important requirement feature, pool that requirement feature. When there are multiple important demand features, a weighted pool is performed on these multiple important demand features.

5. The integrated control method for heterogeneous networks as described in claim 4, characterized in that, The formula for pooling this demand feature is as follows: Pc = Max(Pn normalized to 1) < <n, Where Pc represents the demand features after pooling, Pn represents the set of demand features before pooling, and n represents the index of the most important demand feature. The formula for weighted pooling is: Pc = Σ(Pn*Wn) normalized to one. Where Wn represents the weight.

6. The integrated control method for heterogeneous networks as described in claim 2, characterized in that, The first network entity instruction Iu uses a custom structure description.

7. The integrated control method for heterogeneous networks as described in claim 6, characterized in that, Before the first network entity instruction Iu enters the lower-level Transformer, it also includes: Analyze the importance and priority of the instruction characteristics of the first network entity instruction Iu; Pooling instruction features includes: For instruction characteristics of operation domain, object domain, and parameter domain, joint pooling or independent pooling is performed; In addition, non-repeating instruction features are selected as pooling features.

8. The integrated control method for heterogeneous networks as described in claim 7, characterized in that, Select non-repeating instruction features as pooling features, including: Repeating factors are selected sequentially in ascending order. If the instruction feature corresponding to the selected repeating factor is repeated, the next repeating factor is selected for iteration until instruction features without repetition are obtained.

9. The control device for the integrated control method of heterogeneous networks as described in any one of claims 1-8, characterized in that, The control device includes: The acquisition module is used to obtain the first network demand Rr of network users in real time; The Transformer model takes a first network requirement Rr as input and outputs a second network entity instruction Iv. The Transformer model includes three Transformer layers: the upper Transformer maps the diverse first network requirements Rr to a unified representation of the second network requirements Ru; the middle Transformer maps the second network requirements Ru to a unified representation of the first network entity instruction Iu; and the lower Transformer maps the first network entity instruction Iu to diverse second network entity instructions Iv. The issuing module is used to divide the network entity instruction Iv into network domain instructions and network subdomain instructions, and issue them to the physical network in sequence.

10. A control system, characterized in that, include: The control device as described in claim 9 is used to realize integrated control of heterogeneous networks; The interface components include a management interface component and a network interface component. The management interface component is used to provide a control command interface and an operation interface for the control device, and the network interface component is used to provide network domain and network subdomain interfaces. A database component is provided for storing the control device training dataset and the Transformer dataset, the Transformer dataset including the Transformer model algorithm and the Transformer model parameters.

Images