A network traffic regulation system, method, and storage medium

Abstract

The present invention provides a network traffic regulation system, method and storage medium, which relate to the field of network communication technology; the network traffic regulation system includes: an information collection unit for obtaining historical traffic data of all media plane network elements connected to the control plane network element and the topological relationship of the media plane network elements; an AI model update unit for updating the traffic regulation model of the control plane network element based on the historical traffic data and the topological relationship; a policy generation unit for obtaining the traffic distribution strategy of the control plane network element; and a policy execution unit for dynamically adjusting the media traffic of each media plane network element according to the traffic distribution strategy. The present invention realizes dynamic regulation of network traffic by real-time collection and analysis of network data, combined with the dynamic update of the AI model, can quickly adapt to changes in network conditions, and improves the flexibility and efficiency of traffic regulation.

Description

Technical Field

[0001] The present invention relates to the field of network communication technology, and in particular to a network traffic regulation system, method and storage medium. Background Art

[0002] With the development of 5G networks, 5G services are becoming more and more widespread. Among them, the 5G new communication service is a new type of real-time audio and video communication service derived from IMS data channel technology. During a call, it establishes an IMS data channel for both parties to transmit a variety of multimedia files, such as music, videos, pictures, documents, and mini-programs. The 5G new communication service can also be superimposed with XR technology to derive more personalized applications, such as video XR effects and dynamic XR backgrounds, bringing a new communication experience to users. However, since the 5G new communication service can transmit a variety of different multimedia files, some of which require large network traffic to transmit, while some require small network traffic to transmit. As a result, how to reasonably allocate network traffic has become a technical problem that needs to be solved urgently.

[0003] In related technologies, multimedia files are transmitted through media plane units. In order to ensure the transmission effect of multimedia files such as streaming media files, it is usually necessary to allocate a fixed media plane unit traffic based on the priority of each service. For example, the higher the priority, the greater the allocated traffic. However, fixed traffic allocation may cause certain low-priority services in the media plane unit to be unable to obtain sufficient bandwidth resources for a long time, resulting in their transmission performance being seriously affected, which in turn affects the transmission effect of low-priority multimedia files. Summary of the Invention

[0004] The problem solved by the present invention is how to improve the traffic regulation effect of 5G new communication services to improve the transmission effect of various multimedia files.

[0005] To solve the above problems, the present invention provides a network traffic control system, method and storage medium.

[0006] In a first aspect, the present invention provides a network traffic control system, comprising:

[0007] An information collection unit, configured to obtain historical traffic data of all media plane network elements connected to the control plane network element and the topological relationship of the media plane network elements;

[0008] An AI model updating unit, configured to update a traffic control model of the control plane network element according to the historical traffic data and the topological relationship, wherein the traffic control model includes a deep reinforcement learning sub-model, a fusion sub-model, a graph neural network sub-model, and a Transformer sub-model;

[0009] A strategy generation unit is configured to input the current traffic data of the current time period into the updated traffic control model, perform time series prediction through the fusion sub-model, and obtain a time series prediction result of the current traffic data; perform topological analysis through the graph neural network sub-model to obtain a graph structure of the media plane network element; perform weight distribution through the Transformer sub-model to obtain an allocation weight strategy for each media plane network element; and then perform reinforcement learning through the deep reinforcement learning sub-model based on the time series prediction result, the graph structure, and the allocation weight strategy to obtain a traffic distribution strategy for the control plane network element;

[0010] A policy execution unit is configured to dynamically adjust the media traffic of each media plane network element according to the traffic distribution policy.

[0011] Optionally, the information collection unit is specifically configured to:

[0012] According to a preset collection cycle, the operating status data of all the media plane network elements within a preset time period is obtained from historical data, wherein the operating status data includes CPU occupancy and GPU occupancy;

[0013] Determining network bandwidth occupancy data of the media plane network element based on data flow inflow and outflow information of the network interface of the media plane network element in the historical data;

[0014] Obtaining load data of the media plane network elements according to the signaling throughput data of all the media plane network elements in the historical data;

[0015] The operating status data, the network bandwidth occupancy data, and the load data are used as the historical traffic data.

[0016] Optionally, the AI model updating unit is specifically configured to:

[0017] Inputting the historical traffic data and the topological relationship into the traffic control model for training;

[0018] Verify the traffic distribution strategy output by the traffic control model during training through a digital twin environment and A / B testing;

[0019] When the traffic distribution strategies output by the traffic control model during training meet the test requirements of the digital twin environment and the A / B test, the training of the traffic control model is stopped to obtain the updated traffic control model.

[0020] Optionally, the policy generating unit is specifically configured to:

[0021] Obtaining the operating status data, the network bandwidth occupancy data, and the load data of the media plane network element within the current time period, and inputting the operating status data, the network bandwidth occupancy data, and the load data as the current traffic data into the updated traffic control model;

[0022] Perform feature extraction on the current traffic data through the CNN network of the fusion sub-model to obtain spatial features of the current traffic data;

[0023] Perform feature extraction on the current traffic data according to the BiLSTM network of the fusion sub-model to obtain a time series of the current traffic data;

[0024] According to the spatial features and the time series, a traffic change trend of the media plane network element in a preset future time period is obtained, and the traffic change trend is used as the time series prediction result.

[0025] Optionally, the policy generating unit is specifically configured to:

[0026] Inputting the topological relationship of the media plane network elements in the current time period into the updated traffic control model;

[0027] The graph structure is constructed according to the topological relationship through the graph neural network sub-model, wherein the nodes of the graph structure are the media plane network elements, and the edges of the graph structure are the connection relationships between the media plane network elements.

[0028] Optionally, the policy generating unit is specifically configured to:

[0029] Obtaining the operating status data, the network bandwidth occupancy data, and the load data of the media plane network element within the current time period, and inputting the operating status data, the network bandwidth occupancy data, and the load data as the current traffic data into the updated traffic control model;

[0030] Using the Transformer sub-model, prediction is performed based on the current traffic data to obtain critical values of multiple preset indicators of each media plane network element within a preset future time period;

[0031] A weight value of each of the preset indicators is generated according to the critical value to obtain the allocation weight strategy of each of the media plane network elements.

[0032] Optionally, the policy generating unit is specifically configured to:

[0033] Performing reinforcement learning based on the time series prediction results, the graph structure, and the weight allocation strategy through the deep reinforcement learning sub-model;

[0034] wherein, generating an initial traffic distribution adjustment strategy based on the time series prediction result, the graph structure, and the distribution weight strategy;

[0035] Taking the initial traffic distribution adjustment strategy as an action and the digital twin environment as an environment, calculating a reward value after the action interacts with the environment through a reward function;

[0036] Whether the reinforcement learning of the initial traffic distribution adjustment strategy is completed is determined according to the reward value, and when the reinforcement learning of the initial traffic distribution adjustment strategy is completed, the traffic distribution strategy of the control plane network element is obtained.

[0037] Optionally, the system further includes an early warning unit, which is configured to:

[0038] Obtaining a flow rate change trend within the current time period based on the current flow rate data of the current time period;

[0039] According to the traffic change trend, determining whether there is a sudden increase in traffic on the media plane network element;

[0040] When the traffic surge occurs in the media plane network element, an early warning message is sent.

[0041] In a second aspect, the present invention provides a network traffic control method, comprising:

[0042] Obtaining historical traffic data of all media plane network elements connected to the control plane network element and the topological relationship of the media plane network elements;

[0043] Update the traffic control model of the control plane network element according to the historical traffic data and the topological relationship, wherein the traffic control model includes a deep reinforcement learning sub-model, a fusion sub-model, a graph neural network sub-model, and a Transformer sub-model;

[0044] Input the current traffic data of the current time period into the updated traffic control model, perform time series prediction through the fusion sub-model, and obtain the time series prediction result of the current traffic data; perform topological analysis through the graph neural network sub-model to obtain the graph structure of the media plane network element; perform weight allocation through the Transformer sub-model to obtain the allocation weight strategy for each media plane network element; then perform reinforcement learning through the deep reinforcement learning sub-model based on the time series prediction result, the graph structure, and the allocation weight strategy to obtain the traffic allocation strategy of the control plane network element;

[0045] The media traffic of each media plane network element is dynamically adjusted according to the traffic distribution strategy.

[0046] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the network traffic control method described above is implemented.

[0047] The network traffic regulation system, method, and storage medium of the present invention achieve joint regulation of media plane network elements through an innovative multi-module collaborative mechanism, effectively solving the limitations of the fixed traffic distribution method in the existing technology. The information collection unit regularly collects historical traffic data (such as device CPU / GPU occupancy, short-term or instantaneous traffic, bandwidth occupancy, etc.) and the topological relationship of the device from the media plane network elements, providing a detailed data foundation for subsequent model training; enabling the system to fully understand the historical status of the network and the connection relationship between devices, thereby providing a basis for dynamic traffic allocation. The AI model update unit continuously updates the traffic regulation model based on historical traffic data and topological relationships. The model includes a deep reinforcement learning sub-model, a fusion sub-model, a graph neural network sub-model, and a Transformer sub-model. The synergy of these sub-models enables the system to adapt to different network environments and conditions, thereby improving the flexibility and efficiency of traffic distribution. The policy generation unit uses the fusion sub-model to perform time series predictions, which can predict traffic trends in advance and provide forward-looking guidance for traffic control. The graph neural network sub-model analyzes the topological relationship of devices, which helps to reasonably allocate network resources. The Transformer sub-model allocates weights based on device performance and network conditions to ensure that critical services are given priority and sufficient bandwidth resources. In addition, the generation of these policies is based on dynamic network conditions and real-time data, which also avoids the unfairness of traditional fixed allocation methods. The policy execution unit dynamically adjusts the traffic of each media plane network element based on the traffic distribution policy obtained by the policy generation unit. By monitoring the network status in real time and quickly adjusting the traffic distribution, the policy execution unit can ensure the stability and efficiency of the network. For example, when the load on a media plane network element is too high, the system can transfer part of its traffic to other devices to achieve load balancing.

[0048] The technical solution of the present invention achieves dynamic regulation of network traffic through real-time collection and analysis of network data, combined with dynamic updates of AI models. This allows for rapid adaptation to changing network conditions, improving the flexibility and efficiency of traffic regulation. Using the Transformer sub-model for weight allocation, the system can rationally allocate bandwidth resources based on service priority and device performance, avoiding prolonged bandwidth shortages for low-priority services and enhancing the fairness of traffic regulation. Furthermore, by analyzing the graph structure using the graph neural network sub-model, a deep understanding of the complex topological relationships between media plane network elements is achieved, enabling coordinated traffic allocation across multiple network elements. This joint regulation approach overcomes the limitations of traditional single-source regulation. By comprehensively considering multiple factors such as timing, topology, and weights, it enables more flexible and precise traffic allocation, effectively avoiding prolonged bandwidth shortages for low-priority services. Furthermore, the system can dynamically adjust traffic allocation based on real-time changes in network conditions and service demands, adapting to dynamic changes in network traffic demand and ensuring the transmission quality of various multimedia files. Furthermore, the deep reinforcement learning sub-model performs reinforcement learning based on timing prediction results, graph structure, and weight allocation strategies to optimize traffic allocation strategies, achieving intelligent and automated traffic regulation. By effectively managing network resources, the service quality degradation caused by traffic congestion is reduced, and the user experience when using high-definition video, XR and other services is improved.

[0049] In summary, the present invention realizes intelligent regulation of 5G new communication service traffic through the synergistic effect of information collection, model update, policy generation and policy execution, significantly improves the dynamics, flexibility, fairness and resource utilization of traffic regulation, thereby solving the problems in the existing technology and achieving the goal of improving the traffic regulation effect of 5G new communication services. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A structural block diagram of a network traffic control system according to an embodiment of the present invention;

[0051] Figure 2 This is a schematic diagram of a 5G new communication network architecture according to an embodiment of the present invention;

[0052] Figure 3 This is a schematic diagram of the connection between the control plane and media plane devices according to an embodiment of the present invention;

[0053] Figure 4 The figure is a flow chart of a network traffic control method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0054] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings. Although certain embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as being limited to the embodiments described herein. Instead, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0055] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0056] The term "including" and its variations used in this document are open inclusions, that is, "including but not limited to"; the term "based on" means "based at least in part on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one other embodiment"; the term "some embodiments" means "at least some embodiments"; the term "optionally" means "optional embodiments". The relevant definitions of other terms will be given in the following description. It should be noted that the concepts of "first", "second", etc. mentioned in the present invention are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0057] It should be noted that the modifications of "one" and "multiple" mentioned in the present invention are illustrative rather than restrictive. Those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0058] The names of the messages or information exchanged between multiple devices in the embodiments of the present invention are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0059] 5G new communication services, based on IMS Data Channel (DC), a framework supporting IP-based multimedia communication services, are emerging real-time audio and video communication services. During a call, an IMS DC is established between two parties to transmit a variety of multimedia files, including music, videos, images, documents, and mini-programs. XR (Extended Reality) technology can also be integrated to enable personalized applications such as video effects and dynamic XR backgrounds, as well as personalized voice services such as intelligent customer service and real-time translation. XR technology is an environment created by computer technology and wearable devices that blends the real and virtual worlds and enables human-computer interaction. XR services encompass a variety of technologies, including virtual reality (VR), augmented reality (AR), and mixed reality (MR). These services involve transforming the existing IMS network and adding control and media plane network elements for 5G new communication to establish data channels and basic service processes. However, since 5G new communication services can transmit various multimedia files, especially streaming media files such as high-definition videos, which will occupy a large amount of network traffic, network traffic control and scheduling have become urgent issues to be solved.

[0060] In the prior art, media plane units primarily rely on a fixed traffic allocation method based on service priorities to transmit multimedia files. During this process, the media plane unit first determines the priority of each service. For example, services with high real-time requirements, such as streaming media files, are classified as high-priority services, while services with relatively low real-time requirements, such as ordinary images and documents, are classified as low-priority services. Subsequently, a fixed proportion of traffic is allocated to each service based on pre-set priority rules. The higher the priority, the greater the allocated traffic. Each media plane unit uses its allocated fixed traffic channel to transmit the corresponding multimedia file. During the transmission process, regardless of changes in network conditions, the traffic allocation ratio remains relatively fixed, making it difficult to flexibly adjust according to real-time needs.

[0061] However, due to the limited traffic allocated to low-priority services, they may not receive sufficient bandwidth for extended periods, severely impacting transmission performance, such as slow image loading and interrupted document transfers. Furthermore, network traffic demand is dynamic, and existing technologies are unable to detect and adjust to these changes in a timely manner, making it difficult to adapt to complex network environments. This severely impacts transmission performance, which in turn affects the transmission of low-priority multimedia files.

[0062] In response to the problems existing in the above-mentioned related technologies, this embodiment provides a network traffic control system, method and storage medium.

[0063] Combine Figure 1 As shown, an embodiment of the present invention provides a network traffic control system, including:

[0064] The information collection unit is used to obtain historical traffic data of all media plane network elements connected to the control plane network element and the topological relationship of the media plane network elements.

[0065] Specifically, the information collection unit regularly collects historical traffic data from media plane network elements, including device CPU / GPU usage, short-term or instantaneous data inflow / outflow, and network bandwidth usage. Furthermore, it collects the connectivity and geographic distribution of media plane network elements, constructing a topological diagram of the device cluster, providing basic data support for subsequent traffic regulation.

[0066] For example, in a 5G new communication network, the information collection unit primarily collects data from the following media-plane network elements, such as media servers, video processing devices, media gateways, edge computing nodes, CDN nodes, and signaling processing devices. These devices typically exist as rack-mounted servers, dedicated processing equipment, or network nodes, and are deployed in data centers, communication rooms, or at the network edge. The information collection unit regularly collects data such as CPU / GPU (central processing unit) utilization, traffic changes, and bandwidth usage for each device. For example, during live broadcasts, media servers can generate up to 10 Gbps (Gigabits per second, a unit of data transmission rate often used to describe high-speed network communications), CDN (Content Delivery Network) nodes have an average bandwidth utilization of 5 Gbps, and edge computing nodes process local requests at a rate of 1.2 Gbps. This data provides comprehensive support for the traffic control system, laying the foundation for efficient network resource management and optimization.

[0067] An AI model updating unit is used to update the traffic control model of the control plane network element according to the historical traffic data and the topological relationship, wherein the traffic control model includes a deep reinforcement learning sub-model, a fusion sub-model, a graph neural network sub-model and a Transformer sub-model.

[0068] Specifically, the AI model update unit trains and optimizes the deep reinforcement learning sub-model based on collected historical traffic data and topological relationships to improve its decision-making capabilities in complex network environments. At the same time, the fusion sub-model is trained using historical data to enable more accurate time series predictions. The graph neural network sub-model is updated based on topological relationships to better understand and process the connection relationships between network devices. The Transformer sub-model is also adjusted based on new data to optimize its performance in weight distribution. This comprehensive model update method ensures that the traffic control model can adapt to changing network conditions.

[0069] A strategy generation unit is used to input the current traffic data of the current time period into the updated traffic control model, perform time series prediction through the fusion sub-model, and obtain the time series prediction result of the current traffic data; perform topological analysis through the graph neural network sub-model to obtain the graph structure of the media plane network element; perform weight distribution through the Transformer sub-model to obtain the distribution weight strategy of each media plane network element; and then perform reinforcement learning through the deep reinforcement learning sub-model based on the time series prediction result, the graph structure and the distribution weight strategy to obtain the traffic distribution strategy of the control plane network element.

[0070] Specifically, the policy generation unit first inputs current traffic data into the traffic control model. The fusion sub-model in the traffic control model combines historical data to perform time series predictions, outputting short-term and long-term trends for the current traffic. The graph neural network sub-model then analyzes the topological structure of the media plane network elements, determining the location and relationships of each device in the network. The Transformer sub-model then assigns weights to each media plane network element based on its performance indicators and network conditions, determining its priority in network traffic allocation. Finally, the deep reinforcement learning sub-model integrates the time series prediction results, graph structure, and weight allocation strategy, and uses a reinforcement learning algorithm to generate the optimal traffic distribution strategy to achieve network load balancing and efficient resource utilization.

[0071] A policy execution unit is configured to dynamically adjust the media traffic of each media plane network element according to the traffic distribution policy.

[0072] Specifically, after receiving the traffic distribution policy from the policy generation unit, the policy execution unit adjusts the traffic of each media plane network element in real time according to the policy. For example, when the load of a media plane network element is too high, the policy execution unit will transfer part of its traffic to other media plane network elements in the same province or adjacent areas to reduce the burden on the device. At the same time, it can also dynamically adjust traffic distribution according to the real-time status of the network to ensure stable operation of the network and a good user experience. In addition, the policy execution unit will also monitor the network status in real time so that it can make quick adjustments when necessary to deal with sudden traffic changes.

[0073] The network traffic control system of the present invention realizes the joint control of media plane network elements through an innovative multi-module collaborative mechanism, effectively solving the limitations of the fixed traffic distribution method in the existing technology. The information collection unit regularly collects historical traffic data (such as device CPU / GPU occupancy, short-term or instantaneous traffic, bandwidth occupancy, etc.) and the topological relationship of the device from the media plane network elements, providing a detailed data foundation for subsequent model training; enabling the system to fully understand the historical status of the network and the connection relationship between devices, thereby providing a basis for dynamic traffic distribution. The AI model update unit continuously updates the traffic control model through historical traffic data and topological relationships. The model includes a deep reinforcement learning sub-model, a fusion sub-model, a graph neural network sub-model and a Transformer sub-model. The synergy of these sub-models enables the system to adapt to different network environments and conditions, thereby improving the flexibility and efficiency of traffic distribution. The policy generation unit uses the fusion sub-model to perform time series predictions, which can predict traffic trends in advance and provide forward-looking guidance for traffic control. The graph neural network sub-model analyzes the topological relationship of devices, which helps to reasonably allocate network resources. The Transformer sub-model allocates weights based on device performance and network conditions to ensure that critical services are given priority and sufficient bandwidth resources. In addition, the generation of these policies is based on dynamic network conditions and real-time data, which also avoids the unfairness of traditional fixed allocation methods. The policy execution unit dynamically adjusts the traffic of each media plane network element based on the traffic distribution policy obtained by the policy generation unit. By monitoring the network status in real time and quickly adjusting the traffic distribution, the policy execution unit can ensure the stability and efficiency of the network. For example, when the load on a media plane network element is too high, the system can transfer part of its traffic to other devices to achieve load balancing.

[0074] The technical solution of the present invention achieves dynamic regulation of network traffic through real-time collection and analysis of network data, combined with dynamic updates of AI models. This allows for rapid adaptation to changing network conditions, improving the flexibility and efficiency of traffic regulation. Using the Transformer sub-model for weight allocation, the system can rationally allocate bandwidth resources based on service priority and device performance, avoiding prolonged bandwidth shortages for low-priority services and enhancing the fairness of traffic regulation. Furthermore, by analyzing the graph structure using the graph neural network sub-model, a deep understanding of the complex topological relationships between media plane network elements is achieved, enabling coordinated traffic allocation across multiple network elements. This joint regulation approach overcomes the limitations of traditional single-source regulation. By comprehensively considering multiple factors such as timing, topology, and weights, it enables more flexible and precise traffic allocation, effectively avoiding prolonged bandwidth shortages for low-priority services. Furthermore, the system can dynamically adjust traffic allocation based on real-time changes in network conditions and service demands, adapting to dynamic changes in network traffic demand and ensuring the transmission quality of various multimedia files. Furthermore, the deep reinforcement learning sub-model performs reinforcement learning based on timing prediction results, graph structure, and weight allocation strategies to optimize traffic allocation strategies, achieving intelligent and automated traffic regulation. By effectively managing network resources, the service quality degradation caused by traffic congestion is reduced, and the user experience when using high-definition video, XR and other services is improved.

[0075] In summary, the present invention realizes intelligent regulation of 5G new communication service traffic through the synergistic effect of information collection, model update, policy generation and policy execution, significantly improves the dynamics, flexibility, fairness and resource utilization of traffic regulation, thereby solving the problems in the existing technology and achieving the goal of improving the traffic regulation effect of 5G new communication services.

[0076] Optionally, the information collection unit is specifically configured to:

[0077] According to a preset collection cycle, the operating status data of all the media plane network elements within a preset time period is obtained from historical data, wherein the operating status data includes CPU occupancy and GPU occupancy;

[0078] Determining network bandwidth occupancy data of the media plane network element based on data flow inflow and outflow information of the network interface of the media plane network element in the historical data;

[0079] Obtaining load data of the media plane network elements according to the signaling throughput data of all the media plane network elements in the historical data;

[0080] The operating status data, the network bandwidth occupancy data, and the load data are used as the historical traffic data.

[0081] Specifically, the information collection unit obtains the operating status data of all media plane network elements within a preset time period from the historical data according to the preset collection cycle, wherein these operating status data cover the CPU occupancy and GPU occupancy of the media plane network elements. By analyzing these data, it is possible to understand the computing resource usage of the device, providing an important basis for the subsequent evaluation of the processing capacity and load status of the device. Based on the data flow inflow and outflow information of the network interface of the media plane network element in the historical data, the information collection unit determines the network bandwidth occupancy data of each media plane network element; the data flow inflow and outflow information of the network interface reflects the scale and speed of data transmission between the media plane network element and other parts of the network. By analyzing this information, the bandwidth resources occupied by each media plane network element in the network communication can be accurately calculated, thereby providing key network-level data support for traffic regulation.

[0082] In addition, the information collection unit also obtains load data for each media plane NE based on the signaling throughput data of all media plane NEs in the historical data. Signaling throughput data reflects the workload and efficiency of the media plane NEs when processing signaling messages. By analyzing and processing this data, the device's signaling processing load can be determined, further enriching the assessment of the overall operational status of the media plane NEs.

[0083] Ultimately, the information collection unit integrates the collected operating status data, determined network bandwidth usage data, and acquired load data as the historical traffic data. This historical traffic data provides a comprehensive, detailed, and multi-dimensional data foundation for subsequent AI model updates and traffic control strategy generation, enabling the traffic control system to learn and optimize based on this rich historical information, thereby more accurately predicting and responding to changes in network traffic and achieving precise traffic control.

[0084] It should be noted that in one embodiment of the present invention, the operating status data collected by the information collection unit covers the CPU and GPU usage of the media plane network element. The CPU and GPU here specifically refer to the central processing unit (CPU) and graphics processing unit (GPU) in the media plane network element device. Specifically, the CPU is the core processor in the media plane network element device. Within the media plane network element, the CPU is primarily responsible for tasks such as processing control signaling, media stream protocol processing, resource scheduling, and service logic execution. For example, it processes signaling messages from the control plane network element, makes media stream forwarding and routing decisions, and manages device resource allocation. CPU usage reflects the device's computing resource usage, including the processing workload and resource utilization. By monitoring CPU usage, it is possible to assess whether the device's processing capacity is sufficient and whether there is overload or resource waste. The GPU is the graphics processor in the media plane network element device, specifically designed to accelerate graphics and image processing tasks. In media plane network elements, the GPU primarily handles graphics and image processing tasks related to media streams, such as video encoding and decoding, image rendering, video special effects processing, and XR (extended reality) effect generation. Examples include real-time encoding and decoding of HD video, generating virtual backgrounds, and implementing special effects in videos. GPU usage reflects the device's resource usage for graphics and image processing. By monitoring GPU usage, you can assess the device's load when processing graphics tasks for media streams, ensure proper GPU resource utilization, and avoid delays or lags in media stream processing caused by GPU overload.

[0085] Furthermore, the information collection unit collects CPU and GPU usage data specifically from the central processing units (CPUs) and graphics processing units (GPUs) in media plane network element devices. This data comprehensively reflects the computing resource usage of media plane network elements when processing media streams and executing business logic. By analyzing this data, we can accurately assess the processing capacity and load of the devices, providing a detailed, multi-dimensional data foundation for subsequent traffic control model training and policy formulation. Based on this data, the system can better allocate resources and balance load, improve network resource utilization efficiency, and ensure stable network operation and good performance.

[0086] In this optional embodiment, by obtaining the operating status data, network bandwidth occupancy data and load data for a preset time period according to a preset collection cycle, the historical traffic data of the media plane network element can be collected comprehensively and systematically. These data cover the operating status of the media plane network element in many aspects, such as the CPU and GPU occupancy, the data flow inflow and outflow information of the network interface, and the signaling throughput data, providing a detailed and multi-dimensional data foundation for subsequent traffic control model training and policy formulation. Based on the operating status data, the computing resource usage of the media plane network element can be accurately understood, thereby evaluating its processing capacity. By determining the network bandwidth occupancy data, the system can accurately grasp the bandwidth resources occupied by each media plane network element in network communication, providing key network-level data support for traffic control. By obtaining load data, the evaluation of the overall operating status of the media plane network element is further enriched, which helps to fully understand the load situation of the equipment.

[0087] Furthermore, using this historical traffic data as input to the traffic control model provides rich information for model training and optimization. By analyzing and mining historical traffic data, the traffic control model can more accurately predict and respond to changes in network traffic, achieving precise traffic control. Based on comprehensive and accurate historical traffic data, the system can better allocate resources and balance loads, improving network resource utilization efficiency and ensuring stable network operation and good performance. By effectively managing network resources, the service quality degradation caused by traffic congestion is reduced, and the user experience of services such as HD video and XR is improved.

[0088] To sum up, the setting and function implementation of the information collection unit can comprehensively and accurately collect the historical traffic data of media plane network elements, provide key data support for the traffic control system, help improve the effect of traffic control, and achieve efficient utilization and optimized management of network resources.

[0089] Optionally, the AI model updating unit is specifically configured to:

[0090] Inputting the historical traffic data and the topological relationship into the traffic control model for training;

[0091] Verify the traffic distribution strategy output by the traffic control model during training through a digital twin environment and A / B testing;

[0092] When the traffic distribution strategies output by the traffic control model during training meet the test requirements of the digital twin environment and the A / B test, the training of the traffic control model is stopped to obtain the updated traffic control model.

[0093] Specifically, historical traffic data collected by the information collection unit, including operational status data such as device CPU / GPU usage, network bandwidth usage, and load data, as well as topological information about media plane network elements, is fed into the traffic control model for training. During training, the model continuously adjusts its parameters based on this data to learn patterns in network traffic and the interactions between media plane network elements. The traffic distribution strategies generated during training are validated using a digital twin environment and A / B testing. The digital twin environment simulates a virtual environment that closely resembles the actual network environment. Testing traffic distribution strategies within this environment allows for early observation of the strategy's performance under various possible network conditions, such as high load and bursty traffic, thereby assessing the strategy's feasibility and effectiveness. Meanwhile, A / B testing tests a subset of traffic or devices in the actual network environment and compares the test results with those without the strategy. This intuitively demonstrates the traffic distribution strategy's impact on network performance and user experience. Combining these two verification methods allows for a more comprehensive and accurate assessment of the quality of traffic distribution strategies. If the traffic allocation policy output by the traffic control model during training meets the test requirements in both the digital twin environment and A / B testing, such as effectively achieving network load balancing, improving resource utilization, and ensuring the transmission quality of key services, then training of the traffic control model will be stopped, resulting in an updated traffic control model. This updated model incorporates the knowledge and rules learned from previous training and provides a more accurate and reliable basis for generating subsequent traffic allocation policies.

[0094] For example, the AI (Artificial Intelligence) model update unit uses the collected resource usage of each media plane network element and a trained large model to determine whether the media resources transmitted by some media plane network elements with excessive resource utilization need to be dynamically adjusted and dispatched nearby. If necessary, other media plane network elements in the same province or adjacent regions are selected for media resource transmission to reduce the load on individual media plane network elements. If the load on media plane network elements in the same area is high during certain periods of time, the AI model update unit dispatches media plane network elements in other adjacent areas to share the load.

[0095] Specifically, the control plane NE uses the AI model update unit to collect data from the media plane NEs, including device CPU / GPU usage, short-term (for example, one hour; the specific time period can be adjusted by network operations personnel) or instantaneous data inflow / outflow, bandwidth usage, and short-term / instantaneous signaling throughput. Operations personnel set processing thresholds based on network device conditions. Data exceeding the thresholds is treated as alarm data, while data within the thresholds is treated as normal device operation data. Both types of data are fed into the AI model update unit for training, resulting in an AI model that can automatically warn when signaling and traffic volume of media plane NEs exceed thresholds. The control plane NE periodically sends requests for device status data to the media plane NEs. When the data received from a media plane NE exceeds the threshold, the AI model update unit sends real-time device data requests to the media plane NEs connected to the control plane NE. Based on the collected data from other surrounding media plane NEs, the AI model update unit selects one or more media plane devices to share the signaling and media data traffic of the alarming device based on factors such as device CPU / GPU usage, latency, traffic volume, bandwidth usage, and signaling throughput.

[0096] In addition, for scenarios with instantaneous traffic surges, such as the appearance of popular stars on live broadcasts, the appearance / winning of star players in sports events, etc., we can first collect the operation and maintenance data of the control and media plane equipment that have not yet introduced the AI model update unit to conduct preliminary AI model training. During the actual operation of the network equipment, we can set up an alarm mechanism to receive an alarm notification when the signaling / bandwidth occupancy exceeds 2 times the daily data. The AI model is dynamically regulated, and the operation and maintenance personnel participate in the model regulation. At the same time, a time period for predicted traffic surges is reserved for the AI model update unit. At the reserved time end, the AI model for traffic surges is set up in advance, or the operation and maintenance personnel are predicted in advance to pay attention to traffic changes and participate in AI model regulation.

[0097] In this optional embodiment, by incorporating rich historical traffic data and topological relationships into the model for training, the model can gain a more comprehensive understanding of network operations and traffic patterns, thereby improving the accuracy of the traffic allocation strategies it generates. Furthermore, rigorous validation of strategies using a digital twin environment and A / B testing ensures that only strategies that meet actual needs and achieve the desired results are adopted, further enhancing the reliability of the model. The digital twin environment simulates a variety of possible network conditions, allowing the model to be exposed to a wide range of scenarios during training, thereby enhancing its adaptability to diverse network environments. When the actual network environment changes, the trained and validated model can more quickly and effectively adapt to changes, generating traffic allocation strategies tailored to the current situation and improving the model's practical application. The updated traffic control model can more accurately formulate reasonable traffic allocation plans based on the actual network operation status and service needs. This helps optimize the allocation of network resources, improve resource utilization efficiency, and avoid situations where some devices are overloaded while others are idle, thereby improving the performance and service quality of the entire network.

[0098] Furthermore, accurate and reliable traffic distribution strategies help ensure the transmission quality of critical services and reduce service interruptions and delays caused by traffic congestion. For users, this means a smoother and more stable experience when using services such as HD video and XR. Furthermore, rational resource allocation and effective traffic regulation help maintain stable network operation, reduce the risk of system failures, and improve the reliability and stability of the entire network.

[0099] Optionally, the policy generating unit is specifically configured to:

[0100] Obtaining the operating status data, the network bandwidth occupancy data, and the load data of the media plane network element within the current time period, and inputting the operating status data, the network bandwidth occupancy data, and the load data as the current traffic data into the updated traffic control model;

[0101] Perform feature extraction on the current traffic data through the CNN network of the fusion sub-model to obtain spatial features of the current traffic data;

[0102] Perform feature extraction on the current traffic data according to the BiLSTM network of the fusion sub-model to obtain a time series of the current traffic data;

[0103] According to the spatial features and the time series, a traffic change trend of the media plane network element in a preset future time period is obtained, and the traffic change trend is used as the time series prediction result.

[0104] Specifically, the policy generation unit first obtains the operating status data, network bandwidth usage data, and load data of the media plane network elements for the current time period. This data is consistent with the historical traffic data collected by the information collection unit, including device CPU / GPU usage, network bandwidth usage determined by data inflow and outflow information on network interfaces, and load data derived from signaling throughput data. The acquired operating status data, network bandwidth usage data, and load data are then integrated and input into the updated traffic control model as current traffic data. The fusion sub-model in the traffic control model includes a convolutional neural network (CNN) and a Bidirectional Long Short-Term Memory (BiLSTM) network. The CNN extracts features from the current traffic data, primarily obtaining spatial features. The convolution operation of the CNN automatically learns local spatial patterns and correlations in the data, such as the spatial distribution of traffic between different devices and resource usage patterns. The BiLSTM network extracts features from the current traffic data to obtain time series features. BiLSTM can perform bidirectional modeling on time series data, capturing the temporal dependencies and trends of traffic data, such as rising or falling trends and periodic changes in traffic at different time points. After obtaining the spatial and temporal characteristics of the current traffic data, the fusion sub-model integrates these two features to further analyze and predict traffic trends for media plane network elements within a preset future time period. The predicted results are used as time series forecasts, providing an important basis for the subsequent formulation of traffic allocation strategies. By combining characteristics from both spatial and temporal dimensions, the forecast results can more comprehensively and accurately reflect the possible changes in future traffic, including which devices are likely to experience increases and decreases in traffic, as well as the approximate magnitude and time range of traffic changes.

[0105] In this optional embodiment, the CNN and BiLSTM networks in the fusion sub-model extract the spatial and time series features of the current traffic data, respectively, enabling a more comprehensive capture of the various characteristics and patterns within the traffic data. Spatial features reflect the resource utilization patterns and traffic distribution between network elements on different media planes, while time series features reflect the temporal trends of traffic. Combining these two features to predict traffic trends significantly improves prediction accuracy, enabling the system to understand future traffic changes in advance and providing a more accurate basis for traffic regulation.

[0106] Furthermore, based on the accurate prediction of future traffic trends of media plane network elements, this embodiment enables the system to make corresponding traffic management decisions in advance. For example, if it is predicted that the traffic of a certain device will increase significantly over a period of time in the future and may exceed its processing capacity, then part of the traffic can be adjusted and allocated to other devices in advance to achieve dynamic traffic management. This dynamic adjustment capability helps to avoid network congestion and resource overload, ensuring the stable operation of the network. Accurately predicting traffic trends helps the system to allocate network resources more reasonably. By understanding the changes in traffic demand of each media plane network element in advance, the resource allocation strategy can be dynamically adjusted according to actual conditions, making resource allocation more refined and reasonable. This not only improves resource utilization efficiency and avoids resource waste, but also ensures that critical and high-priority services can obtain sufficient resource support when needed, thereby improving the quality and reliability of network services.

[0107] Optionally, the policy generating unit is specifically configured to:

[0108] Inputting the topological relationship of the media plane network elements in the current time period into the updated traffic control model;

[0109] The graph structure is constructed according to the topological relationship through the graph neural network sub-model, wherein the nodes of the graph structure are the media plane network elements, and the edges of the graph structure are the connection relationships between the media plane network elements.

[0110] Specifically, the policy generation unit takes the topological relationship of the media plane network elements in the current time period as input. The topological relationship describes in detail the connection mode and network structure between the various media plane network elements, such as which devices have direct communication links, the hierarchical position of the devices in the network, etc. Through the graph neural network sub-model, the graph structure of the graph neural network is constructed according to the input topological relationship. In the graph structure, the media plane network elements are abstracted as nodes of the graph, and the connection relationship between the media plane network elements is represented as the edge of the graph. Specifically, each node represents a specific media plane network element, and the attributes of the node may include information such as the device type, performance parameters, and current traffic load of the network element; the edge indicates that there is a communication connection or data transmission path between two media plane network elements. The weight of the edge can be associated with network parameters such as the bandwidth and latency of the connection, so as to reflect the closeness of the connection between different devices and the communication capability.

[0111] In this optional embodiment, the graph structure constructed using the graph neural network sub-model accurately models the network topology of media plane network elements. This allows the traffic control system to intuitively understand the structural characteristics of the network and the interconnection between devices, incorporating network topology information into the traffic control decision-making process. Furthermore, when formulating traffic allocation strategies, the network location of each media plane network element and its connection relationships with other devices can be fully considered, thereby making more reasonable resource allocation decisions. For example, for media plane network elements located on critical network paths or connected to a large number of other devices, their resource needs may be prioritized to ensure smooth communication across the entire network. Based on the graph structure, the system can implement differentiated traffic control strategies for different areas or sub-networks within the network. For example, for relatively independent sub-networks within the topology, local traffic optimization can be performed based on the traffic conditions and topological relationships of the devices within them. However, for core devices connected to multiple sub-networks, traffic control requires a global perspective to ensure stable operation of the entire network.

[0112] By leveraging the graph structure of graph neural networks, traffic control is tailored to the network's actual topology, improving overall network performance and reliability. Effective traffic control helps avoid network congestion, reduces data transmission delays and packet loss, and improves network resource utilization, ensuring stable and efficient network operation and enhancing the user experience.

[0113] Optionally, the policy generating unit is specifically configured to:

[0114] Obtaining the operating status data, the network bandwidth occupancy data, and the load data of the media plane network element within the current time period, and inputting the operating status data, the network bandwidth occupancy data, and the load data as the current traffic data into the updated traffic control model;

[0115] Using the Transformer sub-model, prediction is performed based on the current traffic data to obtain critical values of multiple preset indicators of each media plane network element within a preset future time period;

[0116] A weight value of each of the preset indicators is generated according to the critical value to obtain the allocation weight strategy of each of the media plane network elements.

[0117] Specifically, the policy generation unit obtains the operating status data, network bandwidth usage data, and load data of the media plane network elements within the current time period. This data can reflect the operating status of the media plane network elements in real time, including the device's computing resource usage (such as CPU and GPU utilization), network resource usage (such as the inflow and outflow of data traffic), and signaling processing load. This data is then integrated into current traffic data and input into the traffic control model. The Transformer sub-model in the traffic control model is responsible for in-depth analysis and prediction of current traffic data. Through its powerful sequence processing capabilities and parallel computing advantages, the Transformer sub-model can capture complex patterns and long-term dependencies in the data.

[0118] Specifically, the Transformer sub-model predicts the critical values of multiple preset indicators of each media plane network element within a preset future time period based on current traffic data. These preset indicators include key performance indicators such as device resource utilization, network bandwidth occupancy, and signaling processing delay. In addition, the prediction of critical values helps to identify possible resource bottlenecks or traffic congestion points in advance.

[0119] The critical value refers to a key threshold within a preset time period in the future, that is, the maximum value of a preset indicator within the preset time period in the future, rather than simply the maximum value or the value at the last moment. Specifically, the critical value is predicted based on current traffic data and is used to indicate a key threshold reached by a preset indicator of a media plane network element (such as device resource utilization, network bandwidth occupancy, signaling processing delay, etc.) within a preset time period in the future.

[0120] The purpose of setting a critical value in this optional embodiment is to identify possible resource bottlenecks or traffic congestion points in advance so that the system can take measures to regulate traffic in advance. The critical value is usually set based on statistical analysis of historical data and network operation experience.

[0121] For example, if historical data indicates that a device's CPU utilization may experience performance issues when it exceeds 80%, 80% is set as the CPU utilization threshold. The Transformer sub-model analyzes and predicts current traffic data to estimate whether a preset metric is likely to reach or exceed this threshold within a certain future time period. If a preset metric is predicted to reach or exceed the threshold at any time within the preset time period, the system deems the metric risky and adjusts the weight allocation strategy accordingly to prioritize resource allocation to that media plane network element.

[0122] As another example, if the preset indicator is the network bandwidth utilization rate, the critical value is set to 90%, and the Transformer sub-model predicts that at a certain moment in the future, the bandwidth utilization rate of a certain media plane network element will reach 90%, then 90% is the critical value. If the preset indicator is the CPU utilization rate of the device, the critical value may be set to 80%. If it is predicted that at a certain moment in the future, the CPU utilization rate of a certain device will reach 80%, this 80% is the critical value. In summary, in this optional embodiment, the critical value is a key threshold value used to determine whether a preset indicator may reach or exceed the preset maximum critical value within a preset time period in the future, so that the system can perform traffic regulation in advance.

[0123] After determining the critical values for each preset indicator, the policy generation unit generates weights for each preset indicator based on these critical values. Weights are generated based on the principle that key indicators or those approaching critical values are assigned higher weights to reflect their importance and urgency in network traffic control. By comprehensively considering the weights of each indicator, a weight allocation policy is ultimately formed for each media plane network element. This policy clarifies how much traffic resources each media plane network element should be prioritized for during traffic allocation, providing a basis for subsequent dynamic traffic adjustments.

[0124] In this optional embodiment, the Transformer sub-model uses in-depth analysis and prediction of current traffic data to accurately determine the critical values of key performance indicators for each media plane network element. The weights generated based on these critical values and the weight allocation strategy can achieve more precise resource allocation. For example, devices whose resource utilization is about to reach a critical value can be prioritized for more traffic resources to prevent resource exhaustion and ensure stable network operation.

[0125] Furthermore, the policy generation process dynamically adjusts weight distribution based on real-time traffic data, making traffic control adaptive. When network conditions change, such as a sudden increase in load on a media plane network element, the system can quickly identify and redistribute traffic by adjusting the weighting policy, promptly alleviating network pressure and avoiding congestion. This adaptive traffic control mechanism effectively responds to various dynamic changes in the network, improving network flexibility and reliability. The prediction and weight generation mechanism based on the Transformer sub-model can proactively identify potential performance bottlenecks and optimize them through appropriate resource allocation. This helps improve overall network performance and mitigate performance degradation caused by inappropriate resource allocation. Optimizing traffic distribution can reduce data transmission latency, increase network throughput, and enhance the user experience, especially in application scenarios with high real-time and bandwidth requirements, such as HD video conferencing and XR interaction.

[0126] Finally, precise resource allocation and adaptive traffic control help reduce network overload and resource contention, thereby enhancing system reliability and stability. In the face of traffic bursts or network anomalies, the system can promptly adjust traffic distribution strategies to ensure the continuity of critical services, reduce the risk of system failures, ensure continuous network stability, and enhance the user experience.

[0127] Optionally, the policy generating unit is specifically configured to:

[0128] Performing reinforcement learning based on the time series prediction results, the graph structure, and the weight allocation strategy through the deep reinforcement learning sub-model;

[0129] wherein, generating an initial traffic distribution adjustment strategy based on the time series prediction result, the graph structure, and the distribution weight strategy;

[0130] Taking the initial traffic distribution adjustment strategy as an action and the digital twin environment as an environment, calculating a reward value after the action interacts with the environment through a reward function;

[0131] Whether the reinforcement learning of the initial traffic distribution adjustment strategy is completed is determined according to the reward value, and when the reinforcement learning of the initial traffic distribution adjustment strategy is completed, the traffic distribution strategy of the control plane network element is obtained.

[0132] Specifically, the deep reinforcement learning sub-model uses the time series prediction results, graph structure, and allocation weight strategy as inputs to generate an initial traffic allocation adjustment strategy, allocating traffic resources to each media plane network element. The initial traffic allocation adjustment strategy serves as the action, and the digital twin environment serves as the reinforcement learning environment. The digital twin environment simulates real-world network conditions and traffic conditions, providing a platform for testing and optimizing the initial strategy and predicting the network status and performance indicators after traffic adjustment. A reward function is used to calculate the reward value generated by the interaction between the action and the environment. The reward function comprehensively considers the effectiveness of the traffic allocation strategy, such as network load balancing, resource utilization, and transmission quality of key services. A high reward value indicates a good strategy, while a low reward value indicates that optimization is required. The reward value is used to determine whether reinforcement learning is complete. If the reward value reaches a preset threshold or stabilizes, the initial traffic allocation adjustment strategy is optimized, resulting in the final traffic allocation strategy.

[0133] In this optional embodiment, the deep reinforcement learning sub-model combines time series prediction, graph structure and weight strategy to generate and optimize the initial strategy, improve the rationality of traffic distribution, and enhance network performance. Simulating and evaluating strategies in a digital twin environment can also reduce trial and error costs and risks, improve decision-making quality and reliability, and ensure network stability. Through reinforcement learning, the strategy is continuously adjusted to adapt the system to the dynamic network environment, optimize traffic distribution, and improve resource utilization and business performance. The deep reinforcement learning sub-model enhances the system's intelligent decision-making capabilities, realizes automated and intelligent traffic control, reduces the burden on operation and maintenance personnel, and improves network management efficiency. The optimization and adjustment of reinforcement learning ensures the effectiveness and adaptability of traffic distribution strategies, reduces congestion and overload, ensures business continuity, and enhances network stability. In addition, the optimized traffic distribution strategy ensures the transmission quality of critical and high-priority services, improves the user's network service experience, especially in scenarios such as high-definition video and XR, reduces freezes and delays, and improves satisfaction.

[0134] Optionally, the system further includes an early warning unit, which is configured to:

[0135] Obtaining a flow rate change trend within the current time period based on the current flow rate data of the current time period;

[0136] According to the traffic change trend, determining whether there is a sudden increase in traffic on the media plane network element;

[0137] When the traffic surge occurs in the media plane network element, an early warning message is sent.

[0138] Specifically, the early warning unit obtains the current traffic data of the media plane network elements in the current time period in real time, including operation status data, network bandwidth usage data, load data, etc. Through real-time monitoring and analysis of these data, the traffic change trend is calculated.

[0139] For example, time series analysis methods, such as moving averages and exponential smoothing, can be used to identify short-term fluctuations and long-term trends in traffic data. Furthermore, machine learning algorithms, such as linear regression and decision trees, can be combined to model and predict traffic data, more accurately understanding patterns in traffic fluctuations.

[0140] Based on the traffic trend, the early warning unit determines whether a media plane network element is experiencing a traffic surge. Traffic surge determination can include metrics such as absolute traffic growth, relative growth rate, and growth rate. For example, the average traffic growth rate can be calculated within a preset time window. If this rate exceeds a set threshold, a traffic surge is determined. The threshold can be based on statistical analysis of historical data or dynamically adjusted based on the network's actual operating conditions and service needs. Furthermore, traffic peak detection algorithms, such as wavelet transforms and sliding window extreme value detection, can be incorporated to quickly identify sudden traffic surges. When a media plane network element is identified as experiencing a traffic surge, the early warning unit promptly sends a warning message. The warning message may include key information such as the device identifier of the traffic surge, the scale of the surge, the time of the surge, and the potential impact range. The method for sending the warning message can be selected based on actual needs, such as SMS, email, instant messaging tools, or pop-up windows within the network management system interface, to promptly convey the warning message to network operations and maintenance personnel or relevant automated management systems. This allows valuable time for subsequent traffic regulation and emergency response, allowing proactive measures to be taken to avoid network congestion and performance degradation.

[0141] In this optional embodiment, the early warning unit can promptly detect trends of traffic surges through real-time monitoring and analysis of current traffic data and issue early warnings before problems occur. This allows network operations personnel or the automated management system sufficient time to take appropriate measures, such as adjusting traffic distribution strategies and increasing resource allocation, thereby preventing network congestion and performance issues in advance and improving network stability and reliability. Traffic surges can cause network congestion and impact the normal operation of services. Early warnings enable timely measures such as traffic diversion and rate limiting to ensure smooth network operation, improve service continuity and reliability, and minimize service interruptions or performance degradation caused by network problems. Furthermore, early warnings facilitate the rational arrangement and allocation of resources. Before traffic peaks occur, additional bandwidth resources can be allocated or traffic distribution strategies adjusted in advance to ensure the stable operation of critical services, improve resource utilization efficiency, and avoid resource waste. Early warnings can also trigger responses from the automated management system, enabling automated and intelligent network management, improving management efficiency, reducing the need for manual intervention, and alleviating the workload of operations personnel.

[0142] In a preferred embodiment of the present invention, Figure 2The figure shows the main network elements in the 5G new communication network architecture and their interrelationships. The control plane and media plane network elements are the core components of the architecture. Other key network elements, such as the service application server (AS) and the media processing function (MPF), are also shown. The AI model update unit in the control plane embodies the network's AI capabilities. It obtains real-time parameter information from the media plane through interface 5 for traffic control and optimization.

[0143] Specifically, combined Figure 2 This section introduces the main network elements and their functions. The 5G New Communication Platform (NCP) complies with the 3GPP TS26.114 standard and supports the DC Service Control Function (DCSF). It manages the establishment, maintenance, and release of the DC (Data Channel) and triggers DC services. It also provides a northbound interface for service application systems (ASs) to invoke, enabling flexible service customization and expansion. The Unified Media Function (UMF) complies with the 3GPP TS26.114 standard and supports the DC Media Function (DCMF). It provides various DC service media capabilities, including but not limited to media replication, video stream synthesis, subtitle stream synthesis, background replacement, and virtual portraits. It supports establishing DC channels with terminals and transmits application interaction information, enabling a rich media service experience. The Service Application Server (AS) provides various DC services, such as intelligent translation, XR calling, and intelligent customer service. As the carrier of service logic, it collaborates with control and media plane network elements to provide users with convenient value-added services. MPF (Media Processing Function) is a third-party platform that provides enhanced audio and video media processing capabilities; specific functions include speech-to-text, translation, gesture recognition, etc., enriching the content and form of media services and improving user experience. Figure 3Multiple interfaces (interface 1 to interface 16) are marked in it. These interfaces define the communication paths and interaction methods between different network elements. For example: Interface 5 is the interface between the control plane and the media plane device. The AI model update unit obtains the real-time parameter information of the media plane through this interface. Interfaces such as interface 9 and interface 15 reflect the interactive relationship between the service AS and the control plane and media plane, and are used for service triggering and data transmission. The control plane contains control plane network elements, which are responsible for signaling processing and control function implementation; the AI model update unit is located in the control plane, and realizes intelligent regulation of network traffic by analyzing the real-time data of the media plane network elements. The media plane contains media plane network elements, which are responsible for the processing and transmission of media streams; media plane network elements provide media services, such as forwarding, processing and storage of media streams.

[0144] Among them, such as Figure 2 Other network elements shown include ENUM / DNS, which is used for domain name resolution and number query to help locate users and service servers; HSS / UDM, which is used to store user subscription information and authentication data and support user identity authentication and service authorization; MMTel AS, which is used to provide control and management functions for traditional voice and multimedia telecommunications services; I / S-CSCF, which is responsible for initial session establishment and routing selection, as well as user registration and authentication; P-CSCF / PCF, which serves as the entry point for the user plane and is responsible for session establishment and policy control; EPS / 5GS, which is used to provide connectivity and interoperability with the 4G / 5G core network and support user mobility and session management; UE (user equipment), which is a terminal device used by users, such as smartphones and tablets, used to access the new 5G communication network.

[0145] In this preferred embodiment, by displaying the main network elements in the 5G new communication network architecture and their mutual relationships, the application position and role of the network traffic control method that integrates the control plane network elements and AI technology in the present invention are reflected in the network; the AI model update unit obtains real-time parameter information through the interface with the media plane device to realize intelligent regulation and optimization of network traffic.

[0146] In another preferred embodiment of the present invention, Figure 3The figure shows the connection between a control plane device and multiple media plane devices, embodying the network traffic control architecture of the present invention. Each control plane device contains an AI model update unit, located within the control plane device. This unit is responsible for collecting real-time traffic data from media plane devices, analyzing and making decisions, and generating traffic allocation strategies. Furthermore, the two control plane devices serve as a primary and backup device for each other. One control plane device is the primary device, responsible for primary control and management functions; the other is the backup device, taking over the primary device's functions in the event of a failure, ensuring high system availability and reliability. Each control plane device is connected to multiple media plane devices; these media plane devices are responsible for processing and transmitting media streams. Each control plane device is connected to multiple media plane devices, enabling centralized management and control of these media plane devices. A connection also exists between the two control plane devices for communication and data synchronization between the primary and backup devices. Each media plane device has a direct connection to the control plane device for transmitting media streams and control signaling. The media plane devices provide real-time traffic data, including device operating status, network bandwidth usage, and load, to the AI model update unit of the control plane device.

[0147] In this preferred embodiment, the AI model update unit in the control plane device periodically collects real-time traffic data from connected media plane devices for analysis and prediction. Based on the collected data, the AI model update unit utilizes technologies such as deep reinforcement learning, fusion sub-models, graph neural networks, and Transformers to generate an optimal traffic distribution strategy. Based on real-time traffic data, the AI model update unit dynamically adjusts traffic distribution between media plane devices to achieve network load balancing. When the traffic load of a media plane device is too high, the AI model update unit will transfer part of the traffic to other devices with lower loads to ensure stable operation of the network.

[0148] For example, the AI model update units of two control plane devices set timers of the same duration, with interleaved times (one timer starts half a cycle earlier than the other), take turns making decisions, and share decision data. Collected media resource parameters include, but are not limited to, device CPU / GPU usage, short-term (for example, one hour; the specific time can be adjusted by network operations personnel) or instantaneous data inflow / outflow, network bandwidth usage, and short-term / instantaneous signaling throughput. Data collection and model training are conducted during the network equipment's commercial trial phase. A digital twin environment is established to simulate a user scale of millions. Methods such as weighted round-robin and simulated annealing are used for comparison with traditional algorithm baselines. A / B testing covers at least 10 stress scenarios. During the formal phase, a time series prediction module is first deployed as an early warning system, followed by a phased implementation of a reinforcement learning controller. Ultimately, the trained model is used for network device resource scheduling based on AI recommendations and operations personnel review. During operation, a model drift detection mechanism is established, with real-time visualization of feature weights and heatmaps. Policy rollback triggers are set for real-time monitoring to facilitate network operations adjustments. When the instantaneous burst traffic is too large (2 times or more), media plane device scheduling prioritizes devices in the same province or region whose resources are not fully occupied. If all resources are occupied, eligible devices are selected from adjacent provinces / regions for load sharing.

[0149] This optional embodiment utilizes a composite algorithm architecture. The main framework utilizes an actor-critic architecture based on deep reinforcement learning (DRL), such as DDPG, PPO, and SAC. A CNN-BiLSTM fusion structure is used as the time series prediction component to handle real-time traffic fluctuations in the network. A graph neural network is used as the resource assessment module to streamline the topological relationships between control and media plane devices. A lightweight Transformer is used as the real-time decision-making module to handle instantaneous traffic bursts. A device cluster topology diagram is established spatially, and sliding windows (30s / 1min / 5min / 15min / 30min / 1h) are established temporally. The number of layers is 10, for example, but can be adjusted based on network conditions. Parameters are broadly categorized as basic indicators, traffic characteristics, and environmental parameters, and their types and weights can be adjusted based on the network conditions of individual operators. The dynamic weight allocation system is based on a two-layer LSTM (Long Short-Term Memory) meta-controller architecture. The first LSTM layer predicts critical thresholds for each metric, and the second GRU (Gated Recurrent Unit) layer generates a feature weight allocation strategy. Hourly strategy validation experiments are conducted, and weekly global parameter optimization is performed to adjust parameter weights. For example, basic metrics include CPU utilization (weight 0.15), GPU memory usage (0.18), and inter-node latency (0.22); traffic features include TCP retransmission rate (0.12), UDP jitter (0.08), and QoE mapping parameters (0.15); and environmental parameters include temperature sensor data (0.05), power supply fluctuation (0.03), and link optical attenuation (0.02). Differentiable Neural Architecture Search (DNAS) is used to automatically select features, guide feature importance through a reward function, and control feature redundancy through a gradient penalty term, thereby dynamically adjusting feature weights. The objective function uses a nested approach with multiple loss functions. The core objective layer uses the min (1-99.99% SLA compliance rate × log (resource utilization)) function. The constraint layer is set A bandwidth overlimit penalty is used, a threshold-triggered exponential penalty is used as a signaling storm suppression term, and the PUE² × TCO coefficient is used as an energy consumption-sensitive term. The regularization layer uses adjacent node resource difference regularization as a topology-aware constraint, and a scheduling policy mutation penalty is used as a temporal continuity constraint. Multi-objective Bayesian optimization is used to construct a Pareto surface and a dynamic weight allocation strategy. This strategy uses context-aware weights based on the current network state and guides target priorities through short-term load forecasting. Reward plasticity is generated through reverse engineering based on actual SLAs. Adversarial examples are generated by simulating network storm scenarios. Learning containers are transferred through knowledge transfer across business scenarios. These operations enable dynamic scheduling of media plane network device resources through the control plane's AI model update unit.

[0150] Figure 3 This demonstration demonstrates the connection between the control plane devices and media plane devices in this invention, highlighting the core role of the AI model update unit in the control plane devices in network traffic control. The setup of active and standby control plane devices ensures high system availability. The AI model update unit's intelligent analysis and decision-making enables dynamic regulation of media plane device traffic, optimizing network resource utilization and improving network performance and user experience.

[0151] Combine Figure 4 As shown, the present invention also provides a network traffic control method, comprising:

[0152] Obtaining historical traffic data of all media plane network elements connected to the control plane network element and the topological relationship of the media plane network elements;

[0153] Update the traffic control model of the control plane network element according to the historical traffic data and the topological relationship, wherein the traffic control model includes a deep reinforcement learning sub-model, a fusion sub-model, a graph neural network sub-model, and a Transformer sub-model;

[0154] Input the current traffic data of the current time period into the updated traffic control model, perform time series prediction through the fusion sub-model, and obtain the time series prediction result of the current traffic data; perform topological analysis through the graph neural network sub-model to obtain the graph structure of the media plane network element; perform weight allocation through the Transformer sub-model to obtain the allocation weight strategy for each media plane network element; then perform reinforcement learning through the deep reinforcement learning sub-model based on the time series prediction result, the graph structure, and the allocation weight strategy to obtain the traffic allocation strategy of the control plane network element;

[0155] The media traffic of each media plane network element is dynamically adjusted according to the traffic distribution strategy.

[0156] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned network traffic control method is implemented.

[0157] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the network traffic regulation method described above is implemented.

[0158] In other words, a non-volatile computer-readable storage medium stores a computer program, which, when executed by a processor, causes the processor to perform the following operations:

[0159] Obtaining historical traffic data of all media plane network elements connected to the control plane network element and the topological relationship of the media plane network elements;

[0160] Update the traffic control model of the control plane network element according to the historical traffic data and the topological relationship, wherein the traffic control model includes a deep reinforcement learning sub-model, a fusion sub-model, a graph neural network sub-model, and a Transformer sub-model;

[0161] Input the current traffic data of the current time period into the updated traffic control model, perform time series prediction through the fusion sub-model, and obtain the time series prediction result of the current traffic data; perform topological analysis through the graph neural network sub-model to obtain the graph structure of the media plane network element; perform weight allocation through the Transformer sub-model to obtain the allocation weight strategy for each media plane network element; then perform reinforcement learning through the deep reinforcement learning sub-model based on the time series prediction result, the graph structure, and the allocation weight strategy to obtain the traffic allocation strategy of the control plane network element;

[0162] The media traffic of each media plane network element is dynamically adjusted according to the traffic distribution strategy.

[0163] The advantages of the computer-readable storage medium of the present invention over the prior art are the same as the advantages of the above-mentioned network traffic control method over the prior art, and will not be repeated here.

[0164] Although the present invention is disclosed as above, the scope of protection disclosed by the present invention is not limited thereto. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will fall within the scope of protection of the present invention.

Claims

1. A network traffic control system, characterized in that: include: An information collection unit, configured to obtain historical traffic data of all media plane network elements connected to the control plane network element and the topological relationship of the media plane network elements; An AI model updating unit, configured to update a traffic control model of the control plane network element according to the historical traffic data and the topological relationship, wherein the traffic control model includes a deep reinforcement learning sub-model, a fusion sub-model, a graph neural network sub-model, and a Transformer sub-model; a strategy generating unit, configured to input the current traffic data of the current time period into the updated traffic control model, perform time series prediction through the fusion sub-model, and obtain a time series prediction result of the current traffic data; Specifically comprising: obtaining the operating status data, network bandwidth occupancy data, and load data of the media plane network element in the current time period, and inputting the operating status data, the network bandwidth occupancy data, and the load data as the current traffic data into the updated traffic control model; Perform feature extraction on the current traffic data through the CNN network of the fusion sub-model to obtain spatial features of the current traffic data; Perform feature extraction on the current traffic data according to the BiLSTM network of the fusion sub-model to obtain a time series of the current traffic data; Obtaining a traffic change trend of the media plane network element within a preset future time period based on the spatial characteristics and the time series, and using the traffic change trend as the time series prediction result; The graph neural network sub-model performs topological analysis to obtain the graph structure of the media plane network element; the Transformer sub-model performs weight allocation to obtain the allocation weight strategy for each media plane network element; and the deep reinforcement learning sub-model performs reinforcement learning based on the time series prediction results, the graph structure, and the allocation weight strategy to obtain the traffic allocation strategy for the control plane network element. Specifically comprising: performing reinforcement learning according to the time series prediction results, the graph structure, and the weight allocation strategy through the deep reinforcement learning sub-model; wherein, generating an initial traffic distribution adjustment strategy based on the time series prediction result, the graph structure, and the distribution weight strategy; Taking the initial traffic distribution adjustment strategy as an action and the digital twin environment as an environment, calculating the reward value after the interaction between the action and the environment through a reward function; determining whether reinforcement learning of the initial traffic distribution adjustment strategy is completed according to the reward value, and obtaining the traffic distribution strategy of the control plane network element when reinforcement learning of the initial traffic distribution adjustment strategy is completed; A policy execution unit is configured to dynamically adjust the media traffic of each media plane network element according to the traffic distribution policy.

2. The network traffic control system according to claim 1, characterized in that: The information collection unit is specifically used to: According to a preset collection cycle, the operating status data of all the media plane network elements within a preset time period is obtained from historical data, wherein the operating status data includes CPU occupancy and GPU occupancy; Determining network bandwidth occupancy data of the media plane network element based on data flow inflow and outflow information of the network interface of the media plane network element in the historical data; Obtaining load data of the media plane network elements according to the signaling throughput data of all the media plane network elements in the historical data; The operating status data, the network bandwidth occupancy data, and the load data are used as the historical traffic data.

3. The network traffic control system according to claim 1, characterized in that: The AI model updating unit is specifically used to: Inputting the historical traffic data and the topological relationship into the traffic control model for training; Verify the traffic distribution strategy output by the traffic control model during training through a digital twin environment and A / B testing; When the traffic distribution strategies output by the traffic control model during training meet the test requirements of the digital twin environment and the A / B test, the training of the traffic control model is stopped to obtain the updated traffic control model.

4. The network traffic control system according to claim 2, characterized in that: The strategy generation unit is specifically configured to: Inputting the topological relationship of the media plane network elements in the current time period into the updated traffic control model; The graph structure is constructed through the graph neural network sub-model according to the topological relationship, wherein the nodes of the graph structure are the media plane network elements, and the edges of the graph structure are the connection relationships between the media plane network elements.

5. The network traffic control system according to claim 2, characterized in that: The strategy generation unit is specifically configured to: Obtaining the operating status data, the network bandwidth occupancy data, and the load data of the media plane network element within the current time period, and inputting the operating status data, the network bandwidth occupancy data, and the load data as the current traffic data into the updated traffic control model; Using the Transformer sub-model, prediction is performed based on the current traffic data to obtain critical values of multiple preset indicators of each media plane network element within a preset future time period; A weight value of each of the preset indicators is generated according to the critical value to obtain the allocation weight strategy of each of the media plane network elements.

6. The network traffic control system according to claim 1, characterized in that: It also includes an early warning unit, which is used to: Obtaining a flow rate change trend within the current time period based on the current flow rate data of the current time period; According to the traffic change trend, determining whether there is a sudden increase in traffic on the media plane network element; When the traffic surge occurs in the media plane network element, an early warning message is sent.

7. A network traffic control method, characterized in that: include: Obtaining historical traffic data of all media plane network elements connected to the control plane network element and the topological relationship of the media plane network elements; Updating a traffic control model of the control plane network element according to the historical traffic data and the topological relationship, wherein the traffic control model includes a deep reinforcement learning sub-model, a fusion sub-model, a graph neural network sub-model, and a Transformer sub-model; Inputting the current traffic data of the current time period into the updated traffic control model, performing time series prediction through the fusion sub-model, and obtaining a time series prediction result of the current traffic data; Specifically comprising: obtaining the operating status data, network bandwidth occupancy data, and load data of the media plane network element in the current time period, and inputting the operating status data, the network bandwidth occupancy data, and the load data as the current traffic data into the updated traffic control model; Perform feature extraction on the current traffic data through the CNN network of the fusion sub-model to obtain spatial features of the current traffic data; Perform feature extraction on the current traffic data according to the BiLSTM network of the fusion sub-model to obtain a time series of the current traffic data; Obtaining a traffic change trend of the media plane network element within a preset future time period based on the spatial characteristics and the time series, and using the traffic change trend as the time series prediction result; Performing topological analysis through the graph neural network sub-model to obtain the graph structure of the media plane network element; performing weight allocation through the Transformer sub-model to obtain the allocation weight strategy for each media plane network element; then performing reinforcement learning through the deep reinforcement learning sub-model based on the timing prediction results, the graph structure, and the allocation weight strategy to obtain the traffic allocation strategy for the control plane network element; specifically comprising: performing reinforcement learning through the deep reinforcement learning sub-model based on the timing prediction results, the graph structure, and the allocation weight strategy; wherein, generating an initial traffic distribution adjustment strategy based on the time series prediction result, the graph structure, and the distribution weight strategy; Taking the initial traffic distribution adjustment strategy as an action and the digital twin environment as an environment, calculating the reward value after the interaction between the action and the environment through a reward function; determining whether reinforcement learning of the initial traffic distribution adjustment strategy is completed according to the reward value, and obtaining the traffic distribution strategy of the control plane network element when reinforcement learning of the initial traffic distribution adjustment strategy is completed; The media traffic of each media plane network element is dynamically adjusted according to the traffic distribution strategy.

8. A computer-readable storage medium, characterized in that The storage medium stores a computer program, and when the computer program is executed by the processor, the network traffic regulation method according to claim 7 is implemented.

Images