Satellite network traffic optimization method and system based on traffic prediction

Abstract

The invention provides a satellite network flow optimization method and system based on flow prediction, and the method comprises the steps: obtaining a satellite transmission path set from a satellite network topology, the satellite transmission path set comprises a plurality of satellite transmission paths, and each satellite transmission path comprises a plurality of inter-satellite links; based on the traffic data of each inter-satellite link in the historical time slice, adopting a pre-trained neural network model to predict the traffic data of each inter-satellite link in the next time slice; judging whether to carry out flow optimization or not based on the flow data of the inter-satellite link in the next time slice; if the traffic optimization is carried out, a bypass path between two satellites corresponding to the inter-satellite link is constructed, and the bypass path at least comprises the two satellites in the inter-satellite link on which the traffic optimization is carried out. According to the method, whether congestion occurs is judged by accurately predicting the flow, and the detour path between the two satellites is re-planned after congestion occurs, so that the communication quality is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of satellite networks, and in particular, to a method and system for optimizing satellite network traffic based on traffic prediction. Background Art

[0002] In recent years, global satellite communication networks for user access to communication needs have become a research hotspot. The global satellite communication network, also known as the satellite internetwork or Teledesic, is a global wireless multimedia internetwork that is superior to the traditional internet. This network is mainly composed of a satellite communication network system and can be combined with a terrestrial fiber optic network at the same time, achieving true global coverage, including areas such as vast seas, large deserts, and large plateaus that are difficult to reach by terrestrial networks.

[0003] The global satellite communication network usually consists of hundreds to thousands of low Earth orbit (LEO) satellites. For example, some networks are composed of 840 satellites, forming a network that completely covers the Earth. These satellites operate at high speeds in orbital space, providing uninterrupted communication services for ground users. Users can easily connect to the satellite internetwork through a small antenna (even only a few inches in size). Whether in the office, in the car, on the ship, on the plane, or in a remote desert or ocean, functions such as making calls, sending and receiving emails, browsing the internet, or watching dynamic images can be achieved; compared with traditional communication methods, the global satellite communication network has many advantages. It can achieve three-dimensional, all-weather, and all-region interconnection of all things globally, eliminating the "digital divide". At the same time, due to the relatively small transmission delay and low link loss of low-orbit satellites, the satellite internet has significant advantages in application scenarios that require rapid response. In addition, the satellite internet can also provide high-bandwidth communication services, meet users' needs for high-speed internet access, and has stronger connectivity and anti-destruction capabilities, providing more reliable communication guarantees in emergency situations such as natural disasters.

[0004] However, since each satellite has limited communication resources, when a satellite bears an overly heavy communication task, the communication quality often deteriorates due to excessive pressure, making it difficult to ensure the communication quality. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a method for optimizing satellite network traffic based on traffic prediction to eliminate or improve one or more defects existing in the prior art.

[0006] One aspect of the present invention provides a method for optimizing satellite network traffic based on traffic prediction. The steps of the method include:

[0007] Obtain a set of satellite transmission paths from the satellite network topology. The set of satellite transmission paths includes multiple satellite transmission paths, and each satellite transmission path includes multiple inter-satellite links;

[0008] Based on the traffic data of each inter-satellite link in the historical time slice, use a pre-trained neural network model to predict the traffic data of each inter-satellite link in the next time slice;

[0009] Determine whether to perform traffic optimization based on the traffic data of the inter-satellite link in the next time slice;

[0010] If traffic optimization is to be performed, construct a detour path between the two satellites corresponding to the inter-satellite link. The detour path includes at least two satellites in the inter-satellite link where traffic optimization is performed.

[0011] With the above solution, the present invention first studies the satellite network traffic prediction algorithm. Studying satellite network traffic prediction is of great significance. It helps to optimize resource allocation, improve network efficiency, reduce operating costs, and improve the user experience; by accurately predicting traffic trends, satellite communication resources can be planned and adjusted in advance to ensure sufficient bandwidth during peak demand periods, while avoiding resource waste during low demand periods. When it is found that a certain inter-satellite link is congested in the next time slice, the inter-satellite link can be regarded as congested throughout the time slice. Therefore, when planning the traffic in this time slice, the detour path between the two satellites can be re-planned to avoid this inter-satellite link in advance to achieve smooth traffic scheduling. This traffic optimization system can avoid possible congestion risks in advance and at the same time control the computational amount of the network to ensure communication quality.

[0012] In some embodiments of the present invention, the method is applied to a network controller of a satellite network. The network controller obtains a set of satellite transmission paths from the satellite network topology, and calls the traffic data of each inter-satellite link in the historical time slice to determine whether each inter-satellite link needs to perform traffic optimization. If traffic optimization is to be performed, a detour path between the two satellites corresponding to the inter-satellite link is constructed.

[0013] In some embodiments of the present invention, in the step of using a pre-trained neural network model to predict the traffic data of each inter-satellite link in the next time slice based on the traffic data of each inter-satellite link in the historical time slice, the traffic data of multiple historical time slices is constructed into a prediction vector, and the prediction vector is input into the pre-trained neural network model, and the neural network model outputs the predicted traffic data.

[0014] In some embodiments of the present invention, in the step of predicting the traffic data of each inter-satellite link in the next time slice by using a pre-trained neural network model based on the traffic data of each inter-satellite link in the historical time slices, the neural network model combines a connected graph convolutional network and a gated recurrent unit.

[0015] In some embodiments of the present invention, in the step of combining a connected graph convolutional network and a gated recurrent unit, the number of graph convolutional networks corresponds to the number of historical time slices, the graph convolutional networks and the gated recurrent units are in one-to-one correspondence, the gated recurrent units are connected based on the time order of the historical time slices, each graph convolutional network receives the traffic data of the corresponding historical time slice, and after processing, inputs it into the corresponding gated recurrent unit, and the predicted traffic data is output through the last gated recurrent unit in the order of the multiple gated recurrent units.

[0016] In some embodiments of the present invention, in the step that each graph convolutional network receives the traffic data of the corresponding historical time slice and inputs it into the corresponding gated recurrent unit after processing, the graph convolutional network is used to capture spatial dependence, and the gated recurrent unit is used for temporal dependence.

[0017] In some embodiments of the present invention, the gated recurrent unit adopts a recurrent neural network structure, and calculates the gating signal through the current input information X t and the hidden state H t-1 containing the information of the previous unit, where R t represents the reset gate, Z t represents the update gate, and the formula is as follows:

[0018] R t = σ(X t W xr + H t-1 W hr + b r ),

[0019] Z t = σ(X t W xz + H t-1 W hz + b z ),

[0020] where, W xr , W xz represent the weights during the training process, and b r , b z represent the offsets.

[0021] In some embodiments of the present invention, in the step of determining whether to perform traffic optimization based on the traffic data in the next time slice through the inter-satellite link, the traffic data in the next time slice is compared with a preset congestion threshold. If it is greater than the preset congestion threshold, it is determined that congestion has occurred and traffic optimization is required; if it is not greater than the preset congestion threshold, it is determined that no congestion has occurred and traffic optimization is not required.

[0022] In some embodiments of the present invention, in the step of constructing a detour path between two satellites corresponding to the inter-satellite link, the Dijkstra algorithm or the Bellman-Ford algorithm is used to construct the detour path between two satellites corresponding to the inter-satellite link.

[0023] The second aspect of the present invention further provides a satellite network traffic optimization system based on traffic prediction. The system includes a computer device, the computer device includes a processor and a memory, the memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the system implements the steps implemented by the method described above.

[0024] The third aspect of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps implemented by the aforementioned satellite network traffic optimization method based on traffic prediction.

[0025] The additional advantages, objectives, and features of the present invention will be partially described below and will become partially apparent to those of ordinary skill in the art after studying the following text, or can be learned from the practice of the present invention. The objectives and other advantages of the present invention can be pointed out and obtained specifically in the description and the drawings.

[0026] Those skilled in the art will understand that the objectives and advantages that can be achieved by the present invention are not limited to the above specific descriptions, and the above and other objectives that the present invention can achieve will be more clearly understood according to the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The drawings described herein are used to provide a further understanding of the present invention, form a part of this application, and do not limit the present invention.

[0028] Figure 1 It is a schematic diagram of an embodiment of the satellite network traffic optimization method based on traffic prediction of the present invention;

[0029] Figure 2 It is a schematic diagram of a second embodiment of the satellite network traffic optimization method based on traffic prediction of the present invention;

[0030] Figure 3Schematic diagram of the processing architecture of the neural network model of the present invention;

[0031] Figure 4 Schematic diagram for determining congestion in the present invention. Detailed implementation manners

[0032] To make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with the implementation manners and the drawings. Herein, the illustrative implementation manners of the present invention and their descriptions are used to explain the present invention, but not to limit the present invention.

[0033] Herein, it should also be noted that in order to avoid obscuring the present invention due to unnecessary details, only the structures and / or processing steps closely related to the solution of the present invention are shown in the drawings, while other details less related to the present invention are omitted.

[0034] As Figure 1 and 2 shown, the present invention proposes a satellite network traffic optimization method based on traffic prediction. The steps of the method include:

[0035] Step S100: Obtain a set of satellite transmission paths from the satellite network topology. The set of satellite transmission paths includes multiple satellite transmission paths, and each satellite transmission path includes multiple inter-satellite links;

[0036] In the specific implementation process, each satellite transmission path corresponds to a type of service.

[0037] Step S200: Based on the traffic data of each inter-satellite link in the historical time slice, use a pre-trained neural network model to predict the traffic data of each inter-satellite link in the next time slice;

[0038] In the specific implementation process, the present solution divides time into several time slices, and the network topology within each time slice can be regarded as unchanged. Such modeling can make the topology relatively stable within one time slice, which is convenient for traffic scheduling.

[0039] In some embodiments, the idea of complete time - slice partitioning means dividing time into several time - slices. Within each time - slice, no feeder switching occurs, that is, the inter - satellite topology remains unchanged and the feeder link remains unchanged, and feeder switching only occurs between time - slices. At the same time, we compare the predicted traffic of a certain link with a threshold. If the predicted traffic is higher than the threshold, regardless of whether this value can always be higher than the threshold throughout the time - slice, we will consider that this link is in a congested state in the next time - slice, and actively avoid this link when planning services, so that all traffic can be reasonably allocated. Therefore, a certain link can only be in two states within a certain time - slice, one is congested and the other is not congested. There will be no situation where a link is sometimes congested and sometimes not congested within a complete time - slice. This assumption can reduce the number of snapshots, thus making service switching not too frequent. Since the traffic volume is controlled, the computational amount of the service is reduced, which is beneficial to the stable operation of the entire system.

[0040] Step S300, determine whether to perform traffic optimization based on the traffic data of the inter - satellite link in the next time - slice;

[0041] Step S400, if traffic optimization is to be performed, construct a detour path between the two satellites corresponding to the inter - satellite link, and the detour path includes at least two satellites in the inter - satellite link for which traffic optimization is performed.

[0042] Specifically, check all links in the satellite network to find out which links have traffic higher than the threshold during a certain period in the next time - slice. If there are such links, they can be regarded as congested links in the next time - slice, and all services need to bypass such links.

[0043] After determining that an inter - satellite link needs to perform traffic optimization, construct a detour path between the two satellites. For all satellite transmission paths involving this inter - satellite link, use the detour path to complete data transmission.

[0044] With the above - mentioned solution, the present invention first studies the satellite network traffic prediction algorithm. Studying satellite network traffic prediction is of great significance. It helps to optimize resource allocation, improve network efficiency, reduce operating costs, and improve user experience. By accurately predicting traffic trends, satellite communication resources can be planned and adjusted in advance to ensure sufficient bandwidth during peak demand periods, while avoiding resource waste during low - demand periods. When it is found that a certain inter - satellite link is congested during the next time - slice, this inter - satellite link can be regarded as congested throughout the time - slice. Thus, when planning the traffic within this time - slice, by re - planning the detour path between the two satellites, the congested inter - satellite link can be avoided in advance to achieve smooth traffic scheduling. This traffic optimization system can avoid potential congestion risks in advance and control the computational amount of the network, ensuring communication quality.

[0045] In some embodiments of the present invention, the method is applied to a network controller of a satellite network. The network controller obtains a set of satellite transmission paths from the satellite network topology, and invokes the traffic data of each inter-satellite link in the historical time slice to determine whether each inter-satellite link needs traffic optimization. If traffic optimization is required, a detour path between the two satellites corresponding to the inter-satellite link is constructed.

[0046] In some embodiments of the present invention, in the step of predicting the traffic data of each inter-satellite link in the next time slice by using a pre-trained neural network model based on the traffic data of each inter-satellite link in the historical time slice, the traffic data of multiple historical time slices is constructed into a prediction vector, and the prediction vector is input into the pre-trained neural network model, and the neural network model outputs the predicted traffic data.

[0047] As Figure 3 shown, in some embodiments of the present invention, in the step of predicting the traffic data of each inter-satellite link in the next time slice by using a pre-trained neural network model based on the traffic data of each inter-satellite link in the historical time slice, the neural network model combines a connected graph convolutional network and a gated recurrent unit.

[0048] In some embodiments of the present invention, in the step of combining a connected graph convolutional network and a gated recurrent unit, the number of graph convolutional networks corresponds to the number of historical time slices, the graph convolutional networks and the gated recurrent units are in one-to-one correspondence, the gated recurrent units are connected based on the time order of the historical time slices, each graph convolutional network receives the traffic data of the corresponding historical time slice, and after processing, inputs it into the corresponding gated recurrent unit, and the predicted traffic data is output through the last gated recurrent unit in the order of the multiple gated recurrent units.

[0049] In some embodiments of the present invention, in the step that each graph convolutional network receives the traffic data of the corresponding historical time slice and inputs it into the corresponding gated recurrent unit after processing, the graph convolutional network is used to capture spatial dependence, and the gated recurrent unit is used to capture temporal dependence.

[0050] Adopting the above solution, the neural network algorithm based on the graph convolutional network (GCN) and the gated recurrent unit (GRU) is run to predict the traffic. The GCN is used to learn complex topological structures to capture spatial dependence, while the GRU is used to learn the dynamic changes of traffic data to capture temporal dependence. By capturing the dual dependence of time and space, the trend of traffic development can be well simulated, thus realizing accurate traffic prediction.

[0051] This solution combines GCN and GRU, where GCN is used to learn complex topological structures to capture spatial dependencies, while GRU is used to learn the dynamic changes in traffic data to capture temporal dependencies. The input of this algorithm is to divide the topology G(V, E) according to the ground area (where V represents the set of satellite links, with a total of N, and E represents the connection relationship between satellite links), and each area is based on the traffic dataset of Internet users in a time series dataset, which contains the traffic matrix set X N ={X t-n ,…,X t-1 ,X t}. The output is the predicted value X t+1 of the network traffic for each node in the traffic network in the next time interval. For each link node v, it has its attribute x (i,t) , which represents the satellite network traffic of the link represented by the i-th node v i in the t-th time period, thus forming the traffic matrix X t , and this value is the input of the GCN model. The GCN model uses the structural information of the graph for convolutional calculation, extracts the spatial characteristics of satellite traffic, and obtains a new traffic matrix X' t .

[0052] In some embodiments of the present invention, the gated recurrent unit adopts a recurrent neural network structure, and calculates the gating signal through the current input information X t and the hidden state H t-1 containing the information of the previous unit, where R t represents the reset gate, and Z t represents the update gate, and the formulas are as follows:

[0053] R t =σ(X t W xr +H t-1 W hr +b r ),

[0054] Z t =σ(X t W xz +H t-1 W hz+ b z ),

[0055] where W xr , W xz represent the weights during the training process, and b r , b z represent the offsets.

[0056] In the specific implementation process, after calculating the gating signal, use the reset gate to perform matrix element multiplication, and then combine the result with X t and scale it to the range of (-1, +1) using the activation function tanh. Its candidate hidden state is:

[0057]

[0058] where, W xh and W hh represent the weights during the training process, and b h represents the offset.

[0059] Finally, calculate the output of the hidden state at time step t, which combines the effect of the update gate Z t the previous hidden state H t-1 and the new candidate hidden state of this part to calculate the final output:

[0060]

[0061] Therefore, this solution uses the GCN calculation results with spatial attributes at different historical moments as the input of the GRU. Through the per-unit correlation calculation of the GRU, it extracts the time attributes of the traffic trend and realizes the final prediction matrix X t+1 .

[0062] Specifically, this solution first studied the satellite network traffic prediction algorithm. Studying satellite network traffic prediction is of great significance, which helps to optimize resource allocation, improve network efficiency, reduce operating costs, and enhance user experience. By accurately predicting traffic trends, satellite communication resources can be planned and adjusted in advance to ensure sufficient bandwidth during peak demand periods while avoiding resource waste during low demand periods. This is crucial for maintaining the stability and reliability of the global satellite communication network, especially in the context of growing data transmission demands and complex and changing communication environments. In addition, effective traffic prediction can help operators identify potential security threats and network attacks, enabling them to take preventive measures to ensure communication security. Currently, commonly used traffic prediction algorithms include the historical average model that uses the average traffic information in the historical period as the prediction value, the autoregressive integrated moving average model that fits a parametric model to the observed time series to predict future traffic data, and the support vector regression model that uses historical data to train the model and obtain the relationship between the input and output, and then predicts future traffic data through the trained model. The prediction performances of these models have their own advantages and disadvantages, but none of them can simultaneously capture the temporal and spatial dependencies of the satellite network well. Therefore, there are obvious gaps between the simulation results of the prediction and the actual values, making them unsuitable for high-precision traffic prediction services. To solve this problem, the present invention proposes a traffic prediction algorithm that combines a graph convolutional network (GCN) and a gated recurrent unit (GRU). The GCN is used to learn complex topological structures to capture spatial dependencies, while the GRU is used to learn the dynamic changes of traffic data to capture temporal dependencies. The accuracy of the simulation prediction results using this prediction algorithm can reach over 90%, thus well meeting the requirements of high-precision prediction.

[0063] As Figure 4 shown, in some embodiments of the present invention, in the step of determining whether to perform traffic optimization based on the traffic data in the next time slice of the inter-satellite link, the traffic data in the next time slice is compared with a preset congestion threshold. If it is greater than the preset congestion threshold, it is determined that congestion has occurred and traffic optimization is required; if it is not greater than the preset congestion threshold, it is determined that no congestion has occurred and traffic optimization is not required.

[0064] Specifically, within any arbitrarily specified time slice, it is modeled that it is completely in either a congested or non-congested state. It is not considered that there is a situation where a part of a link is in a congested state and the remaining time is in an idle state within a complete time slice. This assumption can bring the following two benefits: First, it can reduce the number of snapshots, thus preventing the service switching from being too frequent. Second, since the traffic volume is controlled, the computational amount of the service is reduced, which is beneficial to the stable operation of the entire system.

[0065] For links with a value higher than the threshold, these links are directly considered congested within the next entire time slice. This solution can reconstruct a new topology at the end of the current time slice according to this rule and the changes in the actual feeder links.

[0066] Using the above solution, this solution compares the predicted traffic of a certain link with the threshold. If the predicted traffic is higher than the threshold, regardless of whether this value is always higher than the threshold throughout the time slice, this solution will consider that the link is in a congested state within the next time slice, and actively avoid this link when planning services, so that all traffic can be reasonably optimized, thereby reducing the handover frequency caused by congestion.

[0067] Specifically, through the division of time slices, within a time slice, we consider that the inter-satellite links and feeder links remain unchanged. This modeling method can make the topology relatively stable within a time slice, facilitating traffic scheduling and in-depth research. This time interval cannot be set too long or too short. If it is set too long, the accuracy of the traffic optimization system will deteriorate; if it is set too short, routing calculations will be performed frequently, increasing the computational complexity.

[0068] In some embodiments of the present invention, in the step of constructing a detour path between two satellites corresponding to an inter-satellite link, the Dijkstra algorithm or the Bellman-Ford algorithm is used to construct a detour path between two satellites corresponding to the inter-satellite link.

[0069] This solution uses a first rule to construct a detour path between two satellites corresponding to an inter-satellite link. The first setting rule includes:

[0070] Dijkstra algorithm: Find the path with the minimum cost between two points;

[0071] Bellman-Ford algorithm: An algorithm for finding the shortest paths from a single source point to all other vertices, which can handle graphs containing negative-weight edges.

[0072] And / or software-defined network: Select a suitable path according to various factors, not limited to the path cost.

[0073] Dijkstra algorithm: An algorithm for finding the shortest paths from a single source point to all other vertices in a weighted graph. By maintaining a set of vertices to distinguish vertices with determined shortest paths and vertices with undetermined shortest paths, it gradually expands from the source point. Each time, it selects the vertex closest to the source point among the unprocessed vertices and updates the distances of its adjacent vertices until all vertices are processed. This algorithm is applicable to graphs without negative-weight edges and can ensure that the found path is the shortest, thus ensuring that the path selected by routing is the one with the lowest cost.

[0074] And / or Bellman-Ford algorithm: A dynamic routing algorithm used to find the shortest paths from a single source vertex to all other vertices in a weighted graph. It can handle graphs containing negative-weight edges and detect whether there are negative-weight cycles in the graph. It gradually approximates the true shortest paths by iteratively updating distance estimates until no further optimization is possible.

[0075] And / or Software-Defined Network: A technology that utilizes the centralized control feature of the SDN architecture to optimize network traffic management and path selection. It can dynamically monitor the network state and calculate the optimal path based on global information, while supporting multi-path transmission to improve bandwidth utilization and throughput. By integrating advanced technologies such as artificial intelligence and meta-heuristic algorithms, it realizes more efficient, flexible, and intelligent network routing management to adapt to the changing network environment and business requirements.

[0076] Specifically, after proposing a high-precision traffic prediction algorithm, satellite network traffic optimization can then be carried out based on the traffic prediction algorithm. Currently, commonly used routing algorithms include Dijkstra's algorithm and Software-Defined Network algorithm. Among them, Dijkstra's algorithm calculates and selects a routing route with the least path cost. However, when a link becomes congested, this algorithm will not change the link, so it will continuously input traffic to the congested link, and eventually, the link will experience packet loss due to excessive load. The Software-Defined Network algorithm, on the other hand, can select a suitable routing route based on various factors. It can implement routing detours by the network controller updating the flow table of router nodes when a link becomes congested, thereby avoiding the risk of packet loss due to excessive load. Although the Software-Defined Network algorithm can achieve a certain degree of congestion control, it can only be regulated when congestion occurs, so a part of the services have already degraded or been re-routed. Therefore, this algorithm also cannot meet the scenario of large traffic volumes. Currently, common algorithms for satellite network traffic optimization based on the traffic prediction algorithm include ELMDR (Distributed Routing Algorithm Based on Extreme Learning Machine) and HLBR (Load Balancing Routing Algorithm Based on Hybrid Traffic Bypass). Among them, ELMDR considers the traffic distribution density on the earth's surface and uses the machine learning algorithm of extreme learning machine to predict the traffic load of satellite nodes in advance, and the nodes make routing decisions based on the traffic prediction results. HLBR predicts the areas in the network where cascade congestion is likely to occur through prior geographical information and real-time network status, and then calculates the optimal bypass path to make the service traffic bypass the congested areas to achieve load balancing. Both of these algorithms can achieve a certain degree of smooth congestion control and realize traffic load balancing. However, due to the instability of the prediction algorithm, a certain link will be frequently and short-term judged as congested, resulting in the need for continuous re-routing of the entire network topology, which will greatly increase the network's computational workload and burden.

[0077] The beneficial effects of this solution at least include: combining a graph convolutional network (GCN) and a gated recurrent unit (GRU). The GCN is used to learn complex topological structures to capture spatial dependencies, while the GRU is used to learn the dynamic changes of traffic data to capture temporal dependencies. This traffic prediction method can greatly improve the prediction accuracy. At the same time, based on the proposed traffic prediction algorithm, traffic optimization is carried out. The idea of complete time slice division is used for modeling to ensure that a link can only be in two states within a time slice, one is congestion and the other is non-congestion. There will be no situation where a link is sometimes congested and sometimes non-congested within a complete time slice. This reduces the number of snapshots and makes service switching not too frequent. Since the traffic volume is controlled, the computational amount of the service is reduced, thus realizing the early avoidance of possible congestion risks through the traffic prediction algorithm and controlling the network computational amount as much as possible.

[0078] Furthermore, the present invention uses a satellite network with a centralized control architecture. The data plane and the control plane work separately. The network controller in the control plane is responsible for collecting and monitoring network status data, implementing network management, orchestrating services, and issuing transmission policies. The on-board router in the data plane performs matching and forwarding according to the instructions issued by the network controller.

[0079] The present invention uses a satellite network with a centralized control architecture. The data plane and the control plane work separately. The network controller in the control plane is responsible for collecting and monitoring network status data, implementing network management, orchestrating services, and issuing transmission policies. The on-board router in the data plane performs matching and forwarding according to the instructions issued by the network controller.

[0080] In summary, this method proposes a traffic optimization system for a satellite network based on traffic prediction. This method is executed on the network controller of the satellite network. The network controller uses a neural network algorithm to predict the next-stage network load according to the link traffic conditions in the previous time slice. If it is found that a certain inter-satellite link is congested in the next time slice, the inter-satellite link can be regarded as congested throughout the time slice. Thus, when planning the traffic in this time slice, we can avoid this inter-satellite link in advance to achieve smooth traffic scheduling. This traffic optimization system can avoid possible congestion risks in advance and control the computational amount of the network.

[0081] Compared with the prior art, the common algorithms for satellite network traffic optimization based on traffic prediction algorithms currently include ELMDR (Distributed Routing Algorithm Based on Extreme Learning Machine) and HLBR (Load Balancing Routing Algorithm Based on Hybrid Traffic Bypass). Among them, ELMDR considers the traffic distribution density on the earth's surface and uses the extreme learning machine machine learning algorithm to predict the traffic load of satellite nodes in advance. Nodes make routing decisions based on the traffic prediction results. HLBR predicts the areas in the network where cascading congestion is likely to occur through prior geographical information and real-time network status, and then calculates the optimal bypass path to make the service traffic bypass the congested areas to achieve load balancing. Both of these algorithms can achieve a certain degree of smooth congestion control and achieve traffic load balancing. However, due to the instability of the prediction algorithm, a certain link will be frequently and short-term judged as congested, resulting in the need to continuously re-route the entire network topology, which will greatly increase the network's computational complexity and burden. This solution is executed on the network controller of the satellite network. The network controller uses a neural network algorithm to predict the next-stage network load based on the link traffic conditions in the previous time slice. If it is found that a certain inter-satellite link is congested in the next time slice, the inter-satellite link can be regarded as congested throughout the time slice, so as to avoid this inter-satellite link in advance when planning the traffic in this time slice to achieve smooth traffic scheduling. This traffic optimization system can avoid potential congestion risks in advance and at the same time control the network's computational complexity.

[0082] An embodiment of the present invention also provides a satellite network traffic optimization system based on traffic prediction. The system includes a computer device, the computer device includes a processor and a memory, the memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the system implements the steps achieved by the method described above.

[0083] An embodiment of 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, it implements the steps achieved by the aforementioned satellite network traffic optimization method based on traffic prediction. The computer-readable storage medium can be a tangible storage medium, such as a random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, register, floppy disk, hard disk, removable storage disk, CD-ROM, or any other form of storage medium well-known in the technical field.

[0084] Those of ordinary skill in the art should understand that the various exemplary components, systems, and methods described in connection with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Specifically, whether to implement it in hardware or software depends on the specific application and design constraints of the technical solution. Professionals can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the present invention. When implemented in hardware, it can be, for example, an electronic circuit, an application-specific integrated circuit (ASIC), appropriate firmware, a plug-in, a functional card, and so on. When implemented in software, the elements of the present invention are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium or transmitted through a data signal carried in a carrier wave on a transmission medium or a communication link.

[0085] It should be clear that the present invention is not limited to the specific configurations and processes described above and illustrated in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and illustrated as examples. However, the method process of the present invention is not limited to the specific steps described and illustrated. Those skilled in the art can make various changes, modifications, and additions, or change the order between steps after understanding the spirit of the present invention.

[0086] In the present invention, the features described and / or illustrated for one embodiment can be used in the same or a similar manner in one or more other embodiments, and / or combined with the features of other embodiments or replace the features of other embodiments.

[0087] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, various changes and modifications can be made to the embodiments of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A satellite network traffic optimization method based on traffic prediction, characterized in that: The steps of the method include: Acquire a satellite transmission path set from a satellite network topology, wherein the satellite transmission path set includes a plurality of satellite transmission paths, and each satellite transmission path includes a plurality of inter-satellite links; Based on the traffic data of each intersatellite link in the historical time slice, a pre-trained neural network model is used to predict the traffic data of each intersatellite link in the next time slice; Determine whether to perform traffic optimization based on the traffic data of the intersatellite link in the next time slice; If traffic optimization is performed, a detour path between two satellites corresponding to the intersatellite link is constructed, and the detour path includes at least two satellites in the intersatellite link for which traffic optimization is performed.

2. The satellite network traffic optimization method based on traffic prediction according to claim 1 is characterized in that: The method is applied to a network controller of a satellite network. The network controller obtains a set of satellite transmission paths from the satellite network topology, and calls the traffic data of each inter-satellite link in a historical time slice to determine whether each inter-satellite link needs to be traffic tuned. If traffic tuning is required, a detour path between two satellites corresponding to the inter-satellite link is constructed.

3. The satellite network traffic optimization method based on traffic prediction according to claim 1 is characterized in that: In the step of predicting the traffic data of each intersatellite link in the next time slice based on the traffic data of each intersatellite link in the historical time slice using a pre-trained neural network model, the traffic data of multiple historical time slices are constructed into a prediction vector, the prediction vector is input into the pre-trained neural network model, and the neural network model outputs the predicted traffic data.

4. The satellite network traffic optimization method based on traffic prediction according to claim 1 is characterized in that: In the step of predicting the traffic data of each intersatellite link in the next time slice using a pre-trained neural network model based on the traffic data of each intersatellite link in the historical time slice, the neural network model is combined with a connected graph convolutional network and a gated recurrent unit.

5. The satellite network traffic optimization method based on traffic prediction according to claim 4 is characterized in that: In the step of combining connected graph convolutional networks and gated recurrent units, the number of graph convolutional networks corresponds to the number of historical time slices, the graph convolutional networks and gated recurrent units correspond one to one, the gated recurrent units are connected based on the time sequence of the historical time slices, each graph convolutional network receives the traffic data of the corresponding historical time slice, and inputs it into the corresponding gated recurrent unit after processing, and outputs the predicted traffic data through the last gated recurrent unit in the sequence of multiple gated recurrent units.

6. The satellite network traffic optimization method based on traffic prediction according to claim 5 is characterized in that: Each graph convolutional network receives the traffic data of the corresponding historical time slice, and in the step of inputting it into the corresponding gated recurrent unit after processing, the graph convolutional network is used to capture spatial dependency, and the gated recurrent unit is used for temporal dependency.

7. The satellite network traffic optimization method based on traffic prediction according to claim 6 is characterized in that: The gated recurrent unit adopts a recurrent neural network structure, through the current input information X t and a hidden H containing the information of the previous unit t-1 Calculate the gate signal, where R t Represents the control gate, Z t represents the update gate, and the formula is as follows: R t =σ(X t W xr +H t-1 W hr +b r ), Z t =σ(X t W xz +H t-1 W hz +b z ), Among them, W xr , W xz represents the weight during training, b r 、b z Indicates the offset.

8. The satellite network traffic optimization method based on traffic prediction according to any one of claims 1 to 7, characterized in that: In the step of determining whether to perform traffic optimization based on the traffic data of the intersatellite link in the next time slice, the traffic data of the next time slice is compared with a preset congestion threshold. If it is greater than the preset congestion threshold, it is determined that congestion occurs and traffic optimization is required; if it is not greater than the preset congestion threshold, it is determined that congestion does not occur and traffic optimization is not required.

9. The satellite network traffic optimization method based on traffic prediction according to claim 2 is characterized in that: In the step of constructing a circumnavigation path between two satellites corresponding to the inter-satellite link, the Dijkstra algorithm or the Bellman-Ford algorithm is used to construct the circumnavigation path between the two satellites corresponding to the inter-satellite link.

10. A satellite network traffic optimization system based on traffic prediction, characterized in that: The system includes a computer device, which includes a processor and a memory. The memory stores computer instructions. The processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the system implements the steps implemented by the method as described in any one of claims 1 to 9.

Images