Intelligent routing method and device applied to multi-objective optimization in low-orbit satellite optical network
By optimizing routes through geographic grid partitioning and the DDPG algorithm, the problem of high routing computation complexity in low-Earth orbit satellite optical networks is solved, achieving efficient and reliable dynamic path selection and fault response, thereby improving network performance.
Patent Information
- Application Number
- CN202511191391.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-25
- Publication Date
- 2025-11-18
AI Technical Summary
In low-Earth orbit satellite optical networks, routing computation is complex and difficult to meet real-time requirements. Traditional static routing algorithms cannot cope with rapid changes in network status, leading to network performance degradation and link congestion.
The satellite network is divided into abstract regions using geographic grid partitioning technology. An intelligent routing model is constructed by combining the Deep Deterministic Policy Gradient (DDPG) algorithm. Path selection is optimized through reinforcement learning, and routes are dynamically reconstructed when links are interrupted or fail.
It improves network management efficiency and flexibility, enhances the system's robustness in case of link anomalies, ensures communication continuity and reliability, reduces the risk of link interruption, and improves overall communication performance.
Smart Images
Figure CN120979558A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] Embodiments of the present disclosure relate to the field of satellite-to-ground link switching, and in particular to an intelligent routing method and device for multi-objective optimization in a low earth orbit satellite optical network. BACKGROUND
[0002] In a mega low earth orbit satellite optical network, routing computation faces extremely high complexity and challenges. With the continuous expansion of the low earth orbit satellite constellation scale, the number of network nodes increases dramatically, and the satellites move at high speed in orbit, resulting in continuous changes in network topology. This dynamic characteristic makes it difficult for traditional static routing algorithms to cope with the frequent changes in network status, especially in terms of traffic load, link quality and satellite resources. Since routing computation usually needs to be globally optimized within the entire constellation, the complexity and time overhead of computation are extremely large, and it is often difficult to meet the real-time requirements. Especially in high-density satellite groups, routing decisions need to consider a large number of network parameters and constraints, such as link load, delay, bandwidth, computing power of satellite nodes, etc., which makes each routing computation consume a large amount of computing resources and time.
[0003] Another important problem is the excessive division of time slices. In a large-scale satellite network, in order to reduce the computational complexity, the network state is usually divided into multiple time slices for segmented computation. However, excessive division of time slices often leads to deviation between routing decisions and actual network status. Due to the high-speed change of satellite orbits and the time-varying characteristics of network load, excessive division of time slices makes it impossible for routing computation to respond to rapid changes in network status in a timely manner, and it is impossible to ensure real-time accurate service transmission. In this way, even if the routing computation reaches the optimal state within a certain time slice, it may not be able to provide the best transmission path in actual application, resulting in a decline in network performance, and even possible link congestion or service interruption. SUMMARY
[0004] The summary of the present disclosure is intended to introduce the concepts in a simplified form, which will be described in detail in the specific embodiments section below. The summary of the present disclosure is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.
[0005] Some embodiments of the present disclosure propose an intelligent routing method, device, electronic equipment and computer readable medium for multi-objective optimization in a low earth orbit satellite optical network, to solve the technical problems mentioned in the background section above.
[0006] In a first aspect, some embodiments of the present disclosure provide an intelligent routing method for multi-objective optimization in a low-orbit satellite optical network, the method comprising: dividing a low-orbit satellite network into a set of abstract regions according to latitude and longitude ranges by a geographic grid division technique, wherein each abstract region represents different satellite coverage ranges and communication demands; determining a set of satellite nodes corresponding to each abstract region and performing routing strategy initialization processing on the satellite nodes of each abstract region; determining a network structure of an initial multi-objective optimization intelligent routing model, wherein the initial multi-objective optimization intelligent routing model comprises an actor network and an evaluation network; training the initial multi-objective optimization intelligent routing model according to the set of abstract regions and the corresponding sets of satellite nodes to obtain a trained multi-objective optimization intelligent routing model; during a service flow transmission process, if it is detected that a certain communication link cannot be transmitted or a certain satellite node is unavailable, reconstructing a routing strategy for a target abstract region to obtain a reconstructed routing path table, and inputting the reconstructed routing path table into the multi-objective optimization intelligent routing model to obtain an optimized routing strategy.
[0007] In a second aspect, some embodiments of the present disclosure provide an intelligent routing device for multi-objective optimization in a low-orbit satellite optical network, the device comprising: a division unit configured to divide a low-orbit satellite network into a set of abstract regions according to latitude and longitude ranges by a geographic grid division technique, wherein each abstract region represents different satellite coverage ranges and communication demands; an initialization unit configured to determine a set of satellite nodes corresponding to each abstract region and perform routing strategy initialization processing on the satellite nodes of each abstract region; a determination unit configured to determine a network structure of an initial multi-objective optimization intelligent routing model, wherein the initial multi-objective optimization intelligent routing model comprises an actor network and an evaluation network; a training unit configured to train the initial multi-objective optimization intelligent routing model according to the set of abstract regions and the corresponding sets of satellite nodes to obtain a trained multi-objective optimization intelligent routing model; and a reconstruction unit configured to, during a service flow transmission process, if it is detected that a certain communication link cannot be transmitted or a certain satellite node is unavailable, reconstruct a routing strategy for a target abstract region to obtain a reconstructed routing path table, and input the reconstructed routing path table into the multi-objective optimization intelligent routing model to obtain an optimized routing strategy.
[0008] In a third aspect, some embodiments of the present disclosure provide an electronic device, comprising: one or more processors; a storage device having one or more programs stored thereon, when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any of the implementations of the first aspect.
[0009] In a fourth aspect, some embodiments of the present disclosure provide a computer readable medium having stored thereon a computer program, wherein the program, when executed by a processor, implements the method described in any implementation manner of the first aspect.
[0010] The above various embodiments of the present disclosure have the following beneficial effects: Through the application of the intelligent routing method for multi-objective optimization in a low-orbit satellite optical network, some embodiments of the present disclosure first divide the low-orbit satellite network into multiple abstract regions by using a geographic grid technology. According to the latitude and longitude range of the earth's surface, the satellites within the coverage range are classified into independent virtual network units. The region division is dynamically adjusted according to specific geographic conditions and user needs. In high-density areas (such as cities and industrial centers), the grid division is more refined in order to accurately manage communication resources; while in low-density areas, larger grids are used to reduce the redundant management overhead. Through this dynamic grid division method, each region can adapt to changes in satellite orbits and business flow demands, ensuring the optimal configuration and management efficiency of communication resources. On this basis, a deep deterministic policy gradient (DDPG) algorithm is used to build and train an intelligent routing model to optimize the path selection from the source satellite to the target satellite. The model is trained by continuously injecting business flow data, and its input state includes link weights (weighted ratio of delay and traffic), regional network topology, and the current node state; the output is the action decision of selecting the next hop satellite. The reward mechanism design is particularly critical: when the number of routing hops exceeds the number of satellites within the region and the target satellite cannot be found, a negative reward is given to punish inefficient paths; while the link weight is low (short delay and light traffic load), the model will receive a higher reward, encouraging the selection of efficient paths. Through this reinforcement learning process, the model can gradually learn the optimal routing strategy in the network environment, dynamically adapting to network load and link changes. To deal with uncontrollable factors such as link interruption or satellite failure, the algorithm combines real-time network status to exclude faulty nodes and re-plan feasible paths, ensuring the continuity and reliability of communication. In addition, for cross-regional communication needs, the obstacle avoidance algorithm also combines the topology information of adjacent regions for global optimization to avoid overall communication interruption or efficiency decline due to obstacles within a single region. This method has shown significant advantages in the application of giant low-orbit satellite optical networks. Through geographic grid division, the network management efficiency and flexibility are improved, the DDPG routing model realizes dynamic and efficient path selection, and the intelligent obstacle avoidance routing algorithm enhances the robustness of the system in abnormal link conditions. Overall, the present invention provides an innovative solution for complex routing problems in low-orbit satellite networks, especially suitable for large-scale, high-dynamic, and delay-sensitive satellite communication scenarios. BRIEF DESCRIPTION OF DRAWINGS
[0011] The above and other features, advantages, and aspects of embodiments of the present disclosure will become more apparent by describing in detail some embodiments thereof with reference to the annexed drawings in which: like reference numerals refer to like elements throughout. The annexed drawings are intended for purposes of illustration only and shall not limit the scope of the disclosure. The drawings are not necessarily drawn to scale, with an emphasis instead being placed upon illustrating the principles of the embodiments.
[0012] Figure 1 is a flowchart of some embodiments of an intelligent routing method for multi-objective optimization in low earth orbit satellite optical networks according to the present disclosure;
[0013] Figure 2 is a routing reconstruction schematic based on regional obstacle avoidance;
[0014] Figure 3 is a structural schematic of some embodiments of an intelligent routing apparatus for multi-objective optimization in low earth orbit satellite optical networks according to the present disclosure;
[0015] Figure 4 is a structural schematic of an electronic device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION
[0016] Embodiments of the present disclosure will be described herein below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure will be more thoroughly and completely understood. It should be understood that the drawings and embodiments of the present disclosure are only for illustrative purposes and are not intended to limit the scope of protection of the present disclosure.
[0017] It should also be noted that only parts related to the present application are shown in the drawings for the purpose of description. The embodiments in the present disclosure and the features in the embodiments can be combined with each other without conflict.
[0018] It should be noted that the terms "first", "second", and the like in the present disclosure are only used to distinguish different devices, modules, or units, and do not limit the functions performed by these devices, modules, or units or the mutual dependency therebetween.
[0019] It should be noted that the terms "one", "multiple" in the present disclosure are illustrative and not restrictive, and those skilled in the art should understand that "one or more" should be understood unless otherwise explicitly stated in the context.
[0020] The names of messages or information exchanged between the devices in the embodiments of the present disclosure are only for illustrative purposes and do not limit the scope of the messages or information.
[0021] The present disclosure will be described in detail below with reference to the attached drawings and embodiments.
[0022] Figure 1 Flow 100 of some embodiments of the intelligent routing method for multi-objective optimization applied in low-orbit satellite optical networks of some embodiments of the present disclosure. The intelligent routing method for multi-objective optimization applied in low-orbit satellite optical networks includes the following steps:
[0023] Step 101, divide the low-orbit satellite network into a set of abstract regions according to latitude and longitude ranges by a geographic grid division technique.
[0024] In some embodiments, the execution subject (e.g., a computing device) of the intelligent routing method for multi-objective optimization applied in low-orbit satellite optical networks can divide the low-orbit satellite network into a set of abstract regions according to latitude and longitude ranges by a geographic grid division technique. Each abstract region represents different satellite coverage and communication requirements. First, the low-orbit satellite network is divided into several abstract regions according to latitude and longitude ranges by a geographic grid division technique. Each region represents different satellite coverage and communication requirements.
[0025] Step 102, determine the set of satellite nodes corresponding to each abstract region, and perform routing strategy initialization processing on the satellite nodes of each abstract region.
[0026] In some embodiments, the execution subject described above can determine the set of satellite nodes corresponding to each abstract region, and perform routing strategy initialization processing on the satellite nodes of each abstract region. For example, the position relationship between each satellite and the abstract region can be calculated to determine the set of satellite nodes corresponding to the abstract region. This regional design not only effectively abstracts the complex satellite network structure, but also provides a clear training boundary and target for the intelligent routing model. On this basis, a typical region is selected from each latitude as a training group to capture the diversity of different geographical regions in network characteristics, laying a solid foundation for subsequent model optimization.
[0027] In practice, the execution subject described above can perform routing strategy initialization processing on the satellite nodes of each abstract region:
[0028] In the first step, time is divided into time slots in seconds, and a path change rate is introduced in each time slot. During data initialization (node state, link state, etc. of each abstract region), time is finely divided into time slots in seconds for real-time monitoring and updating of the link state of the satellite network. In these time slots, the path change rate λ is introduced as an indicator to measure the frequency of changes in the shortest path from the source node to the target node. The path change marker α further clarifies the link dynamics: if the path changes, α takes the value of 1; if there is no change, it takes the value of 0. In this way, the dynamic characteristics of the links in the satellite network can be accurately captured, providing reliable data input for the training phase and providing a more comprehensive basis for intelligent routing algorithms to deal with complex scenarios.
[0029] In the second step, the delay characteristics corresponding to the satellite node set are divided into communication delay and inter-satellite link setting delay. To enhance the adaptability of model training to network characteristics, the delay characteristics are subdivided into communication delay and LISL (Laser Inter-Satellite Link) setting delay. Communication delay includes propagation delay and node delay, the former is determined by the ratio of link length to the speed of light, and the latter is composed of node processing time and queuing time. While LISL setting delay (Inter-Satellite Link Setting Delay) is closely related to the ATP (Alignment, Acquisition, Tracking and Pointing) process, it is usually about 100ms, reflecting the time overhead of link establishment and adjustment. These delay indicators play a key role in model state input, providing a more detailed calculation dimension for routing optimization.
[0030] In the third step, the link information corresponding to the satellite node set is sent to the corresponding controller to generate the corresponding initial routing strategy. After completing data initialization, the network control architecture ensures smooth execution from data acquisition to routing optimization. The remote control center receives the link information through the ground station and passes it to each SDN controller. The initial routing strategy can refer to the routing strategy between each satellite node of the satellite node set corresponding to the abstract region.
[0031] In the fourth step, in response to receiving the initial routing strategy sent by the controller, the initial routing strategy is synchronized to each satellite node in the satellite node set. These controllers pass the optimized routing strategy to each satellite node in the satellite node set according to the predefined southbound interface protocol. By dynamically updating the transmission direction and path of data flow, the entire intelligent routing system can always maintain efficient and reliable operation in the face of complex and dynamically changing satellite network environments.
[0032] Step 103, determine the network structure of the initial multi-objective optimization intelligent routing model.
[0033] In some embodiments, the execution subject can determine the network structure of the initial multi-objective optimization intelligent routing model. Wherein, the initial multi-objective optimization intelligent routing model includes an actor network and an evaluation network. In the training of the initial multi-objective optimization intelligent routing model, it is mainly composed of deep deterministic policy gradient (DDPG) and distributed multi-agent training mechanism. First, DDPG is a deep reinforcement learning method suitable for continuous action space, which can efficiently handle dynamic routing problems in low-orbit satellite networks. Then, through the distributed training mechanism, each satellite region shares model parameters to achieve a globally optimal routing strategy.
[0034] In terms of model input, the DDPG model takes the satellite network state s of each region as input, including link delay, traffic occupancy ratio, and regional satellite topology information. In order to capture the influence of dynamic changes in link weight on routing performance, the link weight w is defined as the weighted combination of delay and traffic occupancy ratio:
[0035] .
[0036] Wherein, is the weight adjustment coefficient, which is used to flexibly switch between delay-sensitive and traffic-balancing scenarios. Link delay refers to the distance of the link divided by the speed of light; traffic occupancy ratio refers to the ratio of current traffic to maximum traffic.
[0037] In terms of network structure design, the DDPG model (initial multi-objective optimization intelligent routing model) is composed of two main networks: actor network and evaluation network. The actor network is responsible for generating the selection of the next hop satellite a based on the input satellite network state s, while the evaluation network guides the actor network to optimize the action selection strategy by evaluating the Q(s, a) value. In order to better capture the nonlinear characteristics in complex network environments, the network contains two hidden layers, each with 256 neurons, and the activation function uses ReLU to improve the model's expression ability.
[0038] Step 104, according to the above abstract region set and the corresponding each satellite node set, the model training of the initial multi-objective optimization intelligent routing model is carried out, and the trained multi-objective optimization intelligent routing model is obtained.
[0039] In some embodiments, the execution subject can determine the network structure of the initial multi-objective optimization intelligent routing model. Wherein, the initial multi-objective optimization intelligent routing model includes an actor network and an evaluation network. In the training of the initial multi-objective optimization intelligent routing model, it is mainly composed of deep deterministic policy gradient (DDPG) and distributed multi-agent training mechanism. First, DDPG is a deep reinforcement learning method suitable for continuous action space, which can efficiently handle dynamic routing problems in low-orbit satellite networks. Then, through the distributed training mechanism, each satellite region shares model parameters to achieve a globally optimal routing strategy.
[0040] In the training process, the routing task is modeled as a Markov decision process (MDP), where the state s tdenotes the delay d of the current satellite link t and the weight ratio of traffic load f t , i.e., , action a t denotes the selection of the next-hop satellite, the reward function r t guides policy optimization, aiming to maximize the cumulative reward , where is the discount factor.
[0041] DDPG utilizes a policy network and a critic network representing the policy and action value functions, respectively, which are jointly optimized by policy gradient and value iteration. For the policy network, the deterministic policy gradient update rule is adopted:
[0042] .
[0043] , where is the experience replay pool, is the action value function, is the policy function, is the critic network parameter, is the policy network parameter. The critic network updates the weights by the Temporal-Difference (TD) error:
[0044] .
[0045] , where is the target value.
[0046] To guide the model to learn the optimal routing policy, the following reward mechanism is designed: when the number of hops exceeds the number of satellites in the region and the target satellite is not found, the reward is negative , where β>0 is the penalty factor; for links with lower weights, the reward increases inversely , thus guiding the model to prefer low-delay and low-load links.
[0047] During training, by continuously injecting different traffic flows, the experience replay pool is continuously updated in the region divided by geographical grid, and the target network parameter is adjusted by soft update mechanism to improve the generalization ability and convergence speed of the model. As the training deepens, the DDPG routing model gradually learns to dynamically balance link delay and load pressure, and finally generates a globally optimal routing path. Compared with traditional routing algorithms, this method effectively reduces the risk of link interruption and improves the overall communication performance and resource utilization of the giant low-orbit satellite optical network.
[0048] In practice, the above execution subject can train the above initial multi-objective optimization intelligent routing model by the following steps:
[0049] The first step, for each latitude, the following training steps are performed:
[0050] 1. Select a target abstract region corresponding to the latitude from the above abstract region set. The local training of the model is carried out in units of abstract regions, and a typical abstract region (target abstract region) is selected as the training group for each latitude, making full use of the network structure and service characteristics in the abstract region to build a training environment. In each training group, the model is independently trained for the satellite node and link information in the target abstract region.
[0051] 2. Initialize the initial multi-objective optimization intelligent routing model and the network state of the target abstract region. The initial multi-objective optimization intelligent routing model and the network environment (network state) in the target abstract region are initialized, including: defining the link weight of the node (combining the weighted ratio of delay and traffic), the orbit characteristics of the satellite, and the injection rules of the service flow.
[0052] 3. Generate random service flows corresponding to the above target abstract region. Random service flows are initiated from random source satellite nodes in the target abstract region, with the goal of efficiently finding a transmission path to the target satellite node.
[0053] 4. According to the above random service flow and the above network state, the initial multi-objective optimization intelligent routing model is trained to obtain a trained latitude optimization intelligent routing model corresponding to the latitude. In the training, the initial multi-objective optimization intelligent routing model predicts the next hop satellite according to the current state, simulates the actual transmission path of the random service flow and records the relevant state transition. To enhance the model's ability to adapt to dynamic network environments, diverse link conditions are added to each service flow routing, such as adjusting link delay or adding random interruption. By evaluating the results of the service flow, the model obtains positive or negative feedback rewards, with positive rewards when the service flow is successfully transmitted and the path weight is low, and negative rewards when the number of hops exceeds the number of satellites in the region or the path selection fails. The initial multi-objective optimization intelligent routing model improves its understanding and adaptation to the dynamic environment in the region by continuously optimizing the association between state transitions and action decisions.
[0054] This local training method ensures that the routing strategy in each abstract region is targeted, while the robustness of the model is improved through random and diverse service flow scenarios. After multiple iterations, the model can converge to an optimal routing strategy with low delay and high efficiency in the training region, laying the foundation for real-time communication in dynamic low-orbit satellite networks.
[0055] The second step is to determine the latitude optimization intelligent routing model trained for each latitude as the multi-objective optimization intelligent routing model.
[0056] Step 105, during the transmission of the service flow, if it is detected that a certain communication link cannot be transmitted or a certain satellite node is unavailable, the routing strategy of the target abstract region is reconstructed to obtain a reconstructed routing path table, and the reconstructed routing path table is input into the multi-objective optimization intelligent routing model to obtain an optimized routing strategy.
[0057] In some embodiments, the execution subject can reconstruct the routing strategy of the target abstract region during the transmission of the service flow if it is detected that a certain communication link cannot be transmitted or a certain satellite node is unavailable, obtain a reconstructed routing path table, and input the reconstructed routing path table into the multi-objective optimization intelligent routing model to obtain an optimized routing strategy.
[0058] As Figure 2 shown, to cope with the problem of link interruption or satellite node failure that may occur in a dynamic network environment, a routing reconstruction algorithm based on region obstacle avoidance is proposed. The algorithm only reconstructs the routing in the abstract region to ensure the local efficiency and low complexity of the algorithm. Specifically, when the service flow detects that a certain link cannot be transmitted or a certain satellite node is unavailable during transmission, the link or node will be marked as an obstacle, and the topology structure in the region will be dynamically updated.
[0059] In practice, the execution subject can reconstruct the routing strategy of the target abstract region by the following steps:
[0060] First, scan the current network state of the target abstract region to filter out the satellite nodes that are working normally as candidate hop points. In the routing reconstruction phase, the current network state is first scanned to filter out the satellite nodes in the region that are still working normally as candidate hop points.
[0061] Second, select candidate nodes that meet the minimum delay and minimum traffic from each candidate hop point. By analyzing the link characteristics of the candidate hop points, nodes that meet the minimum delay and minimum traffic are preferentially selected to construct a new routing path. Next, by gradually linking the selected hop points, a feasible reconstruction path is dynamically generated, enabling the service flow to bypass the obstacle region and continue to transmit to the target node.
[0062] In the third step, the reconstructed path is generated by linking the selected candidate hop-by-hop, and the routing table of the target abstract region is updated to obtain the reconstructed routing path table. To improve the adaptability and efficiency of the algorithm, the quality of the new path is evaluated in real time during the reconstruction process to ensure that it does not significantly increase the transmission delay or cause uneven network load. After the reconstructed path is successfully established, the algorithm updates the routing table in the region and feeds back the relevant path information to the routing model to continuously optimize the strategy. The obstacle-avoiding reconstruction algorithm balances between dynamics and complexity, effectively enhancing the routing method's ability to cope with dynamic failures in a giant low-orbit satellite optical network. For example, each candidate hop (satellite node) can be connected in turn to obtain a new routing path (reconstructed path). Then, the routing table of the target abstract region can be updated according to each satellite node involved in the reconstructed path to obtain the reconstructed routing path table. The routing table can represent the routing relationship between each satellite node in the satellite node set corresponding to the target abstract region (the connection relationship between satellites and satellites).
[0063] Further reference Figure 3 To implement the methods shown in the above figures, the present disclosure provides some embodiments of intelligent routing devices for multi-objective optimization in a low-orbit satellite optical network, which correspond to the method embodiments shown in Figure 1 The intelligent routing devices for multi-objective optimization in a low-orbit satellite optical network can be applied in various electronic devices.
[0064] As Figure 3As shown, the intelligent routing device 300 for multi-objective optimization in a low-orbit satellite optical network of some embodiments includes a division unit 301, an initialization unit 302, a determination unit 303, a training unit 304, and a reconstruction unit 305. The division unit 301 is configured to divide the low-orbit satellite network into a set of abstract regions according to latitude and longitude ranges by a geographic grid division technique, where each abstract region represents a different satellite coverage range and communication demand. The initialization unit 302 is configured to determine a set of satellite nodes corresponding to each abstract region and perform routing strategy initialization processing on the satellite nodes of each abstract region. The determination unit 303 is configured to determine a network structure of an initial multi-objective optimization intelligent routing model, where the initial multi-objective optimization intelligent routing model includes an actor network and an evaluation network. The training unit 304 is configured to perform model training on the initial multi-objective optimization intelligent routing model according to the set of abstract regions and the corresponding sets of satellite nodes to obtain a trained multi-objective optimization intelligent routing model. The reconstruction unit 305 is configured to, during a service flow transmission process, if it is detected that a certain communication link cannot be transmitted or a certain satellite node is unavailable, reconstruct the routing strategy of a target abstract region to obtain a reconstructed routing path table, and input the reconstructed routing path table into the multi-objective optimization intelligent routing model to obtain an optimized routing strategy.
[0065] It can be understood that the units described in the intelligent routing device 300 for multi-objective optimization in a low-orbit satellite optical network correspond to the respective steps in the method described above. Figure 1 The operations, features, and advantages described above for the method also apply to the intelligent routing device 400 for multi-objective optimization in a low-orbit satellite optical network and the units contained therein, and will not be repeated here.
[0066] Reference is made below to Figure 4 which shows a structural schematic diagram of an electronic device (such as a computing device) suitable for use to implement some embodiments of the present disclosure. Figure 4 The electronic device shown is merely an example and should not impose any limitation on the functions and use range of embodiments of the present disclosure. As Figure 4As shown, the computer device includes a processor, a memory and a network interface connected through a system bus, wherein the memory can include a non-volatile storage medium and an internal memory. The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions which, when executed, can cause the processor to perform any one of the intelligent routing methods for multi-objective optimization in a low-orbit satellite optical network. The processor is configured to provide computing and control capabilities to support the operation of the entire computer device. The internal memory provides an environment for the computer program in the non-volatile storage medium to run, and the computer program, when executed by the processor, can cause the processor to perform any one of the intelligent routing methods for multi-objective optimization in a low-orbit satellite optical network. The network interface is configured to perform network communication, such as sending assigned tasks, etc. Those skilled in the art can understand that Figure 4 The structure shown in the figure is only a block diagram of part of the structure related to the present disclosure, and does not constitute a limitation on the computer device to which the present disclosure is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0067] It should be understood that the processor can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.
[0068] In one embodiment, the processor is configured to run a computer program stored in the memory to implement the following steps: dividing the low-orbit satellite network into a set of abstract regions according to latitude and longitude ranges by a geographic grid division technique, wherein each abstract region represents different satellite coverage and communication requirements; determining a set of satellite nodes corresponding to each abstract region and performing routing strategy initialization processing on the satellite nodes of each abstract region; determining a network structure of an initial multi-objective optimization intelligent routing model, wherein the initial multi-objective optimization intelligent routing model includes an actor network and an evaluation network; training the initial multi-objective optimization intelligent routing model according to the set of abstract regions and the corresponding set of satellite nodes to obtain a trained multi-objective optimization intelligent routing model; during the transmission of a service flow, if it is detected that a communication link cannot be transmitted or a satellite node is unavailable, reconstructing the routing strategy of a target abstract region to obtain a reconstructed routing path table, and inputting the reconstructed routing path table into the multi-objective optimization intelligent routing model to obtain an optimized routing strategy.
[0069] The embodiments of the present disclosure further provide a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program includes program instructions. When the program instructions are executed, the method can be applied to the low-orbit satellite optical network multi-objective optimization intelligent routing method according to the embodiments of the present disclosure.
[0070] The computer readable storage medium can be an internal storage unit of the computer device, for example, a hard disk or a memory of the computer device. The computer readable storage medium can also be an external storage device of the computer device, for example, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc.
[0071] It should be noted that, in this document, the terms "comprising", "including", or any other variant thereof are intended to cover non-exclusive inclusions, so that a process, method, article, or system including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or includes elements inherent to such a process, method, article, or system. Without more limitations, the element defined by the statement "including a" does not exclude the presence of another identical element in the process, method, article, or system including the element.
[0072] The above description is merely exemplary of some preferred embodiments of the present disclosure and of the principles thereof. It is to be understood that the present disclosure is not limited to the specific technical features described above, and that the scope of the present disclosure is not limited to the technical solutions formed by the specific combinations of the above technical features, but also covers other technical solutions formed by the combinations of the above technical features or equivalent features thereof without departing from the above inventive concept. For example, the above technical features are replaced with the technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to) to form technical solutions.
Claims
1. A smart routing method for multi-objective optimization applied in low-Earth orbit satellite optical networks, characterized in that, include: By using geographic grid partitioning technology, the low-Earth orbit satellite network is divided into an abstract set of regions according to latitude and longitude. Each abstract region represents a different satellite coverage area and communication requirements. Determine the set of satellite nodes corresponding to each abstract region, and perform routing policy initialization processing on the satellite nodes of each abstract region; The network structure of the initial multi-objective optimization intelligent routing model is determined, wherein the initial multi-objective optimization intelligent routing model includes: an actor network and an evaluation network; Based on the abstract region set and the corresponding satellite node set, the initial multi-objective optimization intelligent routing model is trained to obtain the trained multi-objective optimization intelligent routing model. During the transmission of service flows, if it is detected that a certain communication link cannot transmit or a certain satellite node is unavailable, the routing strategy of the target abstract area is reconstructed to obtain a reconstructed routing path table, and the reconstructed routing path table is input into the multi-objective optimization intelligent routing model to obtain an optimized routing strategy.
2. The method according to claim 1, characterized in that, The process of determining the set of satellite nodes corresponding to each abstract region and initializing the routing policy for the satellite nodes in each abstract region includes: The time is divided into time slots in seconds, and a path change rate is introduced in each time slot; The delay characteristics corresponding to the satellite node set are divided into communication delay and inter-satellite link setup delay; The link information corresponding to the satellite node set is sent to the corresponding controller so that the controller can generate the corresponding initial routing policy; In response to receiving the initial routing policy sent by the controller, the initial routing policy is synchronized to each satellite node in the satellite node set.
3. The method according to claim 2, characterized in that, The step of training the initial multi-objective optimization intelligent routing model based on the abstract region set and the corresponding satellite node sets to obtain the trained multi-objective optimization intelligent routing model includes: For each dimension, perform the following training steps: Select a target abstract region corresponding to the latitude from the set of abstract regions; Initialize the initial multi-objective optimization intelligent routing model and the network state of the target abstract region; Generate random service flows corresponding to the target abstract region; Based on the random service flow and the network state, the initial multi-objective optimization intelligent routing model is trained to obtain the trained dimension-optimized intelligent routing model corresponding to the dimension. The dimensional optimization intelligent routing model that has been trained in each dimension is identified as the multi-objective optimization intelligent routing model.
4. The method according to claim 3, characterized in that, The process of reconstructing the routing strategy for the target abstract region to obtain a reconstructed routing path table includes: The current network status of the target abstract region is scanned to filter out normally functioning satellite nodes as candidate hop points; Select candidate nodes that meet the requirements of lowest latency and lowest traffic from among the candidate hops; By progressively linking the selected candidate hops, a reconstructed path is generated, and the routing table of the target abstract region is updated to obtain the reconstructed routing path table.
5. A smart routing device for multi-objective optimization in low-Earth orbit satellite optical networks, characterized in that, include: The partitioning unit is configured to divide the low-Earth orbit satellite network into an abstract set of regions according to latitude and longitude ranges using geographic grid partitioning technology. Each abstract region represents a different satellite coverage area and communication requirements. The initialization unit is configured to determine the set of satellite nodes corresponding to each abstract region and perform routing policy initialization processing on the satellite nodes of each abstract region; The determining unit is configured to determine the network structure of an initial multi-objective optimization intelligent routing model, wherein the initial multi-objective optimization intelligent routing model includes: an actor network and an evaluation network; The training unit is configured to train the initial multi-objective optimization intelligent routing model based on the abstract region set and the corresponding satellite node set, so as to obtain the trained multi-objective optimization intelligent routing model. The reconfiguration unit is configured to reconfigure the routing strategy of the target abstract area if a communication link fails to transmit or a satellite node is unavailable during the service flow transmission process, thereby obtaining a reconfigured routing path table, and inputting the reconfigured routing path table into the multi-objective optimized intelligent routing model to obtain an optimized routing strategy.
6. An electronic device, characterized in that, include: One or more processors; A storage device on which one or more programs are stored; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1 to 4.
7. A computer-readable medium, characterized in that, It stores a computer program thereon, wherein the computer program, when executed by a processor, implements the method as described in any one of claims 1 to 4.