Spatial three-dimensional satellite routing method, device and system

Abstract

The invention relates to the technical field of satellite routing, and discloses a spatial three-dimensional satellite routing method, device and system, and the method comprises the steps: constructing a spatial satellite network based on LEO, MEO and GEO three-layer satellite nodes; performing orbit mechanical prediction calculation based on a space satellite network to obtain a link state prediction matrix; inputting the link state prediction matrix into a spatial position coding processor for processing to obtain a spatial time-varying routing graph; generating a routing strategy of the satellite nodes based on the space time-varying routing graph; according to the method, global optimization of end-to-end time delay and energy consumption is achieved under the condition that the network capacity and the link availability are met, and then the cross-track data transmission efficiency is improved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of satellite routing, in particular to a space stereoscopic satellite routing method, device and system. BACKGROUND

[0002] With the rapid development of global satellite communication networks, multi-level satellite constellation deployment is increasingly widespread, and traditional two-dimensional plane routing methods have been unable to meet the needs of modern satellite networks. Traditional two-dimensional routing techniques mainly consider satellite connections within the same orbital plane, ignoring the three-dimensional spatial distribution characteristics of satellite networks, resulting in low efficiency when dealing with cross-orbit communications and failing to fully utilize spatial resources. Especially in complex constellation environments with LEO, MEO and GEO three-layer mixed deployment, satellite nodes are distributed at different orbital altitudes and spatial positions, forming a highly stereoscopic network topology.

[0003] Due to the high-speed movement of satellites and the dynamic changes in their orbits, satellite networks face challenges such as intermittent link visibility and frequent topology reconstruction. Satellites are located at different orbital altitudes, inclinations and phases, resulting in frequent establishment and interruption of inter-satellite links, and the network topology presents high dynamics, making it difficult for traditional routing methods to adapt to such rapid changes. In addition, traditional methods fail to effectively identify and utilize key spatial resources such as orbital intersection points, resulting in cross-orbit communication bottlenecks, link resource waste and path redundancy. SUMMARY

[0004] The application provides a space stereoscopic satellite routing method, device and system, which realizes global optimization of end-to-end delay and energy consumption under the conditions of network capacity and link availability, thereby improving the efficiency of cross-orbit data transmission.

[0005] The first aspect of the application provides a space stereoscopic satellite routing method, which comprises: constructing a space satellite network based on LEO, MEO and GEO three-layer satellite nodes; performing orbital mechanics prediction calculation based on the space satellite network to obtain a link state prediction matrix; inputting the link state prediction matrix into a space position encoding processor for processing to obtain a space time-varying routing graph; generating a routing strategy for satellite nodes based on the space time-varying routing graph; solving the mixed integer programming combined with the routing strategy to obtain an optimal routing decision set.

[0006] The second aspect of the application provides a space stereoscopic satellite routing device, which comprises: A modeling module is configured to construct a space satellite network based on LEO, MEO and GEO three-layer satellite nodes. A prediction module is configured to perform orbit mechanics prediction calculation based on the space satellite network to obtain a link state prediction matrix. A processing module is configured to input the link state prediction matrix into a space position coding processor for processing to obtain a space time-varying routing graph. A calculation module is configured to generate a routing strategy of the satellite node based on the space time-varying routing graph. An optimization module is configured to combine the routing strategy to perform mixed integer programming solution to obtain an optimal routing decision set.

[0007] The third aspect of the application provides an electronic system, comprising a memory and at least one processor, the memory storing instructions; the at least one processor invokes the instructions in the memory to enable the electronic system to perform the space stereoscopic satellite routing method described above.

[0008] Compared with the prior art, the application has the following beneficial effects: by performing three-dimensional space modeling on LEO, MEO and GEO three-layer satellite nodes, fully considering the orbital height difference, constructing a stereoscopic network topology, compared with the traditional two-dimensional plane routing method, more accurately reflecting the actual space distribution characteristics of the satellite network. The orbit mechanics prediction algorithm is used to realize accurate prediction of the future position of the satellite and the link state, construct a link state prediction matrix, make the routing decision have foresight, and effectively cope with the challenge brought by the frequent change of network topology. The space position coding mechanism and the graph attention network are used to extract the stereoscopic topological features of the satellite node, realize the construction of the space time-varying routing graph, and provide comprehensive space-time information for the routing decision. The cross-orbit priority routing mechanism is introduced, the link metric is optimized by comprehensively considering the space priority, the cross-orbit link passing through the orbit intersection area obtains higher priority, and the cross-orbit data transmission efficiency is significantly improved. The mixed integer programming model is applied to globally optimize the routing decision, and under the condition of meeting the network capacity and link availability, the global optimization of end-to-end delay and energy consumption is realized. BRIEF DESCRIPTION OF DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0010] The structure, proportion, size and the like shown in the drawings of the specification are only used to cooperate with the content disclosed in the specification, to be understood and read by those skilled in the art, and do not have technical significance to limit the conditions that the application can be implemented. Any modification of the structure, change of the proportion relationship or adjustment of the size, without affecting the effect that the application can produce and the purpose that the application can achieve, should still fall within the scope covered by the disclosed technology.

[0011] Figure 1 is a flowchart of a spatial stereoscopic satellite routing method provided by an embodiment of the application; Figure 2 is a structural schematic block diagram of a spatial stereoscopic satellite routing device provided by an embodiment of the application; Figure 3 is a structural schematic block diagram of an electronic system provided by an embodiment of the application. DETAILED DESCRIPTION

[0012] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0013] The flowchart shown in the drawings is only an example and does not necessarily include all the contents and operations / steps, nor does it have to be executed in the order described. For example, some operations / steps can be decomposed, combined or partially combined, so the actual execution order may be changed according to the actual situation.

[0014] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in the specification and the appended claims of the application, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0015] It should be further understood that the term "and / or" used in the specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations. Please refer to Figure 1 One embodiment of the spatial stereoscopic satellite routing method in the embodiments of the application includes: Step S100, constructing a spatial satellite network based on LEO, MEO and GEO three-layer satellite nodes; It can be understood that the execution subject of the present application can be a spatial stereoscopic satellite routing device, and can also be a terminal or a server, and the specific embodiments are not limited herein. The server is taken as an example for description of the embodiments of the present application.

[0016] Specifically, the key orbital root number parameters of the three-layer satellite nodes are systematically collected, including the basic data such as orbital altitude, inclination, ascending node right ascension and perigee amplitude. These parameters of each satellite reflect its spatial running state and position change characteristics in the orbit, so a unified data interface is established, and the real-time or predicted orbital root number parameters of all satellites are collected and summarized to form a satellite orbital parameter set through the ground measurement and control system, satellite autonomous ephemeris broadcast or constellation configuration management platform. The satellite orbital parameter set is input into the Keplerian orbit equation set, and according to the classical orbit dynamics theory, the spatial orbit information of each satellite at the current or future specified time is accurately mapped into the earth-centered three-dimensional rectangular coordinate system by using the six-root number rule, and the (X, Y, Z) coordinate points of the LEO, MEO and GEO three-layer satellites in the three-dimensional space are obtained. Based on the three-dimensional space coordinates, the Euclidean distance between any two satellites is calculated to obtain a distance matrix reflecting the spatial relationship between the satellites in the whole network. Each element of the distance matrix corresponds to the actual spatial distance between a pair of satellite nodes. The matrix is compared with the link visibility distance threshold set by the system item by item, and the inter-satellite link set satisfying the physical connectable condition is screened out, and the special links across different orbital layers are distinguished to form the cross-orbit link subset. For the preliminary screened inter-satellite link set and cross-orbit link subset, the time delay characteristics, available bandwidth and energy consumption and other multi-dimensional indexes are comprehensively considered, and each link is comprehensively evaluated to form a link metric matrix. On the basis of the above, the spatial distribution of the three-layer satellite nodes is further analyzed, and the intersection area between the orbits and the satellite nodes near the intersection area are identified. The orbit plane equation and the geometric relationship of three-dimensional coordinates are used to determine the intersection points or intersection areas of the orbit planes, and the areas with cross-orbit communication potential and high routing value are specially marked, and these nodes or areas are given higher weights or priorities in the modeling of the network topology structure. The spatial satellite network is obtained.

[0017] Step S200, performing orbit mechanics prediction calculation based on the spatial satellite network to obtain a link state prediction matrix; Specifically, the perturbed orbit theory is applied to each satellite node for orbit mechanics modeling. In this process, a series of space perturbation factors are considered, such as the earth's non-spherical gravitational field (such as J2 term), solar radiation pressure, atmospheric drag (especially significant for LEO satellites), and third body gravity (moon and sun, etc.). Through the modeling of these physical parameters, a high-precision satellite orbit prediction model is established. The spatial position and velocity of all satellite nodes at the current time are input as initial conditions into the above orbit prediction model, and a high-order numerical integration method (such as the fourth-order Runge-Kutta method) is used to perform time-domain integration calculation on the satellite's dynamic equation set, thereby calculating the orbit evolution trajectory and corresponding spatial coordinates of each satellite within a future set time window, forming the orbit prediction data. On this basis, the satellite node orbit data obtained by prediction is globally traversed and combined two by two, and the spatial relative relationship of any two satellites at any time is calculated, including the Euclidean space distance and relative velocity direction, to obtain the position relationship data set reflecting the satellite dynamic layout. Based on the time-sequenced spatial relationship data, combined with the pre-set link visibility conditions, such as the maximum visible distance of the link, the earth blocking, the satellite self-blocking, and the antenna maximum pointing angle constraints, the state of each pair of satellite nodes is judged on each time slice. If the spatial distance between the satellites is less than the maximum visible distance of the link and meets the antenna pointing restrictions, it is judged that the link is available at that time, otherwise it is unavailable, thereby generating a link availability time sequence for each link, and clearly identifying the establishment and interruption time of the link at future times. For each available link in the link availability time sequence, the dynamic link metric value is calculated based on the relative spatial state and kinematic information of the satellites. These metrics include link propagation delay (calculated according to spatial distance and light speed), Doppler shift (estimated based on relative velocity and carrier frequency), and link stability index (such as link available remaining time ratio), which can reflect the transmission efficiency, signal quality and continuous availability of the link at a specific time. The dynamic metric value of each link at each time is fused with the link availability information in the space-time dimension to form a set of multi-dimensional and dynamically changing link state prediction matrices.

[0018] Step S300, input the link state prediction matrix into the spatial position encoding processor for processing to obtain a spatial time-varying routing graph; It should be noted that the time discretization unit of the space position coding processor processes the link state prediction matrix to map the link state data in the continuous time domain to a set of discrete time series. By uniformly dividing the overall prediction time window into several time slices, for example, the total time interval is divided into K equal time periods, each time point reflects the evolution snapshot of the network structure. Based on the discrete time series, the topology construction unit of the space position coding processor processes the link availability and metric information at each time point, dynamically generates K network topology snapshots according to the link availability judgment matrix and the link metric matrix, and each snapshot describes the spatial structure, node state and link attribute of the whole network at this time. In order to depict the continuity and mutability of the network evolution, the adjacent network topology snapshots are input into the transition probability calculation unit of the space position coding processor, the link state changes between two snapshots are compared and analyzed, and the transition probability matrix of each link and node state from tk to tk+1 is obtained by combining statistical modeling and time series analysis. In the spatial dimension modeling, in order to make the node spatial attribute better serve the subsequent machine learning and graph network algorithm, the position sensitive hash function is applied to the three-dimensional coordinates of each satellite to generate a set of high-dimensional embedding vectors. The embedding vectors of all satellite nodes are input into the graph attention network integrated in the space position coding processor, and under the graph neural network structure, the multi-head attention mechanism is used to weight and aggregate the information between nodes. By designing distance attenuation, spatial position sensitive factor and other mechanisms, the graph attention network can automatically focus on the key areas and important nodes in the network structure, effectively extract the spatio-temporal dependence features of the nodes, and output the hidden state representation reflecting the nodes in the complex space-time relationship. Based on the above obtained discrete network topology snapshot, transition probability matrix depicting time evolution and node hidden state integrating space-time-structure features, these multi-source information are deeply integrated to complete the effective fusion of four-dimensional information, and a space-time-varying routing graph is obtained.

[0019] Step S400, generating a routing strategy of the satellite node based on the space-time-varying routing graph; Specifically, according to the orbit height parameter carried in the space-time-varying routing graph, all satellite nodes are subjected to hierarchical threshold processing, i.e. according to the height data of the nodes, they are accurately divided into three sets of satellite nodes of LEO (low earth orbit), MEO (medium earth orbit) and GEO (geosynchronous earth orbit). Combined with the actual communication attributes of the three types of satellite nodes, orbit level weight coefficients conforming to the network service requirements are respectively assigned to them, wherein the weight coefficient of LEO satellite should be higher than that of MEO and GEO satellites because LEO satellite has low delay, fast link switching but limited coverage; MEO satellite as an intermediate layer has moderate coverage and delay performance, and is assigned a medium weight; GEO satellite has the widest coverage but larger delay and link congestion, so the weight is the lowest. Through scientific allocation, the network routing mechanism can adapt to the advantages and disadvantages of resources of different orbit layers. At the same time, the spatial position weight factor of the orbit intersection area is introduced, and the distance between each satellite node and the orbit intersection area is calculated in real time according to the spatial distance. The distance reflects the proximity of the node to the high routing value area, and provides measurement support for spatial coordination of path selection. The distance factor is nonlinearly transformed by exponential decay or other spatial weight functions, and nodes closer to the distance intersection area are given higher weight to guide the routing strategy to preferentially select link resources that can quickly and efficiently realize cross-orbit forwarding. The spatial position weight factor and the aforementioned orbit level weight coefficient are mathematically combined, such as weighted product or other aggregation methods, to obtain a comprehensive spatial priority that comprehensively reflects the advantages of orbit levels and spatial positions. Based on the comprehensive spatial priority, the original link metric function is prioritized, i.e. the comprehensive spatial priority is used to normalize and correct the multi-dimensional indexes of the original link such as transmission delay, available bandwidth and energy consumption, so that the link closer to the orbit intersection area and in the preferred orbit layer obtains a higher comprehensive score, thereby being preferentially selected in routing calculation. The corrected link metric function can dynamically adapt to the changes of satellite network topology structure and the needs of actual business load, effectively improving the transmission efficiency and robustness of data in the network. According to the corrected link metric function, a group of primary and backup data transmission paths are calculated through multi-path calculation and path diversity strategy, forming an efficient routing strategy oriented to cross-orbit priority.

[0020] Step S500, combined with the routing strategy, a mixed integer programming solution is solved to obtain an optimal routing decision set.

[0021] Specifically, the decision variables are generated based on the modified link metric function in the routing strategy, including selection variables representing whether each link is included in the routing path at a specific time, and data flow allocation variables representing the actual data flow carried by each link. To this end, the flow allocation unit is dynamically configured in combination with the current and predicted traffic demand of the satellite network to form the routing decision variable set and the flow allocation variable set. According to the actual operation requirements and service objectives of the satellite network, the two types of variables and the end-to-end delay energy efficiency weighting coefficient are jointly included in the global optimization objective function construction process, and the transmission efficiency, energy consumption and link time-varying characteristics of the network as a whole are comprehensively considered. The optimization objective is designed to minimize the linear or nonlinear combination of the weighted end-to-end delay and energy consumption of the entire network, and the network performance indicators are balanced flexibly through adjustable coefficients to ensure that the routing strategy meets the diversified task scenarios and actual engineering constraints. While constructing the global optimization objective function, a set of constraint conditions is systematically established according to the topology and service characteristics of the satellite network. These constraints include flow conservation, which requires that the flow in and out of all satellite nodes except the source and destination nodes always remains balanced; link capacity constraint, which ensures that the actual allocated flow of any link does not exceed its maximum bandwidth; link availability constraint, which ensures that only physically reachable and normally functioning links can be selected for routing decisions; cross-orbit path existence constraint, which requires at least one cross-orbit link to be included in the final routing scheme to fully leverage the synergistic value of the three-dimensional network space structure; and path stability constraint, which is used to limit the change amplitude of the routing decision within a continuous time period to avoid additional signaling overhead and service jitter caused by excessive frequent path switching. An efficient mixed integer programming solver is used to input the global optimization objective function and the constraint condition set, and through numerical optimization algorithms such as branch and bound, Lagrange relaxation, and rolling horizon decomposition, an initial routing decision scheme covering the entire network is obtained. To improve the practical applicability and network adaptability of the results, a link pre-screening mechanism is implemented on the initial routing decision scheme to exclude or down-weight links with extremely low link metrics or unstable states, reducing the problem size and subsequent routing fluctuations. At the same time, time domain smoothing processing technology is used to connect and transition the routing decisions of adjacent time segments, reducing the impact of frequent path changes on network operation. After the above process, a set of globally optimal routing decisions is finally obtained.

[0022] In the overall modeling stage of this embodiment, for the global optimization objective function and the complex constraints involved therein such as flow conservation, link capacity, availability, and cross-orbit, the mixed integer programming solver uses Lagrange relaxation processing for the flow conservation constraint which is difficult to directly decompose. The flow conservation condition is embedded in the optimization objective in the form of a multiplier λ to construct a Lagrange function, thereby achieving softening of the constraint and distributed problem decomposition. For the processed Lagrange function, the original global problem is decomposed into several sub-problems that can be independently solved using the structural characteristics of the problem and the natural distribution of network nodes and links. Each sub-problem corresponds to a part of the nodes or local topology, and an initial Lagrange multiplier is assigned to each sub-problem to form an initialized sub-problem set. The built-in branch and bound algorithm of the mixed integer programming solver is used to enumerate, prune, and locally optimize each sub-problem to explore its optimal or near-optimal feasible solution space. The branch and bound process can effectively exclude a large number of decision paths that cannot achieve the global optimum, greatly reducing the computational complexity and improving the convergence speed. After each local solution is completed, the value of the Lagrange multiplier is dynamically adjusted and updated according to the residual between the local optimal solution and the relaxation variable, thereby promoting the convergence of the global Lagrange dual function to the optimum. Through multiple rounds of iterative cycles, the information between sub-problems is gradually coordinated, and the preliminary global solution of the entire network is also continuously optimized until the convergence criteria are met or the maximum number of iterations is reached. After obtaining the global preliminary solution, the feasibility of the global preliminary solution is tested and the boundaries are processed according to the link load balancing standard. The actual load of all links is carefully checked to ensure that no link has overflow, bottleneck, or extreme load imbalance. Individual solutions that exceed the boundaries are modified or redistributed, and the routing selection is adjusted as necessary to meet all actual network constraint conditions. Through the above steps, the initial routing decision scheme is finally obtained.

[0023] In the embodiments of the present application, three-dimensional space modeling is performed on LEO, MEO, and GEO satellite nodes, and the differences in orbital altitudes are fully considered to construct a three-dimensional network topology structure. Compared with traditional two-dimensional plane routing methods, the actual spatial distribution characteristics of the satellite network are more accurately reflected. The orbital mechanics prediction algorithm is used to accurately predict the future positions of the satellites and the state of the links, and a link state prediction matrix is constructed, so that the routing decision has foresight and effectively copes with the challenges brought by frequent changes in network topology. The spatial position coding mechanism and the graph attention network are used to extract the three-dimensional topological features of the satellite nodes, and a spatial time-varying routing graph is constructed, providing comprehensive spatio-temporal information for routing decision. The cross-orbit priority routing mechanism is introduced, and the link metric is optimized by comprehensively considering the spatial priority, so that the cross-orbit links passing through the orbital intersection area have higher priority, significantly improving the cross-orbit data transmission efficiency. The mixed integer programming model is used for global optimization of the routing decision, and the global optimization of the end-to-end delay and energy consumption is realized under the condition of meeting the network capacity and link availability.

[0024] In one specific embodiment, the process of performing step S100 may specifically include the following steps: Orbital elements parameters such as orbital altitude, inclination, right ascension of ascending node, and argument of perigee of LEO, MEO, and GEO satellite nodes are collected to obtain a set of satellite orbital parameters; By inputting the set of satellite orbit parameters into the Kepler orbit equation and performing coordinate transformation, the spatial position coordinates of the three-layer satellite nodes (LEO, MEO, and GEO) in the three-dimensional rectangular coordinate system are obtained. The distance between any two satellites is calculated based on their spatial coordinates to obtain the inter-satellite distance matrix. A threshold judgment is then applied to the inter-satellite distance matrix to obtain the set of inter-satellite links and the subset of inter-orbit links. A comprehensive evaluation of the inter-satellite link set and the inter-orbit link subset is conducted based on link transmission delay, available bandwidth, and energy consumption to obtain a link metric matrix. Spatial location analysis of LEO, MEO, and GEO satellite nodes is performed to obtain the space satellite network.

[0025] Specifically, the satellite orbital root number parameters are collected. The orbital state of each satellite is described by a set of classical orbital root numbers, including orbital height, orbital inclination, ascending node right ascension, and perigee amplitude angle parameters. Among them, the orbital height reflects the average distance of the satellite from the earth's surface, the inclination describes the angle between the satellite orbit plane and the earth's equatorial plane, the ascending node right ascension describes the right ascension coordinate of the point where the orbit plane crosses the earth's equator from south to north, and the perigee amplitude angle describes the angle position of the satellite in the orbit plane from the ascending node to the perigee. Through regular communication with ground control stations, satellite autonomous positioning, inter-satellite transmission and other ways, complete orbital root number parameters of LEO, MEO and GEO satellites are efficiently and real-time collected, and these parameters are time-sequenced and structured managed to form a satellite orbital parameter set. The satellite orbital parameter set is input into the Kepler orbit equation for coordinate transformation, mapping the theoretical orbital parameters to the actual three-dimensional position of the satellite in the physical space. The Kepler equation set is a basic dynamic model for describing the motion of satellites around the earth, which calculates the instantaneous position of the satellite in the earth-centered three-dimensional rectangular coordinate system (i.e. X, Y, Z three-axis space) based on the six root number parameters of the satellite and the current time. Coordinate transformation can eliminate the abstract differences between satellite nodes in different orbital layers and express all satellites in a unified spatial coordinate system. After coordinate transformation, the spatial positions of all LEO, MEO and GEO satellites are mapped to a three-dimensional rectangular coordinate system, reflecting the spatial distribution, orbital level and potential visibility of the satellite networking. Based on the three-dimensional spatial position coordinates, the Euclidean distance between any two satellite nodes is calculated to obtain an inter-satellite distance matrix covering the entire network. Each element of the distance matrix represents the actual spatial straight-line distance between a pair of satellites. On this basis, by setting a maximum visible distance threshold for the link, it is determined whether each pair of satellites has the physical conditions to establish an inter-satellite link at a specific time. When the distance between the satellites is less than or equal to the threshold, it is determined that a direct communication link can be established, otherwise it is considered physically unreachable. Using the orbital layer attributes of the satellite nodes, the inter-satellite link set is divided into intra-orbital layer links and cross-orbital layer links. All satellite pairs that meet the threshold are logically filtered to obtain the inter-satellite link set and the cross-orbital link subset. The inter-satellite link set and the cross-orbital link subset are comprehensively evaluated according to the link transmission delay, available bandwidth and energy consumption. The link transmission delay is mainly obtained by dividing the distance between the two satellites by the speed of light, representing the lower limit of the physical limit of data propagation; the available bandwidth is limited by multiple factors such as the technical parameters of the satellite transmitting and receiving equipment, the current business load, the signal frequency and the spatial interference, and is the key criterion for the concurrent throughput capacity of the link; the energy consumption is related to the communication distance, the transmission power, the link maintenance mechanism and the satellite energy distribution strategy. After integrating these physical, business and energy consumption characteristics, a comprehensive metric score is output for each link by applying weighting or normalization to form a link metric matrix. The spatial positions of the three-layer satellite nodes are analyzed.The algorithm effectively identifies satellite nodes near the orbital intersection area by calculating the intersection or intersection area of different orbital planes and combining the real-time coordinates of satellites in three-dimensional space, and gives these key areas higher communication weights and routing priorities. The orbital intersection determination helps to determine the "high" nodes in the network (i.e. satellites with potential multi-orbit connection), and provides a strategic foothold for dynamic routing, link switching and traffic offloading, so that the network can fully utilize the three-dimensional connectivity in space and continuously maintain high throughput, high reliability and low delay in the face of dynamic topology changes and traffic burst scenarios.

[0026] In a specific embodiment, the process of performing step S200 can specifically include the following steps: Based on the perturbed orbit theory, the orbit mechanics model of each satellite node is modeled to obtain a satellite orbit prediction model; The position and velocity information of the satellite node at the current time is input into the satellite orbit prediction model for integral calculation to obtain orbit prediction data; Based on the orbit prediction data, the relative relationship between any two satellite nodes is calculated to obtain a position relationship data set; According to the position relationship data set and the preset link visibility condition, the state of the inter-satellite link is judged to obtain a link availability time sequence; The dynamic metric of each available link in the link availability time sequence is calculated to obtain a dynamic link metric value; The dynamic link metric value and the link availability information are spatio-temporally fused to obtain a link state prediction matrix.

[0027] Specifically, considering a series of perturbation factors including the non-spherical gravity field of the Earth (such as the J2 term), atmospheric drag, solar radiation pressure, the third body gravity of the Moon and the Sun, and combining with the six elements of satellite orbit description, the high-precision orbit dynamics differential equation set of multiple sources of physical driving is established for LEO, MEO, and GEO satellites respectively. By introducing high-order perturbation terms and the influence of external force field, a satellite orbit prediction model is formed. The spatial position and velocity vector of the satellite node at the current time are taken as the initial conditions, which are input into the above orbit prediction model, and the satellite dynamics equation is time advanced by numerical integration method. The high-precision integration algorithm includes the fourth-order Runge-Kutta method, the Adams-Bashforth method or the adaptive step size integration method, which can control the short-time or long-time prediction error within the range of meters or even smaller under the premise of ensuring the calculation speed. Through continuous integration calculation of the orbit of each satellite, the spatial position, velocity and other key orbit situation data of each satellite node at different times within the specified prediction window in the future are obtained. Based on the orbit prediction data, the relative relationship between any two satellite nodes is calculated, including the calculation of the Euclidean distance, the extraction of the relative velocity vector and its projection processing in the three-dimensional coordinate system, forming the position relationship data set, and identifying the dynamic spatial distance, relative velocity, relative orientation and other attributes between any two satellites in the entire prediction time domain. Based on the position relationship data set, combined with the set link visibility criteria, such as the maximum visible distance of the link, the antenna beam coverage range, the earth or satellite body shielding constraint, etc., the state of the inter-satellite link at each discrete prediction time is judged. When the spatial distance between two satellites is less than the maximum visible distance of the link and is not affected by the earth or the body shielding, the link is determined to be available, otherwise it is not available. Through traversal and judgment, a time series of link availability is generated for each pair of satellite nodes, depicting the establishment and interruption time of the link in the prediction time domain. The multi-dimensional dynamic measurement calculation is carried out for each available link in the link availability time series. The link propagation delay is directly obtained by dividing the spatial distance by the speed of light, which quantifies the physical lower bound delay of the data packet from the source node to the target node. The Doppler shift is calculated by the relative velocity vector and the carrier frequency, which can reflect the signal frequency offset and modulation compensation demand in the high dynamic environment, ensuring the signal integrity of the link transmission and the robustness of the link layer protocol. The link stability index is normalized modeling based on the remaining available time length, historical fluctuation characteristics and business load dynamics. The higher the index, the stronger the availability and business continuity of the link in the future time domain. Through the above dynamic measurement of each time slice and each link, the key indicators such as link propagation delay, Doppler shift and link stability index are output in real time, forming the multi-dimensional dynamic link measurement value reflecting the link quality. Based on the above link dynamic measurement and link availability data, the deep fusion processing of space-time information is carried out. By unifying the dynamic link measurement value and the link availability information in the multi-dimensional time slice and the spatial dimension of the whole network node, a link state prediction matrix is constructed.Each dimension of the matrix corresponds to the source node, the target node, the prediction time slice and the metric index respectively, and can systematically and finely describe the spatio-temporal link availability and performance evolution characteristics of the entire satellite network within the future prediction window. Each element in the four-dimensional matrix contains key attributes such as availability determination, propagation delay, Doppler shift and stability of the link at a specific time.

[0028] In a specific embodiment, the process of performing step S300 can specifically include the following steps: The link state prediction matrix is input into a time discretization unit of the spatial position encoding processor for processing to obtain a discrete time sequence; Based on the discrete time sequence, a topology construction unit of the spatial position encoding processor is used to process the link state at each time point to obtain K network topology snapshots; Adjacent network topology snapshots are input into a transition probability calculation unit of the spatial position encoding processor for processing to obtain a transition probability matrix; The three-dimensional coordinates of the satellite nodes are encoded using a location-sensitive hashing function to obtain an embedding vector; The embedding vector is input into a graph attention network in the spatial position encoding processor for multi-head attention calculation to obtain hidden state features; Four-dimensional information integration is performed based on the network topology snapshots, the transition probability matrix and the hidden state features to obtain a spatial time-varying routing graph.

[0029] Specifically, the link state prediction matrix is input into the time discretization unit of the spatial position encoding processor. By dividing the continuous prediction time domain into several discrete time slices, the dynamic network is snapshoted in fine granularity. The prediction window and time step are preset, and the total time domain is divided into K equally spaced time points. Each time slice corresponds to a "frozen section" of the global network state, enabling the satellite network topology, link availability, and performance metrics to be clearly described in discrete time domain, improving the efficiency of time series modeling and analysis. Based on the discrete time series, the topology construction unit built-in the spatial position encoding processor deeply analyzes the link state data at each time point. According to the link availability, bandwidth, latency, and energy consumption of each time point, the processor comprehensively determines which links can be activated and which nodes are in active, isolated, or intersection state. Through matrix traversal and logical filtering of these time series data, the processor dynamically generates K network topology snapshots, each reflecting the connection structure, spatial distribution, and functional hierarchy of the network at that time. The adjacent network topology snapshots are input into the transition probability calculation unit of the spatial position encoding processor. This unit uses statistical modeling and probability inference algorithms to compare and analyze the changes in node connection state, link availability, and network distribution between each pair of snapshots, and then calculates the probability of link and node state retention or switching from time tk to tk+1, obtaining the transition probability matrix and revealing the dynamic trend and mutation probability of satellite network topology evolution. Meanwhile, in three-dimensional space, the geometric distribution of satellite nodes determines the spatial structure and routing potential of the network. Therefore, the three-dimensional coordinates of each satellite node are encoded using a location-sensitive hash function to generate a spatial embedding vector. This hash function ensures that nodes with similar spatial distances have highly correlated encoding results, allowing the physical adjacency of the network to be converted into geometric similarity in vector space. The embedding vectors of all satellite nodes are input into the graph attention network module integrated in the spatial position encoding processor, which leverages the automatic modeling capabilities of deep learning models for complex spatial relationships. The graph attention network uses a multi-head attention mechanism to weight and aggregate information for each node and its neighbor nodes, and uses spatial distance decay factors and location-sensitive weights to strengthen the correlation effects of nodes with similar spatial proximity and close connections. This mechanism can automatically identify nodes with pivotal roles or cross-layer value in the network and effectively track the changing trends of key node states between different time slices. After multiple rounds of information fusion and nonlinear activation in attention layers, the network outputs the hidden state representation of each satellite node in the spatiotemporal dynamic context. The hidden state integrates the node's own properties, spatial structure features, and surrounding environmental dynamics. Based on the previously generated network topology snapshots, transition probability matrices, and node hidden state representations, a deep integration of four-dimensional information is implemented. The spatial position encoding processor uses information fusion algorithms to unify static spatial structures, dynamic evolution probabilities, and node behavior characteristics into a multi-dimensional tensor structure, achieving spatiotemporal joint optimization of routing decisions.For each network snapshot at each moment, the node hidden state and link attribute are weighted to generate the priority score of path selection, and the transition probability of topology change is combined to dynamically adjust the route stability and connectivity at different moments. Through integration methods such as sliding window, convolution aggregation and multi-scale pooling, the processor extracts global dynamic features from the evolution of network structure and state in K time slices, and fuses local node behavior, spatial location advantage and link time trend, and finally obtains a spatial time-varying routing graph.

[0030] In a specific embodiment, the process of performing step S400 can specifically include the following steps: According to the orbital height parameter of the satellite node in the spatial time-varying routing graph, the satellite node is threshold layered to obtain a LEO satellite node set, a MEO satellite node set and a GEO satellite node set; Based on the coverage range and transmission delay characteristics of different orbital layers, the LEO satellite node set, the MEO satellite node set and the GEO satellite node set are weighted to obtain an orbital level weight coefficient; According to the spatial proximity of the satellite node and the orbital intersection area, the distance of each satellite node is calculated to obtain a spatial position weight factor; The orbital level weight coefficient and the spatial position weight factor are mathematically combined to obtain a comprehensive spatial priority; The comprehensive spatial priority is used to calculate the priority correction of the original link metric to obtain a corrected link metric function; Based on the corrected link metric function, the data transmission path is calculated and the backup path is generated to obtain a routing strategy.

[0031] Specifically, the routing system extracts the orbital height parameter of each satellite node from the spatial time-varying routing graph, which reflects the level of the node's orbit. By setting a reasonable height threshold, all satellite nodes are processed in layers, automatically divided into LEO (Low Earth Orbit), MEO (Medium Earth Orbit) and GEO (Geosynchronous Earth Orbit) three satellite node sets. According to the typical communication characteristics of different orbital layers, each type of node is given a differentiated weight according to its coverage range and transmission delay characteristics. LEO satellite nodes have smaller coverage but lower latency, frequent link switching and other characteristics, MEO satellite nodes are in the middle in terms of coverage and delay performance, and GEO satellite nodes have the widest coverage but the largest transmission delay. Considering the business type and routing target, LEO is assigned a higher weight coefficient to reflect its priority adaptation capability for low-latency services and fast-response services, MEO is assigned a medium weight, and GEO is given the lowest weight to reflect its disadvantage in delay-sensitive scenarios. The spatial proximity of all satellite nodes and orbital intersection regions is analyzed. All orbital intersection points or intersection regions in three-dimensional space are considered as high-value routing resources, and the nearest distance of the nodes to these intersection regions in the spatial coordinate system is used as the measurement standard. Using distance attenuation functions such as Gaussian or exponential form, the spatial position weight factor of each node is dynamically calculated. This factor reflects the spatial potential and communication value of the node in realizing cross-orbit efficient forwarding, and the closer the distance, the greater the weight, which can guide data flow to pass through nodes with high spatial connectivity and redundant potential in network routing decisions, realizing three-dimensional coordination of network resources and path diversification. The orbital level weight coefficient and the spatial position weight factor are mathematically combined to obtain the comprehensive spatial priority, which reflects the global value of the node in the spatial structure and orbital level. The original link metric is modified using the comprehensive spatial priority. The original link metric includes transmission delay, available bandwidth, energy consumption and stability of the link, while the priority modification is to adjust these indicators by weighting the inverse or normalized form of the comprehensive spatial priority, so that the link near the orbital intersection region and in the high weight orbital layer obtains a higher comprehensive score. This modification mechanism can effectively improve the competitiveness of cross-orbit forwarding paths in the routing algorithm, guide data packets to actively choose paths with obvious spatial advantages and sufficient structural redundancy, reduce the communication risk caused by single path congestion or node failure, and improve the global robustness of the routing scheme. Based on the modified link metric function, the intelligent routing system uses multi-path search and dynamic backup mechanism to calculate and maintain the data transmission path in real time. In the main path planning, the routing algorithm uses the modified metric function as the evaluation standard to preferentially select the path with the highest comprehensive weight, the lowest delay, the largest bandwidth and the strongest path stability, ensuring that business data can be efficiently and low-lossly forwarded across multiple orbital layers, and fully exploiting the communication potential of the spatial three-dimensional structure.In the backup path generation process, the system allocates at least one alternative path with large structural difference and few shared nodes for each main path according to the path diversity principle, and combines the link state prediction and congestion determination mechanism to make preparations for link switching and seamless service migration in advance, thereby improving the self-healing ability of the network to high dynamic environment and sudden failure.

[0032] In a specific embodiment, the process of step S500 can specifically include the following steps: Based on the corrected link metric function in the routing strategy, a decision variable set is generated, a traffic allocation unit is configured according to the decision variable set and the satellite network traffic demand, and a traffic allocation variable set is obtained; A global optimization objective function is constructed according to the decision variable set, the traffic allocation variable set and the end-to-end delay energy efficiency weighting coefficient; A constraint condition set is generated according to the traffic conservation, link capacity, link availability, cross-orbit path existence and path stability requirements; The global optimization objective function and the constraint condition set are input into a mixed integer programming solver for solving calculation, and an initial routing decision scheme is obtained; The initial routing decision scheme is subjected to link pre-screening and time domain smoothing processing, and an optimal routing decision set is obtained.

[0033] Specifically, according to the comprehensive performance of each link at each discrete time slice, a set of decision variables is generated for all optional links and their timing states using the modified link metric function. Each decision variable is used to identify whether the corresponding link is activated and used for actual data forwarding at a specific time, and these variables together form a set of routing decision variables that describe the network resource selection and routing feasibility. Real-time monitoring of traffic demand is performed, and the selected links are mapped to different data streams accordingly. Each link is assigned an actual loadable data flow to form a dynamically updated set of traffic allocation variables. A global optimization objective function is constructed based on the set of routing decision variables, the set of traffic allocation variables, and the end-to-end delay-energy efficiency trade-off coefficient. At the same time, a set of constraints is constructed. The flow conservation requirement stipulates that the inflow and outflow of any intermediate node in the network, except the source and destination nodes, must be equal, ensuring continuous data flow in the network without interruption or packet loss. Link capacity constraints ensure that the actual traffic of any link does not exceed its maximum bandwidth set by the physical or protocol, preventing congestion and affecting the overall service quality. Link availability constraints further require that only links that are available, in normal condition, and meet the physical connection requirements at a specific time can participate in routing selection, preventing path failure due to link failure, obstruction, or energy shortage. The cross-orbit path existence constraint requires that at least one link across different orbit levels must exist in the final routing scheme to take advantage of the connectivity and synergy of three-dimensional space networking and ensure that high-priority services can complete end-to-end data transmission across different satellite levels. The path stability requirement suppresses frequent hopping and drastic fluctuations in the route by constraining the routing changes between consecutive time slices, thereby maintaining service continuity, reducing switching delay, reducing signaling overhead, and improving the overall robustness of the network and user experience. After inputting the above optimization objective function and constraint condition set into the mixed integer programming solver, the system automatically performs variable allocation, constraint selection, and local solution combination by comprehensively considering the global link state, traffic demand, physical parameters, and service constraints. The mixed integer programming solver can efficiently handle large-scale, dynamic, and highly coupled variable sets through intelligent branch and bound, relaxation and rolling optimization, parallel iteration, and other algorithmic means, outputting an initial routing decision scheme that meets multiple objectives and constraint conditions. In the link pre-selection stage, real-time health status assessment and historical performance review are performed on all data links involved in routing, and links with high congestion risk, severe fluctuations, or high failure rate are excluded. Only links with good stability, load balancing, and high energy efficiency are retained in the final routing set. Time domain smoothing processing optimizes the transition of routing schemes between consecutive time slices through dynamic sliding windows, multi-step prediction, or path smoothing algorithms, preventing frequent switching of paths due to minor environmental changes, effectively reducing network control load and traffic jitter, and ensuring the smoothness and continuity of end-to-end communication. After link pre-selection and time domain smoothing processing, an optimal set of routing decisions is formed.

[0034] In a specific embodiment, the process of obtaining the initial routing decision scheme by solving the global optimization objective function and the constraint condition set by using the mixed integer programming solver can specifically include the following steps: relaxing the flow conservation constraint in the constraint condition set in the global optimization objective function by using the mixed integer programming solver to obtain a Lagrange function; decomposing the Lagrange function into a plurality of sub-problems, and setting initial Lagrange multipliers for the sub-problems to obtain an initialized sub-problem set; solving the initialized sub-problem set by using a branch and bound method to obtain local optimal solutions of the sub-problems, and updating and iterating the Lagrange multipliers based on the local optimal solutions of the sub-problems to obtain a global preliminary solution; performing feasibility test and boundary processing on the global preliminary solution based on a link load balancing criterion to obtain the initial routing decision scheme.

[0035] Specifically, the global optimization objective function and the flow conservation constraint in the constraint set are relaxed by using a mixed integer programming solver. The mixed integer programming solver uses the Lagrangian relaxation method to relax the flow conservation constraint. The flow conservation equation constraint is converted into a penalty term in the optimization objective function. By introducing a set of Lagrange multipliers, the influence weight of the constraint violation degree on the overall optimization objective is dynamically adjusted, and an adjustable and decomposable Lagrangian function is obtained. The Lagrangian function is decomposed into sub-problems. The problem decomposition is based on the spatial distribution of satellite nodes, the physical division of links, or the time domain blocking strategy. Each sub-problem only focuses on part of the decision variables and constraints, reducing the size and computational complexity of the single problem, and transforming the high-dimensional and strongly coupled original problem into multiple small-scale optimization sub-tasks with good structure. For each decomposed sub-problem, a set of initial Lagrange multipliers is set. These multipliers represent the current "price" of the corresponding relaxed constraint, i.e., the incentive or penalty signal that the network hopes each sub-problem can meet the overall optimization requirements in terms of flow conservation, load balancing, etc. By reasonably initializing these Lagrange multipliers, the sub-problem set can quickly enter the solving state, avoiding falling into a non-optimal or non-convergent solution space in the initial stage. After entering the actual solving stage, each sub-problem is independently input into the branch and bound solving module for optimization. The branch and bound algorithm is a high-efficiency combinatorial optimization technique that significantly reduces the computational load and quickly locks the local optimal solution by systematically dividing the solution space, layering pruning, and gradually approaching the optimal solution. The local optimal solution of the sub-problem reflects the best resource utilization, latency-energy balance, and path scheduling strategy that the network can achieve under the current allocation and constraint combination. With the completion of branch and bound calculation for all sub-problems, the system aggregates the local optimal solutions of all sub-problems and updates the global Lagrange multipliers based on these solutions. According to the Lagrange duality theory, the sub-gradient method or other parallel adaptive iteration strategies are used to dynamically feedback the "penalty" caused by local constraint violation or resource allocation imbalance into the next round of optimization objective. The continuous updating of the Lagrange multipliers guides the system to find a better solution combination within the global feasible region, making the entire optimization system adaptively converge to the balance point of the constraint optimal and objective optimal, and gradually forming a set of global preliminary solutions. Based on the link load balancing standard, the preliminary solutions are systematically checked for feasibility and boundary processing. The real-time load, historical stability, and health status of all links are checked to ensure that no link has traffic overflow, resource bottleneck, or extreme load imbalance. For solutions that exceed the physical boundaries, have extreme allocation, or do not meet the flow balance, the system automatically adjusts, such as reducing the flow on overloaded links, dynamically shutting down or reallocating low utilization links. At the same time, the end-to-end latency, energy consumption, and service requirements of all paths are ensured to meet the requirements of users or tasks, and additional objectives such as path switching and backup link selection are used to further optimize route selection.Through load balancing and boundary feasibility checking, an initial routing decision scheme is finally formed.

[0036] The space stereoscopic satellite routing method in the embodiments of the application is described above, and the space stereoscopic satellite routing device 10 in the embodiments of the application is described below. Please refer to Figure 2 The space stereoscopic satellite routing device 10 in the embodiments of the application includes one embodiment: The modeling module 11 is configured to construct a space satellite network based on LEO, MEO and GEO three-layer satellite nodes. The prediction module 12 is configured to perform orbit mechanics prediction calculation based on the space satellite network to obtain a link state prediction matrix. The processing module 13 is configured to input the link state prediction matrix into a space position coding processor for processing to obtain a space time-varying routing graph. The calculation module 14 is configured to generate a routing strategy of the satellite node based on the space time-varying routing graph. The optimization module 15 is configured to combine the routing strategy to perform mixed integer programming solution to obtain an optimal routing decision set.

[0037] Through the cooperation of the above components, the three-dimensional space modeling of the LEO, MEO and GEO three-layer satellite nodes is performed, the orbital height difference is fully considered, the stereoscopic network topology is constructed, and compared with the traditional two-dimensional plane routing method, the actual space distribution characteristics of the satellite network are more accurately reflected. The orbit mechanics prediction algorithm is used to realize the accurate prediction of the future position of the satellite and the link state, the link state prediction matrix is constructed, the routing decision has foresight, and the challenge brought by the frequent change of network topology is effectively coped with. The space position coding mechanism and the graph attention network are used to extract the stereoscopic topological features of the satellite node, the space time-varying routing graph is constructed, and comprehensive space-time information is provided for the routing decision. The cross-orbit priority routing mechanism is introduced, the link metric is optimized by comprehensively considering the space priority, the cross-orbit link passing through the orbital intersection area obtains higher priority, and the cross-orbit data transmission efficiency is significantly improved. The mixed integer programming model is applied to globally optimize the routing decision, and under the condition of meeting the network capacity and link availability, the global optimization of the end-to-end delay and energy consumption is realized.

[0038] Please refer to Figure 3 , Figure 3 The structure of the electronic system 300 provided in the embodiments of the application is shown in a schematic block diagram, and the electronic system 300 includes a processor 301 and a memory 302, and the processor 301 and the memory 302 are connected through a device bus 303, wherein the memory 302 can include a non-volatile storage medium and an internal memory.

[0039] The non-volatile storage medium can store a computer program. The computer program includes program instructions, which, when executed by the processor 301, can cause the processor 301 to perform any of the above spatial stereoscopic satellite routing methods.

[0040] The processor 301 is configured to provide computing and control capabilities to support the operation of the entire electronic system 300.

[0041] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium, which, when executed by the processor 301, can cause the processor 301 to perform any of the above spatial stereoscopic satellite routing methods.

[0042] Those skilled in the art can understand that, Figure 3 It should be understood that the structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the electronic system 300 involved in the scheme of the present application. The specific electronic system 300 can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0043] It should be understood that the processor 301 can be a central processing unit (CPU), and the processor 301 can also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0044] It should be noted that those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described electronic system 300 can refer to the corresponding process of the foregoing spatial stereoscopic satellite routing method, which will not be described here.

[0045] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0046] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the entire or part of the technical solutions that essentially contribute to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing an electronic system (which can be a personal computer, a server, or a network system, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0047] The above-described embodiments are merely used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for some of the technical features; and these modifications or replacements do not cause the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A spatial three-dimensional satellite routing method, characterized in that, include: A space satellite network is constructed based on three layers of satellite nodes: LEO, MEO, and GEO. Based on the aforementioned space satellite network, orbital mechanics prediction calculations are performed to obtain the link state prediction matrix; The link state prediction matrix is ​​input into the spatial location encoding processor for processing to obtain a spatial time-varying routing map. A routing strategy for satellite nodes is generated based on the aforementioned spatial time-varying routing map; By combining the aforementioned routing strategy with mixed-integer programming, the optimal routing decision set is obtained.

2. The space-based three-dimensional satellite routing method according to claim 1, characterized in that, The construction of a space satellite network based on LEO, MEO, and GEO three-layer satellite nodes includes: Orbital elements parameters such as orbital altitude, inclination, right ascension of ascending node, and argument of perigee of LEO, MEO, and GEO satellite nodes are collected to obtain a set of satellite orbital parameters; The satellite orbit parameter set is input into the Kepler orbit equation for coordinate transformation to obtain the spatial position coordinates of the LEO, MEO, and GEO satellite nodes in the three-dimensional Cartesian coordinate system. Based on the spatial coordinates, the distance between any two satellites is calculated to obtain the set of inter-satellite links and the subset of inter-orbit links; Based on the inter-satellite link set and the inter-orbit link subset, spatial position analysis is performed on the LEO, MEO, and GEO satellite nodes to obtain the space satellite network.

3. The space-based three-dimensional satellite routing method according to claim 1, characterized in that, The orbital mechanics prediction calculation based on the space satellite network, to obtain the link state prediction matrix, includes: Satellite orbit prediction is performed based on the position and velocity information of satellite nodes in the space satellite network to obtain orbit prediction data. Based on the orbit prediction data, the relative relationship between any two satellite nodes is calculated to obtain a positional relationship dataset; Based on the location relationship dataset and preset link visibility conditions, the status of inter-satellite links is determined to obtain a link availability time series. For each available link in the link availability time series, dynamic measurement calculation and spatiotemporal fusion are performed to obtain the link status prediction matrix.

4. The space-based three-dimensional satellite routing method according to claim 1, characterized in that, The step of inputting the link state prediction matrix into a spatial location encoding processor for processing to obtain a spatial time-varying routing map includes: The link state prediction matrix is ​​input into the time discretization unit of the spatial location coding processor for processing to obtain a discrete time series. Based on the discrete time series, the link status at each time point is processed using the topology building unit of the spatial location coding processor to obtain K network topology snapshots; The adjacent network topology snapshots are input into the transformation probability calculation unit of the spatial location encoding processor for processing to obtain the transformation probability matrix; The three-dimensional coordinates of satellite nodes are encoded using a position-sensitive hash function to obtain an embedding vector; The embedded vector is input into the graph attention network in the spatial location encoding processor for multi-head attention computation to obtain the hidden state features; Based on the network topology snapshot, the transition probability matrix, and the hidden state features, a spatial time-varying routing map is obtained by integrating four-dimensional information.

5. The space-based three-dimensional satellite routing method according to claim 1, characterized in that, The routing strategy for generating satellite nodes based on the spatial time-varying routing map includes: Based on the orbital altitude parameters of the satellite nodes in the spatial time-varying routing diagram, threshold layering processing is performed on the satellite nodes to obtain the LEO satellite node set, MEO satellite node set, and GEO satellite node set; Weights are assigned to the LEO satellite node set, MEO satellite node set, and GEO satellite node set to obtain orbital level weight coefficients, and distance is calculated for each satellite node to obtain spatial position weight factors. The orbital level weight coefficient and the spatial location weight factor are mathematically combined to obtain the comprehensive spatial priority, and the comprehensive spatial priority is used to generate the modified link metric function. Based on the modified link metric function, the data transmission path is calculated and backup paths are generated to obtain the routing strategy.

6. The space-based three-dimensional satellite routing method according to claim 1, characterized in that, The process of combining the routing strategy with mixed-integer programming yields the optimal routing decision set, including: Decision variables are generated based on the modified link metric function in the routing strategy to obtain a set of routing decision variables, and a set of traffic allocation variables is constructed based on the set of routing decision variables and the satellite network traffic requirements. A global optimization objective function is constructed based on the routing decision variable set, the traffic allocation variable set, and the end-to-end delay-energy-efficiency tradeoff coefficient. A set of constraints is generated based on requirements for traffic conservation, link capacity, link availability, cross-track path existence, and path stability. A mixed-integer programming solver is used to solve the global optimization objective function and the constraint set to obtain an initial routing decision scheme. The initial routing decision scheme is then subjected to link pre-screening and time-domain smoothing to obtain the optimal routing decision set.

7. The space-based three-dimensional satellite routing method according to claim 6, characterized in that, The method employs a mixed-integer programming solver to solve the global optimization objective function and the constraint set to obtain an initial routing decision scheme, including: The global optimization objective function and the flow conservation constraints in the constraint set are relaxed using a mixed-integer programming solver to obtain the Lagrangian function; The Lagrange function is decomposed into multiple subproblems, and the initial Lagrange multipliers of the subproblems are set to obtain an initialized set of subproblems. The initial set of subproblems is solved by branch and bound to obtain the local optimal solutions of each subproblem. The Lagrange multipliers are then updated and iterated based on the local optimal solutions of each subproblem to obtain the global preliminary solution. The feasibility of the preliminary global solution is verified and boundary conditions are handled based on the link load balancing standard to obtain the initial routing decision scheme.

8. A spatial three-dimensional satellite routing device, characterized in that, For executing the space-based three-dimensional satellite routing method as described in any one of claims 1-7, the space-based three-dimensional satellite routing device comprises: The modeling module is used to construct a space satellite network based on three layers of satellite nodes: LEO, MEO, and GEO. The prediction module is used to perform orbital mechanics prediction calculations based on the space satellite network to obtain the link state prediction matrix; The processing module is used to input the link state prediction matrix into the spatial location coding processor for processing to obtain a spatial time-varying routing map; The calculation module is used to generate routing strategies for satellite nodes based on the spatial time-varying routing map; The optimization module is used to perform mixed-integer programming in conjunction with the routing strategy to obtain the optimal routing decision set.

9. An electronic system, characterized in that, The electronic system includes: a memory and at least one processor, wherein the memory stores instructions; The at least one processor invokes the instructions in the memory to cause the electronic system to execute the space-based three-dimensional satellite routing method as described in any one of claims 1-7.

Images