Hyperbolic space based satellite network information and energy joint transmission routing optimization method

Abstract

The application discloses a satellite network information and energy joint transmission route optimization method based on hyperbolic space and belongs to the technical field of satellite communication.The application extracts the communication state, energy state and visibility evolution characteristics of nodes and links by constructing a time sequence diagram model of joint information and energy, and maps the characteristics to the hyperbolic space for unified representation, aiming at describing the hierarchical topology relationship and communication and energy coupling characteristics evolved with time in a large-scale satellite network. The application can improve the representation accuracy and routing decision efficiency of a complex dynamic satellite network, and provides a new technical approach for solving the path selection, resource allocation and system robustness optimization in the joint information and energy transmission scene.

Description

Technical Field

[0001] This invention belongs to the field of satellite communication technology, and in particular relates to a method for optimizing the joint transmission route of information and energy in satellite networks based on hyperbolic space. Background Technology

[0002] With the rapid development of low-Earth orbit mega-constellations, integrated space-ground networks, and space information network technologies, large-scale satellite networks are gradually evolving from traditional single information transmission systems into comprehensive network systems with both information transmission and energy service capabilities. In these networks, the network scale continues to expand, the number of nodes increases significantly, the orbital hierarchy becomes more complex, link states change rapidly over time, and there is significant heterogeneity between different orbital layers and between satellites and the ground. This makes routing optimization problems more complex than in traditional communication networks. Most existing satellite network routing methods still focus on communication performance, primarily revolving around shortest paths, minimum hop counts, contact graph routing, and reinforcement learning. While these methods can reduce latency, improve delivery success rates, or improve load distribution to some extent, they mostly assume the network only undertakes information transmission tasks and do not fully consider factors such as node remaining energy, receiving node energy gaps, link energy transmission capacity, and the transmission burden of power supply nodes. Therefore, they are difficult to apply to large-scale satellite network routing optimization problems in scenarios involving both information and energy transmission. Meanwhile, large-scale satellite networks typically exhibit distinct hierarchical, core-periphery, and dynamic evolution characteristics in their topology. If traditional Euclidean space representation methods or explicit graph search methods are still used, it is often difficult to effectively maintain the hierarchical structure and cross-layer coupling relationships of the network in low-dimensional space, and it is also difficult to balance modeling accuracy and online solution efficiency. Especially under conditions of rapid topology changes, limited link time windows, and coexistence of communication and energy coupling constraints, existing methods generally suffer from insufficient scalability, inadequate response speed, and poor robustness.

[0003] Hyperbolic space possesses geometric properties such as negative curvature and exponential volume growth, making it suitable for low-distortion representation of complex networks with hierarchical, tree-like, and core-edge structures. Therefore, introducing hyperbolic space representation into the routing problem of large-scale satellite networks is expected to more accurately characterize the hierarchical structure of satellite networks and provide more effective geometric guidance for subsequent path search. However, existing hyperbolic routing research is still mainly focused on communication scenarios and has not yet systematically integrated node energy states, link power supply capabilities, joint edge weight modeling, and joint path feasibility determination into a unified framework. Therefore, it still cannot meet the actual needs of large-scale satellite network joint information and energy transmission scenarios.

[0004] Therefore, there is an urgent need to propose a new routing optimization method that takes into account the hierarchical topology, dynamic characteristics, and joint information and energy transmission constraints of large-scale satellite networks, so as to improve routing efficiency, resource utilization and system robustness in complex dynamic environments. Summary of the Invention

[0005] To address the inherent challenges of large-scale satellite networks in joint information and energy transmission scenarios, such as large node size, complex hierarchical structure, strong link heterogeneity, and highly dynamic topology changes, this invention proposes a hyperbolic space-based routing optimization method for joint information and energy transmission in satellite networks. This method constructs a time-series graph model of joint information and energy transmission, extracts the communication state, energy state, and visibility evolution characteristics of nodes and links, and maps them onto hyperbolic space for unified representation. This aims to characterize the time-evolving hierarchical topology and communication-energy coupling characteristics in large-scale satellite networks. This approach not only improves the accuracy of characterization and routing decision-making efficiency for complex dynamic satellite networks but also provides a new technical approach for solving problems related to path selection, resource allocation, and system robustness optimization in joint information and energy transmission scenarios.

[0006] The first aspect of this invention discloses a satellite network information-energy joint transmission route optimization method based on hyperbolic space, the method comprising: S1: Construct a time-series network model for joint information and energy transmission, representing a large-scale satellite network as a dynamic directed graph structure containing a set of nodes, a set of links and their corresponding characteristics, and characterizing the relationship between communication reachability and energy transmission under a unified graph model; S2: Model the node and link characteristics in the network; node characteristics include node remaining energy, energy gap, load status, and communication resource status; link characteristics include link visibility, effective bandwidth, bit error rate, transmission delay, and energy transmission capacity; achieve a unified representation of joint information and energy transmission status; S3: Establish a hyperbolic space embedding model to map network nodes to hyperbolic space, and define two invertible functions, exponential mapping and logarithmic mapping, for bidirectional transformation between hyperbolic space and local tangent space to obtain the hierarchical topology of large-scale satellite networks. S4: A global embedding representation is constructed based on an energy-aware anchor selection strategy. An anchor set is obtained by comprehensively scoring the node stability, remaining energy, local energy supply capacity and load level, and the anchors are used to perform hyperbolic space interpolation representation of the nodes. S5: Construct an attention mechanism based on the joint features of service quality and energy, perform weighted aggregation on neighbor nodes to obtain the local dynamic representation of the nodes, and enhance the effectiveness of neighborhood information aggregation by jointly considering link latency, link quality, power supply capacity and energy requirements of receiving nodes. S6: Through the exponentially weighted moving average gating update mechanism, the node embedding is updated in the time dimension. When the local visibility, load status, latency level or energy status changes significantly, the current state write intensity is increased, while more historical information is retained when the network status is stable, so as to achieve adaptive modeling of dynamic topology evolution. S7: Construct a joint information and energy edge weight model, and uniformly model the communication delay cost, link transmission quality cost, and energy transmission burden cost to obtain the comprehensive link weight. In the edge weight calculation process, the differences in link type and the visibility window constraint are considered. S8: Path search is based on the edge weight model. It adopts a combination of candidate link screening and heuristic search for routing decisions. At the same time, a hyperbolic greedy backoff mechanism is introduced to ensure path reachability and online real-time performance under conditions of limited search budget or rapid topology changes. S9: Construct a joint optimization objective function to uniformly optimize the node embedding, edge weight estimation, path selection and energy allocation processes. By combining the learning results of joint constraint representation, the fitting effect of link cost and the performance of routing decisions, improve the overall routing efficiency and system robustness in the scenario of joint information and energy transmission in large-scale satellite networks.

[0007] According to the method of the first aspect of the present invention, in S1, a time-series network model for joint information and energy transfer is constructed, specifically including: Large-scale satellite networks at time Represented as a discrete-time graph of joint information and energy: in, and Representing time respectively The set of nodes and the set of directed edges, Represents the node feature matrix, Represents the set of link features, adjacency matrix Satisfy the node At any moment For nodes When reachable, ,otherwise and define nodes At any moment The set of visible successors is: This integrates link visibility, communication reachability, and energy service capabilities into a single time sequence graph model.

[0008] According to the method of the first aspect of the present invention, in S2, the node characteristics and link characteristics in the network are modeled, specifically including: For each directed edge The joint link feature vector is defined as follows: in, Indicates the visibility status of the link. Indicates the effective communication bandwidth under power allocation or time slot switching conditions. Indicates the remaining visible time window. Indicates the effective bit error rate under joint transmission. Indicates length is Grouping on the edge Effective single-hop delay, Indicates link quality metrics. Indicates the link mode. Indicates end-to-end energy transfer efficiency. Indicates the maximum allocatable transmit power. This indicates the capture, pointing, and tracking penalties of the laser link; Effective single-hop delay is defined as: in, Indicates the transmission distance. Represents the speed of light. Indicates the congestion waiting time. This indicates the processing and switching overhead related to the link mode; node At any moment The joint node feature vector is defined as: in, Indicates the amount of remaining communication resources. This represents the normalized buffer state description quantity. This represents the normalized queue state description quantity. Indicates the normalized load level. Indicates local load power consumption. Indicates local energy inflow. This indicates the current energy gap. This indicates a surplus of energy that can be used for relaying and power supply; The normalized visibility ratio, used to characterize the relative sufficiency of currently visible neighbors, is defined as follows: and: The aggregate single-hop latency of a node is defined as follows: in, Indicates allocation to neighbors Normalized aggregate weights; Node energy in time slots The update definition within is: link The amount of harvested energy delivered within a time slot is defined as: in, This represents the energy allocation variable.

[0009] When using a power-division receiving method, we have: When using time-slot switching reception, we have: This enables a complete characterization of the combined information and energy transmission status.

[0010] According to the method of the first aspect of the present invention, in S3, a hyperbolic space embedding model is established, mapping network nodes to hyperbolic space, and the transformation between hyperbolic space and tangent space is realized through exponential mapping and logarithmic mapping, specifically including: use Viponcaré sphere model represents negative curvature hyperbolic space: Among them, the curvature is tangent space at the origin Euclidean computation space; Define the exponential and logarithmic mappings at the origin as follows: Among them, the exponential mapping is used to map the tangent space vector to the hyperbolic space, and the logarithmic mapping is used to pull the hyperbolic space representation back to the tangent space.

[0011] According to the method of the first aspect of the present invention, in S4, a global embedding representation is constructed based on an energy-aware anchor selection strategy, specifically including: For each candidate node Define the anchor point scoring function: in, Indicates the topological stability of nodes. Represents the normalized residual energy. This represents the normalized local energy inflow. This represents the normalized load level; the highest score is selected. Each node constitutes an anchor point set: This allows the anchor point to explicitly respond to the node's energy state; Based on the characteristics of joint nodes, nodes Projecting onto the tangent space, the projection method is as follows: in, Indicates feature normalization, A learnable bias vector; For each anchor point Its hyperbolic coordinates are The corresponding tangent space is represented as: node With anchor point The similarity is defined as: in, For anchor point temperature parameters; node The anchor point interpolation tangent space representation is defined as: Mapping it back to hyperbolic space yields the global representation of the anchor point: This results in a global hyperbolic representation that simultaneously reflects the network hierarchy and the energy support capability of anchor points.

[0012] According to the method of the first aspect of the present invention, in S5, an attention mechanism based on the joint features of service quality and energy is constructed to perform weighted aggregation of neighboring nodes, specifically including: During the neighbor's attention phase, on the other side The energy affordability of an energy source is defined as: in, Indicates allocation on the link On the transmission power, Represents a node energy surplus Small positive constants; For each candidate neighbor, construct a joint service quality and energy feature vector: In this context, an overline or tilde indicates a normalized quantity; Using trainable scoring vectors Pre-select and rate neighbors: Take the highest-rated Top- The candidate neighbors constitute a set: The attention coefficients are obtained by performing softmax normalization on the selected neighbors. This means that neighbor selection depends not only on traditional communication QoS, but also on power supply capacity, receiver power gap, and laser alignment cost. Representing neighboring nodes The hidden representation of the previous time step, then the node The neighborhood tangent space aggregation result is as follows: After mapping back to hyperbolic space, we get: To balance the global pivot point representation with the local neighborhood refinement results, the candidate states are defined in the tangent space as follows: The candidate status then enters the time update phase.

[0013] According to the method of the first aspect of the present invention, in S6, the node embedding is updated in the time dimension through an exponentially weighted moving average gating update mechanism, specifically including: This represents the candidate tangent space state. Represents a node In the historical hidden table of the previous time step, the visible change, load change, and aggregate single-hop latency change are defined as follows: and: This represents the normalized residual energy; Define the net energy change rate as: Based on this, the write strength of the current state is characterized using a single scalar gating function: in, For learnable parameter vectors, For learnable bias, This represents the sigmoid activation function; Based on the above gating strength, perform an exponentially weighted moving average update in the tangent space: Mapping back to hyperbolic space yields the updated node representation: This allows for enhanced contribution of the current candidate state when network topology, traffic load, or energy state changes drastically, while retaining more historical information when the network state is relatively stable, thus enabling lightweight dynamic modeling of time-varying satellite networks.

[0014] According to the method of the first aspect of the present invention, in S7, a joint information and energy edge weight model is constructed to uniformly model the communication delay cost and the energy transmission cost, specifically including: For time Directed edge The communication-related edge weight components are defined as follows: in, Indicates the group length. Indicates the transmission distance. Represents the speed of light. Indicates the effective bandwidth. Indicates the congestion waiting time. Indicates the effective bit error rate. Indicates link quality metrics. This represents the capture, pointing, and tracking penalties of the laser link, and the coefficients are as follows. In the same link type Internal sharing enables lightweight edge weight estimation for type sharing; To represent the edge Energy-related costs, introducing an energy term: The first term represents the transmission burden of the power supply node, the second term represents the remaining unmet energy demand of the receiving node after the current link operation, and the third term represents the fairness adjustment for the reuse of power supply nodes. Normalized cumulative energy overhead is defined as: in, Indicates the current set of power supply nodes The average value; Based on communication edge weights and energy edge weights, the joint information and energy edge weights are defined as follows: in, and These represent the weighting coefficients for the time delay term and the energy term, respectively. Indicates temperature parameter The soft visibility window penalty function.

[0015] According to the method of the first aspect of the present invention, in S8, path search is performed based on the edge weight model, routing decisions are made by combining candidate link screening with heuristic search, and a hyperbolic greedy backoff mechanism is introduced, specifically including: In executing A Before extending the style path, for each outgoing edge Conduct a joint information and energy feasibility assessment, and... Only when the following conditions are met: An edge is considered feasible only if it meets the following conditions: When the route request also requires power service, it must also satisfy the following conditions: Furthermore, when the link mode is a laser link, it also satisfies... Edges that are topologically reachable but do not satisfy the joint constraints of communication and power supply are filtered out. For candidate edges that pass the feasibility screening, define a candidate scoring function: The first five terms characterize communication quality and visible window continuity, while the last three terms characterize the power supply burden, the receiver's remaining energy requirements, and the cost of laser alignment. Based on this, the top K feasible edges with the smallest scores are retained to participate in subsequent expansion. For some paths Its cumulative path cost is defined as: Indicates a partial path The search priority function for the terminal node is: in, Indicates starting from the current node To the target node Landmark-based heuristic residual cost estimation enables online path search with lower bound guidance under the joint edge weights. When the search budget is exhausted or the candidate frontier is sparse, a hyperbolic greedy backoff mechanism is activated, and the local selection potential function is defined as: in, The hyperbolic distance is used to guide the local direction of progress by combining geometric distance and the energy cost of joint communication, so that the retreat phase remains consistent with the overall goal. In A During style expansion and greedy backoff, a taboo set containing visited nodes in the current partial path is maintained to prevent candidate next hops from repeatedly visiting nodes that have already appeared, thereby avoiding loops and obtaining reachable and loop-free feasible paths under the condition of limited search budget.

[0016] According to the method of the first aspect of the present invention, in S9, a joint optimization objective function is constructed to uniformly optimize the node embedding, edge weight estimation, and path selection processes, specifically including: By aligning the communication component in the robust regression-constrained edge weight model with the instantaneous single-hop delay of the observation, the edge-level loss is defined as: in, A set of directed edge observations with timestamps. This represents the observed instantaneous joint single-hop delay. Indicates the Huber threshold; Define the set of triples: Constructing geometric sorting loss: in, Indicates the sorting interval. Indicates tolerance; To maintain the stability of anchor points during time evolution, an anchor point consistency loss is introduced: Ensure that the final embedding of the anchor point is consistent with its anchor point interpolation representation; To ensure that the attention distribution is consistent with the joint candidate score, a soft target distribution is constructed based on the candidate score: in, Define the attention alignment loss for the attention temperature parameter: in, This represents a non-empty set of candidate neighbors containing node and time pairs. To stabilize the gating time update process, a time smoothing loss is defined: When the gating is small, temporal continuity is enhanced; when the gating is large, fast-response topologies and energy mutations can be embedded. For route-level policing, define the source node Departure to Node The recursive form of the smooth joint path cost is: in, Represents a node The precursor collection, Representing smooth temperature, constructing path monitoring loss: This indicates that the route is based on the query set. Indicates the cost of the joint observation path; To ensure that the landmark heuristic function does not underestimate the cost of the remaining path, a reference consistency loss is introduced: This constrains the admissibility of the activation function under joint edge weights; To ensure sufficient energy services, an energy deficit penalty is introduced: in, This indicates the minimum operating energy limit for a node. This indicates the predicted remaining energy at the next moment; To avoid the long-term concentration of charging resources in a few nodes, the following definition is made: And construct fairness loss: in, This represents the set of nodes currently participating in the charging service; The final joint training objective is defined as in, This represents the non-negative weight coefficients of each objective, thereby achieving unified optimization of node representation, edge weight estimation, path search, energy sufficiency, and energy supply fairness.

[0017] A second aspect of this invention discloses a satellite network information and energy joint transmission routing optimization system based on hyperbolic space. The system includes a processing unit configured to perform the following steps: S1: Construct a time-series network model for joint information and energy transmission, representing a large-scale satellite network as a dynamic directed graph structure containing a set of nodes, a set of links and their corresponding characteristics, and characterizing the relationship between communication reachability and energy transmission under a unified graph model; S2: Model the node and link characteristics in the network; node characteristics include node remaining energy, energy gap, load status, and communication resource status; link characteristics include link visibility, effective bandwidth, bit error rate, transmission delay, and energy transmission capacity; achieve a unified representation of joint information and energy transmission status; S3: Establish a hyperbolic space embedding model to map network nodes to hyperbolic space, and define two invertible functions, exponential mapping and logarithmic mapping, for bidirectional transformation between hyperbolic space and local tangent space to obtain the hierarchical topology of large-scale satellite networks. S4: A global embedding representation is constructed based on an energy-aware anchor selection strategy. An anchor set is obtained by comprehensively scoring the node stability, remaining energy, local energy supply capacity and load level, and the anchors are used to perform hyperbolic space interpolation representation of the nodes. S5: Construct an attention mechanism based on the joint features of service quality and energy, perform weighted aggregation on neighbor nodes to obtain the local dynamic representation of the nodes, and enhance the effectiveness of neighborhood information aggregation by jointly considering link latency, link quality, power supply capacity and energy requirements of receiving nodes. S6: Through the exponentially weighted moving average gating update mechanism, the node embedding is updated in the time dimension. When the local visibility, load status, latency level or energy status changes significantly, the current state write intensity is increased, while more historical information is retained when the network status is stable, so as to achieve adaptive modeling of dynamic topology evolution. S7: Construct a joint information and energy edge weight model, and uniformly model the communication delay cost, link transmission quality cost, and energy transmission burden cost to obtain the comprehensive link weight. In the edge weight calculation process, the differences in link type and the visibility window constraint are considered. S8: Path search is based on the edge weight model. It adopts a combination of candidate link screening and heuristic search for routing decisions. At the same time, a hyperbolic greedy backoff mechanism is introduced to ensure path reachability and online real-time performance under conditions of limited search budget or rapid topology changes. S9: Construct a joint optimization objective function to uniformly optimize the node embedding, edge weight estimation, path selection and energy allocation processes. By combining the learning results of joint constraint representation, the fitting effect of link cost and the performance of routing decisions, improve the overall routing efficiency and system robustness in the scenario of joint information and energy transmission in large-scale satellite networks.

[0018] A third aspect of the present invention discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in the first aspect of the present invention.

[0019] A fourth aspect of the present invention discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described in the first aspect of the present invention.

[0020] This invention addresses the challenges of the rapidly increasing number of nodes and the highly dynamic nature of satellite communication networks by designing a large-scale satellite communication network representation model based on hyperbolic space. It comprehensively incorporates node alignment and topology smoothness modeling, and optimizes the representation through stochastic Riemann gradient descent, expanding the application scenarios of hyperbolic geometry in integrated space-ground networks. This invention innovatively maps satellite communication networks to hyperbolic space, significantly improving the accuracy of network representation. This technological breakthrough opens up new ideas and methods for the integrated space-ground integration, routing optimization, and capacity optimization of future large-scale satellite communication networks, providing strong technical support for the further development of the satellite communication field. Attached Figure Description

[0021] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0022] Figure 1 This is a flowchart of a satellite network information and energy joint transmission route optimization method based on hyperbolic space according to an embodiment of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] This invention discloses a satellite network information and energy joint transmission route optimization method based on hyperbolic space (e.g., Figure 1 (As shown).

[0025] S1: Construct a time-series network model for joint information and energy transmission, representing a large-scale satellite network as a dynamic directed graph structure containing a set of nodes, a set of links and their corresponding characteristics, and characterizing the relationship between communication reachability and energy transmission under a unified graph model.

[0026] S11: The large-scale satellite network will be in time... Represented as a discrete-time graph of joint information and energy: in, and Representing time respectively The set of nodes and the set of directed edges, Represents the node feature matrix, Represents the set of link features, adjacency matrix Satisfy the node At any moment For nodes When reachable, ,otherwise and define nodes At any moment The set of visible successors is: This integrates link visibility, communication reachability, and energy service capabilities into a single time sequence graph model.

[0027] S2: Model the node characteristics and link characteristics in the network; node characteristics include node remaining energy, energy gap, load status and communication resource status; link characteristics include link visibility, effective bandwidth, bit error rate, transmission delay and energy transmission capacity; realize a unified representation of joint information and energy transmission status.

[0028] S21: For each directed edge The joint link feature vector is defined as follows: in, Indicates the visibility status of the link. Indicates the effective communication bandwidth under power allocation or time slot switching conditions. Indicates the remaining visible time window. Indicates the effective bit error rate under joint transmission. Indicates length is Grouping on the edge Effective single-hop delay, Indicates link quality metrics. Indicates the link mode. Indicates end-to-end energy transfer efficiency. Indicates the maximum allocatable transmit power. This indicates the capture, pointing, and tracking penalties of the laser link.

[0029] S22: The effective single-hop delay is defined as follows: in: in, Indicates the transmission distance. Represents the speed of light. Indicates the congestion waiting time. This indicates the processing and switching overhead associated with the link mode.

[0030] S23: Node At any moment The joint node feature vector is defined as: in, Indicates the amount of remaining communication resources. This represents the normalized buffer state description quantity. This represents the normalized queue state description quantity. Indicates the normalized load level. Indicates local load power consumption. Indicates local energy inflow. This indicates the current energy gap. This indicates a surplus of energy that can be used for relaying and power supply.

[0031] S24: The normalized visibility ratio, used to characterize the relative sufficiency of currently visible neighbors, is defined as follows: And note: This is to ensure a consistent representation of visible changes during subsequent time update phases.

[0032] S25: The aggregate single-hop delay of a node is defined as follows: in, Indicates allocation to neighbors Normalized aggregate weights.

[0033] S26: Node energy in time slots The update definition within is: link The amount of harvested energy delivered within a time slot is defined as: in, This represents the energy allocation variable.

[0034] S27: When using power-division reception, the following applies: When using time-slot switching reception, we have: This enables a complete characterization of the combined information and energy transmission status.

[0035] S3: Establish a hyperbolic space embedding model, map network nodes to hyperbolic space, and realize the transformation between hyperbolic space and tangent space through exponential and logarithmic mappings. S31: Adopted Viponcaré sphere model represents negative curvature hyperbolic space: Among them, the curvature is tangent space at the origin As a Euclidean computation space.

[0036] S32: Define the exponential and logarithmic mappings at the origin as follows: Among them, the exponential mapping is used to map the tangent space vector to the hyperbolic space, and the logarithmic mapping is used to pull the hyperbolic space representation back to the tangent space, thus providing a unified invertible transformation basis for subsequent anchor point interpolation, neighborhood aggregation and time update.

[0037] S4: A global embedding representation is constructed based on an energy-aware anchor selection strategy. An anchor set is obtained by comprehensively scoring the node stability, remaining energy, local energy supply capacity and load level, and the anchors are used to perform hyperbolic space interpolation representation of the nodes.

[0038] S41: To construct a global anchor reference, for each candidate node Define the anchor point scoring function: in, Indicates the topological stability of nodes. Represents the normalized residual energy. This represents the normalized local energy inflow. This represents the normalized load level; the highest score is selected. Each node constitutes an anchor point set: This allows the anchor point to explicitly respond to the node's energy state.

[0039] S42: Based on the joint node features defined in step 2, the nodes Projecting onto the tangent space, the projection method is as follows: in, Indicates feature normalization, A learnable bias vector; S43, for each anchor point Let its hyperbolic coordinates be... The corresponding tangent space is represented as: node With anchor point The similarity is defined as: in, This refers to the anchor point temperature parameter.

[0040] S44: Node The anchor point interpolation tangent space representation is defined as: Mapping it back to hyperbolic space yields the global representation of the anchor point: This results in a global hyperbolic representation that simultaneously reflects the network hierarchy and the energy support capability of anchor points.

[0041] S5: Construct an attention mechanism based on the joint features of service quality and energy, perform weighted aggregation on neighboring nodes to obtain the local dynamic representation of the nodes, and enhance the effectiveness of neighborhood information aggregation by jointly considering link latency, link quality, power supply capacity and energy requirements of receiving nodes.

[0042] S51: During the neighbor attention phase, the opposite edge The affordability of energy supply is first defined as: in, Indicates allocation on the link On the transmission power, Represents a node energy surplus It is a small positive constant.

[0043] S52: For each candidate neighbor, construct a joint service quality and energy feature vector: In this context, an overline or tilde indicates a normalized quantity.

[0044] S53, utilizing trainable scoring vectors Pre-select and rate neighbors: Take the highest-rated Top- The candidate neighbors constitute a set: Then, perform softmax normalization on the selected neighbors to obtain the attention coefficients: This means that neighbor selection depends not only on traditional communication QoS, but also on power supply capacity, receiver power gap, and laser alignment costs.

[0045] S54: Let Representing neighboring nodes The hidden representation of the previous time step, then the node The neighborhood tangent space aggregation result is as follows: After mapping back to hyperbolic space, we get: Furthermore, to balance the global pivot point representation with the local neighborhood refinement results, the candidate states are defined in the tangent space as follows: The candidate status then enters the time update phase.

[0046] S4: Through an exponentially weighted moving average gating update mechanism, node embeddings are updated in the time dimension. When local visibility, load status, latency level or energy status changes significantly, the current state write intensity is increased, while more historical information is retained when the network status is stable, so as to achieve adaptive modeling of dynamic topology evolution.

[0047] S61: Let This represents the candidate tangent space state obtained from step 4. Represents a node In the historical hidden table of the previous time step, the visible change, load change, and aggregate single-hop latency change are defined as follows: And order: This represents the normalized residual energy.

[0048] S62, defines the net rate of change of energy as: Based on this, the write strength of the current state is characterized using a single scalar gating function: in, For learnable parameter vectors, For learnable bias, This represents the sigmoid activation function.

[0049] S63: Based on the above gating strength, perform an exponentially weighted moving average update in the tangent space: Mapping back to hyperbolic space yields the updated node representation: This allows for enhanced contribution of the current candidate state when network topology, traffic load, or energy state changes drastically, while retaining more historical information when the network state is relatively stable, thus enabling lightweight dynamic modeling of time-varying satellite networks.

[0050] S7: Construct a joint information and energy edge weight model, which unifies the communication delay cost, link transmission quality cost, and energy transmission burden cost to obtain the comprehensive link weight, and considers the differences in link type and the visibility window constraint in the edge weight calculation process.

[0051] S71: Regarding time Directed edge First, the communication-related edge weight components are defined as follows: in, Indicates the group length. Indicates the transmission distance. Represents the speed of light. Indicates the effective bandwidth. Indicates the congestion waiting time. Indicates the effective bit error rate. Indicates link quality metrics. This represents the capture, pointing, and tracking penalties of the laser link, and the coefficients are as follows. In the same link type Internal sharing enables lightweight edge weight estimation with type sharing.

[0052] S72: To characterize the edge Energy-related costs, introducing an energy term: The first term represents the transmission burden of the power supply node, the second term represents the remaining unmet energy demand of the receiving node after the current link operation, and the third term represents the fairness adjustment for the reuse of power supply nodes.

[0053] S73: The normalized cumulative energy overhead in the above equation is defined as follows: in, Indicates the current set of power supply nodes The average value.

[0054] S74: Based on communication edge weights and energy edge weights, the joint information and energy edge weights are defined as follows: in, and These represent the weighting coefficients for the time delay term and the energy term, respectively. Indicates temperature parameter The soft visibility window penalty function smoothly incorporates the visibility window constraint into the edge weight through a soft penalty term, thereby avoiding the edge weight discontinuity problem caused by simple hard constraints, and making the final edge weight explicitly reflect the coupling relationship between communication and energy.

[0055] S8: Path search is based on the edge weight model. It adopts a combination of candidate link screening and heuristic search for routing decisions. At the same time, a hyperbolic greedy backoff mechanism is introduced to ensure path reachability and online real-time performance under conditions of limited search budget or rapid topology changes.

[0056] S81: In executing A Before extending the style path, first examine each outgoing edge. Conduct a joint information and energy feasibility assessment, and... Only when the following conditions are met: An edge is considered feasible only when it meets certain conditions. Furthermore, if the route request also requires power services, the following conditions must also be met: Furthermore, when the link mode is a laser link, the following should also be satisfied: Through the above joint screening, edges that are topologically reachable but do not satisfy the joint constraints of communication and power supply are eliminated.

[0057] S82: Define a candidate scoring function for candidate edges that pass the feasibility screening: The first five parameters characterize communication quality and visible time window continuity, while the latter three characterize the power supply burden, the receiver's remaining energy requirements, and the cost of laser alignment. Based on these parameters, the lowest-scoring Top-ranked parameter is retained. Each feasible edge can participate in subsequent expansions.

[0058] S83: For partial paths Its cumulative path cost is defined as: And order Indicates a partial path The search priority function for the terminal node is: in, Indicates starting from the current node To the target node The landmark heuristic residual cost estimation enables online path search with lower bounds in the sense of joint edge weights.

[0059] S84: When the search budget is exhausted or the candidate frontier is too sparse, a hyperbolic greedy backoff mechanism is initiated, whose local selection potential function is defined as: in, Representing hyperbolic distance, the local forward direction is guided by the geometric distance and the energy cost of joint communication, so that the retreat phase remains consistent with the overall goal.

[0060] S85: In A During style expansion and greedy fallback, a taboo set containing visited nodes in the current partial path is maintained to prevent candidate next hops from repeatedly visiting nodes that have already appeared, thereby avoiding loops and ensuring that reachable and loop-free feasible paths can still be obtained under the condition of limited search budget.

[0061] S9: Construct a joint optimization objective function to uniformly optimize the node embedding, edge weight estimation, path selection and energy allocation processes. By combining the learning results of joint constraint representation, the fitting effect of link cost and the performance of routing decisions, improve the overall routing efficiency and system robustness in the scenario of joint information and energy transmission in large-scale satellite networks.

[0062] S91: First, the communication component in the robust regression-constrained edge weight model is aligned with the instantaneous single-hop delay of the observation. The edge-level loss is defined as: in, A set of directed edge observations with timestamps. This represents the observed instantaneous joint single-hop delay. This represents the Huber threshold.

[0063] S92: To ensure that the learned hyperbolic geometry is consistent with the joint routing preferences, a set of triples is defined: Based on this, a geometric sorting loss is constructed: in, Indicates the sorting interval. Indicates tolerance.

[0064] S93: To maintain the stability of anchor points during time evolution, an anchor point consistency loss is introduced: This ensures that the final embedding of the anchor point is consistent with its anchor point interpolation representation.

[0065] S94: To ensure that the attention distribution is consistent with the joint candidate score, a soft target distribution is first constructed based on the candidate score: in, Let's define the attention alignment loss as the attention temperature parameter: in, This represents a non-empty set of candidate neighbors containing node and time pairs.

[0066] S95: Defines a time smoothing loss to stabilize the gated time update process: When the gating is small, temporal continuity is enhanced; when the gating is large, fast-response topologies and energy mutations can be embedded.

[0067] S96: Defines routing-level policing from the source node. Departure to Node The recursive form of the smooth joint path cost is: in, Represents a node The precursor collection, The smoothed temperature is used to construct the path monitoring loss: in, This indicates that the route is based on the query set. This represents the cost of the joint observation path.

[0068] S97: To ensure that the landmark heuristic function does not underestimate the cost of the remaining path, a reference consistency loss is introduced: This constrains the admissibility of the activation function under joint edge weights.

[0069] S98: To ensure sufficient energy service, an energy deficit penalty is introduced: in, This indicates the minimum operating energy limit for a node. This indicates the predicted remaining energy at the next moment.

[0070] S99, to avoid long-term concentration of charging resources in a few nodes, defines... And construct fairness loss in, This represents the set of nodes currently participating in the charging service; S910, taking into account the above factors, the final joint training objective is defined as follows: in, This represents the non-negative weight coefficients of each objective, thereby achieving unified optimization of node representation, edge weight estimation, path search, energy sufficiency, and energy supply fairness.

[0071] In summary, this invention constructs a time-series graph model of joint information and energy, extracts the communication state, energy state, and visibility evolution characteristics of nodes and links, and maps them to hyperbolic space for unified representation. This aims to characterize the hierarchical topological relationships and communication-energy coupling characteristics that evolve over time in large-scale satellite networks. This approach not only improves the accuracy of characterization and routing decision-making efficiency of complex dynamic satellite networks but also provides a new technical approach to solving problems such as path selection, resource allocation, and system robustness optimization in joint information and energy transmission scenarios.

[0072] Please note that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments have been described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification. The above embodiments only illustrate several implementations of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be pointed out that for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of the present invention should be determined by the appended claims.

Claims

1. A satellite network information and energy joint transmission route optimization method based on hyperbolic space, characterized in that, The method includes: S1: Construct a time-series network model for joint information and energy transmission, representing a large-scale satellite network as a dynamic directed graph structure containing a set of nodes, a set of links and their corresponding characteristics, and characterizing the relationship between communication reachability and energy transmission under a unified graph model; S2: Model the node and link characteristics in the network; node characteristics include node remaining energy, energy gap, load status, and communication resource status; link characteristics include link visibility, effective bandwidth, bit error rate, transmission delay, and energy transmission capacity; achieve a unified representation of joint information and energy transmission status; S3: Establish a hyperbolic space embedding model to map network nodes to hyperbolic space, and define two invertible functions, exponential mapping and logarithmic mapping, for bidirectional transformation between hyperbolic space and local tangent space to obtain the hierarchical topology of large-scale satellite networks. S4: A global embedding representation is constructed based on an energy-aware anchor selection strategy. An anchor set is obtained by comprehensively scoring the node stability, remaining energy, local energy supply capacity and load level, and the anchors are used to perform hyperbolic space interpolation representation of the nodes. S5: Construct an attention mechanism based on the joint features of service quality and energy, perform weighted aggregation on neighbor nodes to obtain the local dynamic representation of the nodes, and enhance the effectiveness of neighborhood information aggregation by jointly considering link latency, link quality, power supply capacity and energy requirements of receiving nodes. S6: Through the exponentially weighted moving average gating update mechanism, the node embedding is updated in the time dimension. When the local visibility, load status, latency level or energy status changes significantly, the current state write intensity is increased, while more historical information is retained when the network status is stable, so as to achieve adaptive modeling of dynamic topology evolution. S7: Construct a joint information and energy edge weight model, and uniformly model the communication delay cost, link transmission quality cost, and energy transmission burden cost to obtain the comprehensive link weight. In the edge weight calculation process, the differences in link type and the visibility window constraint are considered. S8: Path search is based on the edge weight model. It adopts a combination of candidate link screening and heuristic search for routing decisions. At the same time, a hyperbolic greedy backoff mechanism is introduced to ensure path reachability and online real-time performance under conditions of limited search budget or rapid topology changes. S9: Construct a joint optimization objective function to uniformly optimize the node embedding, edge weight estimation, path selection and energy allocation processes. By combining the learning results of joint constraint representation, the fitting effect of link cost and the performance of routing decisions, improve the overall routing efficiency and system robustness in the scenario of joint information and energy transmission in large-scale satellite networks.

2. The satellite network information and energy joint transmission route optimization method based on hyperbolic space according to claim 1, characterized in that, In S1, a time-series network model for joint information and energy transfer is constructed, specifically including: Large-scale satellite networks at time Represented as a discrete-time graph of joint information and energy: in, and Representing time respectively The set of nodes and the set of directed edges, Represents the node feature matrix, Represents the set of link features, adjacency matrix Satisfy when node At any moment For nodes When reachable, ,otherwise and define nodes At any moment The set of visible successors is: This integrates link visibility, communication reachability, and energy service capabilities into a single time sequence graph model.

3. The satellite network information and energy joint transmission route optimization method based on hyperbolic space according to claim 2, characterized in that, In S2, the characteristics of nodes and links in the network are modeled, specifically including: For each directed edge The joint link feature vector is defined as follows: in, Indicates the visibility status of the link. Indicates the effective communication bandwidth under power allocation or time slot switching conditions. Indicates the remaining visible time window. Indicates the effective bit error rate under joint transmission. Indicates length is Grouping on the edge Effective single-hop delay, Indicates link quality metrics, Indicates the link mode. Indicates end-to-end energy transfer efficiency. Indicates the maximum allocatable transmit power. This indicates the capture, pointing, and tracking penalties of the laser link; Effective single-hop delay is defined as: in, Indicates the transmission distance. Represents the speed of light. Indicates the congestion waiting time. This indicates the processing and switching overhead related to the link mode; node At any moment The joint node feature vector is defined as: in, Indicates the amount of remaining communication resources. This represents the normalized buffer state description quantity. This represents the normalized queue state description quantity. Indicates the normalized load level. Indicates local load power consumption. Indicates local energy inflow. This indicates the current energy gap. This indicates a surplus of energy that can be used for relaying and power supply; The normalized visibility ratio, used to characterize the relative sufficiency of currently visible neighbors, is defined as follows: and: The aggregate single-hop latency of a node is defined as follows: in, Indicates allocation to neighbors Normalized aggregate weights; Node energy in time slots The update definition within is: link The amount of harvested energy delivered within a time slot is defined as: in, This represents the energy allocation variable. When using a power-division receiving method, we have: When using time-slot switching reception, we have: This enables a complete characterization of the combined information and energy transmission status.

4. The satellite network information and energy joint transmission route optimization method based on hyperbolic space according to claim 3, characterized in that, In S3, a hyperbolic space embedding model is established to map network nodes to hyperbolic space. The transformation between hyperbolic space and tangent space is achieved through exponential and logarithmic mappings, specifically including: use Viponcaré sphere model represents negative curvature hyperbolic space: Among them, the curvature is tangent space at the origin Euclidean computation space; Define the exponential and logarithmic mappings at the origin as follows: Among them, the exponential mapping is used to map the tangent space vector to the hyperbolic space, and the logarithmic mapping is used to pull the hyperbolic space representation back to the tangent space.

5. The satellite network information and energy joint transmission route optimization method based on hyperbolic space according to claim 4, characterized in that, In S4, a global embedding representation is constructed based on an energy-aware anchor selection strategy, specifically including: For each candidate node Define the anchor point scoring function: in, Indicates the topological stability of nodes. Represents the normalized residual energy. This represents the normalized local energy inflow. This represents the normalized load level; the highest score is selected. Each node constitutes an anchor point set: This allows the anchor point to explicitly respond to the node's energy state; Based on the characteristics of joint nodes, nodes Projecting onto the tangent space, the projection method is as follows: in, Indicates feature normalization, A learnable bias vector; For each anchor point Its hyperbolic coordinates are The corresponding tangent space is represented as: node With anchor point The similarity is defined as: in, For anchor point temperature parameters; node The anchor point interpolation tangent space representation is defined as: Mapping it back to hyperbolic space yields the global representation of the anchor point: This results in a global hyperbolic representation that simultaneously reflects the network hierarchy and the energy support capability of anchor points.

6. The satellite network information and energy joint transmission route optimization method based on hyperbolic space according to claim 5, characterized in that, In S5, an attention mechanism based on the joint features of service quality and energy is constructed to perform weighted aggregation of neighbor nodes, specifically including: During the neighbor's attention phase, on the other side The energy affordability of an energy source is defined as: in, Indicates allocation on the link On the transmission power, Represents a node energy surplus Small positive constants; For each candidate neighbor, construct a joint service quality and energy feature vector: In this context, an overline or tilde indicates a normalized quantity; Using trainable scoring vectors Pre-select and rate the neighbors: Take the highest-rated Top- The candidate neighbors constitute a set: The attention coefficients are obtained by performing softmax normalization on the selected neighbors. This means that neighbor selection depends not only on traditional communication QoS, but also on power supply capacity, receiver power gap, and laser alignment cost. Representing neighboring nodes The hidden representation of the previous time step, then the node The neighborhood tangent space aggregation result is as follows: After mapping back to hyperbolic space, we get: To balance the global pivot point representation with the local neighborhood refinement results, the candidate states are defined in the tangent space as follows: The candidate status then enters the time update phase.

7. The satellite network information and energy joint transmission route optimization method based on hyperbolic space according to claim 6, characterized in that, In S6, the node embedding is updated over time using an exponentially weighted moving average gating update mechanism, specifically including: This represents the candidate tangent space state. Represents a node In the historical hidden table of the previous time step, the visible change, load change, and aggregate single-hop latency change are defined as follows: and: This represents the normalized residual energy; Define the net energy change rate as: Based on this, the write strength of the current state is characterized using a single scalar gating function: in, For learnable parameter vectors, For learnable bias, This represents the sigmoid activation function; Based on the above gating strength, perform an exponentially weighted moving average update in the tangent space: Mapping back to hyperbolic space yields the updated node representation: This allows for enhanced contribution of the current candidate state when network topology, traffic load, or energy state changes drastically, while retaining more historical information when the network state is relatively stable, thus enabling lightweight dynamic modeling of time-varying satellite networks.

8. The satellite network information and energy joint transmission route optimization method based on hyperbolic space according to claim 7, characterized in that, In S7, a joint information and energy edge-weighted model is constructed to unify the modeling of communication latency costs and energy transmission costs, specifically including: For time Directed edge The communication-related edge weight components are defined as follows: in, Indicates the group length. Indicates the transmission distance. Represents the speed of light. Indicates the effective bandwidth. Indicates the congestion waiting time. Indicates the effective bit error rate. Indicates link quality metrics, This represents the capture, pointing, and tracking penalties of the laser link, and the coefficients are as follows. In the same link type Internal sharing enables lightweight edge weight estimation for type sharing; To represent the edge Energy-related costs, introducing an energy term: The first term represents the transmission burden of the power supply node, the second term represents the remaining unmet energy demand of the receiving node after the current link operation, and the third term represents the fairness adjustment for the reuse of power supply nodes. Normalized cumulative energy overhead is defined as: in, Indicates the current set of power supply nodes The average value; Based on communication edge weights and energy edge weights, the joint information and energy edge weights are defined as follows: in, and These represent the weighting coefficients for the time delay term and the energy term, respectively. Indicates temperature parameter The soft visibility window penalty function.

9. The satellite network information and energy joint transmission route optimization method based on hyperbolic space according to claim 8, characterized in that, In S8, path search is based on an edge-weight model, and routing decisions are made by combining candidate link selection with heuristic search. A hyperbolic greedy backoff mechanism is also introduced, specifically including: In executing A Before extending the style path, for each outgoing edge Conduct a joint information and energy feasibility assessment, and... Only when the following conditions are met: An edge is considered feasible only if it meets the following conditions: When the route request also requires power service, it must also satisfy the following conditions: Furthermore, when the link mode is a laser link, it also satisfies... Edges that are topologically reachable but do not satisfy the joint constraints of communication and power supply are filtered out. For candidate edges that pass the feasibility screening, define a candidate scoring function: The first five terms characterize communication quality and visible window continuity, while the last three terms characterize the power supply burden, the receiver's remaining energy requirements, and the cost of laser alignment. Based on this, the top K feasible edges with the smallest scores are retained to participate in subsequent expansion. For some paths Its cumulative path cost is defined as: Indicates a partial path The search priority function for the terminal node is: in, Indicates starting from the current node To the target node Landmark-based heuristic residual cost estimation enables online path search with lower bound guidance under the joint edge weights. When the search budget is exhausted or the candidate frontier is sparse, a hyperbolic greedy backoff mechanism is activated, and the local selection potential function is defined as: in, The hyperbolic distance is used to guide the local direction of progress by combining geometric distance and the energy cost of joint communication, so that the retreat phase remains consistent with the overall goal. In A During style expansion and greedy backoff, a taboo set containing visited nodes in the current partial path is maintained to prevent candidate next hops from repeatedly visiting nodes that have already appeared, thereby avoiding loops and obtaining reachable and loop-free feasible paths under the condition of limited search budget.

10. The satellite network information and energy joint transmission route optimization method based on hyperbolic space according to claim 9, characterized in that, In S9, a joint optimization objective function is constructed to uniformly optimize the node embedding, edge weight estimation, and path selection processes, specifically including: By aligning the communication component in the robust regression-constrained edge weight model with the instantaneous single-hop delay of the observation, the edge-level loss is defined as: in, A set of directed edge observations with timestamps. This represents the observed instantaneous joint single-hop delay. Indicates the Huber threshold; Define the set of triples: Constructing geometric sorting loss: in, Indicates the sorting interval. Indicates tolerance; To maintain the stability of anchor points during time evolution, an anchor point consistency loss is introduced: Ensure that the final embedding of the anchor point is consistent with its anchor point interpolation representation; To ensure that the attention distribution is consistent with the joint candidate score, a soft target distribution is constructed based on the candidate score: in, Define the attention alignment loss for the attention temperature parameter: in, This represents a non-empty set of candidate neighbors containing node and time pairs. To stabilize the gating time update process, a time smoothing loss is defined: When the gating is small, temporal continuity is enhanced; when the gating is large, fast-response topologies and energy mutations can be embedded. For route-level policing, define the source node Departure to Node The recursive form of the smooth joint path cost is: in, Represents a node The precursor collection, Representing smooth temperature, constructing path monitoring loss: This indicates that the route is based on the query set. Indicates the cost of the joint observation path; To ensure that the landmark heuristic function does not underestimate the cost of the remaining path, a reference consistency loss is introduced: This constrains the admissibility of the activation function under joint edge weights; To ensure sufficient energy services, an energy deficit penalty is introduced: in, This indicates the minimum operating energy limit for a node. This indicates the predicted remaining energy at the next moment. To avoid the long-term concentration of charging resources in a few nodes, the following definition is made: And construct fairness loss: in, This represents the set of nodes currently participating in the charging service; The final joint training objective is defined as in, This represents the non-negative weight coefficients of each objective, thereby achieving unified optimization of node representation, edge weight estimation, path search, energy sufficiency, and energy supply fairness.

Images