Constellation dynamic sub-domain satellite-ground management method based on node importance evaluation and distributed state machine driving

Abstract

The application discloses a constellation dynamic sub-domain satellite-ground management method based on node importance evaluation and distributed state machine driving, aims to provide a constellation dynamic sub-domain satellite-ground management method based on node importance evaluation and distributed state machine driving which can be distributedly executed, can be real-time self-adaptive, can simultaneously constrain management overhead and reliability, so that in the case that inter-satellite link topology changes over time, the system can automatically maintain the feasibility of sub-domains (domain capacity, connectivity, diameter / radius, etc.), and can guarantee the network reliability level while significantly reducing the performance fluctuation caused by unstable controller selection points.

Description

Technical Field

[0001] This invention belongs to the field of satellite constellation network and space-ground integrated management and control technology, specifically involving a dynamic domain-based management and intra-domain control node selection method for large-scale satellite constellations, especially multi-orbit, multi-satellite constellations, and a space-ground control strategy generation method based on this method. This method is applicable to the following typical service and system configurations: first, self-organizing management and autonomous network operation and maintenance of inter-satellite links (ISL) for large-scale low-Earth orbit communication constellations; second, domain-based control and cross-domain collaboration for constellation services sensitive to latency, connectivity, and availability (broadband communication, remote sensing data backhaul, emergency communication); third, autonomous collaboration and domain-based scheduling for mission-oriented constellations such as space situational awareness and space debris monitoring; and fourth, hierarchical control / domain-based control and adaptive organization under control overhead constraints for space-ground integrated networks.

[0002] In terms of engineering implementation, this invention can be deployed in: on-board computing units (distributed execution), ground control centers (centralized or semi-centralized execution), and hybrid architectures of space-ground collaboration. Inputs can come from simulation platforms such as STK, orbital mechanics propagators, and TLE / ephemeris data. Outputs can be directly used in space-ground control systems, including domain partitioning tables, domain management node lists, migration / re-domain operation sequences, and control overhead assessment results. Background Technology

[0003] Existing constellation domain partitioning and control schemes commonly employ approaches including: static geometric partitioning (based on orbital planes / latitude and longitude grids), static or quasi-static clustering based on clustering (such as K-means using only location), domain management node selection based on traditional network centrality (only topology closeness / degree), and satellite-to-ground control strategies primarily employing centralized SDN control. These schemes typically exhibit the following shortcomings in large-scale, highly dynamic scenarios (links changing over time):

[0004] First, the domain partitioning is either "static" or "weakly dynamic." Many schemes only perform a few periodic repartitions after the initial partitioning, which is insufficient to cope with changes in the inter-satellite link topology on minute-level or even shorter time scales. This leads to frequent failures of constraints such as intra-domain connectivity and intra-domain diameter, resulting in fluctuations in control quality and reliability.

[0005] Second, the decision-making indicators are too simplistic. Using only geometric distance or topological centrality to select domain management nodes can easily lead to controller selection that is "topologically centered but geometrically biased" or "geometrically centered but topologically unstable," resulting in an increased management radius within the domain, a higher number of hops for forwarding control messages, and ultimately, increased management overhead.

[0006] Third, there is a lack of explicit management overhead modeling and constraint consistency. Many solutions focus on connectivity or latency, but lack calculable models for "intra-domain / inter-domain control message volume, controller management radius, and inter-domain coordination cost," making it difficult to integrate "reliability improvement" and "controllable overhead" into the same decision-making framework.

[0007] Fourth, distributed systems lack scalability. Centralized solutions face computational and communication bottlenecks as the number of satellites increases; purely distributed solutions lack stability if they lack mechanisms to suppress oscillations (such as frequent migrations). Especially with the trend of ultra-large-scale constellations, there is a lack of dynamic domain partitioning mechanisms that balance scalability and stability.

[0008] Fifth, there is insufficient understanding of "local structures within the domain". Even when node importance assessment is introduced, a global perspective is often used, ignoring the local geometric center and local topology within the domain. This leads to inconsistencies between local decisions and global goals, manifesting as non-convergence of migrations or limited improvement in benefits. Summary of the Invention

[0009] The purpose of this invention is to provide a constellation dynamic domain-based satellite-to-ground management method based on node importance assessment and distributed state machine driving, which can be executed in a distributed manner, can be adaptive in real time, and can explicitly constrain management overhead and reliability. This allows the system to automatically maintain the feasibility of domain division (domain capacity, connectivity, diameter / radius, etc.) when the inter-satellite link topology changes over time, and significantly reduce management overhead and performance fluctuations caused by unstable controller selection while ensuring network reliability.

[0010] To achieve the above objectives, this invention constructs a comprehensive method chain of "node importance assessment—initial domain division—dynamic adjustment triggered by state machine (P1 / P2)—selection of integrated domain management nodes—reliability / overhead closed-loop assessment," with the core advantages being:

[0011] At the indicator level, topological and geometric indicators are unified into the AHP+TOPSIS importance assessment framework, and the "intra-domain geometric center caliber" is introduced to make the importance scores more aligned with local domain management decisions. Specifically, "unifying into the AHP+TOPSIS framework" means first constructing an evaluation matrix from multiple indicators, including topological and geometric ones; then determining the relative importance (weight) of each indicator using the Analytic Hierarchy Process (AHP); and finally, using the TOPSIS ranking method, which approximates the ideal solution, to synthesize the multiple indicators into a single proximity score. The "intra-domain geometric center caliber" means that geometric centrality is no longer referenced to the entire constellation center, but rather to the current domain D. k The average position of the internal members is used as a reference, which is more in line with the optimization goals of the coverage radius and forwarding hops of the domain management node (CH) within the domain.

[0012] At the mechanism level, a distributed node state machine (idle / busy) is used to trigger domain adjustments to suppress frequent oscillations while allowing for rapid repair when constraints fail.

[0013] At the controller selection level, a fusion of AHP, geometry, and topological centrality for domain management node selection is proposed, and a "geometric fallback threshold" is set to reduce the cost of management radius.

[0014] At the evaluation level, a computable management overhead decomposition model (comprising three parts: intra-domain, inter-domain, and management radius) and analytical approximation reliability are constructed, enabling the benefit target to be calculated online, thereby supporting real-time or near-real-time optimization.

[0015] The technical solution of this invention is: a constellation dynamic domain-based satellite-to-ground management method driven by node importance assessment and distributed state machine, the specific steps of which are as follows:

[0016] Step S1: Contrast time-series topology construction;

[0017] Step S2: Node importance assessment based on AHP+TOPSIS;

[0018] (1) Calculate the multi-index feature vector for each node: degree centrality Close centrality Second-order neighbor centrality Geometric centrality The second-order neighbor centrality can be expressed as:

[0019] ;

[0020] in, Represents the set of one-hop neighbors of node j; Indicates the degree of node n; This represents the sum of the degrees of node j's neighbors; Represents the set of one-hop neighbors of node i; To represent the second-order neighbor centrality (local centrality) of node i: first, calculate the neighbor degree and Q(j) of each neighbor j, and then sum the results for all the neighbors of i to characterize the connectivity potential of i within a two-hop range.

[0021] (2) To avoid the influence of dimensions, each indicator is normalized by row to unify all indicators. Interval; the original value of the m-th index at node i. The normalization result is:

[0022] ;

[0023] in, and These are the minimum and maximum values ​​of the indicator across all nodes, respectively. The value is a very small positive number to avoid a denominator of 0; m is the index number, m=1…M; i is the node number; This represents the original value of node i on index m;

[0024] (3) Construct the judgment matrix ;in, Indicators Relative indicators The importance of the value (assigned on a scale of 1–9) is assessed, and the judgment is made reasonable through a consistency test (e.g., CR < 0.1); by Obtain the weight vector ,satisfy and Weighted matrix part: for the normalized... We perform weighted average to obtain ; then adopt Calculate proximity: by Constructing the ideal solution With negative ideal solution Calculate the distance from node i to the positive / negative ideal solution. , Proximity , The larger the value, the higher the overall importance of the node;

[0025] ;

[0026] ;

[0027] ;

[0028] (4) When the domain (or candidate domain) is known, for each domain Geometric center within the computational domain Geometric centrality is obtained by normalizing the distance within the domain:

[0029] ;

[0030] in, Let k be the set of nodes in the k-th domain. This indicates that node i belongs to the domain. ; The position vector of node i (can be ECEF / ECI coordinates); For domain The geometric center (centroid) can be taken as ; Let i be the distance from node i to the center of the domain. and These represent the minimum and maximum distances from the center within the domain, respectively. To ensure geometric centrality within the domain, a "1-normalized distance" approach is used, where nodes closer to the domain center have a higher geometric centrality. The larger; To prevent extremely small constants with a denominator of 0;

[0031] This design makes "node importance" more aligned with domain controller selection and domain migration decisions.

[0032] Step S3: Initial domain expansion based on importance seed + constraint;

[0033] According to proximity Select from high to low Several nodes are used as seeds for the domain management node, i.e., CH candidates, and a minimum hop count interval constraint is applied (to avoid uneven domain distribution caused by seeds being too close together). This can be determined based on the total number of nodes N in the constellation and the maximum capacity of a single domain. Determine, for example, take ,in The function is used for rounding up; then, the remaining nodes are attempted to be mounted according to their shortest path distance to the seed node from nearest to farthest, while simultaneously checking constraints;

[0034] Step S4: Dynamic domain adjustment triggered by the distributed state machine;

[0035] Set a state machine for each node ; State continues Steps are used for information gathering, and nodes are in Within each discrete time step, only status / link / domain information is broadcast and statistics are performed; no migration decision is initiated. When the timer expires, the node determines its migration based on the current status. The mapping and subgraph checks whether constraints such as capacity, connectivity, and radius are still satisfied; that is, the mapping and subgraph are used to check whether these constraints are still satisfied. "Idle window length (information collection window)" is measured in discrete time steps (steps) of the algorithm; "satisfies constraints" refers to capacity. Domain subgraph connectivity, and Both conditions must be met simultaneously. Upon expiration, determine if the domain satisfies the constraints.

[0036] Step S5: Check the consistency between the revenue / value function and constraints;

[0037] (1) Construct the profit function Unify reliability and overhead:

[0038]

[0039] in, This is a trade-off factor between reliability and overhead. For the average reliability within the domain, To normalize management overhead; improving reliability will increase Increasing expenses will reduce Domain adjustment operations are considered feasible only if the constraints (capacity, connectivity, diameter / radius) are met; the difference between the payoff of a candidate action and the current payoff is denoted as... ,when When, it means that a certain candidate action can bring a positive incremental return, in all cases where the constraints are satisfied and From the candidates, select Execute the maximum possible action to ensure that every adjustment converges towards the direction of better returns;

[0040] (2) When a domain does not meet the constraints, it is not randomly split or blindly stripped, but the local importance score is calculated within the current domain subgraph, and candidate nodes are sorted according to their low importance. Low-scoring nodes are moved / stripped first, so that the repair has less impact on business-critical nodes and it is easier to restore the feasible domain structure with a limited number of migrations; among them, the importance criterion adopts the AHP+TOPSIS comprehensive score C in step S2. i Calculate: Recalculate proximity only within the current domain subgraph / geometric center. When prioritizing low importance, in the domain to be repaired... Internal calculation of all candidate nodes Sort in ascending order and prioritize migration / stripping. Smaller nodes; if a set of business protection nodes exists You can set a restriction on migration or impose a penalty in the sorting to further reduce the probability of critical nodes being migrated;

[0041] Step S6: Selection of converged domain management node (CH) and calculation of management and control overhead.

[0042] Furthermore, in step S1, the constellation temporal topology is constructed as follows:

[0043] Obtaining constellations at discrete moments satellite position Based on the inter-satellite distance threshold Construct an undirected weighted graph:

[0044] ;

[0045] in: Represents a discrete time step; Indicates time A topological diagram of constellations; This represents a set of nodes, i.e., a set of satellites. ; Indicates time The set of edges can be represented by the set of inter-satellite links; Represents the edge weight function; Indicates the time intervals of nodes i and j. There is a link; , They represent the times of nodes i and j respectively. The position vector; It is the Euclidean norm; Indicates the inter-satellite distance threshold; Equivalence relation: The distance between two stars is less than a threshold if and only if the distance between the two stars is less than a threshold. When the graph is in the same position, connect an undirected edge.

[0046] Furthermore, the constraints in the initial domain of the importance seed + constraint expansion in step S3 are as follows:

[0047] Domain capacity constraints: ;in, Let k be the set of nodes in the k-th domain. This represents the number of nodes within the domain. The maximum manageable node capacity for a single domain (given by on-board computation / control message budget, etc.); constraints Ensure that the domain size does not exceed the manageable limit;

[0048] Intra-domain connectivity constraints: Subgraph connectivity;

[0049] Intra-domain diameter / radius constraint: The shortest path distance to CH does not exceed That is, the connection between any node within the domain and the domain management node. Shortest path hop count Must meet , This represents the maximum allowed management radius / maximum number of hops within the domain, used to limit and control message forwarding latency and management overhead.

[0050] Furthermore, in the dynamic domain adjustment triggered by the distributed state machine in step S4, the determination of whether the domain satisfies the constraints after the expiration is as follows:

[0051] If the conditions are met, the P1 adjustment algorithm (benefit-enhancing migration) is triggered. Under the premise that the domain structure is feasible (constraints are met), a small number of candidate migration actions are enumerated (such as moving node i into an adjacent domain or adjusting the CH), and the profit increment before and after the migration is calculated. ,like If the constraints are still satisfied after migration, it is considered a candidate; finally, execution is performed. The biggest move is to gradually increase overall revenue;

[0052] If the conditions are not met, the P2 repair algorithm (constraint repair / stripping / merging) is triggered. When any constraint failure such as domain capacity / connectivity / radius is detected, repair actions are performed first (stripping nodes, migrating into neighboring domains, merging domains / reselecting CHs if necessary). The goal is to restore feasible domains as quickly as possible within a limited number of operations and prevent infeasible domains from continuing to deteriorate.

[0053] After being triggered, it enters a busy state and continues. To prevent oscillations caused by repeated migrations within a short period, nodes enter a cooldown period after performing a migration / repair: no new migration decisions are triggered within Tbusy time steps, only domain information is synchronized. This mechanism is used to suppress short-cycle repeated migrations (oscillations) and control plane jitter.

[0054] Furthermore, the selection of the converged domain management node (CH) and the calculation of management overhead in step S6 are as follows:

[0055] (1) For each domain Calculate the fusion score of the candidate nodes and select CH:

[0056] ;

[0057] in, , To integrate the weighting coefficients, , ,and ; Node importance score (AHP + TOPSIS proximity); Geometric centrality within the domain; For topological centrality within the domain (such as closeness or betweenness, which can be selected according to the implementation); Indicates only in the domain Select CH from within;

[0058] (2) If the geometric score of the selected node is lower than the threshold , For geometric catch-all threshold, If the candidate nodes are selected based on the fusion score, their Below (If its geometric location deviates from the domain center, which may lead to an excessively large management radius), then the topology center node within the domain is selected as the CH in the fallback to ensure that the coverage radius and cost do not reach extreme values.

[0059] (3) Management expenses adopt a decomposable model:

[0060] ;

[0061] in, This indicates that within a time period, in order to maintain the "domain-cluster head" ( — The additional communication / computing / coordination costs incurred by the "member node" control structure to ensure its normal operation;

[0062] This represents the update / maintenance overhead of the cluster head / control node; it mainly corresponds to the costs of cluster head status broadcasting, heartbeat / reporting between the cluster head and the ground (or upper-level domain controller), and notification and confirmation when cluster head switching (reselection) is triggered.

[0063] This represents intra-domain coordination overhead; it reflects the cost incurred between member nodes and the cluster head within the same domain for task distribution, state synchronization, and resource coordination.

[0064] This represents inter-domain coordination overhead. It corresponds to the cost incurred between cluster heads (or domain controllers) in different domains to maintain global consistency.

[0065] After the above decomposition, any time a node is set as Operations that adjust the domain of a node can all be quantified under the same caliber. , , This allows for the comparison of management costs among different domain selection strategies / cluster head selection strategies, and provides a unified measurement basis for subsequent benefit-cost trade-offs.

[0066] The beneficial effects of this invention are as follows: Structurally, this invention is not a single algorithm point, but a closed-loop chain of domain division—evaluation—re-domain division, with clear data flow and dependencies: the "track / position sequence" on the input side first enters the topology construction module to form... At each moment, the importance assessment module starts from... Read the graph structure and node coordinates, and output the importance score. (Periodic recalculation to reduce computational load). The initial domain segmentation module will... The initial node_domain mapping is established as the basis for seed selection; this mapping serves as a shared data structure for the state machine module, is referenced by all node state machines, and is updated when P1 / P2 is executed.

[0067] The state machine module is structurally organized as "one state machine instance per node + shared domain mapping": each state machine triggers P1 or P2 according to its own timer rhythm, and shares the domain mapping. Atomic updates (migration / stripping) are performed to progressively advance domain structure convergence under distributed conditions. The domain management node selection module does not directly change the domains; instead, it calculates the CH (Choice of Domains) based on the current domain membership set during each evaluation or when management overhead needs to be calculated. Its output serves as the basis for the overhead model. The central point input affects the profit calculation and the next round of migration decisions.

[0068] Finally, the evaluation module calculates the reliability for each domain subgraph. With expenses The results are fed back to the "benefit / value assessment" logic for candidate action selection in the domain-specific algorithm. This structured connection ensures that changes in domain structure alter CH selection and overhead; CH selection and overhead change revenue; and revenue, in turn, influences the next round of domain adjustment, thus forming a closed-loop adaptive mechanism.

[0069] Advantage 1: The importance assessment structure that integrates multiple indicators significantly improves the robustness of domain management decisions;

[0070] This invention combines topological and geometric indicators in a three-part structure of "indicator matrix - AHP weight - TOPSIS proximity," avoiding bias caused by single indicators. Importance scores are used not only for initial CH seed selection but also for candidate ranking during domain-specific algorithm repair, ensuring consistent decision-making across the entire process.

[0071] Advantage 2: The geometric center aperture within the domain ensures that local decisions are consistent with the domain structure, reducing "local optimum conflicts";

[0072] Geometric centrality no longer depends on the global graph center, but rather on the set of members of the current domain and the domain's central point; therefore, the geometric score of the same node in different domains will change with the domain structure. This allows the operation of "migrating a node into a domain" to reflect its impact on domain centrality during the evaluation phase, which is conducive to forming a more stable domain structure and suppressing ineffective migrations.

[0073] Advantage 3: The distributed state machine (idle / busy) structure suppresses oscillations and improves scalability and engineering feasibility;

[0074] The state machine discretizes decision triggers and avoids repeated short-cycle migrations through busy windows. As the number of satellites increases, the domain structure can be gradually repaired / improved without centralized global solutions, while keeping control surface jitter caused by frequent migrations within acceptable limits.

[0075] Advantage 4: The converged domain manager CH selection and geometry fallback significantly reduce the cost of management radius and avoid extreme point selection;

[0076] The domain manager (CH) adopts a two-tier architecture of "linear fusion + threshold backoff": first, the optimal controller is selected based on the fusion score; then, a geometric threshold is used for security verification, and a backoff to the topology center is performed if necessary. This utilizes importance information while avoiding the risk of selecting a controller at the domain edge. This surge reduces overall overhead and improves explainability. Detailed Implementation

[0077] This application provides a constellation dynamic domain-based satellite-to-ground management method based on node importance assessment and distributed state machine driven by the following steps:

[0078] Step S1: Contrast time-series topology construction;

[0079] Obtaining constellations at discrete moments satellite position Based on the inter-satellite distance threshold Construct an undirected weighted graph:

[0080] ;

[0081] in: Represents a discrete time step; Indicates time A topological diagram of constellations; This represents a set of nodes, i.e., a set of satellites. ; Indicates time The set of edges can be represented by the set of inter-satellite links; Represents the edge weight function; Indicates the time intervals of nodes i and j. There is a link; , They represent the times of nodes i and j respectively. The position vector; It is the Euclidean norm; Indicates the inter-satellite distance threshold; Equivalence relation: The distance between two stars is less than a threshold if and only if the distance between the two stars is less than a threshold. When the graph is in the same position, connect an undirected edge.

[0082] Step S2: Node importance assessment based on AHP+TOPSIS;

[0083] (1) Calculate the multi-index feature vector for each node: degree centrality Close centrality Second-order neighbor centrality Geometric centrality The second-order neighbor centrality can be expressed as:

[0084] ;

[0085] in, Represents the set of one-hop neighbors of node j; Indicates the degree of node n; This represents the sum of the degrees of node j's neighbors; Represents the set of one-hop neighbors of node i; To represent the second-order neighbor centrality (local centrality) of node i: first, calculate the neighbor degree and Q(j) of each neighbor j, and then sum the results for all the neighbors of i to characterize the connectivity potential of i within a two-hop range.

[0086] (2) To avoid the influence of dimensions, each indicator is normalized by row to unify all indicators. Interval; the original value of the m-th index at node i. The normalization result is:

[0087] ;

[0088] in, and These are the minimum and maximum values ​​of the indicator across all nodes, respectively. The value is a very small positive number to avoid a denominator of 0; m is the index number, m=1…M; i is the node number; This represents the original value of node i on index m;

[0089] (3) Construct the judgment matrix ;in, Indicators Relative indicators The importance of the value (assigned on a scale of 1–9) is assessed, and the judgment is made reasonable through a consistency test (e.g., CR < 0.1); by Obtain the weight vector ,satisfy and Weighted matrix part: for the normalized... We perform weighted average to obtain ; then adopt Calculate proximity: by Constructing the ideal solution With negative ideal solution Calculate the distance from node i to the positive / negative ideal solution. , Proximity , The larger the value, the higher the overall importance of the node;

[0090] ;

[0091] ;

[0092] ;

[0093] (4) Core point A: the diameter of the geometric center within the domain Unlike the uniform approach of "geometric center of the entire graph," this invention, when the domain (or candidate domain) is known, performs a specific analysis on each domain. Geometric center within the computational domain Geometric centrality is obtained by normalizing the distance within the domain:

[0094] ;

[0095] in, Let k be the set of nodes in the k-th domain. This indicates that node i belongs to the domain. ; The position vector of node i (can be ECEF / ECI coordinates); For domain The geometric center (centroid) can be taken as ; Let i be the distance from node i to the center of the domain. and These represent the minimum and maximum distances from the center within the domain, respectively. To ensure geometric centrality within the domain, a "1-normalized distance" approach is used, where nodes closer to the domain center have a higher geometric centrality. The larger; To prevent extremely small constants with a denominator of 0;

[0096] This design makes "node importance" more aligned with domain controller selection and domain migration decisions.

[0097] Step S3: Initial domain expansion based on importance seed + constraint;

[0098] According to proximity Select from high to low Several nodes are used as seeds for the domain management node, i.e., CH candidates, and a minimum hop count interval constraint is applied (to avoid uneven domain distribution caused by seeds being too close together). This can be determined based on the total number of nodes N in the constellation and the maximum capacity of a single domain. Determine, for example, take , among which This is a rounding function; then, the remaining nodes are attempted to be mounted in order of shortest path distance to the seed node, from nearest to farthest, while simultaneously checking constraints:

[0099] Domain capacity constraints: ;in, Let k be the set of nodes in the k-th domain. This represents the number of nodes within the domain. The maximum manageable node capacity for a single domain (given by on-board computation / control message budget, etc.); constraints Ensure that the domain size does not exceed the manageable limit;

[0100] Intra-domain connectivity constraints: Subgraph connectivity;

[0101] Intra-domain diameter / radius constraint: The shortest path distance to CH does not exceed That is, the connection between any node within the domain and the domain management node. Shortest path hop count Must meet , This represents the maximum allowed management radius / maximum number of hops within the domain, used to limit and control message forwarding latency and management overhead.

[0102] Step S4: Dynamic domain adjustment triggered by the distributed state machine;

[0103] Set a state machine for each node ; State continues Steps are used for information gathering, and nodes are in Within each discrete time step, only status / link / domain information is broadcast and statistics are performed; no migration decision is initiated. When the timer expires, the node determines its migration based on the current status. The mapping and subgraph checks whether constraints such as capacity, connectivity, and radius are still satisfied; that is, the mapping and subgraph are used to check whether these constraints are still satisfied. "Idle window length (information collection window)" is measured in discrete time steps (steps) of the algorithm; "satisfies constraints" refers to capacity. Domain subgraph connectivity, and Both conditions must be met simultaneously. Upon expiration, determine if the domain satisfies the constraints.

[0104] If the conditions are met, the P1 adjustment algorithm (benefit-enhancing migration) is triggered. Under the premise that the domain structure is feasible (constraints are met), a small number of candidate migration actions are enumerated (such as moving node i into an adjacent domain or adjusting the CH), and the profit increment before and after the migration is calculated. ,like If the constraints are still satisfied after migration, it is considered a candidate; finally, execution is performed. The biggest move is to gradually increase overall revenue;

[0105] If the conditions are not met, the P2 repair algorithm (constraint repair / stripping / merging) is triggered. When any constraint failure such as domain capacity / connectivity / radius is detected, repair actions are performed first (stripping nodes, migrating into neighboring domains, merging domains / reselecting CHs if necessary). The goal is to restore feasible domains as quickly as possible within a limited number of operations and prevent infeasible domains from continuing to deteriorate.

[0106] After being triggered, it enters a busy state and continues. To prevent oscillations caused by repeated migrations within a short period, nodes enter a cooldown period after performing a migration / repair: no new migration decisions are triggered within Tbusy time steps, only domain information is synchronized. This mechanism is used to suppress short-cycle repeated migrations (oscillations) and control plane jitter.

[0107] Step S5: Check the consistency between the revenue / value function and constraints;

[0108] (1) Construct the profit function Unify reliability and overhead:

[0109]

[0110] in, This is a trade-off factor between reliability and overhead. For the average reliability within the domain, To normalize management overhead; improving reliability will increase Increasing expenses will reduce Domain adjustment operations are considered feasible only if the constraints (capacity, connectivity, diameter / radius) are met; the difference between the payoff of a candidate action and the current payoff is denoted as... ,when When, it means that a certain candidate action can bring a positive incremental return, in all cases where the constraints are satisfied and From the candidates, select Execute the maximum possible action to ensure that every adjustment converges towards the direction of better returns;

[0111] (2) Core point B: The "low AHP score priority" repair strategy in P2; when a domain does not meet the constraints, it is not randomly split or blindly stripped, but the local importance score is calculated within the current domain subgraph, and candidate nodes are sorted according to low importance priority, and low-scoring nodes are moved / stripped first, so that the repair has less impact on business-critical nodes and it is easier to restore the feasible domain structure with a limited number of migrations; among them, the importance standard adopts the AHP+TOPSIS comprehensive score C in step S2. i Calculate: Recalculate proximity only within the current domain subgraph / geometric center. When prioritizing low importance, in the domain to be repaired... Internal calculation of all candidate nodes Sort in ascending order and prioritize migration / stripping. Smaller nodes; if a set of business protection nodes exists You can set a restriction on migration or impose a penalty in the sorting to further reduce the probability of critical nodes being migrated;

[0112] Step S6: Selection of converged domain management node (CH) and calculation of management and control overhead;

[0113] (1) For each domain Calculate the fusion score of the candidate nodes and select CH:

[0114] ;

[0115] in, , To integrate the weighting coefficients, , ,and ; Node importance score (AHP + TOPSIS proximity); Geometric centrality within the domain; For topological centrality within the domain (such as closeness or betweenness, which can be selected according to the implementation); Indicates only in the domain Select CH from within;

[0116] (2) If the geometric score of the selected node is lower than the threshold , For geometric catch-all threshold, If the candidate nodes are selected based on the fusion score, their Below (If its geometric location deviates from the domain center, which may lead to an excessively large management radius), then the topology center node within the domain is selected as the CH in the fallback to ensure that the coverage radius and cost do not reach extreme values.

[0117] (3) Management expenses adopt a decomposable model:

[0118] ;

[0119] in, This indicates that within a time period, in order to maintain the "domain-cluster head" ( — The additional communication / computing / coordination costs incurred by the "member node" control structure to ensure its normal operation;

[0120] This represents the update / maintenance overhead of the cluster head / controller node; it mainly corresponds to the costs of: cluster head status broadcasting, heartbeats / reporting between the cluster head and the ground (or upper-level domain controller), and notifications and confirmations triggered when a cluster head switchover (re-election) is initiated. Intuitively, when... When the location changes rapidly, cluster head reselection is frequent, or the reporting frequency increases, This will increase significantly. In actual statistics, it can be aggregated by "number of updated messages × message size × number of transmission hops (or latency weight)", and if necessary, the calculation time on the cluster head side (such as importance recalculation, member table refresh) can be added.

[0121] This represents intra-domain coordination overhead; it reflects the cost incurred between member nodes and the cluster head within the same domain for task distribution, state synchronization, and resource coordination. For example, intra-domain state collection (member → ), task assignment ( →Members), intra-domain conflict resolution / queue adjustment, etc. This item is strongly correlated with the intra-domain size and the "diameter / average hop count" of the intra-domain topology: the larger the domain, the longer the intra-domain communication path, or the more unstable the intra-domain link, the higher the frequency / greater cost of intra-domain synchronization, leading to... Growth. In statistical terms, the cumulative transmission cost of control messages within the domain on the domain edge set can usually be taken (e.g., weighted by hop count, latency, or energy consumption).

[0122] This represents inter-domain coordination overhead. It corresponds to the cost incurred between cluster heads (or domain controllers) in different domains to maintain global consistency. Examples include: cross-domain task handover, cross-domain link / load information exchange, and negotiation and confirmation during domain boundary adjustments. This item is typically related to the number of cross-domain connections (inter-domain edges), cross-domain path length, and inter-domain coordination frequency; when domain partitioning is more fragmented and inter-domain interactions are more frequent... This will become the dominant factor. Statistics can be calculated based on the cumulative cost of cross-domain messages at cross-domain edges, or further refined using "number of cross-domain events × number of single negotiation rounds × cost of a single round of messages";

[0123] After the above decomposition, any time a node is set as Operations that adjust the domain of a node can all be quantified under the same caliber. , , This allows for the comparison of management costs among different domain selection strategies / cluster head selection strategies, and provides a unified measurement basis for subsequent benefit-cost trade-offs.

[0124] Example 1

[0125] Scenarios with highly dynamic topology but still containing locally connected components.

[0126] When the inter-satellite link ISL changes frequently over time, but... When a stable local connectivity structure can still be formed within the range, the P1 "benefit-enhancing migration" and fusion CH selection of the present invention can play the most effective role: the domain structure can be gradually adjusted with topology changes, avoiding large-scale migration and control surface oscillations caused by centralized repartitioning.

[0127] Example 2

[0128] Scenarios that are sensitive to management costs and require explicit control over the control surface.

[0129] For example, when satellite-to-ground link resources are limited or onboard computing / communication budgets are tight. The three-component model can be directly used as a design constraint: by adjusting... , With geometric catch-all threshold and domain capacity Parameters such as inter-domain coordination costs enable the system to achieve an interpretable and controllable trade-off between reliability and overhead.

[0130] Example 3

[0131] Scenarios that need protection at critical mission nodes

[0132] When some nodes bear higher business weight (such as gateway stars, aggregation stars, and key measurement and control resource stars), the "low AHP score priority within the domain" repair mechanism in the domain division algorithm can prioritize the movement of low-importance nodes and reduce the disturbance to key nodes.

[0133] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A constellation dynamic domain-based satellite-to-ground control method based on node importance assessment and distributed state machine driven, characterized in that, The specific steps are as follows: Step S1: Contrast time-series topology construction; Step S2: Node importance assessment based on AHP+TOPSIS; (1) Calculate the multi-index feature vector for each node: degree centrality Close centrality Second-order neighbor centrality Geometric centrality The second-order neighbor centrality can be expressed as: ； in, Represents the set of one-hop neighbors of node j; Indicate the degree of node n; This represents the sum of the degrees of node j's neighbors; Represents the set of one-hop neighbors of node i; To represent the second-order neighbor centrality of node i: first, calculate the neighbor degree of each neighbor j and the sum of Q(j), then sum over all the neighbors of i to characterize the connectivity potential of i within a two-hop range; (2) Normalize each indicator by row to unify all indicators. Interval; the original value of the m-th index at node i. The normalization result is: ； in, and These are the minimum and maximum values ​​of the indicator across all nodes, respectively. The value is a very small positive number to avoid a denominator of 0; m is the index number, m=1…M; i is the node number; Let be the original value of node i on index m; (3) Construct the judgment matrix ;in, Indicators Relative indicators The importance of this, and ensuring the rationality of the judgment through consistency checks; by Obtain the weight vector ,satisfy and Weighted matrix part: for the normalized... We perform weighted average to obtain ; then adopt Calculate proximity: by Constructing the ideal solution With negative ideal solution Calculate the distance from node i to the positive / negative ideal solution. , Proximity , The larger the value, the higher the overall importance of the node; ； ； ； (4) When the domains are known, for each domain Geometric center within the computational domain Geometric centrality is obtained by normalizing the distance within the domain: ； in, Let k be the set of nodes in the k-th domain. This indicates that node i belongs to the domain. ; Let i be the position vector of node i; For domain The geometric center can be taken as ; Let i be the distance from node i to the center of the domain; and These represent the minimum and maximum distances from the center within the domain, respectively. To ensure geometric centrality within the domain, a "1-normalized distance" approach is used, where nodes closer to the domain center have a higher geometric centrality. The larger; To prevent extremely small constants with a denominator of 0; Step S3: Initial domain expansion based on importance seed + constraint; According to proximity Select from high to low One node is used as the seed node for domain management, i.e., CH candidate, and a minimum hop count interval constraint is applied; then the remaining nodes are attempted to be mounted in order of the shortest path distance to the seed node from nearest to farthest, while checking the constraints. Step S4: Dynamic domain adjustment triggered by the distributed state machine; Set a state machine for each node ; State continues Steps are used for information gathering, and nodes are in Within each discrete time step, only status / link / domain information is broadcast and statistics are performed; no migration decision is initiated. When the timer expires, the node determines its migration based on the current status. The mapping and domain subgraph are checked to see if the constraints are still satisfied; after the expiration date, it is determined whether the domain satisfies the constraints. Step S5: Check the consistency between the revenue / value function and constraints; (1) Construct the profit function Unify reliability and overhead: in, This is a trade-off factor between reliability and overhead. For the average reliability within the domain, To normalize management overhead; improving reliability will increase Increasing expenses will reduce Domain adjustment operations are considered feasible only if the constraints are met; the difference between the payoff of a candidate action and the current payoff is denoted as... ,when When, it means that a certain candidate action can bring a positive incremental return, in all cases where the constraints are satisfied and From the candidates, select Execute the maximum possible action to ensure that every adjustment converges towards the direction of better returns; (2) When a domain does not meet the constraints, calculate the local importance score within the current domain subgraph, prioritize candidate nodes according to their low importance, and prioritize moving / removing low-scoring nodes to minimize the impact of repairs on critical business nodes and make it easier to restore the feasible domain structure with a limited number of migrations; whereby the importance criterion adopts the AHP+TOPSIS comprehensive score C from step S2. i Calculate: Recalculate proximity only under the current domain subgraph / geometric center within the domain When prioritizing low importance, in the domain to be repaired... Internal calculation of all candidate nodes Sort in ascending order and prioritize migration / stripping. Smaller nodes; if a set of business protection nodes exists You can set a restriction on migration or impose a penalty in the sorting to further reduce the probability of critical nodes being migrated; Step S6: Selection of converged domain management node (CH) and calculation of management and control overhead.

2. The constellation dynamic domain-based satellite-to-ground control method based on node importance assessment and distributed state machine driving as described in claim 1, characterized in that: In step S1, the constellation time-series topology is constructed as follows: Obtaining constellations at discrete moments satellite position Based on the inter-satellite distance threshold Construct an undirected weighted graph: ； in: Represents a discrete time step; Indicates time A topological diagram of constellations; This represents a set of nodes, i.e., a set of satellites. ; Indicates time The set of edges can be represented by the set of inter-satellite links; Represents the edge weight function; Indicates the time intervals of nodes i and j. A link exists; , They represent the times of nodes i and j respectively. The position vector; It is the Euclidean norm; Indicates the inter-satellite distance threshold; Equivalence relation: The distance between two stars is less than a threshold if and only if the distance between the two stars is less than a threshold. When, connect an undirected edge in the graph.

3. The constellation dynamic domain-based satellite-to-ground control method based on node importance assessment and distributed state machine driving as described in claim 1, characterized in that: The constraints in the initial domain of the importance seed and constraint expansion step S3 are as follows: Domain capacity constraints: ;in, Let k be the set of nodes in the k-th domain. This represents the number of nodes within the domain. The maximum manageable node capacity for a single domain; constraints Ensure that the domain size does not exceed the manageable limit; Intra-domain connectivity constraints: Subgraph connectivity; Intra-domain diameter / radius constraint: The shortest path distance to CH does not exceed That is, the connection between any node within the domain and the domain management node. Shortest path hop count Must meet , This represents the maximum allowed management radius / maximum number of hops within the domain, used to limit and control message forwarding latency and management overhead.

4. The constellation dynamic domain-based satellite-to-ground control method based on node importance assessment and distributed state machine driving according to claim 3, characterized in that: In the dynamic domain adjustment triggered by the distributed state machine in step S4, the determination of whether the domain meets the constraints after the expiration is as follows: If the conditions are met, the P1 adjustment algorithm is triggered; provided the domain structure is feasible, a small number of candidate migration actions are enumerated, and the incremental revenue before and after the migration is calculated. ,like If the constraints are still satisfied after migration, it is considered a candidate; finally, execution is performed. The biggest move is to gradually increase overall revenue; If the conditions are not met, the P2 repair algorithm is triggered. When any constraint failure of domain capacity / connectivity / radius is detected, the repair action is executed first. The goal is to restore the feasible domain as soon as possible within a limited number of operations and prevent the infeasible domain from continuing to deteriorate. After being triggered, it enters a busy state and continues. To avoid repeated migrations causing oscillations within a short period of time, nodes enter a cooldown period after performing a migration / repair: no new migration decisions are triggered within Tbusy time steps, and only domain information is synchronized.

5. The constellation dynamic domain-based satellite-to-ground control method based on node importance assessment and distributed state machine driven according to claim 1, characterized in that: The selection of the converged domain management node (CH) and the calculation of management overhead in step S6 are as follows: (1) For each domain Calculate the fusion score of the candidate nodes and select CH: ； in, , To integrate the weighting coefficients, , ,and ; Score the importance of nodes; Geometric centrality within the domain; It is topologically central to the domain; Indicates only in the domain Select CH from within; (2) If the geometric score of the selected node is lower than the threshold , For geometric catch-all threshold, If the candidate nodes are selected based on the fusion score, their Below If so, the topology center node within the domain is selected as the CH in the fallback to ensure that the coverage radius and cost do not reach extreme values; (3) Management expenses adopt a decomposable model: ； in, This represents the additional communication / computing / coordination costs incurred to maintain the normal operation of the "domain-cluster head-member node" control structure within a given time period. This indicates the overhead of updating and maintaining the cluster head / controlling node; Indicates intra-domain coordination costs; This indicates the cost of cross-domain coordination.

Images