Multi-layer dimension reduction star group configuration method for giant satellite constellation task planning

Abstract

The invention discloses a multi-layer dimension reduction star group configuration method for giant satellite constellation task planning, relates to the technical field of aerospace task scheduling, and comprises complex task multi-dimensional characterization, a multi-domain collaborative task scene clustering and star group resource clustering algorithm, and a multi-domain collaborative pre-distribution mechanism and star group parallel conflict reduction method. Through task and satellite bidirectional dimension reduction clustering, pre-distribution screening and conflict reduction, the calculation complexity is significantly reduced, and the problems of dimension explosion and multi-solution oscillation are solved. According to the method, efficient and stable resource allocation under the scene of ten-thousand-level satellites and massive task requests can be realized, and a systematic solution is provided for aerospace task scheduling.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of space mission scheduling, and particularly relates to a multi-layer dimension reduction star cluster configuration method for mega-constellation mission planning. BACKGROUND

[0002] With the rapid development of satellite technology, mega-constellation has become the mainstream development direction of space systems, and future space missions will show significant characteristics of large-scale and high concurrency. Under this background, an innovative method of domain management is adopted, which divides the mega-constellation into multiple logical star domains, and realizes the centralized management and distributed processing of data in each star domain through the domain head node, which can significantly improve the overall management efficiency and operation performance of the mega-constellation system. However, after the division of the domain, the mega-constellation task scheduling still faces multiple challenges such as high-dimensional attributes, super-large-scale concurrency and strong cross-domain coupling. Specifically, the task scheduling needs to consider the coupling conditions of multiple constraints such as load, orbit window, etc., and cope with ten-thousand-level satellites, massive task requests, and complex cross-domain requirements such as the cooperation of optical and SAR satellites for ocean monitoring. The traditional scheduling methods based on heuristic rules or mathematical programming have significant bottlenecks in solving these problems. First, the dimension explosion problem is prominent, and the decision space of task and resource matching increases exponentially with the number of satellites and the scale of tasks, resulting in too long solving time or even unable to complete within a limited time. Second, the problem of heterogeneous constraint coupling is serious, and the multi-dimensional resource requirements of tasks such as timeliness, accuracy and coverage range cause a sharp increase in constraint conflicts, making the global optimization efficiency low. In addition, due to the approximate satellite cluster of resources, a large number of equivalent solutions are easily generated in the pre-allocation stage, resulting in unstable allocation results and difficulty in meeting the needs of normal tasks. Although the existing technology can achieve certain results in local optimization, it cannot systematically break through the joint constraints of high-dimensional coupling and scale explosion, and a new solution is urgently needed.

[0003] In this context, the present application proposes a multi-layer dimension reduction star cluster configuration method for mega satellite constellation task planning, aiming to solve the above technical problems. In view of the solvability problem under the dimension disaster, the present application projects the super high-dimensional task attribute into a low-dimensional space through a resource constraint driven multi-star domain task collaborative clustering algorithm, divides the task group set according to the resource mutual exclusivity, and thus significantly reduces the problem complexity. In view of the coupling constraint disassembly problem, the present application disassembles the task group into independent subsets according to the resource mutual exclusivity, and completes satellite matching through single star domain resource scanning, converts the global NP-Hard problem into parallel sub-problems, and improves the calculation efficiency. In view of the multi-solution stability defect problem, the present application designs a task-oriented parallel conflict reduction algorithm, eliminates the pre-assignment multi-solution through technical means such as task priority based directed conflict graph cutting, configures a dynamically reconfigurable flexible star cluster service unit, and finally realizes the efficient and stable allocation of star domain resources in the super large scale and strong coupling task scene. This innovative scheme not only solves the key bottlenecks of traditional methods in dimension explosion, constraint coupling and multi-solution oscillation, but also provides solid technical support for the normalization and efficient execution of future space tasks. SUMMARY

[0004] The present application proposes a multi-layer dimension reduction star cluster configuration method for mega satellite constellation task planning in view of the dimension explosion, heterogeneous constraint coupling and multi-solution oscillation problems of existing space task scheduling technology in multi-star domain collaborative observation scene. The method realizes efficient and stable star domain resource allocation through complex task multi-dimensional representation, multi-domain collaborative task scene and star cluster resource clustering algorithm, and multi-domain collaborative pre-assignment mechanism and star cluster parallel conflict reduction method.

[0005] To achieve the above purpose, the present application is implemented according to the following technical scheme:

[0006] The multi-layer dimension reduction star cluster configuration method for mega satellite constellation task planning of the present application comprises the following steps:

[0007] S1: Multi-dimensional representation of complex tasks, extraction of key feature parameters of tasks and normalization processing, construction of a unified task description framework;

[0008] S2: Multi-star domain task collaborative clustering algorithm based on resource constraints, aggregation of large-scale concurrent tasks into high-cohesion task communities;

[0009] S3: According to the multi-dimensional similarity of satellite payload, resolution, time availability, etc., the star cluster resource clustering method based on multi-dimensional similarity measurement is adopted to divide the heterogeneous satellite cluster into candidate resource clusters matching the demand;

[0010] S4: Constructing the feature comparison interface of task group and resource cluster, realizing adaptive matching, and outputting the coarse-grained matching result;

[0011] S5: A pre-screening mechanism based on dichotomy is adopted to quickly configure suitable star clusters for the task;

[0012] S6: A star cluster elasticity index measurement method based on a special matrix is used to quantitatively evaluate coverage and redundancy;

[0013] S7: A parallel conflict reduction algorithm based on the weighted Hungarian algorithm is designed to efficiently configure the final response star cluster in response to the resource conflicts caused by highly approximate task requirements.

[0014] The complex task multi-dimensional representation in step S1 includes cooperative observation tasks, relay tracking tasks, and interactive verification tasks. The cooperative observation task extracts key feature parameters of the target area requirements, including target position, cooperative coverage requirement, payload requirement, resolution requirement, and completion timeliness requirement. The relay tracking task extracts target position information, target direction information, target speed information, payload requirement, and effective time requirement. The interactive verification task extracts target position, multi-payload requirement, minimum angle difference, maximum angle difference, and effective time requirement. The interference of scene-specific parameters on resource matching is eliminated through a segmented normalization strategy to construct a unified task description framework. In particular, different types of target areas (point target, rectangular area, circular area, and polygonal area) are extended in a grid manner to generate a standardized coverage matrix, and large task areas are corrected to improve calculation accuracy.

[0015] The multi-star domain task cooperative clustering algorithm based on resource constraints is proposed in step S2 to address the characteristics of a large number of concurrent tasks in a large-scale concurrent task scenario, long time span, and close coupling of multi-star domains. Based on multi-dimensional task features, combined with time overlap and resource demand similarity, the tasks are aggregated into a number of highly cohesive task communities under the premise of meeting the total resource quantity constraint of each star domain, thereby significantly reducing the task scheduling dimension.

[0016] The star cluster resource clustering method based on multi-dimensional similarity measurement is designed in step S3 to address the problem of a large number of heterogeneous satellite resources, a large number of platforms, and scattered resource availability periods. Specifically, the key indicators of satellite resources such as payload, resolution, and time availability are extracted, and an improved clustering feature tree clustering method with multiple levels is used to divide heterogeneous satellites into candidate star clusters that match the task requirements.

[0017] The feature comparison interface between the task group and the resource cluster in step S4 is to compare the task communities output by the cooperative clustering algorithm with the star clusters generated by the resource clustering method to achieve adaptive matching, thereby providing coarse-grained matching results for subsequent star cluster resource pre-allocation and conflict resolution algorithms. The pre-allocation mechanism quickly traverses all satellites and marks the task scenarios related to them to form a number of mutually exclusive candidate satellite groups, significantly compressing the solution space.

[0018] The step S5 based on the bisection method pre-screening mechanism: first, a coverage determination algorithm based on time window bisection method is adopted, the time dimension is taken as the main index, the candidate subsatellite point in the task time window is quickly located by bisection method; then the effective coverage matrix of the task and the redundancy and effective coverage rate are calculated. The bisection method based pre-screening mechanism quickly judges the effective coverage of the satellite, stores the candidate satellite and calculates the effective coverage matrix and coverage rate, and generates a preliminary allocation scheme in order of coverage rate. The coverage determination algorithm based on time window bisection method is proposed, the search range is recursively reduced, and the calculation complexity is significantly reduced. The redundancy index is introduced to quantitatively evaluate the coverage rate and resource utilization efficiency, which provides a basis for conflict reduction. The star cluster redundancy formula is defined to ensure appropriate repeated coverage to improve fault tolerance.

[0019] The step S6 is based on the star cluster coverage rate and the redundancy, and a star cluster elasticity index measurement method based on a special-shaped matrix is studied. The continuous geographic space is converted into a quantifiable analysis calculation unit through a regular grid discretization method, and the satellite observation point data is combined to effectively solve the low efficiency problem of the traditional coverage algorithm in complex terrain and dynamic environment. Through the coverage rate calculation of the discretized grid, the coverage efficiency of the satellite constellation can be dynamically evaluated, and the star cluster resource scheduling is guided.

[0020] The step S7 based on the weighted Hungarian algorithm parallel conflict reduction algorithm: first, a weighted bipartite graph is constructed to depict the resource competition relationship between the satellite clusters; second, a Hungarian algorithm is used to find a assignment method to select one element in each row and each column, so that the total coverage rate is maximized. The parallel conflict reduction method based on the weighted Hungarian algorithm constructs a weighted bipartite graph to depict the resource competition relationship between the satellite clusters, uses the Hungarian algorithm to find the optimal assignment scheme, maximizes the total coverage rate and eliminates the multiple solutions, and configures a dynamically reconfigurable elastic star cluster service unit.

[0021] The multi-layer dimension reduction star cluster configuration system for the mega satellite constellation task planning includes:

[0022] A multi-dimensional representation module is used for multi-dimensional representation of complex tasks, extraction of key feature parameters of the tasks and normalization processing, and construction of a unified task description framework.

[0023] A task clustering module is used for a multi-satellite domain task collaborative clustering algorithm based on resource constraints, and large-scale concurrent tasks are aggregated into high-cohesion task communities.

[0024] A satellite clustering module is used for dividing a heterogeneous satellite cluster into a candidate resource cluster matching the demand according to multi-dimensional similarities such as satellite payload, resolution and time availability.

[0025] The feature comparison module is used for constructing a feature comparison interface of the task group and the resource cluster, realizing adaptive matching, and outputting a coarse-grained matching result.

[0026] The pre-screening module is used for adopting a pre-screening mechanism based on dichotomy to quickly match the task with a suitable star cluster.

[0027] The elastic index measurement module is used for an elastic index measurement method of the star cluster based on a special-shaped matrix to quantitatively evaluate coverage and redundancy.

[0028] The conflict reduction module is used for designing a parallel conflict reduction algorithm based on a weighted Hungarian algorithm to efficiently generate a final response star cluster for the resource conflict caused by the high approximation of the task demand.

[0029] The method has the advantages that:

[0030] The method projects high-dimensional task attributes to a low-dimensional space through a resource constraint driven multi-star domain task collaborative clustering algorithm, and significantly improves the solving efficiency. Strong coupling decoupling is realized through task group set division and satellite resource clustering to solve the global problem, and the optimization efficiency is greatly improved. Multi-solution stability eliminates the allocation conflict of the resource approximate satellite cluster through the design of a task-oriented parallel conflict reduction algorithm to generate a dynamically reconfigurable elastic star cluster service unit. Super large scale applicability supports efficient and stable allocation of ten thousand satellites and a large number of task requests, and is suitable for normal task scenarios.

[0031] The method converts continuous geographic space into quantifiable analysis calculation units based on a rule grid discretization method, and effectively solves the low efficiency problem of traditional coverage algorithms in complex terrain and dynamic environment in combination with satellite observation point data. The coverage efficiency of the satellite constellation is dynamically evaluated through the coverage rate of the discretized grid, and the star cluster resource scheduling is guided. In the multi-dimensional representation of the complex task scenario, the geographic position of the task scenario is normalized as a point target, a polygon target and a circle target. The coverage matrix is constructed for different targets to solve the coverage rate calculation problem.

[0032] The method judges the task coverage and effective coverage rate through the sub-satellite point expansion matrix. The sub-satellite point positions of all over-domain satellites are obtained through the above calculation method for the normal task star domain, the task demand is matched to screen out the satellites meeting the task demand, and the satellite star cluster is established according to the corresponding relationship. For the satellite sub-satellite point, the corresponding grid index is calculated, the satellite coverage matrix is constructed, and the joint coverage rate and star cluster redundancy are calculated to ensure that the appropriate repeated coverage improves the fault tolerance of the star cluster to complete the task.

[0033] In summary, the method realizes efficient and stable allocation of a huge satellite constellation resource in a super large scale and strongly coupled task scenario, and provides a systematic solution for space task scheduling. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 Task scene classification and normalization representation schematic diagram for the present application;

[0035] Figure 2 Target verification area schematic diagram for the present application;

[0036] Figure 3 Task scene clustering feature schematic diagram for the present application;

[0037] Figure 4 Improved clustering feature tree structure schematic diagram for the present application;

[0038] Figure 5 Satellite resource clustering feature schematic diagram for the present application;

[0039] Figure 6 Pre-allocation process flowchart for the present application;

[0040] Figure 7 Constellation and task group matching schematic diagram for the present application;

[0041] Figure 8 Binary-based pre-screening mechanism schematic diagram for the present application;

[0042] Figure 9 Point target task scene coverage matrix construction schematic diagram for the present application;

[0043] Figure 10 Circular area target coverage matrix construction schematic diagram for the present application;

[0044] Figure 11 Polygon target coverage matrix construction schematic diagram for the present application;

[0045] Figure 12 Grid coverage matrix correction schematic diagram for the present application;

[0046] Figure 13 Weighted bipartite graph schematic diagram for the present application. DETAILED DESCRIPTION

[0047] The present application will be further described below in conjunction with the accompanying drawings and specific embodiments, which are illustrative of the present application and are used to explain the present application, but are not limiting of the present application.

[0048] The present application provides a multi-layer dimension reduction constellation configuration method for mega satellite constellation task planning, the core of which is to realize efficient resource allocation in space task scheduling through complex task multi-dimensional representation, multi-domain collaborative task scene and constellation resource clustering algorithm, and multi-domain collaborative pre-allocation mechanism and constellation parallel conflict reduction method. The specific embodiments of the present application are described in detail below in conjunction with the accompanying drawings.

[0049] In the implementation process, first, the complex task needs to be multi-dimensionally characterized to form a unified task description framework. This process is shown in Figure 1 The tasks are divided into three typical task scenarios: cooperative observation, relay tracking, and interactive verification. The key characteristic parameters of cooperative observation tasks include target location, cooperative coverage, payload requirement, resolution, and time window. The key characteristic parameters of relay tracking tasks include target location, direction, speed, payload requirement, and effective time. The key characteristic parameters of interactive verification tasks include target location, multi-payload requirement, minimum / maximum angle difference, and effective time. These task characteristic parameters eliminate the interference of scene-specific parameters on resource matching through a piecewise normalization strategy, thereby constructing a unified task description framework. For example, for a point target task scenario, the geographic location input form is an eight-tuple, containing the coordinates of the four vertices of the grid where the point target is located. For a circular region target, the geographic location input form is a three-tuple, including the coordinates of the origin of the circular region and its radius. For a polygonal region target, the geographic location input form is a 2n-tuple, where n is the number of polygon edges, and n = 4 for a rectangular region target. Through this standardization process, different types of targets are expanded to generate coverage matrices. For large task areas, a correction method is used to improve calculation accuracy, as shown in Figures 9 to 12 .

[0050] Cooperative observation of satellites refers to the coordinated cooperation of multiple satellites (or satellite constellations) in space, time, or function to jointly complete observation objectives that a single satellite cannot efficiently achieve.

[0051] The demand D of a region target cor_area must include the target location Loc a (·) (coordinates of the four corners of a rectangle, coordinates of the center of a circle and its radius, and coordinates of the vertices of a convex polygon), cooperative coverage requirement Cov, payload requirement Pld, resolution requirement Res, and completion timeliness requirement Prd. The region target in cooperative observation is usually a large closed region, and multiple observation targets are characterized by multiple requirements.

[0052] D cor_area = {Loc a (·), Cov, Pld, Res, Prd, …} (1)

[0053] For a region target observation requirement as shown in equation (1), the observation requirement can be further characterized as follows

[0054] Task i = {task_id, req i (·)} (2)

[0055] In equation (2), task_id is the task identifier, and req

[0056] req(·)=(req_id,pld,res,cov,T s ,T e ,cov,loc a (·))

[0057] In the formula, T s With T e The task's time window represents its start and end times; loc(·) represents the task's geographical location, and its data format has multiple representation modes. For a quadrilateral point target, its data format is an octet containing the four vertices of the quadrilateral; for a convex polygon observation area, its data format is a 2n-tuple, where n is the number of sides of the polygon, containing multiple vertex targets of the polygon; for a circular area, its data format is a triple, including the origin coordinates of the circular area and its radius.

[0058] Each point target requirement D in the point group target requirements cor_spot The target point location (Loc) must be included. s (·), coverage frequency requirement Frq, coverage time interval requirement Int, load requirement Pld, resolution requirement Res, etc., while point groups are characterized by a set of point target requirements.

[0059] D cor_spot ={Loc s (·),Frq,Int,Pld,Res,…}

[0060] For the point targets shown in the above formula, their representation can also be uniformly expressed as shown in formula (2). Taking each point target in the formula as a task requirement and assigning it a requirement ID, its form can be expressed as follows:

[0061] req(·)=(req_id,pld,res,cov,T s ,T e ,cov,loc a (·))

[0062] The load requirement PLD and resolution requirement Res are directly obtained from the system input. Due to the special nature of point targets, their coverage is 100%. The observation window (T) for point targets... s ,T eThe coverage frequency requirement Frq and coverage time interval requirement Int are obtained from the system input. For a single-point target with an observation period of T, an observation frequency of Freq, and a coverage time interval requirement Int, completing the observation task can be expressed as multiple task requirements. The difficulty in characterizing these requirements lies in determining the start and end times of the tasks. Based on actual military and civilian task requirements, approximating the observation time can maximize the observation benefits and better represent the actual situation of the observed point. Therefore, solving for the single period of the task requirements is...

[0063]

[0064] For tasks requiring multiple observations, its (T) s ,T e ) can be expressed as (T s +(i-1)T s ,T s +iT). Clearly, such a disassembly setup may not meet the task's coverage time interval requirements. Therefore, further considering the time interval, its observation start and end times can be expressed as...

[0065]

[0066] Relay tracking refers to a technological system that uses multiple satellites (or satellite constellations) to dynamically coordinate in orbit to achieve continuous and seamless observation of moving targets. This system addresses the relay tracking requirements for moving targets. rely The scenario requirements need to be represented by vectors, where each element D in the vector... rely_i To represent the needs of a satellite, there needs to be a clear temporal sequence between vectors. The target in the relay tracking requirement is usually the same moving target. If there are multiple relay tracking targets, there must be multiple requirements to represent them.

[0067] D rely ={D rely_1 D rely_2 ,…,D rely_i}

[0068] For each element D in the vector rely_i It needs to include target location information (Loc). s The target location information includes (·), target orientation information Dir, target velocity information Spd, load requirement Pld, and effective time requirement Prd. The target location information is based on initial situational information during requirement planning, and is updated in subsequent tracking requirements based on the implementation status of previous tracking requirements.

[0069] D rely_i ={Loc s (·),Dir,Spd,Pld,Prd,...}

[0070] In the decomposition of relay tracking tasks, the main challenge is determining the coverage location information and effective time information, especially when multiple steps are required, making the effective input information even more difficult to obtain. Therefore, this invention proposes a task decomposition method based on prediction. For a relay tracking task, it can be decomposed into multiple area target coverage tasks, with the data format shown in the following formula, where req can be expressed as...

[0071] req(·)=(req_id,pld,res,cov,T s ,T e ,cov,loc a (·))

[0072] like Figure 2 As shown, this method uses a multi-location collaborative shooting approach for tracking. For the initial time, i.e., req0, the observation time T s The end time is T. s +d, where d is the observation time interval, and the observation area can be represented as

[0073] <lat0,lon0,lat0+vt d ,lon0+vt d >

[0074] The expression is a quadruple.

[0075] To accelerate constellation deployment, this embodiment of the invention employs a prediction method for constellation generation. Therefore, for subsequent relay tracking tasks, task characterization is also performed, with the observation start and end times expressed as follows:

[0076] <t0+it d ,t0+(it d +1)>,i=0,1...,n

[0077] Its observation target area can be described as

[0078] <lat0+vit d ,,lon0+vit d ,lat0+(i+1)vt d ,lon0+(i+1)vt d >

[0079] However, it is clear that since the target may be non-cooperative, its trajectory may not follow the predicted path. Therefore, a target verification method is needed to determine whether it has deviated from the target. The vertex coordinates of the target verification area are...

[0080] <lat0+vitd -d,,lon0+vit d -d,lat0+(i+1)vt d ,lon0+(i+1)vt d >

[0081] If the task allocation result is within the target verification area and the trajectory has not changed significantly, the constellation will remain unchanged; otherwise, the task constellation will be disbanded and regenerated.

[0082] Satellite cross-verification missions refer to the collaborative observation of the same target area from multiple heterogeneous satellites across multiple dimensions, time phases, and angles, and to improving the reliability and accuracy of the observation results through data fusion and cross-validation. This addresses the target cross-verification requirement D. cross_validation Target location Loc s (·), Multiple load requirements PLD(·), Minimum angle difference Ang min Maximum angular difference Ang max Valid time requirements P In the context of interactive verification requirements, the target is usually the same, and multiple targets are represented by multiple requirements.

[0083] D cross_validation ={Loc s (·),Pld(·),Ang min ,Ang max ,Prd...}

[0084] In this type of task, multiple load requirements can be characterized by splitting the requirements, that is, issuing multiple tasks for computation.

[0085] Next, based on the multidimensional features of the mission and the attributes of satellite resources, bidirectional dimensionality reduction clustering is performed on both the mission and the satellite, such as... Figure 3 and Figure 5 As shown. The core of task clustering lies in improving the application of clustering feature tree algorithms. Tasks have multi-dimensional features (geographical location, time window, resource demand similarity) and constraints such as total resource constraints. Since the distribution range of task time windows varies greatly, a time clustering method is used to determine the possibility of task conflicts, and conflict reduction is only performed on tasks that may have conflicts. The grouping effect is evaluated by CH score, and the optimal stratification is selected to determine the final task set, aggregating large-scale concurrent tasks into highly cohesive task clusters. Satellite clustering divides heterogeneous satellite clusters into candidate resource clusters with mutually exclusive resources based on multi-dimensional similarity such as satellite payload type, orbital factors, resolution, and time availability. The satellite clustering process is basically the same as the task resource clustering process. After constructing a feature comparison interface between task groups and resource clusters, coarse-grained matching results are output, significantly compressing the solution space.

[0086] Clustering metrics mainly consider the following factors:

[0087] (1) Payload resolution constraints of the Giants constellation satellites: Enumerated resources / attributes contain only a limited number of options, such as payload type. These resource / attribute constraints directly limit the composition of the constellation satellites required.

[0088] (2) Task time window: Since the task time is long and the distribution range is not very consistent, it is unnecessary to allocate multiple tasks at the same time to increase the search space. Therefore, the time clustering method is used to judge the possibility of task conflict. Only when the possibility of conflict is high will conflict reduction be run.

[0089] Within each star system, an improved clustering feature tree representing typical mission scenarios is generated. This structure is specialized for the task scenario grouping problem. To explicitly describe this tree structure, the scenario clustering feature (CF) is defined as follows:

[0090] CF = (N,LS,SS)

[0091] Where N represents the number of task scenarios in this CF, LS represents the resource constraint vector of each task scenario in the CF, and SS represents the squared sum of resource constraints of each task scenario in the CF. A CF can contain multiple task scenarios or multiple CFs, depending on the specific circumstances. Figure 4 As shown.

[0092] A clustering feature tree contains two key parameters: the maximum value of the leaf nodes. CF number L Each leaf node CF Sample similarity threshold T .

[0093] When accessing task scenarios sequentially, the branch nodes are first filtered according to the type of payload carried by the task scenario. Then, clustering is performed according to the target point coordinates and time constraints of the task scenario, and the current branch node is evaluated. LS and SS This refers to the current task scenario group description of the node. If the current branch node already has a task scenario or... CF This allows us to find the first clustering metric in the new task scenario that does not exceed the sample similarity threshold. T of CF And add it to it. If it does not exist, then the task scenario will constitute a new one. CF If the search finds a value that satisfies the threshold constraint CF The included task scenarios have exceeded its maximum capacity. CF The number, then the CF The internal task scenarios need to be combined into new ones. CFAfter visiting all task scenarios, the improved clustering feature tree structure shown in the figure above is formed. The same leaf node in this tree contains... CF That is, they belong to the same group.

[0094] Because the improved clustering feature tree has a multi-level clustering structure, the CH score can be used to measure the grouping effect when selecting a specific grouping method, and an appropriate stratification can be chosen to determine the final grouping result. The CH score is calculated as follows:

[0095]

[0096] Among them, B k W represents the discrete matrix between groups. K The discrete matrix within a group is defined as follows:

[0097]

[0098]

[0099] Where K represents the number of clusters, N C represents the total number of data points, where n represents the total number of sample points. q Let n represent the sample points in cluster q. q represents the number of sample points in cluster q, and c represents the center of all datasets.

[0100] The CH score corresponding to the current grouping result can be calculated using the above formula. A larger CH score indicates a smaller covariance within the data of that star region category and a larger covariance between categories. In other words, a larger CH score indicates a better grouping effect within that star region.

[0101] The region contains a large number of satellite resources, and there is coupling between multiple regions for managing mission scenario groups. This means that when selecting satellite constellations for mission scenario groups, it is still necessary to traverse the entire satellite system multiple times, which consumes a lot of storage and computing resources. To address this, a satellite constellation pre-allocation mechanism is proposed that allows for a fast traversal of all satellites and marking them with their relevant mission scenario groups. The pre-allocation process relies on the correspondence between scenario requirements and satellite resources.

[0102] In the pre-allocation mechanism, quickly traversing all satellites and grouping them into relevant mission scenarios is one of the key steps. For example... Figure 6As shown, each satellite domain is divided into multiple resource-exclusive scenario groups based on constraints such as payload type, priority, and orbital requirements that cannot be met by a single satellite. Simultaneously, within each satellite domain, all transit satellites are divided based on factors such as payload type, orbital factors, resolution, and time window, forming multiple resource-exclusive candidate satellite groups. The marked mission scenario groups are moved to the end of the traversal queue, and then an unmarked satellite group is selected. This process is repeated until all satellites are marked. This step achieves preliminary screening of feasible satellite sets for all mission groups and completes coarse-grained matching between satellites and mission scenario groups, improving the computational efficiency of subsequent grouping processes.

[0103] ① Each satellite domain is divided into multiple resource-exclusive scenario groups based on the constraints of payload types, priorities, orbital requirements, etc., which cannot be met by a single satellite.

[0104] ② Within each satellite domain, all transit satellites are divided based on factors such as the type of payload carried by the satellite, orbital factors, resolution, and time window, forming multiple sets of candidate satellites with mutually exclusive resources.

[0105] ③ Move the task scenario set marked in ② to the end of the traversal queue, then take an unmarked satellite set and return to step ②, until all satellites are marked.

[0106] This allows for the rapid search of a preliminary set of feasible satellites for all mission groups, while also completing coarse-grained matching between satellites and mission scenario groups, thus improving the computational efficiency of subsequent grouping processes.

[0107] To address the challenges of high-dimensionality, large number of satellites, and heterogeneous resource attributes in satellite constellations, this study investigates a multi-domain collaborative pre-allocation mechanism for constellation resources and a parallel conflict reduction method for constellations, enabling rapid generation of response constellations based on mission scenarios. For candidate mission groups and resource clusters after pre-clustering: First, a pre-screening mechanism based on binary search is proposed to quickly match suitable constellations for missions, solving the problems of a large number of satellites and a large search space. Second, a constellation resilience index measurement method based on irregular matrices is studied to quantitatively evaluate coverage and redundancy, providing a basis for conflict reduction. Finally, for resource conflicts caused by highly similar mission requirements, a parallel conflict reduction algorithm based on the weighted Hungarian algorithm is designed to efficiently generate the final response constellation, significantly improving mission response speed and resource utilization.

[0108] like Figure 7As shown, there is no possibility of reuse between the clusters and mission groups after clustering and bidirectional matching. Therefore, parallel computing can be used to perform cluster resource matching for each typical mission scenario. A pre-screening mechanism based on binary search is used to determine the effective coverage of each satellite. If a satellite effectively covers the mission, it is stored as a candidate satellite, and its effective coverage matrix and effective coverage rate are calculated. The effective coverage rate of each satellite in the mission is sorted to generate a sequence of candidate satellites with effective coverage rates. Effective coverage rates are calculated from high to low. For each mission, if the effective coverage rate meets the mission requirements and the redundancy reaches 200%, the screening process terminates, and a preliminary allocation scheme is generated.

[0109] like Figure 8 As shown, the coverage determination algorithm based on the time window bisection method recursively narrows the search range, significantly reducing computational complexity. In the data processing stage, the nadir data of each satellite is arranged in ascending order according to the timestamp, generating an ordered sequence. By performing coverage determination on the beginning and end of the nadir data, positions that cannot be covered are excluded, greatly reducing coverage calculations and determinations. The coverage calculation method combines regular grid discretization technology, transforming continuous geographic space into quantifiable and analyzable computational units, dynamically evaluating the coverage performance of the satellite constellation and guiding satellite constellation resource scheduling. The main steps of this method are:

[0110] ① Data processing: The sub-satellite point data of each star is processed according to the timestamp t. i Arrange in ascending order to generate an ordered sequence S = [s0, s1, ..., s2]. n-1 ].

[0111] ② Check the coverage of the beginning and end of the nadir point data. If there is no coverage, calculate whether the intermediate point is covered. If it is covered, verify t. n / 2 With t 3n / 2 If a time point is covered, then continue the binary search to both sides; if it is not covered, then search towards the middle.

[0112] This method can effectively exclude sub-satellite points that cannot be covered, greatly reducing coverage calculations and judgments.

[0113] By employing a regular grid discretization method, the system can transform continuous geographic space into quantifiable computational units. Combined with satellite observation data, this effectively addresses the inefficiency of traditional coverage algorithms in complex terrain and dynamic environments. Through coverage calculations using the discretized grid, the system can dynamically evaluate the coverage performance of satellite constellations and guide constellation resource scheduling.

[0114] In the multidimensional representation of complex task scenarios, the geographical location of the task scenario has been normalized to represent point targets, polygonal targets, and circular targets. How to construct coverage matrices for different targets is an urgent problem to be studied. The following will explain the construction of coverage matrices for different targets.

[0115] For point target tasks, the geographic location input is in the form of [lat, lon]. Expanding this into the surrounding area yields a square grid, such as... Figure 9 As shown, its vertex coordinates are

[0116]

[0117] For a circular target area, its geographic location is input in the form of [lat,lon,R]. Expanding the circular area outwards results in a square grid, such as... Figure 10 As shown, its vertex coordinates are

[0118] [lat-R,lat+R,Lon-R,lon+R]

[0119] For polygonal targets, the geographic location input format is [ <lat i ,lon i > i For an observation area formed by n coordinate points, a rectangular region is obtained by selecting the maximum and minimum values ​​among the observation points, such as... Figure 11 As shown in the triangular observation area, the region enclosed by its latitude and longitude lines can be represented as...

[0120] [min{lat i},max{lat i},min{lon i},max{lon i}]

[0121] To facilitate subsequent calculations, a grid coordinate sequence is generated for the latitude and longitude coordinates [start_lat, end_lat, start_lon, end_lon] decomposed above. Let the grid side length be d, and the number of grid points in the precision and dimensional directions be:

[0122]

[0123] Therefore, the task area can be decomposed into a coordinate sequence.

[0124] lat={lat i |lat i =start_lat+i·d,0,1,...,N lat -1}

[0125] lon={loni |lon i =start_lon + i·d,0,1,...,N lon -1}

[0126] Define the set of grid points as G = {(lat i ,lon j )|lat i ∈lat,lon j To improve computational efficiency, a parallel computing method is adopted for ∈lon}, and multi-threading is used to determine whether the position of the sub-satellite point is within the region;

[0127]

[0128] Where I(·) is an indicator function, its value is 1 when the condition is true, and 0 otherwise. Therefore, the effective grid cover matrix can be obtained as follows:

[0129]

[0130] This method can effectively determine the satellite's coverage of a mission, but when the mission area is too large, the mission area will be perceived as enlarged, resulting in a lower actual effective coverage rate. Figure 12 As shown;

[0131] Clearly, the rectangular observation area extended from the circular observation area has redundant observations at its four corners. Therefore, the standardized grid coverage matrix cannot accurately measure the effective coverage. For some triangular observation areas, even more redundant observation squares are generated, which can easily lead to inaccurate judgment of the effective observation area. Therefore, further processing of the grid coverage matrix is ​​required. This embodiment modifies the matrix in the above formula, as shown in the following formula.

[0132]

[0133] For matrix elements labeled N, their values ​​are always empty to prevent the coverage of that point from affecting the task coverage rate.

[0134] This invention uses an expanded matrix of nadir points to determine mission coverage and effective coverage. The nadir positions of all satellites passing through the domain head of a normal mission can be obtained through the above calculation method. By matching these positions with mission requirements, satellites meeting the requirements can be selected, and a satellite constellation can be established based on the corresponding relationships. For satellite nadir point o... k =(lon) k ,lat k ), calculate its corresponding grid index:

[0135]

[0136] Constructing a satellite coverage matrix Among the elements:

[0137]

[0138] Coverage calculation formula:

[0139]

[0140] Multi-satellite joint coverage calculation

[0141] For the coverage matrix {M} of n satellites sat，(1) ,...,M sat，(n) Construct the joint covering matrix:

[0142]

[0143] Element-level logical OR operation:

[0144]

[0145] Joint coverage calculation

[0146]

[0147] In the process of using multi-satellite coverage, there may be situations where a single task point is repeatedly covered by multiple satellites. A moderate amount of overlapping coverage can improve the redundancy of the satellite constellation and enhance its fault tolerance in completing tasks.

[0148] Therefore, the star cluster redundancy is defined as...

[0149]

[0150] Due to the highly similar resource requirements of various missions and the large number of missions, satellite resource conflicts frequently occur. When a satellite is assigned to multiple constellations simultaneously, it leads to high mission intensity, which is detrimental to the long-term on-orbit operation of the satellite. When time and location conflict simultaneously, the satellite cannot perform two different missions at the same time, resulting in mission failure and causing huge losses.

[0151] The bipartite graph is the core data structure used in this algorithm to characterize the resource competition relationships among satellite groups. Essentially, it is an undirected graph structure G = (V, E), where the vertex set V = {v...} i} represents each satellite group, and the edge set E = {e ij} represents the conflict relationship between star groups. In this algorithm, the generation of conflict relationships follows strict mathematical criteria: for any two groups v p and v q If and only if there exists at least one public satellite s k It belongs to both star groups (i.e., sk ∈ C(t p , r p ) ∩ C(t q , r q ) will a conflict edge e be established between them pq . The formation of this conflict relationship stems from the exclusive use constraint of satellite resources - when the same satellite is demanded by multiple groups simultaneously, a substantial resource competition relationship is formed between these groups. By sorting out the conflict relationships between the satellites and tasks in each candidate subgroup, a weighted bipartite graph can be formed, as shown in Figure 13 . On the left and right sides of the figure are the conflicting satellites and tasks respectively, and its edges are weighted, and the values are calculated by the coverage rate calculation mechanism proposed in the previous section

[0152] In this problem, through the coverage rate calculation mechanism proposed in the previous section, an n×n effective coverage rate matrix C = [c ij about the conflicting tasks and conflicting satellites can be obtained, where c ij represents the negative value of the coverage rate of assigning ground satellite i to task j. Through the Hungarian algorithm, it is hoped to find an assignment method to select exactly one element in each row and each column, so as to maximize the total coverage rate. Its mathematical form can be expressed as

[0153]

[0154] Gradually approach the optimal solution through matrix transformation and augmented path search

[0155] The present invention judges task coverage and effective coverage rate through the sub-satellite point expansion matrix. The domain head of the constellation where the normal task is located obtains the sub-satellite point positions of all satellites passing through the domain through the above calculation method, matches the task requirements to screen out the satellites that meet the task requirements, and establishes a satellite constellation according to the corresponding relationship, and ensures appropriate repeated coverage to improve the fault tolerance ability of the constellation to complete tasks [[ID=​​​​​​

[0157] In the conflict mitigation phase, a parallel conflict mitigation method based on the weighted Hungarian algorithm is used to resolve resource conflicts caused by highly similar mission requirements. When a satellite is simultaneously assigned to multiple constellations, it can lead to excessively high mission intensity, which is detrimental to long-term on-orbit operation. For example... Figure 13 As shown, the bipartite graph serves as the core data structure for depicting resource competition among satellite constellations. The vertex set represents each satellite constellation, while the edge set represents the conflict relationships between them. The generation of conflict relationships follows a strict mathematical criterion: for any two constellations, a conflict edge is established between them only if at least one common satellite belongs to both constellations. The effective coverage matrix between conflicting tasks and conflicting satellites is obtained through a coverage calculation mechanism, where each element represents the negative value of the coverage allocated to task j for satellite i. The Hungarian algorithm gradually approximates the optimal solution through matrix transformations and augmented path search, finding an assignment method that selects exactly one element for each row and each column, maximizing the total coverage. This process ensures appropriate overlapping coverage, improving constellation redundancy and enhancing the constellation's fault tolerance in completing its tasks.

[0158] In summary, this invention utilizes a resource-constrained multi-domain task collaborative clustering algorithm to project high-dimensional task attributes into a low-dimensional space, significantly reducing computational complexity and improving solution efficiency. Strong coupling decoupling is achieved through task group partitioning and satellite resource clustering to realize global problem divide-and-conquer, greatly improving optimization efficiency. Multi-solution stability is achieved by designing a task-oriented parallel conflict reduction algorithm to eliminate allocation conflicts in resource-approximate satellite clusters, generating dynamically reconfigurable elastic constellation service units. Ultra-large-scale applicability supports efficient and stable allocation of tens of thousands of satellites and massive task requests, suitable for routine task scenarios. This invention transforms continuous geographic space into quantifiable computational units using a regular grid discretization method, effectively addressing the inefficiency of traditional coverage algorithms in complex terrain and dynamic environments by combining satellite observation point data. Coverage performance of satellite constellations is dynamically evaluated through discretized grid coverage calculation, guiding constellation resource scheduling. In the multi-dimensional representation of complex task scenarios, the geographical locations of task scenarios are normalized to represent point targets, polygonal targets, and circular targets, and coverage matrices are constructed for different targets to solve the coverage calculation problem.

[0159] The technical solutions of the present invention are not limited to the specific embodiments described above. Any technical modifications made in accordance with the technical solutions of the present invention fall within the protection scope of the present invention.

Claims

1. A multi-level reduced-dimensional constellation configuration method for planning giant satellite constellation missions, characterized in that, Includes the following steps: S1: Perform multi-dimensional representation of complex tasks, extract key feature parameters of the tasks and perform normalization processing to build a unified task description framework. S2: A resource-constrained multi-domain task collaborative clustering algorithm that aggregates large-scale concurrent tasks into highly cohesive task communities; S3: Based on the multidimensional similarity of satellite payload, resolution, and time availability, a clustering method based on multidimensional similarity measurement is used to divide heterogeneous satellite clusters into candidate resource clusters that match the requirements. S4: Construct a feature comparison interface between task groups and resource clusters to achieve adaptive matching and output coarse-grained matching results; S5: Employs a pre-screening mechanism based on binary search to quickly match suitable star clusters for the mission; S6: A method for measuring constellation resilience based on irregular matrices, quantitatively assessing coverage and redundancy; S7: To address resource conflicts caused by highly similar task requirements, a parallel conflict reduction algorithm based on the weighted Hungarian algorithm is designed to efficiently generate the final response constellation.

2. The multi-layered dimensionality reduction constellation configuration method for planning giant satellite constellations according to claim 1, characterized in that: The multidimensional representation of complex tasks in step S1 includes collaborative observation tasks, relay tracking tasks, and interactive verification tasks. The collaborative observation task extracts key feature parameters of regional target requirements, including target location, collaborative coverage requirements, payload requirements, resolution requirements, and completion time requirements. The relay tracking task extracts target location information, target direction information, target velocity information, payload requirements, and effective time requirements. The interactive verification task extracts target location, multiple payload requirements, minimum angle difference, maximum angle difference, and effective time requirements.

3. The multi-layered reduced-dimensional constellation configuration method for planning giant satellite constellations according to claim 1, characterized in that: The multi-star domain task collaborative clustering algorithm based on resource constraints in step S2 is as follows: using multi-dimensional task features as the basic metric, and combining time overlap and resource demand similarity, tasks are aggregated into several highly cohesive task clusters under the premise of satisfying the total resource constraints of each star domain, thereby significantly reducing the task scheduling dimension.

4. The multi-layered reduced-dimensional constellation configuration method for planning giant satellite constellations according to claim 1, characterized in that: The satellite resource clustering method based on multidimensional similarity measurement in step S3 specifically involves: extracting key indicators such as payload, resolution, and time availability of satellite resources, and using a multi-level improved clustering feature tree clustering method to divide heterogeneous satellites into candidate satellite groups that match mission requirements.

5. The multi-layered reduced-dimensional constellation configuration method for planning giant satellite constellations according to claim 1, characterized in that: In step S4, the feature comparison interface between the task group and the resource cluster is constructed by mapping the task group to the satellite resource cluster, thereby providing coarse-grained matching results for the subsequent satellite cluster resource pre-allocation and conflict resolution algorithms.

6. The multi-layered dimensionality reduction constellation configuration method for planning giant satellite constellations according to claim 1, characterized in that: The pre-screening mechanism based on the binary search method in step S5 is as follows: First, a coverage determination algorithm based on the time window binary search method is adopted, with the time dimension as the main index, and the candidate star points within the task time window are quickly located by the binary search method; then, the effective coverage matrix, redundancy, and effective coverage rate of the task are calculated.

7. The multi-layered dimensionality reduction constellation configuration method for planning giant satellite constellations according to claim 1, characterized in that: The method for measuring the constellation elasticity index of the irregular matrix in step S6 is as follows: the observation area of ​​the mission is discretized into an irregular grid to generate an effective coverage matrix; the candidate satellites generate a satellite coverage matrix based on the satellite coverage situation, and the matrix is ​​processed to obtain the effective coverage capability of the satellites for the mission.

8. The multi-layered reduced-dimensional constellation configuration method for planning giant satellite constellations according to claim 1, characterized in that: The parallel conflict reduction algorithm based on the weighted Hungarian algorithm in step S7 is as follows: First, a weighted bipartite graph is constructed to characterize the resource competition relationship among satellite groups; second, an assignment method is found through the Hungarian algorithm to ensure that exactly one element is selected in each row and each column, so as to maximize the total coverage.

9. A multi-layered reduced-dimensional constellation configuration system for planning giant satellite constellation missions, characterized in that, include: The multidimensional representation module is used to perform multidimensional representation of complex tasks, extract key feature parameters of the tasks and perform normalization processing, and build a unified task description framework. The task clustering module is used for resource-constrained multi-domain task collaborative clustering algorithms to aggregate large-scale concurrent tasks into highly cohesive task groups. The satellite clustering module is used to divide heterogeneous satellite clusters into candidate resource clusters that match the requirements based on multi-dimensional similarity such as satellite payload, resolution, and time availability. The feature comparison module is used to build a feature comparison interface between task groups and resource clusters, realize adaptive matching, and output coarse-grained matching results. The pre-screening module is used to quickly match suitable star clusters for the task using a binary search-based pre-screening mechanism. The elasticity index measurement module is used for the star cluster elasticity index measurement method based on irregular matrix, and to quantitatively evaluate coverage and redundancy. The conflict reduction module is designed to address resource conflicts caused by highly similar task requirements. It employs a parallel conflict reduction algorithm based on the weighted Hungarian algorithm to efficiently configure the final response constellation.

Images