A hierarchical task planning method and system for single-unit large-scale heterogeneous constellations
By adopting a single-distributed collaborative task planning method, and using a large language model and RDF triple graph to generate collaborative schemes, the problems of long computation time and insufficient collaborative rules for heterogeneous payloads in the planning of large heterogeneous constellation tasks are solved, and efficient task planning and autonomous collaboration are achieved.
Patent Information
- Application Number
- CN202611261742.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-19
- Publication Date
- 2026-11-17
AI Technical Summary
Existing satellite mission planning methods suffer from long computation times, heavy communication burdens, and poor real-time response capabilities in large heterogeneous constellation scenarios. They are also difficult to adapt to natural language commands and mixed data tables, and the integration of heterogeneous payload coordination rules is insufficient, leading to duplicate task assignments and timing conflicts.
A distributed collaborative task planning method based on single-task assignment is adopted. The task requirements are processed semantically and standardized through a large language model. Satellite clusters are divided for local optimization. Collaborative schemes are generated through RDF triple graphs and SPARQL queries. The Harris Eagle optimization algorithm is combined for intra-cluster optimization and cross-cluster evaluation to generate a global collaborative task plan.
It improves the real-time performance and autonomous coordination capabilities of large heterogeneous constellation mission planning, reduces computational and communication overhead, enhances the automation level of mission planning and the standardization and computability of heterogeneous payload coordination, and strengthens dynamic executability.
Smart Images

Figure CN123117622A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of satellite mission planning technology, specifically to a method and system for hierarchical mission planning of a large heterogeneous constellation with single-unit dispatch. Background Technology
[0002] With the development of large low-Earth orbit satellite constellations, heterogeneous remote sensing payloads, inter-satellite links, and on-board edge computing capabilities, satellite systems are gradually evolving from a centralized control mode with a small number of high-value platforms to a multi-satellite networking, cluster management, satellite-ground collaboration, and autonomous cooperative mode. In missions such as Earth observation, disaster emergency response, smart agriculture, and continuous regional reconnaissance, a single satellite often cannot simultaneously meet the comprehensive requirements of coverage, observation accuracy, response timeliness, mission continuity, and reliability. Multiple satellites of different types are needed to work together according to a specific time sequence and payload coordination rules to complete the mission.
[0003] Existing satellite mission planning methods mainly fall into two categories: centralized planning and distributed planning. Centralized planning typically involves a ground center collecting all satellite status, mission requirements, visible time windows, and resource constraints before uniformly modeling and solving the problem. This approach offers strong global coordination capabilities. However, as the number of missions, satellites, and candidate solutions increases rapidly, the combinatorial search space grows explosively, leading to long computation times, heavy communication burdens, and poor real-time response capabilities. Especially in large-scale heterogeneous constellation scenarios, candidate collaborative solutions involve not only mission allocation but also collaborative relationships between different payloads, time conflicts within the same satellite, capacity constraints, link constraints, and mission priority trade-offs. Centralized solutions struggle to meet the demands for rapid planning and dynamic execution.
[0004] Distributed planning methods can decompose global tasks into multiple local tasks through multi-agent negotiation, market mechanisms, or contract network protocols, thereby improving parallel computing capabilities and system scalability. However, traditional distributed methods still have the following shortcomings: First, task requirements often need to be manually converted into strictly structured inputs, making it difficult to adapt to natural language instructions, temporary tasks, mixed data tables, and abnormal constraints; second, the distributed negotiation process is not sufficiently integrated with heterogeneous payload coordination rules, making it difficult to automatically generate candidate solutions that meet the complex task chains of SAR satellite wide-area perception, hyperspectral imaging, and optical high-resolution imaging and recognition; third, there is a lack of unified constraint expression for intra-cluster local optimization and cross-cluster global evaluation, which easily leads to duplicate task assignments, timing conflicts, and capacity overruns. Therefore, there is an urgent need for a dispatch-based distributed collaborative mission planning technology for large heterogeneous satellite constellations. This technology should be able to automatically convert mission requirements into standardized announcements, divide the constellation into autonomous bidding satellite clusters, combine intra-cluster local optimization with cross-cluster global bidding, and support local replanning when dynamic states change, so as to improve the real-time performance, scalability, executability and autonomous collaborative capability of mission planning. Summary of the Invention
[0005] The purpose of this invention is to propose a hierarchical task planning method and system for large-scale heterogeneous constellations with single-mode dispatch, in order to solve the problems of large-scale centralized solution, insufficient autonomy of distributed negotiation, difficulty in collaborative modeling of heterogeneous payloads, and decreased plan executability caused by dynamic state drift in large-scale heterogeneous constellation task planning. The goal is to reduce the computing power and latency overhead of centralized planning, and improve the real-time performance, executability, and autonomous collaboration level of task planning.
[0006] To achieve the above objectives, the present invention adopts the following technical solution.
[0007] A hierarchical task planning method for a large-scale heterogeneous constellation with single-unit dispatching includes the following steps: The task requirements are semantically parsed, constraints are extracted and standardized to obtain the task announcement, and the task announcement is sent to the corresponding satellite cluster; The large heterogeneous constellation is divided into multiple satellite clusters, and the local state of each satellite cluster is obtained based on the resource status of the member satellites and the planned missions within each satellite cluster. Based on the mission announcement and the local state, the executable time information of the member satellites for the mission is converted into meta-missions, and the meta-missions are combined according to the coordination rules and mission constraints to obtain a set of candidate coordination schemes for each mission. Based on the candidate collaboration scheme set and local state of each satellite cluster, intra-cluster optimization is performed to determine the proposed task and the corresponding candidate collaboration scheme, and a tender document including task, collaboration scheme, execution resources and constraint feasibility information is generated; Based on the bids submitted by multiple satellite clusters, cross-cluster evaluation, duplicate task resolution, and resource conflict repair are carried out to determine the winning satellite clusters and winning collaboration schemes for each task, and generate a global collaborative task plan.
[0008] Furthermore, the task requirements are semantically parsed, constraints extracted, and standardized to obtain a task announcement, which is then sent to the corresponding satellite cluster, including: Receive at least one of natural language task description and structured task data, as well as task planning data, and call a large language model to perform semantic parsing, constraint extraction, anomaly detection and standardization on the task requirements to obtain standardized task information; A mission announcement is generated based on the standardized mission information, and the mission announcement is sent to a satellite cluster with mission execution capabilities.
[0009] Furthermore, a large language model is invoked to perform semantic parsing, constraint extraction, anomaly detection, and standardization on the task requirements, resulting in standardized task information, including: From the task requirements, identify the task object, shooting location, observation time requirements, payload requirements, coordination sequence, and result return requirements to obtain task semantic information; Based on the task semantic information, time constraints, safety time interval constraints, task uniqueness constraints, capacity constraints, reliability constraints, and load matching constraints are extracted, and abnormal task data is standardized to obtain standardized task information.
[0010] Further, generating a mission announcement based on the standardized mission information and sending the mission announcement to a satellite cluster with mission execution capabilities includes: Based on the standardized task information, determine the task identifier, shooting location, observation time requirements, task priority, payload requirements, collaborative observation rules, safety time interval constraints, capacity constraints, reliability requirements, and result return requirements, and generate a task announcement; The mission announcement is used to identify satellite clusters with mission execution capabilities, and the mission announcement is sent to the identified satellite clusters.
[0011] Furthermore, the large heterogeneous constellation is divided into multiple satellite clusters, and the local state of each satellite cluster is obtained based on the resource status and planned missions of the member satellites within each cluster, including: The clustering relationship of member satellites is determined based on at least one of the following: constellation affiliation, orbital relationship, coverage relationship, payload capacity, inter-satellite link status, and mission function relationship, resulting in multiple satellite clusters; Cluster management nodes are determined for the multiple satellite clusters. The cluster management nodes obtain the payload status, visibility window, current load, capacity status, energy status, reliability, link reachability and planned tasks of the member satellites to obtain the local status of each satellite cluster.
[0012] Further, based on the mission announcement and the local state, the executable time information of member satellites for the mission is converted into meta-tasks, and the meta-tasks are combined according to the coordination rules and mission constraints to obtain a set of candidate coordination schemes for each mission, including: Based on the mission announcement and the local state, the visible time window of the member satellite for the mission is converted into a meta-mission, and the meta-mission is associated with the mission, member satellite, payload type, executable time and resource usage information; Based on the collaboration rules and task constraints, the meta-tasks are constrained, matched, and combined to obtain a set of candidate collaboration schemes for each task.
[0013] Further, the meta-tasks are constrained and combined according to the coordination rules and the task constraints to obtain a set of candidate coordination schemes for each task, including: Based on the coordination rules and task constraints, determine the temporal relationships between atomic tasks, as well as load matching, time intervals, link reachability, and resource capacity constraints, and construct a coordination rule graph. The meta-tasks are filtered and combined according to the collaborative rule graph to obtain a set of candidate collaborative schemes that satisfy the temporal relationship, load matching, time interval, link reachability and resource capacity constraints.
[0014] Further, the meta-tasks are filtered and combined according to the collaborative rule graph to obtain a set of candidate collaborative schemes that satisfy the temporal relationship and constraints on load matching, time interval, link reachability, and resource capacity, including: The mission, member satellites, and the meta-mission are represented as RDF resources, and the time information, attribution relationship, execution relationship, and payload type information are represented as RDF predicates to construct an RDF triple graph. Based on the collaborative rule graph, constraint query conditions are generated, and SPARQL query is performed on the RDF triple graph to obtain the candidate collaborative scheme set.
[0015] Furthermore, based on the candidate coordination scheme set and local state of each satellite cluster, intra-cluster optimization is performed to determine the proposed task and the corresponding candidate coordination scheme, generating a tender document that includes task, coordination scheme, execution resources, and constraint feasibility information, including: Discrete index encoding is performed on the candidate collaborative schemes for each task and the case where the corresponding task is not selected to obtain the optimized individual; Based on the local state of each satellite cluster and the mission coverage, mission benefits, mission priority, mission timeliness, mission reliability, load balancing, and constraint feasibility, the optimization individual is iteratively searched to determine the mission to be undertaken and the corresponding candidate collaboration scheme. Based on the proposed mission and the corresponding candidate collaboration schemes, determine the member satellite execution sequence, execution time, payload type, resource consumption, expected benefits, reliability, load status, capacity status, and constraint feasibility information; Based on the other feasible candidate collaborative schemes corresponding to the proposed task, a backup scheme is determined, and a tender document is generated based on the member satellite execution sequence, execution time, payload type, resource consumption, expected benefits, reliability, load status, capacity status, constraint feasibility information, and the backup scheme.
[0016] Furthermore, based on the local state of each satellite cluster and considering factors such as mission coverage, mission benefits, mission priority, mission timeliness, mission reliability, load balancing, and constraint feasibility, the optimized individuals are iteratively searched to determine the tasks to be undertaken and the corresponding candidate collaborative schemes, including: The optimized individuals are evaluated based on task coverage, task benefits, task priority, task timeliness, task reliability, load balancing, and constraint feasibility to obtain a comprehensive evaluation result, and an elite set is determined based on the comprehensive evaluation result. The exploration probability is determined based on the iteration progress, the exploration stage or development stage is determined based on the exploration probability, and the Harris Eagle optimization algorithm is used to update the continuous search position corresponding to the optimized individual, and the updated continuous search position is mapped to the discrete candidate solution index. During the development phase, the discrete candidate scheme index is updated based on the elite set, and local candidate replacement and constraint repair are performed to obtain the task to be undertaken and the corresponding candidate collaborative scheme.
[0017] Furthermore, the discrete candidate scheme index is updated based on the elite set, and local candidate replacement and constraint repair are performed to obtain the task to be undertaken and the corresponding candidate collaborative scheme, including: The replacement priority of each task position is determined based on task benefits, conflict risks, and resource consumption, and the task position to be replaced is determined based on the replacement priority. The candidate collaborative schemes corresponding to the positions of the tasks to be replaced are replaced according to the elite set, and the replacement result is determined according to the comprehensive evaluation results and constraint feasibility before and after the replacement. The replacement results are checked for mission uniqueness, member satellite timing conflicts, and capacity constraints. Candidate collaboration schemes that do not meet the constraints are repaired to obtain the proposed mission and the corresponding candidate collaboration schemes.
[0018] Furthermore, based on the tender documents submitted by multiple satellite clusters, cross-cluster evaluation, duplicate task resolution, and resource conflict remediation are conducted to determine the winning satellite clusters and winning collaborative schemes for each task, and to generate a global collaborative task plan, including: The bid documents are merged according to the task identifier, and the winning utility of each bid result is determined based on task priority, bid benefits, execution time, reliability, resource consumption, load balancing and default risk. Based on the winning bid effect, the winning bid results are determined from the bidding results corresponding to the same task, and resource conflicts between the determined winning bid results are repaired. The winning satellite clusters and winning bid coordination schemes for each task are determined, and a global collaborative task plan is generated.
[0019] Furthermore, based on the winning bid effect, the winning bid results are determined from the bidding results corresponding to the same task, and resource conflicts between the determined winning bid results are repaired. The winning satellite clusters and winning collaboration schemes for each task are determined, and a global collaborative task plan is generated, including: Based on the aforementioned bid effectiveness, execution time, resource consumption, and reliability, the winning bid result is determined from the feasible bid results corresponding to the same task, and the initial winning bid result is obtained. The resource conflicts of shared satellite, shared link and shared ground station in the initial bidding results are verified, the bidding results with resource conflicts are discarded, the alternative solutions are obtained from the bid documents corresponding to the rejected bid results, and the alternative solutions that meet the constraints are determined from the alternative solutions. The initial bidding results are updated based on the alternative solutions, and tasks for which there are no alternative solutions that meet the constraints are re-bid to obtain a global collaborative task plan.
[0020] A hierarchical task planning system for a large-scale heterogeneous constellation with single-unit dispatch capability, comprising: The task announcement generation module is used to perform semantic parsing, constraint extraction and standardization of task requirements to obtain task announcements, and send the task announcements to the corresponding satellite clusters; The local state maintenance module is used to divide a large heterogeneous constellation into multiple satellite clusters and obtain the local state of each satellite cluster based on the resource status of the member satellites and the planned tasks within each satellite cluster. The candidate collaboration scheme generation module is used to convert the executable time information of member satellites for the mission into meta-tasks based on the mission announcement and the local state, and to combine the meta-tasks according to collaboration rules and mission constraints to obtain a set of candidate collaboration schemes for each mission. The intra-cluster optimization module is used to perform intra-cluster optimization based on the candidate coordination scheme set and local state of each satellite cluster, determine the tasks to be undertaken and the corresponding candidate coordination schemes, and generate a tender document that includes task, coordination scheme, execution resources and constraint feasibility information. The global evaluation module is used to conduct cross-cluster evaluation, resolve duplicate tasks, and repair resource conflicts based on the bids submitted by multiple satellite clusters, determine the winning satellite clusters and winning collaborative schemes for each task, and generate a global collaborative task plan.
[0021] The present invention has achieved the following beneficial effects.
[0022] 1. This invention addresses the problems of large search space, high computational and communication overhead in centralized mission planning for large heterogeneous constellations. It organizes mission announcements, bidding, evaluation, and awarding through the Contract Network Protocol (CNP), decomposing global mission planning into intra-cluster local optimization of multiple satellite clusters and cross-cluster coordination by the ground mission management center. This reduces the scale of centralized search and communication overhead, and improves the real-time performance and autonomous collaborative capabilities of large-scale constellation mission planning.
[0023] 2. This invention addresses the problems of heterogeneous sources of data and constraints for natural language tasks and structured tasks, as well as the large workload of manual preprocessing. It uses a large language model to perform semantic parsing, constraint extraction, anomaly detection, and standardization of task requirements, forming a unified task announcement. The hard constraints are satisfied and the solution is selected by rule reasoning, constraint verification, and optimization algorithms, thereby improving the automation level of task requirement access and emergency task processing.
[0024] 3. This invention addresses the problem of the difficulty in unifying the modeling of collaborative rules among heterogeneous payloads such as SAR, hyperspectral, and optical satellites. It converts the visible time windows of member satellites for missions into standard meta-tasks, constructs a collaborative rule graph based on collaborative rules and mission constraints, and filters meta-task combinations that satisfy temporal relationships, payload matching, time intervals, link reachability, and resource capacity constraints through RDF triple graphs and SPARQL queries, thereby improving the standardization and computability of generating candidate schemes for heterogeneous payload collaborative tasks.
[0025] 4. This invention addresses the problem of a large number of discrete candidate collaborative scheme combinations that are subject to task uniqueness, timing conflicts, and capacity constraints. Through discrete index encoding and an optimization process with adaptive exploration, elite guidance, local candidate replacement, and constraint repair mechanisms, the invention optimizes the candidate collaborative schemes of each satellite cluster within the cluster, thereby obtaining a combination of tasks and candidate collaborative schemes that meet the constraints and improving the quality and feasibility of cluster-level bidding schemes.
[0026] 5. This invention addresses the problem of deviation between planned and actual execution states caused by changes in satellite resource status. It updates the satellite cluster resource status through execution feedback and triggers local re-bidding or local re-optimization when state drift or scheme failure occurs. This allows subsequent mission planning to be adjusted based on the updated resource status, reducing scheme failure caused by state drift and improving the dynamic executability of mission plans. Attached Figure Description
[0027] Figure 1 This is a schematic diagram of the overall process of the hierarchical task planning method for a single-type large heterogeneous constellation in the embodiment. Detailed Implementation
[0028] To make the various technical features, advantages, or effects of the present invention more apparent and understandable, detailed descriptions are provided below through embodiments.
[0029] This invention provides a hierarchical task planning method for a large-scale heterogeneous constellation with single-unit dispatch capability. This embodiment uses... Figure 1 The following example illustrates the hierarchical mission planning process for a large heterogeneous constellation. (See also...) Figure 1The ground mission management center negotiates missions with multiple satellite clusters based on the Contract Network Protocol (CNP), with each satellite cluster participating in mission planning as a bidder. The mission planning process includes mission announcement generation, satellite cluster bidding, bid evaluation and awarding, and execution feedback. Specifically, the ground mission management center sends the mission announcement to satellite clusters with mission execution capabilities. Each satellite cluster generates candidate collaborative solutions based on its own resource status and optimizes its intra-cluster bidding, returning the resulting bid documents to the ground mission management center. The ground mission management center evaluates the bid documents submitted by multiple satellite clusters, determines the winning bid, and sends the winning results to the corresponding satellite clusters. The satellite clusters execute the mission according to the winning bid and return execution feedback to the ground mission management center.
[0030] The task understanding and announcement generation process involves semantic parsing, constraint extraction, standardization, and task announcement generation of corresponding task requirements. Cluster A, Cluster B to Cluster N represent different satellite clusters participating in the task planning. Each satellite cluster generates candidate collaborative schemes based on its resource status and performs intra-cluster bidding optimization. The Contract Network Protocol (CNP) is used to carry the interaction between the ground mission management center and each satellite cluster, including task announcements, bid documents, and winning bids. Bid evaluation and awarding are used to evaluate the bid documents submitted by each satellite cluster and determine the winning scheme. After task execution, execution feedback is generated to reflect the task execution status. This embodiment will now describe the task planning process in detail according to the processing order of task announcement generation, satellite cluster local state maintenance, candidate collaborative scheme generation, intra-cluster optimization, and cross-cluster bidding. When this method is executed, the ground mission management center can receive task requirements and task planning data and can interact with satellite nodes in a large heterogeneous constellation. Satellites in a large heterogeneous constellation have constellation affiliation, orbit, payload, visible time window, resource status, and link status information that can be used for task planning. The ground mission management center is responsible for issuing mission announcements and cross-cluster bidding. The cluster management nodes of each satellite cluster are responsible for local state maintenance, candidate collaboration scheme generation, and intra-cluster optimization. The mission planning process sequentially generates mission announcements, candidate collaboration scheme sets, tender documents, and global collaborative mission plans.
[0031] Step S1: Semantic parsing, constraint extraction, and standardization of the task requirements are performed to obtain the task announcement, which is then sent to the corresponding satellite cluster.
[0032] In an optional embodiment of the present invention, step S1 may include: Step S11: Receive at least one of natural language task description and structured task data, as well as task planning data; call a large language model to perform semantic parsing, constraint extraction, anomaly detection, and standardization on task requirements to obtain standardized task information.
[0033] Specifically, the natural language task description can be a natural language instruction to perform observation tasks on multiple disaster-stricken locations in a region within a specified time period; structured task data can include a task list, latitude and longitude coordinates of the imaging task, observation time requirements, task priority, payload type, and coordination rules; task planning data can include a satellite resource table, payload requirement table, visible time window table, time window constraints, current plan table, historical plan data, link reachability table, and capacity status table. The large language model can parse the above data based on preset prompt templates and tool call interfaces, converting task requirements from different sources and with different expressions into a unified data expression.
[0034] When performing semantic parsing, the large language model identifies the task type, task object, shooting location, expected completion time, observation payload, coordination order, and result return requirements from the task requirements to obtain task semantic information. Based on the task semantic information, it extracts the earliest start time, latest completion time, minimum safe time interval, task uniqueness constraint, capacity constraint, reliability constraint, and payload matching constraint. At the same time, it identifies abnormal task data such as missing coordinates, time window conflicts, unclear payload requirements, task duplication, missing task priorities, and tasks exceeding the planning cycle, and standardizes the format of the task data to obtain standardized task information.
[0035] Step S12: Generate a mission announcement based on standardized mission information and send the mission announcement to satellite clusters with mission execution capabilities.
[0036] Specifically, based on standardized task information, the task identifier, shooting location, observation time requirements, task priority, payload requirements, collaborative observation rules, minimum safe time interval, capacity constraints, reliability requirements, and result return requirements are determined, and organized into a task announcement according to a unified field structure. The task announcement can be represented as follows: .in, This indicates a task announcement; Represents a set of tasks; Represents the set of task requirements; Represents a set of constraints; This represents the set of coordination rules. For any task... The task announcement records the task identifier, latitude and longitude of the shooting task, task priority, required payload, time window, collaboration rule identifier, and result return requirements.
[0037] Based on the mission location, payload requirements, observation time requirements, and collaborative observation rules in the mission announcement, satellite clusters with mission execution capabilities are identified, and mission announcements are sent to the identified satellite clusters. The mission announcements provide the mission requirements and constraints needed for subsequent candidate collaborative scheme generation and intra-cluster mission planning.
[0038] In this embodiment, the large language model is used to perform semantic parsing, constraint extraction, anomaly detection, and standardization on multi-source heterogeneous task inputs, but does not directly perform subsequent task hard constraint satisfaction and candidate collaborative scheme selection; task hard constraint satisfaction and candidate collaborative scheme selection are completed based on rule reasoning, constraint verification, and optimization calculation, respectively.
[0039] Step S2: Divide the large heterogeneous constellation into multiple satellite clusters, and obtain the local state of each satellite cluster based on the resource status of the member satellites and the planned missions within each satellite cluster.
[0040] In an optional embodiment of the present invention, step S2 may include: Step S21: Determine the clustering relationship of member satellites based on at least one of constellation affiliation, orbital relationship, coverage relationship, payload capacity, inter-satellite link status, and mission function relationship to obtain multiple satellite clusters.
[0041] Specifically, satellite clusters can be divided according to the same constellation, the same orbital plane, adjacent coverage areas, complementary payload capabilities, inter-satellite link stability, or mission function similarity, so that member satellites belonging to the same cluster relationship are assigned to the corresponding satellite cluster.
[0042] Step S22: Determine cluster management nodes for multiple satellite clusters. The cluster management nodes obtain the payload status, visibility window, current load, capacity status, energy status, reliability, link reachability, and planned tasks of the member satellites to obtain the local status of each satellite cluster.
[0043] Specifically, each satellite cluster corresponds to a cluster management node. The cluster management node can be a member satellite with onboard computing capabilities, communication links, remaining energy, and operational status that meet management requirements, or it can be a logical management node on the ground side. The cluster management node obtains and maintains the member satellite number, payload type, payload operating status, visibility window, currently assigned tasks, capacity limit, occupied capacity, reliability, link reachability, and energy status of the member satellites in the corresponding satellite cluster.
[0044] satellite cluster The local state can be denoted as Local state It consists of the resource status and planned tasks of the member satellites within the corresponding satellite cluster, and serves as the resource status basis for subsequent meta-task generation, candidate collaborative scheme generation, and cluster optimization.
[0045] Step S3: Based on the mission announcement and local status, convert the executable time information of member satellites for the mission into meta-missions, and combine the meta-missions according to the coordination rules and mission constraints to obtain a set of candidate coordination schemes for each mission.
[0046] In an optional embodiment of the present invention, step S3 may include: Step S31: Based on the mission announcement and local status, convert the visible time window of the member satellites for the mission into a meta-mission, and associate the meta-mission with mission, member satellites, payload type, executable time and resource usage information.
[0047] Specifically, the visible time windows formed by member satellites for the mission are standardized, and each executable time window is converted into a meta-task. A meta-task can be represented as: in, Indicates the meta-task number; Indicates the task number; Indicates the member satellite number that is capable of performing the meta-mission; Indicates the load type; and These represent the start and end times of the executable time window, respectively. and These represent the longitude and latitude of the shooting location, respectively. This indicates the capacity, energy, or storage resources consumed in executing a meta-task.
[0048] Step S32: Based on the collaboration rules and task constraints, perform constraint matching and combination on the meta-tasks to obtain a set of candidate collaboration schemes for each task.
[0049] Specifically, for multi-satellite heterogeneous collaborative missions, the temporal relationships between atomic tasks, as well as payload matching, time intervals, link reachability, and resource capacity constraints, are determined based on collaborative rules and mission constraints, and a collaborative rule graph is constructed accordingly. The collaborative rule graph can be represented as follows: .in, Represents a collaborative rule graph; This represents a set of nodes, with each node corresponding to a SAR imaging, hyperspectral imaging, optical imaging, inter-satellite transmission, ground backhaul, or on-board processing atomic task. This represents the set of constraint edges, which indicate the temporal relationships between atomic tasks, as well as constraints on load matching, time intervals, link reachability, and resource capacity.
[0050] The temporal relationships between atomic tasks can include BEFORE, MEET, DURING, and OVERLAP. Constraint edges can be represented as: in, Indicates the identifier of the preceding task; Indicates the identifier for subsequent tasks; Indicates the time sequence relation type; Indicates the time endpoint used for comparison; Represents constraint operators; This indicates the threshold or range corresponding to the constraint.
[0051] The mission, member satellites, and meta-missions are represented as RDF resources, and start_time, end_time, belong_to, execute_by, payload_type, sbj_satellite, and obj_satellite are represented as RDF predicates, constructing an RDF triple graph. Based on the collaborative rule graph, corresponding SPARQL constraint query conditions are generated. Meta-missions in the RDF triple graph are filtered and combined using FILTER and BIND conditions to obtain meta-mission combinations that satisfy constraints on timing relationships, payload matching, time intervals, link reachability, and resource capacity.
[0052] When there are constraint edges in the collaborative rule graph that allow SAR tasks to be executed before optical tasks. At that time, the end time variable "?sarEndTime" corresponding to the SAR mission and the start time variable "?optStartTime" corresponding to the optical mission are obtained, and the SPARQL filtering condition FILTER(?sarEndTime < ?optStartTime) is used to filter the meta-mission combinations that satisfy the sequential relationship. This is done when there is a minimum safe time interval between the SAR mission and the optical mission. When, the corresponding time constraint is Accordingly, the FILTER(?optStartTime-?sarEndTime>?Delta) method is used to filter meta-task combinations that meet the minimum safe time interval requirement. Among these, Indicates the start time of the optical mission; Indicates the end time of the SAR mission; This represents the minimum safe time interval between consecutive tasks.
[0053] Specifically, each combination of meta-tasks that satisfies the collaboration rules and task constraints is transformed into a candidate collaboration scheme. (Task) The set of candidate collaborative schemes can be represented as .in, Indicates task A set of candidate collaborative solutions; Indicates task The corresponding number of candidate collaborative solutions; Indicates task The There are several candidate coordination schemes. Each candidate coordination scheme consists of several member satellite execution segments, and each member satellite execution segment may include member satellite identifiers. Start time End time ,reliability ,load and capacity usage .in, Indicates the sequence number of the member satellite execution segment in the candidate collaborative scheme; Indicates member satellites.
[0054] Step S4: Based on the candidate collaboration scheme set and local state of each satellite cluster, perform intra-cluster optimization to determine the task to be undertaken and the corresponding candidate collaboration scheme, and generate a tender document including task, collaboration scheme, execution resources and constraint feasibility information.
[0055] In an optional embodiment of the present invention, step S4 may include: Step S41: Discrete index encoding is performed on the candidate collaborative schemes for each task and the case where the corresponding task is not selected to obtain the optimized individual.
[0056] Specifically, let the task set be... ,in This indicates the total number of tasks. The corresponding set of candidate collaborative solutions is Using discrete index variables Indicates task The scheme selection result is represented by the discrete index space as follows: .when When this occurs, it indicates that the current satellite cluster is not selecting a mission. ;when When, it indicates that a task has been selected. The One candidate collaborative solution An optimized individual is represented as ,in This represents an optimized entity composed of discrete indexes for each task; Indicates task The corresponding candidate collaborative scheme index. The adaptive Harris Eagle optimization process in this embodiment can be denoted as EAD-HHO. For discrete index variables in EAD-HHO.
[0057] Step S42: Based on the local state of each satellite cluster and the mission coverage, mission benefits, mission priority, mission timeliness, mission reliability, load balancing and constraint feasibility, perform iterative search on the optimization individuals to determine the missions to be undertaken and the corresponding candidate collaboration schemes.
[0058] Specifically, binary decision variables can be used. Describe the selection state of candidate collaborative solutions when the task The When a candidate collaborative scheme is selected ,otherwise The task uniqueness constraint is expressed as: in, Indicates task The The selection status of each candidate collaborative solution; Indicates task The corresponding number of candidate collaborative solutions; This represents a set of tasks.
[0059] When candidate collaborative schemes and candidate collaborative solutions Includes satellites of the same member The execution segment, and does not satisfy or At that time, the two candidate coordination schemes were determined to have timing conflicts among member satellites, and limitations were imposed. .in, and They represent candidate collaborative schemes respectively. The Middle The start and end times of each execution segment; and They represent candidate collaborative schemes respectively. The Middle The start and end times of each execution segment; This indicates the minimum safe time interval required for the same member satellite to perform two consecutive missions.
[0060] Member satellite capacity constraints are expressed as follows: in, Indicates member satellites The initial occupied capacity; Indicates task The Each candidate collaboration scheme for member satellites Capacity usage; Indicates member satellites Capacity limits within the current planning cycle.
[0061] Based on mission coverage, mission benefits, mission priority, mission timeliness, mission reliability, load balancing, member satellite timing conflicts, and capacity exceedance, a comprehensive fitness function is constructed. The comprehensive fitness function can be expressed as: in, Indicates overall fitness; Indicates task coverage; Indicates the rewards for completing the task; Indicates task priority; Indicates the timeliness of the task; Indicates average reliability; This indicates a penalty for unbalanced load. Indicates conflict punishment; This indicates a penalty for exceeding the capacity limit; to This indicates the weight of the corresponding evaluation item, and the weight can be configured according to the task scenario.
[0062] Task coverage can be expressed as: in, Indicates the combination of schemes to be evaluated; Indicates the number of tasks completed; This indicates the total number of tasks.
[0063] The task reward can be expressed as: in, Indicates task Select the The corresponding revenue value for each candidate collaborative solution; This indicates the total number of tasks.
[0064] Task priority can be expressed as: in, Indicates task Select the The priority weights corresponding to each candidate collaborative scheme.
[0065] Task timeliness can be expressed as: in, Indicates task Select the The time satisfaction of each candidate collaborative solution.
[0066] Task reliability can be expressed as: in, Indicates task The The reliability evaluation value of each candidate collaborative scheme.
[0067] Load balancing metrics can be expressed as: in, Indicates member satellites The workload it undertakes; Indicates the standard deviation of the load; This indicates the maximum allowed load.
[0068] The conflict penalty index can be expressed as: in, This indicates the number of task combinations that have time conflicts; This indicates the total number of task combinations.
[0069] The capacity over-limit penalty index can be expressed as: in, Indicates member satellites Corresponding task load; Indicates member satellites The maximum capacity.
[0070] Specifically, an improved Harris Eagle algorithm is used to iteratively search for optimal individuals. Multiple optimal individuals are randomly initialized to form a population, and each individual is evaluated based on a comprehensive fitness function. The consecutive search positions updated by the Harris Eagle algorithm are represented as follows: Regarding the task The corresponding number of candidate collaborative solutions The continuous search positions are mapped to discrete candidate solution indices, and the mapping relationship is expressed as follows: in, Represents a vector of consecutive search positions; Indicates task The corresponding consecutive search positions; and These represent the lower and upper limits of the consecutive search positions, respectively; This represents the index of the discrete candidate scheme after mapping; Indicates task The number of candidate collaborative solutions.
[0071] The exploration probability is adjusted based on the iteration progress. The iteration progress can be represented as... The exploration probability can be expressed as: in, Indicates the first Iteration progress at the next iteration; Indicates the current iteration number; Indicates the maximum number of iterations; Indicates the current exploration probability; Indicates the initial exploration probability; This represents the final probability corresponding to the development stage; This represents the adjustment coefficient. When... During the exploration phase, when During the development phase.
[0072] After each iteration, the optimal individuals in the population are ranked according to their overall fitness. From these top-ranked individuals, a select set of elite individuals is formed. This elite set can be represented as: The number of elites can be expressed as: in, It represents a collection of elites; to This refers to the optimized individuals selected for inclusion in the elite set; Indicates the number of elites; Indicates population size; This represents the adjustment coefficient used to determine the number of elites.
[0073] Specifically, during the development phase, guiding individuals are selected from the elite pool. Furthermore, by guiding individuals to adjust their current search direction, guided updates can be represented as: in, Indicates the consecutive search positions after the guide update; Indicates the current continuous search position; This refers to the guiding individual selected from the elite set; This represents the guiding coefficient.
[0074] During the development phase, the replacement priority of task locations is determined based on task benefits, conflict risks, and resource consumption. Replacement priority can be expressed as: The location of the task to be replaced can be represented as: in, Indicates the task location Replacement priority; Indicates the task location Corresponding revenue evaluation value; Indicates the task location Corresponding conflict risks; Indicates the task location The corresponding resource consumption evaluation value; , and Indicates the weight of the corresponding evaluation item; This indicates the position of the task with the highest priority to be replaced.
[0075] Specifically, elite individuals are selected from the elite set as guiding sources to partially replace candidate collaborative solutions corresponding to the location of the task to be replaced. When the replaced solution satisfies... If the replacement of the candidate cooperative solution can eliminate the infeasibility constraint, then the replacement result is accepted. This represents the optimized individual after the replacement; This represents the optimized individual before the replacement; This represents the comprehensive fitness function.
[0076] The replacement results are verified for task uniqueness, member satellite timing conflicts, and capacity constraints. When constraint conflicts exist, conflict resolution is performed sequentially based on task priority, overall utility, time feasibility, and resource consumption, prioritizing candidate collaborative solutions with higher task priority, higher overall utility, better time feasibility, or lower resource consumption. The optimal index vector is output after reaching the maximum number of iterations or meeting the convergence condition. and will Decoding involves identifying the tasks to be undertaken and the corresponding candidate collaboration solutions.
[0077] Step S43: Determine the execution sequence, execution time, payload type, resource consumption, expected benefits, reliability, load status, capacity status, and constraint feasibility information of member satellites based on the proposed mission and the corresponding candidate collaborative schemes.
[0078] Specifically, based on the member satellite execution segments in the candidate collaborative schemes corresponding to the proposed mission, the execution sequence of the member satellites and the start and end times of each execution segment are determined, and the required payload, resource consumption, expected benefits, average reliability, load variance, remaining capacity, and constraint feasibility markers are determined based on the corresponding candidate collaborative schemes.
[0079] Step S44: Determine the alternative schemes based on the other feasible candidate collaboration schemes corresponding to the proposed task, and generate the tender document based on the member satellite execution sequence, execution time, payload type, resource consumption, expected benefits, reliability, load status, capacity status, constraint feasibility information and alternative schemes.
[0080] Specifically, for satellite clusters whose missions have been determined, feasible candidate coordination schemes that still meet the mission and resource constraints, other than the currently selected scheme, are retained from the candidate coordination scheme set for the corresponding mission as backup schemes. The satellite cluster identifier, mission and candidate scheme index, member satellite execution sequence, start and end times of each execution segment, payload type, resource consumption, expected benefits, average reliability, load variance, remaining capacity, constraint feasibility markers, and backup schemes are organized into a tender document. After each satellite cluster completes intra-cluster optimization, the tender document is submitted to the ground mission management center.
[0081] Step S5 involves cross-cluster evaluation, resolving duplicate tasks, and repairing resource conflicts based on the tender documents submitted by multiple satellite clusters. This process determines the winning satellite clusters and collaborative solutions for each task and generates a global collaborative task plan.
[0082] In an optional embodiment of the present invention, step S5 may include: Step S51: Merge the bid documents according to the task identifier, and determine the winning utility of each bid result based on task priority, bid benefits, execution time, reliability, resource consumption, load balancing and default risk.
[0083] Specifically, the ground mission management center receives bids submitted by each satellite cluster and merges the bids according to mission identifiers to identify duplicate bids submitted by multiple satellite clusters for the same mission. For each bid, a comprehensive evaluation is conducted based on mission priority, bid benefits, execution timeliness, average reliability, resource consumption, load balancing, and default risk to determine the corresponding winning bid utility.
[0084] Step S52: Determine the winning bid results from the bidding results corresponding to the same task based on the winning bid utility, repair resource conflicts between the determined winning bid results, determine the winning satellite clusters and winning collaboration schemes for each task, and generate a global collaborative task plan.
[0085] Specifically, for multiple feasible bids for the same task, the winning bid is determined first based on the utility of the bid. When the utility of multiple bids is close, the winning bid is further determined based on execution time, resource consumption, and reliability. The selection can also be made by considering the impact of candidate bids on subsequent tasks, thus obtaining the initial winning bid.
[0086] The initial bidding results are used to verify resource conflicts related to cross-cluster sharing of satellites, shared links, and shared ground stations. When resource conflicts exist among multiple winning bids, the winning bid with higher utility is retained, while the winning bid with lower utility is removed. Alternative solutions are then read from the bid documents corresponding to the removed bids, and alternative solutions that meet both mission and resource constraints are selected from the alternative solutions.
[0087] The initial bidding results are updated based on alternative solutions that meet the constraints; when none of the candidate solutions in the corresponding bid documents meet the task constraints or resource constraints, the corresponding task is re-bid. After completing the resolution of duplicate tasks and the repair of resource conflicts, the winning satellite clusters and winning collaborative schemes for each task are determined, forming a global collaborative task plan.
[0088] The winning bid results in the global collaborative mission plan can include the mission identifier, winning satellite cluster, candidate scheme number, participating satellites, execution sequence, command type, and constraint verification results. After the ground mission management center distributes the winning bid results to the winning satellite cluster, the corresponding cluster management node determines the onboard execution commands for the member satellites based on the winning collaborative scheme.
[0089] This invention also provides a single-type large-scale heterogeneous constellation hierarchical task planning system for executing the above method, including: The task announcement generation module is used to perform semantic parsing, constraint extraction and standardization of task requirements to obtain task announcements, and send the task announcements to the corresponding satellite clusters; The local state maintenance module is used to divide a large heterogeneous constellation into multiple satellite clusters and obtain the local state of each satellite cluster based on the resource status of the member satellites and the planned tasks within each satellite cluster. The candidate collaboration scheme generation module is used to convert the executable time information of member satellites for the mission into meta-tasks based on the mission announcement and the local state, and to combine the meta-tasks according to collaboration rules and mission constraints to obtain a set of candidate collaboration schemes for each mission. The intra-cluster optimization module is used to perform intra-cluster optimization based on the candidate coordination scheme set and local state of each satellite cluster, determine the tasks to be undertaken and the corresponding candidate coordination schemes, and generate a tender document that includes task, coordination scheme, execution resources and constraint feasibility information. The global evaluation module is used to conduct cross-cluster evaluation, resolve duplicate tasks, and repair resource conflicts based on the bids submitted by multiple satellite clusters, determine the winning satellite clusters and winning collaborative schemes for each task, and generate a global collaborative task plan.
[0090] Although the present invention has been disclosed above with reference to embodiments, it is not intended to limit the present invention. Appropriate modifications or equivalent substitutions made by those skilled in the art to the technical solutions of the present invention should be covered within the protection scope of the present invention, which is defined by the claims.
Claims
1. A hierarchical task planning method for a large-scale heterogeneous constellation with single-unit dispatch, characterized in that, Includes the following steps: The task requirements are semantically parsed, constraints are extracted and standardized to obtain the task announcement, and the task announcement is sent to the corresponding satellite cluster; The large heterogeneous constellation is divided into multiple satellite clusters, and the local state of each satellite cluster is obtained based on the resource status of the member satellites and the planned missions within each cluster. Based on the mission announcement and the local state, the executable time information of the member satellites for the mission is converted into meta-missions, and the meta-missions are combined according to the coordination rules and mission constraints to obtain a set of candidate coordination schemes for each mission. Based on the candidate collaboration scheme set and local state of each satellite cluster, intra-cluster optimization is performed to determine the proposed task and the corresponding candidate collaboration scheme, and a tender document including task, collaboration scheme, execution resources and constraint feasibility information is generated; Based on the bids submitted by multiple satellite clusters, cross-cluster evaluation, duplicate task resolution, and resource conflict repair are carried out to determine the winning satellite clusters and winning collaboration schemes for each task, and generate a global collaborative task plan.
2. The method as described in claim 1, characterized in that, The task requirements are semantically parsed, constraints extracted, and standardized to obtain the task announcement, which is then sent to the corresponding satellite cluster, including: Receive at least one of natural language task description and structured task data, as well as task planning data, and call a large language model to perform semantic parsing, constraint extraction, anomaly detection and standardization on the task requirements to obtain standardized task information; A mission announcement is generated based on the standardized mission information, and the mission announcement is sent to a satellite cluster with mission execution capabilities.
3. The method as described in claim 1, characterized in that, Based on the mission announcement and the local state, the executable time information of member satellites for the mission is converted into meta-missions. These meta-missions are then combined according to coordination rules and mission constraints to obtain a set of candidate coordination schemes for each mission, including: Based on the mission announcement and the local state, the visible time window of the member satellite for the mission is converted into a meta-mission, and the meta-mission is associated with the mission, member satellite, payload type, executable time and resource usage information; Based on the collaboration rules and task constraints, the meta-tasks are constrained, matched, and combined to obtain a set of candidate collaboration schemes for each task.
4. The method as described in claim 3, characterized in that, Based on the collaboration rules and task constraints, the meta-tasks are constrained, matched, and combined to obtain a set of candidate collaboration schemes for each task, including: Based on the coordination rules and task constraints, determine the temporal relationships between atomic tasks, as well as load matching, time intervals, link reachability, and resource capacity constraints, and construct a coordination rule graph. The meta-tasks are filtered and combined according to the collaborative rule graph to obtain a set of candidate collaborative schemes that satisfy the temporal relationship, load matching, time interval, link reachability and resource capacity constraints.
5. The method as described in claim 4, characterized in that, The meta-tasks are filtered and combined according to the collaborative rule graph to obtain a set of candidate collaborative schemes that satisfy the temporal relationship, load matching, time interval, link reachability, and resource capacity constraints, including: The mission, member satellites, and the meta-mission are represented as RDF resources, and the time information, attribution relationship, execution relationship, and payload type information are represented as RDF predicates to construct an RDF triple graph. Constraint query conditions are generated based on the collaborative rule graph, and SPARQL query is performed on the RDF triple graph to obtain the candidate collaborative scheme set.
6. The method as described in claim 1, characterized in that, Based on the candidate coordination scheme set and local state of each satellite cluster, intra-cluster optimization is performed to determine the proposed task and corresponding candidate coordination scheme, generating a tender document that includes task, coordination scheme, execution resources, and constraint feasibility information, including: Discrete index encoding is performed on the candidate collaborative schemes for each task and the case where the corresponding task is not selected to obtain the optimized individual; Based on the local state of each satellite cluster and the mission coverage, mission benefits, mission priority, mission timeliness, mission reliability, load balancing, and constraint feasibility, the optimization individual is iteratively searched to determine the mission to be undertaken and the corresponding candidate collaboration scheme. Based on the proposed mission and the corresponding candidate collaboration schemes, determine the member satellite execution sequence, execution time, payload type, resource consumption, expected benefits, reliability, load status, capacity status, and constraint feasibility information; Based on the other feasible candidate collaborative schemes corresponding to the proposed task, a backup scheme is determined, and a tender document is generated based on the member satellite execution sequence, execution time, payload type, resource consumption, expected benefits, reliability, load status, capacity status, constraint feasibility information, and the backup scheme.
7. The method as described in claim 6, characterized in that, Based on the local state of each satellite cluster and considering factors such as mission coverage, mission benefits, mission priority, mission timeliness, mission reliability, load balancing, and constraint feasibility, an iterative search is performed on the optimized individuals to determine the tasks to be undertaken and the corresponding candidate collaborative schemes, including: The optimized individuals are evaluated based on task coverage, task benefits, task priority, task timeliness, task reliability, load balancing, and constraint feasibility to obtain a comprehensive evaluation result, and an elite set is determined based on the comprehensive evaluation result. The exploration probability is determined based on the iteration progress, the exploration stage or development stage is determined based on the exploration probability, and the Harris Eagle optimization algorithm is used to update the continuous search position corresponding to the optimized individual, and the updated continuous search position is mapped to the discrete candidate solution index. During the development phase, the discrete candidate scheme index is updated based on the elite set, and local candidate replacement and constraint repair are performed to obtain the task to be undertaken and the corresponding candidate collaborative scheme.
8. The method as described in claim 1, characterized in that, Based on the bids submitted by multiple satellite clusters, cross-cluster evaluation, task duplication resolution, and resource conflict remediation are conducted to determine the winning satellite clusters and collaborative solutions for each task, and to generate a global collaborative task plan, including: The bid documents are merged according to the task identifier, and the winning utility of each bid result is determined based on task priority, bid benefits, execution time, reliability, resource consumption, load balancing and default risk. Based on the winning bid effect, the winning bid results are determined from the bidding results corresponding to the same task, and resource conflicts between the determined winning bid results are repaired. The winning satellite clusters and winning bid coordination schemes for each task are determined, and a global collaborative task plan is generated.
9. The method as described in claim 8, characterized in that, Based on the winning bid effect, the winning bid results are determined from the bid results corresponding to the same task, and resource conflicts between the determined winning bid results are repaired. The winning satellite clusters and winning cooperation schemes for each task are determined, and a global collaborative task plan is generated, including: Based on the aforementioned bid effectiveness, execution time, resource consumption, and reliability, the winning bid result is determined from the feasible bid results corresponding to the same task, and the initial winning bid result is obtained. The resource conflicts of shared satellite, shared link and shared ground station in the initial bidding results are verified, the bidding results with resource conflicts are discarded, the alternative solutions are obtained from the bid documents corresponding to the rejected bid results, and the alternative solutions that meet the constraints are determined from the alternative solutions. The initial bidding results are updated based on the alternative solutions, and tasks for which there are no alternative solutions that meet the constraints are re-bid to obtain a global collaborative task plan.
10. A hierarchical task planning system for a large-scale heterogeneous constellation with single-unit dispatch, characterized in that, include: The task announcement generation module is used to perform semantic parsing, constraint extraction and standardization of task requirements to obtain task announcements, and send the task announcements to the corresponding satellite clusters; The local state maintenance module is used to divide a large heterogeneous constellation into multiple satellite clusters and obtain the local state of each satellite cluster based on the resource status of the member satellites and the planned tasks within each satellite cluster. The candidate collaboration scheme generation module is used to convert the executable time information of member satellites for the mission into meta-tasks based on the mission announcement and the local state, and to combine the meta-tasks according to collaboration rules and mission constraints to obtain a set of candidate collaboration schemes for each mission. The intra-cluster optimization module is used to perform intra-cluster optimization based on the candidate coordination scheme set and local state of each satellite cluster, determine the tasks to be undertaken and the corresponding candidate coordination schemes, and generate a tender document that includes task, coordination scheme, execution resources and constraint feasibility information. The global evaluation module is used to conduct cross-cluster evaluation, resolve duplicate tasks, and repair resource conflicts based on the bids submitted by multiple satellite clusters, determine the winning satellite clusters and winning collaborative schemes for each task, and generate a global collaborative task plan.
