Large Language Model-Based Star Cluster Human Intelligence Collaborative Management Method and System

Abstract

A method and system for collaborative control of star cluster human-intelligence based on large language models, which relates to the field of on-orbit satellite management. It solves the problem that existing star clusters rely on manual management and urgently need high-timeliness, integrated, and intelligent collaborative control of star clusters. The system includes: an intelligent interpretation and response unit, a very large-scale star cluster task planning unit, and a very large-scale star cluster high-efficiency simulation and deduction unit; the intelligent interpretation and response unit obtains text information and extracts keywords through text extraction, key element extraction, and event detection of data; obtains the information theme category, event type, and corresponding command content according to the keywords, and obtains a task planning set according to the command content; the very large-scale star cluster task planning unit performs collaborative task planning and scheduling according to the task planning set, and outputs the scheduling result to the very large-scale star cluster high-efficiency simulation and deduction unit; receives the scheduling result and conducts real-time simulation. It is applied to the field of artificial intelligence collaborative control of star clusters.

Description

Technical Field

[0001] The present invention relates to the field of on-orbit satellite management, and particularly to a method for collaborative control of satellite constellations based on large language models for human intelligence. Background Art

[0002] In recent years, large-scale constellation networks represented by Starlink and OneWeb have been under construction and have initially formed capabilities.

[0003] Currently, the management of on-orbit satellites is mainly carried out by ground stations issuing specific instructions. When satellite users initiate task requirements, ground engineers manually check the current satellite platform status, payload status, and task progress status according to the task requirements, and feedback task information to users in the form of statistical data or reports. When users initiate collaborative task requirements, ground engineers need to perform task planning and solution, and manually formulate a satellite constellation collaboration plan by combining relevant domain knowledge and experience. The advantage of this management method is that both the technical solution and the management mode are very mature, and the control ability of satellites is strong. However, with the increase in the number of satellites and their wide distribution, relying solely on manual labor, the control work will become extremely heavy, wasting a great deal of material and human resources, and significantly increasing the control time, making it unable to cope with the rapidly changing space environment in the future.

[0004] The main difference between ultra-large-scale satellite constellations and large-scale constellations is that the number of satellites in ultra-large-scale satellite constellations will exceed ten thousand, while the number of satellites in constellations is relatively small, only a few hundred. Therefore, the control of constellations is mainly reflected in the task application level, and the control method is relatively mature. Due to the large-scale increase in the number of satellites in ultra-large-scale constellations, the control information will also increase exponentially, including but not limited to massive information such as satellite status, attitude and orbit parameters, coverage characteristics, communication links, payload status, backup orbits, enemy status, collision probability, task status, and user attribution. When facing massive data, the decision-making layer needs computer assistance for processing. Therefore, there is an urgent need to develop remote sensing satellite constellation control technology based on the dialogue mode.

[0005] In 2022, the advent of ChatGPT has widely expanded the potential application scope of large models in the military field. In addition to the above applications, it may also be used for automatic target recognition and surveillance, battlefield environment support, driverless vehicles, drone systems and autonomous weapon system control, massive military data management, situation awareness, path planning, and so on. Palantir is launching the Palantir Artificial Intelligence Platform (AIP), which is designed to run large language models such as GPT-4 and alternatives on private networks. In one of its promotional videos, Palantir demonstrated how the military can use AIP to wage war. AIP allows operators to quickly obtain information such as the location, equipment, and intentions of enemy forces by interacting with a chatbot and formulate multiple attack plans, including using drones for reconnaissance, interfering with enemy communications, and launching missiles. AIP can also provide operators with real-time feedback, explanations, and suggestions, as well as the ability to monitor and control AI agents and other AI activities. The AIP artificial intelligence platform has brought about great changes to the defense and military fields by using large models.

[0006] High-timeliness, integrated, and intelligent constellation human-machine collaborative management and control is the future trend. There is an urgent need to conduct research on self-developed remote sensing constellation human-machine collaborative management and control technology. Its goal is to be task- and demand-oriented, intelligently and efficiently coordinate satellite resources and time, give full play to the capabilities of complex satellite systems, integrate massive commercial, defense, and military data, manage the data accessible to each model, improve the data usage efficiency of the model, provide detailed spatial information and perfect action options for commanders, optimize the decision-making process and shorten the decision-making time, and assist commanders in monitoring the entire mission process. Summary of the Invention

[0007] In view of the problem that the existing constellations rely on manual management and there is an urgent need for high-timeliness, integrated, and intelligent constellation human-machine collaborative management and control, the present invention proposes a constellation human-machine collaborative management and control system based on a large language model. The system includes:

[0008] An intelligent interpretation and response unit, a super-large-scale constellation mission planning unit, and a super-large-scale constellation high-efficiency simulation and deduction unit;

[0009] The intelligent interpretation and response unit interprets and identifies based on aerospace professional knowledge, obtains the text information of the dialogue input through text extraction, key element extraction, and event detection of data, and extracts keywords; obtains the information theme category, event type, and corresponding command content according to the keywords, and obtains a mission planning set according to the command content;

[0010] The ultra-large scale satellite constellation mission planning unit is used to traverse all remote sensing satellites according to the mission planning set obtained by the intelligent interpretation response module, conduct collaborative mission planning and scheduling for the traversed satellite constellation, and output the scheduling result to the ultra-large scale satellite constellation high-efficiency simulation and deduction unit;

[0011] The ultra-large scale satellite constellation high-efficiency simulation and deduction unit receives the scheduling result and conducts real-time simulation.

[0012] Furthermore, a preferred method is also provided. The intelligent interpretation response unit includes an inquiry task module, a command and control task module, and a decision-making task module;

[0013] The inquiry task module is used to match and query the input keywords with the massive data in the ultra-large scale satellite constellation high-efficiency simulation and deduction unit, obtain user inquiry information from satellite status, attitude and orbit parameters, coverage characteristics, communication links, payload status, enemy status, collision probability, mission status, and user attribution, and conduct intelligent reply;

[0014] The command and control task module is used to decompose the input keywords to obtain the standard instruction form, and use it as the command and control input to the ultra-large scale satellite constellation high-efficiency simulation and deduction module, and drive the attitude and orbit control module, rocket module, thruster module, antenna module, camera module, space environment module, measurement and control coverage module, navigation module, sensor module, and routing calculation module in the ultra-large scale satellite constellation high-efficiency simulation and deduction unit to conduct real-time simulation;

[0015] The decision-making task module decomposes the input keywords into a mission planning set, including an observation resource set, an observation target set, and a constraint set.

[0016] Furthermore, a preferred method is also provided. The massive data in the ultra-large scale satellite constellation high-efficiency simulation and deduction unit includes: satellite status, attitude and orbit parameters, coverage characteristics, communication links, payload status, enemy status, collision probability, mission status, and user attribution.

[0017] Furthermore, a preferred method is also provided. The standard instruction includes: execution time, instruction ID, instruction execution component name, mission name, instruction name, and instruction content.

[0018] Furthermore, a preferred method is also provided. The execution time is the UTC time, and the format is: year-month-day-hour-minute-second.

[0019] Furthermore, a preferred method is also provided. The mission planning process of the ultra-large scale satellite constellation mission planning unit is specifically as follows:

[0020] Traverse all remote sensing satellites;

[0021] Forecast the orbit of the remote sensing satellite during the mission time according to the traversal result, and calculate the attitude of the remote sensing satellite and the camera pointing to obtain the observable area;

[0022] Traverse all target sets and calculate the visibility of the remote sensing satellite to the targets;

[0023] Judge according to the visibility of the remote sensing satellite to the targets for the constraint conditions input by the user, and screen out the list of available observation satellites;

[0024] Use the mission planning method to screen out the optional observation plans in the observation list, and feedback the plans in text form through the intelligent interpretation response unit;

[0025] Make a decision selection and confirmation according to the feedback result, and after confirming the decision, feedback the strategy to the simulation system in the form of command for real-time simulation.

[0026] Furthermore, a preferred method is also provided. The step of using the mission planning method to screen out the optional observation plans in the observation list includes:

[0027] Divide the observation tasks of the satellite constellation into continuous observation tasks and discontinuous observation tasks. Among them, for discontinuous observation tasks, only one optimal observation satellite needs to be selected to execute the task, and for continuous observation tasks, the observation sequence needs to be optimized and sorted.

[0028] Based on the same inventive concept, the present invention also proposes a satellite constellation human-machine collaborative control method based on a large language model. The method is implemented based on the system described in any one of the above, and the method includes:

[0029] When the user's demand is input in the form of a text dialogue, the intelligent interpretation response unit extracts key elements from the dialogue history, aerospace domain knowledge, and preset key elements, and classifies the user's demand according to the key elements;

[0030] After classification, extract keywords, obtain the information theme category, event type, and corresponding command content according to the keywords, and obtain the mission planning set according to the command content;

[0031] The ultra-large-scale satellite constellation mission planning unit traverses all remote sensing satellites according to the mission planning set obtained by the intelligent interpretation response module, conducts collaborative mission planning and scheduling for the traversed satellite constellation, and outputs the scheduling result to the ultra-large-scale satellite constellation high-efficiency simulation and deduction unit;

[0032] The ultra-large-scale satellite constellation high-efficiency simulation and deduction unit receives the scheduling result for real-time simulation.

[0033] Based on the same inventive concept, the present invention also provides a computer device, including a memory and a processor. A computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the method for star cluster human-intelligence collaborative management and control based on the large language model as described above.

[0034] A computer-readable storage medium is used to store a computer program, and the computer program executes the method for star cluster human-intelligence collaborative management and control based on the large language model in any one of the above.

[0035] The beneficial effects of the present invention are as follows:

[0036] The present invention solves the problem of urgently needing high-efficiency, integrated, and intelligent star cluster human-intelligence collaborative management and control.

[0037] For the star cluster human-intelligence collaborative management and control system based on the large language model proposed by the present invention, through the intelligent interpretation and response unit, the system can utilize the text analysis ability of the large language model to quickly and accurately interpret and respond to the dialogue input, so as to achieve efficient star cluster management. The ultra-large-scale star cluster task planning unit traverses all remote sensing satellites and uses the task planning set obtained by the intelligent interpretation and response unit to realize the collaborative task planning and scheduling of the star cluster, enabling the entire system to have integrated management capabilities. The ultra-large-scale star cluster high-efficiency simulation and deduction unit can receive the scheduling results of the task planning unit and perform real-time star cluster simulation and deduction, so that the system can evaluate and optimize the task planning in a virtual environment and improve the efficiency of star cluster management and control. The intelligent interpretation and response unit performs interpretation and recognition based on aerospace professional knowledge. Through text extraction of data, key element extraction, and event detection, the system can more accurately understand and process information related to star cluster management. The system can continuously analyze the dialogue input and extract key information to quickly respond to real-time changing situations, making the star cluster management more flexible.

[0038] In a star cluster human-intelligence collaborative management and control system based on a large language model proposed by the present invention, the intelligent interpretation and response unit uses the large language model and aerospace professional knowledge to interpret the text information input in the conversation, extract keywords, analyze the information theme category and event type, obtain the corresponding accusation content, and finally obtain a task planning set. The ultra-large-scale star cluster task planning unit traverses all remote sensing satellites according to the task planning set provided by the intelligent interpretation and response unit, realizes the collaborative task planning and scheduling of the star cluster, and outputs the scheduling result to the ultra-large-scale star cluster high-efficiency simulation and deduction unit. The ultra-large-scale star cluster high-efficiency simulation and deduction unit receives the scheduling result of the task planning unit, evaluates the performance and effectiveness of the star cluster through real-time simulation and deduction, provides feedback information for the system, and supports the optimization and adjustment of decision-making. The purpose of this system is to solve the problem of existing star cluster manual management. By introducing a large language model, the star cluster management becomes more efficient, integrated, and intelligent. Through the collaborative planning, real-time simulation and deduction, and intelligent interpretation of the system, it aims to improve the overall management level of the star cluster and ensure that the star cluster can effectively and flexibly execute tasks in a complex environment. The design goal of the system is to make the star cluster management more adaptable to the modern needs through an automated and intelligent way, and improve the efficiency of satellite task planning, collaboration, and execution.

[0039] The present invention is applied to the field of star cluster artificial intelligence collaborative management and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a flowchart of the star cluster human-intelligence collaborative management and control method based on the large language model described in Embodiment 1;

[0041] Figure 2 It is a flowchart of the control instruction generation and feedback of the intelligent interpretation and response unit described in Embodiment 2;

[0042] Figure 3 It is a task planning flowchart of the ultra-large-scale star cluster task planning unit described in Embodiment 6. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] To make the purpose, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, embodiments of the present invention.

[0044] Embodiment 1. Refer to Figure 1 This embodiment will be described. The star cluster human-intelligence collaborative management and control system based on the large language model described in this embodiment, the system includes:

[0045] An intelligent interpretation and response unit, an ultra-large-scale star cluster task planning unit, and an ultra-large-scale star cluster high-efficiency simulation and deduction unit;

[0046] The intelligent interpretation and response unit interprets and identifies based on aerospace professional knowledge. By extracting text from data, extracting key elements, and detecting events, it obtains the text information of the dialogue input and extracts keywords. Based on the keywords, it obtains the information theme category, event type, and corresponding accusation content, and obtains the task planning set according to the accusation content.

[0047] The ultra-large-scale satellite constellation task planning unit is used to traverse all remote sensing satellites according to the task planning set obtained by the intelligent interpretation and response module, perform collaborative task planning and scheduling on the traversed satellite constellation, and output the scheduling result to the ultra-large-scale satellite constellation high-efficiency simulation and deduction unit.

[0048] The ultra-large-scale satellite constellation high-efficiency simulation and deduction unit receives the scheduling result for real-time simulation.

[0049] In this embodiment, through the intelligent interpretation and response unit, the system can utilize the text analysis ability of the large language model to quickly and accurately interpret and respond to the dialogue input, thereby realizing efficient satellite constellation management. The ultra-large-scale satellite constellation task planning unit traverses all remote sensing satellites and utilizes the task planning set obtained by the intelligent interpretation and response unit to achieve collaborative task planning and scheduling of the satellite constellation, enabling the entire system to have an integrated management ability. The ultra-large-scale satellite constellation high-efficiency simulation and deduction unit can receive the scheduling result of the task planning unit and perform real-time satellite constellation simulation and deduction, so that the system can evaluate and optimize the task planning in a virtual environment and improve the efficiency of satellite constellation control. The intelligent interpretation and response unit interprets and identifies based on aerospace professional knowledge. By extracting text from data, extracting key elements, and detecting events, the system can more accurately understand and process information related to satellite constellation management. The system can quickly respond to real-time changing situations by continuously parsing the dialogue input and extracting key information, making the satellite constellation management more flexible.

[0050] In this embodiment, the intelligent interpretation and response unit uses a large language model to interpret the text information input in the dialogue by leveraging aerospace expertise, extract keywords, analyze the information theme category and event type, obtain the corresponding accusation content, and finally obtain the mission planning set. The ultra-large-scale satellite constellation mission planning unit traverses all remote sensing satellites according to the mission planning set provided by the intelligent interpretation and response unit to achieve collaborative mission planning and scheduling of the satellite constellation, and outputs the scheduling result to the ultra-large-scale satellite constellation high-efficiency simulation and deduction unit. The ultra-large-scale satellite constellation high-efficiency simulation and deduction unit receives the scheduling result of the mission planning unit, evaluates the performance and effectiveness of the satellite constellation through real-time simulation and deduction, provides feedback information for the system, and supports the optimization and adjustment of decisions. The purpose of this system is to solve the problem of manual management of existing satellite constellations. By introducing a large language model, the management of satellite constellations becomes more efficient, integrated, and intelligent. Through the collaborative planning, real-time simulation and deduction, and intelligent interpretation of the system, it aims to improve the overall management level of the satellite constellation and ensure that the satellite constellation can effectively and flexibly execute tasks in a complex environment. The design goal of the system is to make the management of satellite constellations more adaptable to modern requirements through automated and intelligent means, and improve the efficiency of satellite mission planning, collaboration, and execution.

[0051] Embodiment 2. See Figure 2 This embodiment is described. This embodiment further limits the satellite constellation human-machine collaborative control system based on a large language model described in Embodiment 1. The intelligent interpretation and response unit includes an inquiry task module, an accusation task module, and a decision-making task module;

[0052] The inquiry task module is used to match and query the input keywords with the massive data in the ultra-large-scale satellite constellation high-efficiency simulation and deduction unit, obtain the user inquiry information from the satellite status, attitude and orbit parameters, coverage characteristics, communication link, payload status, enemy status, collision probability, mission status, and user attribution, and make an intelligent response;

[0053] The accusation task module is used to decompose the input keywords to obtain the standard instruction form, and use it as an accusation input to the ultra-large-scale satellite constellation high-efficiency simulation and deduction module, and drive the attitude and orbit control module, rocket module, thruster module, antenna module, camera module, space environment module, TT&C coverage module, navigation module, sensor module, and routing calculation module in the ultra-large-scale satellite constellation high-efficiency simulation and deduction unit to perform real-time simulation;

[0054] The decision-making task module decomposes the input keywords into a mission planning set, including an observation resource set, an observation target set, and a constraint set.

[0055] In this embodiment, the inquiry task module obtains information related to the user's keywords from multiple aspects by matching and querying a large amount of data, enabling intelligent responses to the user's inquiries and improving the interaction efficiency between the system and the user. The accusation task module converts the keywords input by the user into a standard instruction form and drives the large-scale constellation high-efficiency simulation and deduction unit for real-time simulation. This helps to simulate the actual behavior of the constellation in a virtual environment and provide more intuitive and real-time information support for the user.

[0056] Through the decomposition and transformation of the accusation task module, real-time simulation driving of multiple modules (such as attitude and orbit control module, rocket module, thruster module, etc.) in the large-scale constellation high-efficiency simulation and deduction unit is achieved, making the functions of the system more comprehensive and complete. The intelligent interpretation module decomposes the keywords input by the user into a task planning set, including an observation resource set, an observation target set, and a constraint set, which helps the system to more comprehensively understand and plan constellation tasks.

[0057] This embodiment further refines the functions of the system, enabling the system to more accurately respond to the user's inquiries, execute the user's accusations, and support decision-making tasks through real-time simulation. By mapping the keywords input by the user to specific constellation data and task planning sets, the system more comprehensively understands the user's needs, improving the intelligence, real-time performance, and multi-module comprehensive support of the constellation management and control system. The overall goal is to improve the management efficiency of the system, making constellation management more intelligent, flexible, and adaptable.

[0058] Embodiment 3: This embodiment further defines the large language model-based constellation human-machine collaborative management and control system described in Embodiment 2. The massive data in the large-scale constellation high-efficiency simulation and deduction unit includes: satellite status, attitude and orbit parameters, coverage characteristics, communication links, payload status, enemy status, collision probability, task status, and user attribution.

[0059] In this embodiment, by integrating massive data in multiple aspects such as satellite status, attitude and orbit parameters, coverage characteristics, communication links, payload status, enemy status, collision probability, task status, and user attribution, the system can more comprehensively and integrally understand the operating environment and key indicators of the constellation, thereby improving the accuracy of decision-making and the ability of global optimization. By using the large-scale constellation high-efficiency simulation and deduction unit, the system can real-time simulate the operating conditions of the constellation, including changes in various states, the effects of task execution, etc. This helps to timely discover potential problems, optimize task planning, and provide real-time decision support. The comprehensive utilization of various data provides sufficient intelligent decision-making basis for the system. The system can analyze based on real-time data and historical data to make more intelligent and adaptable decisions, improving the efficiency of constellation collaborative management and control.

[0060] The core of this embodiment is to integrate data from multiple aspects into the high - efficiency simulation and deduction unit of a super - large - scale satellite constellation. This includes data in multiple dimensions such as the status of satellites, attitude and orbit parameters, communication links, etc. Through data integration, the system forms a comprehensive and detailed understanding of the operating environment of the satellite constellation. The high - efficiency simulation and deduction unit of the super - large - scale satellite constellation is responsible for real - time simulation of the various states of the satellite constellation. This process involves real - time calculation and update of massive amounts of data to reflect the dynamic evolution of the satellite constellation under different conditions. Based on the integrated data, the system can use large language models for intelligent decision - making. The system can learn historical data, analyze real - time data, and make intelligent decisions that conform to the overall goal.

[0061] Embodiment 4. This embodiment further defines the satellite constellation human - intelligence collaborative management and control system based on the large language model described in Embodiment 2. The standard instructions include: execution time, instruction ID, instruction execution component name, task name, instruction name, and instruction content.

[0062] Using a standard instruction format that includes information such as execution time, instruction ID, instruction execution component name, task name, instruction name, and instruction content helps improve the scalability and interoperability of the system. All instructions are defined according to the same specification, simplifying the management and maintenance of the system. The standard instructions containing information such as execution time enable the task execution time to be clearly defined, improving the accuracy of task control. This helps the system execute tasks according to the established plan and requirements, reducing errors and uncertainties. Each instruction has a unique instruction ID, and through this identifier, it is convenient to track and monitor the execution of the instruction. This is very important for the operation status monitoring and fault troubleshooting of the system.

[0063] In this embodiment, a unified and standard instruction format is designed for the system. This format covers key information such as execution time, instruction ID, instruction execution component name, task name, instruction name, and instruction content. Such standardization helps improve the readability and maintainability of the system while ensuring the consistency of instructions. By clearly defining the execution time in the standard instruction, the system can accurately control the start and execution timing of tasks based on this information. This is achieved through timestamps or other time representation methods to ensure that the system executes tasks according to the established plan. Each instruction has a unique instruction ID, ensuring the uniqueness of each instruction in the system. This unique identifier enables the system to accurately track, monitor, and manage instructions.

[0064] Embodiment 5. This embodiment further defines the satellite constellation human - intelligence collaborative management and control system based on the large language model described in Embodiment 4. The execution time is in UTC time, and the format is: year - month - day - hour - minute - second.

[0065] In this embodiment, UTC time is used as the standard for execution time, ensuring the same time benchmark is used globally. This is very important for the constellation human-intelligence collaborative control system because the constellation may involve different time zones, and a unified time format facilitates collaborative operations. The UTC time format is clearly defined, avoiding confusion and misunderstandings that may be caused by different time zones. This ensures that the time used in the constellation control system is consistent and clear, reducing potential errors. UTC is the world standard time and is not affected by time zones. This makes it easier for the constellation human-intelligence collaborative control system to integrate with other systems because other systems can also adopt the same UTC time standard.

[0066] By using UTC time, the system can achieve global collaborative operations. Regardless of where the components in the constellation are located, they can work collaboratively based on the same time standard. This is very important for the collaboration and overall operational consistency of the constellation. The format of the execution time is clearly defined, avoiding errors that may be caused by time zone differences or inconsistent time formats. This helps improve the accuracy and reliability of the system. Adopting the UTC time standard and the specified time format helps the system integrate with other external systems, reducing potential time processing problems during the integration process and improving the pluggability and interoperability of the system.

[0067] Embodiment Six: This embodiment further defines the constellation human-intelligence collaborative control system based on the large language model described in Embodiment One. The task planning process of the ultra-large-scale constellation task planning unit is specifically as follows:

[0068] Traverse all remote sensing satellites;

[0069] According to the traversal results, predict the orbits of the remote sensing satellites during the mission time, and calculate the attitude of the remote sensing satellites and the camera pointing to obtain the observable area;

[0070] Traverse all target sets and calculate the visibility of the remote sensing satellites to the targets;

[0071] Based on the visibility of the remote sensing satellites to the targets, judge the user input constraints, and filter out the list of available observation satellites;

[0072] Use the task planning method to filter out the optional observation plans in the observation list, and feedback the plans in text form through the intelligent interpretation response unit;

[0073] Make a decision selection and confirmation based on the feedback results. After confirming the decision, feedback the strategy to the simulation system in the form of command and control for real-time simulation.

[0074] In this embodiment, by traversing all remote sensing satellites and target sets, the system comprehensively utilizes a large amount of remote sensing data and target information. This ensures that the basic data for mission planning is comprehensive and helps to plan satellite missions more precisely. By predicting the orbits of remote sensing satellites and calculating the attitudes and camera pointings, the system can provide real-time observable area information. This helps to detect and utilize emergencies in a timely manner, improving the real-time performance and accuracy of mission planning. Following the constraints input by the user, the system takes into account the user's requirements in mission planning, such as time windows, observation directions, etc. This ensures that the generated list of observation satellites and plans meet the user's expectations.

[0075] This embodiment adopts a mission planning method to intelligently screen out optional observation plans, and uses a large language model for interpretation and feedback, making the system more intelligent and user-friendly and reducing the user's cognitive burden. The decisions confirmed by the user are fed back to the simulation system in the form of commands to achieve real-time simulation. This helps to verify the feasibility and effectiveness of the plan, improving the practicality and reliability of the system. Traverse all remote sensing satellites to obtain data such as orbits, attitudes, and camera pointings to provide basic information for mission planning. By calculating the visibility between remote sensing satellites and targets, determine which targets can be observed, thereby screening out a list of available observation satellites. According to the constraints input by the user, screen the visibility list to generate a list of observation satellites that meet the user's requirements. Use the mission planning method to further screen the observation list to generate optional observation plans to ensure the optimal mission planning results. Intelligently interpret the generated plans through a large language model and feedback them to the user in text form to provide intuitive and easy-to-understand information. After the user confirms the decision, feed back the strategy to the simulation system in the form of commands for real-time simulation verification to ensure the feasibility of the plan. By comprehensively utilizing a large amount of information and considering constraints, the purpose of the system is to generate an optimal constellation mission plan to achieve efficient utilization of satellite resources. Through the steps of intelligent interpretation and user decision confirmation, the system aims to provide a user-friendly interface and interaction method, enabling users to understand and participate in the mission planning process. Through real-time simulation feedback, the system ensures the real-time performance of the plan and guarantees the reliability of the plan through verification to meet the needs of practical applications. Through the intelligent interpretation of the large language model, the system provides users with detailed and easy-to-understand mission planning information to support the intelligent selection of users in the decision-making process.

[0076] Embodiment 7. This embodiment further limits the constellation human-machine collaborative control system based on a large language model described in Embodiment 6. The use of the mission planning method to screen out optional observation plans in the observation list includes:

[0077] The observation tasks of the satellite constellation are divided into continuous observation tasks and discontinuous observation tasks. Among them, for discontinuous observation tasks, only one optimal observation satellite needs to be selected to execute the tasks, and for continuous observation tasks, the observation sequence needs to be optimized and sorted.

[0078] In this embodiment, by dividing the observation tasks into continuous and discontinuous observations, the system can more effectively optimize satellite resources. For discontinuous observation tasks, selecting one optimal observation satellite reduces the scattered use of resources; for continuous observation tasks, optimizing the sorting can maximize the observation efficiency and ensure seamless connection between tasks. Different types of observation tasks may require different processing methods. By separately processing continuous and discontinuous observation tasks, the system can better adapt to the characteristics of different tasks and improve flexibility. For discontinuous observation tasks, the system can flexibly allocate tasks according to the performance and position characteristics of the optimal observation satellite to meet the task requirements. For continuous observation tasks, the system can achieve more intelligent task scheduling through optimizing the sorting. The optimization of the sorting for continuous observation tasks helps to improve the observation efficiency, ensure the full utilization of satellite resources in the time series, reduce idle time, and maximize the satisfaction of observation requirements.

[0079] Dividing the observation tasks into continuous and discontinuous observations and adopting different task planning strategies according to the characteristics of different types of tasks. For discontinuous observation tasks, the system selects one optimal observation satellite through the task planning method, considering factors such as satellite performance, position, and task requirements. For continuous observation tasks, the system arranges the observation sequence through the optimization sorting algorithm to maximize the observation efficiency and ensure smooth transition between tasks. For discontinuous observation tasks, the system formulates a flexible task allocation strategy to ensure that the optimal observation satellite can effectively execute the tasks. For continuous observation tasks, the system adopts an intelligent task scheduling strategy to achieve an efficient observation sequence. By differentiating the processing of continuous and discontinuous observation tasks, the system aims to more effectively utilize the satellite resources in the satellite constellation and improve the overall observation efficiency. By classifying the task types and adopting different task planning strategies, the system aims to improve the adaptability and flexibility to different task types and ensure that the system can perform well in various scenarios. Through optimizing the sorting and intelligent task scheduling, the system is committed to improving the efficiency of continuous observation tasks and ensuring that more observation tasks can be completed within a limited time.

[0080] Embodiment 8. The method for collaborative control of satellite constellation human-machine intelligence based on a large language model according to this embodiment is implemented based on the system described in any one of Embodiments 1 to 7, and the method includes:

[0081] When the user's requirements are input in the form of text dialogue, the intelligent interpretation and response unit extracts key elements from the dialogue history, aerospace domain knowledge, and preset key elements, and classifies the user's requirements according to the key elements;

[0082] After classification, keyword extraction is performed. Based on the keywords, information theme categories, event types, and corresponding accusation contents are obtained. According to the accusation contents, a task planning set is obtained;

[0083] The ultra-large-scale constellation task planning unit traverses all remote sensing satellites according to the task planning set obtained by the intelligent interpretation response module, performs collaborative task planning and scheduling on the traversed constellation, and outputs the scheduling result to the ultra-large-scale constellation high-efficiency simulation and deduction unit;

[0084] The ultra-large-scale constellation high-efficiency simulation and deduction unit receives the scheduling result and performs real-time simulation.

[0085] Embodiment Nine. A computer device described in this embodiment includes a memory and a processor. A computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the constellation human-machine collaborative management and control method based on the large language model according to Embodiment Eight.

[0086] Embodiment Ten. A computer-readable storage medium described in this embodiment is used to store a computer program, and the computer program executes the constellation human-machine collaborative management and control method based on the large language model according to Embodiment Eight.

[0087] Embodiment Eleven. This embodiment provides a specific example for the constellation human-machine collaborative management and control system based on the large language model described in Embodiment One, and is also used to explain Embodiments Two to Seven. Specifically:

[0088] The initiator of the large-scale constellation management and control process is generally the constellation commander. By comprehensively judging information such as the status of constellation satellites and attitude and orbit parameters, and making a judgment on the space situation, a command instruction is issued for management and control operations. As the constellation develops towards ultra-large scale, diverse payloads, and complex tasks, the management and control information shows exponential growth, presenting characteristics such as diversification, high timeliness, and large data volume. Traditional demand information interpretation and recognition still face multi-level manual analysis and decomposition, with problems such as understanding deviations caused by insufficient personnel understanding levels and long task issuance times, bringing many difficulties to subsequent demand determination, collaborative planning, and resource layout. Using the remote sensing constellation human-machine collaborative management and control system based on the large language model can avoid the cumbersome feature selection process, automatically abstract features, learn corresponding parameters, capture complex features, and quickly, accurately, and intelligently understand and recognize the issued emergency task requirements.

[0089] The remote sensing constellation human-machine collaborative management and control system based on the large language model is divided into three units, namely the intelligent interpretation response unit, the ultra-large-scale constellation task planning unit, and the ultra-large-scale constellation high-efficiency simulation and deduction unit;

[0090] The intelligent interpretation and response unit interprets the management and control tasks described in natural language into task instructions that can be understood by machines. To quickly, accurately, and intelligently understand and identify the requirements of the emergency tasks issued, the intelligent interpretation and response unit interprets and identifies based on aerospace professional knowledge. Through text extraction, key element extraction, and event detection of data, it effectively obtains the text information of the dialogue input, extracts keywords, obtains the information theme category, event type, and corresponding command content, and then generates standardized command information that can be automatically recognized by the satellite constellation deduction and intelligent planning system.

[0091] Table 1 Instruction Structure

[0092]

[0093] When the user's requirements are input in text form for dialogue, the intelligent interpretation and response unit extracts key elements based on the trained retrieval model, including dialogue history, aerospace domain knowledge, and preset key elements, and classifies the user input events according to the key elements. The user's requirements are decomposed into three task types: inquiry tasks, command tasks, and decision-making tasks, and keyword extraction is performed respectively on this basis. For inquiry tasks, the intelligent interpretation and response unit matches and queries the user input keywords with the massive data in the ultra-large scale satellite constellation high-efficiency simulation deduction unit, and obtains the user's inquiry information from satellite status, attitude and orbit parameters, coverage characteristics, communication links, payload status, enemy status, collision probability, task status, and user attribution, and gives an intelligent reply; for command tasks, the intelligent interpretation and response unit decomposes the user input keywords to obtain the standard instruction form shown in Table 1, and uses it as a command input to the ultra-large scale satellite constellation high-efficiency simulation deduction unit to drive the real-time intervention simulation of the attitude and orbit control module, rocket module, thruster module, antenna module, camera module, space environment module, TT&C coverage module, navigation module, sensor module, and routing calculation module; for decision-making tasks, the intelligent interpretation and response unit decomposes the user input into a task planning set according to keywords, including an observation resource set, an observation target set, and a constraint set, to support the solution of the ultra-large scale satellite constellation task planning unit.

[0094] Based on the observation task planning set (observation resource set, observation target set, constraint set) extracted from the dialogue, the ultra-large scale satellite constellation task planning unit is used for satellite constellation collaborative task planning and scheduling, providing perfect decision options for the commander, thereby optimizing the decision-making process and shortening the decision-making time, and assisting the commander in monitoring the entire task process.

[0095] The prerequisite for the satellite to observe a mission is that the satellite has a visible time window for the observation mission within the specified time. The mission planning method is to select the best time window from all the visible time windows of the satellite for the target. Therefore, for intelligent decision-making planning, it is necessary to first perform a loop calculation on the visibility of the satellites under management for the target. Using the task planning set (observation resource set, observation target set, constraint set) parsed by the intelligent interpretation response unit as the input, first traverse all remote sensing satellites, predict their orbits during the mission time, calculate their attitudes and camera pointing directions to determine the observable area, then traverse all target sets, calculate the visibility of the satellite to the target, and on this basis, judge according to the constraints input by the user to screen out a list of available observation satellites. The observation tasks of the satellite constellation are divided into continuous observation tasks and discontinuous observation tasks. The difference is that for discontinuous observation tasks, only one optimal observation satellite needs to be selected to execute the task, while for continuous observation tasks, it is necessary to optimize and sort the observation sequence. After using the task planning method to determine the optional observation plans respectively, use the intelligent interpretation response unit to feedback the plans to the user in text form. The user makes a decision selection and confirmation according to the feedback decision list. After the user confirms the decision, the strategy is fed back to the ultra-large-scale satellite constellation high-efficiency simulation and deduction unit in the form of command and control (as shown in Table 1) for real-time simulation, so as to facilitate the user to observe the mission results.

[0096] The above further describes the technical solution provided by the present invention in conjunction with the accompanying drawings to highlight the advantages and beneficial effects, and is not used as a limitation to the present invention. Any modifications, combinations of implementation manners, improvements, equivalent replacements, etc. based on the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A star cluster human-intelligence collaborative control system based on a large language model, characterized in that The system includes: an intelligent interpretation and response unit, a very large-scale satellite constellation mission planning unit, and a very large-scale satellite constellation high-efficiency simulation and deduction unit; The intelligent interpretation and response unit interprets and identifies based on aerospace professional knowledge. Through text extraction, key element extraction, and event detection of data, it obtains the text information of the dialogue input and extracts keywords; according to the keywords, it obtains the information theme category, event type, and corresponding accusation content, and obtains the mission planning set according to the accusation content; The very large-scale satellite constellation mission planning unit is used to traverse all remote sensing satellites according to the mission planning set obtained by the intelligent interpretation and response module, perform collaborative mission planning and scheduling on the traversed satellite constellation, and output the scheduling result to the very large-scale satellite constellation high-efficiency simulation and deduction unit; The very large-scale satellite constellation high-efficiency simulation and deduction unit receives the scheduling result for real-time simulation; The intelligent interpretation and response unit includes an inquiry task module, an accusation task module, and a decision task module; The inquiry task module is used to match and query the input keywords with the massive data in the very large-scale satellite constellation high-efficiency simulation and deduction unit, obtain inquiry information from satellite status, attitude and orbit parameters, coverage characteristics, communication links, payload status, enemy status, collision probability, mission status, and user attribution, and perform intelligent reply; The accusation task module is used to decompose the input keywords to obtain the standard instruction form, and use it as an accusation input to the very large-scale satellite constellation high-efficiency simulation and deduction module. By driving the attitude and orbit control module, rocket module, thruster module, antenna module, camera module, space environment module, TT&C coverage module, navigation module, sensor module, and routing calculation module in the very large-scale satellite constellation high-efficiency simulation and deduction unit for real-time simulation; The decision task module decomposes the input keywords into a mission planning set, including an observation resource set, an observation target set, and a constraint set.

2. The star cluster human-machine collaborative control system based on the large language model according to claim 1, wherein The massive data in the very large-scale satellite constellation high-efficiency simulation and deduction unit includes: satellite status, attitude and orbit parameters, coverage characteristics, communication links, payload status, enemy status, collision probability, mission status, and user attribution.

3. The star cluster human-machine collaborative control system based on the large language model according to claim 1, characterized in that, The standard instruction includes: execution time, instruction ID, instruction execution component name, mission name, instruction name, and instruction content.

4. The star cluster human-machine collaborative control system based on the large language model according to claim 3, wherein, The execution time is the UTC time, and the format is: year-month-day-hour-minute-second.

5. The star cluster human-machine collaborative control system based on a large language model according to claim 1, characterized in that, The mission planning process of the very large-scale satellite constellation mission planning unit is specifically as follows: Traverse all remote sensing satellites; According to the traversal result, predict the orbit of the remote sensing satellite during the mission time, and calculate the attitude of the remote sensing satellite and the camera pointing to obtain the observable area; Traverse all target sets and calculate the visibility of the remote sensing satellite to the target; Judge according to the visibility of the remote sensing satellite to the target for the input constraint conditions, and filter out the list of available observation satellites; Use the mission planning method to filter out the optional observation plans in the observation list, and feedback the plan in text form through the intelligent interpretation and response unit; Make a decision selection and confirmation according to the feedback result, and after confirming the decision, feedback the strategy to the simulation system in the form of an accusation for real-time simulation.

6. The star cluster human-machine collaborative control system based on the large language model according to claim 5, characterized in that, The use of the mission planning method to filter out the optional observation plans in the observation list includes: The observation tasks of the satellite constellation are divided into continuous observation tasks and discontinuous observation tasks. Among them, only one optimal observation satellite needs to be selected to perform the discontinuous observation task, and the continuous observation task requires optimizing the sorting of the observation sequence.

7. The star cluster human-machine collaborative control method based on large language models is characterized in that, The method is implemented based on the system described in any one of claims 1 to 6, and the method includes: When the user's requirements are input in the form of text dialogue, the intelligent interpretation and response unit extracts key elements from the dialogue history, aerospace domain knowledge, and preset key elements, and classifies the user's requirements according to the key elements; After classification, keyword extraction is performed. According to the keywords, the information theme category, event type, and corresponding accusation content are obtained, and a task planning set is obtained according to the accusation content; The ultra-large-scale satellite constellation task planning unit traverses all remote sensing satellites according to the task planning set obtained by the intelligent interpretation and response module, performs collaborative task planning and scheduling on the traversed satellite constellation, and outputs the scheduling result to the ultra-large-scale satellite constellation high-efficiency simulation and deduction unit; The ultra-large-scale satellite constellation high-efficiency simulation and deduction unit receives the scheduling result for real-time simulation.

8. A computer device, characterized in that: It includes a memory and a processor. A computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the method for collaborative management and control of satellite constellation human intelligence based on the large language model according to claim 7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program, and the computer program executes the method for collaborative management and control of satellite constellation human intelligence based on the large language model according to claim 7.

Images