Multi-satellite intelligent task planning system and method based on large language model

Abstract

The invention discloses a multi-satellite intelligent task planning system and method based on a large language model, and the system comprises a heuristic algorithm module construction module which is used for constructing a heuristic algorithm module based on a greedy algorithm; the large model optimization module is used for carrying out deep analysis and optimization on key functions in the heuristic algorithm module through a large language model to obtain an optimized heuristic algorithm module; the task generation module is used for generating a task plan through the optimized heuristic algorithm module; and the task execution module is used for guiding the satellite to execute the task according to the task plan. According to the method, an initial task planning scheme is quickly generated by utilizing a heuristic algorithm, and key functions (such as task sorting and stripe selection) are deeply optimized by applying a large language model, so that the planning accuracy and efficiency are improved.

Description

Technical Field

[0001] The present invention belongs to the field of satellite mission planning, and in particular relates to a multi-satellite intelligent mission planning system and method based on a large language model. Background Art

[0002] Satellite mission planning refers to the process of making arrangements for satellite orbits, observation targets, and time to complete specific observation missions. This process is crucial for effectively utilizing satellite resources, improving observation efficiency and data quality, and plays a key role in land and resources surveys, environmental monitoring, disaster prevention and mitigation, and other fields.

[0003] With the increase in the number of satellites and the diversification of application requirements, the amount of information required for mission planning has increased significantly, and new satellite mission planning technologies are urgently needed for planning.

[0004] Although existing satellite mission planning techniques are effective to a certain extent, they still have the following major shortcomings:

[0005] Slow processing speed: A large number of tasks and complex constraints make traditional algorithms (such as genetic algorithms and particle swarm optimization) slow in processing speed, and they are not agile enough for emergencies that require rapid response.

[0006] Lack of flexibility and adaptability: Traditional algorithms are often designed for static problems. It is difficult to make effective adjustments quickly to dynamically changing task requirements (such as urgent task insertion). They also lack the ability to learn and continuously optimize, cannot cope with different task contexts, and lack broad adaptability.

[0007] Optimization limitations: Many traditional algorithms (such as simulated annealing and genetic algorithms) are prone to falling into local optimal solutions and are difficult to reach global optimal solutions. In addition, parameter settings rely on experience and are complex to adjust, which affects the versatility and efficiency of the algorithm. Summary of the invention

[0008] In order to solve the above technical problems, the present invention proposes a multi-satellite intelligent mission planning system and method based on a large language model to solve the problems existing in the above-mentioned prior art.

[0009] To achieve the above object, the present invention provides a multi-satellite intelligent mission planning system based on a large language model, comprising:

[0010] Heuristic algorithm module building module, used to build heuristic algorithm modules based on greedy algorithm;

[0011] A large model optimization module, used to perform in-depth analysis and optimization on key functions in the heuristic algorithm module through a large language model to obtain an optimized heuristic algorithm module;

[0012] A task generation module is used to generate task planning through the optimized heuristic algorithm module;

[0013] The mission execution module is used to guide the satellite to execute the mission according to the mission plan.

[0014] Preferably, the heuristic algorithm module comprises a first building unit, a second building unit and a modular unit;

[0015] The first construction unit is used to construct a data set for implementing planning and an evaluation index for evaluating planning schemes;

[0016] The second construction unit is used to construct a basic algorithm based on a greedy algorithm;

[0017] The modularization unit is used to modularize the basic algorithm to obtain a modular algorithm.

[0018] Preferably, the large model optimization module includes a marking unit, an optimization unit and an evaluation unit;

[0019] The marking unit is used to construct an evaluation function according to the evaluation index, mark the evaluation function and the function to be optimized in the modular algorithm through a decorator, and obtain a specification file;

[0020] The optimization unit is used to optimize the function to be optimized in the specification file based on the data set to obtain a preliminary heuristic algorithm module;

[0021] The evaluation unit is used to evaluate the preliminary heuristic algorithm module based on the data set and the evaluation index, and obtain the optimized heuristic algorithm module when the evaluation is qualified.

[0022] Preferably, the multi-satellite intelligent mission planning system also includes a feedback module, which is used to evaluate the effectiveness of the satellite's mission execution and feed back to the large model optimization module to participate in the in-depth analysis and optimization of key functions.

[0023] The present invention also provides a multi-satellite intelligent mission planning method based on a large language model, comprising the following steps:

[0024] Construct heuristic algorithm module based on greedy algorithm;

[0025] Performing in-depth analysis and optimization on key functions in the heuristic algorithm module through a large language model to obtain an optimized heuristic algorithm module;

[0026] Generate mission plans through optimized heuristic algorithm modules;

[0027] The satellite is guided to perform the mission according to the mission plan.

[0028] Preferably, before constructing the heuristic algorithm module, the method further includes: constructing a data set for implementing the planning and an evaluation index for evaluating the planning scheme, wherein the data set includes a target data set and a satellite data set.

[0029] Preferably, the method for constructing a heuristic algorithm module includes: constructing a basic algorithm based on a greedy algorithm, and modularizing the basic algorithm to obtain a heuristic algorithm module.

[0030] Preferably, the method for obtaining the optimized heuristic algorithm module comprises:

[0031] Construct an evaluation function according to the evaluation index, mark the evaluation function and the function to be optimized in the modular algorithm through a decorator, and obtain a specification file;

[0032] Based on the data set, optimize the function to be optimized in the specification file to obtain a preliminary heuristic algorithm module;

[0033] The preliminary heuristic algorithm module is evaluated based on the data set and evaluation indicators, and the optimized heuristic algorithm module is obtained when the evaluation is qualified.

[0034] Preferably, the present invention also evaluates the effectiveness of the satellite in executing the mission, and feeds the evaluation results back to the large model optimization module to participate in the in-depth analysis and optimization of key functions.

[0035] Compared with the prior art, the present invention has the following advantages and technical effects:

[0036] The present invention discloses a multi-satellite intelligent mission planning system and method based on a large language model, including: constructing a heuristic algorithm module based on a greedy algorithm; deeply analyzing and optimizing key functions in the heuristic algorithm module through a large language model to obtain an optimized heuristic algorithm module; generating a mission plan through the optimized heuristic algorithm module; and guiding satellites to perform missions according to the mission plan. The present invention uses a heuristic algorithm to quickly generate an initial mission planning scheme and applies a large language model to deeply optimize key functions (such as task sorting and strip selection), thereby improving the accuracy and efficiency of planning, avoiding optimization limitations, and making the present invention more adaptable. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0038] Figure 1 A schematic diagram of the architecture of a multi-satellite intelligent mission planning system according to an embodiment of the present invention;

[0039] Figure 2 A flowchart of a multi-satellite intelligent mission planning method according to an embodiment of the present invention;

[0040] Figure 3 A flowchart for implementing the large model optimization module of an embodiment of the present invention;

[0041] Figure 4 A selection function optimization score curve diagram of an embodiment of the present invention;

[0042] Figure 5 This is a ranking function optimization curve diagram according to an embodiment of the present invention. DETAILED DESCRIPTION

[0043] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0044] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0045] Embodiment 1

[0046] like Figure 1 As shown, in this embodiment, a multi-satellite intelligent mission planning system based on a large language model is provided, including:

[0047] Heuristic algorithm module building module, used to build heuristic algorithm modules based on greedy algorithm;

[0048] The large model optimization module is used to perform in-depth analysis and optimization on key functions in the heuristic algorithm module through a large language model to obtain an optimized heuristic algorithm module.

[0049] A task generation module is used to generate task planning through the optimized heuristic algorithm module;

[0050] The mission execution module is used to guide the satellite to perform the mission according to the mission plan.

[0051] Further, the heuristic algorithm module includes a first building unit, a second building unit and a modular unit;

[0052] The first construction unit is used to construct a data set for implementing the plan and an evaluation index for evaluating the plan scheme;

[0053] The second construction unit is used to construct a basic algorithm based on the greedy algorithm;

[0054] The modular unit is used to modularize the basic algorithm to obtain a modular algorithm.

[0055] Further, the large model optimization module includes a labeling unit, an optimization unit, and an evaluation unit;

[0056] The marking unit is used to construct an evaluation function according to the evaluation index, mark the evaluation function and the function to be optimized in the modular algorithm through the decorator, and obtain the specification file;

[0057] The optimization unit is used to optimize the function to be optimized in the specification file based on the data set to obtain a preliminary heuristic algorithm module;

[0058] The evaluation unit is used to evaluate the preliminary heuristic algorithm module based on the data set and the evaluation index, and obtain the optimized heuristic algorithm module when the evaluation is qualified.

[0059] Furthermore, the multi-satellite intelligent mission planning system also includes a feedback module, which is used to evaluate the effectiveness of the satellite's mission execution and feed back to the large model optimization module to participate in the in-depth analysis and optimization of key functions.

[0060] Among them, the heuristic algorithm module uses a greedy algorithm as the basis to quickly generate a preliminary task plan based on task urgency and resource availability.

[0061] The large model optimization module uses a well-trained large language model (such as GPT) to further analyze and optimize the key functions in the heuristic algorithm. It mainly optimizes the two functions of task sorting and strip selection that have a greater impact on the final planning results, and enhances the accuracy and efficiency of decision-making through model reasoning. It conducts self-learning, continuously learns from new task execution results, and adjusts the optimization strategy.

[0062] The mission execution module guides the satellite to perform specific missions according to the final optimized mission plan, including the scheduling, management and resource allocation of observation missions, etc. It monitors the mission execution status, updates the mission execution status in real time, and feeds back to the optimization module to adjust the plan when necessary.

[0063] like Figure 2 As shown, this embodiment also provides a multi-satellite intelligent mission planning method based on a large language model, including the following steps:

[0064] Construct heuristic algorithm module based on greedy algorithm;

[0065] Through the large language model, the key functions in the heuristic algorithm module are deeply analyzed and optimized to obtain the optimized heuristic algorithm module;

[0066] Generate mission plans through optimized heuristic algorithm modules;

[0067] Guide the satellite to perform its mission according to the mission plan.

[0068] The specific implementation steps are shown in Table 1:

[0069] Table 1

[0070]

[0071]

[0072] In step 1.1, the required dataset should contain two parts, namely the target dataset and the satellite dataset.

[0073] The target dataset should contain the geographic information of all targets, i.e. the coordinates, height, boundary and area (if any) of the targets in the specified coordinate system, as well as the constraints of the targets, such as the number of observations required for the targets, the resolution achieved by the observations, the required solar altitude angle during observations, the available observation time window and the required time interval between two observations. This part of the data should be saved in csv or pkl (recommended) format for subsequent reading.

[0074] Satellite data sets should include the motion information of all satellites, i.e., sub-satellite point trajectory, satellite altitude, satellite orbit period, and satellite constraints, such as satellite resolution and the upper limit of the cumulative working time of each satellite orbit. This part of the data should be saved in csv or pkl (recommended) format for subsequent reading.

[0075] In addition, the evaluation indicators in step 1.1 should be designed based on the type of tasks to be performed, should have the characteristics of quantitative comparison of control variables, and have a relatively high degree of differentiation, which can clearly distinguish the advantages and disadvantages of different schemes. For example, if the task completion level is used as the evaluation indicator, the time range T can be specified, and the score can be given according to the number of effective observations or observation areas completed within the T time (the higher the completion level, the higher the score); at the same time, the task completion time can also be used as the evaluation indicator, and a certain number or all of the required observations or observation areas must be completed, and the time required is examined, and the score is assigned according to the length of time (the less time required, the higher the score).

[0076] In step 1.2, the basic algorithm constructed should meet the following requirements:

[0077] The algorithm input should be the dataset in step 1.1;

[0078] The output of the algorithm should be a satellite scheduling plan, stored in csv format, and used as input for subsequent evaluation;

[0079] The basic algorithm constructed does not require the following:

[0080] There is no requirement for the algorithm language, but Python is recommended as the algorithm language

[0081] There are no requirements for the content, form, and complexity of the algorithm, but to facilitate subsequent operations, it is recommended to use an algorithm with relatively simple logic

[0082] In step 1.3, the so-called "modularizing the algorithm" means abstracting some functions in the main program and decomposing them into multiple sub-functions by defining functions. These sub-functions are responsible for a single function and collaborate through data transmission to finally complete the output of the overall program. Each sub-function should have clear responsibilities and explain its functions, input parameters and output results in detail through comments. This method not only makes the program structure clearer and easier to maintain and modify, but also provides a good foundation for the subsequent learning and optimization of large models.

[0083] Furthermore, before constructing the heuristic algorithm module, it also includes: constructing a data set for implementing the planning and an evaluation index for evaluating the planning scheme, and the data set includes a target data set and a satellite data set.

[0084] Furthermore, the method for constructing a heuristic algorithm module includes: constructing a basic algorithm based on a greedy algorithm, and modularizing the basic algorithm to obtain a heuristic algorithm module.

[0085] Furthermore, the method for obtaining the optimized heuristic algorithm module includes:

[0086] Construct an evaluation function based on the evaluation index, mark the evaluation function and the function to be optimized in the modular algorithm through decorators, and obtain the specification file;

[0087] Based on the data set, optimize the function to be optimized in the specification file to obtain a preliminary heuristic algorithm module;

[0088] The preliminary heuristic algorithm module is evaluated based on the data set and evaluation indicators, and the optimized heuristic algorithm module is obtained when the evaluation is qualified.

[0089] Furthermore, the present invention will also evaluate the effectiveness of the satellite's mission execution and feed the evaluation results back to the large model optimization module to participate in the in-depth analysis and optimization of key functions.

[0090] Embodiment 2

[0091] The above method is illustrated as follows:

[0092] First, define point targets and area targets.

[0093] Point target: refers to a target object without considering the area of ​​the target, only with position coordinate information and the requirement of the number of observations. The score is calculated based on the ratio of the number of observations to the required number of observations.

[0094] Area target: A target object with an area, defined by the coordinate information of the area boundary, and the score is calculated based on the ratio of the observed area to the target area.

[0095] Consider a task planning problem, T a represents the regional target set, T p Represents a set of point targets. Each target t∈T in the target set a or t∈T p It has complex constraints and priority weights. The mission execution object (satellite) is a set of execution units S, and the accessibility of each execution unit s∈S to the target t depends on the mission planning time and its constraints.

[0096] The completion of the task goal depends on the effective coverage of the goal by multiple execution units s∈S. Different goal types have different evaluation criteria:

[0097] For a point target t∈T p ,The evaluation criteria depends on the relationship between the effective ,coverage number and its required coverage number.

[0098] For a regional target t∈T a ,The evaluation criteria depends on the ratio between the effective coverage area and its total area.

[0099] set up:

[0100] n is the number of point targets, n=|T p |,

[0101] m is the number of regional targets, m = |T a |,

[0102] f i represents the effective coverage times of the i-th point target,

[0103] F i is the number of coverages required for the i-th point target,

[0104] P i,p is the priority score of the i-th point target,

[0105] a j represents the effective coverage area of ​​the jth regional target,

[0106] A j is the total area of ​​the jth regional target,

[0107] Pj,a is the priority score of the j-th regional target.

[0108] The goal of task planning is to maximize an evaluation index P, which is defined as the weighted sum of all point targets and regional targets:

[0109]

[0110] in:

[0111] The first summation term represents the coverage contribution of the point target, which reflects the ratio of the effective coverage degree of each point target to its required coverage times, multiplied by the corresponding priority score;

[0112] The second summation term represents the coverage contribution of the regional targets, reflecting the ratio of the effective coverage area of ​​each regional target to its total area, multiplied by the corresponding priority score.

[0113] The availability of execution units s∈S and access to targets may be subject to a variety of complex constraints, such as time, resources, geometric location, etc. Such constraints will affect the effective coverage of the target, thereby affecting the maximization process of the evaluation index P of the overall task planning.

[0114] Next, consider a specific case. In the present embodiment, it is assumed that 450 point targets and 50 regional targets (a total of 500 initial targets) scattered in different parts of the world and a total of 60 targets (50 point targets and 10 regional targets) dynamically added over time are carried out mission planning, and the execution object is a total of 200 satellites, and the mission targets and satellites all have more complex constraints, and different mission targets have different priorities (shown as having different scores). The data set used is the access information of the satellite to each target, and the evaluation index P calculation method for a certain planning scheme is:

[0115]

[0116] Among them, n and m are the total number of point targets and area targets respectively, and f i With F i are the effective coverage times and required coverage times for the i-th point target, respectively, P i,p is the score for the i-th point target, a i With A i are the effective coverage area of ​​the i-th regional target and the total area of ​​the regional target, P i,a is the score for the i-th regional target.

[0117] After completing the heuristic algorithm module, two functions, task sorting and strip selection, were selected as the objects of optimization of the large language model. The large model used is GPT-3.5-turbo, which is accessed and interacted with through API. The number of previously generated algorithm samples obtained for each round of prompt words is 4. The optimization process is recorded through tensorboard. A total of six different scenes are selected to form the training set. The total score is calculated by weighting the scene scores according to different difficulties. The computing environment is Windows 11 operating system, the language version is python 3.11, and the computing device is an 8-core 4GB processor. First, the strip selection function is optimized, and the optimization curve is as follows: Figure 4 shown.

[0118] The initial score is the score of the original basic heuristic algorithm on the test set, which reaches the highest value after 1.35 hours of iteration. In the early and middle stages, the optimization score increases rapidly. This shows that there is a large room for optimization of this function, and it is still possible to further improve it by extending the number of iterations. The final generated result of the strip selection function is as follows:

[0119]

[0120]

[0121] It can be found that the optimized function focuses on adjusting the selection strategy of point targets, and is more inclined to select SAR satellites with smaller width and higher resolution. This is also in line with the actual logic, that is, there is no determination of the coverage area for point targets, so satellites with smaller width are preferred, and satellite resources with large width are reserved for regional targets. However, there are also some unreasonable places, such as simply adding the solar altitude angle and deflection angle, giving priority to strips with long duration, etc., which shows that there is still room for improvement in this function.

[0122] After completing the evolution of the strip selection function, the task sorting function was further evolved and iterated, using the same configuration as before. The resulting optimization curve is shown in Figure 5 shown.

[0123] Among them, the initial score is the final result of the selection function optimization, which reaches the highest value after 0.86 hours. The improvement of the task sorting function is also obvious, with a major breakthrough. The final generation result of the task sorting function is as follows:

[0124]

[0125]

[0126]

[0127] It can be seen that the generated strategy tends to arrange point targets first, then select effective and unplanned information from the access information of each task, calculate its potential scoring ability, and finally sort all tasks according to the product of potential scoring ability and scoring factor. This strategy is a partial modification of the original strategy, which has achieved good improvement results.

[0128] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A multi-satellite intelligent mission planning system based on a large language model, characterized in that: include: Heuristic algorithm module building module, used to build heuristic algorithm modules based on greedy algorithm; A large model optimization module, used to perform in-depth analysis and optimization on key functions in the heuristic algorithm module through a large language model to obtain an optimized heuristic algorithm module; A task generation module is used to generate task planning through the optimized heuristic algorithm module; The mission execution module is used to guide the satellite to execute the mission according to the mission plan.

2. The multi-satellite intelligent mission planning system based on a large language model according to claim 1, characterized in that: The heuristic algorithm module includes a first building unit, a second building unit and a modular unit; The first construction unit is used to construct a data set for implementing planning and an evaluation index for evaluating planning schemes; The second construction unit is used to construct a basic algorithm based on a greedy algorithm; The modularization unit is used to modularize the basic algorithm to obtain a modular algorithm.

3. The multi-satellite intelligent mission planning system based on a large language model according to claim 1, characterized in that: The large model optimization module includes a marking unit, an optimization unit and an evaluation unit; The marking unit is used to construct an evaluation function according to the evaluation index, mark the evaluation function and the function to be optimized in the modular algorithm through a decorator, and obtain a specification file; The optimization unit is used to optimize the function to be optimized in the specification file based on the data set to obtain a preliminary heuristic algorithm module; The evaluation unit is used to evaluate the preliminary heuristic algorithm module based on the data set and the evaluation index, and obtain the optimized heuristic algorithm module when the evaluation is qualified.

4. The multi-satellite intelligent mission planning system based on a large language model according to claim 1, characterized in that: The multi-satellite intelligent mission planning system also includes a feedback module, which is used to evaluate the effectiveness of satellite mission execution and feed back to the large model optimization module to participate in the in-depth analysis and optimization of key functions.

5. A multi-satellite intelligent mission planning method based on a large language model, characterized in that: The following steps are involved: Construct heuristic algorithm module based on greedy algorithm; Performing in-depth analysis and optimization on key functions in the heuristic algorithm module through a large language model to obtain an optimized heuristic algorithm module; Generate mission plans through optimized heuristic algorithm modules; The satellite is guided to perform the mission according to the mission plan.

6. The multi-satellite intelligent mission planning method based on a large language model according to claim 5 is characterized in that: Before constructing the heuristic algorithm module, the method also includes: constructing a data set for implementing the planning and an evaluation index for evaluating the planning scheme, wherein the data set includes a target data set and a satellite data set.

7. The multi-satellite intelligent mission planning method based on a large language model according to claim 5, characterized in that: The method for constructing a heuristic algorithm module includes: constructing a basic algorithm based on a greedy algorithm, and modularizing the basic algorithm to obtain a heuristic algorithm module.

8. The multi-satellite intelligent mission planning method based on a large language model according to claim 5, characterized in that: The method for obtaining the optimized heuristic algorithm module comprises: Construct an evaluation function according to the evaluation index, mark the evaluation function and the function to be optimized in the modular algorithm through a decorator, and obtain a specification file; Based on the data set, the functions to be optimized in the specification file are optimized to obtain a preliminary heuristic algorithm module, wherein the functions to be optimized include a stripe selection function and a task sorting function; The preliminary heuristic algorithm module is evaluated based on the data set and evaluation indicators, and the optimized heuristic algorithm module is obtained when the evaluation is qualified.

Images