A multifunctional satellite constellation configuration design optimization method based on large language model

Abstract

This invention discloses a multifunctional satellite constellation configuration design optimization method based on a large language model, belonging to the field of spacecraft system design technology. The method includes: setting design variables and fixed variables, where the design variables include the orbital inclination and orbital altitude of satellites in the satellite constellation; constructing an evaluation function based on coverage percentage, ground pixel resolution, average number of visible satellites, and geometric dilution of precision; iteratively generating solutions based on the design variables and fixed variables using the large language model, evaluating the solutions using the evaluation function, and obtaining an optimal satellite constellation configuration design scheme based on the evaluation results. Through the above technical solution, the present invention can provide a satellite constellation configuration design optimization method that takes into account multiple optimization aspects.

Description

Technical Field

[0001] The present invention belongs to the technical field of spacecraft system design, and in particular relates to a multi-functional satellite constellation configuration design optimization method based on a large language model. Background Art

[0002] A multi-satellite, multi-functional constellation is a system of small satellites operating in formation, using distributed cooperative control and wireless communication technologies to operate in specific configurations in orbit or deep space, meeting diverse operational requirements. These satellites collaborate to perform tasks that are difficult for a single, traditional satellite to accomplish alone. However, since the design of large-scale, multi-functional constellations increases exponentially with the number of satellites, and often faces complex constraints, it is a typical NP-hard problem, making it difficult to find an optimal solution within a finite timeframe. The design challenge lies in finding high-quality feasible and suboptimal solutions within a finite timeframe. During the conceptual design phase of a constellation, traditional constellation design algorithms primarily include simulation comparison, geometric analysis, and methods based on manually designed optimization algorithms. However, these algorithms are prone to local optima. Furthermore, large-scale constellations are typically large and expensive to deploy. Traditional algorithms often focus on single-function designs, resulting in low satellite utilization. To improve resource utilization and conserve orbital resources and deployment costs, it is necessary to explore multi-functional constellation design methods that enable multiple satellites to serve multiple purposes. Therefore, in order to overcome the limitations of traditional algorithms for constellations, it is necessary to develop an intelligent search optimization algorithm that can explore the existence of optimization space and has generalization capabilities, so as to realize the design and autonomous optimization of satellite constellation configuration schemes in the overall design stage, and provide a scientific basis and reference for the overall design of satellite constellation systems. Summary of the Invention

[0003] In order to solve the above technical problems, the present invention proposes a multi-functional satellite constellation configuration design optimization method based on a large language model to solve the problems existing in the above-mentioned prior art.

[0004] To achieve the above objectives, the present invention provides a multi-functional satellite constellation configuration design optimization method based on a large language model, comprising:

[0005] Setting design variables and fixed variables, wherein the design variables include orbital inclinations and orbital altitudes of satellites in the satellite constellation;

[0006] Construct an evaluation function based on coverage percentage, ground pixel resolution, average number of visible stars, and geometric dilution of precision;

[0007] Based on design variables and fixed variables, solutions are iteratively generated through a large language model, and the solutions are evaluated through an evaluation function. Based on the evaluation results, the optimal satellite constellation configuration design scheme is obtained.

[0008] Optionally, the target optimization function is

[0009] find x=[h,i]

[0010]

[0011] st i min ≤i≤i max ,h min ≤h≤h max ,C r =100

[0012] Among them, w1, w2, w3, and w4 are the weights of the corresponding targets, F(x) is the evaluation function, and C r is the coverage percentage, is the dimensionless ground pixel resolution after normalization, N is the average number of visible stars, GDOP is the geometric dilution of precision, x is the design variable, h is the orbit altitude, and i is the orbit inclination, where the subscripts min and max represent the corresponding minimum and maximum values.

[0013] Optionally, the process of obtaining the coverage percentage includes:

[0014] The coverage angle, center angle, and coverage bandwidth are calculated based on the orbital altitude of a single satellite and the corresponding sub-satellite point; the longitude and latitude of the ground target and the longitude and latitude of the sub-satellite point are obtained based on the coverage angle, center angle, and coverage bandwidth; the geocentric angle is calculated based on the longitude and latitude of the ground target and the longitude and latitude of the sub-satellite point; the geocentric angle and coverage angle are judged; if the geocentric angle is less than or equal to the coverage angle, the ground target is included in the coverage range of the current satellite;

[0015] Make a grid map of the observation area, use the grid points as feature points, and use the feature points as ground targets to determine whether the ground targets are included in the coverage range of the current satellite. Obtain the total time that all feature points in the observation area are covered by one or more satellites, and divide the total time by the total simulation time to get the coverage percentage.

[0016] Optionally, the process of obtaining the ground pixel resolution includes:

[0017] The ground pixel resolution of the payload is calculated based on the orbit altitude, the pixel size of the satellite payload, and the focal length of the satellite's optical system, and the ground pixel resolution of the payload is normalized to obtain a normalized dimensionless ground pixel resolution.

[0018] Optionally, the process of obtaining the average number of visible stars includes:

[0019] Using the geocentric rectangular coordinate system as the reference system, obtain the coordinates of the ground observation station and the satellite. Based on the coordinates of the ground observation station and the satellite, obtain the observation vector from the ground observation station to the satellite. Convert the observation vector from the geocentric rectangular coordinate system to the station-centered coordinate system based on the ground observation station. Calculate based on the converted observation vector to obtain the satellite elevation angle. When the preset altitude cutoff angle is less than the satellite elevation angle, the satellite is visible to the ground observation station.

[0020] Make a grid map of the observation area, use the grid points as feature points, and use the feature points as ground observation stations to determine whether the satellite is visible to the ground observation station. Get the sum of the number of visible stars at all feature points at each moment, and divide the sum of the number of visible stars by the total simulation time to get the average number of visible stars.

[0021] Optionally, the process of obtaining the geometric dilution of precision includes:

[0022] According to the satellite elevation angle and satellite azimuth angle, a coefficient matrix is ​​obtained. According to the coefficient matrix, the mean square error of the three-dimensional position and clock bias is obtained. According to the mean square error of the three-dimensional position and clock bias, the geometric precision factor is calculated.

[0023] Optionally, fixed parameters include the total number of satellites, the number of orbital planes, phase parameters, and the longitude and latitude range of the observation area.

[0024] The process of obtaining the optional optimal satellite constellation configuration design includes:

[0025] Initializing design variables, optimizing the design variables using the large language model as an optimizer to obtain a solution, wherein the solution is a new constellation configuration scheme;

[0026] Evaluate the solution through the evaluation function to obtain the evaluation score;

[0027] Based on the evaluation scores, the solutions are screened to obtain the optimal solution. Meta-prompts are generated based on the optimal solution and the task description. The meta-prompts are predicted and optimized through the large language model to generate new solutions.

[0028] Repeat the solution generation, evaluation and screening steps until the maximum number of iterations is reached, and then the solution with the best evaluation score is obtained as the optimal satellite constellation configuration design solution.

[0029] Compared with the prior art, the present invention has the following advantages and technical effects:

[0030] To address the problem of large-scale, multi-functional constellation configuration design, traditional algorithms are prone to local optimality due to the exponential increase in parameters and complex constraints. This paper proposes a constellation configuration design optimization algorithm based on a large language model. This algorithm can generate and autonomously find the optimal solution to the problem with zero sample count, achieving better optimization results than traditional design methods and possessing stronger generalization capabilities. Large-scale constellations are typically large and costly to deploy, while traditional single-function constellation configuration design methods result in low satellite utilization. This paper proposes a multi-functional constellation configuration design method, which improves satellite utilization and saves orbital resources and deployment costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0032] Figure 1 Schematic diagram of a satellite ground coverage model according to an embodiment of the present invention;

[0033] Figure 2 Schematic diagram of a satellite visibility analysis model according to an embodiment of the present invention;

[0034] Figure 3 Schematic diagram of the distribution of satellites involved in the positioning solution when calculating GDOP according to an embodiment of the present invention;

[0035] Figure 4 This is an overview diagram of the large language model optimization algorithm framework according to an embodiment of the present invention;

[0036] Figure 5 This is an optimized flow chart of an embodiment of the present invention;

[0037] Figure 6 Schematic diagram of optimization results of the large language model optimization method according to an embodiment of the present invention;

[0038] Figure 7 Schematic diagram of optimization results of the improved genetic algorithm according to an embodiment of the present invention. DETAILED DESCRIPTION

[0039] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this 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.

[0040] 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.

[0041] To overcome the problem of traditional algorithms being prone to local optimality in large-scale multi-functional satellite constellation configuration design due to exponential parameter increases and complex constraints, the present invention discloses a multi-functional satellite constellation configuration design optimization method based on a large language model, which is a highly efficient constellation configuration design method. The problem to be solved is to explore the existence of optimization space and construct an intelligent search optimization algorithm with generalization capabilities to achieve a constellation configuration design that integrates navigation enhancement and remote sensing, so that the coverage percentage of the target observation area is as high as possible, the pixel resolution of the payload and the navigation GDOP are as small as possible, the average number of visible satellites in the target area is as large as possible, and the cost of satellite constellation design is reduced, orbital resources are conserved, and resource utilization is improved.

[0042] The present invention is achieved through the following technical solutions.

[0043] This invention discloses a multifunctional satellite constellation configuration design optimization method based on a large language model. Based on the Walker-δ constellation configuration, the method determines the initial conditions for the satellite constellation optimization design problem. It then establishes a satellite constellation ground coverage analysis model, a satellite constellation ground observation resolution model, a visibility analysis model, and a GDOP model. The method then uses a large language model optimization algorithm to autonomously optimize the orbital altitude and inclination using coverage percentage, ground pixel resolution, number of visible satellites, and GDOP as objective functions. Furthermore, a linear weighting method is used to transform the multi-objective problem into a single-objective one, thereby efficiently obtaining a set of optimal configuration solutions that meet engineering requirements.

[0044] Specifically, the present invention discloses an efficient constellation configuration design method, which includes the following steps.

[0045] 1. Optimization model establishment

[0046] (1) Design variables: Set the design variables to the satellite's orbital height h = [h1,h2,…,h n ], orbital inclination i=[i1,i2,…,i n ], n is the total number of satellites in the constellation. Since the present invention adopts the Walker-δ constellation configuration, the orbital height h and orbital inclination i in the constellation are unified, which can be abbreviated as x=[h,i].

[0047] (2) Constraints: Track height constraint h∈[h min ,h max ], orbital inclination constraint i∈[i min ,i max ], the target area has no leakage coverage constraint, that is, the cumulative coverage percentage should be 100%.

[0048] (3) Optimization objectives: Among them C r(x) is the cumulative coverage percentage of the satellite to the target area within 24 hours, R p (x) is the ground pixel resolution of satellite earth observation, N(x) is the average number of visible stars in the target area, and GDOP(x) is the geometric dilution of precision that measures the navigation performance in the target area.

[0049] The above optimization target calculation method is as follows.

[0050] a. Coverage percentage calculation method

[0051] First, establish a satellite coverage analysis model. The satellite constellation coverage model is as follows: Figure 1 As shown. According to the satellite orbit height h and the minimum observation angle σ, the coverage angle d of a single satellite is calculated. σ , center angle α σ and coverage bandwidth S wσ On this basis, the geocentric angle θ corresponding to the ground target and the satellite is obtained to determine whether the point is within the coverage of the satellite.

[0052] Assume that the orbital height of satellite S at a certain moment is h, and the corresponding sub-satellite point is G. According to formula (1), the coverage angle d under the minimum observation angle σ is calculated respectively: σ , central angle α σ and coverage bandwidth S wσ .

[0053]

[0054] S wσ =2a E ·d σ

[0055] Among them, a E represents the radius of the earth, and P represents a point on the earth.

[0056] The intersection of the line connecting the spacecraft and the Earth's center of mass with the Earth's surface is called the sub-satellite point. The curve formed by the series of sub-satellite points formed by the spacecraft in orbit is called the sub-satellite point trajectory. The sub-satellite point is usually expressed in longitude and latitude, and can be calculated using the satellite's orbital elements. If the satellite's orbital elements at the current moment are known, the longitude and latitude of the sub-satellite point can be expressed as

[0057]

[0058] Among them, λ S is the longitude of the satellite subsatellite point, is the latitude of the satellite's subsatellite point; i is the orbital inclination, u=ω+f is the latitude argument, ω is the perigee argument, f is the true anomaly; Ω is the right ascension of the orbital ascending node, α G It is the Greenwich mean sidereal hour angle, which is affected by the Earth's rotation.

[0059] The Julian date starts at noon Greenwich on July 1, 4713 BC. From this epoch, the Julian date increases by 1 for each mean solar day. The Julian date can be used to easily calculate the length of time between two dates. Given the Coordinated Universal Time (year (yr), month (mon), day (day), hour (hr), minute (min), second (sec)), the corresponding Julian date JD is:

[0060]

[0061] The corresponding calculation method of Greenwich mean sidereal time is defined as

[0062] α G =67310.54841+(876600+8640184.812866)T UTI +0.093104T 2 UTI -6.2×10 -6 T 3 UTI

[0063] Where T UTI is the Julian century relative to 12:00 UT1 on January 1, 2000, calculated as

[0064]

[0065] Where JD is the corresponding Julian day.

[0066] Based on the above coverage angle, center angle, coverage bandwidth and longitude and latitude of the sub-satellite point, it is assumed that the longitude of the current ground target is λ w , latitude is According to formula (3), calculate the geocentric angle θ between the ground target and the satellite; if θ≤d σ , it means that the ground target is included in the coverage of satellite S.

[0067]

[0068] A grid map is drawn at preset latitude and longitude intervals within the observation area, and the obtained grid points are used as feature points; the coverage percentage C is used. r As an indicator of the constellation's Earth observation coverage performance, it is equal to the total time that all feature points in the observation area are covered by one or more satellites divided by the total simulation time. The present invention uses simulation software to calculate the cumulative coverage percentage, which is the percentage of the target area covered by satellite sensors within a specified time.

[0069] b. Calculation method of ground pixel resolution

[0070] According to the satellite orbit height h, the payload pixel size d and the optical system focal length f, the ground pixel resolution R of the payload is obtained according to formula (4): p As an indicator of the constellation's Earth observation resolution performance:

[0071]

[0072] According to formula (5), R p Normalize and get dimensionless ground pixel resolution

[0073]

[0074] c. Calculation method of the number of visible stars

[0075] like Figure 2 The model for satellite visibility analysis is shown in Figure 1. The angle between the observation vector between the ground observation station and the satellite and the horizontal plane defined by the easting and northing axes in the station-centered coordinate system is called the satellite elevation angle. The positioning performance of satellite navigation systems is influenced by multiple metrics, with satellite visibility being a core indicator. Its determination relies on comparing the satellite's elevation angle with a preset cutoff elevation angle. Specifically, a satellite is considered visible when its elevation angle reaches or exceeds this preset cutoff value.

[0076] The analysis process is as follows:

[0077] First, take the geocentric rectangular coordinate system as the reference system and assume that the coordinates of the ground observation station O are (x o ,y o ,z o ), let the coordinates of the satellite be (x i ,y i ,z i ). Through the coordinates of these two points, the observation vector (Δx, Δy, Δz) from the ground observation station O to the satellite can be calculated. The calculation method is shown in formula (6).

[0078]

[0079] Next, assume that the current latitude and longitude of the ground observation station O is (α, β). The Earth-centered Earth-fixed coordinate system, i.e., the Earth-centered rectangular coordinate system, is first rotated about the Z axis by (90° + α) and then about the X axis by (90° - β). This converts the Earth-centered Earth-fixed coordinate system into the station-centered coordinate system. Assume that the observation vector in the station-centered coordinate system is [Δe Δn Δu] T According to the matrix rotation formula, the coordinate rotation matrix T can be obtained as shown in formula (7):

[0080]

[0081] According to the matrix transformation principle, the observation vector in the station center coordinate system is obtained as shown in formula (8):

[0082]

[0083] It is known that a satellite is visible to a ground observation station only when the preset altitude cutoff angle is less than the satellite altitude angle. Therefore, the satellite altitude angle ξ at the current moment must also be calculated. The calculation formula is shown in Equation (9).

[0084]

[0085] The average number of visible stars is obtained by calculating the sum of the number of visible stars at each moment for all feature points in the target area and dividing it by the total simulation time.

[0086] d. Geometric Dilution of Precision

[0087] like Figure 3 The figure shows the distribution of satellites participating in the positioning solution when calculating GDOP. From a geometric perspective, GDOP means that the volume of the geometric volume formed by the unit vector pointing from the user to the four satellites in space is inversely proportional to the GDOP value. Therefore, the larger the volume of the geometric volume formed by the satellites participating in the positioning solution and the user, the wider and more evenly distributed the satellites are, the lower the GDOP value and the higher the positioning accuracy. Conversely, if the angles from the user to the satellites are very close, the GDOP value will increase, and the navigation positioning accuracy will also decrease.

[0088] The calculation method is as follows:

[0089] The coefficient matrix H of the observation equation can be obtained using the satellite elevation angle ξ and the satellite azimuth angle α, as shown in Equation (10).

[0090]

[0091] Where n represents the total number of satellites involved in the positioning solution, and the subscripts 1, 2, ..., n represent the satellite numbers corresponding to the parameters.

[0092] so,

[0093]

[0094] Where D 11 、D 22 、D 33 、D 44 Represents the three-dimensional position (x, y, z) and clock deviation of the feature point in the observation area The mean square error.

[0095] The value of GDOP can be expressed as formula (12):

[0096]

[0097] 2. Optimization process description

[0098] The large language model used in this paper is ChatGPT-3.5-turbo. This method can generate a variety of large models that are suitable for different task requirements and application scenarios. The optimization method using a large language model as an optimizer uses the large model as a black box to optimize the constellation configuration. Its main advantage is that task requirements can be described in natural language, and the code is more readable and versatile than traditional algorithms.

[0099] An overview of the large language model optimization algorithm is shown in the figure below: Figure 4 The specific steps are as follows:

[0100] (1) The LLM initially randomly generates n solutions based on the mission description, where the solutions include different numerical designs for different satellites in the constellation. In each subsequent iteration, the LLM randomly generates m solutions based on the mission description and historical solutions.

[0101] (2) Pass the newly generated solution to the evaluation function for performance evaluation and obtain the corresponding evaluation score;

[0102] (3) Compare the scores of the new solution with the original n solutions, retain the n best solutions and write them into the solution library;

[0103] (4) Combine the solution library and task requirements to generate a new meta-prompt;

[0104] (5) LLM then generates m new, better solutions based on the task description and historical solutions, and repeats step (1);

[0105] (6) Until the number of optimization steps reaches the maximum, the optimal solution is output, and the satellites under the constellation are configured through the optimal solution to achieve the above-mentioned multi-objective optimization design.

[0106] like Figure 5 As shown, the specific implementation steps of the present invention are as follows:

[0107] (1) Determine the fixed parameters of the Walker-δ constellation configuration, including the total number of satellites T, the number of orbital planes P, the phase parameter F, and the longitude and latitude range of the area to be observed; take the orbital inclination i and orbital altitude h as design variables, and determine the value range of each design variable [i min ,i max ],[h min ,h max ]; On this basis, the coverage percentage of earth observation is C r, payload ground pixel resolution R P , the average number of visible satellites N and the geometric dilution of precision GDOP are used as performance indicators, and the mathematical model of the multi-objective optimization design problem of the satellite constellation is established as shown in formula (13):

[0108] find x=[h,i]

[0109]

[0110] st i min ≤i≤i max ,h min ≤h≤h max ,C r =100

[0111] (2) Based on the above-mentioned method, the Gaussian constellation orbit dynamics equation, satellite constellation ground coverage analysis model, satellite constellation ground observation resolution model, visibility analysis model and precision index model are established respectively. Then, the simulation platform built by STK is used to calculate the four performance indicators.

[0112] (3) Using a large language model as an optimization method to generate a new constellation configuration scheme;

[0113] (4) The evaluation function takes the maximum coverage percentage and average number of visible stars and the minimum ground pixel resolution and GDOP as optimization objectives. A single-objective optimization function, namely the evaluation function F(x), is constructed through a weighted method, as shown in Equation (14). This converts the multi-objective problem into a single-objective optimization problem. The new solution is then passed to the evaluation function to obtain the corresponding score.

[0114] find x=[h,i]

[0115]

[0116] st i min ≤i≤i max ,h min ≤h≤h max , C r =100

[0117] Among them, w1, w2, w3, and w4 are the weights of the corresponding targets.

[0118] (5) Retain n optimal solutions based on the scores of the new solution and the original solution, and generate a meta-hint together with the task description and pass it to the LLM. The task description mainly includes some constraints and requirements for the output of the large model. The specific constraints include the range of orbital altitude and inclination. The output requirement is the number of better solutions generated according to the above goals and the specific format of the LLM output.

[0119] (6) Determine whether the number of analysis model calls has reached the maximum value. If not, return to step 3 to continue the optimization process. Otherwise, the optimization process terminates and outputs a set of satellite constellation designs with the optimal target value. This achieves a high coverage percentage of the constellation over the target observation area and a small pixel resolution for the payload, thereby reducing the computational and design costs of the satellite constellation.

[0120] In order to better demonstrate the effectiveness and engineering practicality of the present invention, the present invention will be further explained below by taking a specific multi-functional satellite constellation optimization design problem as an example.

[0121] In this case, a low-orbit Walker constellation was established, with a total of 18 satellites, 3 orbital planes, 6 satellites per orbital plane, and a phase factor of 1. Each satellite carries a sensor with a half-angle of 5°. The pixel size of the Earth observation payload is d = 10 μm and the focal length is f = 500 mm. The longitude and latitude of the target area to be observed range from 73°E to 135°E and 3°N to 53°N. The TLE ephemeris data of the BeiDou-3 satellite was then imported into the current STK scene to analyze the navigation enhancement effect of the Walker constellation on the BDS-3 constellation. During the simulation, the orbital altitude h and orbital inclination i were used as optimization variables, with the value ranges of orbital altitude h∈[500km, 2000km] and orbital inclination i∈[30°, 60°], respectively. During the optimization process, the large model generated 10 solutions based on the mission requirements in the first round. After that, the number of new solutions generated in each optimization was 4, and the maximum number of model calls was 20. The designed evaluation function is: The optimization results of the multifunctional satellite constellation configuration design are as follows: Figure 6 At the same time, the present invention also uses the improved genetic algorithm as the control group to conduct experiments, and the obtained optimization results are shown in Figure 7 The comparison results of the large language model optimization method and the improved genetic algorithm are shown in Table 1.

[0122] Table 1

[0123]

[0124] from Figure 6-7 As can be seen from the results, the optimization method using the large language model achieved superior results compared to the improved genetic algorithm. It offers significant advantages in handling complex optimization problems, overcoming the vulnerability of traditional algorithms to falling into local optima and demonstrating greater adaptability and flexibility. Furthermore, its advantages in code generalization were even more pronounced. The solution derived from the large model optimization algorithm was ultimately selected as the final design.

[0125] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A multifunctional satellite constellation configuration design optimization method based on a large language model, characterized in that: include: Setting design variables and fixed variables, wherein the design variables include orbital inclinations and orbital altitudes of satellites in the satellite constellation; Construct an evaluation function based on coverage percentage, ground pixel resolution, average number of visible stars, and geometric dilution of precision; Based on the design variables and fixed variables, a solution is iteratively generated through a large language model. The solution is evaluated through an evaluation function, and based on the evaluation results, the optimal satellite constellation configuration design is obtained. The objective optimization function is in, are the weights of the corresponding targets, F(x) is the evaluation function, is the coverage percentage, is the dimensionless ground pixel resolution after normalization, N is the average number of visible stars, GDOP is the geometric dilution of precision, is the design variable, is the orbit height, is the orbital inclination, where the subscripts min and max represent the corresponding minimum and maximum values; The process of obtaining the optimal satellite constellation configuration design includes: Initializing design variables, optimizing the design variables using the large language model as an optimizer to obtain a solution, wherein the solution is a new constellation configuration scheme; Evaluate the solution through the evaluation function to obtain the evaluation score; Based on the evaluation scores, the solutions are screened to obtain the optimal solution. Meta-prompts are generated based on the optimal solution and the task description. The meta-prompts are predicted and optimized through the large language model to generate new solutions. Repeat the solution generation, evaluation and screening steps until the maximum number of iterations is reached, and then the solution with the best evaluation score is obtained as the optimal satellite constellation configuration design solution.

2. The method according to claim 1, characterized in that The process of obtaining coverage percentage includes: The coverage angle, center angle, and coverage bandwidth are calculated based on the orbital altitude of a single satellite and the corresponding sub-satellite point; the longitude and latitude of the ground target and the longitude and latitude of the sub-satellite point are obtained based on the coverage angle, center angle, and coverage bandwidth; the geocentric angle is calculated based on the longitude and latitude of the ground target and the longitude and latitude of the sub-satellite point; the geocentric angle and coverage angle are judged; if the geocentric angle is less than or equal to the coverage angle, the ground target is included in the coverage range of the current satellite; Make a grid map of the observation area, use the grid points as feature points, and use the feature points as ground targets to determine whether the ground targets are included in the coverage range of the current satellite. Obtain the total time that all feature points in the observation area are covered by one or more satellites, and divide the total time by the total simulation time to get the coverage percentage.

3. The method according to claim 1, characterized in that The process of obtaining ground pixel resolution includes: The ground pixel resolution of the payload is calculated based on the orbit altitude, the pixel size of the satellite payload, and the focal length of the satellite's optical system, and the ground pixel resolution of the payload is normalized to obtain a normalized dimensionless ground pixel resolution.

4. The method according to claim 1, wherein The process of obtaining the average number of visible stars includes: Using the geocentric rectangular coordinate system as the reference system, obtain the coordinates of the ground observation station and the satellite. Based on the coordinates of the ground observation station and the satellite, obtain the observation vector from the ground observation station to the satellite. Convert the observation vector from the geocentric rectangular coordinate system to the station-centered coordinate system based on the ground observation station. Calculate based on the converted observation vector to obtain the satellite elevation angle. When the preset altitude cutoff angle is less than the satellite elevation angle, the satellite is visible to the ground observation station. Make a grid map of the observation area, use the grid points as feature points, and use the feature points as ground observation stations to determine whether the satellite is visible to the ground observation station. Get the sum of the number of visible stars at all feature points at each moment, and divide the sum of the number of visible stars by the total simulation time to get the average number of visible stars.

5. The method according to claim 4, characterized in that The process of obtaining the geometric dilution of precision includes: According to the satellite elevation angle and satellite azimuth angle, a coefficient matrix is ​​obtained. According to the coefficient matrix, the mean square error of the three-dimensional position and clock bias is obtained. According to the mean square error of the three-dimensional position and clock bias, the geometric precision factor is calculated.

6. The method according to claim 1, characterized in that Fixed variables include the total number of satellites, the number of orbital planes, phase parameters, and the longitude and latitude range of the observation area.

Images