A Site Selection Method for Lunar Antarctic Research Stations Based on a Multimodal Large Model
By constructing a multimodal large model for site selection of a lunar south pole research station, the problems of insufficient multimodal data fusion and adaptability were solved, and efficient and accurate site selection decisions were achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- PLA PEOPLES LIBERATION ARMY OF CHINA STRATEGIC SUPPORT FORCE AEROSPACE ENG UNIV
- Filing Date
- 2026-02-03
- Publication Date
- 2026-06-02
AI Technical Summary
Existing technologies struggle to effectively integrate multimodal data, lack adaptability and high-level understanding capabilities, and are ill-suited for precise site selection in the complex environment of the lunar south pole.
Construct a multimodal large model, unify data representation through multimodal semantic grids, use the multimodal large model for data fusion and adaptive decision-making, generate site selection evaluation weights and parameters, and perform closed-loop optimization.
It achieves deep fusion and unified modeling of multi-source data, improves the objectivity and adaptability of site selection decisions, and enhances the accuracy and efficiency of site selection in extreme environments.
Smart Images
Figure CN122133057A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of deep space exploration engineering planning and artificial intelligence technology, specifically to a method for selecting a lunar south pole research station based on a multimodal large model. Background Technology
[0002] Due to its unique geographical location and environmental conditions, the lunar south pole region has become a focal point for international lunar exploration and the construction of future lunar bases. This region has extremely complex terrain, with numerous permanently shadowed (PSR) and permanently illuminated areas, containing valuable volatile resources that may exist in the form of water ice, making it of immense scientific exploration and resource utilization value. China, the United States, the European Space Agency, and others have all planned or implemented exploration missions to the lunar south pole, with the establishment of long-term research stations as one of their core objectives.
[0003] Selecting the optimal site for a lunar research station, especially in the harsh and constrained Antarctic region, is a complex systems engineering project involving the fusion of multi-source heterogeneous data and multi-objective integrated decision-making. Site selection must take into account many factors, including engineering feasibility (such as terrain slope, illumination continuity, and communication visibility), resource availability (such as water ice distribution), scientific value (such as geological features and proximity to permanently shadowed areas), and long-term operational risks (such as temperature stability).
[0004] Currently, researchers both domestically and internationally have proposed various methods for selecting lunar landing sites or bases. Early studies were mostly based on expert knowledge, using thresholds to set for single or a few key factors (such as permanently shadowed areas, hydrogen abundance, and slope) and performing overlay analysis to screen potential areas of interest. With the enrichment of data sources, multi-factor comprehensive analysis techniques combining more constraints (such as Earth visibility and landing ellipse size) have emerged. In recent years, to address the complexity of site selection problems, researchers have begun to introduce computer intelligence methods, such as using hierarchical clustering and K-means clustering to classify terrain regions, or constructing deep learning models based on convolutional neural networks (CNNs) to quantitatively assess site suitability.
[0005] However, existing technical solutions still have significant limitations: First, traditional methods struggle to effectively integrate multimodal and heterogeneous data, such as high-resolution remote sensing images, temporal illumination, 3D terrain, and communication links, for unified representation and joint reasoning. Second, in most methods, the weighting of different site selection indicators heavily relies on the prior experience of domain experts, exhibiting strong subjectivity and lacking the ability to dynamically and adaptively adjust to specific tasks and environments. Finally, facing the highly nonlinear and multi-coupled complex environment of the lunar south pole, existing solutions based on linear superposition or shallow machine learning models lack sufficient modeling and comprehensive decision-making capabilities, making it difficult to accurately capture and balance various conflicting constraints and objectives.
[0006] Therefore, there is an urgent need to develop an intelligent site selection method that can deeply integrate multimodal data, possess high-level understanding and reasoning capabilities, and adaptively optimize the decision-making process, in order to support the efficient, scientific, and reliable planning of future lunar Antarctic research stations. Summary of the Invention
[0007] To achieve the objectives of this invention, this application provides a method for selecting a lunar south pole research station based on a multimodal large model, comprising: Step S1: Encapsulate the evaluation models and calculation methods involved in the site selection process of the lunar research station into callable tool functions to form a site selection tool library; Step S2: Acquire multi-source data of the lunar south polar region, unify the multi-source data to the same coordinate system and resolution, construct a multimodal semantic raster representing different site selection factors, and form a comprehensive cost model; Step S3: Encode the environmental features represented by the multimodal semantic grid and the preset addressing task target into a cue vector using an addressing encoder; Step S4: Calculate the similarity between the prompt vector and the preset embedding vectors of each indicator tool capability using a multimodal large model, select the K tool chains with the highest matching degree as candidates, and generate the initial values of the location evaluation weight parameters and tool parameters corresponding to the current task using the multimodal large model. Step S5: Based on the weight parameters and initial parameter values, comprehensively evaluate the candidate locations within the target region; organize the evaluation results into structured feedback information and input it into the multimodal large model, which adaptively adjusts the weights and parameters, and iteratively executes the evaluation and adjustment process until the preset convergence condition is met; Step S6: Output the final research station site selection result that meets the convergence conditions, and trigger the dynamic update of the site selection process when new exploration data or mission objective change instructions are received.
[0008] In some specific embodiments, the tool functions include: an input data interface, an output metric score or risk assessment result, and weight or threshold parameters that can be adjusted by a large model.
[0009] In some specific embodiments, in step S1, the site selection tool library includes at least one of the following tool types: illumination continuity assessment tool, temperature stability assessment tool, terrain stability assessment tool, communication visibility analysis tool, scientific value potential assessment tool, and comprehensive risk assessment tool.
[0010] In some specific embodiments, a sexuality grid, a temperature risk grid, a communication cost grid, and a scientific value grid are used.
[0011] In some specific embodiments, in step S3, the cue vector includes at least one of the following feature codes: Antarctic environmental complexity features, temperature and light suitability features, communication accessibility features, water ice content features, and permanent shadow region proximity features.
[0012] In some specific embodiments, in step S4, the similarity is determined according to the following formula: In the formula, The cue vector representing the current location selection task; B k Indicates the first The capability embedding vector of a toolchain.
[0013] In some specific embodiments, in step S4, the site selection evaluation weight parameter corresponds to at least one of the following evaluation indicators: energy availability indicator, terrain constructability indicator, communication continuity indicator, scientific value indicator, and risk constraint indicator.
[0014] In some specific embodiments, in step S5, the structured feedback information includes: the comprehensive cost value of each candidate location, the comparison results of each individual indicator value with the preset threshold, and the specific reasons for not meeting the location constraints.
[0015] In some specific embodiments, in step S5, the adaptive adjustment operation performed by the multimodal large model includes at least one of the following: enhancing the penalty for the weights corresponding to indicators that exceed the threshold, adjusting the sensitivity parameters of relevant evaluation tools, and rebalancing the weight relationships between multiple conflicting indicators.
[0016] In some specific embodiments, in step S5, the preset convergence condition is any one of the following: the change in the comprehensive cost value is less than a preset threshold, all candidate positions meet the preset location constraints, and the number of iterations reaches a preset maximum value.
[0017] The beneficial effects of the above technical solution are as follows: Compared with existing technologies, the lunar south pole research station site selection method based on a multimodal large model provided by this invention has the following significant advantages: 1. Achieved deep fusion and unified modeling of multi-source heterogeneous data: By constructing a multimodal semantic grid, data of different dimensions, sources and formats (such as remote sensing images, digital elevation models, time-series illumination data, water ice inversion data, etc.) of terrain, illumination, temperature, communication, and scientific value are unified under the same spatial framework for representation. This solves the problem that traditional methods are difficult to effectively integrate and collaboratively utilize multimodal data, and provides high-quality and structured input for subsequent comprehensive reasoning.
[0018] 2. It reduces the over-reliance on prior human experience in site selection decisions, improving the objectivity and adaptability of the decisions: Utilizing the semantic understanding and reasoning capabilities of a multimodal large-scale model, it dynamically calculates the matching degree between site selection task prompts and various evaluation toolchains, and automatically generates indicator weights and initial parameter values adapted to specific task objectives and current environmental characteristics. This process reduces subjective bias caused by manually setting fixed weights, enabling site selection criteria to flexibly adapt to different regional characteristics and task requirements.
[0019] 3. Enhanced modeling and integrated decision-making capabilities for the complex nonlinear coupled environment of Antarctica: The multimodal large model can capture and handle the complex nonlinear interactions between factors such as illumination, topography, and temperature. Through a closed-loop optimization mechanism of "evaluation-feedback-adjustment," the model can continuously learn and optimize decision-making strategies, thereby finding a better balance between conflicting engineering constraints and scientific objectives, and improving the scientific rigor and robustness of site selection decisions in extremely complex environments.
[0020] 4. Significantly improves the intelligence and efficiency of site selection assessment: This method combines expert knowledge, physical models, and large-scale model reasoning capabilities to form a toolchain that can be dynamically invoked, combined, and optimized. This automates multiple stages from data analysis and indicator calculation to comprehensive evaluation, reduces repetitive manual analysis workload, and enables rapid and refined screening and assessment of broad areas, providing efficient tool support for task planning.
[0021] 5. Excellent scalability and mission adaptability: The site selection index toolkit and the cue vector-based mission parsing mechanism exhibit strong modularity. By updating the toolkit or adjusting mission cue, this method can easily incorporate new site selection factors, adapt to updated exploration data, or be migrated for base site selection missions on the surfaces of other celestial bodies such as Mars, demonstrating broad application prospects.
[0022] In summary, by introducing a multimodal large model, this invention effectively solves key problems in the site selection of lunar Antarctic research stations, such as the difficulty of multimodal data fusion, strong decision-making subjectivity, and insufficient ability to model complex environments. It provides a more intelligent, adaptive, and efficient site selection technology approach, which has important practical value for promoting the early planning and demonstration of deep space exploration missions. Attached Figure Description
[0023] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0024] Figure 1This is a flowchart illustrating a method for selecting a lunar Antarctic research station based on a multimodal large model, as provided in one embodiment of the present invention. Detailed Implementation
[0025] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0026] Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar symbols denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the invention, and should not be construed as limiting the invention.
[0027] Example 1 One embodiment of the present invention provides a method for site selection of a lunar Antarctic research station based on a multimodal large model, comprising: Step S1: Encapsulate the evaluation models and calculation methods involved in the site selection process of the lunar research station into callable tool functions to form a site selection tool library; In one specific embodiment of the present invention, the tool function includes: an input data interface, an output indicator score or risk assessment result, and weight or threshold parameters that can be adjusted by a large model.
[0028] In a specific embodiment of the present invention, in step S1, the site selection tool library includes at least one of the following tool types: illumination continuity assessment tool, temperature stability assessment tool, terrain stability assessment tool, communication visibility analysis tool, scientific value potential assessment tool, and comprehensive risk assessment tool.
[0029] The evaluation models and calculation methods involved in the site selection process for lunar research stations are encapsulated into callable tool functions, forming a site selection tool library: Each utility function should include at least: Input: Regional raster data, time series data, and constraints; Output: Individual indicator score or risk assessment result; Parameter interface: Weights or threshold parameters that can be adjusted by the large model.
[0030] Step S2: Acquire multi-source data of the lunar south polar region, unify the multi-source data to the same coordinate system and resolution, construct a multimodal semantic raster representing different site selection factors, and form a comprehensive cost model; In a specific embodiment of the present invention, the multi-source data acquired in step S2 includes: the digital elevation model and slope grid of the Lunar Orbiter Laser Altimeter (LOLA), digital orthophotos of Chang'e-2, remote sensing images of the Lunar Reconnaissance Orbiter Wide Angle Camera (WAC) and Narrow Angle Camera (NAC), time-series data of polar illumination intensity and inversion data of permanent shadow distribution based on the LOLA digital elevation model, Earth visibility data, and inversion data of polar water ice content based on the Lunar Prospector neutron spectrometer.
[0031] In one specific embodiment of the present invention, a sexuality grid, a temperature risk grid, a communication cost grid, and a scientific value grid are used.
[0032] Acquire multi-source data of the lunar south polar region, including: Digital elevation model and slope grid of the Lunar Orbiter Laser Altimeter (LOLA); Digital orthophotos from Chang'e-2 (7m / pixel), remote sensing images from the Lunar Reconnaissance Orbiter Wide Angle Camera (LRO WAC, 100m / pixel) and the Lunar Reconnaissance Orbiter Narrow Angle Camera (LRO NAC, 0.5-2 m / pixel); Polar illumination intensity time series data based on LOLA DEM; Inversion data of permanent shadow area distribution based on LOLA DEM; Ground visibility data based on LOLA DEM; Polar water ice content inversion data based on the Lunar Prospector neutron spectrometer.
[0033] The above data is projected onto the same coordinate system and resolution, and a multimodal semantic raster is constructed: A unified comprehensive cost model is formed: in, The corresponding indicator weights.
[0034] Step S3: Encode the environmental features represented by the multimodal semantic grid and the preset addressing task target into a cue vector using an addressing encoder; In a specific embodiment of the present invention, in step S3, the prompt vector includes at least one of the following feature codes: Antarctic environmental complexity features, temperature and light suitability features, communication accessibility features, water ice content features, and permanent shadow region proximity features.
[0035] Environmental features and task objectives are encoded into cue vectors using a location-selective encoder. The cue vector includes: Characteristics of the complexity of the Antarctic environment; The temperature is moderate, the light intensity is high, and the communication reachability is high. The moon has a high water ice content; There is a permanently shadowed area nearby.
[0036] Perform a consistency check on the prompt vector. If a missing or abnormal vector is found, return to step S2.
[0037] Step S4: Calculate the similarity between the prompt vector and the preset embedding vectors of each indicator tool capability using a multimodal large model, select the K tool chains with the highest matching degree as candidates, and generate the initial values of the location evaluation weight parameters and tool parameters corresponding to the current task using the multimodal large model. In a specific embodiment of the present invention, in step S4, the site selection evaluation weight parameter corresponds to at least one of the following evaluation indicators: energy availability indicator, terrain constructability indicator, communication continuity indicator, scientific value indicator, and risk constraint indicator.
[0038] The large model calculates the similarity between the suggestion vector and the embedding vectors of each indicator tool capability: In the formula, The cue vector representing the current location selection task; B k Indicates the first The toolchain's capabilities are embedded in a vector. This method measures the similarity of vector directions, with values ranging from [value range missing]. The closer the value is to 1, the more similar they are; the closer it is to -1, the less similar they are; and 0 indicates that they are perpendicular (irrelevant).
[0039] Under the inference control of a multimodal large model, several toolchains that best match the current lunar south pole site selection mission are automatically selected from multiple candidate site evaluation toolchains for subsequent comprehensive evaluation and optimization calculations.
[0040] Select the K combinations of indicators with the highest matching degree as candidates.
[0041] After completing the Top-K toolchain screening, the multimodal large model further generates corresponding site selection evaluation weight parameters and initial values of calculation parameters for the screened toolchains, so as to achieve adaptive site selection evaluation for the specific environment and mission objectives of the lunar south pole.
[0042] enter: The hint vector is used to characterize the characteristics of the lunar south pole environment and the constraints of the site selection mission; The Top-K toolchain candidate set; Capability embedding vectors for each toolchain; The preset site selection task objectives and constraints.
[0043] Based on the above inputs, the multimodal large model performs joint inference on multiple indicators involved in site selection assessment, generating a set of weight parameters for each site selection assessment indicator: Each weight parameter corresponds to a different site selection evaluation index, which may include: energy availability index; terrain constructability index; communication continuity index; scientific value index; and risk constraint index.
[0044] The weight parameters reflect the relative importance of each indicator in the current lunar south pole site selection mission. The large model generates initial values of the weights and threshold parameters of each indicator in combination with the mission objectives.
[0045] Step S5: Based on the weight parameters and initial parameter values, comprehensively evaluate the candidate locations within the target region; organize the evaluation results into structured feedback information and input it into the multimodal large model, which adaptively adjusts the weights and parameters, and iteratively executes the evaluation and adjustment process until the preset convergence condition is met; In a specific embodiment of the present invention, in step S5, the structured feedback information includes: the comprehensive cost value of each candidate location, the comparison result of each individual index value with a preset threshold, and the specific reason type for not meeting the location constraints.
[0046] In a specific embodiment of the present invention, in step S5, the adaptive adjustment operation performed by the multimodal large model includes at least one of the following: enhancing the penalty for the weights corresponding to indicators that exceed the threshold, adjusting the sensitivity parameters of relevant evaluation tools, and rebalancing the weight relationships between multiple conflicting indicators.
[0047] In a specific embodiment of the present invention, in step S5, the preset convergence condition is any one of the following: the change in the comprehensive cost value is less than a preset threshold, all candidate positions meet the preset location constraints, and the number of iterations reaches a preset maximum value.
[0048] S5.1 Initial Site Selection Assessment Based on the output set of weight parameters and the initial values of toolchain parameters, a site selection evaluation is performed on multiple candidate locations within the lunar south pole target area to obtain the initial comprehensive score results for each candidate location.
[0049] S5.2 Multi-indicator Comprehensive Evaluation Based on the initial scoring results, calculate the comprehensive cost of each candidate position: in: Indicates the first The cost or risk value corresponding to each site selection evaluation indicator; This represents the corresponding weight parameters generated in step S4.3. The comprehensive cost value is used to reflect the overall applicability of the candidate position under the current task constraints.
[0050] S5.3 Feedback Organization and Large Model Parameter Tuning The site selection assessment results are converted into structured feedback information, which includes: The comprehensive cost of each candidate position; Whether each individual indicator exceeds the preset threshold; The specific types of reasons why the location constraints are not met.
[0051] The feedback information is sent as new input data to the multimodal large model.
[0052] S5.4 Multimodal Large Model Adaptive Adjustment Based on the feedback information, the multimodal large model adaptively adjusts the weight parameters and toolchain parameters to generate an updated set of weight parameters and initial parameter values.
[0053] The adaptive adjustment includes at least: Increase the penalty for the weights corresponding to indicators that exceed the threshold; Adjust the sensitive parameters of relevant assessment tools; Balancing the relative importance of conflicting indicators such as energy, communications, and security.
[0054] S5.5 Iterative Execution and Convergence Judgment Based on the updated weight parameters and initial parameter values, repeat steps S5.1 to S5.4 to form a closed-loop optimization process of "evaluation-feedback-adjustment".
[0055] The closed-loop optimization process terminates when any of the following conditions are met: The change in the overall cost is less than the preset convergence threshold; All candidate locations satisfy the preset location constraints; The preset maximum number of iterations has been reached.
[0056] Step S6: Output the final research station site selection result that meets the convergence conditions, and trigger the dynamic update of the site selection process when new exploration data or mission objective change instructions are received.
[0057] After terminating the closed-loop optimization, the final research station site selection results are output, including the optimal or near-optimal candidate locations and their corresponding comprehensive scores.
[0058] Output the coordinates, range, and confidence level of the candidate research station area; Update newly added detection data online; Trigger relocation when mission objectives change.
[0059] This application constructs a multimodal semantic grid to unify data from different dimensions, sources, and formats (such as remote sensing images, digital elevation models, time-series illumination data, and water ice inversion data) into a single spatial framework for representation. This solves the problem that traditional methods struggle to effectively integrate and collaboratively utilize multimodal data, providing high-quality, structured input for subsequent comprehensive reasoning.
[0060] This application leverages the semantic understanding and reasoning capabilities of a multimodal large model to dynamically calculate the matching degree between site selection task prompts and various evaluation toolchains, and automatically generates indicator weights and initial parameter values adapted to specific task objectives and current environmental characteristics. This process reduces subjective bias caused by manually setting fixed weights, enabling site selection criteria to flexibly adapt to different regional characteristics and task requirements.
[0061] The multimodal large model constructed in this application can capture and handle the complex nonlinear interactions between factors such as illumination, terrain, and temperature. Through a closed-loop optimization mechanism of "evaluation-feedback-adjustment", the model can continuously learn and optimize decision-making strategies, thereby finding a better balance between conflicting engineering constraints and scientific objectives, and improving the scientific rigor and robustness of site selection decisions in extremely complex environments.
[0062] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
[0063] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. The embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing terminal equipment to cause a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1The steps of the functions specified in one or more boxes. Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the invention. Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.
[0064] The methods and apparatus provided by the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.
[0065] In the description of this specification, references to terms such as "an embodiment," "some embodiments," "example," "specific example," or "a specific embodiment" or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.
[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.
Claims
1. A method for site selection of a lunar south pole research station based on a multimodal large model, characterized in that, include: Step S1: Encapsulate the evaluation models and calculation methods involved in the site selection process of the lunar research station into callable tool functions to form a site selection tool library; Step S2: Acquire multi-source data of the lunar south polar region, unify the multi-source data to the same coordinate system and resolution, construct a multimodal semantic raster representing different site selection factors, and form a comprehensive cost model; Step S3: Encode the environmental features represented by the multimodal semantic grid and the preset addressing task target into a cue vector using an addressing encoder; Step S4: Calculate the similarity between the prompt vector and the preset embedding vectors of each indicator tool capability using a multimodal large model, select the K tool chains with the highest matching degree as candidates, and generate the initial values of the location evaluation weight parameters and tool parameters corresponding to the current task using the multimodal large model. Step S5: Based on the weight parameters and initial parameter values, comprehensively evaluate the candidate locations within the target region; organize the evaluation results into structured feedback information and input it into the multimodal large model, which adaptively adjusts the weights and parameters, and iteratively executes the evaluation and adjustment process until the preset convergence condition is met; Step S6: Output the final research station site selection result that meets the convergence conditions, and trigger the dynamic update of the site selection process when new exploration data or mission objective change instructions are received.
2. The method for selecting a lunar south pole research station based on a multimodal large model according to claim 1, characterized in that, The tool functions include: an input data interface, output indicator scores or risk assessment results, and weight or threshold parameters that can be adjusted by the large model.
3. The method for selecting a lunar south pole research station based on a multimodal large model according to claim 1, characterized in that, In step S1, the site selection tool library includes at least one of the following tool types: illumination continuity assessment tool, temperature stability assessment tool, terrain stability assessment tool, communication visibility analysis tool, scientific value potential assessment tool, and comprehensive risk assessment tool.
4. The method for selecting a lunar south pole research station based on a multimodal large model according to claim 1, characterized in that, In step S2, the multimodal semantic grid includes at least one of the following: terrain cost grid, energy availability grid, temperature risk grid, communication cost grid, and scientific value grid.
5. The method for selecting a lunar south pole research station based on a multimodal large model according to claim 1, characterized in that, In step S3, the prompt vector includes at least one of the following feature codes: Antarctic environmental complexity features, temperature and light suitability features, communication accessibility features, water ice content features, and permanent shadow region proximity features.
6. The method for selecting a lunar south pole research station based on a multimodal large model according to claim 1, characterized in that, In step S4, the similarity is determined according to the following formula: In the formula, The cue vector representing the current location selection task; B k Indicates the first The capability embedding vector of a toolchain.
7. The method for selecting a lunar south pole research station based on a multimodal large model according to claim 1, characterized in that, In step S4, the site selection evaluation weight parameter corresponds to at least one of the following evaluation indicators: energy availability indicator, terrain constructability indicator, communication continuity indicator, scientific value indicator, and risk constraint indicator.
8. The method for selecting a lunar south pole research station based on a multimodal large model according to claim 1, characterized in that, In step S5, the structured feedback information includes: the comprehensive cost value of each candidate location, the comparison results of each individual indicator value with the preset threshold, and the specific reasons for not meeting the location constraints.
9. The method for selecting a lunar south pole research station based on a multimodal large model according to claim 1, characterized in that, In step S5, the adaptive adjustment operation performed by the multimodal large model includes at least one of the following: enhancing the penalty for the weights corresponding to indicators that exceed the threshold, adjusting the sensitivity parameters of relevant evaluation tools, and rebalancing the weight relationships between multiple conflicting indicators.
10. The method for selecting a lunar south pole research station based on a multimodal large model according to claim 1, characterized in that, In step S5, the preset convergence condition is any one of the following: the change in the comprehensive cost value is less than a preset threshold, all candidate positions meet the preset location constraints, and the number of iterations reaches a preset maximum value.