An offshore unmanned underwater vehicle mine detection test planning method based on a large model

Abstract

The application discloses a large model-based mine detection test planning method for offshore unmanned underwater vehicles, and belongs to the technical field of artificial intelligence and underwater equipment test planning. The method constructs a domain knowledge enhanced large model, receives natural language or structured test requirements, fuses marine environment, vehicle performance and test rule constraints, forms a structured planning constraint set through semantic understanding and extraction, generates a layered test scheme through a scenario layer, a planning layer, an environment adaptation layer and an evaluation layer, verifies path accessibility, energy consumption, time length, detection coverage and risk indicators by importing the scheme into a simulation module, forms structured correction information according to simulation feedback, drives the large model to iteratively optimize the scheme until the preset constraints are met. The application realizes the transformation of test planning from artificial guidance to demand driving, automatic generation and closed-loop optimization, effectively improves planning efficiency, executability and adaptability to complex marine environments, and is suitable for the whole-process planning of mine detection tests for offshore unmanned underwater vehicles.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence and underwater equipment test planning technology, specifically a test planning method for mine detection by unmanned underwater vehicles based on a large model. Background Technology

[0002] With the widespread application of unmanned underwater vehicles (UUVs) in tasks such as mine countermeasures, underwater target search, and marine environment detection, mine detection test planning in complex marine environments has become a core aspect of equipment testing organization and performance verification. Such tests require the simultaneous consideration of multiple factors, including mission objectives, mine types, marine environment, platform performance, search strategies, risk control, and assessment indicators. They are characterized by numerous constraints, strong environmental coupling, and complex planning processes.

[0003] Current planning for mine detection experiments using unmanned underwater vehicles (UUVs) primarily relies on human experience, rule templates, or single optimization algorithms. While manual planning, dependent on expert judgment, offers high flexibility, it suffers from high workload, low efficiency, and inconsistent planning results when facing complex sea conditions and multi-constraint tasks. Template-based planning can improve efficiency for repetitive tasks, but its adaptability to changes in task input, environmental scenarios, and equipment characteristics is limited, making it difficult to meet the personalized planning needs of complex scenarios. Traditional optimization models can only optimize single indicators such as path, energy consumption, and time, failing to uniformly address the experimental requirements of natural language processing, the marine environment, platform boundaries, and risk constraints, still requiring significant manual adjustments to the proposed solutions.

[0004] Existing planning methods have significant limitations in complex marine environments. Key environmental factors such as ocean currents, temperature and salinity structure, background noise, water depth, and seabed topography cannot be modeled in conjunction with mission requirements, platform performance, and experimental rules. This can easily lead to problems such as exceeding energy consumption limits, exceeding time limits, insufficient detection coverage, entering risky areas, and unreachable tracks during actual execution, directly affecting the effectiveness of the experiment. At the same time, existing methods generally suffer from high cost of requirement analysis, weak ability to fuse multi-source constraints, incomplete scheme output, and a disconnect between planning and simulation verification. They cannot achieve automatic iterative correction of the scheme and are unable to meet the requirements of efficient, feasible, and adaptive experimental planning.

[0005] In summary, a large-scale model-based experimental planning method for mine detection by unmanned underwater vehicles is proposed to address the above issues. Summary of the Invention

[0006] 1. The technical problem to be solved by the present invention

[0007] The purpose of this invention is to propose a large-scale model-based method for planning mine detection experiments using unmanned underwater vehicles (UUVs) to address the following problems in existing technologies: (1) Existing mine detection test planning is difficult to uniformly analyze and collaboratively model the needs of natural language test, marine environment, equipment performance and rule constraints. It has problems such as high cost of demand structure processing, weak ability to integrate multi-source conditions, incomplete scheme output and reliance on manual intervention, resulting in insufficient planning efficiency and scheme consistency and completeness.

[0008] (2) The test planning scheme and the simulation verification process are separated from each other. It is impossible to automatically correct the indicators such as path, energy consumption, duration, coverage and risk based on the simulation evaluation results. The scheme has poor feasibility and environmental adaptability, and it is difficult to meet the test constraints and engineering implementation requirements in complex marine environments.

[0009] 2. Technical Solution To address the problems of low planning efficiency, insufficient environmental adaptability, weak constraint fusion capability, and difficulty in ensuring the executability of schemes caused by the reliance on manual experience, rule templates, or single optimization models in existing mine detection test planning processes, this invention provides the following technical solution: a large-scale model-based method for planning mine detection tests of unmanned underwater vehicles (UUVs). This method constructs a domain knowledge-enhanced large-scale model for mine detection test planning, performs semantic understanding and structured constraint extraction of test requirements, and combines marine environmental information, UUV platform capability parameters, and mission boundary conditions to generate test schemes that include mission scenarios, path planning, environmental adaptation, and evaluation indicators. Furthermore, simulation verification results are used to provide feedback correction and iterative optimization of the test schemes, thereby improving test planning efficiency, scheme completeness, constraint satisfaction, and environmental adaptability. This invention aims to transform the planning of mine detection tests of UUVs from a manually-led, static approach to a demand-driven, automatically generated, and closed-loop optimized approach, providing an executable, correctable, and scalable planning method for mine detection tests in complex marine test scenarios.

[0010] Specifically, the following steps are included: S1. Data Acquisition: Receive initial test requirements from users in natural language or structured forms, and acquire data including corresponding marine environmental data, unmanned underwater vehicle equipment performance parameters, and test rule constraints; S2. Constraint Analysis and Structured Construction: Construct a knowledge-enhanced large model to jointly analyze the initial requirements of the experiment, marine environmental data, equipment performance parameters and experimental rule constraints, extract the experimental objectives, platform capability boundaries, environmental limitations and evaluation indicators, and form a unified set of structured planning constraints; S3. Generation of hierarchical test plans: Based on the structured planning constraint set, a hierarchical test plan including a scenario layer, a planning layer, an environment adaptation layer, and an evaluation layer is generated through a large model to form a draft test plan; S4. Simulation Verification and Feasibility Assessment: Import the draft test plan into the simulation verification module to quantitatively assess path accessibility, expected energy consumption, mission duration, detection coverage, and risk area crossing to obtain feasibility results; S5. Structured Feedback and Iterative Optimization: If the evaluation results do not meet the preset constraints, structured feedback information is generated and sent back to the large model to iteratively correct the experimental scheme until the constraints are met or the convergence condition is reached. S6. Planning Results Output: The output includes waypoint sequence, speed control, environmental adaptation suggestions, risk list and simulation summary, and meets the preset constraints.

[0011] Preferably, the initial requirements for the test described in S1 include at least: the test sea area, the model of the unmanned underwater vehicle, the type of mine, the detection target, the performance indicators, and the test duration constraints; the marine environmental data include at least: ocean currents, temperature and salinity distribution, background noise, seabed topography, water depth, and prohibited / high-risk areas.

[0012] Preferably, the construction of the knowledge-enhancing large model described in S2 specifically includes: S2.1. Constructing an experimental planning knowledge base: Establishing a domain knowledge base related to the experimental planning of mine detection by unmanned underwater vehicles at sea. The domain knowledge base shall include at least knowledge of unmanned underwater vehicle platforms, mine detection missions, marine environment, experimental rules and constraints, and historical experimental cases. S2.2. Constructing training samples for experimental planning: Based on the domain knowledge base, construct experimental planning sample data for training and adapting large models; the input samples can be experimental task descriptions in natural language or structured parameter tables; the output samples should include at least the experimental objectives, area scope, platform constraints, environmental limitations, recommended search patterns, path planning results, and evaluation metrics. S2.3. Domain Adaptation and Knowledge Enhancement: Based on the general large model, domain adaptation is performed using the test planning sample data to enable the large model to have a professional understanding of the test planning scenario for mine detection by unmanned underwater vehicles; by introducing domain knowledge base content, the large model's ability to identify the relationship between equipment terminology, environmental elements, test constraints and planning rules is enhanced, thereby improving the accuracy and consistency of the model in test requirement analysis and scheme generation.

[0013] S2.4. Output of planning task capability: The domain-adapted large model has the following capabilities: semantic understanding of test requirements in natural language form; extraction of structured planning constraints from test requirements, environmental data and equipment parameters; generation of test plan drafts based on structured planning constraints; and correction and optimization of test plan drafts based on simulation verification feedback.

[0014] Preferably, the set of structured programming constraints described in S2 is uniformly represented as:

[0015] In the formula, Represents the set of constraints for structured programming; Indicates the scope of the test area; Indicates the task time window; Indicates the target detection area; This indicates the maximum speed of an unmanned underwater vehicle. This indicates the economic speed of an unmanned underwater vehicle. Indicates the maximum battery life; Indicates the minimum turning radius; Indicates the detection coverage threshold; Indicates the detection probability threshold; Indicates the upper limit of energy consumption; Indicates a group within a no-navigation zone; This represents a set of high-risk areas; Indicates the environmental noise level; This indicates the level of impact of ocean currents.

[0016] Preferably, the outputs of each layer of the hierarchical experimental planning draft described in S3 are as follows: The scenario layer includes the test objective, mine deployment assumptions, area scope, time window, and test background. The planning layer includes search mode, waypoint sequence, survey line spacing, speed range, flight distance, and mission sequence. The environmental adaptation layer includes strategies for adapting to strong currents, high noise, and complex terrain, as well as trajectory correction, speed adjustment, and risk marking. The assessment layer includes coverage, detection probability, estimated energy consumption, total duration, risk items, and feasibility evaluation.

[0017] Preferably, the simulation verification module in S4 uses quantitative indicators; the task duration is specifically expressed as follows:

[0018] In the formula, Indicates the total duration of the task. Indicates sailing time. Indicates the time of the exploration operation. Indicates turning and maneuvering time. This indicates the return time. Furthermore, the travel time can be expressed based on the waypoint sequence and the speed of each segment. The projected energy consumption is specifically expressed as follows:

[0019] In the formula, This indicates the estimated total energy consumption. This represents the average power of the i-th flight segment. This represents the duration of the i-th flight segment; Indicates the speed of the unmanned underwater vehicle at sea. The base power, This indicates the level of ocean current impact in the sea area corresponding to the i-th segment. This represents the ocean current influence coefficient, used to characterize the additional impact of environmental disturbances on energy consumption. The coverage rate of the detection area is expressed as:

[0020] In the formula, Indicates the coverage rate of the target area. Indicates the actual effective coverage area. This represents the total area of ​​the target detection region.

[0021] The comprehensive evaluation objective function for traversing the risk area is expressed as:

[0022] In the formula, This represents the comprehensive evaluation target value of the experimental planning scheme; Indicates the total distance of the planned route; Indicates the reference voyage distance; This indicates the projected total energy consumption; Indicates the upper limit of energy consumption; Indicates the total duration of the task; Indicates the maximum duration of the task; Indicates risk score; Indicates the coverage rate of the target area; This represents the weight coefficient corresponding to each evaluation indicator.

[0023] As a preferred embodiment, the iteration termination condition in S5 is:

[0024] In the formula, This indicates the preset minimum coverage requirement; This represents the maximum permissible risk threshold; the iteration termination condition can be expressed as the difference between the combined target values ​​of adjacent iterations:

[0025] In the formula, This represents the comprehensive evaluation target value corresponding to the k-th iteration. Indicates the kth The comprehensive evaluation target value corresponding to 1 iteration This indicates the preset convergence threshold.

[0026] Preferably, the S3 generation scheme follows the following principles: safety constraints take precedence, platform capability constraints take precedence over efficiency constraints; when constraints conflict, priority is given to satisfying no-fly zones, minimum turning radius, maximum range, and time window.

[0027] Compared with existing technologies, the present invention provides a method for planning mine detection experiments using an unmanned underwater vehicle based on a large model, which has the following advantages: 1. Improve the efficiency of experimental requirements analysis and planning generation. This invention utilizes a domain knowledge-enhanced large model to perform semantic understanding and structured constraint extraction of experimental requirements. It can uniformly transform experimental inputs in natural language or structured form into a set of structured planning constraints that can be used for planning, thereby reducing reliance on manual understanding, organization, and transformation, and improving the efficiency of experimental requirements analysis and scheme generation.

[0028] 2. Enhanced collaborative modeling capabilities under multi-source conditions. This invention unifies and integrates the analysis of test mission requirements, marine environmental conditions, performance parameters of the unmanned underwater vehicle platform, and test rule constraints, enabling the test planning process to simultaneously consider mission objectives, equipment capability boundaries, environmental impacts, and safety limitations, thereby improving the completeness and consistency of test planning under multi-source heterogeneous conditions.

[0029] 3. Improve the completeness and feasibility of experimental plans. This invention adopts a hierarchical experimental plan generation method that combines scenario layer, planning layer, environment adaptation layer and evaluation layer. It can simultaneously generate task description, path planning results, environment adaptation suggestions and evaluation index information, avoiding the problem of traditional methods that only generate local paths or single parameter plans, thereby improving the completeness and engineering feasibility of experimental plans.

[0030] 4. Enhance the adaptability of the test scheme to complex marine environments. This invention incorporates marine environmental factors such as ocean currents, background noise, seabed topography, and water depth changes during the scheme generation process. By adjusting the track distribution, speed settings, and key area operation strategies through an environmental adaptation layer, the test scheme can enhance its adaptability to changes in complex marine environments and reduce the impact of environmental disturbances on the mission implementation effect.

[0031] 5. Improve the constraint satisfaction and correctability of the test plan. This invention uses simulation verification and iterative optimization mechanisms to verify the generated test plan draft in terms of path reachability, expected energy consumption, task duration, target area coverage, and risk items. Items that do not meet constraints are returned to the large model in a structured feedback form for correction, thereby enabling the test plan to be continuously optimized in the closed-loop process, improving the degree of constraint satisfaction and overall correctability of the plan.

[0032] 6. Improve the reviewability and reusability of test results output. The final test planning results output by this invention not only include execution results such as waypoint sequences, search mode parameters, and motion control parameters, but also supporting results such as mission scenario descriptions, environmental adaptation suggestions, risk assessment lists, and simulation verification summaries. This facilitates pre-implementation review, in-implementation reference, and post-implementation review by test organizers, thereby improving the reviewability, reusability, and application value of the test planning results.

[0033] In summary, this invention realizes the transformation of mine detection test planning for unmanned underwater vehicles from manual and static formulation to demand-driven, automatic generation and closed-loop optimization, which can improve test planning efficiency, scheme integrity, constraint satisfaction and environmental adaptability in complex marine test scenarios. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the overall process of the sea mine detection test planning method based on a large model for unmanned underwater vehicles described in this invention. Figure 2 This is a schematic diagram illustrating the construction of the domain knowledge enhancement model for experiment planning described in this invention; Figure 3 This is a schematic diagram illustrating the generation of the stratified test scheme described in this invention; Figure 4 This is a schematic diagram of the iterative optimization of the test scheme based on simulation feedback as described in this invention; Figure 5 This is a schematic diagram of the test sea area and functional area in an embodiment of the present invention; Figure 6 This is a schematic diagram comparing the test tracks before and after optimization in an embodiment of the present invention. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0036] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0037] Example 1, please refer to Figures 1 to 6 As shown: To address the problems mentioned in the technical solutions, this application provides a method for planning mine detection experiments using an unmanned underwater vehicle based on a large model, such as... Figure 1As shown, the mine detection test planning method based on a large model described in this invention follows the overall process of "requirement input - data acquisition - constraint analysis - scheme generation - simulation verification - iterative optimization - result output". Through the coordinated processing of test requirements, marine environmental information, performance parameters of the unmanned underwater vehicle equipment, and test rule constraints, the automatic planning of mine detection test schemes is achieved.

[0038] Specifically, the method includes the following steps: 1. Requirements Input Step: Receive initial test requirements submitted by users. These requirements support both natural language descriptions and structured forms, and must include at least the test sea area, unmanned underwater vehicle model, mine type, detection target, performance indicators, and test duration constraints. This step completes the collection, format validation, and preprocessing of the input requirements.

[0039] 2. Data Acquisition Steps: Based on the initial requirements of the experiment, acquire basic data related to the experiment planning. The basic data includes at least user requirement data, experiment knowledge base data, real-time or forecasted marine environmental data, and historical experiment data and simulation data, which are used to provide standardized input for the subsequent planning process.

[0040] 3. Constraint Resolution Steps: The initial experimental requirements and basic data are jointly analyzed using a large-scale model to extract experimental objectives, platform capability boundaries, environmental constraints, and evaluation indicators, forming a structured planning constraint set. This structured planning constraint set includes at least the experimental area range, mission time window, maximum speed, endurance limit, minimum turning radius, target detection coverage, energy consumption limit, and no-fly zones. The process of constructing the domain knowledge-enhanced large-scale model for experimental planning can be combined with… Figure 2 The content shown will be explained.

[0041] 4. Scheme Generation Steps: Based on the structured programming constraint set, a test scheme for mine detection by an unmanned underwater vehicle is generated using a large model. The test scheme includes at least a mission scenario, path planning results, environmental adaptation suggestions, and test evaluation indicators, used to form an executable draft test plan. The hierarchical generation process of the test scheme can be combined with... Figure 3 The content shown will be explained.

[0042] 5. Simulation verification steps: Import the draft test plan into the simulation verification module to verify the path accessibility, expected energy consumption, mission duration, detection coverage, and crossing of key risk areas of the plan, and obtain the feasibility assessment results of the plan.

[0043] 6. Iterative optimization steps: When the feasibility assessment results of the proposed scheme do not meet the preset constraints, the corresponding feedback information is re-input into the large model to revise and optimize the draft experimental plan until the preset constraints are met; wherein, the feedback information includes at least energy consumption exceeding limits, insufficient coverage, path conflict, and environmental risk. The closed-loop optimization process based on simulation feedback can be combined with... Figure 4 The content shown will be explained.

[0044] 7. Results Output Steps: Output the final test planning results that meet the preset constraints. The final test planning results shall include at least the waypoint sequence, speed and motion control parameters, test implementation instructions, risk assessment list and simulation verification summary.

[0045] Through the above process, this invention realizes the transformation of mine detection test planning for unmanned underwater vehicles from being dominated by human experience to being demand-driven, automatically generated, and optimized in a closed loop, thereby improving the efficiency of test planning, the completeness of the plan, and the adaptability to complex marine environments.

[0046] Specifically, this embodiment includes: S1: Construction of a domain knowledge-enhanced large model for experiment planning; like Figure 2 As shown, this invention first constructs a domain knowledge-enhanced large model for the planning of mine detection experiments for unmanned underwater vehicles, which is used to realize the understanding of experimental requirements, extraction of structured constraints, generation of planning drafts, and iterative correction based on simulation feedback.

[0047] Specifically, the construction of the domain knowledge enhancement big model includes the following: 1. Construct a test planning knowledge base: Establish a domain knowledge base related to the test planning of mine detection by unmanned underwater vehicles. The domain knowledge base shall include at least knowledge of unmanned underwater vehicle platforms, mine detection missions, marine environment, test rules and constraints, and historical test cases.

[0048] 2. Construct training samples for experimental planning: Based on the aforementioned domain knowledge base, construct experimental planning sample data for training and adapting large models. Input samples can be experimental task descriptions in natural language or structured parameter tables; output samples should include at least the experimental objective, region scope, platform constraints, environmental limitations, recommended search patterns, path planning results, and evaluation metrics.

[0049] 3. Domain Adaptation and Knowledge Enhancement: Based on the general large model, domain adaptation is performed using the aforementioned test planning sample data, enabling the large model to possess a professional understanding of the test planning scenario for mine detection by unmanned underwater vehicles. By introducing domain knowledge base content, the large model's ability to identify the relationships between equipment terminology, environmental factors, test constraints, and planning rules is enhanced, thereby improving the accuracy and consistency of the model in test requirement analysis and scheme generation.

[0050] 4. Output of planning task capabilities: The domain-adapted large model has the following capabilities: semantic understanding of test requirements in natural language form; extraction of structured planning constraints from test requirements, environmental data and equipment parameters; generation of test plan drafts based on structured planning constraints; and correction and optimization of test plan drafts based on simulation verification feedback.

[0051] Through the above steps, a large model that can serve as a test plan for mine detection by unmanned underwater vehicles is constructed, providing basic support for subsequent constraint analysis, scheme generation, and closed-loop optimization.

[0052] S2: Analysis of multi-source experimental conditions and representation of structured constraints; After completing the construction of the domain knowledge-enhanced large model for test planning, we further analyze the test requirements, marine environmental information, performance parameters of unmanned underwater vehicles, and test rule constraints to form a set of structured planning constraints that can be used for subsequent scheme generation and optimization.

[0053] Specifically, the multi-source experimental condition analysis and structured constraint representation includes the following: 1. Test Requirement Analysis: A domain-adapted large model is used to perform semantic understanding and key information extraction on the user-input test requirements. These requirements include at least the test task objectives, test area, platform type, mine type, performance indicators, and time requirements.

[0054] 2. Marine Environmental Condition Analysis: Analyze marine environmental data related to the experimental mission to extract key environmental features affecting the experimental plan. The marine environmental data includes at least information such as ocean current velocity and direction, temperature and salinity distribution, background noise intensity, seabed topography, water depth variations, and obstacle area distribution.

[0055] 3. Platform performance parameter analysis: Extract and normalize the performance parameters of the unmanned underwater vehicle equipment. The equipment performance parameters include at least the maximum speed, economic speed, endurance, minimum turning radius, payload capacity, working range of detection equipment, and control accuracy.

[0056] 4. Rule Constraint Analysis: The experimental rules, area boundaries, and safety restrictions are analyzed to form a set of explicit constraints in the planning process. These rule constraints include at least no-navigation zones, sensitive sea area avoidance requirements, mission duration limits, energy consumption limits, safety distance restrictions, and regulatory requirements.

[0057] 5. Construction of Structured Programming Constraint Set: The results of the above-mentioned test requirement analysis, marine environmental condition analysis, platform performance parameter analysis, and rule constraint analysis are uniformly integrated to form a structured programming constraint set. To facilitate the unified representation, storage, and subsequent retrieval of multi-source test conditions, this invention represents the structured programming constraint set as follows:

[0058] In the formula, Represents the set of constraints for structured programming; Indicates the scope of the test area; Indicates the task time window; Indicates the target detection area; This indicates the maximum speed of an unmanned underwater vehicle. This indicates the economic speed of an unmanned underwater vehicle. Indicates the maximum battery life; Indicates the minimum turning radius; Indicates the detection coverage threshold; Indicates the detection probability threshold; Indicates the upper limit of energy consumption; Indicates a group within a no-navigation zone; This represents a set of high-risk areas; Indicates the environmental noise level; This indicates the level of impact of ocean currents.

[0059] The structured planning constraint set includes at least the test area range, mission time window, target detection area, maximum and economic speed, endurance limit, minimum turning radius, detection coverage threshold, detection probability threshold, energy consumption limit, no-navigation zones and high-risk zones, environmental noise level, and ocean current impact level. This structured planning constraint set serves as a unified input for subsequent hierarchical test scheme generation and simulation feedback correction, enabling collaborative modeling and consistent application of test mission objectives, platform capability boundaries, environmental influencing factors, and rule constraints within the same constraint framework.

[0060] 6. Constraint Representation Output: The structured programming constraint set is output as a unified data representation for use in subsequent experimental scheme generation steps. This unified data representation can be a parameter table, a structured field set, or a machine-readable configuration file, ensuring consistent use by the large model during scheme generation and simulation feedback correction.

[0061] Based on the above set of structured programming constraints, we can further combine... Figure 3 The process shown generates a stratified test plan.

[0062] S3: Generation of tiered test schemes Based on the structured planning constraint set formed in step S2, a test plan for mine detection by an unmanned underwater vehicle is generated using a domain knowledge-enhanced large model. This results in a hierarchical test plan that satisfies mission objectives, platform capability boundaries, environmental adaptability requirements, and safety constraints. To facilitate a unified representation of the hierarchical structure of the generated plan, this invention represents the draft test plan as follows:

[0063] in, This indicates the draft of the experimental plan; This indicates the desired layer generation result; This indicates the result generated by the planning layer; This indicates the result of generating the environment adaptation layer; This represents the results generated by the evaluation layer. The outputs of all the above layers together constitute a complete draft experimental plan, which is used for subsequent simulation verification and iterative optimization.

[0064] like Figure 3 As shown, the hierarchical test plan generation process includes scenario generation, planning generation, environment adaptation generation, and evaluation generation. The outputs of each layer together constitute a complete test plan draft.

[0065] 1. Scenario Generation: Based on the test mission objectives, target sea area information, mine type, and mission background constraints, generate test mission scenario content, which includes at least the test mission objectives, target mine type and deployment assumptions, test area range and start and end locations, mission time window, and test background conditions.

[0066] 2. Planning Layer Generation: Based on the output of the scenario layer and the set of structured planning constraints, the core planning content for the unmanned underwater vehicle (UUV) to perform the test mission is generated. This includes at least the search mode selection result, waypoint sequence, survey line spacing, speed range, estimated mission duration, and target area coverage path. The search mode can be selected as parallel search, sector search, reciprocating search, or other search methods suitable for mine detection missions, depending on the mission objective and regional characteristics. The planning layer is used to form the basic trajectory and operational procedures for the UUV to perform the mission. To facilitate a unified representation of the trajectory results generated by the planning layer, the waypoint sequence of the UUV is represented as follows:

[0067] in, Represents a sequence of waypoints. Indicates the first Spatial coordinates of each waypoint These represent the coordinates of the waypoints within the test sea area. Based on this sequence of waypoints, the total distance of the planned route can be expressed as:

[0068] in, This indicates the total voyage of the unmanned underwater vehicle during its test mission. This represents the Euclidean distance between two adjacent waypoints.

[0069] The above representation can formally describe the trajectory path generated by the planning layer as a task execution path formed by connecting multiple waypoints in sequence, and provide basic input for subsequent estimation of total task duration, calculation of expected energy consumption and simulation verification.

[0070] 3. Environment Adaptation Layer Generation: Based on the analysis results of marine environmental conditions, environmental adaptation adjustments are made to the track and operational parameters output by the planning layer. This includes at least suggestions for avoiding strong current areas, suggestions for detection compensation in high-noise areas, suggestions for track correction in complex terrain areas, suggestions for speed adjustment in different sea areas, and the marking results of key risk areas.

[0071] 4. Evaluation Layer Generation: Based on the output results of the scenario layer, planning layer, and environment adaptation layer, generate evaluation layer content for subsequent simulation verification and experimental evaluation, which includes at least the target area coverage index, detection probability index, expected energy consumption index, total mission duration index, risk item list, and feasibility evaluation description.

[0072] 5. Drafting the Experimental Plan: The outputs from the scenario layer, planning layer, environment adaptation layer, and evaluation layer are organized to form a draft experimental plan. This draft plan includes at least mission description information, trajectory parameter information, environment adaptation information, and evaluation index information, and is used for subsequent simulation verification steps.

[0073] In a preferred embodiment, when there are restrictions such as energy consumption upper limit, time window, minimum turning radius or no-navigation zone in the structured planning constraint set, the large model prioritizes satisfying explicit constraint conditions when generating the planning layer and environment adaptation layer content; when there are conflicts between different constraints, the experimental planning draft is generated according to the principle of prioritizing safety constraints and platform capability constraints over efficiency constraints.

[0074] S4: Simulation Verification and Iterative Optimization After the draft test plan is formed, it is imported into the simulation verification module to evaluate the feasibility, constraint satisfaction, and environmental adaptability of the generated plan. When the evaluation results do not meet the preset requirements, the simulation feedback results are returned to the large model to revise and optimize the draft test plan, forming a closed-loop iterative planning process based on simulation feedback.

[0075] like Figure 4 As shown, the simulation verification and iterative optimization process includes five stages: scheme import, simulation evaluation, feedback generation, scheme correction, and termination judgment.

[0076] 1. Scheme Import: Input the draft test plan generated in step S3 into the simulation verification module. The draft test plan shall include at least the mission description information, waypoint sequence, search mode parameters, speed range, environmental adaptation strategy and evaluation index information.

[0077] 2. Simulation Evaluation: A feasibility assessment is conducted on the draft test plan, including at least path reachability, estimated energy consumption, total mission duration, target area coverage, detection probability, minimum turning radius compliance, passage through restricted or high-risk areas, and track deviation under environmental disturbances. Through this assessment, simulation verification results of the draft test plan are obtained to determine whether it meets the mission implementation requirements and platform execution constraints. To facilitate the quantitative evaluation of mission execution time, the total test mission duration is expressed as:

[0078] in, Indicates the total duration of the task. Indicates sailing time. Indicates the time of the exploration operation. Indicates turning and maneuvering time. This indicates the return time. Further, the travel time can be expressed based on the waypoint sequence and the speed of each segment as follows:

[0079] in, This indicates the segment distance between two adjacent waypoints. This represents the planned speed for the i-th flight segment. Using this representation, the track distribution, speed settings, and mission process in the draft test plan can be transformed into calculable mission duration indicators, providing a quantitative basis for subsequent judgments on whether the mission time window constraints are met.

[0080] To facilitate the quantitative assessment of energy consumption during the execution of experimental tasks, this invention expresses the expected total energy consumption as follows:

[0081] in, This indicates the estimated total energy consumption. This represents the average power of the i-th flight segment. This represents the duration of the i-th segment. In a preferred embodiment, to further reflect the impact of environmental factors such as ocean currents on energy consumption, the estimated total energy consumption can also be expressed as:

[0082] in, Indicates the speed of the unmanned underwater vehicle at sea. The base power, This indicates the level of ocean current impact in the sea area corresponding to the i-th segment. The current influence coefficient is used to characterize the additional impact of environmental disturbances on energy consumption. Using this representation, the track distribution, speed settings, and environmental factors in the draft test plan can be transformed into calculable energy consumption indicators, providing a quantitative basis for subsequent assessments of whether energy consumption constraints are met.

[0083] To facilitate a quantitative evaluation of the coverage effect of the target detection area, this invention expresses the target area coverage rate as:

[0084] in, Indicates the coverage rate of the target area. Indicates the actual effective coverage area. This represents the total area of ​​the target detection zone. In one implementation, the actual effective coverage area can be estimated based on the planned path of the unmanned underwater vehicle, the effective detection width of the sonar, and the distribution of key detection areas, and is used to measure the adequacy of the coverage of the target area by the generated test plan.

[0085] To facilitate the comprehensive evaluation and ranking of different experimental planning schemes, this invention further constructs a comprehensive evaluation objective function, expressed as:

[0086] in, This represents the comprehensive evaluation target value of the experimental planning scheme; Indicates the total distance of the planned route; Indicates the reference voyage distance; This indicates the projected total energy consumption; Indicates the upper limit of energy consumption; Indicates the total duration of the task; Indicates the maximum duration of the task; Indicates risk score; Indicates the coverage rate of the target area; This represents the weight coefficient corresponding to each evaluation indicator.

[0087] In the aforementioned comprehensive evaluation objective function, total flight distance, estimated total energy consumption, total mission duration, and risk score are considered cost terms, with smaller values ​​indicating a better solution. Target area coverage is considered a benefit term, with larger values ​​indicating a better solution; therefore, it is introduced as a negative term in the objective function. By uniformly weighting the above multiple indicators, this invention can comprehensively evaluate different test planning schemes and select the test planning scheme that meets the constraints and has better overall performance.

[0088] In a preferred embodiment, the weighting coefficient can be adjusted according to the task's focus. When the task emphasizes battery life, it can be appropriately increased. The value of can be appropriately increased when the task emphasizes sufficient coverage of the target area. The value of can be appropriately increased when the task execution environment is complex and the risk is high. The value of .

[0089] 3. Feedback Generation: When simulation verification results show that the draft test plan has constraint conflicts, insufficient performance, or excessive risk items, structured feedback information is generated. The structured feedback information includes at least the name of the unmet constraint, its corresponding deviation value, location or corresponding sea area segment, risk level, and suggested correction direction. The unmet constraints include at least energy consumption exceeding limits, insufficient coverage, mission duration exceeding limits, path conflict, crossing of restricted navigation areas, and environmental risk. To facilitate a unified representation and subsequent use of simulation verification results, this invention represents the structured feedback information as follows:

[0090] in, Represents a set of structured feedback information; Indicates energy consumption deviation; Indicates the deviation in task duration; Indicates the deviation in target area coverage; This indicates a bias in the risk scoring; This represents the set of constraint conflict terms. In one implementation, each deviation term can be determined based on the difference between the simulation evaluation results and preset constraints. For example, when the expected total energy consumption exceeds the energy consumption limit, When the total task duration exceeds the preset time limit, When the coverage rate of the target area is lower than the preset coverage threshold, When the planned route crosses a restricted navigation area or a critical risk area, the set of constraint conflict terms... The system records the corresponding conflict type and the area where it occurred. For example, when the ocean current in a certain sea area is too strong, causing the expected energy consumption to exceed the preset threshold, the feedback information can mark the area as a high-energy-consumption risk zone and prompt adjustments to the course or a reduction in operational density; when the coverage of a target area is lower than the set threshold, the feedback information can prompt the addition of survey lines or a reduction in the spacing between survey lines.

[0091] By using the structured feedback information representation method described above, multiple deviation results generated during the simulation verification stage can be uniformly transformed into feedback inputs that can be further processed by the large model, thereby providing a clear basis for the subsequent revision of the experimental planning draft.

[0092] 4. Scheme Revision: The structured feedback information is re-input into the domain knowledge-enhanced large model. The large model, combined with the original structured planning constraint set, is used to revise the draft experimental plan. The revision methods include at least adjusting waypoint distribution, changing the search mode, modifying survey line spacing, adjusting speed ranges, reallocating key detection areas, and adding environmental risk avoidance strategies.

[0093] 5. Termination Judgment: When the revised draft test plan meets the preset constraints, the iterative optimization process terminates, and the final test plan result is output; if the preset constraints are still not met, the feedback generation and scheme revision steps continue. The preset constraints include at least the following: expected energy consumption does not exceed the preset energy consumption limit; total task duration does not exceed the preset time limit; target area coverage reaches a preset threshold; the planned path meets the platform's minimum turning radius and accessibility requirements; the planned path does not cross restricted airspace; and key environmental risk items are below a set threshold. To facilitate a unified determination of the iterative optimization termination conditions, this invention expresses the termination conditions as follows:

[0094] in, This indicates the preset minimum coverage requirement; This represents the maximum permissible risk threshold. In a preferred embodiment, if the improvement of the draft experimental plan falls below a preset threshold after multiple iterations, the iteration process can be terminated, and the current optimal feasible solution can be output to avoid ineffective repeated optimization.

[0095] Furthermore, during continuous multi-round iterative optimization, if the change in the comprehensive evaluation target value between two adjacent iterations is less than a preset convergence threshold, the iterative process can be determined to have converged, and the optimization process can be terminated. The determination condition can be expressed as follows:

[0096] in, This represents the comprehensive evaluation target value corresponding to the k-th iteration. Indicates the kth The comprehensive evaluation target value corresponding to 1 iteration This indicates a preset convergence threshold. Through the aforementioned termination determination method, this invention ensures that the output experimental planning results not only meet basic constraints such as energy consumption, duration, coverage, and risk, but also maintain overall performance stability after multiple rounds of feedback correction, thereby improving the executability and optimization efficiency of the final experimental planning scheme.

[0097] S5: Output of Planning Results After completing simulation verification and iterative optimization in step S4, the final test planning results that meet the preset constraints are output for use in the implementation, mission verification and test support of the unmanned underwater vehicle mine detection test.

[0098] Specifically, the planning results output includes the following: 1. Output of Execution Results: Output the core planning results used to guide the unmanned underwater vehicle in carrying out test missions, including at least the waypoint sequence, target area search path, search mode and its parameters, speed range and segment action control parameters, mission start and end positions and mission execution sequence, and key detection area allocation results.

[0099] 2. Supporting Results Output: Output supporting results for test organization, implementation instructions, and risk management, including at least test mission scenario description, marine environment adaptation recommendations, risk assessment list, marking results of restricted navigation areas and high-risk areas, test support requirements description, and simulation verification summary.

[0100] 3. Results Organization and Output Format: The execution results and supporting results are organized according to a unified format to form the final test planning results. The final test planning results can be output in the form of parameter tables, text descriptions, structured configuration files, or a combination thereof, to adapt to different test implementation and task management needs.

[0101] 4. Result Retrieval Method: The final test planning results can be retrieved by test planners, ground control terminals, or simulation verification systems for pre-test verification, test process reference, and subsequent review and analysis. Through a unified data organization method, consistent transfer of test planning results can be achieved between scheme generation, execution verification, and implementation support.

[0102] Example 2: Based on Embodiment 1, but with some differences, the following describes the proposed method for planning mine detection experiments using a large-scale unmanned underwater vehicle, in conjunction with specific examples and accompanying drawings. The specific details are as follows.

[0103] This embodiment takes the mine detection test planning of an unmanned underwater vehicle in a complex hydrological area of ​​the East China Sea as an example to illustrate the mine detection test planning method of the unmanned underwater vehicle based on a large model described in this invention.

[0104] 1. Test Scenario Setting like Figure 5 As shown, the test area in this embodiment is a nearshore test area in the East China Sea, which is characterized by complex ocean currents, a thermocline, seabed undulations, and strong background noise. The test area is further divided into test area boundaries, a no-navigation zone, a strong current zone, a high-noise zone, and a key detection zone. The test subject is a certain type of unmanned underwater vehicle (UUV) equipped with multi-beam side-scan sonar and forward-looking imaging sonar; the test objective is to verify the UUV's target detection capability, positioning accuracy, and mission endurance reliability against bottom mines and moored mines in complex environments.

[0105] In one exemplary implementation, the following test condition parameters may be used:

[0106] The parameters described above are merely illustrative examples used to demonstrate the implementation of the present invention in specific experimental scenarios and do not constitute a limitation on the scope of protection of the present invention.

[0107] 2. Test requirement input and constraint analysis; In this embodiment, the user-inputted test requirements can be expressed as: "Conduct a mine detection test of a certain type of unmanned underwater vehicle in a certain sea area of ​​the East China Sea to verify the detection performance of bottom mines and moored mines in complex hydrological environments. The target area detection coverage rate is required to be no less than 90%, the total test duration is no more than 12 hours, and energy consumption and environmental risks should be minimized." Using the domain knowledge-enhanced large model described in this invention, the above-mentioned experimental requirements are semantically parsed, and combined with marine environmental data, platform performance parameters, and experimental rule constraints, a set of structured planning constraints is formed.

[0108] In this embodiment, the structured planning constraint set includes at least the test area range, key detection area location, no-navigation area location, maximum speed and economic speed, endurance limit, minimum turning radius, coverage threshold, total mission duration limit, energy consumption limit, high current risk area and high noise interference area.

[0109] Through the above analysis, the experimental task requirements in natural language form are transformed into unified constraint inputs that can be used for subsequent scheme generation and simulation optimization.

[0110] 3. Generation of stratified test plans; After obtaining the set of structured programming constraints, domain knowledge is used to enhance the large model and generate hierarchical experimental schemes. For example... Figure 3 As shown, the test plan generated in this embodiment includes a scenario layer, a planning layer, an environment adaptation layer, and an evaluation layer.

[0111] (1) Scenario layer: Construct the test background and mission scenario, and set a scenario in which bottom mines and moored mines are mixed in the target sea area, in which bottom mines are deployed in a dispersed manner and moored mines are deployed in a local cluster manner, forming a typical detection mission background.

[0112] (2) Planning layer: Based on the mission objectives and area scope, the target area search path, waypoint sequence, survey line spacing, speed range, and mission timeline of the unmanned underwater vehicle are generated. In one exemplary embodiment, the initial planning scheme adopts a composite search mode combining plowshare and sector search to generate a waypoint sequence covering the target area and provide a complete mission time schedule for deployment, search, identification, return, and recovery.

[0113] (3) Environmental adaptation layer: Based on ocean currents, thermocline, background noise and seabed topography, the track generated by the planning layer is modified to adapt to the environment. For example, reduce speed and adjust local course in strong current areas; appropriately reduce the spacing between survey lines in high noise areas; and increase safe avoidance distance in areas with large seabed undulations.

[0114] (4) Evaluation layer: Generate evaluation indicators for subsequent simulation verification, including target area coverage, detection probability, positioning accuracy, false alarm rate, total mission duration and endurance utilization.

[0115] Based on the outputs of the above layers, a draft experimental plan is formed.

[0116] 4. Simulation Verification and Iterative Optimization The draft test plan is imported into the simulation verification module, and execution is simulated in a digital twin environment. For example... Figure 4 As shown, the simulation verification module includes at least a dynamic model of an unmanned underwater vehicle, a marine environmental field model, and a mine target characteristic model.

[0117] In this embodiment, the initial planning scheme is simulated and evaluated, and the following initial evaluation results are obtained:

[0118] The simulation results show that the initial planning scheme does not meet the preset requirements of "total duration not exceeding 12 hours" and "coverage rate not less than 90%", and has problems of high energy consumption and significant environmental risks. Therefore, iterative optimization is required.

[0119] In this embodiment, the large model automatically generates correction suggestions based on simulation feedback, including at least: bypassing local segments near strong current areas; adjusting the spacing of survey lines in high-noise areas; reducing the speed range of some segments; redistributing the path density of key detection areas; and smoothing local turning waypoints.

[0120] The revised plan was then evaluated again in the simulation verification module. After the second round of optimization, the optimized planning scheme results are shown in the table above. The comparison shows that the optimized scheme is superior to the initial planning scheme in terms of duration, energy consumption, coverage, and environmental risk.

[0121] 5. Results and Implementation Effectiveness; After completing simulation verification and iterative optimization, the final test planning results are output. The final test planning results include at least the mission scenario specification, the optimal sequence of exploration waypoints and track map, search mode and speed control parameters, marine environment adaptation suggestions, risk assessment list and emergency plan, and a summary of simulation verification and optimization records.

[0122] like Figure 6 As shown, compared with the initial planning scheme, the optimized planning scheme effectively avoids areas with strong currents and high risks, while improving the coverage of the target area and reducing the expected energy consumption and total mission duration.

[0123] This demonstrates that the large-model-based mine detection test planning method for unmanned underwater vehicles described in this invention can achieve closed-loop processing of test requirements analysis, automatic scheme generation, simulation verification, and iterative correction under complex marine environmental conditions, thereby improving test planning efficiency, scheme feasibility, and environmental adaptability.

[0124] The core of this invention is not simply generating text using a large model, but rather constructing a closed-loop method for planning mine detection experiments for unmanned underwater vehicles. This method integrates experiment requirement analysis, environmental and equipment constraint fusion, hierarchical scheme generation, simulation verification, and feedback correction into a single technical process to achieve automatic planning and iterative optimization of the experiment scheme.

[0125] Please refer to the above work process. Figures 1 to 6 .

[0126] It should be noted that the term "comprising" or any other variation thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus 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 apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0127] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for planning mine detection experiments using an unmanned underwater vehicle based on a large model, characterized in that, Includes the following steps: S1. Data Acquisition: Receive initial test requirements from users in natural language or structured forms, and acquire data including corresponding marine environmental data, unmanned underwater vehicle equipment performance parameters, and test rule constraints; S2. Constraint Analysis and Structured Construction: Construct a knowledge-enhanced large model to jointly analyze the initial requirements of the experiment, marine environmental data, equipment performance parameters and experimental rule constraints, extract the experimental objectives, platform capability boundaries, environmental limitations and evaluation indicators, and form a unified set of structured planning constraints; S3. Generation of hierarchical test plans: Based on the structured planning constraint set, a hierarchical test plan including a scenario layer, a planning layer, an environment adaptation layer, and an evaluation layer is generated through a large model to form a draft test plan; S4. Simulation Verification and Feasibility Assessment: Import the draft test plan into the simulation verification module to quantitatively assess path accessibility, expected energy consumption, mission duration, detection coverage, and risk area crossing to obtain feasibility results; S5. Structured Feedback and Iterative Optimization: If the evaluation results do not meet the preset constraints, structured feedback information is generated and sent back to the large model to iteratively correct the experimental scheme until the constraints are met or the convergence condition is reached. S6. Planning Results Output: The output includes waypoint sequence, speed control, environmental adaptation suggestions, risk list and simulation summary, and meets the preset constraints.

2. The method for planning a mine detection test based on a large model using an unmanned underwater vehicle according to claim 1, characterized in that, The initial requirements for the test described in S1 shall include at least the following: test sea area, unmanned underwater vehicle model, mine type, detection target, performance indicators, and test duration constraints; the marine environmental data shall include at least the following: ocean currents, temperature and salinity distribution, background noise, seabed topography, water depth, and prohibited / high-risk areas.

3. The method for planning a mine detection test based on a large model using an unmanned underwater vehicle according to claim 1, characterized in that, The construction of the knowledge-enhancing large model described in S2 specifically includes: S2.

1. Constructing an experimental planning knowledge base: Establishing a domain knowledge base related to the experimental planning of mine detection by unmanned underwater vehicles at sea. The domain knowledge base shall include at least knowledge of unmanned underwater vehicle platforms, mine detection missions, marine environment, experimental rules and constraints, and historical experimental cases. S2.

2. Constructing training samples for experimental planning: Based on the domain knowledge base, construct experimental planning sample data for training and adapting large models; the input samples can be experimental task descriptions in natural language or structured parameter tables; the output samples should include at least the experimental objectives, area scope, platform constraints, environmental limitations, recommended search patterns, path planning results, and evaluation metrics. S2.

3. Domain Adaptation and Knowledge Enhancement: Based on the general large model, the experimental planning sample data is used to perform domain adaptation, so that the large model has the professional understanding of the experimental planning scenario for mine detection by unmanned underwater vehicles. S2.

4. Output of planning task capability: The domain-adapted large model has the following capabilities: semantic understanding of test requirements in natural language form; extraction of structured planning constraints from test requirements, environmental data and equipment parameters; generation of test plan drafts based on structured planning constraints; and correction and optimization of test plan drafts based on simulation verification feedback.

4. The method for planning a mine detection test based on a large model using an unmanned underwater vehicle according to claim 1, characterized in that, The structured programming constraint set described in S2 is uniformly represented as: In the formula, Represents the set of constraints for structured programming; Indicates the scope of the test area; Indicates the task time window; Indicates the target detection area; This indicates the maximum speed of an unmanned underwater vehicle. This indicates the economic speed of an unmanned underwater vehicle. Indicates the maximum battery life; Indicates the minimum turning radius; Indicates the detection coverage threshold; Indicates the detection probability threshold; Indicates the upper limit of energy consumption; Indicates a group within a no-navigation zone; This represents a set of high-risk areas; Indicates the environmental noise level; This indicates the level of impact of ocean currents.

5. The method for planning a mine detection test based on a large model using an unmanned underwater vehicle according to claim 1, characterized in that, The outputs of each layer of the hierarchical experimental planning draft described in S3 are as follows: The scenario layer includes the test objective, mine deployment assumptions, area scope, time window, and test background. The planning layer includes search mode, waypoint sequence, survey line spacing, speed range, flight distance, and mission sequence. The environmental adaptation layer includes strategies for adapting to strong currents, high noise, and complex terrain, as well as trajectory correction, speed adjustment, and risk marking. The assessment layer includes coverage, detection probability, estimated energy consumption, total duration, risk items, and feasibility evaluation.

6. The method for planning a mine detection test based on a large model for unmanned underwater vehicles according to claim 1, characterized in that, The simulation verification module in S4 uses quantitative indicators; the task duration is specifically expressed as follows: in, Indicates the total duration of the task. Indicates sailing time. Indicates the time of the exploration operation. Indicates turning and maneuvering time. This indicates the return time. Furthermore, the travel time can be expressed based on the waypoint sequence and the speed of each segment. The projected energy consumption is specifically expressed as follows: In the formula, This indicates the estimated total energy consumption. This represents the average power of the i-th flight segment. This represents the duration of the i-th flight segment; Indicates the speed of the unmanned underwater vehicle at sea. The base power below, This indicates the level of ocean current impact in the sea area corresponding to the i-th segment. This represents the ocean current influence coefficient, used to characterize the additional impact of environmental disturbances on energy consumption. The coverage rate of the detection area is expressed as: In the formula, Indicates the coverage rate of the target area. Indicates the actual effective coverage area. This represents the total area of ​​the target detection region. The comprehensive evaluation objective function for traversing the risk area is expressed as: In the formula, This represents the comprehensive evaluation target value of the experimental planning scheme; Indicates the total distance of the planned route; Indicates the reference voyage distance; This indicates the projected total energy consumption; Indicates the upper limit of energy consumption; Indicates the total duration of the task; Indicates the maximum duration of the task; Indicates risk score; Indicates the coverage rate of the target area; This represents the weight coefficient corresponding to each evaluation indicator.

7. The method for planning a mine detection test based on a large model for unmanned underwater vehicles according to claim 1, characterized in that, The iteration termination condition in S5 is: in, This indicates the preset minimum coverage requirement; This represents the maximum permissible risk threshold; the iteration termination condition can be expressed as the difference between the combined target values ​​of adjacent iterations: in, This represents the comprehensive evaluation target value corresponding to the k-th iteration. Indicates the kth The comprehensive evaluation target value corresponding to 1 iteration This indicates the preset convergence threshold.

8. The method for planning a mine detection test based on a large model using an unmanned underwater vehicle according to claim 1, characterized in that, The S3 generation scheme follows the following principles: safety constraints take precedence, and platform capability constraints take precedence over efficiency constraints; when there is a conflict of constraints, priority is given to satisfying no-fly zones, minimum turning radius, maximum range, and time window.

Images