Underwater marshalling path planning method and system thereof

Abstract

The application relates to an underwater formation path planning method and system, which comprises the following steps: based on environment data of a water area search area, a target area in the water area search area is rasterized and a detection capability matrix is generated; based on the detection capability matrix, a minimum open covering center point set of a target area in a detection range of a formation as a whole is generated; the minimum open covering center point set is taken as a traversal object, an initial candidate path sequence population is generated through a greedy algorithm, and a multi-objective optimization algorithm is used to perform iterative optimization on paths in the initial candidate path sequence population, so that a global optimal path is generated; the global optimal path is decomposed into independent paths of each platform in the formation; based on preset constraint conditions, all the independent paths are smoothed, actual application paths of each platform in the formation are obtained, and underwater formation path planning is completed. The application uses real-time calculation of environment information, thereby saving training time, hardware cost and the trouble of data labeling.

Description

Technical Field

[0001] This application relates to the field of underwater path planning, specifically to an underwater grouping path planning method and system. Background Technology

[0002] With the rapid development of intelligent technologies, efficient path planning in complex open sea areas has become an important direction in marine exploration research. The search results for underwater targets in open sea areas are affected by the dynamic changes in marine environmental factors and the effectiveness of sonar detection, typically involving complex variables in the marine environment and parameter inputs from various sensors. Open sea areas refer to algorithms capable of adapting to any size ocean region, irregular boundary sea areas, and fluctuations in ocean elements at any time. Furthermore, the information about the planned sea area and the initial deployment information are unknown before planning. Therefore, there are high demands on the robustness, accuracy, and generalization ability of the planning device to obtain rapid response search capabilities under complex dynamic environments and uncertain conditions of the grouped objects.

[0003] Existing path planning methods can be broadly categorized into two types: rule-based traditional planning methods and optimization algorithms based on machine learning (evolutionary learning, reinforcement learning, etc.). Traditional planning methods generate paths using fixed rules, offering advantages such as computational simplicity and high efficiency, like zigzag search and zigzag search. However, in complex open ocean environments, rule-based methods struggle to utilize environmental information and adapt to situations with irregular geometric shapes in the area being explored, typically resulting in low efficiency in practical applications. While path planning methods based on reinforcement learning can better utilize environmental information and features, they often suffer from excessively high computational complexity and training costs, making them generally difficult to apply to open, uncertain, and dynamically complex environments. Summary of the Invention

[0004] This application provides an underwater grouping path planning method and system, which can solve the problems of high computational complexity, high training cost, and difficulty in adapting to path planning in irregular areas of open, dynamic, and complex environments in the prior art.

[0005] In a first aspect, embodiments of this application provide an underwater grouping path planning method, which includes:

[0006] Based on the environmental data of the water search area, the target area in the water search area is rasterized and a detection capability matrix is ​​generated;

[0007] Based on the detection capability matrix, a set of minimum open coverage center points of the target area is generated, with the group as a whole, and the group includes multiple platforms.

[0008] Using the set of minimum open cover center points as the traversal object, an initial candidate path sequence population is generated through a greedy algorithm, and a multi-objective optimization algorithm is used to iteratively optimize the paths in the initial candidate path sequence population to generate the globally optimal path.

[0009] The globally optimal path is decomposed into independent paths for each platform in the formation;

[0010] Based on preset constraints, all independent paths are smoothed to obtain the actual application path of each platform in the formation, thus completing the underwater formation path planning.

[0011] In conjunction with the first aspect, in one implementation, based on environmental data of the water search area, the target area within the water search area is rasterized and a detection capability matrix is ​​generated, specifically including:

[0012] Interpolate the environmental data of the water search area into a continuous underwater environment model;

[0013] The target area in the water search area is discretized into multiple grid cells using a rasterization method;

[0014] The detection performance value of each grid cell is calculated using a continuous underwater environment model, and a detection capability matrix is ​​generated.

[0015] In conjunction with the first aspect, in one implementation, based on the detection capability matrix, a set of minimum open coverage center points for the target area is generated, with the group as a whole, specifically including:

[0016] Initialize all grid points in the rasterized target area to an uncovered state to create an initial target point set;

[0017] Based on the detection capability matrix, the grid point with the highest detection efficiency value is selected from all grid points not covered in the target area as the center point, and this center point is added to the target point set;

[0018] Calculate the detection range when the group passes through the center point as a whole, and mark all grid points within the intersection of the center point, the detection range, and the target area as covered;

[0019] Repeat the above adding and marking process until all grid points in the target area are covered, thus completing the generation of the minimum open coverage center point set.

[0020] In conjunction with the first aspect, in one implementation, using the set of minimum open cover center points as the traversal object, an initial candidate path sequence population is generated through a greedy algorithm, and a multi-objective optimization algorithm is used to iteratively optimize the paths in the initial candidate path sequence population to generate the globally optimal path, specifically including:

[0021] Using the set of minimum open cover center points as the traversal object, an initial candidate path sequence population containing greedy paths and random paths is generated based on a greedy algorithm;

[0022] A multi-objective optimization loss function is constructed based on the total path length and path smoothness of each path in the initial candidate path sequence population.

[0023] A multi-objective optimization algorithm is used to iteratively optimize the initial candidate path sequence population based on the objective function, generating multiple dominant paths;

[0024] Choose the path that is weakly dominant over all other paths from among all dominant paths, and select it as the globally optimal path.

[0025] In conjunction with the first aspect, in one implementation, the multi-objective optimization loss function is:

[0026] F(C) = (L(C), S(C));

[0027] Where L(C) is the total path length of each path, and S(C) is the path smoothness.

[0028] In conjunction with the first aspect, in one implementation, a multi-objective optimization algorithm is used to iteratively optimize the initial candidate path sequence population based on the objective function to generate multiple dominant paths, specifically including:

[0029] Perform mutation operations on several paths in the initial candidate path sequence population to generate several new paths;

[0030] The multi-objective optimization algorithm is used to compare each new path with all paths in the current initial candidate path sequence population one by one based on the multi-objective optimization loss function, and iteratively optimizes to find the dominant path until the initial candidate path sequence population converges.

[0031] In conjunction with the first aspect, in one implementation, the globally optimal path is decomposed into independent paths for each platform in the formation, specifically including:

[0032] Remove redundant intermediate points with unchanged direction from the globally optimal path to generate a simplified path;

[0033] Remove points in the simplified path where the distance between adjacent points is less than the set shortest turning distance, and generate the final path that satisfies the turning distance constraint;

[0034] Based on the final path, the coordinates of the path points of each platform are calculated using platform-related parameters, generating an independent path for each platform in the formation.

[0035] In conjunction with the first aspect, in one implementation, the platform-related parameters include platform spacing coefficient, heading angle, and formation angle.

[0036] In conjunction with the first aspect, in one implementation, all independent paths are smoothed based on preset constraints to obtain the actual application path for each platform in the formation, thus completing the underwater formation path planning. Specifically, this includes:

[0037] Based on the preset minimum turning angle, adjust the coordinates of the path points in all independent paths so that the turning angles of all independent paths meet the minimum turning angle constraint.

[0038] Based on the speed range of each platform, a unified time interval is calculated, and the coordinates of the path points corresponding to platforms that do not meet the speed constraints are adjusted so that each platform can complete the movement from its current point to its next point within the unified time interval.

[0039] Based on the adjusted path coordinates of each platform, the actual application path of each platform in the formation is generated, and the underwater formation path planning is completed.

[0040] Secondly, embodiments of this application provide an underwater formation path planning system, comprising: a first module, a second module, a third module, a fourth module, and a fifth module; the first module is used to rasterize the target area in the water search area based on environmental data of the water search area and generate a detection capability matrix; the second module is used to generate a set of minimum open coverage center points for the detection range of the formation as a whole relative to the target area based on the detection capability matrix, wherein the formation includes multiple platforms; the third module is used to generate an initial candidate path sequence population using a greedy algorithm with the minimum open coverage center point set as the traversal object, and to perform iterative optimization on the paths in the initial candidate path sequence population using a multi-objective optimization algorithm to generate a globally optimal path; the fourth module is used to decompose the globally optimal path into independent paths for each platform in the formation; the fifth module is used to smooth all independent paths based on platform-related parameters to obtain the actual application paths of each platform in the formation, thus completing the underwater formation path planning.

[0041] The beneficial effects of the technical solutions provided in this application include:

[0042] This application provides an underwater formation path planning method and system. It utilizes real-time environmental information to rasterize the search area and generate a detection capability matrix, avoiding the need for pre-collecting large amounts of training data and saving training time, hardware costs, and data labeling. Based on the detection capability matrix, it generates a minimum open coverage center point set, achieving complete coverage of the target area with the minimum number of center points, reducing the complexity of formation path planning. Furthermore, it uses a greedy algorithm to construct an initial candidate path population and combines iterative path optimization with a multi-objective optimization algorithm, ensuring the globally optimal path maintains coverage efficiency while reducing energy loss during platform movement. After decomposing the globally optimal path into independent paths for each platform, it performs smoothing based on preset constraints, thereby maintaining the overall coverage performance of the formation while enabling each platform in the formation to safely and smoothly execute its mission based on the actual application path. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of the underwater grouping path planning method of this application;

[0044] Figure 2 This is a flowchart illustrating step 102 of this application;

[0045] Figure 3 This is a flowchart illustrating step 103 of this application;

[0046] Figure 4 This is a schematic diagram illustrating the coordinate calculation of the platform in this application;

[0047] Figure 5 This is a comparison chart of the path graph results of this application and rule-based path planning methods represented by integrated parallel search;

[0048] Figure 6 This is a comparison chart of the cumulative detection probabilities of this application and rule-based path planning methods, such as integrated parallel search.

[0049] Figure 7 This is a comparison chart of the calculation time for this application and the genetic method. Detailed Implementation

[0050] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0051] This application provides an underwater grouping path planning method and system, which can solve the problems of high computational complexity, high training cost, and difficulty in adapting to path planning in irregular areas of open, dynamic, and complex environments in the prior art.

[0052] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0053] In a first aspect, embodiments of this application provide an underwater grouping path planning method, which includes:

[0054] 101: Based on the environmental data of the water search area, the target area in the water search area is rasterized and a detection capability matrix is ​​generated;

[0055] 102: Based on the detection capability matrix, generate a set of minimum open coverage center points of the target area for each group as a whole, wherein the group includes multiple platforms;

[0056] 103: Using the set of minimum open cover center points as the traversal object, an initial candidate path sequence population is generated through a greedy algorithm, and a multi-objective optimization algorithm is used to iteratively optimize the paths in the initial candidate path sequence population to generate the globally optimal path;

[0057] 104: Decompose the globally optimal path into independent paths for each platform in the formation;

[0058] 105: Based on preset constraints, smooth all independent paths to obtain the actual application path of each platform in the formation, and complete the underwater formation path planning.

[0059] In this application, firstly, based on real-time environmental data of the water search area, such as underwater acoustic propagation characteristics and background noise distribution, the target area is rasterized and discretized to form a two-dimensional grid structure, with each grid cell corresponding to a fixed sea area block. The detection capability matrix is ​​averaged over time to calculate the sensor detection efficiency value of each grid cell, thus constructing a complete detection capability matrix. Secondly, based on this detection capability matrix, the set of minimum open coverage center points for the target area when the group is considered as a whole is determined; that is, a set of center points is found such that when the group is at any center point, its detection range can completely cover a specified portion of the target area, forming an efficient coverage area division. Then, these center points are used as the traversal objects. An initial candidate path sequence population is generated using a greedy algorithm. Then, the Pareto-2 element multi-objective optimization (P-2opt) algorithm (or k-opt algorithm) is used to encode dynamic environmental constraints into the optimization process. The initial population is optimized through iterative multi-objective trade-offs to generate a globally optimal path sequence. Next, the globally optimal path is decomposed into independent path sequences for each platform in the formation, ensuring reasonable path connection between platforms in collaborative operations and avoiding path intersections or conflicts. Finally, based on preset platform physical characteristic constraints, each independent path is smoothed, and the coordinates of path points are adjusted to eliminate abrupt turns, making the path conform to actual navigation capabilities, and ultimately generating a formation path that can be directly used for execution.

[0060] Therefore, this application utilizes real-time environmental information calculations to rasterize the water search area and generate a detection capability matrix, avoiding the need to collect large amounts of training data in advance, thus saving training time, hardware costs, and the hassle of data labeling. Based on the detection capability matrix, a minimum open coverage center point set is generated to achieve complete coverage of the target area with the minimum number of center points, reducing the complexity of group path planning. Furthermore, a greedy algorithm is used to construct an initial candidate path population, and a multi-objective optimization algorithm is combined to iteratively optimize the path, ensuring that the globally optimal path reduces energy loss during platform movement while maintaining coverage efficiency. After decomposing the globally optimal path into independent paths for each platform, smoothing is performed based on preset constraints, thereby enabling each platform in the group to safely and smoothly execute its mission based on the actual application path while maintaining the overall coverage performance of the group.

[0061] Based on the above embodiments, in this embodiment, the target area in the water search area is rasterized and a detection capability matrix is ​​generated based on the environmental data of the water search area, specifically including steps 1011 to 1013:

[0062] Step 1011: Interpolate the environmental data of the water search area into a continuous underwater environment model.

[0063] In this embodiment, the environmental data of the water search area includes discrete point data such as those collected by underwater sonar arrays or buoy sensors, including underwater acoustic propagation attenuation coefficient, background noise spectrum distribution, temperature and salinity gradient, as well as water flow, water depth, and underwater acoustic characteristics.

[0064] Based on environmental data from the search area, a continuous underwater environment model is constructed using a bilinear interpolation algorithm. This model fits discrete data points into a smooth three-dimensional field, effectively characterizing the spatiotemporal continuous variation of the underwater environment and avoiding detection blind spots caused by sparse data points.

[0065] Step 1012: Discretize the target area in the water search area into multiple grid cells using a rasterization method.

[0066] In this embodiment, a fixed-size rasterization method is used to discretize the target region into multiple grid cells (i,j), and the width of each grid cell is... and height The system is dynamically configured according to actual task requirements, thereby dividing the continuous water area into multiple homogeneous regions, with each grid cell representing a fixed sea area unit.

[0067] Step 1013: Calculate the detection performance value of each grid cell using a continuous underwater environment model, and generate a detection capability matrix.

[0068] In this step, based on a continuous underwater environment model, the detection performance value of each grid cell (i,j) is calculated, specifically evaluating the signal-to-noise ratio and detection probability of the target signal within the cell to generate an initial detection capability matrix. In the time dimension, the continuously acquired detection capability matrix M is averaged to eliminate instantaneous noise interference caused by tidal or weather fluctuations, ensuring spatial distribution stability. Finally, the detection capability matrix M is stored as a two-dimensional array, with each position (i,j) corresponding to a fixed region block. The platform's detection capability in that region can be represented by the value of the detection capability matrix at the corresponding position. express.

[0069] The mapping between the actual environmental coordinates (x, y) and the grid position of that point is achieved using the formula:

[0070] accomplish;

[0071] in The maximum value of the ordinate of the rectangular area covered by the detection capability matrix. For the length and width of the grid size, , In This represents the Gaussian function (rounded down).

[0072] Based on the above embodiments, in this embodiment, based on the detection capability matrix, a minimum open coverage center point set is generated for the group as a whole, whose detection range covers the target area. This specifically includes steps 1021 to 1024. Here, open coverage in this application refers to an open set covering the target area (also called the area to be detected), where each open set has the property that when the group is located at a certain point in the open set, the group's detection capability exactly covers every point in the open set. This point is called the center point of the open set, and the center points of all open sets in the open coverage are called all center points of the open coverage.

[0073] Each open set in the open coverage is not a simple geometric shape, but a dynamically generated region based on the detection capability matrix. When the group of platforms is centered on a specific point, the actual detection range of its sensors under specific environmental conditions can completely cover all points within that open set, and the center point is the optimal deployment location for the group. This dynamic distribution based on actual detection performance allows the open coverage to match the irregular boundaries and environmental fluctuations of complex open sea areas, avoiding the inefficiency of path planning caused by fixed grid division in traditional methods.

[0074] in:

[0075] Step 1021: Initialize all grid points of the rasterized target area to an uncovered state to create an initial target point set.

[0076] Specifically, step 1021 discretizes the target detection area into a regular grid structure using rasterization technology. Each grid point corresponds to a specific coordinate location in the sea area, facilitating accurate tracking of the coverage status. The system initializes the state of all grid points to uncovered, that is, it explicitly marks each point as not yet covered by any group's detection capability using a standard data structure, ensuring that subsequent coverage calculations start from a unified starting point. Simultaneously, an empty target point set S is initialized. This set serves as the core output of the algorithm, used to dynamically accumulate all center points that constitute the minimum open coverage. These center points directly determine the optimal deployment position of the group. Initializing the target point set S to an empty initial state avoids interference from historical data, providing a clean starting environment for subsequent steps, enabling the algorithm to efficiently construct the minimum coverage scheme while ensuring that the coverage process is free of redundancy or omissions, meeting the dynamic detection requirements of complex open sea areas.

[0077] Step 1022: Based on the detection capability matrix, select the grid point with the highest detection efficiency value from all grid points not covered in the target area as the center point, and add the center point to the target point set.

[0078] Specifically, step 1022 dynamically evaluates the detection efficiency of all uncovered grid points in the target area through the detection capability matrix. This matrix is ​​used as system input data and is pre-calculated based on real-time marine environmental parameters. Each grid point corresponds to a comprehensive detection efficiency value, which quantifies the acoustic wave propagation efficiency and detection coverage when grouped at that point, reflecting the impact of the actual marine environment on detection capabilities.

[0079] The system iterates through all grid points marked as uncovered, compares the detection efficiency value point by point, and selects the point P with the highest value as the current center point. The detection efficiency value of this point represents the maximum coverage area that the group can achieve in the current uncovered area. After selection, the coordinates of this point are directly added to the initial target point set S.

[0080] Step 1022 ensures that the selection of each center point is based on objective environmental data, avoiding coverage deviations caused by human experience. At the same time, the initial target point set S serves as a dynamic accumulation container, continuously recording the sequence of center points that constitute the minimum open coverage, providing input data for subsequent coverage calculations, enabling the algorithm to achieve efficient and non-redundant coverage planning in complex open sea areas.

[0081] Step 1023: Calculate the detection range when the group passes through the center point as a whole, and mark all grid points in the intersection of the center point, the detection range, and the target area as covered.

[0082] Specifically, step 1023 calculates the actual detection range coverage area A of the group at the selected center point. In practice, the detection range is not a fixed geometric shape, but is dynamically adjusted according to real-time ocean conditions. For example, in shallow sea areas, due to enhanced sound wave reflection, the coverage area is closer to a circle; in deep sea or strong current areas, the coverage area is elliptical, with its major axis aligned with the current direction and its minor axis perpendicular to the current, so as to accurately reflect the actual path of sound wave propagation.

[0083] The system then performs spatial intersection calculations between this dynamically generated detection range and the rasterized target area, identifies the coordinates of all overlapping grid points (including boundary points) using a standard raster intersection algorithm, and updates the status of these points (including the center point itself) from "uncovered" to "covered".

[0084] Step 1024: Repeat the above adding and marking process until all grid points in the target area are covered, thus completing the generation of the minimum open coverage center point set.

[0085] Specifically, step 1024 continuously executes steps 1022 and 1023 through an automated loop mechanism to achieve complete coverage of the target area. In each iteration, the system dynamically evaluates the effectiveness value of the currently uncovered grid points based on the real-time updated detection capability matrix, automatically selecting the point with the highest effectiveness value as the new center point; then, the detection range calculation in step 1023 is performed to generate a dynamic coverage area, and the grid points (including the center point itself) at the intersection of this area and the target area are uniformly marked as covered.

[0086] During the iteration process, the coverage progress is monitored in real time. When no uncovered points are detected in the grid state array, the loop automatically terminates. At this point, the accumulated point sequence in the initial target point set S constitutes the center point set of the minimum open coverage. This set selects the optimal coverage point (i.e., the most efficient point) in the currently uncovered area in each iteration, ensuring the minimization of the number of coverage points. This avoids redundancy or omissions caused by traditional fixed grid coverage, enabling the open coverage scheme to adapt to the irregular coastlines and real-time environmental fluctuations of complex open sea areas.

[0087] Based on the above embodiments, in this embodiment, the minimum open cover center point set is used as the traversal object, an initial candidate path sequence population is generated by a greedy algorithm, and a multi-objective optimization algorithm is used to iteratively optimize the paths in the initial candidate path sequence population to generate the globally optimal path, specifically including steps 1031 to 1034:

[0088] Step 1031: Using the set of minimum open cover center points as the traversal object, an initial candidate path sequence population containing greedy paths and random paths is generated based on a greedy algorithm.

[0089] Specifically, step 1031 constructs an initial candidate path sequence population based on the minimum open cover center point set. The system will use the user-input path starting point... The initial position is set, and the initial population size N is dynamically set according to the actual coverage of the target area. The point indices in the minimum open coverage center point set are arranged in a completely disordered manner to generate N-1 random paths. This ensures that each path forms a unique sequence based on a random combination of center points, thereby fully covering the exploration space of path planning in the initial stage.

[0090] Simultaneously, a greedy algorithm is added to the initial candidate path sequence population to generate one greedy path: from... Starting from the point, calculate the Euclidean distance between the current point and the unvisited center points using the Cartesian coordinates of the nautical chart coordinate system. Continuously select the point with the smallest distance as the next point to visit, until all center points have been traversed, forming a path sequence:

[0091] ;

[0092] Where n is the total number of center points in the minimum open-cover center point set. This greedy path, as an elite individual, is directly included in the initial candidate path sequence population. Because it achieves the shortest cumulative path length and the highest traversal continuity in a local range, it can effectively suppress the long-distance jumps and repeated visits common in random paths, providing a high-quality initial solution for subsequent multi-objective optimization. By ensuring the global exploration depth through N-1 random paths and providing a local optimal solution with one greedy path, the system achieves the fusion of the breadth of random exploration and the greedy strategy in the initial stage. This ensures that the path sequence covers a wide range of feasible solutions from inefficient to efficient, and also includes an efficient subset, improving the convergence efficiency of the algorithm in complex open sea dynamic exploration tasks and avoiding the increase in iterations and waste of computational resources due to insufficient initial population quality.

[0093] Step 1032: Construct a multi-objective optimization loss function based on the total path length and path smoothness of each path in the initial candidate path sequence population.

[0094] In this embodiment, the multi-objective optimization loss function is:

[0095] F(C) = (L(C), S(C)); where L(C) is the total path length of each path and S(C) is the path smoothness.

[0096] Specifically, step 1032 constructs a multi-objective optimization loss function by quantifying the total length and directional continuity of the path. This function serves as the evaluation criterion for subsequent path iterative optimization, and is associated with the physical efficiency and navigation stability of the path.

[0097] The total path length L(C) is the sum of the Euclidean distances between all consecutive points in the path sequence C, that is:

[0098] ;in This represents the chart coordinates of the i-th point on the path. This value reflects the physical distance of the voyage, and the optimization goal is to minimize it to reduce energy consumption and mission duration.

[0099] The path smoothness S(C) is calculated by the dot product of the vectors of consecutive flight segments, as shown in the formula:

[0100] ,in This represents the coordinates of the i-th point in the current path. This represents finding the dot product of vectors, where vectors ( and ) represent the heading vectors of the i-th segment and the (i+1)-th segment, respectively.

[0101] To incorporate maximizing S(C) into the minimization optimization framework, the loss function is F(C) = (L(C), -S(C)), that is, the algorithm pursues the minimization of L(C) and the maximization of S(C) simultaneously during iteration. In actual ocean exploration, a path with high smoothness can effectively suppress the turning frequency and amplitude of the ship, and a smooth path can maintain the continuity and quality of the detection data by reducing the number of turns; at the same time, the minimization of L(C) ensures that the overall path is compact and avoids redundant trajectories.

[0102] This design enables the algorithm to preferentially select a path that is both short and smooth in complex open waters, avoiding the problem of impracticable paths caused by neglecting smoothness in traditional methods, and providing an evaluation basis for subsequent multi-objective optimization.

[0103] Step 1033: Use a multi-objective optimization algorithm to perform iterative optimization on the initial candidate path sequence population based on this objective function to generate multiple non-dominated paths.

[0104] Specifically, step 1033 performs iterative optimization on the initial candidate path sequence population through a multi-objective optimization algorithm. The core lies in dynamically performing mutation operations and realizing population evolution based on the loss function.

[0105] First, perform mutation operations on several paths in the initial candidate path sequence population to generate several new paths: In each iteration, the system randomly selects m paths from the current initial candidate path sequence population, and for each selected path, two different points (excluding the starting point) are selected and (the indices satisfy 0 < i < j ≤ n, where n is the total number of path points), ensuring that the starting point is not involved in the mutation as a fixed starting point to avoid path logic breakage.

[0106] Subsequently, the system reversely recombines the subsequence between indices i and j (including and ) to generate a new path: .

[0107] For example, in the path sequence [P0, P1, P2, P3, P4, P5], if i = 2 and j = 4, the new path becomes [P0, P1, P4, P3, P2, P5]. This operation improves the topological diversity of the path by changing the connection order of the segments.

[0108] Next, the Pareto-2 element multi-objective optimization algorithm is used to compare each new path with all paths in the current initial candidate path sequence population one by one based on the multi-objective optimization loss function, and the dominant path is iteratively optimized until the initial candidate path sequence population converges: The system calculates the loss value F(C') of the new path C' based on the loss function F(C)=(L(C), -S(C)). The loss value F(C') of the new path C' is compared with the F(C) of all paths in the current initial candidate path sequence population in a multi-objective manner. If the newly obtained path by mutation is dominant over some paths in the current population, the new path C' is added to the population, and all paths dominated by C' are deleted to ensure that the population size remains constant at N. In this implementation, if F(C') ≤ F(C) and at least one objective function value is less than F(C), then C' is said to be dominant over C. In this implementation, if the path C’ satisfies L(C’) ≤ L(C) and S(C’) ≥ S(C), and at least one of the above two inequalities does not take the equal sign (that is, at least one of L(C’)<L(C) and S(C’)>S(C) holds), then C' is said to be dominant over C.

[0109] In this embodiment, the iteration termination condition is: when the change rate of the loss value of the optimal path in the population is less than the set threshold in consecutive multiple iterations, it is determined to converge, or the maximum number of iterations T is reached. max =c×n (c is an adaptive constant, and n is the total number of center points in the set of center points of the minimum open cover).

[0110] Step 1034: Select the path that weakly dominates all other paths from all the dominant paths as the global optimal path.

[0111] Specifically, step 1034 selects a path that weakly dominates all other paths from the set of dominant paths generated by multi-objective optimization iteration as the global optimal path for output. The determination of this path is based on the multi-objective comparison mechanism of the loss function F(C)=(L(C), -S(C)): The system traverses all the dominant paths in the population and screens out the path C* that satisfies the weak dominance condition, that is, for any other path C in the population, it satisfies L(C*) ≤ L(C) (the total path length is not inferior to other paths) and S(C*) ≥ S(C) (the path smoothness is not lower than other paths). When the initial population size N is large enough and the iteration process converges, this path constitutes the Pareto optimal solution, and this kind of path sequence can synchronously optimize the navigation efficiency and data quality: the minimization of the total path length directly reduces the ship fuel consumption and mission duration, and the improvement of the path smoothness significantly reduces the turning frequency and amplitude, ensuring continuous acquisition of high-quality detection data in complex open sea areas.

[0112] Based on the above embodiments, in this embodiment, the global optimal path is decomposed into independent paths for each platform in the formation, specifically including steps 1041 to step 1043:

[0113] Step 1041: Remove redundant intermediate points with unchanged direction in the globally optimal path to generate a simplified path.

[0114] Specifically, step 1041 automatically removes redundant intermediate points from the globally optimal path through heading continuity analysis, generating a simplified path. For example, the system calculates a path sequence C=[P0,P1,...,P...]. n Each point P in ] i (i from 1 to n-1) Perform heading detection: Calculate the forward and backward vectors, and compare the angle between the two vectors. If the direction is determined to be unchanged, then P i Removed. This process is executed iteratively to ensure that all points in the simplified path meet the direction change condition. For example, multiple consecutive points are merged into a single point because they have the same heading, simplifying the path sequence from [P0,P1,P2,P3] to [P0,P3].

[0115] This simplified operation effectively reduces the number of path points, avoiding redundant turning commands during straight-line navigation, thereby reducing the computational load on the navigation system and improving real-time response capabilities. At the same time, the integrity of path coverage remains unaffected, as redundant points only exist on straight sections. Even after removal, the detection equipment can still continuously acquire high-quality data at key turning points, providing a simpler and more efficient navigation reference for dynamic detection missions.

[0116] Step 1042: Remove points in the simplified path where the distance between adjacent points is less than the set minimum turning distance, and generate the final path that satisfies the turning distance constraint.

[0117] Specifically, step 1042 dynamically verifies the distance between adjacent points to ensure that the final path meets the turning distance constraint and conforms to the ship's dynamics characteristics. The system simplifies the path sequence C=[P0,P1,...,P...]. n Based on this, traverse excluding the starting point. Each point outside (i=1 to n), calculate its relationship with the previous point. The Euclidean distance between them; if this distance is less than the set minimum turning distance, then... Remove from the path.

[0118] After the removal operation, the path sequence is automatically reassembled. For example, in the path [P0,P1,P2,P3], if ||P1-P0|| <d min If P1 is deleted, the new path is [P0, P2, P3]; the generated final path is the group path, which represents the trajectory of the center point when the group moves as a whole.

[0119] Step 1043: Based on the final path, calculate the path point coordinates of each platform using platform-related parameters to generate an independent path for each platform in the formation.

[0120] In this embodiment, the platform-related parameters include platform spacing coefficient, heading angle, and formation angle.

[0121] Specifically, step 1043 dynamically calculates the independent coordinates of each platform in the group based on each point on the grouping path. The system calculates the coordinates of each platform at each point P on the grouping path. i Based on the platform's detection capabilities and the spacing coefficient input by the user, the distance d between the platform and the group center point is calculated.

[0122] The coordinates of the platform at the current point are calculated based on the current formation heading θ and the angle α between the line connecting the platform and the center in the preset formation and the formation heading. The current formation heading θ is calculated by taking the current point P... i With the next point P i+1 The coordinate difference vector is obtained, representing the real-time forward direction of the formation at that point. The preset formation is the user-inputted or system-default platform relative position relationship.

[0123] Based on θ and α, the system calculates the coordinates of each platform at its current location: Where (cos(θ+α),sin(θ+α)) are unit vectors, ensuring that the platform position is offset by a distance d to the specified angular direction. This calculation process generates an independent coordinate sequence for each platform in the formation, ensuring that the formation maintains the preset formation during movement, avoiding mutual interference or overlap between platforms, and dynamically adjusting the spacing to adapt to different detection capabilities, so that the path point coordinates truly reflect the actual formation layout requirements.

[0124] Based on the above embodiments, in this embodiment, all independent paths are smoothed according to preset constraints to obtain the actual application path of each platform in the formation, thus completing the underwater formation path planning. Specifically, this includes steps 1051 to 1053:

[0125] Step 1051: Based on the preset minimum turning angle, adjust the coordinates of the path points in all independent paths so that the turning angles of all independent paths meet the minimum turning angle constraint.

[0126] Specifically, in step 1051, the system traverses each internal point P in all independent paths. i (i=1 to n-1), perform turning angle constraint checks and adjustments for each independent path. For each point Pi, calculate its relationship with the previous point P based on the chart coordinate system. i-1 The last point P i+1 If the resulting turning angle θ is less than the preset minimum turning angle α, minThen P i Along line segment P i-1 -P i+1 The midpoint M is moved so that the turning angle θ ≥ α. min The system repeatedly traverses all points on the independent paths, continuously checking and correcting points that do not meet the constraints, until the turning angles of all internal points meet the minimum turning angle constraint requirements.

[0127] Step 1052: Based on the speed range of each platform, calculate a unified time interval and adjust the coordinates of the path points corresponding to the platforms that do not meet the speed constraints, so that each platform can complete the movement from its current point to its next point within the unified time interval.

[0128] Specifically, step 1052 ensures that all platforms in the formation arrive at each point on the formation path synchronously during navigation by dynamically calculating a unified time interval and adjusting the coordinates of the platform path points, thereby maintaining formation alignment and detection continuity.

[0129] The system iterates through each point i in the grouping path except for the endpoint, and for each platform k (k=1 to m) in each independent path, calculates the current point P based on the nautical chart coordinate system. i,k To the next point P i+1,k Euclidean distance d k And based on the preset speed range (v) of platform k min,k to v max,k Determine the time constraints for each platform: t k,min =d k / v max,k (Shortest time at maximum speed), t k,max =d k / v min,k (Longest time at minimum speed).

[0130] The unified time t required for this segment of group navigation is set to t for all platforms. k,min Maximum value: This ensures that even the slowest platform can complete a segment within time t, providing a benchmark for synchronization.

[0131] For platform k, if t≤t k,max Then platform k at speed v k =d k / t navigation, ensuring arrival at the next point within time t;

[0132] If t>t k,max This indicates that the minimum speed of platform k is too low to complete the segment within time t. The system updates the coordinates of the next point of the platform to: based on the current point P. i,kStarting from the origin, the heading of platform k at the current point along the obtained path, with platform k's maximum speed v max,k The coordinates reached at a travel time t. This adjustment process is performed iteratively at the waypoints until the travel time of all platforms satisfies t ≤ t. k,max .

[0133] Step 1053: Based on the adjusted path point coordinates of each platform, generate the actual application path for each platform in the formation, and complete the underwater formation path planning.

[0134] Specifically, step 1053, based on the adjusted coordinates of each platform's path points in step 1052, directly generates a complete sequence of actual application paths for each platform in the formation. This sequence is formed by strictly arranging the coordinates of each platform's location on the formation path in chronological order, and the coordinates have passed the minimum turning angle constraint in step 1051 (ensuring the turning angle ≥ α). min The dynamic optimization is completed by combining the speed synchronization constraints of step 1052 (ensuring that all platforms arrive at each point synchronously).

[0135] In summary, this application integrates the detection performance calculation and path planning into a single end-to-end model, directly utilizes real-time marine environmental data to dynamically optimize path generation, and optimizes the performance of the path planning part, enabling it to generate paths that better meet the needs of practical applications more efficiently.

[0136] Compared to traditional solutions, this application calculates sensor detection performance based on actual environmental data in step 101 as the basis for path planning, making the path planned by this application more consistent with actual environmental factors. In steps 102, 103, and 104, this application innovatively combines multi-objective optimization technology, evolutionary learning technology, and the 2-opt algorithm based on the detection performance calculation results. This not only achieves intelligent generation of the optimal path but also simplifies the cumbersome steps of traditional evolutionary learning algorithms, making the calculation speed and the stability of the generated path much faster than other evolutionary learning or reinforcement learning algorithms.

[0137] In step 105, this application further smooths the path based on platform-related parameters input by the user, which can ensure that the path generated by this invention can be applied in practical applications, making the practical application value of this application far exceed that of similar algorithms.

[0138] This application demonstrates a significantly higher adaptability to irregular regions, particularly non-convex or multi-connected regions, compared to rule-based path planning algorithms, thus giving it greater practical application value.

[0139] See Figures 5-7In irregular polygonal regions, the path generated in this application is compared with rule-based path planning methods, such as integrated parallel search. The path graph results and cumulative detection probabilities are as follows: Figure 5 , Figure 6 As shown in the figure, the comparison results demonstrate that the present application has better detection performance and adaptability to irregular polygonal regions compared with rule-based methods.

[0140] Meanwhile, the path generation time of this application is compared with the time based on the evolutionary learning algorithm, for example... Figure 7 As shown, Figure 7 This demonstrates that the proposed algorithm has higher computational efficiency compared to other evolutionary learning-based algorithms.

[0141] Secondly, embodiments of this application provide an underwater formation path planning system, comprising: a first module, a second module, a third module, a fourth module, and a fifth module; the first module is used to rasterize the target area in the water search area based on environmental data of the water search area and generate a detection capability matrix; the second module is used to generate a set of minimum open coverage center points for the detection range of the formation as a whole relative to the target area based on the detection capability matrix, wherein the formation includes multiple platforms; the third module is used to generate an initial candidate path sequence population using a greedy algorithm with the minimum open coverage center point set as the traversal object, and to perform iterative optimization on the paths in the initial candidate path sequence population using a multi-objective optimization algorithm to generate a globally optimal path; the fourth module is used to decompose the globally optimal path into independent paths for each platform in the formation; the fifth module is used to smooth all independent paths based on platform-related parameters to obtain the actual application paths of each platform in the formation, thus completing the underwater formation path planning.

[0142] The functions of each module in the above-mentioned underwater formation path planning system correspond to the steps in the above-mentioned underwater formation path planning method embodiment, and their functions and implementation processes will not be described in detail here.

[0143] Thirdly, embodiments of this application provide an underwater grouping path planning device, which can be a personal computer (PC), laptop computer, server, or other device with data processing capabilities.

[0144] In this embodiment of the application, the underwater grouping path planning device may include a processor, a memory, a communication interface, and a communication bus.

[0145] The communication bus can be of any type and is used to interconnect the processor, memory, and communication interface.

[0146] The communication interface includes input / output (I / O) interfaces, physical interfaces, and logical interfaces used for interconnecting devices within the underwater swarming and path planning equipment, as well as interfaces used for interconnecting the underwater swarming and path planning equipment with other devices (such as other computing devices or user equipment). Physical interfaces can be Ethernet interfaces, fiber optic interfaces, ATM interfaces, etc.; user equipment can be displays, keyboards, etc.

[0147] Memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical storage, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0148] The processor can be a general-purpose processor, which can call the underwater grouping path planning program stored in the memory and execute the underwater grouping path planning method provided in the embodiments of this application. For example, the general-purpose processor can be a central processing unit (CPU). The method executed when the underwater grouping path planning program is called can be referred to in the various embodiments of the underwater grouping path planning method of this application, and will not be repeated here.

[0149] Fourthly, embodiments of this application also provide a computer-readable storage medium.

[0150] The present application has a computer-readable storage medium storing an underwater grouping path planning program, wherein when the underwater grouping path planning program is executed by a processor, it implements the steps of the underwater grouping path planning method as described above.

[0151] The method implemented when the underwater formation path planning program is executed can be referred to in the various embodiments of the underwater formation path planning method of this application, and will not be repeated here.

[0152] It should be noted that the sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0153] The terms "comprising" and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. The terms "first," "second," and "third," etc., are used to distinguish different objects, etc., and do not indicate a sequence, nor do they limit "first," "second," and "third" to different types.

[0154] In the description of the embodiments of this application, terms such as "exemplary," "for example," or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary," "for example," or "for instance" is intended to present the relevant concepts in a concrete manner.

[0155] In the description of the embodiments of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more.

[0156] In some processes described in the embodiments of this application, multiple operations or steps are included in a specific order. However, it should be understood that these operations or steps may not be executed in the order they appear in the embodiments of this application, or they may be executed in parallel. The sequence number of the operation is only used to distinguish different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.

[0157] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device to execute the methods described in the various embodiments of this application.

[0158] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. An underwater grouping path planning method, characterized in that, It includes: Based on the environmental data of the water search area, the target area in the water search area is rasterized and a detection capability matrix is ​​generated. Specifically, this includes: interpolating the environmental data of the water search area into a continuous underwater environment model; discretizing the target area in the water search area into multiple grid cells using a rasterization method; calculating the detection performance value of each grid cell through the continuous underwater environment model, and generating a detection capability matrix. Based on the detection capability matrix, a set of minimum open coverage center points of the target area is generated, with the group as a whole, and the group includes multiple platforms. Using the set of minimum open cover center points as the traversal object, an initial candidate path sequence population is generated through a greedy algorithm, and a multi-objective optimization algorithm is used to iteratively optimize the paths in the initial candidate path sequence population to generate the globally optimal path. The globally optimal path is decomposed into independent paths for each platform in the formation; Based on preset constraints, all independent paths are smoothed to obtain the actual application path for each platform in the formation, thus completing the underwater formation path planning. Specifically, this includes: adjusting the coordinates of path points in all independent paths based on preset minimum turning angles so that the turning angles of all independent paths meet the minimum turning angle constraints; calculating a unified time interval based on the speed range of each platform and adjusting the coordinates of path points corresponding to platforms that do not meet the speed constraints so that each platform completes its movement from its current point to its next point within the unified time interval; and generating the actual application path for each platform in the formation based on the adjusted path point coordinates of each platform, thus completing the underwater formation path planning.

2. The underwater grouping path planning method as described in claim 1, characterized in that, Based on the detection capability matrix, a set of minimum open coverage center points for the target area is generated, with each group as a whole, specifically including: Initialize all grid points in the rasterized target area to an uncovered state to create an initial target point set; Based on the detection capability matrix, the grid point with the highest detection efficiency value is selected from all grid points not covered in the target area as the center point, and this center point is added to the target point set; Calculate the detection range when the group passes through the center point as a whole, and mark all grid points within the intersection of the center point, the detection range, and the target area as covered; Repeat the above adding and marking process until all grid points in the target area are covered, thus completing the generation of the minimum open coverage center point set.

3. The underwater grouping path planning method as described in claim 1, characterized in that, Using the set of minimum open cover center points as the traversal object, an initial candidate path sequence population is generated through a greedy algorithm. Then, a multi-objective optimization algorithm is used to iteratively optimize the paths in the initial candidate path sequence population to generate the globally optimal path. Specifically, this includes: Using the set of minimum open cover center points as the traversal object, an initial candidate path sequence population containing greedy paths and random paths is generated based on a greedy algorithm; A multi-objective optimization loss function is constructed based on the total path length and path smoothness of each path in the initial candidate path sequence population. A multi-objective optimization algorithm is used to iteratively optimize the initial candidate path sequence population based on the objective function, generating multiple dominant paths; Choose the path that is weakly dominant over all other paths from among all dominant paths, and select it as the globally optimal path.

4. The underwater grouping path planning method as described in claim 3, characterized in that, The multi-objective optimization loss function is: F(C)=(L(C), S(C)); Where L(C) is the total path length of each path, and S(C) is the path smoothness.

5. The underwater formation path planning method as described in claim 3, characterized in that, A multi-objective optimization algorithm is used to iteratively optimize the initial candidate path sequence population based on the objective function, generating multiple dominant paths, specifically including: Perform mutation operations on several paths in the initial candidate path sequence population to generate several new paths; The multi-objective optimization algorithm is used to compare each new path with all paths in the current initial candidate path sequence population one by one based on the multi-objective optimization loss function, and iteratively optimizes to find the dominant path until the initial candidate path sequence population converges.

6. The underwater grouping path planning method as described in claim 1, characterized in that, The globally optimal path is decomposed into independent paths for each platform in the formation, specifically including: Remove redundant intermediate points with unchanged direction from the globally optimal path to generate a simplified path; Remove points in the simplified path where the distance between adjacent points is less than the set shortest turning distance, and generate the final path that satisfies the turning distance constraint; Based on the final path, the coordinates of the path points of each platform are calculated using platform-related parameters, generating an independent path for each platform in the formation.

7. The underwater grouping path planning method as described in claim 6, characterized in that: The platform-related parameters include the platform spacing coefficient, heading angle, and formation angle.

8. An underwater grouping path planning system, characterized in that, It includes: The first module is used to rasterize the target area in the water search area and generate a detection capability matrix based on the environmental data of the water search area. The second module is used to generate, based on the detection capability matrix, a set of minimum open coverage center points of the target area with respect to the detection range of a group as a whole, wherein the group includes multiple platforms. The third module is used to generate an initial candidate path sequence population using a greedy algorithm with the minimum open cover center point set as the traversal object, and to perform iterative optimization on the paths in the initial candidate path sequence population using a multi-objective optimization algorithm to generate the globally optimal path. The fourth module is used to decompose the globally optimal path into independent paths for each platform in the formation; The fifth module is used to smooth all independent paths based on platform-related parameters, obtain the actual application paths of each platform in the formation, and complete the underwater formation path planning. The first module is also used to interpolate the environmental data of the water search area into a continuous underwater environment model; to discretize the target area in the water search area into multiple grid cells using a rasterization method; to calculate the detection performance value of each grid cell using the continuous underwater environment model, and to generate a detection capability matrix; The fifth module is also used to adjust the coordinates of path points in all independent paths based on a preset minimum turning angle, so that the turning angles of all independent paths meet the minimum turning angle constraint; calculate a unified time interval based on the speed range of each platform, and adjust the coordinates of path points corresponding to platforms that do not meet the speed constraint, so that each platform completes the movement from its current point to its next point within the unified time interval; and generate the actual application path for each platform in the formation based on the adjusted path point coordinates of each platform, thus completing the underwater formation path planning.

Images