A method for task allocation for obstacle avoidance of an unmanned underwater vehicle

Abstract

The present application relates to the technical field of offshore obstacle laying, and particularly relates to a task allocation method for unmanned underwater vehicle obstacle laying, comprising: acquiring chart data of a target sea area; performing environment preprocessing based on the chart data, identifying an obstacle area and defining a safety zone, and generating an anchor point set in the safety zone by using a distance-driven sampling method; calculating obstacle-avoiding shortest path distances between each anchor point and each laying point based on anchor point positions and laying point positions; clustering the laying points by using an anchor point-aware capacity-constrained clustering method, constructing a target function, and solving the target function under unmanned underwater vehicle load constraints to obtain cluster division results and associated anchor points corresponding to each cluster; and calculating obstacle-avoiding shortest paths based on the cluster division results and the associated anchor points, and planning execution paths for each unmanned underwater vehicle. The present application achieves reasonable clustering of laying points and association of anchor points, and improves the overall efficiency of unmanned underwater vehicle laying tasks.

Description

Technical Field

[0001] This invention relates to the field of marine obstacle deployment technology, and in particular to a task allocation method for deploying obstacles using an unmanned underwater vehicle. Background Technology

[0002] Unmanned underwater vehicles (UUVs) play a crucial role in obstacle deployment missions at sea. This task requires planning a safe and efficient deployment route for the carrier platform and UUVs in a discrete marine environment to drop all obstacles within a designated area. In practice, the target area may need to cover dozens of deployment points. Since each UUV carries a limited number of deployment points, deploying them point-by-point would be inefficient. Therefore, it is necessary to divide the deployment points into clusters of no more than four points, with each cluster being independently deployed by a single UUV. This approach fully utilizes the UUV's payload while reducing the complexity of carrier platform scheduling and recovery.

[0003] When performing missions, unmanned underwater vehicles (UUVs) must depart from the safe zone where their carrier platform is located, traverse complex seabed terrain to reach the target point, and then safely return. Nautical charts contain inaccessible areas such as islands, reefs, shoals, and restricted deployment zones. Using straight-line distances directly would lead to infeasible or even dangerous solutions. Therefore, obstacle avoidance path search algorithms must be introduced to ensure that the UUV's trajectory is reachable under realistic nautical chart conditions and strictly avoids obstacles, thus obtaining a safe and near-optimal path. Existing technologies primarily focus on the spatial compactness between deployment points, neglecting the crucial constraint that the UUV must depart from and return from the safe zone. This results in clustering results that may minimize distances within clusters, but the overall location of the clusters is far from the safe zone, significantly increasing the UUV's round-trip distance and reducing mission efficiency.

[0004] Chinese patent application CN114897215A discloses a method for allocating multi-UAV reconnaissance tasks based on unsupervised learning discrete pigeon flocking optimization. This method flexibly groups reconnaissance targets using cosine similarity clustering, employs K-Means to group targets into designated clusters based on cosine similarity, and then uses the discrete pigeon flocking algorithm to optimize the fitness function to achieve multi-UAV reconnaissance task allocation. Cluster analysis reduces track overlap. This method is applied to aerial UAV scenarios. UAVs have high maneuverability in three-dimensional space, while unmanned underwater vehicles (UUVs) are subject to multiple constraints in the underwater environment, such as seabed topography, water depth, and ocean currents, resulting in significantly lower maneuverability compared to aerial UAVs. Furthermore, UUVs require consideration of the launch and recovery location of the carrier platform, leading to higher actual execution costs for task allocation. Summary of the Invention

[0005] In view of this, the present invention provides a task allocation method for unmanned underwater vehicles to deploy obstacles, so as to solve the technical problems of excessively long task execution range and low task allocation efficiency in the prior art when allocating task allocation for unmanned underwater vehicles to deploy obstacles.

[0006] The technical solution of this invention is implemented as follows: This invention provides a task allocation method for unmanned underwater vehicles to deploy obstacles, including: S1. Obtain nautical chart data of the target sea area, location information of the obstacles to be deployed, and the number and payload limit of the unmanned underwater vehicles; S2. Based on the nautical chart data, perform environmental preprocessing, identify obstacle areas and define safe zones, and generate a set of anchor points within the safe zones using a distance-driven sampling method. The anchor points are candidate release and recovery locations for the unmanned underwater vehicle. S3. Based on the anchor point location and the deployment point location, calculate the shortest obstacle avoidance path distance between each anchor point and each deployment point; S4. The deployment points are clustered using an anchor-aware capacity-constrained clustering method. The cluster compactness and accessibility from the cluster to the safe zone are optimized simultaneously in the objective function. The weight coefficients of the two optimization objectives are adaptively determined according to the spatial distribution characteristics of the deployment points. The cluster division results and the associated anchor points of each cluster are obtained under the load constraints of the unmanned underwater vehicle. S5. Based on the cluster division results and associated anchor points, a path search algorithm is used to calculate the shortest obstacle avoidance paths between each cluster and the safe zone, as well as between each cluster, and to assign task clusters and plan execution paths for each unmanned underwater vehicle.

[0007] Based on the above technical solutions, preferably, the environmental preprocessing in step S2 includes: constructing water depth partitions according to the nautical chart matrix, dividing the sea area into five categories: land, shallow water, medium water, deep water and deep sea, and determining the navigable area; Construct an obstacle area mask to mark impassable areas, including islands, shoals, land, and no-navigation zones, using a binary mask representation; define a safe zone polygon as the range for the carrier platform to release the unmanned underwater vehicle, defined by a set of coordinate points; construct a deployment area mask based on water depth categories to exclude unsuitable waters for deployment.

[0008] Based on the above technical solutions, preferably, the distance-driven sampling method in step S2 includes: calculating the centroid of all deployment points; selecting the point closest to the centroid from the boundary of the safe zone as the first anchor point; sequentially selecting the boundary point farthest from the existing anchor point as the next anchor point, and adopting the farthest point sampling strategy; repeating the above steps until K anchor points are generated, wherein the number of anchor points K is determined by dividing the total number of deployment points by the upper limit of the load.

[0009] Based on the above technical solutions, preferably, step S3 specifically includes: S31. Obtain the target deployment point set according to the task requirements, filter out points that do not meet the water depth and area conditions, and obtain the set of valid deployment points located within the allowed deployment area. Each deployment point corresponds to one obstacle to be deployed. S32. Traverse the set of anchor points and the set of effective deployment points, use a path search algorithm to calculate the shortest obstacle avoidance path distance between each anchor point and each deployment point, and store the distance data in the anchor point distance matrix.

[0010] Based on the above technical solutions, preferably, step S4 specifically includes: S41. Calculate the spatial distribution characteristic parameters of the deployment point set based on the anchor point distance matrix and the obstacle avoidance shortest path distance matrix between deployment points; S42. Based on the spatial distribution characteristic parameters, an adaptive weight calculation strategy is used to determine the weight coefficients for intra-cluster compactness and cluster-to-safe-zone reachability. S43. Establish an anchor-aware capacity-constrained clustering model, introduce decision variables, construct a dual-objective weighted optimization objective function, and set constraints. S44. For each cluster and its associated anchor point, the actual distance from the anchor point to the cluster is calculated using the weighted quantile distance method based on cluster morphology. S45. Solve the clustering model using a mixed-integer linear programming solver, and output the cluster partitioning results, the center point of each cluster, and the associated anchor point of each cluster.

[0011] Based on the above technical solutions, preferably, the adaptive weight calculation strategy in step S42 includes: calculating the average distance from all deployment points to the nearest anchor point to characterize the overall distance relationship between the deployment point set and the safe zone; calculating the average distance between deployment points to characterize the internal dispersion of the deployment point set; calculating the ratio of the two average distances as the distance ratio; and using the Sigmoid function to map the distance ratio to a weight coefficient for the reachability of the cluster to the safe zone, wherein the weight coefficient for the compactness within the cluster is 1 minus the reachability weight coefficient.

[0012] Based on the above technical solution, preferably, the decision variables introduced in step S43 include: a central decision variable regarding whether the representation point is selected as the cluster center, an allocation decision variable regarding whether the representation point is assigned to a certain cluster, and an anchor point association decision variable regarding whether the representation cluster is associated with the anchor point; the objective function is: ; Where P is the set of valid deployment points, and i and j are the indices of the deployment points. Let be the shortest obstacle avoidance path distance between points i and j. To assign decision variables, This is the intra-cluster compactness weighting coefficient. For the k-th anchor point, For the i-th deployment point, This represents the shortest obstacle avoidance path distance from the anchor point to the deployment point. As anchor points for decision variables, Let i be the central decision variable, representing whether point i is selected as the cluster center. is the accessibility weighting coefficient from the cluster to the safe zone, and K is the number of unmanned underwater vehicles.

[0013] Based on the above technical solutions, preferably, the constraints in step S43 include: (1) each point must be assigned to a cluster; (2) each point can only be assigned to the selected center point; (3) the capacity of each cluster does not exceed the upper limit of the unmanned underwater vehicle's payload; (4) the number of clusters does not exceed the number of unmanned underwater vehicles; (5) each cluster must be associated with an anchor point, and if a certain deployment point is the center of a cluster, then there must be one and only one anchor point associated with it.

[0014] Based on the above technical solutions, preferably, the weighted quantile distance method based on cluster morphology in step S44 includes: sorting the distances from all points within the cluster to the anchor point, obtaining the nearest point distance, median distance, and farthest point distance; assigning different weights to the three quantile distances using a decreasing weight function; calculating the normalized spatial dispersion of the cluster; and multiplying the weighted quantile distance by the spatial dispersion correction factor to obtain the actual distance from the anchor point to the cluster, calculated using the following formula: ; Where Q is the set of quantiles, including the quantile of 0.0 corresponding to the nearest point, the quantile of 0.5 corresponding to the median point, and the quantile of 1.0 corresponding to the farthest point. The weight of the q-th quantile. anchor point to cluster The qth quantile distance This is the dispersion penalty coefficient. For clusters The normalized spatial discreteness.

[0015] Based on the above technical solutions, preferably, step S5 specifically includes: S51. Construct a passable mask based on a two-dimensional discrete nautical chart to identify passable and impassable areas, define an adjacency set in eight directions including diagonal and direct neighbors, and set edge weights for direct neighbor movement and diagonal movement. S52. Obtain the center point of each cluster as the cluster representative point, and use the path search algorithm to calculate the shortest obstacle avoidance path and distance between the anchor point of the safe zone and the cluster representative point, between the cluster representative points and the anchor points within the safe zone, respectively, to obtain the distance matrix and path polyline. S53. Based on the cluster division results, associated anchor points, distance matrix and path polyline, assign a task cluster to each unmanned underwater vehicle and plan the execution path from the associated anchor point, through each deployment point in the cluster and back to the associated anchor point.

[0016] The present invention has the following advantages over the prior art: (1) The task allocation method provided by this invention, through the anchor point perception mechanism and adaptive weight optimization strategy, considers both intra-cluster compactness and accessibility from cluster to safe zone during the clustering stage, so that the task allocation result conforms to the actual execution scenario of unmanned underwater vehicles. This method comprehensively considers load constraints, obstacle avoidance path search and spatial distribution characteristics, realizes reasonable clustering of deployment points and anchor point association, improves the overall efficiency of unmanned underwater vehicle deployment tasks, and reduces the complexity of carrier platform scheduling and recovery.

[0017] (2) This invention uses a distance-driven sampling method to generate a set of anchor points at the boundary of the safe zone. By calculating the centroid of the deployment points and combining it with the farthest point sampling strategy, the anchor points are evenly distributed at the boundary of the safe zone and cover the main operation direction. Compared with the calculation of the complete point-to-point distance matrix, this method reduces the computational complexity, significantly reduces the computational amount of path search, and ensures that the anchor point positions can effectively cover the deployment task requirements.

[0018] (3) This invention uses a weighted quantile distance method combined with a spatial dispersion correction factor to calculate the actual distance from the anchor point to the cluster, comprehensively considering the distances to the nearest, median, and farthest points of the cluster, as well as the spatial distribution morphology of the cluster. This method can distinguish between compact clusters and elongated strip clusters, accurately depicting the actual cost for an unmanned underwater vehicle to enter the cluster from the anchor point and visit each deployment point, avoiding the shortcomings of a single distance metric that ignores cluster depth or is affected by outliers. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of the task allocation method for the unmanned underwater vehicle to deploy obstacles according to the present invention. Figure 2This is a flowchart illustrating the generation of anchor points on the boundary of a safe zone according to the present invention. Figure 3 This is a flowchart of the clustering solution process of the present invention; Figure 4 This is a flowchart of the route planning process of the present invention. Detailed Implementation

[0021] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the 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.

[0022] like Figure 1 As shown, the present invention provides a task allocation method for unmanned underwater vehicles to deploy obstacles, including: S1. Obtain nautical chart data of the target sea area, location information of the obstacles to be deployed, and the number and payload limit of the unmanned underwater vehicles; S2. Based on the nautical chart data, perform environmental preprocessing, identify obstacle areas and define safe zones, and generate a set of anchor points within the safe zones using a distance-driven sampling method. The anchor points are candidate release and recovery locations for the unmanned underwater vehicle. S3. Based on the anchor point location and the deployment point location, calculate the shortest obstacle avoidance path distance between each anchor point and each deployment point; S4. The deployment points are clustered using an anchor-aware capacity-constrained clustering method. The cluster compactness and accessibility from the cluster to the safe zone are optimized simultaneously in the objective function. The weight coefficients of the two optimization objectives are adaptively determined according to the spatial distribution characteristics of the deployment points. The cluster division results and the associated anchor points of each cluster are obtained under the load constraints of the unmanned underwater vehicle. S5. Based on the cluster division results and associated anchor points, a path search algorithm is used to calculate the shortest obstacle avoidance paths between each cluster and the safe zone, as well as between each cluster, and to assign task clusters and plan execution paths for each unmanned underwater vehicle.

[0023] In one embodiment of the present invention, step S1 includes: acquiring a nautical chart matrix of the target sea area, wherein the nautical chart matrix is ​​a two-dimensional discretized sea area map containing water depth zones and obstacle information; acquiring the location information of the obstacle deployment points to be deployed, wherein each deployment point corresponds to one obstacle to be deployed; and acquiring the number K and payload limit of the unmanned underwater vehicles. The payload limit is the maximum number of obstacles that each unmanned underwater vehicle can carry. The nautical chart matrix is ​​a rasterized representation, with each grid point corresponding to a rectangular area of ​​the real sea area. The grid resolution is typically 50 meters to 200 meters, determined based on mission accuracy requirements and computing resources.

[0024] The nautical chart matrix divides the sea area into five categories based on water depth partitions: land, shallow water, mid-water, deep water, and deep sea. Land is marked as non-navigable, shallow water (0-200 meters) as navigable, mid-water (200-500 meters) as navigable, deep water (500-1000 meters) as navigable, and deep sea (greater than 1000 meters) as navigable. This classification standard references the water depth classification conventions in actual navigational regulations while also considering the typical operating depth range of unmanned underwater vehicles (UUVs). The chart data also includes information on obstacle areas, identifying impassable areas such as islands, shoals, land, and no-navigation zones. These areas are represented in the chart matrix by specific markers or independent mask layers. The chart data further includes the coordinates of safety zone polygons, which define the range within which the platform can maneuver and release UUVs. Safety zones are typically located far from shore, in moderately deep water, and free of obstacles, allowing the platform free maneuverability to select suitable locations for releasing and recovering UUVs. The size and shape of the safe zone are determined based on mission requirements and sea conditions. It is usually a polygonal area whose boundary is defined by a set of ordered coordinate points.

[0025] In one embodiment of the present invention, such as Figure 2 As shown, step S2 includes two stages: environmental preprocessing and anchor point generation.

[0026] The environmental preprocessing stage includes: constructing water depth zones based on the nautical chart matrix, dividing the sea area into five categories: land, shallow water, mid-water, deep water, and deep sea, and determining navigable areas; and constructing obstacle area masks. The chart identifies impassable areas such as islands, shoals, land, and no-navigation zones, using a binary mask representation, where W and H represent the width and height of the chart, respectively. Representing grid points Passage is permitted. This indicates that the point is an obstacle or a no-navigation zone. This mask is used for subsequent obstacle avoidance constraints on the unmanned underwater vehicle's trajectory. During path search, all paths that pass through grid points marked as 0 on the obstacle mask are considered infeasible.

[0027] Define the safe zone polygon as the area within which the carrier platform can maneuver and release unmanned underwater vehicles, defined by a set of coordinate points. Define, where This represents the number of vertices in the safe zone polygon. This area is unobstructed and has suitable water depth, making it suitable as a launch and return base for unmanned underwater vehicles. The platform remains within the safe zone, and release and retrieval points can be selected within this area, defining the available areas for these points. A deployment area mask is constructed based on water depth categories. The mask excludes areas that are too deep or too shallow, as well as prohibited deployment zones such as ultra-deep water or restricted areas. The mask's construction rules are as follows: if a grid point's water depth category is land or deep sea, it is marked as 0 to indicate that deployment is not allowed; if the water depth category is shallow, medium, or deep water, it is marked as 1 to indicate that deployment is allowed. This ensures the legality of deployment points and that obstacles deployed by the unmanned underwater vehicle fall within the effective area. Obstacles may not be able to be deployed properly in excessively shallow water, and may exceed the obstacle's working depth range in excessively deep water.

[0028] In the anchor point generation stage, a distance-driven sampling method is used to generate K anchor points on the safety zone boundary as candidate release and recovery locations for the unmanned underwater vehicle. The distance-driven sampling method, based on a farthest-point sampling strategy, ensures that the anchor points are evenly distributed on the safety zone boundary, covering different directions, thus providing diverse anchor point selections for subsequent clustering. Specific steps include: calculating the centroid of all deployment points. ,in , n is the total number of deployment points. Let C be the coordinates of the i-th deployment point. The centroid represents the geometric center of the set of deployment points, reflecting the spatial center of gravity of the deployment task. The point closest to the centroid C on the boundary of the safety zone is selected as the first anchor point. ,Right now ,in Represents the set of boundary points of the safe zone. Let p represent the Euclidean distance, and p be a candidate point on the safe zone boundary. Selecting the boundary point closest to the centroid as the first anchor point ensures the first anchor point is as close as possible to the deployment point set, reducing the initial range. The next anchor point is then selected sequentially from the boundary points furthest from existing anchor points, employing a furthest-point sampling strategy. Specifically, the... Anchor points The selection rules are That is, selecting from an existing set of anchor points. The boundary point with the largest nearest anchor point is selected. This strategy disperses anchor points as much as possible along the boundary, preventing them from clustering in a local area and thus improving the coverage of anchor points in different orientations. Repeat the above steps until K anchor points are generated.

[0029] The method for determining the number of anchor points K is as follows: Where n is the total number of deployment points. This represents the upper limit for the payload of the unmanned underwater vehicle. The number of anchor points should at least cover the expected number of clusters, i.e. However, it should not be less than the minimum number of anchor points. This ensures the clustering algorithm has sufficient flexibility. Output anchor point set. Each anchor point Coordinates This indicates the specific location on the boundary of the safe zone.

[0030] In one embodiment of the present invention, step S3 includes: S31. Obtain the target deployment point set according to the task requirements, filter out points that do not meet the water depth and area conditions, and ensure that all candidate points are within the allowable deployment area to obtain the effective deployment point set. ,in Let represent the i-th valid deployment point, where n is the total number of valid deployment points, and each point corresponds to one obstacle to be deployed. Specifically, for each valid deployment point... ,examine If the value is 1, then remove the point from the set of valid placement points.

[0031] S32. Traverse the set of anchor points and the set of valid placement points, and calculate the value of each anchor point using a path search algorithm. To each deployment point Shortest path distance for obstacle avoidance between Store distance data in the anchor point distance matrix. Among them To determine the effective number of deployment points, .

[0032] The path search algorithm uses A Algorithm, A The algorithm is a heuristic search algorithm that introduces a heuristic function based on Dijkstra's algorithm. Used to estimate the current node The remaining distance to the target node t is used to guide the search process towards the target node more quickly. In this embodiment, an eight-way heuristic function is employed. ,in For the node The formula for estimating the eight-way distance to the target point t is as follows: ; and They are nodes The difference between the coordinates of the target point t in the x and y directions. Let be the coordinates of node b. The coordinates of the target point t represents the weight coefficients of the heuristic function, with values ​​ranging from [1.0, 1.2]. When At that time, the heuristic function is consistent, A The algorithm guarantees finding the optimal path; when When the value is slightly greater than 1.0, the heuristic function becomes inconsistent, resulting in a faster search speed but potentially sacrificing a small amount of optimality. This heuristic function guarantees optimality and significantly accelerates the search process compared to Dijkstra's algorithm. The algorithm, guided by a heuristic function, reduces unnecessary node expansion. The computational complexity is O(n log n). Next A The search, with a time complexity of O(n) for each search on the sparse graph. Where V is the set of grid points after discretization of the nautical chart. This represents the number of elements in set V, i.e., the total number of nautical chart grid points. Compared to the complete Distance matrix needs The computational complexity of the anchor point distance matrix is ​​significantly reduced in this second calculation, which enables the algorithm to process a large set of deployment points within a reasonable time.

[0033] In one embodiment of the present invention, such as Figure 3 and 4 As shown, step S4 employs an anchor-aware capacity-constrained clustering method. Based on traditional capacity-constrained clustering, it introduces an anchor-association mechanism and adaptive weight optimization, ensuring that the clustering results consider not only intra-cluster compactness but also the reachability from clusters to safe zones. Specifically, this includes: S41. Calculate the spatial distribution characteristic parameters of the deployment point set based on the anchor point distance matrix and the shortest obstacle avoidance path distance matrix between deployment points.

[0034] Spatial distribution characteristic parameters are used for subsequent adaptive weight calculation, reflecting the spatial distribution characteristics of the deployment point set relative to the safety zone. The average distance from all deployment points to the nearest anchor point is calculated. This parameter characterizes the overall distance relationship between the set of deployment points and the safe zone. The calculation formula is: ; in From the anchor point distance matrix This is obtained from the data. For each deployment point, find the nearest anchor point, and then average the distances to the nearest anchor points for all deployment points to obtain the result. .like A larger value indicates that the overall deployment point is far from the safe zone, requiring the unmanned underwater vehicle to travel a longer distance to reach the deployment area; if A smaller distance indicates that the deployment point cluster is close to the safe zone, and the unmanned underwater vehicle's range is relatively short. Calculate the average distance between deployment points. This parameter characterizes the degree of internal dispersion of the deployment point set. The calculation formula is: ; in Deployment point and The shortest path distance for obstacle avoidance between them, via A The algorithm calculates the distance between all pairs of points. Since the computational cost of calculating the distance between all pairs of points is... The computational cost is high when the number of deployment points is large. To reduce the computational load, random sampling estimation is used, i.e. ,in For m pairs of points sampled randomly, it is recommended Random sampling methods can significantly reduce computational costs while maintaining estimation accuracy. If... A larger value indicates that the deployment points are scattered, and more attention needs to be paid to the compactness within clusters during clustering; if A smaller value indicates that the deployment points are densely distributed, and more consideration can be given to the accessibility of the cluster to the safe zone when clustering.

[0035] S42. Based on spatial distribution characteristic parameters, an adaptive weighting calculation strategy is used to determine the weight coefficients for intra-cluster compactness and cluster-to-safe-zone reachability. The ratio of the two average distances is calculated as the distance ratio: ; in To smooth the term and avoid division by zero errors, the value is... The kilometer value, set empirically based on typical sea area scales, is equivalent to 100 meters. In most application scenarios, this effectively avoids numerical instability caused by an excessively small denominator. The distance ratio *r* reflects the relative relationship between the distance from the deployment point to the safe zone and the distance between deployment points. It is a dimensionless parameter that can be compared across different mission scenarios. The Sigmoid function is used to map the distance ratio to a weighted coefficient for the reachability of the cluster to the safe zone. ; Weighting coefficients for intra-cluster compactness .in, The steepness coefficient controls the rate of change of the Sigmoid function. The value is determined based on experimental debugging. As a balance threshold, when the distance ratio At that time, weight balance Recommended value This value is determined based on statistical analysis of numerous task scenarios. When the average distance from the deployment point to the safe zone is approximately 1.5 times the average distance between deployment points, both optimization objectives should be given equal weight. The Sigmoid function implements a smooth, non-linear mapping from the distance ratio to the weights, avoiding abrupt changes in weights. When the distance ratio... For example The output of the Sigmoid function Approaching 1, at this point When the distance ratio is close to 0, cluster optimization prioritizes the accessibility of clusters to the safe zone. Since deployment points are close together but far from the safe zone, minimizing the unmanned underwater vehicle's (UUV) round-trip distance to the safe zone should be a priority. For example The output of the Sigmoid function Between 0.3 and 0.7, the weights of the two optimization objectives are relatively balanced, and the clustering results comprehensively consider both intra-cluster compactness and accessibility from clusters to safe zones. When the distance ratio... For example The output of the Sigmoid function Approaching 0, at this point When the value is close to 1, cluster optimization places more emphasis on intra-cluster compactness because the deployment points are close to the safe zone but scattered from each other, so reducing intra-cluster travel should be given priority.

[0036] S43. Establish an anchor-aware capacity-constrained clustering model. Introduce decision variables including: If point i is chosen as the center of a cluster, then ,otherwise , is the central decision variable, used to characterize which deployment points are selected as the center point of the cluster. If point j is assigned to the cluster to which center i belongs, then ,otherwise , which is the decision variable for allocation, used to characterize which cluster each deployment point is assigned to. If the cluster association anchor point is centered at point i ,but ,otherwise , where is the anchor point association decision variable, used to characterize which anchor point each cluster is associated with as the release and reclamation location. The objective function for the bi-objective weighted optimization is constructed as follows: ; in, The set of valid deployment points, where i and j are the indices of the deployment points. Let be the shortest obstacle avoidance path distance between points i and j. For the assignment decision variable, it represents whether point j is assigned to the cluster to which center i belongs; This is the intra-cluster compactness weighting coefficient. For the k-th anchor point, For the i-th deployment point, The shortest obstacle avoidance path distance from the anchor point to the deployment point is calculated from the anchor point distance matrix in step S3. Obtain from, For anchor point association decision variables, it indicates whether a cluster centered at point i is associated with the anchor point. , Let i be the central decision variable, representing whether point i is selected as the cluster center. Let K be the reachability weight coefficient from the cluster to the safe zone, and K be the number of unmanned underwater vehicles. The first term of the objective function... This represents the sum of internal distances across all clusters, measuring intra-cluster compactness. It is used when a placement point j is assigned to the cluster containing center i (i.e., ...). This point contributes distance to the objective function. Summing over all assignment relationships yields the total intra-cluster distance. A smaller intra-cluster distance indicates a more compact distribution of points within the cluster, resulting in a shorter range for the unmanned underwater vehicle (UUV) when deploying within the cluster. The second term of the objective function... This represents the sum of distances from all cluster centers to their associated anchor points, measuring the reachability of a cluster to a safe zone. When a deployment point i is selected as a cluster center (i.e., ...), ... And the cluster is associated with anchor points (Right now When ), the cluster contributes distance to the objective function. Summing over all clusters yields the total distance from each cluster to the safe zone. The smaller the distance from a cluster to the safe zone, the shorter the journey for the unmanned underwater vehicle (UUV) from the safe zone to the cluster, and the higher the mission efficiency.

[0037] The constraints set include: (1) Each point must be assigned to a cluster, and the constraint is as follows: This constraint ensures that each deployment point is assigned to a specific cluster, and only to one cluster, avoiding omissions or duplicate assignments. (2) Each point can only be assigned to the selected center point, with the constraint condition being... This constraint ensures that only the selected center point can accept assignments from other points; if point i is not selected as the center (i.e., ...), the other points will be assigned to it. If ), then no point can be assigned to point i (i.e., all points cannot be assigned to point i). (3) The capacity of each cluster shall not exceed the upper limit of the payload of the unmanned underwater vehicle, and the constraint is: ,in This represents the maximum payload capacity of each unmanned underwater vehicle, i.e., the maximum number of obstacles it can carry. This constraint ensures that the number of deployment points within each cluster does not exceed the maximum payload capacity, if point i is selected as the center (i.e., ...). If the number of points assigned to that cluster does not exceed [a certain threshold], then the number of points assigned to that cluster will not exceed [ If point i is not selected as the center (i.e. If the number of clusters does not exceed the number of unmanned underwater vehicles, then the constraint is automatically satisfied. (4) The number of clusters does not exceed the number of unmanned underwater vehicles, and the constraint is: This constraint ensures that the number of generated clusters does not exceed the number of available unmanned underwater vehicles (UUVs), avoiding a situation where there are not enough UUVs to perform all cluster tasks. (5) Each cluster must be associated with an anchor point. If a deployment point is the cluster center, then there must be one and only one anchor point associated with it. The constraint is as follows: This constraint ensures that each cluster has one and only one associated anchor point if point i is selected as the cluster center (i.e., ...). If point i is not selected as the cluster center (i.e., ...), then one of the K anchor points must be chosen as the associated anchor point of the cluster; if point i is not selected as the cluster center (i.e., ...), then ... If it is an anchor point, then it cannot be associated with any anchor point.

[0038] S44. For each cluster and its associated anchor points, calculate the actual distance from the anchor points to the cluster using a weighted quantile distance method based on cluster morphology. Sort the distances from all points within the cluster to the anchor points and obtain the nearest point distance, median point distance, and farthest point distance. Specifically, for clusters... and its associated anchor points Calculate the deployment point within the cluster. Distance to anchor point Then sort these distances in ascending order to obtain ,in For clusters Number of deployment points within the area. Distance to the nearest point is... This represents the distance from the nearest deployment point within the cluster to the anchor point. The median distance is... , representing the median distance within the cluster to the anchor point, where This indicates rounding up. The distance to the farthest point is... , representing the distance of the placement point farthest from the anchor point within the cluster. A decreasing weighting function is used to assign different weights to the three quantile distances; the weight calculation formula is as follows: ,in For the set of quantiles, The decay exponent controls the rate of weight decrease. The values ​​were determined through experimental debugging, aiming to maintain a reasonable distribution of weights among the nearest, median, and farthest points, reflecting the dominance of the nearest point while not completely ignoring the influence of the median and farthest points. This decreasing weight distribution aligns with the spatial access characteristics of an unmanned underwater vehicle (UUV) entering a cluster from an anchor point. The UUV first reaches the nearest point within the cluster, then visits other points sequentially, and finally reaches the farthest point. Therefore, the distance from the nearest point has the greatest impact on the total range, while the distance from the farthest point has the least impact. The normalized spatial dispersion of the cluster is then calculated. ,in Anchor point to cluster The average distance between all points within the area. The average distance between global deployment points is used for normalization. Normalized spatial dispersion. To measure the spatial distribution of a cluster, if the distances from each point within the cluster to the anchor point are relatively small, then... A smaller size indicates a compact cluster; if the distances from each point within the cluster to the anchor point vary significantly, then... A relatively large size indicates that the clusters are elongated or dispersed. Through... Normalize to make As a dimensionless parameter, it can be compared across different task scenarios. Multiplying the weighted quantile distance by the spatial dispersion correction factor yields the actual distance from the anchor point to the cluster, calculated using the following formula: ; Where Q is the set of quantiles, including the quantile of 0.0 corresponding to the nearest point, the quantile of 0.5 corresponding to the median point, and the quantile of 1.0 corresponding to the farthest point. The weight of the q-th quantile. anchor point to cluster The qth quantile distance This is the dispersion penalty coefficient. The value is determined based on experimental debugging, aiming to keep the spatial dispersion effect on distance moderate, so as to distinguish between compact clusters and dispersed clusters without causing the distance of dispersed clusters to increase excessively. For clusters The normalized spatial dispersion. Spatial dispersion correction factor. Used to distinguish between compact, clump-like clusters and slender, strip-like clusters. If the cluster is compact, When the cluster is small, the correction factor is close to 1, and the weighted quantile distance remains essentially unchanged; if the cluster is elongated, The value is relatively large, with a correction factor greater than 1, amplifying the weighted quantile distance and reflecting the additional cost for unmanned underwater vehicles to penetrate deep into slender clusters. This formula combines the weighted quantile mechanism and spatial dispersion correction to characterize the actual cost of entering a cluster from an anchor point, including not only the distance to the cluster but also the additional access cost caused by the spatial distribution of the cluster.

[0039] S45. Use a mixed-integer linear programming solver to solve the clustering model, outputting the cluster partitioning results, the centroid of each cluster, and the associated anchor point of each cluster. First, construct the distance matrix. ,in , indicating the deployment point and The shortest path distance for obstacle avoidance and the anchor point distance matrix between them. ,in , indicating anchor point to the deployment point The shortest path distance for obstacle avoidance. This distance is obtained through A. Alternatively, the Dijkstra algorithm can be used to calculate the deployment points on a discrete nautical chart, strictly avoiding obstacles and no-deployment zones. Then, based on the spatial distribution characteristics of the deployment points, the calculation is performed in step S41. and The adaptive weight formula in step S42 is used to calculate... and Next, an integer linear programming model is established, including decision variables. The objective function is a bi-objective weighted optimization function, with constraints including point assignment constraints, capacity constraints, cluster number constraints, and anchor point association constraints. Finally, a mixed-integer linear programming solver is invoked for optimization. Commonly used solvers include Gurobi, CPLEX, and SCIP. These solvers employ methods such as branch and bound and cutting planes, enabling them to solve medium-sized integer linear programming problems in a reasonable time. The output shows the cluster partitioning results. ,in The actual number of clusters generated, per cluster Including no more than There are several deployment points. Output the center point of each cluster, denoted as . ,in The center point is the placement point selected as the center within the cluster, which satisfies... Output the associated anchor points for each cluster. This indicates which anchor point is best suited for releasing the unmanned underwater vehicle from the cluster, and the associated anchor point satisfies... anchor point When dealing with large-scale problems, solving integer linear programming problems can be slow. Column generation or branch-and-price strategies can be used to accelerate the process. Heuristic methods can also be combined to provide an initial solution and shorten the convergence time. A heuristic method could be a fast clustering algorithm based on a greedy strategy. For example, starting from each anchor point, the algorithm sequentially selects the nearest unassigned placement point until the load limit is reached. This greedy solution is then used as the initial feasible solution for the hybrid integer linear programming solver. The solver optimizes from this initial solution, leading to faster convergence to the optimal or near-optimal solution.

[0040] This step introduces an anchor point association mechanism and an adaptive weight optimization strategy, ensuring that the clustering results consider not only the distances between deployment points within a cluster but also the distance from the cluster to the safe zone. The adaptive weight strategy dynamically adjusts the trade-offs of optimization objectives based on the spatial relationship between deployment points and the safe zone. When deployment points are far from the safe zone, accessibility weights are automatically increased, making clustering optimization more inclined to select cluster partitioning schemes closer to the safe zone; conversely, when deployment points are close to the safe zone, compactness weights are automatically increased, making clustering optimization more inclined to select cluster partitioning schemes with smaller intra-cluster distances. The weighted quantile distance method comprehensively considers the distances from the nearest, median, and farthest points within a cluster to the anchor point, and incorporates the spatial dispersion of the cluster for correction, more accurately reflecting the actual cost of unmanned underwater vehicles (UUVs) accessing clusters. For elongated clusters, due to the larger spatial dispersion, the distances are amplified, causing the clustering algorithm to tend to avoid generating elongated clusters; for compact clusters, the distances remain relatively constant, causing the clustering algorithm to tend to generate compact clusters. These technical solutions improve the overall efficiency of task allocation, reduce the total range of the UUV, and increase the task completion speed.

[0041] In one embodiment of the present invention, step S5 includes: S51. Constructing a passable mask based on a two-dimensional discrete nautical chart. The chart indicates navigable and impassable areas, where W and H represent the width and height of the chart, respectively. Representing grid points Passage is permitted. This indicates that the point is an obstacle or a no-entry zone. The adjacency set N is defined as the set of eight directions including diagonal and direct neighbors, for a grid point. Its eight adjacent grid points are , , Set edge weights for direct adjacent moves and diagonal moves, where nodes are walkable cells and edges represent moves between adjacent cells. Edge weights Take the raster resolution, where The edge weight for moving two adjacent grid points (i.e., horizontal or vertical movement) is... The edge weight for a diagonal move (i.e., a diagonal shift) is , This represents the actual length of a single grid cell, i.e., the grid resolution, in kilometers or meters. If a line segment connecting two grid points crosses a restricted area or obstacle, the edge is deleted, and passage is not permitted. Specifically, for diagonal movement edges, it is checked whether the line segment connecting two diagonal grid points passes through a restricted area. If one or both of the two adjacent directly adjacent grid points of the two diagonal grid points are obstacles, the diagonal movement is considered infeasible, and the edge is deleted. This approach prevents unmanned underwater vehicles from navigating through diagonal gaps in obstacles, improving path safety.

[0042] S52. Obtain the center point of each cluster as the cluster representative point, denoted as... ,in , This represents the actual number of clusters generated. Cluster representative points are used to represent the location of the entire cluster, simplifying subsequent path search calculations. A is used. The algorithm calculates the shortest obstacle avoidance paths and distances between anchor points in the safe zone and cluster representative points, between cluster representative points, and between anchor points within the safe zone, obtaining the distance matrix and path polylines. Let the set of candidate anchor points within the safe zone be... Cluster represents the point set as Calculate and cache From anchor point a in the safe zone to the cluster representative point The shortest path distance and the corresponding polyline path. The path polyline is a sequence of grid point coordinates, representing the path from point a to point b. The complete path. For all combinations of anchor points and cluster representative points, calculation is required. Next A Search. Calculate. Cluster representative point arrive The shortest path distance and the corresponding polyline path. For all cluster representative point pairs, it is necessary to calculate Next A Search. Calculate. Anchor points within the safe zone arrive The distance, which can be used for maneuver planning of the carrier platform between different anchor points within the safe zone, needs to be calculated for all anchor point pairs. Next A Search. The typical complexity on sparse graphs is O(n log n). , where V is the node set. At a fixed resolution, for Pairwise computations of this scale can be processed in parallel or in batches, accelerated through multithreading or distributed computing. This distance and path information is used for subsequent task allocation and scheduling optimization, as well as recycling route planning.

[0043] S53. Based on the clustering results, associated anchor points, distance matrix, and path polylines, assign a task cluster to each unmanned underwater vehicle. According to the clustering results, each cluster... An anchor point has been associated. , will cluster Assigned to the corresponding unmanned underwater vehicle Plan the execution path starting from the associated anchor point, passing through each deployment point within the cluster, and returning to the associated anchor point. Regarding the access order of each deployment point within the cluster, since each cluster contains at most... Given a set of placement points, the optimal placement order can be obtained by solving the intra-cluster traveling salesman problem using state compression dynamic programming or enumeration methods. The time complexity of the state compression dynamic programming method is O(n log n). ,when At this time, the computational cost is very small, and it can be solved in milliseconds. The enumeration method directly enumerates all possible access orders, totaling... Such arrangements, when At this point, there are only 24 possible permutations. We can enumerate them all and calculate the total distance for each permutation, then select the permutation with the smallest total distance as the optimal access order. The execution path includes starting from the associated anchor point... The path to the first deployment point within the cluster, the paths between deployment points within the cluster, and the path back to the associated anchor point from the last deployment point within the cluster. The path. All paths use the shortest obstacle avoidance path, through A. The algorithm calculates and ensures that the unmanned underwater vehicle's trajectory is reachable under realistic nautical chart conditions and strictly avoids obstacles. Specifically, it sets up clusters... The optimal access order is The execution path is ,in This represents the shortest obstacle avoidance path between two points, via A. The algorithm calculates the total length of the execution path. This length indicates the length of the unmanned underwater vehicle. Execution Cluster Total mission range. Based on the unmanned underwater vehicle's speed. Execution time can be estimated. It is used for task scheduling and time planning.

[0044] Through the above steps, this invention realizes a complete task allocation method that comprehensively considers load constraints, cluster optimization, and obstacle avoidance path search. Starting from nautical chart data and deployment point location information, this method, through steps such as environmental preprocessing, anchor point generation, distance calculation, cluster optimization, and path planning, ultimately assigns task clusters and plans execution paths for each unmanned underwater vehicle (UUV). This invention incorporates an anchor point awareness mechanism, including the constraint that UUVs must depart from and return from a safe zone into the cluster optimization objective. An adaptive weighting strategy dynamically balances intra-cluster compactness and cluster accessibility to the safe zone based on the task scenario. A weighted quantile distance method considers the spatial morphological characteristics of the clusters, making the task allocation results more consistent with the actual execution scenario and improving the overall efficiency of UUV obstacle deployment tasks. Compared to traditional clustering methods that only consider intra-cluster compactness, this invention's method can significantly reduce the total range of UUVs, especially in scenarios where deployment points are far from the safe zone. Compared to traditional clustering methods based on single-point distance, the weighted quantile distance method of this invention can more accurately reflect the spatial morphology of clusters, avoid generating elongated clusters, and thus further reduce the range of unmanned underwater vehicles within clusters.

[0045] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for task allocation of obstacle placement by an unmanned underwater vehicle, characterized in that, include: S1. Obtain nautical chart data of the target sea area, location information of the obstacles to be deployed, and the number and payload limit of the unmanned underwater vehicles; S2. Based on the nautical chart data, perform environmental preprocessing, identify obstacle areas and define safe zones, and generate a set of anchor points within the safe zones using a distance-driven sampling method. The anchor points are candidate release and recovery locations for the unmanned underwater vehicle. S3. Based on the anchor point location and the deployment point location, calculate the shortest obstacle avoidance path distance between each anchor point and each deployment point; S4. The deployment points are clustered using an anchor-aware capacity-constrained clustering method. The cluster compactness and accessibility from the cluster to the safe zone are optimized simultaneously in the objective function. The weight coefficients of the two optimization objectives are adaptively determined according to the spatial distribution characteristics of the deployment points. The cluster division results and the associated anchor points of each cluster are obtained under the load constraints of the unmanned underwater vehicle. S5. Based on the cluster division results and associated anchor points, a path search algorithm is used to calculate the shortest obstacle avoidance paths between each cluster and the safe zone, as well as between each cluster, and to assign task clusters and plan execution paths for each unmanned underwater vehicle.

2. The task allocation method for deploying obstacles by an unmanned underwater vehicle as described in claim 1, characterized in that, The environmental preprocessing described in step S2 includes: constructing water depth partitions based on the nautical chart matrix, dividing the sea area into five categories: land, shallow water, medium water, deep water, and deep sea, and determining navigable areas; Construct an obstacle area mask to mark impassable areas, including islands, shoals, land, and no-navigation zones, using a binary mask representation; define a safe zone polygon as the range for the carrier platform to release the unmanned underwater vehicle, defined by a set of coordinate points; construct a deployment area mask based on water depth categories to exclude unsuitable waters for deployment.

3. The task allocation method for deploying obstacles by an unmanned underwater vehicle as described in claim 1, characterized in that, The distance-driven sampling method described in step S2 includes: calculating the centroid of all deployment points; selecting the point closest to the centroid from the boundary of the safe zone as the first anchor point; sequentially selecting the boundary point farthest from the existing anchor points as the next anchor point, using the farthest point sampling strategy; repeating the above steps until K anchor points are generated, where the number of anchor points K is determined by dividing the total number of deployment points by the upper limit of the load.

4. The task allocation method for deploying obstacles by an unmanned underwater vehicle as described in claim 1, characterized in that, Step S3 specifically includes: S31. Obtain the target deployment point set according to the task requirements, filter out points that do not meet the water depth and area conditions, and obtain the set of valid deployment points located within the allowed deployment area. Each deployment point corresponds to one obstacle to be deployed. S32. Traverse the set of anchor points and the set of effective deployment points, use a path search algorithm to calculate the shortest obstacle avoidance path distance between each anchor point and each deployment point, and store the distance data in the anchor point distance matrix.

5. The task allocation method for deploying obstacles by an unmanned underwater vehicle as described in claim 1, characterized in that, Step S4 specifically includes: S41. Calculate the spatial distribution characteristic parameters of the deployment point set based on the anchor point distance matrix and the obstacle avoidance shortest path distance matrix between deployment points; S42. Based on the spatial distribution characteristic parameters, an adaptive weight calculation strategy is used to determine the weight coefficients for intra-cluster compactness and cluster-to-safe-zone reachability. S43. Establish an anchor-aware capacity-constrained clustering model, introduce decision variables, construct a dual-objective weighted optimization objective function, and set constraints. S44. For each cluster and its associated anchor point, the actual distance from the anchor point to the cluster is calculated using the weighted quantile distance method based on cluster morphology. S45. Solve the clustering model using a mixed-integer linear programming solver, and output the cluster partitioning results, the center point of each cluster, and the associated anchor point of each cluster.

6. The task allocation method for deploying obstacles by an unmanned underwater vehicle as described in claim 5, characterized in that, The adaptive weight calculation strategy in step S42 includes: calculating the average distance from all deployment points to the nearest anchor point to represent the overall distance relationship between the deployment point set and the safe zone; calculating the average distance between deployment points to represent the internal dispersion of the deployment point set; calculating the ratio of the two average distances as the distance ratio; and using the Sigmoid function to map the distance ratio to the weight coefficient of the cluster's reachability to the safe zone, wherein the weight coefficient of the cluster's compactness is 1 minus the reachability weight coefficient.

7. The task allocation method for deploying obstacles by an unmanned underwater vehicle as described in claim 5, characterized in that, The decision variables introduced in step S43 include: a central decision variable indicating whether a representation point is selected as a cluster center, an allocation decision variable indicating whether a representation point is assigned to a cluster, and an anchor-related decision variable indicating whether a cluster is associated with an anchor point; the objective function is: ； Where P is the set of valid deployment points, and i and j are the indices of the deployment points. Let be the shortest obstacle avoidance path distance between points i and j. To assign decision variables, This is the intra-cluster compactness weighting coefficient. For the k-th anchor point, For the i-th deployment point, This represents the shortest obstacle avoidance path distance from the anchor point to the deployment point. As anchor points for decision variables, Let i be the central decision variable, representing whether point i is selected as the cluster center. is the accessibility weighting coefficient from the cluster to the safe zone, and K is the number of unmanned underwater vehicles.

8. The task allocation method for deploying obstacles by an unmanned underwater vehicle as described in claim 5, characterized in that, The constraints in step S43 include: (1) each point must be assigned to a cluster; (2) each point can only be assigned to the selected center point; (3) the capacity of each cluster does not exceed the upper limit of the unmanned underwater vehicle's payload; (4) the number of clusters does not exceed the number of unmanned underwater vehicles; (5) each cluster must be associated with an anchor point, and if a deployment point is the center of a cluster, then there must be one and only one anchor point associated with it.

9. The task allocation method for deploying obstacles by an unmanned underwater vehicle as described in claim 5, characterized in that, The weighted quantile distance method based on cluster morphology described in step S44 includes: sorting the distances from all points within a cluster to the anchor point, obtaining the nearest point distance, median distance, and farthest point distance; assigning different weights to the three quantile distances using a decreasing weight function; calculating the normalized spatial dispersion of the cluster; and multiplying the weighted quantile distance by the spatial dispersion correction factor to obtain the actual distance from the anchor point to the cluster, calculated using the following formula: ； Where Q is the set of quantiles, including the quantile of 0.0 corresponding to the nearest point, the quantile of 0.5 corresponding to the median point, and the quantile of 1.0 corresponding to the farthest point. The weight of the q-th quantile. anchor point to cluster The qth quantile distance This is the dispersion penalty coefficient. For clusters The normalized spatial discreteness.

10. The task allocation method for deploying obstacles by an unmanned underwater vehicle as described in claim 1, characterized in that, Step S5 specifically includes: S51. Construct a passable mask based on a two-dimensional discrete nautical chart to identify passable and impassable areas, define an adjacency set in eight directions including diagonal and direct neighbors, and set edge weights for direct neighbor movement and diagonal movement. S52. Obtain the center point of each cluster as the cluster representative point, and use the path search algorithm to calculate the shortest obstacle avoidance path and distance between the anchor point of the safe zone and the cluster representative point, between the cluster representative points and the anchor points within the safe zone, respectively, to obtain the distance matrix and path polyline. S53. Based on the cluster division results, associated anchor points, distance matrix and path polyline, assign a task cluster to each unmanned underwater vehicle and plan the execution path from the associated anchor point, through each deployment point in the cluster and back to the associated anchor point.

Images