Anti-unmanned ship unmanned aerial vehicle swarm dynamic clustering algorithm and saturation attack path planning method

Abstract

The invention discloses an anti-unmanned ship unmanned aerial vehicle swarm dynamic clustering algorithm and a saturation attack path planning method, and belongs to the technical field of path planning. The invention discloses an anti-unmanned ship unmanned aerial vehicle swarm dynamic clustering algorithm and saturation attack path planning method. The method comprises the steps of constructing a multi-dimensional situation awareness index system; constructing a clustering decision function; and establishing a cluster structure dynamic adjustment mechanism, and carrying out collaborative scheduling optimization on the cluster structure. According to the invention, the problem of decision error caused by incapability of timely obtaining accurate information when communication is interrupted in the prior art is solved. According to the method, dynamic changes of the bee colony targets can be more flexibly coped with, it is ensured that each bee colony target is covered with the corresponding interception cluster, even if part of communication is interrupted, all clusters can still make decisions autonomously according to local information, certain interception capacity is maintained, basic interception tasks continue to be executed under the condition that communication is limited, and the communication efficiency is improved. It is ensured that interception preparation is completed before the bee colony reaches the defense target, and defense failure caused by response delay is effectively avoided.

Description

Technical Field

[0001] This invention relates to the field of path planning technology, specifically to a dynamic clustering algorithm for anti-unmanned surface vessel (USV) swarms and a saturation attack path planning method. Background Technology

[0002] Swarm operations, comprised of unmanned surface vessels (USVs) and unmanned aerial vehicles (UAVs), have become a significant threat to modern operations. These swarms are characterized by their large numbers, high mobility, strong coordination, and low cost, enabling them to penetrate traditional defense systems through saturation attacks and inflict devastating damage on targets.

[0003] Chinese patent CN120447617A discloses a method and system for UAV swarm cooperative path planning based on deep reinforcement learning, relating to the field of UAV swarm cooperative technology. The method includes: acquiring environmental state information; determining a preliminary path within a pre-constructed state space based on a sampling algorithm; and performing local smoothing and obstacle avoidance on the preliminary path; using deep reinforcement learning to determine the UAV's action commands at each time step and correcting the preliminary path in real time; and achieving overall swarm cooperative movement through periodic exchange of local information between the UAVs. This UAV swarm cooperative path planning method and system improves the accuracy of path planning and the cooperativeness of UAV actions.

[0004] In threat interception scenarios, the aforementioned patents address the issue that communication may be disrupted due to interference or environmental factors. When communication is interrupted, accurate information may not be obtained in a timely manner, leading to decision-making errors. Therefore, these patents do not meet existing requirements. To address this, we propose a dynamic clustering algorithm for anti-unmanned surface vessels and drone swarms, along with a saturation attack path planning method. Summary of the Invention

[0005] The purpose of this invention is to provide a dynamic clustering algorithm and saturation attack path planning method for anti-unmanned surface vessel (USV) swarms. This method can more flexibly respond to the dynamic changes of swarm targets, ensuring that each swarm target has a corresponding interception cluster coverage. Even if some communication is interrupted, each cluster can still make autonomous decisions based on local information to maintain a certain interception capability. It can continue to perform basic interception tasks under communication constraints, ensuring that interception preparation is completed before the swarm reaches the defense target. This effectively avoids defense failure caused by response delay, reduces the probability of interception units being exposed in high-risk areas, avoids path conflicts, ensures that multiple interception units arrive at the interception point simultaneously, improves the interception effect against highly mobile swarms, and solves the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a dynamic clustering algorithm for anti-unmanned surface vessel (USV) swarms, comprising:

[0007] Construct a multi-dimensional situational awareness indicator system, which includes swarm density indicators, swarm movement speed and direction indicators, swarm attack priority indicators, and interception unit performance matching indicators.

[0008] The analytic hierarchy process (AHP) is used to assign weights to a multi-dimensional situational awareness indicator system and to construct a clustered decision function.

[0009] A dynamic adjustment mechanism for cluster structure is established based on the cluster decision function and the K-means algorithm, and a collaborative communication mechanism between cluster structures is established to perform collaborative scheduling optimization of the cluster structure.

[0010] Preferably, the construction of the multi-dimensional situational awareness indicator system specifically includes:

[0011] The bee colony density index is obtained by fusing data collected from radar, photoelectric, and sonar sensors.

[0012] The Kalman filter algorithm is used to process the position data of the bee colony at multiple consecutive moments to predict the movement speed vector and direction angle of the bee colony. Then, by calculating the cosine similarity between the movement speed vectors of the bee colony and the interception unit, the speed and direction matching degree is obtained, which is the bee colony movement speed and direction index.

[0013] The priority index for swarm attacks is determined based on the distance between the swarm and the target, the type of equipment carried by the swarm, and the movement trend of the swarm.

[0014] The performance matching index of the interception unit is determined based on the degree of adaptation of the interception unit's range, movement speed, interception success rate, and the characteristics of the bee colony.

[0015] Preferably, the step of using the analytic hierarchy process (AHP) to assign weights to the multi-dimensional situational awareness indicator system and constructing a clustered decision function specifically includes:

[0016] The importance of each indicator in the multi-dimensional situational awareness indicator system is compared pairwise to obtain the assessment results of importance.

[0017] A judgment matrix is ​​constructed based on the evaluation results, and the consistency ratio of the judgment matrix is ​​calculated. When the consistency ratio of the judgment matrix is ​​less than 0.1, the judgment matrix is ​​deemed reasonable and valid.

[0018] Based on the effective judgment matrix, the weights of each indicator are calculated using the eigenvalue method, and the sum of the weights of all indicators is 1.

[0019] By combining the weights of each indicator with the quantified indicator values, a clustering decision function is constructed.

[0020] Preferably, the step of calculating the weights of each indicator based on the effective judgment matrix using the eigenvalue method specifically includes:

[0021] For indicators at the same level, perform pairwise comparisons, calculate the average value of each comparison pair, and construct a first-level indicator judgment matrix and a second-level sub-indicator judgment matrix under each first-level indicator based on the average value.

[0022] Calculate the largest eigenvalue and consistency index of the judgment matrix. Based on the determined average random consistency index, calculate the consistency ratio. If the consistency ratio is <0.1, the judgment matrix passes the consistency test.

[0023] The eigenvectors of the judgment matrix are calculated using the eigenvalue method, and these eigenvectors are the initial weights of the corresponding indicators.

[0024] Preferably, the step of establishing a dynamic cluster structure adjustment mechanism based on the clustering decision function and the K-means algorithm, and using this mechanism to dynamically adjust the cluster structure, specifically includes:

[0025] Calculate the local density and distance index of each interception unit;

[0026] The interception units are sorted according to the product of local density and distance index, and the top N interception units in the product ranking are selected as the initial cluster structure.

[0027] The matching degree between each interception unit and the target beehive corresponding to each initial cluster structure is calculated based on the clustering decision function. The interception unit with the highest matching degree is assigned to the corresponding cluster to complete the initial clustering.

[0028] Establish a dynamic adjustment mechanism for cluster structure, and use this mechanism to dynamically adjust the cluster structure.

[0029] Preferably, the establishment of the cluster structure dynamic adjustment mechanism specifically includes:

[0030] Real-time monitoring of bee colony density changes; when the number of bee colonies in a unit area exceeds a preset threshold, the original cluster is automatically split into multiple sub-clusters, the initial center of the sub-clusters is recalculated, and interception units are allocated.

[0031] When the bee colony is detected to have split into multiple sub-colonies, the situational data of each sub-colonie is re-collected, the clustering decision function is calculated, and the cluster boundaries and centers are adjusted to ensure that each sub-colonie corresponds to at least one interception cluster.

[0032] When an interception unit fails, the system automatically searches for the performance matching degree of other interception units in the same cluster and assigns the task of the failed unit to the normal unit with the highest matching degree.

[0033] If there are no available normal units in the same cluster, calculate the clustering decision function matching degree between the cluster and the adjacent clusters, merge it into the adjacent cluster with the highest matching degree, and then redistribute the task scope of the interception units in the cluster.

[0034] The attack path planning method based on the anti-unmanned surface vessel (USV) drone swarm dynamic clustering algorithm includes the following steps:

[0035] S1: Construct a multi-objective optimization model with the shortest interception time, lowest path risk, and highest coordination efficiency as optimization objectives;

[0036] S2: Solve the multi-objective optimization model, and perform fuzzy comprehensive evaluation on the optimal path set based on the solution results to obtain the optimal path;

[0037] S3: Receive real-time data from radar, photoelectric, and sonar sensors and compare it with the initial clustering results and path planning results;

[0038] S4: When the straight-line distance between the actual position of the bee colony and the initial predicted position exceeds the set threshold, the position deviation is determined to be excessive.

[0039] S5: When the difference between the actual performance of the interception unit and the preset performance exceeds 20%, the state deviation is determined to be excessive.

[0040] S6: When the difference between the actual number of bee colonies intercepted and the expected number of interceptions exceeds 30% of the expected value, the effect deviation is judged to be excessive.

[0041] S7: Make appropriate corrections to the deviations that exceed the standards based on the comparison results.

[0042] Preferably, the calculation of the target with the shortest interception time specifically includes:

[0043] The path time for each interception unit is calculated to obtain the target with the shortest interception time. The path time consists of the base path time and the swarm movement correction time. The base path time is the time it takes for the interception unit to travel from its current position to the initial interception point.

[0044] The path length is determined based on the moving speed of the interception unit. When the moving speed of the interception unit is lower than a preset threshold, the shortest path is selected. When the speed is higher than the threshold, the path length is increased according to the distribution density of the risk area.

[0045] Collect location data of the bee colony over the past 10 time points, establish a movement trajectory model, and use the movement trajectory model to predict the position offset of the bee colony during the movement of the interception unit;

[0046] The additional time required for the interception unit is calculated based on the position offset, thus obtaining the swarm movement correction time.

[0047] Preferably, the step of correcting the deviation exceeding the standard based on the comparison results specifically includes:

[0048] When the analysis indicates that the position deviation exceeds the standard, the Kalman filter algorithm is called again to predict the movement trajectory of the bee colony in the next 30 seconds, and the target point of the interception path is adjusted according to the new trajectory.

[0049] When the analysis indicates that the state deviation exceeds the standard, if it is a single interception unit failure, retrieve the performance data of other units in the same cluster and assign the task of the failure unit to the unit with the highest performance matching degree; if it is multiple unit failures, split or merge the related cluster structure.

[0050] When the analysis shows that the performance deviation exceeds the standard, analyze the reasons for the deviation. If it is due to the number of bee colonies exceeding the expectation, increase the number of interception units for the clusters corresponding to high-priority bee colonies. If the interception fails due to excessive path risk, adjust the risk coefficient weight in the path planning and regenerate the path plan.

[0051] Preferably, the step of solving the multi-objective optimization model and obtaining the optimal path by performing fuzzy comprehensive evaluation on the optimal path set based on the solution results specifically includes:

[0052] Multiple random path schemes are generated as an initial population, and the process is iterated.

[0053] During the iteration process, the middle segments of two path schemes are randomly selected for exchange, and the direction of the path segments is randomly adjusted to generate a new population;

[0054] For each path scheme in the population, the fitness value is calculated according to the optimization objective. After the calculation is completed, the Pareto level is divided by non-dominated sorting, and the path schemes with higher levels are retained.

[0055] The Pareto optimal path set is obtained by sorting the path schemes with higher congestion levels and iterating continuously on the path schemes with higher levels.

[0056] The optimal path set is evaluated using fuzzy comprehensive analysis, and the path scheme with the highest score is taken as the final path planning result.

[0057] Compared with the prior art, the beneficial effects of the present invention are:

[0058] This invention utilizes a dynamic clustering mechanism based on multi-dimensional situational awareness to adjust the cluster structure in real time according to swarm density, movement trends, threat levels, and interception unit performance. This allows for more flexible responses to dynamic changes in swarm targets, ensuring that each swarm target has a corresponding interception cluster coverage. This gives the anti-UAV swarm dynamic clustering algorithm a certain degree of distributed characteristics. Even with partial communication interruptions, each cluster can still make autonomous decisions based on local information, maintaining a certain level of interception capability. It continues to perform basic interception tasks even under communication constraints. Simultaneously, through cluster structure optimization and collaborative scheduling, it avoids resource redundancy or shortage in local areas, ensuring that interception preparation is completed before the swarm reaches the defense target. This effectively avoids defense failure due to response delays. A multi-target optimization model for saturated attack scenarios is constructed with the shortest interception time, lowest path risk, and highest collaborative efficiency as optimization objectives. This reduces the probability of interception units being exposed in high-risk areas, avoids path conflicts, ensures that multiple interception units arrive at the interception point simultaneously, and improves the interception efficiency against highly mobile swarms. Attached Figure Description

[0059] Figure 1 This is a flowchart of the anti-unmanned surface vessel (USV) drone swarm dynamic clustering algorithm of the present invention;

[0060] Figure 2 This is a flowchart of the attack path planning method based on the anti-unmanned surface vessel (USV) swarm dynamic clustering algorithm of the present invention. Detailed Implementation

[0061] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0062] To address the issue that existing technologies may experience communication disruptions or interruptions due to environmental factors in threat interception scenarios, potentially leading to inaccurate information acquisition and decision-making errors, please refer to [link to relevant documentation]. Figures 1-2 This embodiment provides the following technical solution:

[0063] The dynamic clustering algorithm for anti-unmanned surface vessel (USV) swarms includes:

[0064] Construct a multi-dimensional situational awareness indicator system, which includes swarm density indicators, swarm movement speed and direction indicators, swarm attack priority indicators, and interception unit performance matching indicators.

[0065] The analytic hierarchy process (AHP) is used to assign weights to a multi-dimensional situational awareness indicator system and to construct a clustered decision function.

[0066] A dynamic adjustment mechanism for cluster structure is established based on the cluster decision function and the K-means algorithm. A collaborative communication mechanism between cluster structures is also established to optimize the collaborative scheduling of cluster structures. When the interception pressure of a certain cluster structure exceeds the threshold, a collaborative request is automatically sent to the adjacent cluster structure. The adjacent cluster structure decides whether to respond based on its own resource redundancy. After responding, some of the bee colony interception tasks are transferred to the collaborative cluster structure by adjusting the cluster boundary. During the transfer process, it is necessary to ensure that the coverage of the two cluster structures does not overlap and there are no defense blind spots. After the task transfer is completed, the interception pressure of the two cluster structures is recalculated to ensure that both are below the threshold.

[0067] Construct a multi-dimensional situational awareness indicator system, specifically including:

[0068] The swarm density index is obtained by fusing data collected by radar, photoelectric and sonar sensors. It represents the total number of unmanned surface vessels and drones within a unit defense area. The specific calculation method is to count the total number of swarm units within a unit area.

[0069] The Kalman filter algorithm is used to process the position data of the bee colony at multiple consecutive moments to predict the movement speed vector and direction angle of the bee colony. Then, by calculating the cosine similarity between the movement speed vectors of the bee colony and the interception unit, the speed and direction matching degree is obtained, which is the index of the bee colony's movement speed and direction. The higher the speed and direction matching degree similarity, the more consistent the movement directions of the two are.

[0070] The attack priority index of the bee swarm is determined based on the distance between the bee swarm and the defense target, the type of equipment carried by the bee swarm, and the movement trend of the bee swarm. The closer the bee swarm is to the defense target, the more offensive equipment it carries, and the more it moves towards the defense target, the higher the attack priority. The priority is quantified using a 1-10 point system, with 10 points being the highest priority.

[0071] The performance matching index of the interception unit is determined based on the degree of adaptation of the interception unit's range, moving speed, interception success rate and the characteristics of the bee colony. If the interception unit's range can cover the bee colony's moving range, its moving speed is greater than the bee colony's moving speed, and its interception success rate is high, then the performance matching degree is higher. Similarly, a value between 0 and 1 is used for quantification.

[0072] The Analytic Hierarchy Process (AHP) is used to assign weights to a multi-dimensional situational awareness indicator system and to construct a clustered decision function, specifically including:

[0073] The importance of each indicator in the multi-dimensional situational awareness indicator system is compared pairwise to obtain the assessment results of importance. For example, the difference in importance between the swarm attack priority indicator and the swarm density indicator is judged. If the attack priority indicator is more important, it is assigned a higher value in the judgment matrix.

[0074] A judgment matrix is ​​constructed based on the evaluation results, and the consistency ratio of the judgment matrix is ​​calculated. When the consistency ratio of the judgment matrix is ​​less than 0.1, the judgment matrix is ​​deemed reasonable and valid.

[0075] Based on the effective judgment matrix, the weights of each indicator are calculated using the eigenvalue method, and the sum of the weights of all indicators is 1.

[0076] By combining the weights of each indicator with the quantified indicator values, a clustering decision function is constructed. The clustering decision function is used to calculate the cluster matching degree between the interception unit and the target bee colony. The larger the matching degree value, the stronger the adaptability between the interception unit and the target bee colony, and the more suitable it is to be grouped into the same cluster.

[0077] Based on the effective judgment matrix, the weights of each indicator are calculated using the eigenvalue method, specifically including:

[0078] Calculate the average assignment for each comparison pair, and construct the primary indicator judgment matrix and the secondary sub-indicator judgment matrix under each primary indicator based on the average assignment.

[0079] Calculate the maximum eigenvalue and consistency index of the judgment matrix. Based on the determined average random consistency index, calculate the consistency ratio. If the consistency ratio is <0.1, the judgment matrix passes the consistency test. If the consistency ratio is ≥0.1, it needs to be re-evaluated and the assigned values ​​adjusted until the consistency ratio is <0.1. For example, when the consistency ratio of the indicator judgment matrix is ​​0.12, find the comparison pair with the largest assignment difference, and reassign the values ​​based on the evaluation criteria to reduce the assignment deviation until the consistency ratio meets the requirements.

[0080] The eigenvectors of the judgment matrix are calculated using the eigenvalue method, and these eigenvectors represent the initial weights of the corresponding indicators. For example, for the judgment matrix of the first-level indicators, the eigenvector [w] is calculated. A w B w C w D [w] represents the initial weights of the four primary indicators; for the judgment matrix of the secondary sub-indicators, such as the judgment matrix under A, the eigenvector [w] is calculated. A1 w A2 ], which are the initial weights of A1 and A2. The initial weights are normalized to ensure that the sum of the weights of indicators at the same level is 1. Finally, the global weights of the second-level sub-indicators are calculated for subsequent clustering decision function calculations.

[0081] A dynamic cluster structure adjustment mechanism is established based on the clustering decision function and the K-means algorithm. This mechanism is then used to dynamically adjust the cluster structure, specifically including:

[0082] Calculate the local density and distance index of each interception unit. That is, count the number of other interception units within a preset cutoff distance centered on the interception unit. The more units there are, the higher the local density. Calculate the distance index of each interception unit, which is the distance between the interception unit and all interception units with a local density higher than its own. Select the minimum distance among them as the distance index of the interception unit.

[0083] The interception units are sorted according to the product of local density and distance index, and the top N interception units in the product ranking are selected as the initial cluster structure. N is determined according to the total number of interception units in the defense area, usually 1 / 5 to 1 / 3 of the total number.

[0084] The matching degree between each interception unit and the target beehive corresponding to each initial cluster structure is calculated based on the clustering decision function. The interception unit with the highest matching degree is assigned to the corresponding cluster to complete the initial clustering.

[0085] Establish a dynamic adjustment mechanism for cluster structure, and use this mechanism to dynamically adjust the cluster structure.

[0086] Establish a dynamic adjustment mechanism for cluster structure, specifically including:

[0087] Real-time monitoring of bee colony density changes; when the number of bee colonies in a unit area exceeds a preset threshold, the original cluster is automatically split into 2 to 3 sub-clusters, the initial center of the sub-clusters is recalculated, and interception units are allocated.

[0088] When the bee colony is detected to have split into multiple sub-colonies, the situational data of each sub-colonie is re-collected, the clustering decision function is calculated, and the cluster boundaries and centers are adjusted to ensure that each sub-colonie corresponds to at least one interception cluster.

[0089] When an interception unit fails, the system automatically searches for the performance matching degree of other interception units in the same cluster and assigns the task of the failed unit to the normal unit with the highest matching degree.

[0090] If there are no available normal units in the same cluster, calculate the clustering decision function matching degree between the cluster and the adjacent clusters, merge it into the adjacent cluster with the highest matching degree, and then redistribute the task scope of the interception units in the cluster.

[0091] The attack path planning method based on the anti-unmanned surface vessel (USV) drone swarm dynamic clustering algorithm includes the following steps:

[0092] S1: Construct a multi-objective optimization model with the shortest interception time, lowest path risk, and highest coordination efficiency as optimization objectives;

[0093] S2: Solve the multi-objective optimization model, and perform fuzzy comprehensive evaluation on the optimal path set based on the solution results to obtain the optimal path;

[0094] S3: Receive real-time data from radar, photoelectric, and sonar sensors and compare it with the initial clustering results and path planning results;

[0095] S4: When the straight-line distance between the actual position of the bee colony and the initial predicted position exceeds the set threshold, the position deviation is determined to be excessive.

[0096] S5: When the difference between the actual performance of the interception unit and the preset performance exceeds 20%, the state deviation is determined to be excessive.

[0097] S6: When the difference between the actual number of bee colonies intercepted and the expected number of interceptions exceeds 30% of the expected value, the effect deviation is judged to be excessive.

[0098] S7: Make appropriate corrections to the deviations that exceed the standards based on the comparison results.

[0099] The calculation of the target with the shortest interception time specifically includes:

[0100] The path time for each interception unit is calculated to obtain the target with the shortest interception time. The path time consists of the base path time and the swarm movement correction time. The base path time is the time it takes for the interception unit to travel from its current position to the initial interception point.

[0101] The path length is determined based on the movement speed of the interception unit. When the movement speed of the interception unit is lower than a preset threshold, the shortest path is selected.

[0102] When the speed exceeds the threshold, the path length is increased according to the distribution density of the risk area to avoid high-risk areas.

[0103] Collect location data of the bee colony over the past 10 time points, establish a movement trajectory model, and use the movement trajectory model to predict the position offset of the bee colony during the movement of the interception unit;

[0104] The additional time required for the interception unit is calculated based on the position offset, and the swarm movement correction time is obtained to ensure that the corrected interception time matches the real-time position of the swarm.

[0105] Based on the comparison results, appropriate corrections are made to the deviations that exceed the standards, specifically including:

[0106] When the analysis indicates that the position deviation exceeds the standard, the Kalman filter algorithm is called again to predict the movement trajectory of the bee colony in the next 30 seconds. The target point of the interception path is adjusted according to the new trajectory. At the same time, the time and risk changes after the path adjustment are calculated to ensure that the adjusted path still meets the optimization objective.

[0107] When the analysis indicates that the state deviation exceeds the standard, if it is a single interception unit failure, the performance data of other units in the same cluster are retrieved, and the task of the failure unit is assigned to the unit with the highest performance matching degree. The matching degree is calculated based on a comprehensive score of interception range, speed and success rate. A score ≥80 is considered suitable. If multiple units fail, the relevant cluster structure is split or merged. When splitting, ensure that the number of interception units in each sub-cluster is not less than 3. When merging, ensure that the defense range of the new cluster does not exceed the maximum range of the interception unit.

[0108] When the analysis shows that the effect deviation exceeds the standard, analyze the cause of the deviation. If it is due to the number of bee colonies exceeding the expectation, increase the number of interception units for the corresponding clusters of high-priority bee colonies by 20% to 50%. If the interception fails due to excessive path risk, adjust the risk coefficient weight in the path planning, increase the weight of environmental risk or bee colony counterattack risk by 0.1-0.2, and regenerate the path plan.

[0109] The multi-objective optimization model is solved, and the optimal path is obtained by fuzzy comprehensive evaluation of the optimal path set based on the solution results. Specifically, this includes:

[0110] Multiple random path schemes are generated as an initial population, and the process is iterated.

[0111] During the iteration process, the middle segments of two path schemes are randomly selected for exchange, and the direction of the path segments is randomly adjusted to generate a new population;

[0112] For each path scheme in the population, the fitness value is calculated according to the optimization objective. After the calculation is completed, the Pareto level is divided by non-dominated sorting, and the path schemes with higher levels are retained.

[0113] The path schemes with higher levels of congestion are sorted to ensure population diversity and avoid local optima. The Pareto optimal path set is obtained by iterating continuously on the path schemes with higher levels of congestion.

[0114] The optimal path set is evaluated using fuzzy comprehensive analysis, and the path scheme with the highest score is taken as the final path planning result.

[0115] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0116] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. A dynamic clustering algorithm for anti-unmanned surface vessel (USV) swarms, characterized in that, include: Construct a multi-dimensional situational awareness indicator system, which includes swarm density indicators, swarm movement speed and direction indicators, swarm attack priority indicators, and interception unit performance matching indicators. The analytic hierarchy process (AHP) is used to assign weights to a multi-dimensional situational awareness indicator system and to construct a clustered decision function. A dynamic adjustment mechanism for cluster structure is established based on the cluster decision function and the K-means algorithm, and a collaborative communication mechanism between cluster structures is established to perform collaborative scheduling optimization of the cluster structure.

2. The anti-unmanned surface vessel (USV) swarm dynamic clustering algorithm according to claim 1, characterized in that, The construction of the multi-dimensional situational awareness indicator system specifically includes: The bee colony density index is obtained by fusing data collected from radar, photoelectric, and sonar sensors. The Kalman filter algorithm is used to process the position data of the bee colony at multiple consecutive moments to predict the movement speed vector and direction angle of the bee colony. Then, by calculating the cosine similarity between the movement speed vectors of the bee colony and the interception unit, the speed and direction matching degree is obtained, which is the bee colony movement speed and direction index. The priority index for swarm attacks is determined based on the distance between the swarm and the target being defended, the type of equipment carried by the swarm, and the swarm's movement trend. The performance matching index of the interception unit is determined based on the degree of adaptation of the interception unit's range, movement speed, interception success rate, and the characteristics of the bee colony.

3. The anti-unmanned surface vessel (USV) swarm dynamic clustering algorithm according to claim 1, characterized in that, The method of employing the analytic hierarchy process (AHP) to assign weights to the multi-dimensional situational awareness indicator system and constructing a clustered decision function specifically includes: The importance of each indicator in the multi-dimensional situational awareness indicator system is compared pairwise to obtain the importance assessment results; A judgment matrix is ​​constructed based on the evaluation results, and the consistency ratio of the judgment matrix is ​​calculated. When the consistency ratio of the judgment matrix is ​​less than 0.1, the judgment matrix is ​​deemed reasonable and valid. Based on the effective judgment matrix, the weights of each indicator are calculated using the eigenvalue method, and the sum of the weights of all indicators is 1. By combining the weights of each indicator with the quantified indicator values, a clustering decision function is constructed.

4. The anti-unmanned surface vessel (USV) swarm dynamic clustering algorithm according to claim 3, characterized in that, The weights of each indicator are calculated using the eigenvalue method based on the effective judgment matrix, specifically including: For indicators at the same level, perform pairwise comparisons, calculate the average value of each comparison pair, and construct a first-level indicator judgment matrix and a second-level sub-indicator judgment matrix under each first-level indicator based on the average value. Calculate the largest eigenvalue and consistency index of the judgment matrix. Based on the determined average random consistency index, calculate the consistency ratio. If the consistency ratio is less than 0.1, the judgment matrix passes the consistency test. The eigenvectors of the judgment matrix are calculated using the eigenvalue method, and these eigenvectors are the initial weights of the corresponding indicators.

5. The anti-unmanned surface vessel (USV) swarm dynamic clustering algorithm according to claim 1, characterized in that, The aforementioned dynamic adjustment mechanism for cluster structure based on clustering decision functions and the K-means algorithm, and the dynamic adjustment of cluster structure using this mechanism, specifically includes: Calculate the local density and distance index of each interception unit; The interception units are sorted according to the product of local density and distance index, and the top N interception units in the product ranking are selected as the initial cluster structure. The matching degree between each interception unit and the target beehive corresponding to each initial cluster structure is calculated based on the clustering decision function. The interception unit with the highest matching degree is assigned to the corresponding cluster to complete the initial clustering. Establish a dynamic adjustment mechanism for cluster structure, and use this mechanism to dynamically adjust the cluster structure.

6. The anti-unmanned surface vessel (USV) swarm dynamic clustering algorithm according to claim 5, characterized in that, The aforementioned dynamic adjustment mechanism for cluster structure specifically includes: Real-time monitoring of bee colony density changes; when the number of bee colonies in a unit area exceeds a preset threshold, the original cluster is automatically split into multiple sub-clusters, the initial center of the sub-clusters is recalculated, and interception units are allocated. When the bee colony is detected to have split into multiple sub-colonies, the situational data of each sub-colonie is re-collected, the clustering decision function is calculated, and the cluster boundaries and centers are adjusted to ensure that each sub-colonie corresponds to at least one interception cluster. When an interception unit fails, the system automatically searches for the performance matching degree of other interception units in the same cluster and assigns the task of the failed unit to the normal unit with the highest matching degree. If there are no available normal units in the same cluster, calculate the clustering decision function matching degree between the cluster and the adjacent clusters, merge it into the adjacent cluster with the highest matching degree, and then redistribute the task scope of the interception units in the cluster.

7. An attack path planning method based on an anti-unmanned surface vessel (USV) swarm dynamic clustering algorithm, characterized in that, Includes the following steps: S1: Construct a multi-objective optimization model with the shortest interception time, lowest path risk, and highest coordination efficiency as optimization objectives; S2: Solve the multi-objective optimization model, and perform fuzzy comprehensive evaluation on the optimal path set based on the solution results to obtain the optimal path; S3: Receive real-time data from radar, photoelectric, and sonar sensors and compare it with the initial clustering results and path planning results; S4: When the straight-line distance between the actual position of the bee colony and the initial predicted position exceeds the set threshold, the position deviation is determined to be excessive. S5: When the difference between the actual performance of the interception unit and the preset performance exceeds 20%, the state deviation is determined to be excessive. S6: When the difference between the actual number of bee colonies intercepted and the expected number of interceptions exceeds 30% of the expected value, the effect deviation is judged to be excessive. S7: Make appropriate corrections to the deviations that exceed the standards based on the comparison results.

8. The attack path planning method based on the anti-unmanned surface vessel (USV) swarm dynamic clustering algorithm according to claim 7, characterized in that, The calculation of the target with the shortest interception time specifically includes: The path time of each interception unit is calculated to obtain the target with the shortest interception time. The path time consists of the base path time and the swarm movement correction time. The base path time is the time it takes for the interception unit to travel from its current position to the initial interception point. The path length is determined based on the moving speed of the interception unit. When the moving speed of the interception unit is lower than a preset threshold, the shortest path is selected. When the speed is higher than the threshold, the path length is increased according to the distribution density of the risk area. Collect location data of the bee colony over the past 10 time points, establish a movement trajectory model, and use the movement trajectory model to predict the position offset of the bee colony during the movement of the interception unit; The additional time required for the interception unit is calculated based on the position offset, thus obtaining the swarm movement correction time.

9. The attack path planning method based on the anti-unmanned surface vessel (USV) swarm dynamic clustering algorithm according to claim 7, characterized in that, The correction of deviations exceeding the standard based on the comparison results specifically includes: When the analysis indicates that the position deviation exceeds the standard, the Kalman filter algorithm is called again to predict the movement trajectory of the bee colony in the next 30 seconds, and the target point of the interception path is adjusted according to the new trajectory. When the analysis indicates that the state deviation exceeds the standard, if it is a single interception unit failure, retrieve the performance data of other units in the same cluster and assign the task of the failure unit to the unit with the highest performance matching degree; if it is multiple unit failures, split or merge the related cluster structure. When the analysis shows that the performance deviation exceeds the standard, analyze the reasons for the deviation. If it is due to the number of bee colonies exceeding the expectation, increase the number of interception units for the clusters corresponding to high-priority bee colonies. If the interception fails due to excessive path risk, adjust the risk coefficient weight in the path planning and regenerate the path plan.

10. The attack path planning method based on the anti-unmanned surface vessel (USV) swarm dynamic clustering algorithm according to claim 9, characterized in that, The process of solving the multi-objective optimization model and then performing a fuzzy comprehensive evaluation of the optimal path set based on the solution results to obtain the optimal path specifically includes: Multiple random path schemes are generated as an initial population, and the process is iterated. During the iteration process, the middle segments of two path schemes are randomly selected for exchange, and the direction of the path segments is randomly adjusted to generate a new population; For each path scheme in the population, the fitness value is calculated according to the optimization objective. After the calculation is completed, the Pareto level is divided by non-dominated sorting, and the path schemes with higher levels are retained. The Pareto optimal path set is obtained by sorting the path schemes with higher congestion levels and iterating continuously on the path schemes with higher levels. The optimal path set is evaluated using fuzzy comprehensive analysis, and the path scheme with the highest score is taken as the final path planning result.

Images