An unmanned ship group self-organizing optimal task subgroup generation method suitable for near electromagnetic spectrum reconnaissance scene

Abstract

The application discloses a kind of unmanned ship group self-organizing optimal task subgroup generation methods suitable for near electromagnetic spectrum reconnaissance scene, including ground station randomly selects a node from unmanned ship group as initial candidate cluster head, sends tactical task and related parameters, calculates own weighted metric function value using heuristic algorithm, and sends result back to initial candidate cluster head node, selects the node with the maximum weighted metric function value as the candidate cluster head node of second round, carries out second round iteration, until candidate cluster head node is stable and unchanged, iteration ends;Cluster head node will decompose tactical task into individual task executable by each node according to task decomposition algorithm module.The application greatly reduces the computing burden of central control system, effectively solves the "dimension disaster" caused by the nonlinear growth of cluster size, can guarantee that the grouping of unmanned ship node is more targeted and effective, and can also improve the success rate and efficiency of task completion.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of cluster coordination electromagnetic spectrum confrontation, and particularly relates to a method for generating an optimal task sub-cluster of an unmanned ship cluster in a close electromagnetic spectrum reconnaissance scenario. BACKGROUND

[0002] In the field of unmanned cluster electromagnetic spectrum reconnaissance technology, according to the specific electromagnetic spectrum reconnaissance tactical task issued by the operational command, the unmanned cluster forms a corresponding task sub-cluster, which is an extremely important link in cluster coordination. The existing sub-cluster generation mechanism is mostly based on centralized or distributed topology structure, and adopts traditional precise programming for optimization to obtain the optimal sub-cluster. The generation mechanism under the centralized topology structure has poor robustness and cannot cope with unexpected situations, typical scenarios such as task re-planning, node failure and node being tracked by firepower. The distributed topology causes redundancy of the cluster network resources to some extent. The traditional precise programming will inevitably cause the "dimension disaster" problem with the upgrading of the cluster scale, and the optimization parameters of the current optimization algorithm do not consider the specific scene of electromagnetic spectrum reconnaissance, so the generalization ability is poor. Therefore, according to the specific tactical scene of electromagnetic spectrum reconnaissance, it is of great research value to design a generation mechanism of the task sub-cluster under a specific system topology architecture. SUMMARY

[0003] The purpose of the present application is to provide a method for generating an optimal task sub-cluster of an unmanned ship cluster in a close electromagnetic spectrum reconnaissance scenario.

[0004] Technical scheme: The present application comprises the following steps:

[0005] (1) The ground station randomly selects a node from the unmanned ship cluster as an initial candidate cluster head, and sends the tactical task and related parameters to the selected node. If the randomly selected node receives these instructions, it will send its coordinates and task parameters to other nodes;

[0006] (2) Other nodes will calculate their weighted metric function values according to the task parameters and their own information using heuristic algorithms, and send the results back to the initial candidate cluster head node. The initial candidate cluster head node will sort the weighted metric function values of each node, and select the node with the largest weighted metric function value as the candidate cluster head node for the second round, and then perform the second round of iteration. This process continues until the candidate cluster head node is stable and unchanged, and the iteration ends;

[0007] (3) The cluster head node will decompose the tactical task into individual tasks executable by each node according to the task decomposition algorithm module.

[0008] Further, the cluster head election process and task instruction transmission process in the unmanned ship group in step (1) comprises: the ground station initiates the task instruction by randomly selecting a node as an initial candidate cluster head, if the node receives the task instruction, it will send its coordinates and task parameters to other nodes, thereby forming a cluster head, which is responsible for coordinating the actions of other nodes to complete the task.

[0009] Further, the calculation process of step (2) is:

[0010] Assuming there are N nodes, the base station randomly selects a node as the first round candidate cluster head, and broadcasts its information as a cluster head and the number of other cluster head nodes to the cluster head node, and then the cluster head node announces itself as the cluster head of this round to other N-1 nodes, the communication content is the information of "becoming a cluster head" and the number of its own ID;

[0011] Calculate the clustering metric weight between itself and the cluster head, join the node with the largest clustering metric weight connected to itself, and send a response message and coordinate message to the corresponding cluster head, and the cluster head receives the clustering metric weight of the member, selects the member with the highest clustering metric weight as the next round candidate cluster head, and iterates in this way;

[0012] According to the clustering loss function, the iteration ends when the sum of the error squares of each sample distance from the cluster center point no longer decreases.

[0013] Further, the node is according to the clustering metric weight formula:

[0014] R=Q1d mi +Q2I+Q3S+Q4T+Q5E m

[0015] Where Q * is the weight coefficient, d mi is the distance between the node and the task area center, I is the node direction finding accuracy, S is the sensitivity, T is the health index, and E m is the current fuel of the cluster head.

[0016] Further, step (3) comprises: the cluster head node decomposes the tactical level task into atomic tasks executable by individual nodes according to the preset strategy, decomposes the task instruction based on timing, uses a timing diagram to depict tactics, outputs an atomic task sequence file, and initiates a tender within the cluster.

[0017] Further, the calculation process of step (3) is:

[0018] (3.1) According to the task content, the actual task scene is imagined, the corresponding task implementation rules and plan scope are made, the task purpose and environment planning route of the whole cluster are set according to the task needs, the mathematical model of the task track route and the working mode of all unmanned nodes in the whole task process are established, and the optimal task model is found by theoretical derivation;

[0019] (3.2) According to the mathematical model of the theory and the preset parameters of the task, the actual track route and the corresponding working mode of each unmanned node required in the actual task execution process are calculated;

[0020] (3.3) The actual track route and the corresponding working mode of each unmanned node are decomposed into corresponding cluster task sequences, and each atomic task in the sequence is initiated in the cluster.

[0021] Further, the step (3.1) uses the area diagonal coordinates to determine the search area, adopts a regional search method, that is, according to the number of unmanned boats, the larger target area is divided into a corresponding number of small areas, and then each unmanned boat is responsible for searching and path planning in the corresponding small area.

[0022] Further, the regional search method calculation process includes:

[0023] The diagonal coordinates of the rectangular search area are determined, the rectangular search range is determined, the number of search sub-areas is determined according to the number of unmanned boats and unmanned boat formations, and the unmanned boats are distributed to each sub-area for regional search;

[0024] The search radius of the unmanned boat and the unmanned boat is determined based on the sensitivity, and the searchable area range is determined.

[0025] Further, the searchable area range calculation method is:

[0026] The unmanned boat search radar detection area is circular, if the unmanned boat attitude angle change is not considered, the detection range of the unmanned boat is a circle with a radius of r,

[0027] The left upper corner coordinates (x L , y L ) and the right lower corner coordinates (x R , y R ) of the rectangular search area are determined, the number of unmanned boat formations is M and N respectively, the number of search sub-areas is determined, and the unmanned boats are distributed to each sub-area for regional search task execution by using a specific algorithm, and the length and width of the unmanned boat sub-area are L and W respectively;

[0028] If , the unmanned boat does not need to move, and can search all the area range by being fixed at the center of the sub-area; if Then the unmanned vehicle needs to plan the path according to the following rules, so that the whole area is covered.

[0029] Further, the minimum turning radius constraint of the unmanned vehicle is:

[0030] The minimum turning radius R is equal to the detection range radius r;

[0031] The minimum number of turns of the unmanned vehicle is The unmanned vehicle formation carries out linear-right turn-linear area search, and repeats it until the target area reconnaissance or monitoring task is completed.

[0032] When W=2nr, the unmanned vehicle can just scan the area, and there is no redundant area scanned.

[0033] When 2nr

[0034] Beneficial effects: compared with the prior art, the present application has the following significant advantages:

[0035] The present application is based on a decentralized wireless network architecture, driven by a collaborative electromagnetic spectrum counter-tactical task, and the task subgroup of the unmanned vehicle group is generated by self-organizing negotiation of a distributed intelligent decision-making software system; in the process of realizing the self-organizing action of the cluster, a distributed optimization strategy is used, which greatly reduces the computing burden of the central control system, effectively solves the "dimension disaster" caused by the nonlinear growth of the cluster size, and makes this mechanism have good control effect in large-scale unmanned node cluster tasks; in the process of realizing the heuristic algorithm of the cluster, through multi-level judgment procedures, the grouping of the unmanned vehicle nodes can be more targeted and effective, and the success rate and efficiency of task completion can be improved; in the self-organizing process, the distance from the task area center, sensitivity, direction-finding accuracy, remaining oil quantity, and health index are considered, which are combined and respectively given different weights, so that the optimization result reaches the optimal effect. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 It is a schematic diagram of an unmanned cluster system;

[0037] Figure 2 It is a schematic diagram of information flow generated by the task subgroup of the unmanned vehicle group. DETAILED DESCRIPTION

[0038] The technical solutions of the present application will be further described below in combination with the drawings.

[0039] The present application comprises the following steps:

[0040] Step 1: The ground station first randomly selects a node from the unmanned boat group as an initial candidate cluster head. The ground station sends the node a tactical task (such as area search) and related parameters (such as task area and task start time). If the randomly selected node receives these instructions, it will send its coordinates and task parameters to other nodes.

[0041] The cluster head election process and task instruction delivery process in the unmanned boat group are as follows: the ground station in the unmanned boat group initiates the task instruction by randomly selecting a node as an initial candidate cluster head. If the node receives the task instruction, it will send its coordinates and task parameters to other nodes, thereby forming a cluster head. This cluster head will be responsible for coordinating the actions of other nodes to complete the task. The method of randomly selecting nodes is an effective and fair intra-group random selection method to ensure the equal nature of each individual in the distributed structure.

[0042] Step 2: Other nodes will calculate their weighted metric function values using heuristic algorithms based on task parameters and their own information, and send the results back to the initial candidate cluster head node. The initial candidate cluster head node will sort the weighted metric function values of each node and select the node with the largest weighted metric function value as the candidate cluster head node for the second round, then proceed to the second round of iteration. This process will continue until the candidate cluster head node stabilizes and the iteration ends. In this step, each node calculates its weighted metric function value based on task parameters and its own information, and sends the result back to the initial candidate cluster head node. Then, the initial candidate cluster head node sorts the weighted metric function values of each node and selects the node with the largest weighted metric function value as the candidate cluster head node for the second round, then proceeds to the second round of iteration. This process will continue until the candidate cluster head node stabilizes and the iteration ends. This algorithm requires some prior parameters and information, such as task parameters and node information, which are included in the task instruction. The convergence speed of this algorithm may be affected by factors such as the number of nodes, initial value, and selection of weighted metric function, so these factors must be reasonably set and adjusted.

[0043] Step 3: The final cluster head node will decompose the tactical task into individual tasks executable by each node according to the task decomposition algorithm module. This task decomposition algorithm follows a set of rules, which vary depending on the tactical task of electromagnetic spectrum reconnaissance. Taking an area search task as an example, the specific steps are as follows: First, the algorithm uses the diagonal coordinates of the area to determine the search area and uses the total number of unmanned surface vessels (USVs) to divide it into sub-regions, thus achieving comprehensive search of the entire area. Taking a rectangular search area as an example, the algorithm adopts a sub-regional search method. Specifically, based on the number of USVs, a larger target area is divided into a corresponding number of smaller areas, and each USV is responsible for searching and planning the path within its respective sub-region. Therefore, this algorithm is a distributed collaborative decision-making algorithm that can help multiple USVs collaborate to complete search tasks, improving search efficiency and accuracy.

[0044] like Figure 2 As shown, in order to achieve the above design, the technical solution provided by the present invention is as follows:

[0045] (1) As Figure 1 The 002 unmanned cluster nodes are based on a decentralized wireless communication network architecture, and a distributed auxiliary decision-making system is deployed inside each unmanned vessel node.

[0046] (2) Figure 1 As shown, the 001 combat command issues an area search task, randomly selecting 002. The command triggers a dynamic cluster head formation algorithm, and 002 generates a cluster head node through dynamic negotiation within the cluster. The negotiation mechanism, using a clustering algorithm as an example, is implemented as follows: Assuming there are N nodes, the base station first randomly selects a node as the first round of candidate cluster head and broadcasts its status as the cluster head, along with the IDs of other cluster head nodes, to the cluster head node. Then, the cluster head node announces itself as the cluster head for this round to the other N-1 nodes, communicating with the information "becoming a cluster head" and its own ID number. Other nodes then proceed according to the clustering metric weight formula:

[0047] R = Q1d mi +Q2I+Q3S+Q4T+Q5E m

[0048] Calculate the clustering metric weight between itself and the cluster head, add the node with the largest clustering metric weight on its adjacent edge, and send a response message and coordinate message to the corresponding cluster head, where Q... * d represents the weighting coefficient. mi I is the distance between the node and the center of the mission area, S is the sensitivity, T is the health index, and E is the distance between the node and the center of the mission area. m This indicates the current oil level in the cluster head.

[0049] After the cluster head receives the clustering metric weights of the members, it selects the member with the highest clustering metric weight as the candidate cluster head for the next round, and iterates in this way; finally, the iteration ends when the clustering loss function, i.e., the sum of the squared errors of each sample from the center point of the cluster, no longer decreases.

[0050] (3) The cluster head node decomposes the tactical level task into atomic tasks executable by individual nodes according to a preset strategy, uses a timing diagram to depict the tactics by decomposing the task instructions based on timing, outputs an atomic task sequence file, and initiates a bidding within the cluster. The specific implementation is as follows: first, according to the task content, the actual task scenario is imagined, the corresponding task implementation rules and plan range are formulated, the task purpose and environmental planning route of the overall cluster are set according to the task needs, the mathematical model of the task track route and working mode of all unmanned nodes in the entire task process is established, the optimal task model is found by theoretical derivation, and the actual track route and corresponding working mode of each unmanned node required in the actual task execution process are calculated according to the mathematical model of the theory and the preset parameters of the task. Finally, as shown in Table 1, the corresponding cluster task sequence is decomposed, and each atomic task in the sequence is initiated for bidding within the cluster.

[0051] Table 1: Atomic task sequence file format

[0052]

[0053] Taking regional search as an example, first, the algorithm uses the diagonal coordinates of the region to determine the search region, and uses the total number of unmanned boats to divide the number of sub-regions, thereby realizing the coverage search of the entire region.

[0054] The algorithm adopts a regional search method. Specifically, according to the number of unmanned boats, a larger target region is divided into a corresponding number of small regions, and then each unmanned boat is responsible for searching and path planning in the corresponding small region. Therefore, the algorithm is a distributed cooperative decision-making algorithm, which can help multiple unmanned boats to cooperatively complete the search task and improve the search efficiency and accuracy.

[0055] The actual method is as follows:

[0056] (3.1) The diagonal coordinates of the rectangular search region are known, and the rectangular search range is determined. The number of search sub-regions is determined according to the number of unmanned boats and unmanned boat formations, and the unmanned boats are allocated to each sub-region for regional search.

[0057] (3.2) According to step (3.1), the search radius of the unmanned boat is determined based on the sensitivity of the unmanned boat, and then the range of the searchable region is determined.

[0058] The detection area of the unmanned vehicle search radar is circular, and if the unmanned vehicle attitude angle is not considered, the detection range of the unmanned vehicle is a circle with a radius of r.

[0059] The rectangular search area is known to have upper left corner coordinates (x L , y L ) and lower right corner coordinates (x R , y R ), and the rectangular search range is determined. The number of unmanned vehicle formations is M and N, respectively, and the number of search sub-areas (MN search sub-areas) is determined, and the unmanned vehicle is allocated to each sub-area using a specific algorithm to perform the area search task. The length and width of the unmanned vehicle sub-area are L and W, respectively.

[0060] If , the unmanned vehicle does not need to move and can search all the areas by being fixed at the center of the sub-area; if , the unmanned vehicle needs to be path planned according to the following rules to cover the entire area.

[0061] To solve the problem of the minimum turning radius constraint of the unmanned vehicle, researchers have proposed many different algorithms, such as serial methods and fusion methods. In the serial method, the unmanned vehicle will advance along the circular path at the turning point until it reaches the next parallel line. In the fusion method, the search paths on multiple parallel lines are fused together to form a continuous search path.

[0062] Therefore, the minimum turning radius R is equal to the detection range radius r. As shown in the figure, the left side of the search boundary is the inside of the search area. Therefore, the minimum number of turns of the unmanned vehicle is

[0063] The unmanned vehicle formation performs straight-line-right-turn-straight-line area search, and this is repeated until the target area reconnaissance or monitoring task is completed. When W = 2nr, the unmanned vehicle can scan the area exactly, and there is no extra area scanned; when 2nr < W < 2(n+1)r, considering the turning process without gaps, there is an area of S = [2(n+1)r-W]·L that is scanned redundantly.

[0064] The area search algorithm can achieve area scanning of the unmanned vehicle group in different areas by changing the range of the area search and the radius of the unmanned vehicle search.

Claims

1. A method for generating self-organized optimal task subgroups of unmanned surface vessel swarms suitable for close-range electromagnetic spectrum reconnaissance scenarios, characterized in that, The method includes the following steps: (1) The ground station randomly selects a node from the unmanned surface vessel swarm as the initial candidate cluster head. The ground station sends tactical tasks and related parameters to the selected node. If the randomly selected node receives these instructions, it sends its coordinates and task parameters to other nodes. (2) Other nodes will use heuristic algorithms to calculate their own weighted metric function values ​​based on task parameters and their own information, and send the results back to the initial candidate cluster head node. The initial candidate cluster head node sorts the weighted metric function values ​​of each node and selects the node with the largest weighted metric function value as the candidate cluster head node for the second round. Then, the second round of iteration is carried out. This process continues until the candidate cluster head node is stable and the iteration ends. The calculation process is as follows: Assuming there are N nodes, the base station randomly selects a node as the first round of candidate cluster head and broadcasts the information that it has become the cluster head, as well as the ID numbers of other cluster head nodes, to the cluster head node. Then, the cluster head node announces to the other N-1 nodes that it has become the cluster head in this round, and the communication content is the information of "becoming the cluster head" and its own ID number. Calculate the clustering metric weight between itself and the cluster head, add the node with the largest clustering metric weight on the edge connected to itself, and send a response message and coordinate message to the corresponding cluster head. After receiving the clustering metric weight of the member, the cluster head selects the member with the highest clustering metric weight as the candidate cluster head for the next round, and so on iteratively. The iteration ends when the sum of squared errors between each sample and the center of its cluster no longer decreases, according to the clustering loss function. The formula for the node clustering metric weight is expressed as: ， in These are the weighting coefficients. It is the distance between the node and the center of the task area. It refers to the accuracy of nodal orientation finding. It's about sensitivity. It's a health index. This is the current oil level in the cluster head; (3) The cluster head node will decompose the tactical task into individual tasks that can be executed by each node according to the task decomposition algorithm module; This includes the cluster head node decomposing tactical-level tasks into atomic tasks that can be executed by individual nodes according to a pre-defined strategy, using time-based decomposition of task instructions, using time sequence diagrams to characterize tactics, outputting atomic task sequence files, and initiating bidding within the cluster.

2. The method for generating self-organized optimal task subgroups of unmanned surface vessels in close-range electromagnetic spectrum reconnaissance scenarios according to claim 1, characterized in that, The cluster head selection process and command transmission process in step (1) of the unmanned surface vessel swarm include: the ground station randomly selects a node as the initial candidate cluster head to initiate the mission command. If the node receives the mission command, it will send its coordinates and mission parameters to other nodes to form a cluster head. This cluster head is responsible for coordinating the actions of other nodes to complete the mission.

3. The method for generating self-organized optimal task subgroups of unmanned surface vessels in close-range electromagnetic spectrum reconnaissance scenarios according to claim 1, characterized in that, The specific process of step (3) is as follows: (3.1) Based on the task content, imagine the actual task scenario, formulate the corresponding task implementation rules and planning scope, set the overall cluster's task objectives and environmental planning routes according to the task requirements, establish a mathematical model of the task trajectory routes and working modes that all unmanned nodes should have during the entire task process, and find the optimal task model through theoretical derivation. (3.2) Based on the theoretical mathematical model and the preset parameters of the task, calculate the actual flight path and corresponding working mode of each unmanned node required during the actual task execution. (3.3) Decompose the actual flight path and corresponding working mode of each unmanned node into a corresponding cluster task sequence, and initiate bidding for each atomic task in the sequence within the cluster.

4. The method for generating self-organized optimal task subgroups of unmanned surface vessel swarms in close-range electromagnetic spectrum reconnaissance scenarios according to claim 3, characterized in that, Step (3.1) uses the diagonal coordinates of the region to determine the search area and adopts a sub-regional search method. That is, according to the number of unmanned surface vessels, the larger target area is divided into a corresponding number of small areas, and then each unmanned surface vessel is responsible for searching and planning the path for the corresponding small area.

5. The method for generating self-organized optimal task subgroups of unmanned surface vessels in close-range electromagnetic spectrum reconnaissance scenarios according to claim 4, characterized in that, The calculation process for the region-based search method includes: Given the diagonal coordinates of the rectangular search area, determine the rectangular search range. Given the number of unmanned surface vessels (USVs) and the number of USVs in the formation, determine the number of search sub-regions and assign USVs to each sub-region for area search. The search radius of an unmanned surface vessel (USV) is determined based on its sensitivity, thereby determining the range of areas it can search.

6. The method for generating self-organized optimal task subgroups of unmanned surface vessels in close-range electromagnetic spectrum reconnaissance scenarios according to claim 5, characterized in that, The method for calculating the searchable area is as follows: The detection area of ​​the unmanned surface vessel (USV) search radar is circular. If we disregard changes in the USV's attitude angle, the USV's detection range is a circle with a radius of [missing information]. The circle, The coordinates of the top left corner of the rectangular search area and the coordinates of the bottom right corner Determine the rectangular search area, and let the number of unmanned surface vessels in the formation be respectively... The number of search sub-regions is determined, and the unmanned surface vessel (USV) is assigned to each sub-region using a specific algorithm to perform the region search task. Let the length and width of each USV sub-region be... and ; like Then the unmanned surface vessel does not need to move; it can search the entire area from its fixed position in the center of the sub-region. If Then the unmanned surface vessel needs to plan its path according to the following rules to ensure that the entire area is covered.

7. The method for generating self-organized optimal task subgroups of unmanned surface vessel swarms in close-range electromagnetic spectrum reconnaissance scenarios according to claim 6, characterized in that, The minimum turning radius constraint for unmanned surface vessels is: The minimum turning radius R is equal to the detection range radius r; The minimum number of turns for an unmanned surface vessel is The unmanned surface vessel formation conducts a straight-turn-right-straight-line area search, repeating this process until the reconnaissance or monitoring mission of the target area is completed; when At that time, the unmanned surface vessel was able to scan the area precisely, without any extra areas being scanned; when When considering a turning process without gaps, there is The area was scanned redundantly.

Images