Task-driven unmanned aerial vehicle ad hoc network clustering method

Through the task-driven drone ad hoc network clustering method, the UTMOPSO algorithm is used to optimize the resource configuration of the drone cluster, which solves the shortage of resource requirements and task completion time in the existing technology, and realizes efficient and flexible task execution and Resource utilization.

CN119052886APending Publication Date: 2024-11-29EAST CHINA INST OF COMPUTING TECH

Patent Information

Application Number
CN202411092567.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-09
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

The existing drone ad hoc networking clustering method cannot adjust resource requirements and task completion time according to actual task scenarios, and the optimization parameters are limited and cannot fully meet the task execution needs.

Method used

A task-driven self-networking clustering method is proposed. The task cluster is formed based on task information through UTMOPSO algorithm, and the energy, task arrival time and resource configuration in the cluster are optimized to ensure maximum resource utilization in the cluster and adapt to different Requirements for task scenarios.

Benefits of technology

The resource optimization configuration of the drone cluster is realized, the task execution efficiency and system flexibility are improved, and the working mode of the drone in the cluster can be adjusted according to the urgency and priority of the task, and adapted to different task scenarios and environmental changes.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119052886A_ABST
    Figure CN119052886A_ABST
Patent Text Reader

Abstract

The invention relates to a task-driven unmanned aerial vehicle ad hoc network clustering method, which comprises the following steps that: an unmanned aerial vehicle cluster initializes nodes, distributes tasks according to task location information, and enters a task cluster establishment stage or a region search stage; the unmanned aerial vehicle cluster is clustered according to pre-input task location information and is divided into a plurality of reconnaissance clusters, and the reconnaissance clusters go to a plurality of task locations for reconnaissance to find a task target; after target information is obtained, the unmanned aerial vehicle node enters a task cluster establishment stage, and a task cluster is established by using a UTMOPSO algorithm; member nodes in the cluster calculate own weights and exchange weight information in the network; and after the task cluster is established, entering a task execution stage. The problem that the resource demand and the task completion time cannot be adjusted according to the actual task scene in the unmanned aerial vehicle ad hoc network clustering networking method is solved, the resource optimization configuration of the unmanned aerial vehicle cluster is realized, the resource demand and the task completion time requirement under different task scenes are met, and the efficiency and the flexibility of the unmanned aerial vehicle ad hoc network are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to unmanned aerial vehicle (UAV) technology, and in particular to a method for generating clusters of unmanned aerial vehicle (UAV) clusters. Background Art

[0002] As the scale of drone clusters expands, the increase in the number of nodes poses a huge challenge to the communication and management of drone self-organizing networks. In actual work, drones performing the same task need to communicate frequently in real time, while the frequency of information exchange between drones performing different tasks is low, and the locations of different tasks may be far apart, and the needs they face are also different. In a space, drone clusters perform tasks around different target locations, so the distribution is not uniform. Therefore, dividing a large number of drone clusters into different drone cluster structures can not only reduce the communication interference between drone nodes of different tasks, but also help ground personnel to centrally manage and assign tasks to drone clusters. Clustering algorithms are the key part of achieving collaboration, collaborative work and task execution within the cluster. Clustering algorithms are used to group drones within a cluster into different clusters or groups to achieve task division and data processing synergy. By reasonably designing clustering algorithms, the collaborative working mode in the cluster can be optimized, the task execution efficiency can be improved, conflicts and collisions can be reduced, and the overall performance can be improved.

[0003] Faced with increasingly complex mission scenarios, traditional clustering algorithms such as LEACH and WCA are difficult to meet the needs of missions within the scenario. Existing clustering algorithms are often used to reduce the energy consumption and communication overhead of drone ad hoc networks. The node with the highest weight is obtained based on the current node battery energy, node relevance, node mobility and node distance of each node as the cluster head, and the nodes within its effective communication range are members of the cluster. However, the above algorithms only guarantee the communication and energy stability of drone ad hoc networks, while ignoring the constraints of the mission scenario on the resources, quantity and execution time of the mission drones.

[0004] Patent publication number CN114885379A discloses a large-scale UAV cluster adaptive clustering networking method. It includes the following steps: Step 1, according to the communication radius of the UAV and the deployment location of the cluster, a communication connectivity matrix is ​​constructed to calculate the communication connectivity and communication connectivity node set of each UAV; Step 2, using the communication connectivity as an evaluation index, a preliminary cluster head node is selected through a cyclic iteration method to initialize the clustering of the UAV cluster; Step 3, the communication load of the cluster head node in the initial clustering scheme is calculated, and the clustering scheme is load balanced and optimized according to the communication load capacity of a single UAV; Step 4, the communication energy consumption between the sub-nodes in the cluster and the cluster head node is calculated according to the communication distance, and the communication energy consumption of the clustering scheme obtained in step 3 is optimized to obtain the final clustering networking scheme. This patent mainly solves the problems of uneven load and energy consumption of UAV clusters, but does not consider whether the UAV carries other resources, and cannot fully schedule and utilize the resources of the UAV cluster.

[0005] Patent publication number CN110312292A discloses a method for dynamic weighted cluster head election for unmanned aerial vehicle self-organizing networks. The patent first numbers all drones, and obtains the current node battery energy, node relevance, node mobility, and node distance sum of each node, and then obtains the weight of each node. Then, the node with the largest weight is selected as the first cluster head, and the nodes within its effective communication distance are cluster members. Except for the cluster head and cluster members of the first cluster, the remaining nodes again elect the largest weight as the cluster head of the second cluster, and the nodes within the effective communication distance of the cluster head are cluster members, and so on, until all nodes are clustered. However, this patent only considers the energy and mobility indicators of drone nodes when clustering, and the weight value of each parameter depends on empirical parameters, so the drone cluster obtained cannot well meet the execution requirements of related tasks.

[0006] It can be seen that the existing UAV self-organizing network clustering method only focuses on the life cycle of the network, but does not fully consider whether the UAV cluster can meet the resource requirements of the task and the task completion time. In addition, the optimization parameters are limited and cannot be adjusted according to the actual mission scenario. Summary of the invention

[0007] Aiming at the problem that the existing UAV self-organizing network clustering method cannot adjust the resource demand and task completion time according to the actual mission scenario, a task-driven UAV self-organizing network clustering method is proposed. This method forms corresponding UAV task clusters according to the task information in the scenario, and achieves the overall optimization in terms of average energy within the cluster, task cluster arrival time, and task cluster resources, thereby achieving optimal resource allocation of UAV clusters to meet the resource demand and task completion time requirements in different mission scenarios, thereby improving the efficiency and flexibility of UAV self-organizing networks.

[0008] The technical solution of the present invention is:

[0009] A task-driven UAV self-organizing network clustering method comprises the following steps:

[0010] Step 1: Initialize the nodes of the drone cluster and assign tasks according to the mission location information. If there is mission target information in advance, go to step 3 to enter the mission cluster formation stage; if there is no mission information, go to step 2 to enter the regional search stage;

[0011] Step 2: The drone cluster is clustered according to the pre-input mission location information, and divided into multiple reconnaissance clusters. The reconnaissance clusters go to multiple mission locations to conduct reconnaissance and search for mission targets. If the mission target is found, go to step 3. If not, repeat the reconnaissance until the mission target is found.

[0012] Step 3: When a UAV finds the mission target and obtains the target information, the UAV node enters the mission cluster formation stage, and the node is transformed into a temporary mission cluster head. It sends a mission cluster formation application in the original cluster. After receiving the application, other non-task cluster nodes will send a joining request to the temporary cluster head. The request message contains the node's own resources and location information; the temporary cluster head performs task clustering according to the received node information, selects suitable cluster members, and sends a success reply to them, and sends a rejection reply to the remaining unselected nodes; if the overall resources in the reconnaissance cluster cannot meet the mission requirements, a cross-cluster request is made, and a formation application is sent to other reconnaissance clusters. After receiving the information, other clusters use the UTMOPSO algorithm to form a task cluster;

[0013] Step 4: The member nodes in the cluster calculate their own weights and exchange weight information within the network; when the drone detects a node with a larger weight than its own recorded weight, it records the weight value, sets the node as the cluster head and forwards the cluster head information; after the network is stable, the node with the largest weight in the network is the cluster head;

[0014] Step 5: After the task cluster is formed, it enters the task execution phase. The task cluster executes the task and then sends it to the ground station or other clusters based on the completion result. If the task fails, it will be handed over to other clusters to continue to complete the task. If the task is successful, the task cluster will be transformed into a reconnaissance cluster and re-enter the task area search phase, repeating step 2. When the drone cluster finds a new target, it enters step 3 until the drone cluster completes all tasks in the area.

[0015] Furthermore, the task cluster of step 3 is formed as follows:

[0016] Step 3.1: Initialize parameters; set the maximum number of loops MaxIter, the number of particles p pa, establish a mapping table for drone ID based on the collected drone information, and u Determine the particle dimension dim so that N u = dim; Initialize the position p of particle pa pa and speed v pa , particle pa(0≤pa≤p pa ) The jth (0≤j≤dim) dimension is 1, indicating that the drone with mapping ID j joins the cluster;

[0017] Step 3.2: Calculate the dominance of the particle set; calculate the average residual energy and free resources of the two optimization objectives according to the initial particle position and the fitness function; calculate the dominance relationship between particles according to the value; extract the solution set that does not dominate each other between particles and put it into the storage library, which stores the particle information of all non-dominated solutions;

[0018] Step 3.3: loop; number of loops Iter = Iter + 1;

[0019] Step 3.4: Update the weight factor w;

[0020]

[0021] Among them, Wmax and Wmin are the maximum weight and minimum weight experience values;

[0022] Step 3.5: Use the roulette wheel selection algorithm to randomly select a Pareto optimal solution in the repository as the global optimal solution gbest, which is used to update the particle speed;

[0023] Step 3.6: According to the historical optimal solution pbest of particle pa pa , that is, the position with the smallest fitness value among all the positions experienced by particle pa, and the global optimal solution gbest update the particle speed and map the particle speed; update the particle speed:

[0024] v pa =w*v pa +c1*rand(n)*(pbest pa -p pa )+c2*rand(n)*(gbest-p pa )

[0025] Among them, c1 and c2 are acceleration coefficients, and their values ​​are usually [1.5, 2];

[0026] The particle velocity v pa Mapping to the new velocity v′ pa :

[0027]

[0028] Step 3.7: Update particle position p′ pa , and perform random mutation; if v′ pa (j)>σ, then the particle position is 1, otherwise the particle position in this dimension is 0, where σ is the mapping threshold, and its value is;

[0029]

[0030] Calculate the mutation factor p according to the number of iterations Iter, and select particles for bit mutation according to p;

[0031]

[0032] Among them, mu is the variation factor;

[0033] Step 3.8: Calculate the fitness value FT, calculate the optimal solution, which is the minimum value of FT, and update the particle's historical optimal solution pbest according to the dominance relationship pa ;

[0034] Step 3.9: Update the repository Rep;

[0035] Step 3.10: If the number of iterations Iter is less than the maximum number of iterations MaxIter, continue iterating and jump to step 3.1, otherwise jump to step 3.11;

[0036] Step 3.11: Based on the Pareto solutions of the repository, calculate the arrival time of each solution to the task target, and select the solution member with the minimum arrival time as the cluster member.

[0037] Furthermore, the calculation method of the cluster members in step 3.11 is as follows:

[0038] The total number of drones in the entire network is set to N u , the number of known tasks is N t , the set of tasks is The set of clusters in the network is V c , assuming that there are N c clusters, the number of drones in the mth cluster is The set of drones in the mth cluster is but This method uses a The matrix to record the node's ownership, x i,m =1 means that node i, i.e. drone i, joins cluster m;

[0039] In a heterogeneous drone cluster, each drone carries different resources. Assume that the drone cluster contains N qresources, then the resource vector of a single UAV i is The number of resource types required for the same task is also N q The resource requirement vector of the e-th task is Drone i location The location of the mission objective e Therefore, the constraints mainly include:

[0040] Constraint 1: For pending tasks Corresponding task cluster The mission conditions must be met to complete the mission, so each resource of the drone in the cluster The sum must be greater than the task's resource requirement vector:

[0041]

[0042] Constraint 2: When there are multiple tasks that require the formation of multiple clusters, the same drone i can only belong to one cluster, that is, after multiple cluster head nodes initiate cluster formation applications, drone i can only choose one cluster to join; therefore, the constraints are:

[0043]

[0044] Constraint 3: In order to ensure the completion rate of the task, the energy E of the members in the cluster should be equal to i Maximization is calculated using the average energy within the cluster:

[0045]

[0046] For the convenience of calculation, the average energy can be taken as a negative value, and the maximization problem can be transformed into a minimization problem:

[0047]

[0048] Constraint 4: Since the number of resources in the drone network is limited, in order to ensure that the number of tasks to be executed is sufficient, when the drones form a cluster, the amount of excess resources should be minimized while satisfying the above equation, as shown in the following equation:

[0049]

[0050] Constraint 5: In order to assume that the time for a drone cluster to complete a task is fixed, the key factor that determines the time to complete a task is the arrival time of the drone cluster. The arrival time of the drone cluster depends on the position information of the shortest drone. Therefore, the farthest distance from the members of the drone cluster to the mission target determines the arrival time of the drone cluster. It is expected that the arrival time of the drone cluster is the minimum, so:

[0051]

[0052] In summary, the UTMOPSO fitness function and constraints are as follows:

[0053]

[0054] Under the multi-constraint conditions of the above formula, find the cluster member corresponding to the minimum fitness function value minFT.

[0055] The beneficial effects of the present invention are:

[0056] 1) The task-driven UAV self-organizing network clustering method of the present invention can dynamically adjust the distribution and configuration of clusters according to the specific requirements of the task, so that the UAV cluster can more effectively meet the resource requirements and task completion time requirements of the task. This can optimize the execution process of the task and improve the completion efficiency and quality of the task.

[0057] 2) Through the adaptive scheduling algorithm, the present invention realizes the dynamic allocation and scheduling of the drone cluster to ensure the maximum resource utilization of the drones in the cluster. This can make full use of the resources of the drone cluster and improve the overall performance and efficiency of the network.

[0058] 3) The present invention can flexibly adjust the working mode and task execution strategy of the UAVs in the cluster according to factors such as the urgency and priority of the task. This can improve the adaptability and flexibility of the system, so that the UAV self-organizing network can better adapt to different mission scenarios and environmental changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a task cluster formation flow chart of the unmanned aerial vehicle self-organizing network clustering method of the present invention;

[0060] Figure 2 The present invention is a flow chart of cluster member recommendation based on task information for the unmanned aerial vehicle self-organizing network clustering method. DETAILED DESCRIPTION

[0061] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0062] The purpose of the present invention is to propose a task-driven UAV self-organizing network clustering method to achieve optimal resource allocation of UAV clusters and effective execution of tasks, thereby solving the limitations of existing UAV self-organizing network clustering methods and providing a more flexible and efficient network organization method for the application of UAV self-organizing networks in different mission scenarios;

[0063] Compared with clusters obtained by weighted calculation, the present invention can simultaneously optimize multiple conflicting objectives, such as the average energy of the cluster, the arrival time, and the number of resources carried by the node. This makes it possible to better balance the relationship between various indicators during cluster formation and node allocation, thereby improving the performance and efficiency of the system.

[0064] refer to Figure 1 As shown, Figure 1 This is a flow chart of task cluster formation of the present invention. This embodiment provides a process of the present invention, and its specific steps include:

[0065] Step 1: Initialize the nodes of the drone cluster and assign tasks according to the mission location information. If there is mission target information in advance, go to step 3 to enter the mission cluster formation stage; if there is no mission information, go to step 2 to enter the regional search stage;

[0066] Step 2: The drone cluster is divided into multiple reconnaissance clusters according to the pre-entered mission location information. The reconnaissance clusters go to multiple mission locations to conduct reconnaissance and search for mission targets. If the mission target is found, go to step 3. If not, repeat the reconnaissance until the mission target is found.

[0067] Step 3: When a drone finds the mission target and obtains the target information, the drone node enters the mission cluster formation stage. The node is transformed into a temporary mission cluster head and sends a mission cluster formation application in the original cluster. After receiving the application, other non-task cluster nodes will send a joining request to the temporary cluster head. The request message contains the node's own resources and location information. The temporary cluster head enters step 3.1 to cluster the task according to the received node information, selects suitable cluster members, and sends a success reply to them, and sends a rejection reply to the remaining unselected nodes. If the overall resources in the reconnaissance cluster cannot meet the needs of the task, a cross-cluster request is made and a formation application is sent to other reconnaissance clusters. After receiving the information, other clusters use the UTMOPSO (Unmanned Aerial Vehicle Task-based Multi-Objective Particle SwarmOptimization) algorithm to form a task cluster.

[0068] Step 4: The member nodes in the cluster calculate their own weights and exchange weight information within the network. When the drone detects a node with a larger weight than its own recorded weight, it records the weight value, sets the node as the cluster head and forwards the cluster head information. After the network is stable, the node with the largest weight in the network is the cluster head.

[0069] Step 5: After the task cluster is formed, it enters the task execution phase. The task cluster executes the task and then sends it to the ground station or other clusters based on the completion result. If the task fails, it will be handed over to other clusters to continue to complete the task. If the task succeeds, the task cluster will be transferred to the reconnaissance cluster and re-enter the task area search phase, repeating step 2. When the drone cluster finds a new target, it enters step 3 until the drone cluster completes all tasks in the area.

[0070] refer to Figure 2 As shown, Figure 2 This is a flow chart of cluster member recommendation based on task information in the present invention. This process is included in the task cluster formation of the embodiment, and the specific technical solution is as follows: Steps 3.1-3.11:

[0071] Step 3.1: Initialize parameters; set the maximum number of loops MaxIter, the number of particles p pa , establish a mapping table for drone ID based on the collected drone information, and u Determine the particle dimension dim so that N u = dim; Initialize the position p of particle pa pa and speed v pa , particle pa(0≤pa≤p pa ) The jth (0≤j≤dim) dimension is 1, indicating that the drone with mapping ID j joins the cluster;

[0072] Step 3.2: Calculate the dominance of the particle set; calculate the average residual energy and free resources of the two optimization objectives according to the initial particle position and the fitness function; calculate the dominance relationship between particles according to the value; extract the solution set that does not dominate each other between particles and put it into the storage library, which stores the particle information of all non-dominated solutions;

[0073] Step 3.3: loop; number of loops Iter = Iter + 1;

[0074] Step 3.4: Update the weight factor w;

[0075]

[0076] Among them, Wmax and Wmin are the maximum weight and minimum weight experience values;

[0077] Step 3.5: Use the roulette wheel selection algorithm to randomly select a Pareto optimal solution in the repository as the global optimal solution gbest, which is used to update the particle speed;

[0078] Step 3.6: According to the historical optimal solution pbest of particle pa pa , that is, the position with the smallest fitness value among all the positions experienced by particle pa, and the global optimal solution gbest update the particle speed and map the particle speed; update the particle speed:

[0079] v pa =w*v pa +c1*rand(n)*(pbest pa -p pa )+c2*rand(n)*(gbest-p pa )

[0080] Among them, c1 and c2 are acceleration coefficients, and their values ​​are usually [1.5, 2];

[0081] The particle velocity v pa Mapping to the new velocity v′ pa :

[0082]

[0083] Step 3.7: Update particle position p′ pa , and perform random mutation; if v′ pa (j)>σ, then the particle position is 1, otherwise the particle position in this dimension is 0, where σ is the mapping threshold, and its value is;

[0084]

[0085] Calculate the mutation factor p according to the number of iterations Iter, and select particles for bit mutation according to p;

[0086]

[0087] Among them, mu is the variation factor;

[0088] Step 3.8: Calculate the fitness value FT, calculate the optimal solution, which is the minimum value of FT, and update the particle's historical optimal solution pbest according to the dominance relationship pa ;

[0089] Step 3.9: Update the repository Rep;

[0090] Step 3.10: If the number of iterations Iter is less than the maximum number of iterations MaxIter, continue iterating and jump to step 3.1, otherwise jump to step 3.11;

[0091] Step 3.11: Based on the Pareto solutions of the repository, calculate the arrival time of each solution to the task target, and select the solution member with the minimum arrival time as the cluster member.

[0092] The calculation method of cluster members in step 3.11 is as follows:

[0093] The total number of drones in the entire network is set to N u , the known number of tasks is Nt, and the set of tasks is The set of clusters in the network is V c , assuming that there are N c clusters, the number of drones in the mth cluster is The set of drones in the mth cluster is but The present invention uses a The matrix to record the node's ownership, x i,m =1 indicates that node (i.e., drone) i joins cluster m.

[0094] In a heterogeneous drone cluster, each drone carries different resources. Assume that the drone cluster contains N q resources, then the resource vector of a single UAV i is The number of resource types required for the same task is also N q The resource requirement vector of the e-th task is Drone i location The location of the mission objective e Therefore, the constraints mainly include:

[0095] (1) For pending tasks Corresponding task cluster The mission conditions must be met to complete the mission, so each resource of the drone in the cluster The sum must be greater than the task's resource requirement vector:

[0096]

[0097] (2) When there are multiple tasks that require the formation of multiple clusters, the same UAV i can only belong to one cluster, that is, after multiple cluster head nodes initiate cluster formation applications, UAV i can only choose one cluster to join. Therefore, the constraints are:

[0098]

[0099] (3) In order to ensure the completion rate of the task, we hope to make the energy E of the members in the cluster equal toi To maximize, we use the average energy within the cluster to calculate:

[0100]

[0101] For the convenience of calculation, we can take a negative value for the average energy and transform the maximization problem into a minimization problem:

[0102]

[0103] (4) Since the number of resources in the UAV network is limited, in order to ensure that the number of tasks to be executed is sufficient, we hope that when the UAVs form a cluster, the amount of excess resources can be minimized while satisfying the above equation, as shown in the following equation:

[0104]

[0105] (5) In order to assume that the time it takes for a drone cluster to complete a task is fixed, the key factor that determines the time it takes to complete a task is the arrival time of the drone cluster. The arrival time of the drone cluster depends on the location information of the shortest drone. Therefore, the farthest distance from the members of the drone cluster to the mission target determines the arrival time of the drone cluster. We expect the arrival time of the drone cluster to be the shortest, so we have:

[0106]

[0107] In summary, the UTMOPSO fitness function and constraints are as follows:

[0108]

[0109] The purpose of this algorithm is to find the cluster member that satisfies the minimum fitness function value minFT under the multi-constraint conditions of the above formula.

[0110] The above-mentioned embodiment only expresses one implementation mode of the present invention, and its description is relatively specific and detailed, but it cannot be understood as limiting the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be based on the attached claims.

Claims

1. A task-driven UAV self-organizing network clustering method, characterized in that: The following steps are involved: Step 1: Initialize the nodes of the drone cluster and assign tasks according to the mission location information. If there is mission target information in advance, go to step 3 to enter the mission cluster formation stage; if there is no mission information, go to step 2 to enter the regional search stage; Step 2: The drone cluster is clustered according to the pre-input mission location information, and divided into multiple reconnaissance clusters. The reconnaissance clusters go to multiple mission locations to conduct reconnaissance and search for mission targets. If the mission target is found, go to step 3. If not, repeat the reconnaissance until the mission target is found. Step 3: When a UAV finds the mission target and obtains the target information, the UAV node enters the mission cluster formation stage, and the node is transformed into a temporary mission cluster head. It sends a mission cluster formation application in the original cluster. After receiving the application, other non-task cluster nodes will send a joining request to the temporary cluster head. The request message contains the node's own resources and location information; the temporary cluster head performs task clustering according to the received node information, selects suitable cluster members, and sends a success reply to them, and sends a rejection reply to the remaining unselected nodes; if the overall resources in the reconnaissance cluster cannot meet the mission requirements, a cross-cluster request is made, and a formation application is sent to other reconnaissance clusters. After receiving the information, other clusters use the UTMOPSO algorithm to form a task cluster; Step 4: The member nodes in the cluster calculate their own weights and exchange weight information within the network; when the drone detects a node with a larger weight than its own recorded weight, it records the weight value, sets the node as the cluster head and forwards the cluster head information; after the network is stable, the node with the largest weight in the network is the cluster head; Step 5: After the task cluster is formed, it enters the task execution phase. The task cluster executes the task and then sends it to the ground station or other clusters based on the completion result. If the task fails, it will be handed over to other clusters to continue to complete the task. If the task is successful, the task cluster will be transformed into a reconnaissance cluster and re-enter the task area search phase, repeating step 2. When the drone cluster finds a new target, it enters step 3 until the drone cluster completes all tasks in the area.

2. The task-driven UAV self-organizing network clustering method according to claim 1, characterized in that: The task cluster of step 3 is formed as follows: Step 3.1: Initialize parameters; set the maximum number of loops MaxIter, the number of particles p pa , establish a mapping table for drone ID based on the collected drone information, and u Determine the particle dimension dim so that N u =dim; Initialize the position p of particle pa pa and speed v pa , particle pa(0≤pa≤p pa ) The jth (0≤j≤dim) dimension is 1, indicating that the drone with mapping ID j joins the cluster; Step 3.2: Calculate the dominance of the particle set; calculate the average residual energy and free resources of the two optimization objectives according to the initial particle position and the fitness function; calculate the dominance relationship between particles according to the value; extract the solution set that does not dominate each other between particles and put it into the storage library, which stores the particle information of all non-dominated solutions; Step 3.3: loop; number of loops Iter = Iter + 1; Step 3.4: Update the weight factor w; Among them, Wmax and Wmin are the maximum weight and minimum weight experience values; Step 3.5: Use the roulette wheel selection algorithm to randomly select a Pareto optimal solution in the repository as the global optimal solution gbest, which is used to update the particle speed; Step 3.6: According to the historical optimal solution pbest of particle pa pa , that is, the position with the smallest fitness value among all the positions experienced by particle pa, and the global optimal solution gbest update the particle speed and map the particle speed; update the particle speed: v pa =w*v pa +c1*rand(n)*(pbest pa -p pa )+c2*rand(n)*(gbest-p pa ) Among them, c1 and c2 are acceleration coefficients, and their values ​​are usually [1.5, 2]; The particle velocity v pa Mapping to the new velocity v′ pa : Step 3.7: Update particle position p′ pa , and perform random mutation; if v′ pa (j)>σ, then the particle position is 1, otherwise the particle position in this dimension is 0, where σ is the mapping threshold, and its value is; Calculate the mutation factor p according to the number of iterations Iter, and select particles for bit mutation according to p; Among them, mu is the variation factor; Step 3.8: Calculate the fitness value FT, calculate the optimal solution, which is the minimum value of FT, and update the particle's historical optimal solution pbest according to the dominance relationship pa ; Step 3.9: Update the repository Rep; Step 3.10: If the number of iterations Iter is less than the maximum number of iterations MaxIter, continue iterating and jump to step 3.1, otherwise jump to step 3.11; Step 3.11: Based on the Pareto solutions of the repository, calculate the arrival time of each solution to the task target, and select the solution member with the minimum arrival time as the cluster member.

3. The task-driven UAV self-organizing network clustering method according to claim 2, characterized in that: The calculation method of cluster members in step 3.11 is as follows: The total number of drones in the entire network is set to N u , the number of known tasks is N t , the set of tasks is The set of clusters in the network is V c , assuming that there are N c clusters, the number of drones in the mth cluster is The set of drones in the mth cluster is but This method uses a The matrix to record the node's ownership, x i,m =1 means that node i, i.e. drone i, joins cluster m; In a heterogeneous drone cluster, each drone carries different resources. Assume that the drone cluster contains N q resources, then the resource vector of a single UAV i is The number of resource types required for the same task is also N q The resource requirement vector of the e-th task is Drone i location The location of the mission objective e Therefore, the constraints mainly include: Constraint 1: For pending tasks Corresponding task cluster The mission conditions must be met to complete the mission, so each resource of the drone in the cluster The sum must be greater than the task's resource requirement vector: Constraint 2: When there are multiple tasks that require the formation of multiple clusters, the same drone i can only belong to one cluster, that is, after multiple cluster head nodes initiate cluster formation applications, drone i can only choose one cluster to join; therefore, the constraints are: Constraint 3: In order to ensure the completion rate of the task, the energy E of the members in the cluster should be equal to i Maximization is calculated using the average energy within the cluster: For the convenience of calculation, the average energy can be taken as a negative value, and the maximization problem can be transformed into a minimization problem: Constraint 4: Since the number of resources in the drone network is limited, in order to ensure that the number of tasks to be executed is sufficient, when the drones form a cluster, the amount of excess resources should be minimized while satisfying the above equation, as shown in the following equation: Constraint 5: In order to assume that the time for a drone cluster to complete a task is fixed, the key factor that determines the time to complete a task is the arrival time of the drone cluster. The arrival time of the drone cluster depends on the position information of the shortest drone. Therefore, the farthest distance from the members of the drone cluster to the mission target determines the arrival time of the drone cluster. It is expected that the arrival time of the drone cluster is the minimum, so: In summary, the UTMOPSO fitness function and constraints are as follows: Under the multi-constraint conditions of the above formula, find the cluster member corresponding to the minimum fitness function value minFT.

Citation Information

Patent Citations

  • Multi-task dynamic allocation method for unmanned aerial vehicle cluster

    CN115202402A

  • Multi-unmanned aerial vehicle task allocation method based on mean shift clustering algorithm optimization

    CN116595864A

  • Method for constructing unmanned bee colony survivability topological structure based on connected graph

    CN118382120A

Cited By

  • Clustering type unmanned aerial vehicle group spectrum resource planning method and system and storage medium

    CN119743842A

  • Clustering method and device of unmanned aerial vehicle, electronic equipment and computer program product

    CN120358570A

  • UAV clustering method, device, electronic device and computer program product

    CN120358570B

  • Unmanned aerial vehicle cluster simulation scheduling method for time-sensitive target task

    CN122110763A

  • An unmanned aerial vehicle cluster simulation scheduling method for time-sensitive target tasks

    CN122110763B