Fish swarm algorithm-based cluster reconstruction method and system under large-scale damage

Abstract

The invention discloses a fish swarm algorithm-based cluster reconstruction method and system under large-scale damage, and the method specifically comprises the steps: cluster head selection and foraging signal release: after an unmanned plane cluster is damaged, cluster head nodes are selected again, and each cluster head node simulates the formation of a cruise formation in a space; cluster points are gathered and moved according to concentration, non-cluster-head nodes in a cluster sense the distribution state of other nodes in the communication range of the non-cluster-head nodes, local density information is calculated, and the nodes move to areas with low density according to the local density information; the basic behavior of the fish school is converted into an interaction rule between clusters, cluster head nodes dynamically fuse the interaction rule, the global optimal moving direction and the target position of the cluster are determined, and self-organizing reconstruction is achieved. The cluster destroyed in a large scale can be reconstructed without depending on a far-end control center, the communication quality of the whole network is optimized, and the reliability and effectiveness of the whole system are improved.

Description

Technical Field

[0001] This invention relates to the field of network communication technology, and in particular to a cluster reconstruction method and system based on the fish swarm algorithm under large-scale damage. Background Technology

[0002] Large-scale drone swarm reconfiguration technology aims to enable drone swarms to autonomously adjust their formation or organizational structure to maintain overall performance in the event of partial failure, mission changes, or environmental shifts. Traditional centralized or precise planning methods face three major bottlenecks in this regard: scalability, robustness, and real-time performance. However, fish communities in nature have already solved this problem.

[0003] Fish swarm algorithms, by simulating the autonomous behavior of individual fish and their group interaction rules, promise to provide a highly parallel and robust solution for the control of self-organized drone swarms. However, the basic fish swarm algorithm suffers from slow convergence speed in the later stages, susceptibility to getting trapped in local optima, and sensitivity to parameter settings. In chaotic and highly dynamic environments after damage, these problems are amplified, potentially leading to inefficient reconstruction or failure. Summary of the Invention

[0004] The purpose of this invention is to provide a cluster reconstruction method and system based on the fish swarm algorithm under large-scale damage, which ensures that each cluster in a multi-level network can dynamically adjust its target location according to the communication quality, optimize the overall network communication quality, and improve the reliability and effectiveness of the overall system.

[0005] The technical solution to achieve the purpose of this invention is: a cluster reconstruction method based on the fish swarm algorithm under large-scale damage, comprising the following steps:

[0006] Step 1: Cluster head selection and foraging signal release: After the drone swarm is damaged, a new cluster head node is elected, and each cluster head node simulates the formation of a cruising formation in space;

[0007] Step 2: Cluster aggregation and movement by concentration: Non-cluster head nodes in the cluster perceive the distribution status of other nodes within their communication range and calculate local density information. Each node moves to the area with lower density based on the local density information.

[0008] Step 3: Multi-cluster action, autonomous reconstruction according to fish school behavior rules: The basic behavior of the fish school is converted into interaction rules between clusters. Each cluster head node dynamically integrates the interaction rules to determine the global optimal movement direction and target position of the cluster, thereby achieving self-organized reconstruction.

[0009] Furthermore, step 1 is specifically as follows:

[0010] After the hierarchical structure of a large drone swarm is disrupted, the remaining scattered drone nodes elect a cluster head node according to the cluster head algorithm.

[0011] Each cluster head node establishes intra-cluster routes within its own cluster and cruises according to the set cruise trajectory;

[0012] Each cluster head node uses its own position as a pre-center to simulate the formation of a cruising formation in space.

[0013] Furthermore, the cluster aggregation described in step 2 involves, after the cluster head is selected, issuing a foraging signal centered on the cluster head. The cluster points calculate the target location based on the cluster head's position information and move towards the area with the lowest concentration, as detailed below:

[0014] For each cluster head node, multiple candidate target locations are defined within its communication radius;

[0015] The number of drone nodes within a set distance range of each candidate target location is counted and quantified as the node concentration at that candidate target location;

[0016] Non-cluster head nodes within the cluster acquire node concentration information for each candidate target location within the communication range, and autonomously move to the candidate target location with the lowest node concentration based on the concentration gradient, so as to achieve initial aggregation and distribution balance of nodes.

[0017] Furthermore, step 3 is specifically as follows:

[0018] Each cluster head node uses its onboard sensors to evaluate the quality of the communication link in its local area in real time, and calculates multiple communication quality indicators, including signal strength, signal-to-noise ratio, transmission delay and network throughput.

[0019] The three behavioral rules of fish schools—directionality, dispersion, and alignment—are transformed into interaction rules between clusters.

[0020] Based on the real-time evaluation results of the communication link quality, each cluster head node dynamically integrates interaction rules, calculates and selects the globally optimal movement direction and target position for its cluster, and drives the entire cluster to self-organize and reconstruct in space to continuously optimize and restore global network communication performance.

[0021] Furthermore, the directional rule is mapped as follows: the cluster head drives the cluster to move towards other clusters or regions with higher communication quality assessment;

[0022] Clusters are described by the following formula. Cluster The tendency force :

[0023]

[0024] in, Cluster The surrounding neighboring clusters, It is a cluster and cluster The distance between them; It is from a cluster Pointer cluster , unit vector; Cluster and cluster The signal power function between them adopts an exponential decay model.

[0025] Furthermore, the dispersion rule is mapped as follows: when the distance between this cluster and other clusters is too close, causing increased communication interference, the cluster head drives this cluster to actively move away from the other cluster;

[0026] Clusters are described by the following formula. and cluster Dispersion forces between :

[0027]

[0028] in, This represents the adjustable repulsive force gain coefficient.

[0029] Furthermore, the alignment rule is mapped as follows: the cluster head adjusts the direction and speed of its own movement to keep the movement of its own cluster coordinated with that of its neighboring clusters;

[0030] Clusters are described by the following formula. and cluster Alignment force between :

[0031]

[0032] in, It is a cluster The speed of movement, It is a cluster The speed of movement, It is an adjustable alignment force gain coefficient.

[0033] Furthermore, the cluster head nodes dynamically integrate interaction rules to determine the globally optimal movement direction and target position for their respective clusters, specifically as follows:

[0034] Calculate the resultant force :

[0035]

[0036] in, , , These are coefficients used to adjust the weights of the three types of fish behavior, respectively.

[0037] Update speed:

[0038] According to the principles of dynamics: acceleration Assuming quality Then the acceleration equals the net force; velocity updates should be based on acceleration:

[0039]

[0040] in, Indicates the simulation or control cycle; , These represent the clusters before and after the update, respectively. speed;

[0041] Update location:

[0042]

[0043] middle, , These represent the clusters before and after the update, respectively. The location.

[0044] A cluster reconstruction system based on the fish swarm algorithm under large-scale damage is disclosed. This system implements the aforementioned cluster reconstruction method based on the fish swarm algorithm under large-scale damage. The system includes a cluster head selection module, a cluster aggregation module, and multi-cluster actions.

[0045] The cluster head selection module is used for foraging signal dissemination. Specifically, after the drone swarm is damaged, a new cluster head node is elected, and each cluster head node simulates the formation of a cruising formation in space.

[0046] The cluster aggregation module is used for moving according to concentration. Specifically, non-cluster head nodes in the cluster sense the distribution status of other nodes within their communication range and calculate local density information. Each node moves to the area with lower density based on the local density information.

[0047] The multi-cluster action module is used for autonomous reconstruction according to the fish school behavior rules. Specifically, it transforms the basic behavior of the fish school into interaction rules between clusters. Each cluster head node dynamically integrates the interaction rules to determine the global optimal movement direction and target position of its own cluster, thereby achieving self-organized reconstruction.

[0048] A computer device includes a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the cluster reconstruction method based on the fish swarm algorithm under large-scale damage.

[0049] Compared with the prior art, the significant advantages of this invention are:

[0050] (1) Based on the biological characteristics of fish schools, analyze the aggregation and dispersion patterns of fish schools, draw on the advantage that fish schools can quickly reorganize after being dispersed, combine various behavioral patterns of fish schools such as foraging, gathering, and following, and sort out the description formulas corresponding to the behavioral rules for multi-level architectures to realize an accurate and reasonable cluster movement model based on fish school behavior.

[0051] (2) Ensure that each cluster in a multi-level network can dynamically adjust its target location based on communication quality, optimize the overall network communication quality, and improve the reliability and effectiveness of the overall system;

[0052] (3) It is suitable for large-scale cluster networks with multiple levels and can reconstruct clusters that have been destroyed on a large scale without relying on a remote control center. Attached Figure Description

[0053] Figure 1 This is a flowchart of the cluster reconstruction method based on the fish swarm algorithm under large-scale damage according to the present invention.

[0054] Figure 2 This is a schematic diagram of the cluster head foraging signal release status provided in Embodiment 1 of the present invention.

[0055] Figure 3 This is a schematic diagram of the cluster aggregation and concentration movement state provided in Embodiment 1 of the present invention. Detailed Implementation

[0056] This invention addresses the clustering structure of large UAV swarms, optimizing the application of the fish swarm algorithm in multi-level, multi-cluster networks. It combines various behavioral patterns such as foraging (individuals finding optimal positions), swarming (forming stable groups), and following (maintaining group connectivity) to address multi-polar architectures. It outlines descriptive formulas for three behavioral rules: directional (clusters tend to move towards other clusters, especially areas with better communication quality), dispersion (clusters need to maintain a certain distance to avoid collisions or interference), and alignment (cluster movement direction should be consistent with the direction of surrounding clusters to ensure coordination). This results in a cluster movement model based on fish swarm behavior, ensuring that each cluster in a multi-level network can dynamically adjust its target position according to communication quality, optimizing the overall network communication quality and improving the reliability and effectiveness of the overall system.

[0057] Combination Figure 1 This invention discloses a cluster reconstruction method based on the fish swarm algorithm under large-scale damage, comprising the following steps:

[0058] Step 1: Cluster head selection and foraging signal release: After the drone swarm is damaged, a new cluster head node is elected, and each cluster head node simulates the formation of a cruising formation in space;

[0059] Step 2: Cluster aggregation and movement by concentration: Non-cluster head nodes in the cluster perceive the distribution status of other nodes within their communication range and calculate local density information. Each node moves to the area with lower density based on the local density information.

[0060] Step 3: Multi-cluster action, autonomous reconstruction according to fish school behavior rules: The basic behavior of the fish school is converted into interaction rules between clusters. Each cluster head node dynamically integrates the interaction rules to determine the global optimal movement direction and target position of the cluster, thereby achieving self-organized reconstruction.

[0061] As a specific example, step 1 is as follows:

[0062] After the hierarchical structure of a large drone swarm is disrupted, the remaining scattered drone nodes elect a cluster head node according to the cluster head algorithm.

[0063] Each cluster head node establishes intra-cluster routes within its own cluster and cruises according to the set cruise trajectory;

[0064] Each cluster head node uses its own position as a pre-center to simulate the formation of a cruising formation in space.

[0065] As a specific example, the cluster aggregation described in step 2 involves, after the cluster head is selected, a foraging signal is issued with the cluster head as the center. The cluster points calculate the target position based on the cluster head position information and move towards the area with the lowest concentration, as detailed below:

[0066] For each cluster head node, multiple candidate target locations are defined within its communication radius;

[0067] The number of drone nodes within a set distance range of each candidate target location is counted and quantified as the node concentration at that candidate target location;

[0068] Non-cluster head nodes within the cluster acquire node concentration information for each candidate target location within the communication range, and autonomously move to the candidate target location with the lowest node concentration based on the concentration gradient, so as to achieve initial aggregation and distribution balance of nodes.

[0069] As a specific example, step 3 is as follows:

[0070] Each cluster head node uses its onboard sensors to evaluate the quality of the communication link in its local area in real time, and calculates multiple communication quality indicators, including signal strength, signal-to-noise ratio, transmission delay and network throughput.

[0071] The three behavioral rules of fish schools—directionality, dispersion, and alignment—are transformed into interaction rules between clusters.

[0072] Based on the real-time evaluation results of the communication link quality, each cluster head node dynamically integrates interaction rules, calculates and selects the globally optimal movement direction and target position for its cluster, and drives the entire cluster to self-organize and reconstruct in space to continuously optimize and restore global network communication performance.

[0073] As a specific example, the directional rule is mapped as follows: the cluster head drives the cluster to move towards other clusters or regions with higher communication quality assessment;

[0074] Clusters are described by the following formula. Cluster The tendency force :

[0075]

[0076] in, Cluster The surrounding neighboring clusters, It is a cluster and cluster The distance between them; It is from a cluster Pointer cluster , unit vector; Cluster and cluster The signal power function between them adopts an exponential decay model.

[0077] As a concrete example, the dispersion rule is mapped as follows: when the distance between a cluster and other clusters is too close, causing increased communication interference, the cluster head drives the cluster to actively move away from the other cluster;

[0078] Clusters are described by the following formula. and cluster Dispersion forces between :

[0079]

[0080] in, This represents the adjustable repulsive force gain coefficient.

[0081] As a specific example, the alignment rule is mapped as follows: the cluster head adjusts the direction and speed of its own movement to keep the movement of its own cluster coordinated with that of its neighboring clusters;

[0082] Clusters are described by the following formula. and cluster Alignment force between :

[0083]

[0084] in, It is a cluster The speed of movement, It is a cluster The speed of movement, It is an adjustable alignment force gain coefficient.

[0085] As a specific example, the cluster head nodes dynamically merge interaction rules to determine the globally optimal movement direction and target position for their respective clusters, specifically as follows:

[0086] Calculate the resultant force :

[0087]

[0088] in, , , These are coefficients used to adjust the weights of the three types of fish behavior, respectively.

[0089] Update speed:

[0090] According to the principles of dynamics: acceleration Assuming quality Then the acceleration equals the net force; velocity updates should be based on acceleration:

[0091]

[0092] in, Indicates the simulation or control cycle; , These represent the clusters before and after the update, respectively. speed;

[0093] Update location:

[0094]

[0095] middle, , These represent the clusters before and after the update, respectively. The location.

[0096] This invention also provides a cluster reconstruction system based on the fish swarm algorithm under large-scale damage. This system is used to implement the aforementioned cluster reconstruction method based on the fish swarm algorithm under large-scale damage. The system includes a cluster head selection module, a cluster aggregation module, and multi-cluster actions.

[0097] The cluster head selection module is used for foraging signal dissemination. Specifically, after the drone swarm is damaged, a new cluster head node is elected, and each cluster head node simulates the formation of a cruising formation in space.

[0098] The cluster aggregation module is used for moving according to concentration. Specifically, non-cluster head nodes in the cluster sense the distribution status of other nodes within their communication range and calculate local density information. Each node moves to the area with lower density based on the local density information.

[0099] The multi-cluster action module is used for autonomous reconstruction according to the fish school behavior rules. Specifically, it transforms the basic behavior of the fish school into interaction rules between clusters. Each cluster head node dynamically integrates the interaction rules to determine the global optimal movement direction and target position of its own cluster, thereby achieving self-organized reconstruction.

[0100] The present invention also provides a computer device, including: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the cluster reconstruction method based on the fish swarm algorithm under large-scale damage.

[0101] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0102] Example 1

[0103] Combination Figure 2 This embodiment provides a basic flowchart of cluster aggregation, which mainly includes the following processes:

[0104] After a scattered drone swarm elects a cluster head using a cluster head algorithm and the intra-cluster routing is completed, the cluster head cruises along a pre-defined circular trajectory, using its own position as a pre-center. The formation of a cruise fleet is simulated in space. Figure 1 As shown, each point on the circle represents the target location of the member node. , ...... Corresponding number , Set the current number of nodes in the cluster, and send a message containing the target location information to the temporary cluster members.

[0105] Combination Figure 3 It moves according to concentration. Within the communication radius of the drone. Each of the goals Count the number of nodes around the location. (If the number of nodes cannot be obtained, the signal strength in each target direction can be used as a substitute). This signal strength is then used as the target location concentration. The UAV node obtains the target location within its communication range. drone concentration And according to step size To the target location with the lowest concentration move.

[0106] Example 2

[0107] This embodiment provides multi-cluster actions that are autonomously reconstructed according to fish school behavior rules, including various behavior rules and formulas such as tendency, dispersion, and alignment.

[0108] (1) The homing tendency refers to the tendency of a cluster to gravitate towards a cluster region with better communication quality. Communication quality parameters are calculated by analyzing the signal strength, signal-to-noise ratio, delay, and throughput of surrounding clusters, ultimately yielding the homing force of the surrounding clusters towards this cluster. The following formula can be used to describe the homing tendency of a cluster. Cluster The tendency force :

[0109]

[0110] in, Cluster The surrounding neighboring clusters, It is a cluster and cluster The distance between them; It is from a cluster Pointer cluster , unit vector; Cluster and cluster The signal power function between them adopts an exponential decay model.

[0111] (2) The dispersion refers to the need for clusters to maintain a certain distance to avoid collisions and interference. The dispersion force of the surrounding clusters on the current cluster is calculated using the distance information between the current cluster and its neighboring clusters. The following formula describes the dispersion force of the clusters. and cluster Dispersion forces between :

[0112]

[0113] in, This represents the adjustable repulsion gain coefficient. This ensures that clusters do not get too close, avoiding communication conflicts.

[0114] (3) The alignment refers to the fact that the direction of movement of the cluster should be consistent with the direction of the surrounding clusters to ensure coordination. The alignment force of the cluster is obtained by measuring the direction and movement speed of the surrounding clusters, and the cluster is described by the following formula. and cluster Alignment force between :

[0115]

[0116] in, It is a cluster The speed of movement, It is a cluster The speed of movement, It is an adjustable alignment force gain coefficient.

[0117] Target location: The cluster head adjusts the target location based on communication quality and the behavior of other fish groups (other clusters). Each cluster will move towards the area with the best communication quality. The target location update calculation takes into account the following factors: current communication quality, distance and communication quality of surrounding clusters, current cluster position and target location.

[0118] Autonomous Action Reconfiguration: Clusters optimize communication quality within the network through interactions. Each cluster dynamically selects the optimal movement direction and target location based on its communication quality assessment with other clusters.

[0119] First, calculate the resultant force. :

[0120]

[0121] in, , , These are coefficients used to adjust the weights of the three types of fish behavior, respectively.

[0122] Update speed:

[0123] According to the principles of dynamics: acceleration Assuming quality Then the acceleration equals the net force; velocity updates should be based on acceleration:

[0124]

[0125] in, Indicates the simulation or control cycle; , These represent the clusters before and after the update, respectively. speed;

[0126] Update location:

[0127]

[0128] middle, , These represent the clusters before and after the update, respectively. The location.

[0129] Repeat the appeal steps to implement a cluster movement model based on fish swarm behavior, ensuring that each cluster in the network can dynamically adjust its target position according to communication quality, thereby optimizing the overall network communication quality.

[0130] In summary, this invention, based on the biological characteristics of fish schools, analyzes the aggregation and dispersion patterns of fish schools, draws on the advantage of fish schools' rapid reorganization after dispersal, and combines various behavioral patterns such as foraging, swarming, and following. For multi-level architectures, it outlines descriptive formulas corresponding to these behavioral rules, achieving an accurate and reasonable cluster movement model based on fish school behavior. This ensures that each cluster in a multi-level network can dynamically adjust its target position according to communication quality, optimizing the overall network's communication quality and improving the overall system's reliability and effectiveness. This invention is applicable to large-scale, multi-level cluster networks and can reconstruct clusters that have suffered large-scale destruction without relying on a remote control center.

Claims

1. A method for cluster reconstruction based on fish swarm algorithm in large-scale damage, characterized in that, The method comprises the following steps: Step 1, cluster head selection and foraging signal release: after the unmanned aerial vehicle cluster is damaged, a cluster head node is re-elected, and each cluster head node simulates the formation of a cruise formation in space; Step 2, cluster point aggregation and movement according to concentration: the non-cluster head nodes in the cluster perceive the distribution state of other nodes within their communication range, and calculate local density information, and each node moves to an area with lower density according to the local density information; Step 3, multi-cluster action and autonomous reconstruction according to fish school behavior rules: the basic behaviors of fish schools are converted into interaction rules between clusters, each cluster head node dynamically fuses the interaction rules to determine the global optimal moving direction and target position of the cluster, and realizes self-organizing reconstruction.

2. The method of claim 1, wherein the method is characterized by, The step 1 is specifically as follows: After the hierarchical structure of the large unmanned aerial vehicle cluster is damaged, the scattered unmanned aerial vehicle nodes left over select a cluster head node according to a cluster head algorithm; Each cluster head node establishes an intra-cluster routing in the cluster to which it belongs, and cruises according to a set cruise trajectory; Each cluster head node takes its own position as a pre-center to simulate the formation of a cruise formation in space.

3. The method of claim 1, wherein the method is characterized by: The cluster point aggregation of step 2 is as follows: For each cluster head node, a plurality of candidate target positions are defined within the communication radius thereof; The number of unmanned aerial vehicle nodes within the distance range set for each candidate target position is counted and quantified as the node concentration of the candidate target position; The non-cluster head nodes in the cluster obtain the node concentration information of each candidate target position within the communication range, and autonomously move to the candidate target position with the lowest node concentration according to the concentration gradient, so as to realize the preliminary aggregation and distribution balance of the nodes.

4. The method of claim 1, wherein the method is characterized by: The step 3 is specifically as follows: Each cluster head node evaluates the communication link quality of the local area through the sensor carried thereon in real time, and calculates a plurality of communication quality indexes including signal strength, signal-to-noise ratio, transmission delay and network throughput; The three behavior rules of fish schools, i.e. tropotaxis, dispersion and alignment, are converted into interaction rules between clusters; Based on the real-time evaluation result of the communication link quality, each cluster head node dynamically fuses the interaction rules, calculates and selects the global optimal moving direction and target position of the cluster, and drives the entire cluster to perform self-organizing reconstruction in space, so as to continuously optimize and restore the global network communication performance.

5. The method of claim 4, wherein the method is characterized by, The tropotaxis rule is mapped as: the cluster head drives the cluster to move in the direction of other clusters or areas with higher communication quality evaluation; The cluster is described by the following equation The cluster tends to : ； where, denotes the cluster around the cluster is the cluster and the cluster distance between the cluster is a unit vector from the cluster to the cluster denotes the signal power function between the cluster and the cluster with an exponential decay model.​ 6. The method of claim 5, wherein the method is characterized by, The dispersion rule is mapped as: when the cluster is too close to other clusters, causing communication interference to be intensified, the cluster head drives the cluster to actively move away from the cluster; The dispersion forces between the clusters and the clusters are described by the following formula : ； wherein represents an adjustable repulsive force gain coefficient.

7. The method of claim 6, wherein the method is a large-scale damage-based flocking algorithm-based cluster reconstruction method, characterized by, The alignment rule is mapped as: the cluster head adjusts the motion direction and speed of the cluster, so that the motion state of the cluster and the adjacent cluster is kept coordinated; The alignment force between the clusters and the cluster is described by the following equation : ； wherein, is the velocity of motion of the cluster of clusters is the velocity of motion of the cluster of clusters is the adjustable alignment force gain coefficient.

8. The method of claim 7, wherein the method is characterized by, Each cluster head node dynamically fuses the interaction rules to determine the global optimal moving direction and target position of the cluster, specifically as follows: calculating the resultant force : ； wherein, , , are coefficients for adjusting the weights of the three fish behaviors, respectively; Update speed: According to the principle of dynamics: acceleration , assuming mass , then acceleration is equal to the resultant force; velocity update should be based on acceleration: ； wherein denotes a simulation or control cycle; , denote the velocities of the clusters before and after the update, respectively ; Update position: ； In, , respectively represent the position of the cluster before and after the update.

9. A cluster reconstruction system based on fish swarm algorithm in large scale damage, characterized in that, The system is used to realize the cluster reconstruction method based on fish school algorithm under large-scale damage according to any one of claims 1-8, and the system comprises a cluster head selection module, a cluster point aggregation module and a multi-cluster action module. The cluster head selection module is used for foraging signal release, and specifically, after the unmanned aerial vehicle cluster is destroyed, cluster head nodes are re-elected, and each cluster head node simulates the formation of a cruising formation in space; The cluster point aggregation module is used for moving according to concentration, and specifically, non-cluster head nodes in the cluster perceive the distribution state of other nodes within the communication range, and calculate local density information, and each node moves to a region with lower density according to the local density information; The multi-cluster action module is used for autonomous reconstruction according to fish school behavior rules, and specifically, the basic behaviors of fish schools are converted into interaction rules between clusters, each cluster head node dynamically fuses the interaction rules to determine the global optimal moving direction and target position of the cluster, and self-organizing reconstruction is realized.

10. A computer device, comprising: The method comprises: A memory and a processor are communicatively connected, the memory stores computer instructions, and the processor executes the computer instructions to perform the method according to any one of claims 1-8.

Images