Dynamic clustering routing method for high-dynamic unmanned aerial vehicle group network

Abstract

The invention relates to the technical field of unmanned aerial vehicle group networks, in particular to a dynamic clustering routing method for a high-dynamic unmanned aerial vehicle group network, which is characterized by comprising the following steps of: dividing the whole unmanned aerial vehicle group network into a centralized-distributed hierarchical domain architecture, dividing the network into a plurality of clusters, and enabling each cluster to consist of a cluster head and a plurality of cluster members, the cluster head can be recombined into a higher-level cluster; according to the method, a layered and domain-divided network architecture is provided, cluster heads are dynamically elected in combination with a grey wolf optimization algorithm, real-time self-adaptive clustering and load balancing routing of the unmanned aerial vehicle group in a high-speed mobile environment are realized through a double-cluster-head switching and reselection mechanism, the cluster heads and paths are ensured to be optimal, routing is optimized through a genetic algorithm, relay forwarding is reduced, and the routing efficiency is improved. The communication overhead is reduced, the service life of the network is prolonged, the robustness of the system is enhanced by the alpha-beta-delta leader model and the iterative surrounding strategy, the time delay is reduced and the transmission success rate is improved by the fuzzy clustering, and the real-time reliable communication requirement in a high dynamic scene is met.

Description

Technical Field

[0001] The present invention relates to the technical field of drone swarm networks, and in particular to a dynamic clustering routing method for a highly dynamic drone swarm network. Background Art

[0002] In recent years, with breakthroughs in flight control, communication electronics, navigation and positioning, and artificial intelligence, drone technology has developed rapidly. Drones have the advantages of small size, autonomous control, low operating costs, and low operational risks. Drones have gradually expanded from their original military-specific equipment to civilian and commercial fields, becoming an indispensable and important tool in modern society. The progress of drone technology is not only reflected in the improvement of single-machine performance, but also in the enhancement of multi-machine coordination capabilities, which has made the concept of drone swarm collaborative work gradually move from theory to practice. Compared with a single drone, the advantages of drone swarms are more significant. When performing tasks, drone swarms can achieve wider area coverage with the help of distributed collaboration mode, greatly improving the efficiency of task execution. At the same time, drone swarms have excellent robustness, even if a single drone Even if a machine fails, it will not cause the entire system to fail, ensuring the continuous advancement of the mission. In the face of complex and changing environments, drone swarms can also flexibly respond to various challenges through collaborative perception and decision-making mechanisms, demonstrating strong environmental adaptability and mission execution capabilities. Nowadays, drone swarms can complete complex tasks that are difficult for single drones to achieve with their advantages such as easy deployment and strong maneuverability, such as large-scale area monitoring, traffic management, efficient logistics distribution, precision agriculture management, and disaster emergency response. Drone swarm collaborative technology has made great progress and laid a solid foundation for various applications. However, with the increase in application demand, people have also put forward higher requirements for the collaborative control, path planning and communication optimization of drone swarms, especially the communication needs in highly dynamic environments. Drone swarm communication technology faces many challenges.

[0003] A highly dynamic UAV swarm refers to a UAV system whose topology, communication links, and mission requirements change rapidly during mission execution. This high dynamism is usually caused by a variety of complex factors, such as the rapid movement of UAVs, dynamic interference from environmental obstacles and malicious nodes, volatility of communication links, and real-time adjustment of mission objectives. In a highly dynamic environment, the network topology and communication conditions of the UAV swarm may change at any time. Traditional routing protocols (such as AODV, OLSR, and GPSR) often fail to meet the real-time and stability requirements due to their lack of sufficient flexibility and adaptability. This can significantly reduce the mission execution efficiency of the UAV swarm and may even cause serious consequences such as communication interruption, mission failure, or system crash. In particular, when the UAV nodes in the network are highly mobile and the topology changes frequently, the communication links between nodes are extremely easy to break, and the original clustering method can no longer maintain the operation of the network, resulting in large-scale node failure. This not only affects the communication coverage capability of the system, but also causes network regional fragmentation, which in turn affects data exchange in the UAV swarm network and reduces the communication performance of the entire UAV swarm network.

[0004] The main challenge of highly dynamic drone swarm routing algorithms is how to achieve efficient routing planning and multi-factor trade-off optimization in a highly dynamic environment. For drone swarm clustered networks, the dynamic changes of nodes require the network to be re-clustered in a timely manner to adapt to the new topology. At the same time, the dynamic changes in network task requirements or the increase in data flow require the routing algorithm to have rapid response and load balancing capabilities. For this highly dynamic, large-scale drone swarm network environment, traditional wireless routing methods and clustering principles have become difficult to apply. Some existing routing methods are also difficult to ensure good communication quality in highly dynamic scenarios, and sometimes communication in the network can only be achieved through flooding. In order to solve this problem, scholars at home and abroad have gradually begun to study dynamic clustering routing methods for drone swarm networks. By re-clustering the drone swarm network, a clustering strategy suitable for the current node distribution and network status is made, and load balancing is achieved through a reasonable routing algorithm to ensure the real-time and reliability of node communication.

[0005] The integration of dynamic clustering and intelligent routing methods has gradually become a new research field. For the dynamic clustering and routing decision-making problems in highly dynamic UAV swarm networks, existing routing methods have problems such as link loss between nodes and high energy consumption. Nowadays, there is an increasing amount of research on optimizing network routing problems. Most of the research aims to reduce blocking rate, reduce latency, and improve computational efficiency to improve network performance. To solve the above problems, we propose a dynamic clustering routing method for highly dynamic UAV swarm networks. Summary of the Invention

[0006] In view of the shortcomings of the existing technology, the present invention provides a dynamic clustering routing method for a highly dynamic drone swarm network to solve the problems raised in the background technology.

[0007] The above technical objectives of the present invention are achieved through the following technical solutions:

[0008] A dynamic cluster routing method for a highly dynamic drone swarm network includes the following steps:

[0009] S1. The entire drone swarm network is divided into a centralized-distributed hierarchical and domain-specific architecture. The network is divided into several clusters, each consisting of a cluster head and multiple cluster members. Cluster heads can further form higher-level clusters. When a service arrives, it is determined to be intra-cluster routing or inter-cluster routing based on the cluster affiliation of the source node and the destination node of the service. Intra-cluster routing is determined by the intra-cluster controller and forwarded directly. Inter-cluster routing is determined by the inter-cluster controller reading the inter-cluster routing table to determine the destination cluster, selecting the optimal cluster head forwarding path to the destination cluster head, and then the cluster head activates the intra-cluster controller to forward the service.

[0010] S2. Dynamic cluster head election based on the gray wolf optimization algorithm: Each node is evaluated according to the weighted fitness function value that combines the node's residual energy, congestion, connectivity, and average distance to neighboring nodes. The flight energy consumption, data transmission energy consumption, and data reception and aggregation energy consumption are comprehensively calculated to obtain the node's residual energy. Based on this, the state distance is introduced to simulate the encirclement and attack process of wolf hunting. The node's state vector is updated through multiple iterations, and the nodes with the highest fitness are finally selected as cluster head candidates.

[0011] S3. Based on the selected cluster head candidates, the dual cluster head (α-β) centers are used as the initial cluster centers to perform membership clustering on the remaining nodes: Initialize the membership matrix, calculate the distance between each node and the dual cluster head centers, and iteratively update the membership and new cluster centers until the clustering results converge;

[0012] S4. When a routing request arrives, all possible paths are encoded as chromosomes, and the quality of the paths is evaluated using the same weighted fitness function. The population is optimized through genetic operators such as roulette wheel selection, single-point crossover, and random gene mutation to select the optimal route.

[0013] S5. During network operation, a cluster head load threshold is set. When the current cluster head load exceeds the threshold, the β node will temporarily take over the cluster head responsibilities and switch back after the original cluster head load recovers. When both α and β are not suitable as cluster heads, local re-election will be performed within the cluster first, and global cluster head re-election will be performed if necessary to ensure the balance of cluster head responsibilities and network stability.

[0014] Preferably, the fitness function Fi is calculated as follows:

[0015] Fi =w1×E i +w2×C i +w3×N i +w4×D i

[0016] Among them, E i 、C i 、N i 、D i are the residual energy, congestion, connectivity and average distance of node i respectively, and w1 to w4 are weight coefficients.

[0017] Preferably, the remaining energy E i Calculate as follows:

[0018]

[0019] Among them, m i represents the mass of node i, t represents the flight time, V i Indicates the average speed during the flight time, E elec is the energy consumed to transmit 1 bit of data, ε fs and ε mp Respectively represent the energy loss coefficient of the power amplifier circuit under different channel transmission models.

[0020] Preferably, the node congestion degree C i Calculate as follows:

[0021]

[0022] Among them, k j represents the length of the i-th data packet in node i, k all Indicates the maximum data capacity of node i.

[0023] Preferably, the node connectivity N i Calculate as follows:

[0024]

[0025] The connectivity of a node represents the ratio of the number of neighboring nodes of the node to the total number of nodes in the network.

[0026] Preferably, the average distance D i Calculate as follows:

[0027]

[0028] Where D represents the average distance between neighboring nodes within the communication range of the node.

[0029] Preferably, the iterative update of the gray wolf optimization adopts the following formula:

[0030] X i (t+1)=X j -A i ×D ij

[0031] A i =2a i ×r 1i -a i

[0032]

[0033] In each iteration, the top three nodes in terms of fitness are used as α, β, and δ, and the status of other nodes are updated according to the following formula:

[0034]

[0035] When the iteration is completed, the fitness value of each node is calculated again, and the nodes are classified according to the final fitness value. The first c nodes with the highest fitness value (c represents the number of clusters) are marked as α nodes as cluster heads, the next c nodes are marked as β nodes, and the next c nodes are marked as δ nodes. The remaining nodes are marked as ω nodes.

[0036] Preferably, the clustering of the ω nodes is calculated using the membership matrix U and dual cluster head centers:

[0037]

[0038] And according to the formula:

[0039]

[0040] Iterate and update until convergence.

[0041] Preferably, the genetic algorithm route optimization includes:

[0042] Chromosomes are sequences of possible path nodes;

[0043] Roulette wheel selection, single-point crossover and random gene mutation operators are used;

[0044] The mutation rate is set low to balance diversity and convergence.

[0045] Preferably, the adaptive intra-cluster dual-cluster head switching mechanism is:

[0046] When the load of cluster head node α exceeds the preset threshold, node β is temporarily promoted to cluster head;

[0047] When the original α load returns to below the threshold, it switches back to α again;

[0048] When both α and β are unsuitable, a local re-election is triggered within the cluster, and a global re-election of the cluster head of the entire network is performed when necessary.

[0049] In summary, the present invention mainly has the following beneficial effects:

[0050] 1. The present invention realizes real-time adaptive clustering and energy efficiency optimization of drone swarms in high-speed mobile and frequently changing topology environments through a hierarchical centralized-distributed architecture and dynamic cluster head election and load-balancing routing based on gray wolf optimization and genetic algorithm. On the one hand, the multi-factor weighted fitness function (covering residual energy, congestion, connectivity and distance) ensures that the optimal cluster head is selected in each iteration, and combines dual cluster head switching with local / global reselection mechanism to significantly improve the stability and balance of the cluster structure. On the other hand, the cluster head forwards only when cross-cluster communication is necessary, and optimizes the path through genetic operators to minimize communication overhead and energy consumption, thereby extending the network life and maintaining efficient routing performance.

[0051] 2. In addition, the present invention uses the α-β-δ three-level leader model and the "state distance" iterative encirclement strategy to give the system extremely strong robustness and fault tolerance. Even if some nodes fail or encounter interference, the cluster head can be quickly reconstructed and routing continuity can be guaranteed. Combined with fuzzy membership clustering and multi-factor fitness optimization, the optimal path significantly reduces transmission delay and improves the success rate of data packets, meeting the real-time and reliability requirements in highly dynamic application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 A centralized-distributed hierarchical and domain-based high-dynamic drone swarm network architecture and routing forwarding diagram provided by an embodiment of the present invention;

[0053] Figure 2 A flow chart of a highly dynamic UAV swarm network clustering method based on dynamic cluster head election and node clustering provided by an embodiment of the present invention;

[0054] Figure 3 An overall flow chart of a highly dynamic UAV swarm network routing decision method based on dynamic clustering provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the described embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] The following examples are intended to illustrate the present invention but are not intended to limit the scope of protection of the present invention. The conditions in the examples may be further adjusted according to specific conditions. Simple improvements to the method of the present invention within the scope of the present invention are also within the scope of protection claimed in the present invention.

[0057] Example 1

[0058] refer to Figure 1-Figure 3 ,A dynamic cluster routing method for a highly dynamic UAV swarm network includes the following steps:

[0059] S1. The entire drone swarm network is divided into a centralized-distributed hierarchical and domain-specific architecture. The network is divided into several clusters, each consisting of a cluster head and multiple cluster members. Cluster heads can further form higher-level clusters. When a service arrives, it is determined to be intra-cluster routing or inter-cluster routing based on the cluster affiliation of the source node and the destination node of the service. Intra-cluster routing is determined by the intra-cluster controller and forwarded directly. Inter-cluster routing is determined by the inter-cluster controller reading the inter-cluster routing table to determine the destination cluster, selecting the optimal cluster head forwarding path to the destination cluster head, and then the cluster head activates the intra-cluster controller to forward the service.

[0060] S2. Dynamic cluster head election based on the gray wolf optimization algorithm: Each node is evaluated according to the weighted fitness function value that combines the node's residual energy, congestion, connectivity, and average distance to neighboring nodes. The flight energy consumption, data transmission energy consumption, and data reception and aggregation energy consumption are comprehensively calculated to obtain the node's residual energy. Based on this, the state distance is introduced to simulate the encirclement and attack process of wolf hunting. The node's state vector is updated through multiple iterations, and the nodes with the highest fitness are finally selected as cluster head candidates.

[0061] S3. Based on the selected cluster head candidates, the dual cluster head (α-β) centers are used as the initial cluster centers to perform membership clustering on the remaining nodes: Initialize the membership matrix, calculate the distance between each node and the dual cluster head centers, and iteratively update the membership and new cluster centers until the clustering results converge;

[0062] S4. When a routing request arrives, all possible paths are encoded as chromosomes, and the quality of the paths is evaluated using the same weighted fitness function. The population is optimized through genetic operators such as roulette wheel selection, single-point crossover, and random gene mutation to select the optimal route.

[0063] S5. During network operation, a cluster head load threshold is set. When the current cluster head load exceeds the threshold, the β node will temporarily take over the cluster head responsibilities and switch back after the original cluster head load recovers. When both α and β are not suitable as cluster heads, local re-election will be performed within the cluster first, and global cluster head re-election will be performed if necessary to ensure the balance of cluster head responsibilities and network stability.

[0064] Figure 1 This is a diagram of the centralized-distributed hierarchical and domain-based high-dynamic drone swarm network architecture and routing forwarding, specifically including:

[0065] A centralized-distributed hierarchical and domain-based high-dynamic UAV swarm network architecture is provided. This solves the problems of large network scale, strong network dynamics, difficulty in collecting real-time global network status statistics, and excessive overhead of traditional flooding methods in high-dynamic UAV swarm networks. To meet the real-time and changing needs of tasks in dynamic scenarios, the centralized-distributed hierarchical and domain-based high-dynamic UAV swarm network architecture is divided into several clusters. Each cluster consists of a cluster head and multiple members. The cluster heads can form a higher-level network, enabling the system to meet the real-time and highly dynamic environment while achieving high routing efficiency.

[0066] When a service arrives, the cluster to which it belongs will determine whether the task is an intra-cluster routing task or a cross-domain routing service based on the clustering relationship between the source node and the destination node of the service. When it is determined to be an intra-cluster routing service, only the intra-cluster routing controller is responsible for routing decision output and directly forwards routing decisions within the cluster. When it is determined to be an inter-cluster routing service, the inter-cluster routing controller is responsible for routing control decision output, reading the inter-cluster routing table, determining the cluster where the destination node is located, and finding the optimal cluster head routing forwarding method. When the cluster head of the destination node is reached, the intra-cluster routing controller is activated to be responsible for routing decision output and find the optimal intra-cluster forwarding method within the cluster.

[0067] Figure 2 This is a flow chart of a highly dynamic UAV swarm network clustering method based on dynamic cluster head election and node clustering, specifically including:

[0068] After the network is established, if the network needs to be re-clustered, the cluster head election is first performed according to the dynamic cluster head election strategy. In each iteration, the fitness value of each drone node is calculated according to Formula 1, and then the state distance of each node is calculated and updated to the state distance matrix. Next, the network state is updated according to Formula 11. When the maximum number of iterations is reached, the iteration ends, the fitness values ​​of all nodes are calculated, and the appropriate α-β-δ nodes are selected independently to each cluster. The remaining ω nodes will calculate the cluster center of each cluster based on the selected cluster head, and update the membership matrix based on Formula 13 to calculate the degree of membership of each node to each cluster. When the maximum number of iterations is reached or the iteration stop condition is met, a cluster cluster label is assigned to each node according to the FCM algorithm result. At this time, the entire clustering process is completed. If the network does not need to be re-clustered, it continues to run according to the current clustering strategy until the current strategy can no longer meet the network operation requirements, and then re-clustering is performed.

[0069] Figure 3 This is the overall flow chart of the routing decision method for highly dynamic UAV swarm networks based on dynamic clustering, which specifically includes:

[0070] After the drone swarm network is established, the initial network is clustered by the dynamic clustering method. The network will operate normally according to the clustering strategy. During operation, each cluster will regularly detect whether the cluster head of the cluster is in a high-load state. If the cluster head is already in a high-load state, the cluster head will detect whether the network meets the conditions for re-clustering of the entire network, including that more than half of the cluster heads are high-loaded or have low energy. If so, the entire network will be re-clustered according to the proposed dynamic clustering method. If there is no need for full network re-clustering, it will detect whether the β node can replace the α node as the cluster head. If so, the β node will be switched to the cluster head node of the cluster. If not, the cluster head is re-elected according to the fitness function value sorting within the cluster. When there is a data packet to be forwarded in the network, a chromosome population is established based on the source node and the destination node, and the fitness value of each chromosome is calculated according to Formula 1. The excellent parent chromosome is selected by roulette using Formula 14, and the next generation population is generated through single-point crossover and mutation, and the population is updated. If the maximum number of iterations is reached, the algorithm ends, otherwise the next generation fitness value is calculated and the iteration continues. Through continuous iteration, the algorithm gradually evolves the chromosome with the highest fitness (i.e., routing path), and the data packet will be forwarded according to the path.

[0071] Example 2

[0072] refer to Figure 1-Figure 3Based on the hierarchy of gray wolf groups and combined with the drone swarm network, drones are divided into four levels. The first layer of the cluster is the leader in each cluster, that is, the cluster head node, called α node. In the cluster, α node has management capabilities and cross-cluster transmission capabilities. The second layer is the backup node of α, called β node. β node mainly shares the task load of α to prevent α node congestion and excessive consumption. When the α node in the cluster becomes vacant, such as death, failure, or leaving the range of the cluster, β node will take over the position of α node. The status of β node in the cluster is second only to α node, and it has the same function as α node. The third layer is δ node. δ node is mainly responsible for maintaining information within the cluster, such as collecting status information of nodes within the cluster and analyzing whether a new cluster head election is needed within the cluster. The bottom layer is ω node, which is mainly responsible for sending, relaying and receiving data within the cluster.

[0073] In order to elect the most suitable node as the cluster head in the network, this strategy comprehensively considers the factors affecting network performance, including the remaining energy E of the drone, the congestion C of the drone, the connectivity N of the drone node, and the average distance D between the drone and its neighboring nodes. The fitness function value Fi of the drone node i is calculated according to formula (1) to comprehensively evaluate the pros and cons of the node as the cluster head, providing a basis for cluster head election.

[0074] F i =w1×E i +w2×C i +w3×N i +w4×D i (1)

[0075] The reference factors in formula (1) are defined as follows:

[0076] The energy consumption of node i includes the flight energy consumption, transmission energy consumption and reception energy consumption of the node. The flight energy consumption of the UAV mainly depends on the mass and flight speed of the UAV. Therefore, the flight energy consumption of the UAV is shown in formula (2), where m i represents the mass of node i, t represents the flight time, V i Express the average speed during the flight time:

[0077]

[0078] The transmission energy consumption of the UAV is related to the distance between the transmitter and the receiver and the amount of data. The transmission energy consumption is calculated as shown in formula (3):

[0079]

[0080] Among them E elec is the energy consumed to transmit 1 bit of data, ε fs and ε mpThey represent the energy loss coefficient of the power amplifier circuit under different channel transmission models, the influence of the channel mode selection on the transmission distance, is the transmission distance threshold.

[0081] The calculation formula for the energy consumption of receiving data is shown in formula (4), where ρ represents the aggregation efficiency factor, E da Energy consumed to aggregate 1 bit of data:

[0082]

[0083] In summary, the energy consumption calculation formula is shown in formula (5), where: Represents the initial total energy of the drone:

[0084]

[0085] The node congestion reflects the current communication load of the node. The node with lower congestion can have a larger space for processing data and avoid communication congestion. The calculation formula of the node congestion is shown in formula (6), where k j represents the length of the jth data packet in node i, k all Indicates the maximum data capacity of node i.

[0086]

[0087] The connectivity of a node represents the ratio of the number of neighboring nodes of the node to the total number of nodes in the network. Nodes with higher connectivity can communicate better with other nodes and improve the overall connectivity of the network. The calculation formula of node connectivity is shown in formula (7):

[0088]

[0089] The last evaluation factor, average distance D, represents the average distance between neighboring nodes within the communication range of a node. Nodes with smaller average distances can reduce the energy consumption of data transmission. The calculation formula for the average distance is shown in (8):

[0090]

[0091] The dynamic cluster head election strategy based on the gray wolf optimization algorithm refers to the hunting process of wolves, including tracking and approaching prey, chasing and surrounding, and attacking prey, and determines the optimal node division of labor through multiple iterations.

[0092] For the entire drone swarm, the present invention defines a population matrix, including the remaining energy E, congestion C, node connectivity N, and average distance D from neighboring nodes of each drone, so as to calculate the fitness of the drone. At the same time, the "state distance" is defined to measure the degree of difference between the states of the nodes. The calculation formula of the "state distance" is shown in formula (9):

[0093] D ij =|C i ×X j -X i | (9a)

[0094] C i =2r i (9b)

[0095] According to the calculated “state distance”, the node will update the optimal state. Specifically, the population will iterate to the node with better comprehensive performance to select the cluster head with the best comprehensive performance. The state vector formula of node i in the t+1th iteration is shown in formula (11):

[0096] X i (t+1)=X j -A i ×D ij (10a)

[0097] A i =2a i ×r 1i -a i (10b)

[0098]

[0099] In each iteration, the top three nodes in terms of comprehensive performance are selected as α nodes, β nodes, and δ nodes. Other nodes will update their status based on the status information of these three nodes. The calculation formula is shown in (11):

[0100]

[0101] When the iteration is completed, the fitness value of each node is calculated again, and the nodes are classified according to the final fitness value. The first c nodes with the highest fitness value (c represents the number of clusters) are marked as α nodes as cluster heads, the next c nodes are marked as β nodes, and the next c nodes are marked as δ nodes. The remaining nodes are marked as ω nodes.

[0102] This embodiment addresses the issue of link stability between nodes in highly dynamic scenarios and proposes a dynamic node clustering algorithm based on fuzzy C-means (FCM). Based on the proposed GWO-based dynamic cluster head election strategy, c α nodes, β nodes, and δ nodes are elected (where c represents the number of clusters). These nodes are grouped together based on proximity to form an α-β-δ cluster manager combination for each cluster. FCM clustering is then performed on the remaining ω nodes.

[0103] The node clustering algorithm proposed in the present invention uses membership to represent the relationship between each data point, thereby determining the cluster to which each data point belongs. First, the membership matrix U=[u ij ] n×c , where u ij Indicates the degree of membership of the i-th node to the j-th cluster, and n represents the total number of nodes surviving in the current network. In addition, this scheme adopts the α-β node dual cluster head optimization mechanism to prevent the cluster head node from being overloaded. Therefore, the location information of the α node and the β node needs to be taken into account in the calculation of the cluster center. For cluster j, and the dual cluster heads (x jα ,y jα ,z jα )、(x jβ ,y jβ ,z jβ ) The calculation formula of its cluster center is shown in formula (12):

[0104]

[0105] In each iteration, the core task is to update the membership matrix U and the cluster centers. First, the distance d from each node to each cluster center is calculated according to formula (8b): ij , get the distance matrix, and then use formula (13) to update the membership matrix and recalculate the cluster center v according to the distance information j :

[0106]

[0107]

[0108] When the iteration is completed, each surviving ω node is assigned a cluster label according to the final membership matrix, and forms a new cluster with the corresponding α-β-δ nodes to complete the final cluster division.

[0109] On the one hand, this embodiment provides a load balancing routing algorithm based on an improved genetic algorithm (GA), which takes into account factors such as the load degree and residual energy of the cluster head node in real time, reduces the load and energy consumption of the cluster head, and improves the survival time of the network. The algorithm first encodes the routing path, and each possible routing path is encoded as a chromosome. Specifically, the chromosome is composed of a series of genes, each gene represents a node in the drone swarm network, and the order of arrangement of the genes represents the forwarding order of data packets between nodes, that is, the routing path.

[0110] In the process of UAV swarm routing, the design of the fitness function needs to comprehensively consider multiple factors to ensure that the selected routing path can achieve load balancing and meet the network requirements of low latency and high reliability. The fitness function F i The calculation is the same as formula (1), the value of the fitness function F i The larger it is, the better the routing path is, and the more likely the node will choose it.

[0111] Genetic algorithms simulate the biological evolution process through genetic operations such as selection, crossover, and mutation, and continuously optimize the chromosomes in the population to find the optimal solution. The selection method adopts the roulette wheel selection method. In the roulette wheel selection method, the probability of each chromosome being selected is proportional to its fitness value. Specifically, assuming that there are M chromosomes (i.e., routing paths) in the population, the fitness of the kth chromosome is F k , then the probability of being selected is P k As shown in formula (14):

[0112]

[0113] The proposed algorithm uses a single-point crossover method. The specific operation is as follows: First, a crossover point is randomly selected. Then, the two parent chromosomes (i.e., routing paths) are swapped at the crossover point to obtain two new daughter chromosomes. If the two parent chromosomes are [n1, n2, n3, n4] and [m1, m2, m3, m4], respectively, and the crossover point is a second gene, the daughter chromosomes after the crossover are [n1, n2, m3, m4] and [m1, m2, n3, n4], respectively. The crossover operation can increase the diversity of the population and help the algorithm escape the local optimal solution. The mutation operation randomly changes some genes in the chromosome to introduce new genetic information, further increasing the diversity of the population. In a drone swarm network, the mutation operation can be implemented by randomly replacing a node in the routing path.

[0114] A routing decision method for highly dynamic drone swarm networks based on dynamic cluster head election and node clustering is proposed. In order to effectively alleviate the high load problem of cluster head nodes, an adaptive intra-cluster dual-cluster head switching mechanism and an intra-cluster re-election mechanism are proposed for drone swarm networks. When the load of the cluster head node exceeds a certain threshold, the cluster head of this cluster will be replaced by the β node obtained by the clustering strategy until the load of the original α node returns to a lower level. At the same time, in order to reduce the overhead of frequent clustering of the overall network, an intra-cluster re-election mechanism is proposed. The mechanism uses local election first and global election second to ensure that when both α node and β node in a cluster in the network are not suitable as cluster heads, a new cluster head suitable for forwarding can still be elected within the cluster.

[0115] Although the embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that, unless otherwise defined, the technical or scientific terms used in the present invention shall have the usual meanings understood by persons having ordinary skills in the field to which the present invention belongs, and the words "include" or "comprise" and the like used in the present invention mean that the elements or objects preceding the word include the elements or objects listed after the word and their equivalents. Although the embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations may be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention shall be defined by the appended claims and their equivalents.

Claims

1. A dynamic cluster routing method for a highly dynamic UAV swarm network, characterized by: The following steps are involved: S1. The entire drone swarm network is divided into a centralized-distributed hierarchical and domain-specific architecture. The network is divided into several clusters, each consisting of a cluster head and multiple cluster members. Cluster heads can further form higher-level clusters. When a service arrives, it is determined to be intra-cluster routing or inter-cluster routing based on the cluster affiliation of the source node and the destination node of the service. Intra-cluster routing is determined by the intra-cluster controller and forwarded directly. Inter-cluster routing is determined by the inter-cluster controller reading the inter-cluster routing table to determine the destination cluster, selecting the optimal cluster head forwarding path to the destination cluster head, and then the cluster head activates the intra-cluster controller to forward the service. S2. Dynamic cluster head election based on the gray wolf optimization algorithm: Each node is evaluated according to the weighted fitness function value that combines the node's residual energy, congestion, connectivity, and average distance to neighboring nodes. The flight energy consumption, data transmission energy consumption, and data reception and aggregation energy consumption are comprehensively calculated to obtain the node's residual energy. Based on this, the state distance is introduced to simulate the encirclement and attack process of wolf hunting. The node's state vector is updated through multiple iterations, and the nodes with the highest fitness are finally selected as cluster head candidates. S3. Based on the selected cluster head candidates, the dual cluster head (α-β) centers are used as the initial cluster centers to perform membership clustering on the remaining nodes: Initialize the membership matrix, calculate the distance between each node and the dual cluster head centers, and iteratively update the membership and new cluster centers until the clustering results converge; S4. When a routing request arrives, all possible paths are encoded as chromosomes, and the quality of the paths is evaluated using the same weighted fitness function. The population is optimized through genetic operators such as roulette wheel selection, single-point crossover, and random gene mutation to select the optimal route. S5. During network operation, a cluster head load threshold is set. When the current cluster head load exceeds the threshold, the β node will temporarily take over the cluster head responsibilities and switch back after the original cluster head load recovers. When both α and β are not suitable as cluster heads, local re-election will be performed within the cluster first, and global cluster head re-election will be performed if necessary to ensure the balance of cluster head responsibilities and network stability.

2. A dynamic cluster routing method for a highly dynamic UAV swarm network according to claim 1, characterized in that: The fitness function Fi is calculated as follows: F i =w1×E i +w2×C i +w3×N i +w4×D i Among them, E i 、C i 、N i 、D i are the residual energy, congestion, connectivity and average distance of node i respectively, and w1 to w4 are weight coefficients.

3. The method for dynamic cluster routing in a highly dynamic UAV swarm network according to claim 1, characterized in that: The remaining energy E i Calculate as follows: Among them, m i represents the mass of node i, t represents the flight time, V i Indicates the average speed during the flight time, E elec is the energy consumed to transmit 1 bit of data, ε fs and ε mp Respectively represent the energy loss coefficient of the power amplifier circuit under different channel transmission models.

4. The method for dynamic cluster routing in a highly dynamic UAV swarm network according to claim 1, characterized in that: The node congestion C i Calculate as follows: Among them, k j represents the length of the jth data packet in node i, k all Indicates the maximum data capacity of node i.

5. The method for dynamic cluster routing in a highly dynamic UAV swarm network according to claim 1, characterized in that: The node connectivity N i Calculate as follows: The connectivity of a node represents the ratio of the number of neighboring nodes of the node to the total number of nodes in the network.

6. The method for dynamic cluster routing in a highly dynamic UAV swarm network according to claim 1, characterized in that: The average distance D i Calculate as follows: Where D represents the average distance between neighboring nodes within the communication range of the node.

7. The method for dynamic cluster routing in a highly dynamic UAV swarm network according to claim 1, characterized in that: The iterative update of the gray wolf optimization adopts the following formula: X i (t+1)=X j -A i ×D ij A i =2a i ×r 1i -a i In each iteration, the top three nodes in terms of fitness are used as α, β, and δ, and the status of other nodes are updated according to the following formula: When the iteration is completed, the fitness value of each node is calculated again, and the nodes are classified according to the final fitness value. The first c nodes with the highest fitness value (c represents the number of clusters) are marked as α nodes as cluster heads, the next c nodes are marked as β nodes, and the next c nodes are marked as δ nodes. The remaining nodes are marked as ω nodes.

8. The method for dynamic cluster routing in a highly dynamic UAV swarm network according to claim 1, characterized in that: The clustering of the ω nodes is calculated using the membership matrix U and the dual cluster head centers: And according to the formula: Iterate and update until convergence.

9. The method for dynamic cluster routing in a highly dynamic UAV swarm network according to claim 1, characterized in that: The route optimization of the genetic algorithm includes: Chromosomes are sequences of possible path nodes; Roulette wheel selection, single-point crossover and random gene mutation operators are used; The mutation rate is set low to balance diversity and convergence.

10. The method for dynamic cluster routing in a highly dynamic UAV swarm network according to claim 1, characterized in that: The adaptive intra-cluster dual-cluster head switching mechanism is: When the load of cluster head node α exceeds the preset threshold, node β is temporarily promoted to cluster head; When the original α load returns to below the threshold, it switches back to α again; When both α and β are unsuitable, a local re-election is triggered within the cluster, and a global re-election of the cluster head of the entire network is performed when necessary.

Images