Unmanned aerial vehicle cluster network topology control method in high confrontation environment

Abstract

The invention discloses an unmanned aerial vehicle cluster network topology control method in a high confrontation environment, and belongs to the technical field of wireless communication. According to the method, the unmanned aerial vehicle cluster topology is optimized by using two stages of global clustering and local control, and the specific process comprises the following steps: (1) initialization setting of each parameter of an algorithm and setting of a cluster head node of an unmanned aerial vehicle cluster; (2) executing a global clustering method; (3) the unmanned aerial vehicle cluster enters a continuous cluster maintenance stage; and (4) periodically executing a local topology control method. According to the method, the advantages of a clustering type topology control method and a centralized topology control method in the prior art are integrated, and the expandability of an unmanned aerial vehicle system and accurate control over node positions are considered through the method of clustering first and then centralized control; compared with a distributed topology control method in the prior art, information interaction between neighbors is reduced, and energy consumption of unmanned aerial vehicle nodes is reduced; and a reconnection mechanism is provided for the offline node, and the robustness of the unmanned aerial vehicle cluster network is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communication, and particularly relates to a method for controlling the network topology of an unmanned aerial vehicle (UAV) cluster in a highly adversarial environment. Background Art

[0002] In the military field, UAVs, with the advantages of being able to perform tasks in complex environments without direct human risk, are used to execute key tasks such as reconnaissance, surveillance, and target tracking. By carrying advanced reconnaissance equipment, UAVs can penetrate into enemy areas and obtain important intelligence information, providing strong support for military decision-making.

[0003] In the past decade, the UAV system in China has developed rapidly and has become the main combat equipment of the air combat force and a key part of the systemized and intelligent combat. Advanced UAV systems such as high-altitude, high-speed, and long-endurance UAVs will show outstanding strategic penetration capabilities and persistent combat capabilities in future high-risk areas and high-intensity combat areas. However, the UAV cluster network still faces challenges such as dynamic time-varying network topology and possible strong communication interference, which affect the effect of the UAV cluster in performing specific tasks, and may even cause some nodes in the cluster to be unable to communicate with surrounding nodes, resulting in an unconnected topology of the cluster.

[0004] The above technical problems make topology control play a crucial role in the UAV cluster network. Topology control defines the deployment method of UAV nodes and ensures normal communication between nodes, which is the basis for studying other problems of the UAV network. Especially for UAV networks operating in unknown environments, how to improve the connectivity, fault tolerance, and survivability of the network becomes the primary problem to be solved. The goals to be achieved by other topology control algorithms include ensuring stable connectivity in the high-speed dynamic UAV network environment, keeping an appropriate distance between UAVs to avoid collisions, and meeting the signal-to-noise ratio constraint conditions to ensure the quality of service of the communication link; at the same time, maximizing the network coverage to effectively execute tasks, optimizing the energy consumption of UAVs, and extending the overall operation time of the network. Another aspect of topology control research is topology mapping, that is, mapping UAVs from the current topology to a pre-set topology shape. The key problem here is the matching of the UAV positions from the current topology to the target topology.

[0005] In addition to controlling the topology of the UAV cluster, the topology control algorithm also affects the network protocols adopted by UAVs. For example, in terms of cooperation with the MAC layer, the topology control algorithm can optimize the transmission power, reduce the energy consumption of UAVs, and extend the network lifetime. At the same time, the topology control algorithm can also help UAVs manage the neighbor list of nodes, reducing conflicts and interference between data. In terms of cooperation with the network layer, the topology control algorithm can enhance the link stability. It can control the movement mode of UAVs to ensure strong neighborhood relationships and sufficient link durations among UAVs within time slots, reduce retransmissions, and provide a stable neighbor list for the routing protocol to update the routing path. The reasonable cooperation between the topology control algorithm and the network protocol can make the UAV network more efficient and robust.

[0006] In the prior art, topology control methods can be divided into the following three categories.

[0007] (1) Centralized topology control method: In the centralized topology control method, there is a centralized control node, and other nodes act as ordinary nodes to receive instructions from the centralized control node. The centralized method has the characteristics of relatively simple overall architecture and easy implementation. However, since the overall topology of the group is controlled by a single control node, the entire network may not work due to the single-point failure of the control node. At the same time, when the cluster expands, the computational complexity of the control node will also increase significantly, which will lead to poor scalability of the cluster.

[0008] (2) Distributed topology control method: The core idea of the distributed topology control method is that the role of each node is equal. Nodes can independently collect surrounding information and make decisions based on this information. The distributed method does not require a centralized control node with extremely strong computing power. However, the distributed topology control method highly depends on a large amount of exchanged information such as position and speed between neighbor nodes, so it requires small delays in information transmission and reception and small positioning errors between nodes, otherwise nodes will make decisions using inaccurate information.

[0009] (3) Clustering topology control method: In the clustering method, the nodes in the network are divided into multiple clusters. There is a cluster head node in each cluster to manage other member nodes within the cluster. Generally speaking, clustering control can be roughly divided into three stages: cluster formation, cluster head selection, and cluster maintenance. However, the existing clustering algorithms in the prior art cannot accurately control the positions of UAV nodes.

[0010] Currently, there is little research on topology control in high-adversity environments in the prior art, and there is a lack of design for the reconnection mechanism of UAV node disconnection. Summary of the Invention

[0011] The object of the present invention is to overcome the defects of the above-mentioned prior art, provide a method for controlling the network topology of an unmanned aerial vehicle (UAV) cluster in a high-countermeasure environment, provide a reconnection mechanism for dropped UAV nodes, and improve the robustness of the network topology of the UAV cluster.

[0012] The technical problem proposed by the present invention is solved as follows:

[0013] A method for controlling the network topology of an unmanned aerial vehicle (UAV) cluster in a high-countermeasure environment, comprising the following steps:

[0014] Step 1: Initialize and set the parameters of the algorithm, and set the cluster head nodes of the UAV cluster;

[0015] According to the flight area range of the UAV cluster and the number of UAV nodes, preset the number of cluster head nodes; evenly distribute the cluster head nodes within the flight area range, and randomly initialize the positions of the UAV member nodes through a random function;

[0016] Step 2: Global clustering;

[0017] Step 2-1: Cluster formation;

[0018] All UAV member nodes select a cluster to join according to the principle of closer distance, and receive the control messages sent by the cluster head nodes;

[0019] Step 2-2: Redundant cluster head election;

[0020] Comprehensively consider various factors for weighted scoring, and select the UAV member node with the highest weighted score as the redundant cluster head node;

[0021] Step 3: The UAV cluster enters a continuous cluster maintenance stage;

[0022] The cluster maintenance stage includes cluster head handover scenarios, node addition scenarios, node departure scenarios, and node failure scenarios, and corresponding execution processes are set for the four scenarios;

[0023] Step 4: Periodically execute local topology control

[0024] Encode the coordinate information of the positions of the UAV nodes in the current cluster as the initial particles, and use the particle swarm optimization algorithm for optimization; after each round of execution of the particle swarm optimization algorithm, use the crossover, mutation, and selection operations in the genetic algorithm to update the particles, and use the updated particles and randomly generated new particles as the new generation population to execute the next round of the particle swarm optimization algorithm; when the current iteration number reaches the maximum value, the particle swarm optimization algorithm outputs the positions of the current UAV nodes;

[0025] In the particle swarm optimization algorithm, the fitness function Fitness is expressed as:

[0026]

[0027] Among them, x1, x2…, x n represent the position coordinate information of n UAV nodes within the cluster corresponding to the current particle; inf represents infinity, and B d represents the end-to-end delay reward term, and B th represents the throughput reward term, and B dis represents the node-to-node distance reward term, and P nei represents the neighbor number penalty term, and B tar represents the approaching target location reward term; α1 to α5 are weight coefficients and need to satisfy the following relationship:

[0028] |α1| + |α2| + |α3| + |α4| + |α5| = 1

[0029] is the neighbor number penalty term P nei : A penalty is imposed when the average number of neighbors exceeds the set range of the required number of node neighbors; the end-to-end delay reward term B d is calculated as shown in the following formula:

[0030]

[0031] Among them, d pre is the delay prediction value of the intra-cluster topology, and d thr is the set delay threshold value;

[0032] The throughput reward term B th is calculated as shown in the following formula:

[0033]

[0034] Among them, th pre is the throughput prediction value of the intra-cluster topology, and th thr is the set throughput threshold value;

[0035] The node-to-node distance reward term B dis : A penalty is imposed when the average distance between nodes exceeds the set range of the required node-to-node distance;

[0036] The approaching target location reward term B tar : A reward is imposed when the distance between a UAV node and the set target detection node is less than the threshold value;

[0037] In summary, the optimization problem solved by the particle swarm algorithm is described as:

[0038]

[0039] Furthermore, the specific process of step 2-2 is as follows:

[0040] For any UAV member node, the factors considered in the redundant cluster head election process include:

[0041] (1) The remaining energy E of the current UAV member node i ;

[0042] (2) The average distance D between the current UAV member node and other UAV member nodes in the cluster i ;

[0043] (3) The ideal node degree difference Δ i ;

[0044] (4) The backbone network node degree C i ;

[0045] Calculate the weighted scores of each UAV member node, and the identity ID of the UAV member node corresponding to the maximum weighted score i is:

[0046]

[0047] Among them, argmax represents the identity ID of the UAV member node corresponding to the maximum weighted score, E0 is the initial energy of the current UAV member node, and λ1, λ2, λ3, and λ4 respectively represent the weights of the remaining energy, average distance, ideal node degree difference, and backbone network node degree, satisfying the following relationship:

[0048] λ1 + λ2 + λ3 + λ4 = 1 (7)

[0049] The cluster head node elects the UAV member node corresponding to the maximum weighted score as the redundant cluster head node.

[0050] Furthermore, in step 2-2, the specific calculation process for the factor average distance is as follows:

[0051] The distance d i，j (t) between UAV member nodes i and j is:

[0052]

[0053] Among them, t represents the current moment, and x i (t), y i (t), and zi(t) respectively represent the x-axis, y-axis, and z-axis position coordinates of UAV member node i at the current moment, and x j (t), y j (t), and z j (t) respectively represent the x-axis, y-axis, and z-axis position coordinates of UAV member node j at the current moment;

[0054] Calculate the average distance D between the current UAV member node and other UAV member nodes in the cluster i :

[0055]

[0056] where N is the number of other UAV member nodes in the cluster, and 1 ≤ j ≤ N.

[0057] Furthermore, in step 2-2, the specific calculation process for the factor of the ideal node degree difference is as follows:

[0058] The ideal number of members K in a cluster is:

[0059]

[0060] where M is the total number of UAV nodes in the UAV cluster, and calculate the ideal node degree difference β of the current UAV member node i as:

[0061] β i = |n i - K| (4)

[0062] where n i is the number of neighbor nodes of the current UAV member node;

[0063] Normalize β i to obtain the normalized ideal node degree difference Δ i :

[0064]

[0065] Furthermore, in step 3, the cluster head switching scenarios include: (1) Every time a set period passes, the cluster head node calculates its own weighted score, and the cluster head node checks whether the weighted score of the redundant cluster head node is higher; if the weighted score of the redundant cluster head node is higher, then a cluster head switch is required, and a message for switching the cluster head node is broadcast; (2) The redundant cluster head node does not receive the heartbeat information of the cluster head node periodically, subjectively determines that the cluster head node has gone offline, and the redundant cluster head node switches itself to the cluster head node and broadcasts;

[0066] The process of cluster head switching is as follows: After receiving the cluster head switching message of the current cluster, the UAV member node modifies the cluster head node number it belongs to; for the second scenario, after a set time, if the original cluster head node re-establishes contact with the newly switched cluster head node, the original cluster head node modifies its own role to a UAV member node and is regarded as rejoining this cluster; if the original cluster head node has not been online all the time, it is determined that the original cluster head node has gone offline, triggering the node failure scenario and executing the node failure process.

[0067] Furthermore, in step 3, the node joining scenarios include: (1) at initialization, all drone member nodes select a cluster to join; (2) drone member nodes leave the existing cluster and join a new cluster; (3) failed drone member nodes re-search for a cluster head node that can communicate and apply to join the cluster;

[0068] The process of node joining is as follows: after the drone member node finds the cluster it needs to join, it sends an application to join to the cluster head node. The cluster head node updates its own cluster member information and returns the cluster head declaration information. The drone member node joining the new cluster needs to modify the cluster head node number to which it belongs and calculate the weighted score in order to participate in the next round of redundant cluster head node election. If the drone member node has not received the cluster head declaration information after the set time, it will continue to look for the cluster head node to send an application to join until it successfully joins a cluster.

[0069] Furthermore, in step 3, the node leaving scenario means that when the drone member node meets the following conditions at the same time, it chooses to leave the current cluster and join another cluster: (1) the number of drone member nodes in the current cluster is too large; (2) the distance to another cluster head node is closer, and the distance difference is higher than the set distance threshold; (3) it is not a redundant cluster head node;

[0070] The process of node leaving is as follows: when a drone member node leaves the current cluster, it sends an application to join the new cluster head node and executes the node joining process; it sends a node leaving message to the original cluster head node, and the original cluster head node deletes the information of the drone member node.

[0071] Furthermore, in step 3, the node failure scenario refers to the temporary inability of a drone node to communicate with other drone nodes, resulting in a temporary failure, or a permanent failure; drone nodes in the drone cluster use a heartbeat information mechanism to check whether there is a node failure;

[0072] The node failure process is as follows:

[0073] The cluster head node can actively detect the failure of drone member nodes through the heartbeat mechanism, but the drone member nodes cannot actively detect the failure of other drone nodes; if the cluster head node finds that a node has failed, it first sends a message of node failure in the backbone network; the message data part contains the number of the lost drone member node, which is used to notify other cluster head nodes; the cluster head node sends the information of the lost node to the drone member nodes, so that all drone member nodes have the information of the lost node, so that all drone member nodes join the search for the lost node;

[0074] If drone node A finds that it cannot communicate normally with all neighboring nodes, it will determine that it is offline; the current drone node A changes its clustering status to unclustered, and tries to communicate with any drone node at maximum power; if the current drone node A successfully communicates with a drone node B, the successful drone node B will forward the message to retrieve the current drone node A; all drone nodes in the drone cluster will know that the lost node has been found, and the lost node itself will also find a cluster to join and set its role as a drone member node;

[0075] When the redundant cluster head node actively finds that the cluster head node is offline, it executes the cluster head switching process; if it finds that it is offline, it converts its role to a drone member node and executes the node failure process.

[0076] The beneficial effects of the present invention are:

[0077] (1) The method of the present invention combines the advantages of the clustering topology control method and the centralized topology control method in the prior art, and takes into account the scalability of the UAV system and the precise control of the node position through the method of clustering first and then centralized control;

[0078] (2) Compared with the distributed topology control method in the prior art, the method of the present invention reduces the information interaction between neighbors and reduces the energy consumption of drone nodes;

[0079] (3) The method described in the present invention provides a reconnection mechanism for offline nodes, thereby improving the robustness of the drone cluster network. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 Switching pictures for drone characters;

[0081] Figure 2 It is the flow chart of the local topology control algorithm;

[0082] Figure 3 This is a comparison diagram of average path lengths in Example 1;

[0083] Figure 4 This is a throughput comparison chart in Example 1;

[0084] Figure 5 This is a comparison chart of successful submission rates in Example 2. DETAILED DESCRIPTION

[0085] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0086] This embodiment provides a method for controlling the topology of a drone cluster network in a highly confrontational environment, which uses two stages, global clustering and local control, to optimize the topology of the drone cluster. The steps are as follows:

[0087] (1) Initialize and set each parameter of the algorithm, and set the cluster head nodes of the UAV cluster; (2) Execute the global clustering method; (3) The UAV cluster enters the continuous cluster maintenance phase; (4) Periodically execute the local topology control method.

[0088] Step 1 is as follows:

[0089] In the initialization phase, it is necessary to accurately preset the number of cluster head nodes according to the flight area range of the UAV cluster and the number of UAV nodes. The cluster head nodes are roughly evenly distributed within the flight area range, and the positions of the UAV member nodes are randomly initialized through a random function.

[0090] Step 2 is as follows:

[0091] Step 2-1, Cluster formation

[0092] All UAV member nodes need to select a cluster to join according to the principle of closer distance and receive the control messages sent by the cluster head nodes.

[0093] Step 2-2, Redundant cluster head election

[0094] The election method of the redundant cluster head is to conduct a weighted score considering multiple factors, and select the UAV member node with the highest weighted score as the redundant cluster head node;

[0095] For any UAV member node, the factors considered during the redundant cluster head election include:

[0096] (1) The remaining energy E of the current UAV member node i .

[0097] (2) The average distance between the current UAV member node and other UAV member nodes in the cluster:

[0098] The distance d between UAV member nodes i and j i，j (t) is:

[0099]

[0100] where t represents the current moment, and x i (t), y i (t) and z i (t) respectively represent the x-axis, y-axis, and z-axis position coordinates of the UAV member node i at the current moment, and x j (t), y j (t) and z j (t) respectively represent the x-axis, y-axis, and z-axis position coordinates of the UAV member node j at the current moment.

[0101] The average distance D between the current UAV member node and other UAV member nodes within the cluster can be calculated i as follows:

[0102]

[0103] where N is the number of other UAV member nodes within the cluster, and 1 ≤ j ≤ N.

[0104] (3) Ideal node degree difference:

[0105] The ideal number of members K within a cluster is:

[0106] where M is the total number of UAV nodes within the UAV cluster. From this, the ideal node degree difference β of the current UAV member node can be calculated i as follows:

[0107] β i = |n i - K| (4)

[0108] where n i is the number of neighbor nodes of the current UAV member node.

[0109] Normalize β i to obtain the normalized ideal node degree difference Δ i :

[0110]

[0111] (4) Backbone network node degree: The backbone network refers to the network formed by connecting cluster head nodes. The backbone network node degree refers to the number C of other cluster head nodes that the current UAV member node can communicate with i .

[0112] In the present invention, the calculation method of weighted scoring can be expressed by formula (6), where E0 is the initial energy of the current UAV member node, and the identity identifier id of the UAV member node corresponding to the maximum weighted score i is as follows:

[0113]

[0114] where argmax represents the identity identifier of the UAV member node corresponding to the maximum value, and λ1, λ2, λ3, and λ4 respectively represent the weights of the remaining energy, average distance, ideal node degree difference, and backbone network node degree, satisfying the following relationship:

[0115] λ1 + λ2 + λ3 + λ4 = 1 (7)

[0116] The cluster head node elects it as a redundant cluster head node, and the node handover is as follows Figure 1 as shown

[0117] Step 3 is specifically as follows:

[0118] The cluster maintenance phase includes four scenarios: cluster head handover, node joining, node leaving, and node failure.

[0119] (1). For the cluster head handover scenario:

[0120] The cluster head node may not be able to continue working due to too little remaining energy or other reasons. At this time, the redundant cluster head node needs to switch to the role of the cluster head node and take over the work of the original cluster head node. The scenarios where cluster head handover occurs include:

[0121] (1) Every time a fixed period of time T passes, the cluster head node calculates its own weighted score, and the cluster head node checks whether the weighted score of the redundant cluster head node is higher; if the weighted score of the redundant cluster head node is higher, then a cluster head handover is required, and a message for switching the cluster head node is broadcast.

[0122] (2) Normally, the cluster head node periodically sends heartbeat messages to the redundant cluster head node. If the redundant cluster head node does not receive the heartbeat information periodically, it subjectively considers that the cluster head node has gone offline (regardless of whether the cluster head is objectively offline), and the redundant cluster head node switches itself to the cluster head node and broadcasts.

[0123] The process of cluster head handover is as follows: After the UAV member node receives the cluster head handover message of the current cluster, it modifies the cluster head node number to which it belongs. In the second scenario, regardless of whether the original cluster head node is objectively offline, after a period of time, if the original cluster head node re-establishes contact with the newly switched cluster head node, to avoid having two cluster head nodes in the same cluster at the same time, the original cluster head node should modify its own role to that of a UAV member node and be regarded as rejoining this cluster. If the original cluster head node has not been online all the time, it is determined that the original cluster head node has indeed had a problem, and the node failure process is executed.

[0124] (2). For the node joining scenario

[0125] The scenarios where node joining occurs include:

[0126] (1) During initialization, all UAV member nodes need to select a cluster to join.

[0127] (2) During the operation of the UAV cluster, a cluster handover may occur, that is, the UAV member node needs to leave the existing cluster and join a new cluster.

[0128] (3) The failed node re-searches for a communicable cluster head node and applies to join the cluster.

[0129] In the scenario of a node joining, after the UAV member node finds the cluster it needs to join, it sends a join request message to the cluster head node. The cluster head node updates its cluster member information and returns a cluster head declaration message. The UAV member node that joins the new cluster needs to modify the cluster head node number it belongs to and calculate a weighted score to participate in the next round of redundant cluster head node election. If the UAV member node does not receive the cluster head declaration message after a set period of time, it will continuously search for the cluster head node to send a join request message until it successfully joins a cluster.

[0130] (3) For the node leaving scenario

[0131] Since the UAV member node is in constant motion, the UAV member node may move to a place far from the current cluster head node it belongs to, which makes the communication between these two nodes consume more energy. At this time, the UAV member node can consider joining another cluster to avoid this situation.

[0132] When a UAV member node meets the following conditions simultaneously, it can choose to leave the current cluster and join another cluster:

[0133] (1) The number of UAV member nodes in the current cluster is too large;

[0134] (2) The distance to another cluster head node is closer, and the difference between them is higher than the set distance threshold (to avoid excessive cluster switching);

[0135] (3) It is not a redundant cluster head node.

[0136] When the UAV member node leaves the current cluster, it sends a join request message to the new cluster head node and executes the node joining process. Then it sends a node leaving message to the original cluster head node, and the original cluster head node deletes the information of this UAV member node. In this process, the original cluster head node will not attempt to check whether this UAV member node has successfully joined another cluster, and will not return any information after receiving the leaving message of this UAV member node, but only deletes the information of this UAV member node in its own UAV member node information table.

[0137] If for some unexpected reasons, this UAV member node fails to join the new cluster successfully, or the original cluster head node does not receive the leaving information of the UAV member node, the original cluster head node will still keep the information of this UAV member node, but will not receive any information from this UAV member node. Therefore, the original cluster head node will consider this UAV member node to be invalid and enter the node invalidation workflow. In this workflow, the cluster heads will exchange information to find the cluster that this UAV member node has joined, which means that this UAV member node is not invalid and has been found.

[0138] If the redundant cluster head node moves to a position far from other members, then it will be replaced by other UAV member nodes in the next redundant cluster head election and itself will resume the role of a UAV member node. After that, the node can choose to leave the current cluster and join a new cluster.

[0139] (4) For the node failure scenario

[0140] The natural environment and communication conditions faced by the UAV cluster during operation may be relatively complex, which means that UAV nodes may either temporarily fail to communicate with other nodes, resulting in temporary failure; or may permanently fail due to unexpected situations (such as collisions). This requires the UAV cluster to be able to handle different node failure situations.

[0141] UAV nodes in the UAV cluster check for node failures through a heartbeat information mechanism. Since the cluster head node is more important, the heartbeat information interaction between the cluster head node and the redundant cluster head node is more frequent than that between the cluster head node and the UAV member node.

[0142] Different node roles behave differently when dealing with disconnections. The cluster head node can actively detect the failure of UAV member nodes through the heartbeat mechanism, while UAV member nodes do not actively detect the failure of other UAV nodes. Once the cluster head node detects a node failure, it will first send a node failure message in the backbone network. The message data part will contain the number of the lost member node to notify other cluster head nodes. The cluster head node will send the information of the lost node to the UAV member nodes, so that all UAV member nodes have the information of the lost node and all UAV member nodes in the network join the search for the lost node.

[0143] If a drone node A (regardless of its role) finds that it cannot communicate properly with all its neighbor nodes, it will consider itself offline. At this time, the current drone node A modifies its clustering status to unclustered and attempts to communicate with any drone node at the maximum power. If the current drone node A successfully communicates with a drone node B (node B discovers that the node number of this node is the lost node A and its status is unclustered), the successfully communicating drone node B will forward the message to retrieve the current drone node A (if node B is a member node, it will report the message to the cluster head node, and the cluster head node will forward it within the backbone network). The message to retrieve the node has the same message type as the node failure message, but fills in the information of the retrieved node in the data part and supplements the new cluster head sequence number information. After completing this process, all nodes in the drone cluster will know that the lost node has been retrieved, and the lost node itself can also find a suitable cluster to join. Regardless of the role of the lost node before it was lost, it is set as a drone member node after rejoining the drone cluster.

[0144] When the redundant cluster head node actively discovers that the cluster head node is offline, it will execute the cluster head switching process described above; if it discovers that it is offline, it will convert its role to a drone member node and execute the above node failure process. At the same time, to ensure that there are redundant cluster heads in the cluster, a new redundant cluster head election needs to be carried out.

[0145] Step 4 is specifically as follows:

[0146] Based on the particle swarm algorithm to achieve local topology control, the schematic diagram of its process is as Figure 2 shown; the coordinate information of the positions of the drones in the current cluster is encoded as the initial particles, and the particle swarm algorithm is used for optimization. After each round of the particle swarm algorithm is executed, the particles are updated using the crossover, mutation, and selection operations in the genetic algorithm, and these particles and randomly generated new particles are used as the new generation population to execute the next round of the particle swarm algorithm. When the current iteration number reaches the maximum value, the algorithm outputs the positions of the current drone nodes.

[0147] In the particle swarm algorithm, the fitness function Fitness is expressed as:

[0148]

[0149] where x1, x2…, x n represents the coordinate information of the positions of n drone nodes in the cluster corresponding to the current particle; inf represents infinity, B d represents the end-to-end delay reward term, B th represents the throughput reward term, B dis represents the node-to-node distance reward term, P nei represents the neighbor number penalty term, B tarIndicates the reward item for approaching the target location; α1 to α5 are weight coefficients, and the following relationship needs to be satisfied:

[0150] |α1| + |α2| + |α3| + |α4| + |α5| = 1 (9)

[0151] Next, the fitness function will be explained:

[0152] When the topology within the cluster represented by the particle is not connected, directly set the fitness of the particle to infinity. When the topology represented by any particle is connected, this unconnected scheme will be eliminated by the algorithm. This is to ensure the bottom line of the algorithm output result, that is, the calculated topology is connected. When the topology is connected, the topology is evaluated by several other dimensions:

[0153] P nei Is the neighbor number penalty term. The required range of the average neighbor number of nodes can be adjusted according to the actual situation, and a penalty is imposed when the average neighbor number exceeds the set required range.

[0154] B d Represents the end-to-end delay reward term. The calculation method of the end-to-end delay reward term is shown in Equation (10):

[0155]

[0156] Among them, d pre Is the delay prediction value of the topology within the cluster, and d thr Is the set delay threshold. When d pre Exceeds d thr Set this item to infinity, meaning that such a topology does not meet the requirements and abandon this scheme in the next round of algorithm iteration.

[0157] B th Is the throughput reward term. The calculation method of the throughput reward term is shown in Equation (11):

[0158]

[0159] Among them, th pre Is the throughput prediction value of the topology within the cluster, and th thr Is the set throughput threshold.

[0160] B dis Is the node - to - node distance reward term. The setting of the required range of the node - to - node distance should be set according to the scenario in the actual application. When the average node - to - node distance exceeds the set required range, a penalty is imposed.

[0161] B tarIt is a reward item for approaching the target location. When the distance between a drone node and the set target detection node is less than the threshold value, a reward item is added to the fitness function to encourage the node to approach the target location.

[0162] In summary, the optimization problem finally solved by the particle swarm algorithm can be written as:

[0163]

[0164] Embodiment 1

[0165] Under the performance simulation parameter settings shown in Table 1, the average path length of the method described in this embodiment is relatively small compared to the Adaptive K-means algorithm, which means that the network latency is small, as Figure 3 shown. Under the same parameter settings, the throughput of the network topology formed by the method for controlling the network topology of the UAV cluster in the high-confrontation battlefield environment described in this embodiment is higher, as Figure 4 shown.

[0166] Table 1 Performance simulation parameter settings

[0167]

[0168] Embodiment 2

[0169] Under the performance simulation parameter settings shown in Table 2, in order to study the anti-destruction characteristics of the topology generated by the algorithm, the following scenarios were specially set up for simulation experiments: A sending node was set at the lower left corner of the simulation area, and a receiving node was set at the upper right corner. The UAV network was used as a relay in the simulation area to be responsible for forwarding the messages sent by the sending node to the receiving node. After the UAV network topology was determined, a certain node failure rate was used to make some randomly selected nodes in the topology no longer work properly, and then it was checked whether the receiving node could receive the information sent by the sending node. In each scenario with different failure rates, multiple experiments were repeated to simulate the situation where different nodes failed under the same failure rate, and the probability that the receiving node could successfully receive the message (hereinafter referred to as the successful delivery rate) was calculated respectively, and then the average value of 1000 experimental data was taken.

[0170] As Figure 5 shown, in this scenario, when the node failure rate is less than 15%, the method for controlling the network topology of the UAV cluster in the high-confrontation battlefield environment described in this embodiment can still ensure a successful delivery rate of more than 90%. When the node failure rate reaches 50%, the method of the present invention can still maintain a successful delivery rate of 29%, while the successful delivery rate of the Adaptive K-means method is only 2.4%.

[0171] Table 2 Anti-destruction experiment parameter settings

[0172]

Claims

1. A method for controlling the network topology of an unmanned aerial vehicle (UAV) cluster in a high confrontation environment, characterized in that, It includes the following steps: Step 1: Initialize and set the parameters of the algorithm, and set the cluster head nodes of the UAV cluster; According to the flight area range of the UAV cluster and the number of UAV nodes, preset the number of cluster head nodes; Evenly distribute the cluster head nodes within the flight area range, and randomly initialize the positions of the UAV member nodes through a random function; Step 2: Global clustering; Step 2-1: Cluster formation; All UAV member nodes select a cluster to join according to the principle of closer distance and receive the control messages sent by the cluster head nodes; Step 2-2: Redundant cluster head election; Perform weighted scoring considering multiple factors, and select the UAV member node with the highest weighted score as the redundant cluster head node; Step 3: The UAV cluster enters the continuous cluster maintenance stage; The cluster maintenance stage includes four scenarios: cluster head switching scenario, node joining scenario, node leaving scenario, and node failure scenario, and corresponding execution processes are set for the four scenarios; Step 4: Periodically execute local topology control Encode the coordinate information of the positions of the UAV nodes in the current cluster as the initial particles, and use the particle swarm optimization algorithm for optimization; after each round of the particle swarm optimization algorithm is executed, use the crossover, mutation, and selection operations in the genetic algorithm to update the particles, and use the updated particles and randomly generated new particles as the new generation population to execute the next round of the particle swarm optimization algorithm; when the current iteration number reaches the maximum value, the particle swarm optimization algorithm outputs the positions of the current UAV nodes; In the particle swarm optimization algorithm, the fitness function Fitness is expressed as: Among them, x1, x2,..., x n represent the position coordinate information of n UAV nodes within the cluster corresponding to the current particle; inf represents infinity, and B d represents the end-to-end delay reward term, and B th represents the throughput reward term, and B dis represents the inter-node distance reward term, and P nei represents the neighbor number penalty term, and B tar represents the reward term for approaching the target location; α1 to α5 are weight coefficients, and the following relationships need to be satisfied: |α1|+|α2|+|α3|+|α4|+|α5|＝1 For the neighbor number penalty term P nei : Penalty is imposed when the average neighbor number exceeds the set range of the node neighbor number requirement; End-to-end delay reward item B d is calculated as shown in the following formula: Among them, d pre is the delay prediction value of the intra-cluster topology, and d thr is the set delay threshold value; Throughput Reward Item B th The calculation method is shown in the following formula: Among them, th pre is the throughput prediction value of the intra-cluster topology, and th thr is the set throughput threshold value; Node - to - node distance reward term B dis : A penalty is imposed when the average distance between nodes exceeds the required range of the set node - to - node distance; Proximity to Target Location Reward Item B tar : Apply a reward when the distance between a drone node and a set target detection node is less than the threshold value; In summary, the optimization problem solved by the particle swarm optimization algorithm is described as:

2. The method for controlling the network topology of an unmanned aerial vehicle cluster in a high confrontation environment according to claim 1, wherein The specific process of Step 2-2 is: For any UAV member node, the factors considered during the redundant cluster head election process include: (1) The remaining energy E of the current UAV member node i ; (2) The average distance D between the current UAV member node and other UAV member nodes within the cluster i ; (3) Ideal node degree difference Δ i ; (4) Backbone network node degree C i ; Calculate the weighted scores of each UAV member node, and the identity ID of the UAV member node corresponding to the maximum weighted score i is as follows: Among them, argmax represents the identity identifier of the UAV member node corresponding to the maximum weighted score, E0 is the initial energy of the current UAV member node, and λ1, λ2, λ3, and λ4 respectively represent the weights of the remaining energy, average distance, ideal node degree difference, and backbone network node degree, and satisfy the following relationship: λ1 + λ2 + λ3 + λ4 = 1 (7) The cluster head node elects the UAV member node corresponding to the maximum weighted score as the redundant cluster head node.

3. The method for controlling the network topology of an unmanned aerial vehicle cluster in a high confrontation environment according to claim 1, wherein, In Step 2-2, the specific calculation process for the factor average distance is: The distance d between the UAV member nodes i and j i，j (t) is as follows: Among them, t represents the current moment, x i (t), y i (t) and z i (t) respectively represent the x-axis, y-axis and z-axis position coordinates of the UAV member node i at the current moment, x j (t), y j (t) and z j (t) respectively represent the x-axis, y-axis and z-axis position coordinates of the UAV member node j at the current moment; Calculate the average distance D between the current UAV member node and other UAV member nodes within the cluster i : Among them, N is the number of other UAV member nodes in the cluster, and 1 ≤ j ≤ N.

4. The method for controlling the network topology of an unmanned aerial vehicle cluster in a high confrontation environment according to claim 1, wherein In Step 2-2, the specific calculation process for the factor ideal node degree difference is: The ideal number of members K in a cluster is: Where M is the total number of UAV nodes in the UAV cluster, and the ideal node degree difference β of the current UAV member node is calculated i as follows: β i = |n i - K| (4) where n i is the number of neighbor nodes of the current UAV member node; For β i perform normalization to obtain the normalized ideal node degree difference Δ i :

5. The method for controlling the network topology of an unmanned aerial vehicle cluster in a high confrontation environment according to claim 1, wherein, In Step 3, the cluster head switching scenario includes: (1) Without passing through the set period, the cluster head node calculates its own weighted score, and the cluster head node checks whether the weighted score of the redundant cluster head node is higher; if the weighted score of the redundant cluster head node is higher, then cluster head switching is required, and a cluster head switching node message is broadcast; (2) The redundant cluster head node does not receive the heartbeat information of the cluster head node periodically, subjectively determines that the cluster head node has gone offline, and the redundant cluster head node switches itself to the cluster head node and broadcasts; The process of cluster head switching is as follows: after receiving the cluster head switching message of the current cluster, the UAV member node modifies the cluster head node number to which it belongs; for scenario (2), after a set time, if the original cluster head node reconnects with the new cluster head node after switching, the original cluster head node modifies its role to a UAV member node and is considered to have rejoined the cluster; If the original cluster head node has not been online, it is determined that the original cluster head node is offline, triggering the node failure scenario and executing the node failure process.

6. The method for controlling the network topology of an unmanned aerial vehicle cluster in a high confrontation environment according to claim 1, wherein In step 3, the node joining scenarios include: (1) at initialization, all drone member nodes select a cluster to join; (2) drone member nodes leave the existing cluster and join a new cluster; (3) failed drone member nodes re-search for a cluster head node that can communicate and apply to join the cluster; The process of node joining is as follows: after the drone member node finds the cluster it needs to join, it sends an application to join to the cluster head node. The cluster head node updates its own cluster member information and returns the cluster head declaration information. The drone member node joining the new cluster needs to modify the cluster head node number to which it belongs and calculate the weighted score in order to participate in the next round of redundant cluster head node election. If the drone member node has not received the cluster head declaration information after the set time, it will continue to look for the cluster head node to send an application to join until it successfully joins a cluster.

7. The method for controlling the network topology of an unmanned aerial vehicle cluster in a high confrontation environment according to claim 1, wherein, In step 3, the node leaving scenario means that when the drone member node meets the following conditions at the same time, it chooses to leave the current cluster and join another cluster: (1) the number of drone member nodes in the current cluster is too large; (2) the distance to another cluster head node is closer, and the distance difference is higher than the set distance threshold; (3) it is not a redundant cluster head node; The process of node leaving is as follows: when a drone member node leaves the current cluster, it sends an application to join the new cluster head node and executes the node joining process; it sends a node leaving message to the original cluster head node, and the original cluster head node deletes the information of the drone member node.

8. The method for controlling the network topology of an unmanned aerial vehicle cluster in a high confrontation environment according to claim 1, wherein, In step 3, the node failure scenario refers to the temporary inability of a drone node to communicate with other drone nodes, resulting in a temporary failure, or a permanent failure. The drone nodes in the drone cluster use the heartbeat information mechanism to check whether there is a node failure. The node failure process is as follows: The cluster head node can actively detect the failure of the drone member node through the heartbeat mechanism, but the drone member node cannot actively detect the failure of other drone nodes; if the cluster head node finds that a node fails, it first sends a node failure message in the backbone network; the message data part contains the number of the lost drone member node, which is used to notify other cluster head nodes; The cluster head node sends the information of the lost node to the drone member nodes, so that all drone member nodes have the information of the lost node and join in the search for the lost node; If the UAV node A finds that it cannot communicate normally with all its neighbor nodes, it determines that it has dropped offline; the current UAV node A modifies its clustering status to unclustered and attempts to communicate with any UAV node at the maximum power; if the current UAV node A successfully communicates with a UAV node B, the successfully communicating UAV node B will forward the message to retrieve the current UAV node A; all UAV nodes in the UAV cluster will learn that the lost node has been retrieved, and the lost node itself also finds a cluster to join and sets its role as a UAV member node; When the redundant cluster head node actively discovers that the cluster head node has dropped offline, it executes the cluster head handover process; If it finds that it has dropped offline, it converts its role to a UAV member node and executes the node failure process.

Images