A suboptimal cluster head selection method for UAV swarm data links based on K-means

Abstract

This invention provides a method for selecting suboptimal cluster heads in a UAV swarm data link based on K-means clustering, comprising the following steps: the UAV swarm data link performs self-checks and automatically forms a network; based on the K-means clustering algorithm, the terminal devices of the UAV swarm are clustered; terminal device characteristic parameters are loaded, and based on these parameters, the DIST value of each terminal device within the cluster relative to itself and other terminal devices is calculated, generating a DIST matrix; based on the DIST matrix of the terminal device clusters, the terminal devices are ranked as suboptimal cluster heads, and a suboptimal cluster head is selected; the suboptimal cluster head ranking is refreshed at set intervals, and the cluster heads are updated; the terminal device characteristic parameters of the normally operating terminal device cluster are stored and sent to the central control center as training samples for the UAV swarm. The beneficial effects of this invention are: real-time identification of suboptimal cluster heads in the UAV swarm; reduced network outage risk; and enhanced network robustness.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) swarm technology, and in particular relates to a method for suboptimal selection of cluster heads in UAV swarm data links based on K-means. Background Technology

[0002] Currently, UAV swarm data links with cluster heads experience the problem of cluster heads disappearing abnormally in complex environments. The UAV swarm data link is a crucial component of the UAV system, primarily responsible for transmitting control information, payload-free communication, and payload communication, and for networking communication in complex environments. The physical channel between more than two nodes is called a swarm communication link.

[0003] The latest UAV swarm data link supports multi-point networking and self-networking; simultaneously, the application layer supports the AODV algorithm, the hello handshake protocol, and data queuing algorithms, facilitating rapid and robust UAV swarm LAN networking. This application aims to solve the problem of poor networking flexibility in traditional UAV data links when cluster heads are lost in highly dynamic and highly adversarial environments, especially meeting the needs of time-sensitive targets and cross-platform real-time collaboration. As the underlying communication pillar, it supports seamless links across multiple domains including land, sea, air, space, cyber, and electromagnetic. When a cluster head goes offline abnormally, this invention can effectively and quickly restore cluster head communication. Summary of the Invention

[0004] In view of this, the present invention aims to propose a suboptimal cluster head selection method based on K-means UAV swarm data link to solve the problem of the inability to quickly select cluster heads when a local single UAV loses network connection during the flexible networking operation of UAV swarm data link.

[0005] To achieve the above objectives, the technical solution of the present invention is implemented as follows: A method for suboptimal cluster head selection in UAV swarm data links based on K-means includes the following steps: S1. After powering on, the drone cluster data link performs a self-test and forms a network on its own; S2. Based on the K-means clustering algorithm, the terminal devices of the UAV swarm are clustered to form terminal device clusters; S3. The terminal loads the terminal characteristic parameters and, based on the terminal characteristic parameters, calculates the DIST value of each terminal in the terminal cluster relative to itself and other terminals, and generates the DIST matrix. S4. Based on the DIST matrix of the terminal clusters, rank the suboptimal cluster heads of the terminal clusters, generate a suboptimal cluster head table, and select the cluster heads. S5. At set intervals, repeat steps S2 to S4 to refresh the ranking of the second-best cluster heads and update the second-best cluster heads. S6. Store the terminal characteristic parameters of the terminal cluster during normal operation and send them to the central control center as learning and training samples for the UAV cluster. In step S3, the terminal characteristic parameters include distance parameters, importance parameters, and importance weighting coefficients.

[0006] Furthermore, in step S2, based on the K-means clustering algorithm, the terminal devices of the UAV swarm are clustered to form terminal device clusters, including: S21. Select K initial cluster center endpoints; S22. Assign each terminal to the nearest cluster center terminal; S23. Calculate and update the cluster center terminal; S24. Repeat steps S22 to S23 until the cluster center terminal is stable.

[0007] Furthermore, in step S3, the calculation expression for DIST is as follows: ; In the formula, Let j be the DIST value of terminal j relative to terminal i. Here is the distance parameter value for terminal i. Let j be the distance parameter value of the terminal. For the importance parameter value of terminal i, For the importance parameter value of terminal j, The weighting factor for the importance of terminal j relative to terminal i; The expression for calculating the DIST matrix is ​​as follows: ; In the formula, Let DIST be a matrix containing k endpoints. The elements in the matrix represent the DIST values ​​between two endpoints.

[0008] Furthermore, in step S4, based on the DIST matrix of the terminal clusters, the suboptimal cluster heads of the terminals are ranked, a suboptimal cluster head table is generated, and the suboptimal cluster heads are selected, including: S41. Obtain the DIST value of each terminal as the center; S42. Rank each terminal based on its DIST value, with each terminal as the center. The smaller the DIST value, the higher the ranking. Generate a suboptimal cluster head table. S43, the top-ranked terminal, was selected as the second-best cluster head.

[0009] Furthermore, in step S41, the DIST value centered on each terminal is obtained, including: The minimum DIST value in the first row of the DIST matrix is ​​used as the DIST value centered on terminal 1. The minimum DIST value in the second row of the DIST matrix is ​​used as the DIST value centered on terminal 2. Similarly, the DIST value of each terminal is obtained as the center.

[0010] Furthermore, the terminal device characteristic parameters are obtained through training by the UAV swarm and change in real time as the terminal device's operating environment and configuration change.

[0011] Compared with existing technologies, the suboptimal cluster head selection method based on K-means UAV swarm data link described in this invention has the following advantages: By using the K-means algorithm and real-time updated terminal characteristic parameter values, suboptimal cluster head terminals are selected. Each terminal prepares suboptimal cluster head information. If a cluster head terminal loses network access at a certain moment, the prepared suboptimal cluster head terminal is immediately selected as the actual cluster head terminal, saving the cluster head voting time after network outage and enhancing network robustness. Attached Figure Description

[0012] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a schematic diagram of the initiator entering the network according to a conventional embodiment of the present invention; Figure 2 This is a schematic diagram of the response terminal entering the network according to a conventional embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the initiation of communication between the terminal and the terminal in a conventional embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the communication link establishment between terminal devices according to a conventional embodiment of the present invention; Figure 5 This is a schematic diagram illustrating the maintenance of the UAV cluster data link neighbor table and routing protocol according to a conventional embodiment of the present invention; Figure 6 This is a schematic diagram of the initial state of cluster head selection according to an embodiment of the present invention; Figure 7 This is a schematic diagram of the first cluster head selection according to an embodiment of the present invention; Figure 8 This is a schematic diagram of the second cluster head selection according to an embodiment of the present invention; Figure 9 This is a schematic diagram of the third cluster head selection according to an embodiment of the present invention; Figure 10 This is a schematic diagram illustrating the network setup as described in an embodiment of the present invention; Figure 11 This is a schematic diagram of the K-means algorithm clustering described in an embodiment of the present invention; Figure 12 This is a schematic diagram illustrating the selection of suboptimal cluster heads in K-means clustering according to an embodiment of the present invention; Figure 13 This is an illustration of the updated neighbor table, routing table, and suboptimal cluster head representation as described in the embodiments of the present invention; Figure 14 This is an illustration of the cluster head offline update neighbor table, routing table, and suboptimal cluster head representation as described in the embodiments of the present invention. Detailed Implementation

[0013] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0014] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0015] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0016] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0017] Example 1 like Figures 1 to 14 As shown, a method for suboptimal cluster head selection in UAV swarm data links based on K-means includes the following steps: S1. After powering on, the drone cluster data link performs a self-test and forms a network on its own; S2. Based on the K-means clustering algorithm, the terminal devices of the UAV swarm are clustered to form terminal device clusters; S3. The terminal loads the terminal characteristic parameters and, based on the terminal characteristic parameters, calculates the DIST value of each terminal in the terminal cluster relative to itself and other terminals, and generates the DIST matrix. S4. Based on the DIST matrix of the terminal clusters, rank the suboptimal cluster heads of the terminal clusters, generate a suboptimal cluster head table, and select the cluster heads. S5. At set intervals, repeat steps S2 to S4 to refresh the ranking of the second-best cluster heads and update the second-best cluster heads. S6 stores the terminal characteristic parameters of the drone cluster during normal operation and sends them to the central control center as learning and training samples for the drone cluster.

[0018] Specifically as follows: I. Data Link Networking for Unmanned Aerial Vehicle (UAV) Clusters The UAV swarm data link channel access control protocol adopts a queuing protocol, possessing capabilities such as temporary, dynamic, and rapid network formation, distributed dynamic access, support for simultaneous communication by multiple nodes, QoS guarantees that prioritize various services, and immediate channel access for the highest priority services. Key technologies include: a distributed, fast, and secure network access authentication algorithm; a channel busy / idle level detection and prediction algorithm based on time-frequency analysis; a multi-channel distributed access algorithm with conflict decomposition; a multi-channel access algorithm based on channel busy / idle level and prioritizing services; and a variable factor backoff algorithm based on load feedback and real-time backoff window updates.

[0019] The drone swarm data link repeatedly searches for a cluster head using known access points and selects one based on their feedback response values. The cluster head stores and can adjust its routing table. Access is completed using the routing table stored in the control cluster head.

[0020] Conventional access rules: Unmanned aerial vehicle (UAV) swarm data link networking utilizes the AODV algorithm, an adaptive routing protocol based on distance-vector routing, for routing selection in wireless ad hoc networks (MANETs). The implementation process of the AODV protocol is as follows: like Figure 1 As shown, a node broadcasts an RREQ (Route Request) message: when a node needs to send a data packet to a destination, it broadcasts an RREQ message to find a route to the destination. This RREQ message contains the source node address, the destination node address, and a unique sequence number to prevent message loops.

[0021] like Figure 2 As shown, a node responds to an RREQ message: When a node receives an RREQ message, it checks its routing table. If there is no route to the destination, it forwards the RREQ message to its neighboring nodes. If the node already has a route to the destination, it sends an RREP (Route Reply) message to the source node.

[0022] like Figure 3 As shown, nodes broadcast RREP messages: when a destination node receives an RREQ message, it sends an RREP message to the source node, containing the shortest path to the source node. Each node caches the RREP message before forwarding it for later use.

[0023] Nodes maintain routing tables: Each node maintains a routing table to store routing information to other nodes. Whenever a node receives an RREP message, it updates its routing table and broadcasts the updated routing information to its neighboring nodes.

[0024] like Figure 4 As shown, nodes periodically send HELLO messages: Nodes periodically send HELLO messages to check if neighboring nodes are still present. If a node does not receive HELLO messages from its neighboring nodes several times in a row, it considers the node to be offline and updates its routing table.

[0025] The main advantages of the AODV protocol are its adaptability and efficiency. Because AODV only establishes routes when needed, it effectively reduces control traffic in the network. Furthermore, AODV is adaptive, dynamically adjusting routes based on network topology and traffic load.

[0026] like Figure 5 As shown, the core idea of ​​the drone swarm data link networking protocol is to reduce routing overhead in the network through on-demand routing. When a source node needs to send data to a destination node, it sends a route request (RREQ) to surrounding nodes. If a node knows how to reach the destination node, it sends a route response (RREP) to the source node. The source node will use the information in the route response to send data to the destination node. In this process, each node maintains a routing table, recording the shortest path to the destination node and the next-hop node.

[0027] II. Selection of Cluster Head for UAV Swarm Data Link Networking Step 1: As Figure 6 As shown, during network initialization, nodes are assigned unique IDs, calculate their own weights, and broadcast hello packets within time slots. The main contents of the hello packet are: the node's ID, identity, weight, and geographical location information (longitude, latitude, and altitude). Step 2: As Figure 7 As shown, the weights of all neighboring nodes (including the current node) are compared. The node with the smallest weight is selected as the cluster head node, and a hello packet is broadcast. Step 3: As Figure 8 As shown, determine whether the selected cluster head node is a valid cluster head. If it is, proceed to Step 5; otherwise, select a cluster head a second time. Step 4: As Figure 9 As shown, determine whether the selected cluster head node is a valid cluster head. If it is, proceed to Step 5; otherwise, select a cluster head a third time. Step 5: As Figure 10 As shown, the cluster head was successfully selected, and a hello packet was broadcast. Step 6: As Figure 10 As shown, determine the gateway node; Step 7: As Figure 10 As shown, a backbone network is established; Step 8: The network is complete, and the cluster head broadcasts information about all cluster members within its cluster.

[0028] III. Specific Steps Step 1: Based on the K-means clustering algorithm, cluster the terminal devices of the drone swarm to form terminal device clusters.

[0029] Step 2: The terminal loads the terminal characteristic parameters, and based on the terminal characteristic parameters, calculates the DIST value of each terminal in the terminal cluster relative to itself and other terminals, and generates the DIST matrix.

[0030] Terminal characteristic parameters include distance parameters, importance parameters, and importance weighting coefficients.

[0031] The terminal device characteristic parameters are obtained through training by the UAV swarm and change in real time as the terminal device's operating environment and configuration change.

[0032] The expression for calculating the DIST value of terminal j relative to terminal i is as follows: ; The expression for terminal i relative to itself and the various DIST values ​​of other terminals is as follows: ; In the formula, Let j be the DIST value of terminal j relative to terminal i. Here is the distance parameter value for terminal i. Let j be the distance parameter value of the terminal. For the importance parameter value of terminal i, For the importance parameter value of terminal j, The weighting factor for the importance of terminal j relative to terminal i; The expression for the DIST matrix is ​​as follows: ; In the formula, Let DIST be a matrix containing k endpoints. The elements in the matrix represent the DIST values ​​between two endpoints.

[0033] Step 3: Based on the DIST matrix of the terminal clusters, rank the suboptimal cluster heads of the terminal clusters, generate a suboptimal cluster head table, and select the suboptimal cluster heads.

[0034] The minimum DIST value in the first row of the DIST matrix is ​​used as the DIST value centered on terminal 1; the minimum DIST value in the second row of the DIST matrix is ​​used as the DIST value centered on terminal 2; and so on, obtaining the DIST value centered on each terminal. Based on the DIST values ​​centered on each terminal, the terminals are ranked, with smaller DIST values ​​ranking higher. The terminal ranked first is selected as the cluster head.

[0035] The expression for ranking the second-best cluster heads is as follows: ; ; In the formula, The DIST value centered on terminal r is here. This means ranking the DIST values ​​of terminal r relative to itself and other terminal r, and taking the minimum value as the DIST value centered on terminal r. The DIST value of the cluster head, here. This indicates that the DIST values ​​of each terminal are ranked, and the minimum value is taken as the DIST value of the cluster head.

[0036] Step 4: Refresh the ranking of the second-best cluster heads every set time interval and update the cluster heads.

[0037] Step 5: Store the terminal characteristic parameters of the terminal cluster during normal operation and send them to the central control center as learning and training samples for the drone cluster.

[0038] Example 1: like Figure 13 and Figure 14As shown, the terminal devices refresh every 0.1ms. If the cluster head unexpectedly loses network access, the LAN immediately selects a suboptimal cluster head, with the green terminal device being the suboptimal cluster head. The parameter table is updated in real time to prepare for the next suboptimal cluster head selection. In the parameter table, all terminal devices in the network are numbered sequentially (A~Z). The terminal numbers that can be directly reached between terminal devices form the neighbor table; the terminal forwarding information and the terminal devices that perform forwarding actions form the routing table; if the current cluster head unexpectedly goes offline, a suboptimal cluster head terminal device is recommended according to the suboptimal cluster head selection method. In the suboptimal cluster head table, the earlier a terminal device appears, the more likely it is to become the new cluster head.

[0039] Example 2 The present invention can also have a second embodiment, as long as the solution can achieve the purpose of the present invention.

[0040] A method for suboptimal cluster head selection in UAV swarm data links based on K-means includes the following steps: A1. Define the parameters of the K-means algorithm; A2. Define the characteristics of the terminal; A3. Define terminal signals using terminal characteristics; A4. Using the known sample training data, obtain the parameter values ​​suitable for the k-means algorithm; A5. Use the K-means algorithm to cluster the samples in the dataset; A6. Calculate the terminal signal, rank the cluster heads based on the terminal signal, and update the cluster center of the terminal. A7. Measure the sample distance and mark noise points or classify samples; A8. Adjust the time for refreshing cluster head rankings; A9. Wait for the cluster head ranking, update the parameter table in real time, select the second-best cluster head in real time, and determine whether the cluster center has changed. If so, skip to step A5. A10. Determine if the time to refresh the cluster head ranking has been reached. If so, proceed to step A5. A11. Determine whether the work has ended. If so, store the variable value and pass it down to the central control unit. Otherwise, skip to step A8.

[0041] The specific implementation method is as follows: 1. UAV swarm data link networking access rules: The UAV swarm data link channel access control protocol adopts a queuing protocol, possessing capabilities such as temporary, dynamic, and rapid network formation, distributed dynamic access, support for simultaneous communication by multiple nodes, QoS guarantees that prioritize various services, and immediate channel access for the highest priority services. Key technologies include: a distributed, fast, and secure network access authentication algorithm; a channel busy / idle level detection and prediction algorithm based on time-frequency analysis; a multi-channel distributed access algorithm with conflict decomposition; a multi-channel access algorithm based on channel busy / idle level and prioritizing services; and a variable factor backoff algorithm based on load feedback and real-time backoff window updates.

[0042] The drone swarm data link repeatedly searches for a cluster head using known access points and selects one based on their feedback response values. The cluster head stores and can adjust its routing table. Access is completed using the routing table stored in the control cluster head.

[0043] Conventional access rules: Unmanned aerial vehicle (UAV) swarm data link networking utilizes the AODV algorithm, an adaptive routing protocol based on distance-vector routing, for routing selection in wireless ad hoc networks (MANETs). The implementation process of the AODV protocol is as follows: like Figure 1 As shown, a node broadcasts an RREQ (Route Request) message: when a node needs to send a data packet to a destination, it broadcasts an RREQ message to find a route to the destination. This RREQ message contains the source node address, the destination node address, and a unique sequence number to prevent message loops.

[0044] like Figure 2 As shown, a node responds to an RREQ message: When a node receives an RREQ message, it checks its routing table. If there is no route to the destination, it forwards the RREQ message to its neighboring nodes. If the node already has a route to the destination, it sends an RREP (Route Reply) message to the source node.

[0045] like Figure 3 As shown, nodes broadcast RREP messages: when a destination node receives an RREQ message, it sends an RREP message to the source node, containing the shortest path to the source node. Each node caches the RREP message before forwarding it for later use.

[0046] Nodes maintain routing tables: Each node maintains a routing table to store routing information to other nodes. Whenever a node receives an RREP message, it updates its routing table and broadcasts the updated routing information to its neighboring nodes.

[0047] like Figure 4 As shown, nodes periodically send HELLO messages: Nodes periodically send HELLO messages to check if neighboring nodes are still present. If a node does not receive HELLO messages from its neighboring nodes several times in a row, it considers the node to be offline and updates its routing table.

[0048] The main advantages of the AODV protocol are its adaptability and efficiency. Because AODV only establishes routes when needed, it effectively reduces control traffic in the network. Furthermore, AODV is adaptive, dynamically adjusting routes based on network topology and traffic load.

[0049] like Figure 5 As shown, the core idea of ​​the drone swarm data link networking protocol is to reduce routing overhead in the network through on-demand routing. When a source node needs to send data to a destination node, it sends a route request (RREQ) to surrounding nodes. If a node knows how to reach the destination node, it sends a route response (RREP) to the source node. The source node will use the information in the route response to send data to the destination node. In this process, each node maintains a routing table, recording the shortest path to the destination node and the next-hop node.

[0050] 2. Selection of standard UAV swarm data link networking cluster head Step 1: As Figure 6 As shown, during network initialization, nodes are assigned unique IDs, calculate their own weights, and broadcast hello packets within time slots. The main contents of the hello packet are: the node's ID, identity, weight, and geographical location information (longitude, latitude, and altitude). Step 2: As Figure 7 As shown, the weights of all neighboring nodes (including the current node) are compared. The node with the smallest weight is selected as the cluster head node, and a hello packet is broadcast. Step 3: As Figure 8 As shown, determine whether the selected cluster head node is a valid cluster head. If it is, proceed to Step 5; otherwise, select a cluster head a second time. Step 4: As Figure 9 As shown, determine whether the selected cluster head node is a valid cluster head. If it is, proceed to Step 5; otherwise, select a cluster head a third time. Step 5: As Figure 10 As shown, the cluster head was successfully selected, and a hello packet was broadcast. Step 6: As Figure 10 As shown, determine the gateway node; Step 7: As Figure 10 As shown, a backbone network is established; Step 8: The network is complete, and the cluster head broadcasts information about all cluster members within its cluster.

[0051] 3. Specific Implementation Steps K-means algorithm: This invention selects the K-means clustering algorithm for possible cluster head processing.

[0052] This algorithm, taking into account the requirements of UAV swarm data link network characteristics, improves the standard K-means algorithm into a new New K-means algorithm, and modifies the standard distance iteration to distance and signal criticality iteration.

[0053] The K-means algorithm is an algorithm that divides unlabeled data into several categories. It has great advantages in network signal state identification. When the data is too large and there are too many classification variables, traditional classification methods are inefficient and cannot meet the requirements. Clustering algorithms can perfectly solve such problems. Clustering analysis refers to dividing a dataset into different clusters according to specific criteria (distance criteria), so that data objects in the same cluster are as similar as possible, and data objects in different clusters are as different as possible. After clustering, data from the same class will be as concentrated as possible, while data from different classes will be as separated as possible. Finally, by using an improved version of the K-means algorithm and adding importance conditions, the data can be quickly ranked to complete the selection of suboptimal cluster heads.

[0054] Original K-means algorithm: Dataset k cluster centers are Among them, the cluster center set The following conditions must be met: , ( (Represents non-empty) ; ; The K-means algorithm uses Euclidean distance for calculation. Its basic idea is to calculate the spatial distance between two data points using their coordinates in Euclidean space. The definition is: ; From the perspective of a certain terminal, all other terminals are reduced to different fixed Euclidean distances. The corresponding fixed parameters at a certain moment are obtained:

[0055] After traversing a certain terminal, we obtain: ; The K-means mathematical formula for the UAV data link terminal cluster of the present invention can be obtained.

[0056] This invention increases the importance of the signal and converts it to Euclidean distance: ; After using the added weighted signal, all K-cluster classes Since the performance of the cluster head can be calculated, the K-means algorithm can be used to divide the complex signal into K simple cluster heads. Then, by comparing the K clusters, the most likely suboptimal cluster head can be selected.

[0057] Define the parameters of the K-means algorithm: definition : Represents the importance weighting coefficient of terminal i to terminal j, with a minimum of 0.001 and a maximum of 1. This parameter indicates the importance of terminal i to other terminal units. This is a parameter indicating the importance of communication between terminal j and other terminals.

[0058] like Figure 11 As shown, the red dots are the core objects because of their... - The neighborhood consists of at least the samples initially set upon startup. Black samples are non-core objects. All samples that can be directly accessed by the density of core objects are within the supersphere centered on the red core object. If a sample is not within the refresh range, it cannot be directly accessed by density.

[0059] All terminal devices are connected to form a network. To put it simply, black dots represent ordinary terminal devices, red dots represent network cluster heads, and connecting lines indicate that the devices are connected. It should be noted that the drone swarm data link network supports frame skipping.

[0060] The K-means clustering definition: The set of samples connected by the highest density, derived from density reachability relationships, is the final category, or cluster, of our clusters. The method used by K-means is simple: it arbitrarily selects a kernel object without a category as a seed, then finds the set of all samples that are density-reachable from this kernel object; this is a cluster. It then continues by selecting another kernel object without a category and finding the density-reachable sample set, thus obtaining another cluster, and so on, until all kernel objects have a category.

[0061] The processing steps of this invention are as follows: Define the parameters for the K-means algorithm; define the characteristics of the endpoints; use known sample training data to obtain parameter values ​​suitable for the K-means algorithm; define and queue all endpoint signals using the endpoint characteristics, i.e., compare the network performance based on the endpoint as a cluster head, with the best endpoint number at the front; use the K-means algorithm to cluster the samples in the dataset; update the cluster centers of each endpoint; measure sample distances and label noise points or classify samples; repeat this process until the cluster centers no longer change or the pre-defined maximum number of iterations is reached; if a cluster head unexpectedly disconnects, the local area network selects the second-best cluster head, and the parameter table is updated in real time; dynamically adjust the ranking refresh time using the endpoint characteristics; refresh the latest parameters and aggregate all data as future training data.

[0062] Specifically as follows: first step: Terminal characteristic definition: Assuming terminal A is used as an example, after powering on, there is no communication between all terminals, and all distance weights are set to 100 kilometers. What is the distance value between terminal A and terminal B? The distance between terminal A and terminal C Copy the distance value of terminal K sequentially. .

[0063] In its first calculation, terminal A calculates its relative distances to all other terminals, obtaining the initial mean vector of the terminals, expressed as follows: ; In the formula, Let be the initial mean vector. Let n be the relative distance vector between the terminal and the terminal. Utilizing the characteristics of the terminal devices, define all terminal device signals and queue them. Queuing means that after the parameters between the terminals are aggregated, the terminal device with the strongest aggregation point is designated as the second-best terminal device (number 1), the next strongest as number 2, and so on, as shown in the following expression: ; In the formula, to quickly aggregate samples, a suboptimal cluster head unit is selected, and a... parameter, Let i be the terminal signal value. Indicates terminal i, This represents the distance of terminal j. This indicates the importance of terminal i to terminal j. Let be the square of the distance between end device i and end device j. Let `dist` be the square of the importance of terminal `i` to terminal `j`. The formula means that when terminal `i` is active, `dist`, a vector, must be calculated for all other terminal devices. Only after calculating `dist` can queuing comparisons be performed. Within a domain, if there are N endpoints, there will be N×(N-1) dist values. The value is achieved in the communication test. The value is usually obtained through training.

[0064] Step Two: This invention modifies the k-means algorithm by weighting the distance values ​​between endpoints, proposing a new k-means algorithm that adds an importance parameter between each endpoint. The absolute distance calculation is transformed into a weighted distance calculation. For each sample point in the set, the weighted distance value of all points in its neighborhood is calculated, and points with a density greater than or equal to a threshold are added to a preset target set. This action is the aggregation action of the K-means algorithm, which aggregates each sample in the data D. Clustering is performed using the following expression: ; ; In the formula, represents all calculated values, which are related to In this domain, each endpoint needs to test all other endpoints. The overall data is a two-dimensional matrix. Extracting the values ​​directly related to endpoint i results in a vector Cbj. Then, the minimum value argmin, bj, is selected. All changes are attributed to... Domain; arg min: argument of the minimum, when the function When taking the minimum value The present invention aims to find the second-best aggregator number after aggregating all terminal devices.

[0065] This indicates that the calculation is for terminal number j, which affects the coefficient values ​​of all other terminal numbers. Represents a cluster vector set, Indicates terminal i, This represents the distance of terminal j. This indicates the importance of terminal j. This represents the importance weighting coefficient of terminal i to terminal j.

[0066] Step 3: Update the cluster center of each terminal machine using the following expression: ; In the formula, a certain terminal is defined as terminal j, and the above formula holds true when terminal j is observed. This represents the calculated value of the suboptimal cluster head of terminal j. This represents the absolute value calculated between all terminals and terminal j. This is the value calculated by all terminal devices and terminal j. This indicates that each terminal x generates a sample.

[0067] Step 4: The weighted distance metric algorithm is used to calculate the distance between a sample and the core object samples. In K-means, a certain distance metric is used to measure the distance between samples. At the same time, all samples are traversed, and a threshold is set. Data points with distances less than the threshold are marked as noise points, while others are assigned to the same cluster as their original set.

[0068] Step 5: Repeat steps two, three, and four until the cluster center no longer changes or the pre-given maximum number of iterations is reached, at which point the algorithm terminates.

[0069] To compute the suboptimal cluster head selection, the K-means algorithm needs to calculate the relative distances of all endpoints (this algorithm assumes that all endpoints belong to the domain of this discussion). However, if the distance between the two endpoints is too great, or they do not communicate directly, their density can be considered sparse, and the algorithm may even consider the endpoint not to belong to a cluster. In this case, appropriately adjusting the coefficients can ensure its participation in clustering; through different iterations, it can be ensured that all endpoints participate in the calculation. Generally, distance coefficients and importance coefficients are added to upper and lower constraints.

[0070] like Figure 12 As shown, due to the importance calculation, the terminal machines are color-coded to indicate their importance. (Red indicates cluster head terminal machines, cyan indicates higher importance, and black indicates general terminal machines).

[0071] According to calculations ; This is the action of the aggregation signal. After aggregation is completed, it will select the most suitable suboptimal cluster head, the second most suitable suboptimal cluster head, and so on.

[0072] Step 6: The ranking refresh time can be adjusted by parameters. Currently, the simulation can be set to several time intervals such as 0.1ms, 1ms, 5ms, and 10ms, with the default parameter updating once every 1ms.

[0073] like Figure 13 As shown, refresh the terminal ID number and update the routing table, including the suboptimal cluster head position.

[0074] Example 1: How the K-means algorithm works: like Figure 13 and Figure 14As shown, the terminal device operates every 0.1ms. If the cluster head unexpectedly goes offline, the local area network immediately selects the next best frame header. The green terminal device is responsible for selecting the next best cluster head, updating the parameter table in real time, and preparing for the next selection. In the parameter table, the neighbor table refers to all terminal devices in the network, numbered sequentially (A~Z in this example), and the terminal numbers that can be directly reached between terminals form the neighbor table; the routing table refers to the information forwarded by the terminal devices, and the routing table is composed of terminal devices that can forward information; the next best cluster head table refers to the terminal device recommended by the algorithm if the current cluster head goes offline unexpectedly, with those appearing earlier being more likely to become the new cluster head.

[0075] Advantages and benefits of this invention: By using the K-means algorithm and real-time updated terminal characteristic parameter values, suboptimal cluster head terminals are selected. Each terminal prepares suboptimal cluster head information. If a cluster head terminal loses network access at a certain moment, the prepared suboptimal cluster head terminal is immediately selected as the actual cluster head terminal, saving the cluster head voting time after network outage and enhancing network robustness.

[0076] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for suboptimal cluster head selection in UAV swarm data links based on K-means, characterized in that: Includes the following steps: S1. After powering on, the drone cluster data link performs a self-test and forms a network on its own; S2. Based on the K-means clustering algorithm, the terminal devices of the UAV swarm are clustered to form terminal device clusters; S3. The terminal loads the terminal characteristic parameters and, based on the terminal characteristic parameters, calculates the DIST value of each terminal in the terminal cluster relative to itself and other terminals, and generates the DIST matrix. S4. Based on the DIST matrix of the terminal clusters, rank the suboptimal cluster heads of the terminal clusters, generate a suboptimal cluster head table, and select the suboptimal cluster heads. S5. At set intervals, repeat steps S2 to S4 to refresh the ranking of the second-best cluster heads and update the second-best cluster heads. S6. Store the terminal characteristic parameters of the terminal cluster during normal operation and send them to the central control center as learning and training samples for the UAV cluster. In step S3, the terminal characteristic parameters include distance parameters, importance parameters, and importance weighting coefficients; In step S3, the calculation expression for DIST is as follows: ； In the formula, Let j be the DIST value of terminal j relative to terminal i. Here is the distance parameter value for terminal i. Let j be the distance parameter value of the terminal. For the importance parameter value of terminal i, For the importance parameter value of terminal j, The weighting factor for the importance of terminal j relative to terminal i; The expression for calculating the DIST matrix is ​​as follows: ； In the formula, Let DIST be a matrix containing k endpoints. The elements in the matrix represent the DIST values ​​between two endpoints.

2. The method for suboptimal selection of cluster heads in a K-means UAV swarm data link according to claim 1, characterized in that: In step S2, based on the K-means clustering algorithm, the terminal devices of the UAV swarm are clustered to form terminal device clusters, including: S21. Select K initial cluster center endpoints; S22. Assign each terminal to the nearest cluster center terminal; S23. Calculate and update the cluster center terminal; S24. Repeat steps S22 to S23 until the cluster center terminal is stable.

3. The method for suboptimal selection of cluster heads in a K-means UAV swarm data link according to claim 1, characterized in that: In step S4, based on the DIST matrix of the terminal clusters, the suboptimal cluster heads of the terminal clusters are ranked, a suboptimal cluster head table is generated, and the suboptimal cluster heads are selected, including: S41. Obtain the DIST value of each terminal as the center; S42. Rank each terminal based on its DIST value, with each terminal as the center. The smaller the DIST value, the higher the ranking. Generate a suboptimal cluster head table. S43, the top-ranked terminal, was selected as the second-best cluster head.

4. The method for suboptimal selection of cluster heads in a K-means UAV swarm data link according to claim 3, characterized in that: In step S41, the DIST value centered on each terminal is obtained, including: The minimum DIST value in the first row of the DIST matrix is ​​used as the DIST value centered on terminal 1. The minimum DIST value in the second row of the DIST matrix is ​​used as the DIST value centered on terminal 2. Similarly, the DIST value of each terminal is obtained as the center.

5. The method for suboptimal selection of cluster heads in a K-means UAV swarm data link according to claim 1, characterized in that: The terminal device characteristic parameters are obtained through training by the UAV swarm and change in real time as the terminal device's operating environment and configuration change.

Images