Unmanned aerial vehicle cluster elastic formation control method and system and computer equipment

Abstract

The invention discloses an unmanned aerial vehicle cluster elastic formation control method and system and computer equipment, and the method comprises the steps: embedding a local auction mechanism in each unmanned aerial vehicle, calculating a safe neighbor set, and updating bidding through employing a maximum consensus mechanism, achieves the decentralized task distribution between unmanned aerial vehicles, only needs the local communication between neighbors in each auction, and improves the auction efficiency. Global coordination is not needed, and the communication overhead is greatly reduced. On the basis of keeping a communication topology, an artificial potential field method based on a Leader-Follower formation theory is introduced, a small number of leaders guide the whole formation to move through a predetermined track or a task target, and other followers dynamically keep and adjust the formation by using attraction and repulsive force in the artificial potential field. The method is especially suitable for unmanned aerial vehicle cluster formation performance, high-altitude flight tasks and other scenes. The lightweight distributed algorithm developed by using the characteristics of the unmanned cluster provides technical support for a subsequent unmanned cluster formation method.

Description

Technical Field

[0001] The present invention belongs to the technical field of UAV cluster formation control, and in particular relates to a UAV cluster flexible formation control method, system and computer equipment. Background Art

[0002] A drone swarm is an autonomous, coordinated system of multiple drones. Through distributed control, intelligent collaboration, and network communications, drones can collaborate to perform complex tasks. Compared to individual drones, drone swarms offer greater adaptability, flexibility, and robustness, and are widely applicable in a variety of fields, including military, agriculture, emergency rescue, logistics, and smart cities.

[0003] As a large-scale system, maintaining communication connectivity is a crucial issue for unmanned swarms. Currently, methods for maintaining connectivity in unmanned swarm formations primarily focus on topology control and optimization. Topology control methods, such as the Delaunay triangulation algorithm, construct a triangulated network using a set of planar UAV points as vertices, satisfying the Delaunay condition. Delaunay triangulations typically have optimal maximum and minimum angles, reducing communication link instability. However, facing practical challenges such as the high-speed motion of unmanned swarm systems and the addition and removal of nodes, Delaunay triangulation requires frequent reconstruction, dramatically increasing computational complexity in three-dimensional space. This results in low algorithm efficiency and high storage overhead, making it unsuitable for real-time computation. Optimization algorithms, such as those based on the Laplacian matrix, maintain the communication topology of unmanned swarm systems by maximizing the second smallest eigenvalue of the Laplacian matrix. However, in large-scale UAV formations (100+ nodes), the computational overhead of calculating the overall communication network eigenvalues ​​is high, necessitating the use of approximate methods. The Laplacian matrix only reflects the current network topology and cannot directly predict future topology changes. Maximizing the second smallest eigenvalue is not equivalent to shortest path optimization. When drones move at high speed or undergo large-scale topology changes, network topology optimization requires frequent recalculation of the Laplace matrix, which affects real-time performance. Summary of the Invention

[0004] In response to the above technical problems, the present invention provides a method, system and computer equipment for controlling the flexible formation of a swarm of drones.

[0005] The technical solution adopted by the present invention to solve its technical problem is: A method for controlling a flexible formation of a drone swarm, comprising the following steps: S100: Calculate the safe neighbor set of each agent and initialize the request, maximum bid set and token vector of each drone in the auction algorithm; S200: Each agent updates the maximum bid set through the token collection mechanism; S300: Determine the winner based on the maximum bid set, delete the edge corresponding to the highest bid, and repeat S200 and S300. After multiple rounds of auctions, the final overall network connectivity graph that meets the pre-set k-connectivity is obtained; S400: Define the center of the drone cluster, determine the leader based on the Euclidean distance between the drone and the center of the drone cluster, and define the remaining drones as followers. For the target area, initialize the key parameters of the artificial potential field algorithm, including the initial position of the leader, the coordinate range of the danger zone, and the position of the final target point; S500: Calculate the specific location information of the UAV cluster flight area and the size range of its structure. When the UAV cluster reaches the location of the final target point, the formation flight ends. If it does not reach the location of the final target point, proceed to the next step. S600: Based on the key parameters in the artificial potential field algorithm in S400, the repulsive force and attractive force on the unmanned swarm system are calculated to obtain the total resultant force. The formation flight information of the leader at the next moment is then calculated based on the artificial potential field method, including position information, speed information, and yaw angle information. S700: Each follower follows the leader to perform formation flight, forms and maintains the desired formation, and returns to S500 until the formation flight ends.

[0006] Preferably, S100 includes: S110: Each drone i first checks its directly connected neighbors , and calculate the second smallest eigenvalue of the network's Laplace matrix in a local range. By analyzing the changes in the second smallest eigenvalue, it is determined whether deleting an edge (i, j) will destroy the connectivity of the entire network. Finally, all nodes j corresponding to edges that can be safely deleted form a set, namely the safe neighbor set ; S120: Initialize the bidding request of each agent, the request format is vector form , where i is the number of the current drone, Indicates the edge that the drone wants to delete. is the corresponding bid amount; the drone performs different operations according to the situation of its own safe neighbor set: if drone i still has safe neighbors, that is, , then select a target neighbor from the safe neighbors as the edge to be deleted and submit a deletion request; if UAV i has no safe neighbors, that is, , it will not actually participate in the bidding, but only serve as an information transmission node. At this time, the request it submits is , which means it does not mention deleting any edges and does not set a bid amount; S130: Initialize the maximum bidding set of each agent and token vector , Mainly used to record each drone's own bidding information and ensure that this information can be shared with its neighbors Exchange and share, if drone i still has safe neighbors, that is , then use your own bidding information as the initial value of the maximum bidding set, that is, , which means that the current maximum bidding set only contains the drone’s own bidding data; if drone i has no safe neighbors, that is, , then it will not participate in the bidding. In this case, the maximum bidding set is an empty set, that is, Token vector The format is , the length of the vector is equal to the total number of drones in the cluster. Only the value at index i is 1, indicating that the drone currently numbered i has participated in the bidding, while the other positions are all 0, indicating that other drones have not yet submitted bids.

[0007] Preferably, S200 includes: The initial state of the token vector of the drone numbered i is , the token will gradually spread as the information between neighbors is exchanged. In this process, each agent i will collect token information from its neighbor nodes and continuously update its own token vector: ; in, represents a logical OR operation, which means that agent i will continuously synchronize the token status of its neighbors to ensure that it obtains all visible bidding information; When the token vector of the node is 0, that is, , indicating that the drone is still bidding and the auction continues. When the token vectors of all nodes are all 1, that is, When all drones have completed bidding, the system enters the termination state and the auction ends. During the bidding process, each agent i will apply the maximum consensus algorithm to update its maximum bidding set. Specifically, agent i will compare its own bid value with that of its neighbors and select the one with the largest third element in the bid value as the new bid set. ,in, is the bidding set of agent m at time t; represents the neighbor set of node i at time t, The third element in the three-dimensional vector [initiator, edge to be deleted, bid] corresponds to the bid value. Agent i collects the bidding sets of its neighbors, merges all visible bidding information, finds the bidding information corresponding to the highest bid value, and uses it as the current maximum bidding set. By iteratively propagating bidding information, all agents gradually synchronize the bidding results.

[0008] Preferably, S300 includes: Each agent determines the winner based on the maximum bid set , the expression is Once the request with the highest bid is accepted, the edge will be removed from the network topology; the winning bid is determined in a decentralized manner, and each agent makes decisions based only on its own information and that of its neighbors. A winner is determined in each auction round. After multiple rounds of auctions, the final connectivity graph is determined when the request format of each drone is r=[i, None, 0], where i is the number of each drone, None means there is no edge to be deleted, and 0 means the bid is zero, i.e. no bid is made. The network structure is dynamically adjusted while maintaining global connectivity, and the relative positions of each drone are also determined accordingly.

[0009] Preferably, S400 includes: For a small UAV cluster of 4 or fewer, the one with the closest Euclidean distance to the target point is selected as the leader. More than 4 drones, define the drone cluster center as , For the position of each drone in the final overall network connectivity graph, the radius r of the drone cluster center is defined according to the actual scenario requirements. When the Euclidean distance between the drone position and the center point is less than r, all drones that meet this condition are classified as leaders, and the remaining drones are regarded as followers. After that, the roles of leader and follower will not change. The leader plays a core guiding role in the task execution process, determining the overall movement direction of the formation. Followers adjust their own trajectories based on the leader's movement status and the location information of their neighbors to maintain formation stability and ensure network connectivity. Initialize the key parameters of the artificial potential field algorithm, including the initial position of the leader, the coordinate range of the danger zone, and the position of the final target point. At the same time, set the core weight parameters in the potential field calculation, including the strength of the attractive potential field, the effective radius of the repulsive potential field, and the drone's sensitivity to obstacle avoidance. Reasonable configuration of these parameters will directly affect the drone's motion path, enabling it to move smoothly towards the target area.

[0010] Preferably, S600 includes: S610: Calculate the repulsive and attractive forces acting on the drones based on the repulsive and attractive potential functions to analyze the impact of the environment on their motion. The repulsive force originates from obstacles, dangerous areas, and interactions between drones, preventing collisions and maintaining a safe distance. It increases as distance decreases. The attractive force, determined by the target point or formation requirements, guides the drones toward the target area, ensuring cluster coherence. The combined total force of the two determines the drone's motion trend and influences trajectory adjustment. S620: Based on the artificial potential field method, the resultant force is calculated according to the gradient change of the potential function, so that the UAV moves along the direction of the resultant force, and then the position information, speed and yaw angle of the leader at the next moment are calculated.

[0011] Preferably, the repulsive force and attractive force acting on the UAV are calculated based on the repulsive force potential function and the attractive force potential function in S610, and the resultant force is calculated according to the gradient change of the potential function in S620, specifically: The attractive potential function is , the repulsive force potential function is , find the negative gradient of the potential function, and we get the attraction and repulsive force The sum of the two is the magnitude of the resultant force in the artificial potential field ,in, is the attraction gain coefficient, is the repulsive force gain coefficient, where is the distance between the drone and the obstacle, The radius of obstacles affecting the drone.

[0012] Preferably, S700 includes: Based on the leader's information and relying on the leader-follower formation theory, the followers maintain a fixed distance from the leader at a predetermined yaw angle. When the formation needs to change, only the following distance and yaw angle need to be adjusted. Each follower controls the trajectory according to the rules to achieve synchronous formation flight. The leader does not need to pay attention to the follower status. When the leader moves, the followers adjust accordingly under the influence of the potential field and auction algorithm. When the error between the actual distance and the expected value is within the allowable range or close to zero, the formation is considered to be achieved, and finally the process returns to S500 to determine whether to end the formation flight.

[0013] A flexible formation control system for a swarm of unmanned aerial vehicles (UAVs), comprising an auction algorithm initialization module, a maximum bid set update module, a network connectivity graph determination module, a role division module, a formation flight end judgment module, a formation flight information determination module, and a formation flight module; The auction algorithm initialization module is used to calculate the safe neighbor set of each agent and initialize the request, maximum bid set and token vector of each drone in the auction algorithm; The maximum bid set update module is used by each agent to update the maximum bid set through the token collection mechanism; The network connectivity graph determination module is used to determine the winner based on the maximum bid set, delete the edge corresponding to the highest bid, and repeatedly execute the auction algorithm initialization module and the maximum bid set update module. After multiple rounds of auctions, the final overall network connectivity graph that meets the pre-set k-connectivity is obtained; The role division module is used to define the center of the drone cluster. The leader is determined based on the Euclidean distance between the drone and the center of the drone cluster. The remaining drones are defined as followers. For the target area, the key parameters of the artificial potential field algorithm are initialized, including the initial position of the leader, the coordinate range of the danger zone, and the location of the final target point. The formation flight end judgment module is used to calculate the specific location information of the drone cluster flight area and the size range of its structure. When the drone cluster reaches the final target point, the formation flight ends. If it does not reach the final target point, it proceeds to the next step. The formation flight information determination module is used to calculate the repulsive and attractive forces acting on the unmanned swarm system based on the key parameters of the artificial potential field algorithm in the role division module, obtain the total force, and then calculate the formation flight information of the leader at the next moment, including position information, speed information, and yaw angle information, using the artificial potential field method; The formation flight module is used for each follower to follow the leader to perform formation flight, form and maintain the desired formation formation, and return to the formation flight end judgment module until the formation flight ends.

[0014] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of a method for controlling a flexible formation of a drone cluster when executing the computer program.

[0015] The above-mentioned flexible formation control method, system and computer equipment of drone swarms can achieve target allocation, formation reconstruction and link optimization without a central controller by embedding a local auction mechanism in each drone. Each round of auction only requires local communication between neighbors, which greatly reduces communication overhead. On this basis, an artificial potential field method based on the Leader-Follower formation theory is introduced. A small number of leaders guide the overall movement of the formation, and the remaining followers use the attractive and repulsive forces in the artificial potential field to dynamically maintain and adjust the formation. It is particularly suitable for scenarios such as drone swarm formation performances and high-altitude flight missions. The lightweight distributed algorithm developed by the present invention utilizes the characteristics of the unmanned swarm itself to provide technical support for subsequent unmanned swarm formation methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1This is a flow chart of a method for controlling flexible formation of a drone swarm in one embodiment of the present invention; Figure 2 This is a flow chart of a method for controlling flexible formation of a drone swarm in another embodiment of the present invention; Figure 3 This is an example of processing a fixed unmanned swarm network in one embodiment of the present invention, where (a) is the original formation network diagram of 50 drones, and (b) is the final network connectivity diagram after the auction; Figure 4 This is a schematic diagram of a distributed auction algorithm in one embodiment of the present invention; Figure 5 This is a schematic diagram of the artificial potential field algorithm based on the Leader-Follower formation theory in one embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described in detail below with reference to the accompanying drawings.

[0018] In one embodiment, Figure 1 and Figure 2 As shown, a method for controlling a flexible formation of a drone swarm is provided, the method comprising the following steps: S100: Calculate the safe neighbor set of each agent and initialize the request, maximum bid set and token vector of each drone in the auction algorithm; S200: Each agent updates the maximum bid set through the token collection mechanism; S300: Determine the winner based on the maximum bid set, delete the edge corresponding to the highest bid, and repeat S200 and S300. After multiple rounds of auctions, the final overall network connectivity graph that meets the pre-set k-connectivity is obtained; S400: Define the center of the drone cluster, determine the leader based on the Euclidean distance between the drone and the center of the drone cluster, and define the remaining drones as followers. For the target area, initialize the key parameters of the artificial potential field algorithm, including the initial position of the leader, the coordinate range of the danger zone, and the position of the final target point; S500: Calculate the specific location information of the UAV cluster flight area and the size range of its structure. When the UAV cluster reaches the location of the final target point, the formation flight ends. If it does not reach the location of the final target point, proceed to the next step. S600: Based on the key parameters in the artificial potential field algorithm in S400, the repulsive force and attractive force on the unmanned swarm system are calculated to obtain the total resultant force. The formation flight information of the leader at the next moment is then calculated based on the artificial potential field method, including position information, speed information, and yaw angle information. S700: Each follower follows the leader to perform formation flight, forms and maintains the desired formation, and returns to S500 until the formation flight ends.

[0019] This flexible UAV swarm formation control method utilizes distributed control and the maximum bidding principle, while ensuring network connectivity. It also incorporates an artificial potential field method based on leader-follower formation theory. This combination makes the formation process more flexible and more efficient, making it ideal for multi-agent collaborative formation tasks.

[0020] The goal of S100 is to identify edges that can be deleted without affecting the overall connectivity of the drone cluster. Each drone in the cluster submits a request to delete an edge, along with a bid. Based on graph theory, the entire system can be represented as an undirected graph, with each drone in the system numbered.

[0021] In one embodiment, S100 includes: S110: Each drone i first checks its directly connected neighbors , and calculate the second smallest eigenvalue of the network's Laplace matrix in a local range. By analyzing the changes in the second smallest eigenvalue, it is determined whether deleting an edge (i, j) will destroy the connectivity of the entire network. Finally, all nodes j corresponding to edges that can be safely deleted form a set, namely the safe neighbor set ; S120: Initialize the bidding request of each agent, the request format is vector form , where i is the number of the current drone, Indicates the edge that the drone wants to delete. is the corresponding bid amount; the drone performs different operations according to the situation of its own safe neighbor set: if drone i still has safe neighbors, that is, , then select a target neighbor from the safe neighbors as the edge to be deleted and submit a deletion request; at the same time, the bid amount Set to the maximum connection distance in the current system; if drone i has no safe neighbors, , it will not actually participate in the bidding, but only serve as an information transmission node. At this time, the request it submits is , which means it does not mention deleting any edges and does not set a bid amount; S130: Initialize the maximum bidding set of each agent and token vector , Mainly used to record each drone's own bidding information and ensure that this information can be shared with its neighbors Exchange and share, if drone i still has safe neighbors, that is , then use your own bidding information as the initial value of the maximum bidding set, that is, , which means that the current maximum bidding set only contains the drone’s own bidding data; if drone i has no safe neighbors, that is, , then it will not participate in the bidding. In this case, the maximum bidding set is an empty set, that is, Token vector Used to mark whether the drone has submitted a bid and used for subsequent information synchronization. The format is The length of the vector is equal to the total number of drones in the swarm. Only the value at index i is 1, indicating that the drone with index i has already participated in the bidding, while all other positions are 0, indicating that other drones have not yet submitted bids. When the token vector is initialized, only the position of agent i itself is 1, while all other positions are 0.

[0022] like Figure 3 As shown, S200 formally enters the auction cycle. In one embodiment, S200 includes: The initial state of the token vector of the drone numbered i is , the token will gradually spread as the information between neighbors is exchanged. In this process, each agent i will collect token information from its neighbor nodes and continuously update its own token vector: ; in, represents a logical OR operation, which means that agent i will continuously synchronize the token status of its neighbors to ensure that it obtains all visible bidding information; When the token vector of the node is 0, that is, , indicating that the drone is still bidding and the auction continues. When the token vectors of all nodes are all 1, that is, When all drones have completed bidding, the system enters the termination state and the auction ends. During the bidding process, each agent i will apply the maximum consensus algorithm to update its maximum bidding set. Specifically, agent i will compare its own bid value with that of its neighbors and select the one with the largest third element in the bid value as the new bid set. ,in, is the bidding set of agent m at time t; represents the neighbor set of node i at time t, The third element in the three-dimensional vector [initiator, edge to be deleted, bid] corresponds to the bid value. Agent i collects the bidding sets of its neighbors, merges all visible bidding information, finds the bidding information corresponding to the highest bid value, and uses it as the current maximum bidding set. By iteratively propagating bidding information, all agents gradually synchronize the bidding results.

[0023] like Figure 4 As shown, the goal of S300 is to delete the edge that actually wins and update the network topology. In one embodiment, S300 includes: Each agent determines the winner based on the maximum bid set , the expression is Once the request with the highest bid is accepted, the edge will be removed from the network topology; the winning bid is determined in a decentralized manner, and each agent makes decisions based only on its own information and that of its neighbors. A winner is determined in each auction round. After multiple rounds of auctions, the final connectivity graph is determined when the request format of each drone is r=[i, None, 0], where i is the number of each drone, None means there is no edge to be deleted, and 0 means the bid is zero, i.e. no bid is made. The network structure is dynamically adjusted while maintaining global connectivity, and the relative positions of each drone are also determined accordingly.

[0024] In the control strategy of S400, in order to ensure the coordinated movement and task execution efficiency of the entire system, the drones are first divided into roles. In one embodiment, S400 includes: For a small UAV cluster of 4 or fewer, the one closest to the target point in Euclidean distance is selected as the leader; More than 4 drones, define the drone cluster center as , For the position of each drone in the final overall network connectivity graph, the radius r of the drone cluster center is defined according to the actual scenario requirements. When the Euclidean distance between the drone position and the center point is less than r, all drones that meet this condition are classified as leaders, and the remaining drones are regarded as followers. After that, the roles of leader and follower will not change. For a given target area, it is necessary to initialize the key parameters of the artificial potential field algorithm, including the initial position of the leader, the coordinate range of the danger zone, and the position of the final target point. At the same time, the core weight parameters in the potential field calculation are set, including the strength of the attractive potential field, the effective radius of the repulsive potential field, and the drone's sensitivity to obstacle avoidance. Reasonable configuration of these parameters will directly affect the drone's motion path, enabling it to smoothly advance to the target area.

[0025] S500 is the flag for determining the end of formation flight. When the entire drone cluster system reaches the target location, the formation flight mission ends. If it does not reach the target location, the next action is taken.

[0026] like Figure 5 As shown, in one embodiment, S600 includes: S610: Calculate the repulsive and attractive forces acting on the drone based on the repulsive and attractive potential functions to analyze the impact of the environment on its motion. The repulsive force originates from the interaction between obstacles, dangerous areas, and drones, preventing collisions and maintaining a safe distance. It increases as distance decreases (generally, the magnitude of the repulsive force is inversely proportional to the distance between the drone and obstacles or other drones. That is, the closer the distance, the greater the repulsive force, forcing drones to stay away from dangerous areas or maintain a reasonable distance between each other). The attractive force is determined by the target point or formation requirements, guiding the drones to move towards the target area and ensuring cluster coherence. The combined total force of the two determines the drone's motion trend and influences trajectory adjustment. S620: Based on the artificial potential field method, the resultant force is calculated according to the gradient change of the potential function, so that the UAV moves along the direction of the resultant force, and then the position information, speed and yaw angle of the leader at the next moment are calculated.

[0027] In one embodiment, the repulsive force and attractive force acting on the drone are calculated based on the repulsive force potential function and the attractive force potential function in S610, and the resultant force is calculated according to the gradient change of the potential function in S620, specifically: The attractive potential function is , the repulsive force potential function is , find the negative gradient of the potential function, and we get the attraction and repulsive force The sum of the two is the magnitude of the resultant force in the artificial potential field ,in, is the attraction gain coefficient, is the repulsive force gain coefficient, where is the distance between the drone and the obstacle, The reasonable configuration of these parameters will directly affect the trajectory of the drone, allowing it to smoothly move toward the target area while effectively avoiding obstacles and dangerous areas during flight, avoiding the problem of falling into local minima, and ultimately achieving safe and efficient coordinated flight of drone swarms.

[0028] In one embodiment, S700 includes: Based on the leader's information and relying on the leader-follower formation theory, the followers maintain a fixed distance from the leader at a predetermined yaw angle. When the formation needs to change, only the following distance and yaw angle need to be adjusted. Each follower controls the trajectory according to the rules to achieve synchronous formation flight. The leader does not need to pay attention to the follower status. When the leader moves, the followers adjust accordingly under the influence of the potential field and auction algorithm. When the error between the actual distance and the expected value is within the allowable range or close to zero, the formation is considered to be achieved, and finally the process returns to S500 to determine whether to end the formation flight.

[0029] The above-mentioned flexible formation control method for drone swarms integrates a distributed auction mechanism with an artificial potential field method based on leader-follower formation theory within a flexible formation control framework. The auction algorithm primarily addresses communication topology issues. By calculating a set of safe neighbors and updating bids using a maximum consensus mechanism, it achieves decentralized task allocation among drones. In large-scale drone systems, the auction mechanism allows each drone to autonomously select and maintain key communication links that meet k-connectivity requirements by simply exchanging information with its local neighbors, eliminating the need for global coordination and effectively reducing the communication burden. While maintaining the communication topology, leader-follower formation control is introduced. A small number of leaders guide the overall formation movement through predetermined trajectories or mission objectives, while the majority of followers dynamically adjust based on the attractive forces (towards the leader or target) and repulsive forces (to avoid collisions or obstacles) generated by the artificial potential field method. This ensures k-connectivity while maintaining a reasonable formation structure. Given a network of drone nodes, which can be viewed as an undirected graph, k-connectivity means that at least k edges must be removed for the entire network to become disconnected.

[0030] The combination of a distributed auction algorithm and an artificial potential field method based on the Leader-Follower theory can achieve efficient collaboration and flexible reconstruction of drone clusters in dynamic environments while maintaining network connectivity and formation structure stability, providing a reliable solution for the autonomous control of large-scale intelligent unmanned systems.

[0031] In one embodiment, a flexible formation control system for a swarm of unmanned aerial vehicles includes an auction algorithm initialization module, a maximum bid set update module, a network connectivity graph determination module, a role division module, a formation flight end judgment module, a formation flight information determination module, and a formation flight module; The auction algorithm initialization module is used to calculate the safe neighbor set of each agent and initialize the request, maximum bid set and token vector of each drone in the auction algorithm; The maximum bid set update module is used by each agent to update the maximum bid set through the token collection mechanism; The network connectivity graph determination module is used to determine the winner based on the maximum bid set, delete the edge corresponding to the highest bid, and repeatedly execute the auction algorithm initialization module and the maximum bid set update module. After multiple rounds of auctions, the final overall network connectivity graph that meets the pre-set k-connectivity is obtained; The role division module is used to define the center of the drone cluster. The leader is determined based on the Euclidean distance between the drone and the center of the drone cluster. The remaining drones are defined as followers. For the target area, the key parameters of the artificial potential field algorithm are initialized, including the initial position of the leader, the coordinate range of the danger zone, and the location of the final target point. The formation flight end judgment module is used to calculate the specific location information of the drone cluster flight area and the size range of its structure. When the drone cluster reaches the final target point, the formation flight ends. If it does not reach the final target point, it proceeds to the next step. The formation flight information determination module is used to calculate the repulsive and attractive forces acting on the unmanned swarm system based on the information in the role division module, obtain the total force, and then calculate the formation flight information of the leader at the next moment, including position information, speed information, and yaw angle information, using the artificial potential field method; The formation flight module is used for each follower to follow the leader to perform formation flight, form and maintain the desired formation formation, and return to the formation flight end judgment module until the formation flight ends.

[0032] For the specific definition of a drone swarm elastic formation control system, please refer to the definition of a drone swarm elastic formation control method above, which will not be repeated here. The various modules in the above-mentioned drone swarm elastic formation control system can be implemented in whole or in part by software, hardware, and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above modules.

[0033] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of a method for controlling a flexible formation of a drone cluster when executing the computer program.

[0034] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described method embodiments. Any reference to memory, storage, database, or other media used in the embodiments provided herein may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0035] The above is a detailed introduction to the flexible formation control method, system and computer equipment for drone swarms provided by the present invention. This article uses specific examples to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the core idea of ​​the present invention. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present invention, the present invention can also be improved and modified in several ways, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

Claims

1. A method for controlling flexible formation of drone swarms, characterized in that: The method comprises the following steps: S100: Calculate the safe neighbor set of each agent and initialize the request, maximum bid set and token vector of each drone in the auction algorithm; S200: Each agent updates the maximum bid set through the token collection mechanism; S300: Determine the winner based on the maximum bid set, delete the edge corresponding to the highest bid, and repeat S200 and S300. After multiple rounds of auctions, the final overall network connectivity graph that meets the pre-set k-connectivity is obtained; S400: Define the center of the drone cluster, determine the leader based on the Euclidean distance between the drone and the center of the drone cluster, and define the remaining drones as followers. For the target area, initialize the key parameters of the artificial potential field algorithm, including the initial position of the leader, the coordinate range of the danger zone, and the position of the final target point; S500: Calculate the specific location information of the UAV cluster flight area and the size range of its structure. When the UAV cluster reaches the location of the final target point, the formation flight ends. If it does not reach the location of the final target point, proceed to the next step. S600: Based on the key parameters in the artificial potential field algorithm in S400, the repulsive force and attractive force on the unmanned swarm system are calculated to obtain the total resultant force. The formation flight information of the leader at the next moment is then calculated based on the artificial potential field method, including position information, speed information, and yaw angle information. S700: Each follower follows the leader to perform formation flight, forms and maintains the desired formation, and returns to S500 until the formation flight ends.

2. The method according to claim 1, characterized in that S100 includes: S110: Each drone i first checks its directly connected neighbors , and calculate the second smallest eigenvalue of the network's Laplace matrix in a local range. By analyzing the changes in the second smallest eigenvalue, it is determined whether deleting an edge (i, j) will destroy the connectivity of the entire network. Finally, all nodes j corresponding to edges that can be safely deleted form a set, namely the safe neighbor set ; S120: Initialize the bidding request of each agent, the request format is vector form , where i is the number of the current drone, Indicates the edge that the drone wants to delete. is the corresponding bid amount; the drone performs different operations according to the situation of its own safe neighbor set: if drone i still has safe neighbors, that is, , then select a target neighbor from the safe neighbors as the edge to be deleted and submit a deletion request; if UAV i has no safe neighbors, that is, , it will not actually participate in the bidding, but only serve as an information transmission node. At this time, the request it submits is , which means it does not mention deleting any edges and does not set a bid amount; S130: Initialize the maximum bidding set of each agent and token vector , Mainly used to record each drone's own bidding information and ensure that this information can be shared with its neighbors Exchange and share, if drone i still has safe neighbors, that is , then use your own bidding information as the initial value of the maximum bidding set, that is, , which means that the current maximum bidding set only contains the drone’s own bidding data; if drone i has no safe neighbors, that is, , then it will not participate in the bidding. In this case, the maximum bidding set is an empty set, that is, Token vector The format is , the length of the vector is equal to the total number of drones in the cluster. Only the value at index i is 1, indicating that the drone currently numbered i has participated in the bidding, while the other positions are all 0, indicating that other drones have not yet submitted bids.

3. The method according to claim 2, characterized in that S200 includes: The initial state of the token vector of the drone numbered i is , the token will gradually spread as the information between neighbors is exchanged. In this process, each agent i will collect token information from its neighbor nodes and continuously update its own token vector: ； in, represents a logical OR operation, which means that agent i will continuously synchronize the token status of its neighbors to ensure that it obtains all visible bidding information; When the token vector of the node is 0, that is, , indicating that the drone is still bidding and the auction continues. When the token vectors of all nodes are all 1, that is, When all drones have completed bidding, the system enters the termination state and the auction ends. During the bidding process, each agent i will apply the maximum consensus algorithm to update its maximum bidding set. Specifically, agent i will compare its own bid value with that of its neighbors and select the one with the largest third element in the bid value as the new bid set. ,in, is the bidding set of agent m at time t; represents the neighbor set of node i at time t, The third element in the three-dimensional vector [initiator, edge to be deleted, bid] corresponds to the bid value. Agent i collects the bidding sets of its neighbors, merges all visible bidding information, finds the bidding information corresponding to the highest bid value, and uses it as the current maximum bidding set. By iteratively propagating bidding information, all agents gradually synchronize the bidding results.

4. The method according to claim 3, characterized in that S300 includes: Each agent determines the winner based on the maximum bid set , the expression is Once the request with the highest bid is accepted, the edge will be removed from the network topology; the winning bid is determined in a decentralized manner, and each agent makes decisions based only on its own information and that of its neighbors. A winner is determined in each auction round. After multiple rounds of auctions, the final connectivity graph is determined when the request format of each drone is r=[i, None, 0], where i is the number of each drone, None means there is no edge to be deleted, and 0 means the bid is zero, i.e. no bid is made. The network structure is dynamically adjusted while maintaining global connectivity, and the relative positions of each drone are also determined accordingly.

5. The method according to claim 4, characterized in that S400 includes: For a small UAV cluster of 4 or fewer, the one closest to the target point in Euclidean distance is selected as the leader; More than 4 drones, define the drone cluster center as , For the position of each drone in the final overall network connectivity graph, the radius r of the drone cluster center is defined according to the actual scenario requirements. When the Euclidean distance between the drone position and the center point is less than r, all drones that meet this condition are classified as leaders, and the remaining drones are regarded as followers. After that, the roles of leader and follower will not change. The leader plays a core guiding role in the task execution process, determining the overall movement direction of the formation. Followers adjust their own trajectories based on the leader's movement status and the location information of their neighbors to maintain formation stability and ensure network connectivity. Initialize the key parameters of the artificial potential field algorithm, including the initial position of the leader, the coordinate range of the danger zone, and the position of the final target point. At the same time, set the core weight parameters in the potential field calculation, including the strength of the attractive potential field, the effective radius of the repulsive potential field, and the drone's sensitivity to obstacle avoidance. Reasonable configuration of these parameters will directly affect the drone's motion path, enabling it to move smoothly towards the target area.

6. The method according to claim 5, characterized in that S600 includes: S610: Calculate the repulsive and attractive forces acting on the drones based on the repulsive and attractive potential functions to analyze the impact of the environment on their motion. The repulsive force originates from obstacles, dangerous areas, and interactions between drones, preventing collisions and maintaining a safe distance. It increases as distance decreases. The attractive force, determined by the target point or formation requirements, guides the drones toward the target area, ensuring cluster coherence. The combined total force of the two determines the drone's motion trend and influences trajectory adjustment. S620: Based on the artificial potential field method, the resultant force is calculated according to the gradient change of the potential function, so that the UAV moves along the direction of the resultant force, and then the position information, speed and yaw angle of the leader at the next moment are calculated.

7. The method according to claim 6, characterized in that In S610, the repulsive force and attractive force acting on the UAV are calculated based on the repulsive force potential function and the attractive force potential function, and in S620, the resultant force is calculated according to the gradient change of the potential function. Specifically, The attractive potential function is , the repulsive force potential function is , find the negative gradient of the potential function, and we get the attraction and repulsive force The sum of the two is the magnitude of the resultant force in the artificial potential field ,in, is the attraction gain coefficient, is the repulsive force gain coefficient, where is the distance between the drone and the obstacle, The radius of obstacles affecting the drone.

8. The method according to claim 7, characterized in that The S700 includes: Based on the leader's information and relying on the leader-follower formation theory, the followers maintain a fixed distance from the leader at a predetermined yaw angle. When the formation needs to change, only the following distance and yaw angle need to be adjusted. Each follower controls the trajectory according to the rules to achieve synchronous formation flight. The leader does not need to pay attention to the follower status. When the leader moves, the followers adjust accordingly under the influence of the potential field and auction algorithm. When the error between the actual distance and the expected value is within the allowable range or close to zero, the formation is considered to be achieved, and finally the process returns to S500 to determine whether to end the formation flight.

9. A flexible formation control system for drone swarms, characterized by: It includes an auction algorithm initialization module, a maximum bidding set update module, a network connectivity graph determination module, a role division module, a formation flight end judgment module, a formation flight information determination module, and a formation flight module; The auction algorithm initialization module is used to calculate the safe neighbor set of each agent and initialize the request, maximum bid set and token vector of each drone in the auction algorithm; The maximum bid set update module is used by each agent to update the maximum bid set through the token collection mechanism; The network connectivity graph determination module is used to determine the winner based on the maximum bid set, delete the edge corresponding to the highest bid, and repeatedly execute the auction algorithm initialization module and the maximum bid set update module. After multiple rounds of auctions, the final overall network connectivity graph that meets the pre-set k-connectivity is obtained; The role division module is used to define the center of the drone cluster. The leader is determined based on the Euclidean distance between the drone and the center of the drone cluster. The remaining drones are defined as followers. For the target area, the key parameters of the artificial potential field algorithm are initialized, including the initial position of the leader, the coordinate range of the danger zone, and the location of the final target point. The formation flight end judgment module is used to calculate the specific location information of the drone cluster flight area and the size range of its structure. When the drone cluster reaches the final target point, the formation flight ends. If it does not reach the final target point, it proceeds to the next step. The formation flight information determination module is used to calculate the repulsive and attractive forces acting on the unmanned swarm system based on the information in the role division module, obtain the total force, and then calculate the formation flight information of the leader at the next moment, including position information, speed information, and yaw angle information, using the artificial potential field method; The formation flight module is used for each follower to follow the leader to perform formation flight, form and maintain the desired formation formation, and return to the formation flight end judgment module until the formation flight ends.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

Images