A cooperative control method for drone swarms for target search in a dynamic obstacle environment

Abstract

The present invention relates to a collaborative control method for a drone cluster for target search in a dynamic obstacle environment. Compared with the existing technology, the method solves the defects of low target search efficiency and poor obstacle avoidance performance of drone clusters in complex and dynamic environments. The present invention includes the following steps: initialization of the drone cluster; dynamic environment perception and obstacle information acquisition; collaborative optimization of target search by multiple subgroups of drones; drone cluster path planning and obstacle avoidance processing; and updating of drone scheduling positions. The present invention combines real-time environmental perception, data fusion technology, and an obstacle avoidance mechanism based on force fields to ensure the safety and stability of drone clusters in complex environments, and avoid possible collisions and path planning errors of drones in dynamic obstacle environments. The path optimization method based on the cost matrix effectively reduces the total time and flight distance for task completion, further improving the efficiency of task execution.

Description

Technical Field

[0001] The present invention relates to the technical field of drone cluster control, and in particular to a drone cluster collaborative control method for target search in a dynamic obstacle environment. Background Art

[0002] With the rapid development of drone technology, the application of drone swarms in complex environments has become a research hotspot. In particular, how to efficiently coordinate drone swarms to complete target search missions in environments with dynamic obstacles has become a pressing issue. In practical applications, dynamic obstacles (such as wind-driven objects, moving vehicles, or people) can disrupt drone flight paths.

[0003] Existing drone target search technologies mostly focus on static environments or assume that obstacles remain constant. However, in real-world dynamic environments, obstacles and targets often change over time, posing significant challenges to drone target search and path planning. Traditional path planning methods often rely on static maps or preset obstacle locations. Therefore, they are unable to make effective real-time adjustments to the unknown and constantly changing dynamic obstacles, resulting in low search efficiency and poor mission completion quality.

[0004] To address this challenge, researchers have recently proposed collaborative control methods based on swarm intelligence, attempting to improve the efficiency and accuracy of target search through the coordinated cooperation of multiple drones. While these methods address the low efficiency of individual drones to some extent, existing technologies still have many shortcomings in dynamic obstacle environments. Specifically, existing drone swarm control strategies often ignore the impact of dynamic environmental changes and are unable to quickly and flexibly adjust paths and update decisions based on real-time obstacle information and target position changes.

[0005] To address the problem of drone swarm target search in dynamic obstacle environments, Phung et al. proposed a motion-encoding-based particle swarm algorithm to solve the problem of searching for moving targets. Wang et al. proposed a drone swarm algorithm based on the bat algorithm to detect dynamically intruding targets in oil fields. Duan et al. proposed a dynamic discrete pigeon swarm optimization method based on the pigeon swarm algorithm to solve the planning problem of drone swarms performing search and attack missions. Zheng et al. proposed a human-machine collaborative algorithm to address the problem of capturing fugitives.

[0006] While various collaborative control strategies have been proposed, balancing the requirements of target search and obstacle avoidance, while maintaining efficient collaboration among clusters in dynamic environments, remains a technical challenge. Existing methods often rely too heavily on static search patterns or focus too closely on local areas, easily falling into local optimal solutions and failing to effectively address the dynamic changes of targets and obstacles in complex environments.

[0007] Therefore, how to optimize the target search efficiency of drone swarms in complex dynamic obstacle environments through flexible collaborative control strategies and ensure coordination and cooperation among swarm members is a key issue that needs to be addressed in current drone swarm target search technology. Existing technologies lack efficient and practical solutions for drone swarm collaborative control methods in dynamic obstacle environments. Summary of the Invention

[0008] The purpose of the present invention is to solve the defects of the existing technology of low target search efficiency and poor obstacle avoidance performance of drone clusters in complex and dynamic environments, and to provide a drone cluster collaborative control method for target search in a dynamic obstacle environment to solve the above problems.

[0009] In order to achieve the above object, the technical solution of the present invention is as follows:

[0010] A method for cooperative control of a swarm of unmanned aerial vehicles (UAVs) for target search in a dynamic obstacle environment comprises the following steps:

[0011] Initialization of the UAV swarm: Initialize the search area according to the mission requirements, set the initial flight parameters of the UAV swarm, and establish the swarm communication network;

[0012] Dynamic environment perception and obstacle information acquisition: Utilize drone sensors to collect real-time information about the surrounding environment, generate real-time environmental maps through data fusion technology, and promptly feed back environmental data to each drone;

[0013] Collaborative optimization of target search among multiple drone sub-swarms: The drone cluster is divided into multiple sub-swarms, each responsible for target search tasks in different areas. The flight missions and path planning of the sub-swarms are dynamically adjusted based on target distribution and environmental information.

[0014] Drone swarm path planning and obstacle avoidance: Calculate the optimal movement path for each drone, use a real-time obstacle avoidance mechanism, and adjust the path based on sensor feedback during flight.

[0015] Update of UAV scheduling position: The scheduling system performs dynamic scheduling based on the current status of the cluster. After the UAV is dispatched to the new position, the UAV cluster collaborative target search step continues until the target search mission is successfully completed.

[0016] The initialization of the drone cluster includes the following steps:

[0017] According to the relative positions between the UAV clusters, an initial plane coordinate system is established in the target working area. When setting the initial coordinate system C in the target working area, it is assumed that the target area R is:

[0018] R=[x min ,x max ]×[y min ,y max ],

[0019] Among them, x max 、x min and represent the upper and lower boundaries of the x-axis, y max 、y min and represent the upper and lower boundaries of the y-axis respectively;

[0020] Set the initial flight parameters of each drone, including: the number of sub-drone clusters N s It represents the number of neutron drone clusters in the target search of multi-sub-swarm collaborative optimization of drones, and the historical archive size A h Indicates the size of the archive storing historical environmental information, and the initial speed v0 of the drone;

[0021] The setting formula for initial speed and flight mode is:

[0022] v0=MinSpeed+(MaxSpeed-MinSpeed)·β(β∈[0,1]),

[0023] Among them, MinSpeed ​​and MaxSpeed ​​are the minimum and maximum speeds of the drone respectively, and β is a random factor between [0,1], which is used to simulate the randomness of speed;

[0024] Assume that the drone cluster communication network is implemented through the wireless communication protocol Wi-Fi, and its topology is represented by graph theory:

[0025] Each drone is considered as a node in the graph. If the distance d between two drones ij Less than the communication threshold d th , then there is an edge between the two drones. The formula for constructing the topological structure is as follows:

[0026] G=(V,E),V={1,2,…,N},E={(i,j)∣∣d ij ≤d th}

[0027] Among them, G is the graph structure of the communication network, V is all the nodes in the drone cluster, E is the connected edge set, d ijis the distance between UAV i and UAV j, and d th is the set communication distance threshold;

[0028] According to the target search task of the UAV cluster in a dynamic obstacle environment, a fitness value function f(x) is specified to represent the signal strength of the search target:

[0029] Assuming that the target signal strength is calculated based on the target position x and environmental conditions ∈, the fitness function is expressed as:

[0030]

[0031] Where S(x) is the signal strength of the target at position x, d(x,x goal ) is from the current position x to the target position x goal distance, ∈(x) represents the degree of environmental obstacles or interference at that location, α is the attenuation coefficient, which determines the degree of environmental impact, and x represents the regional coordinate position.

[0032] The fitness value f(x) is used to guide the adjustment of the search strategy. The stronger the target signal strength, the greater the fitness, indicating that the search priority in this area is higher;

[0033] Initialize the location set of the drone cluster:

[0034] Set the initial position set P init When the location is selected according to the regional division and the distribution of the target, it is assumed that the UAV cluster is in the target area R = [x min ,x max ]×[y min ,y max ] distribution, the initialization position P of each UAV i It is set by the following formula:

[0035] P init ={P1,P2,…,P N},P i =(x i ,y i ),x i ∈[x min ,x max ],y i ∈[y min ,y max ]

[0036] Where N is the number of drones in the cluster, P i is the initial position of the i-th UAV, P initThe position set is the initial state of the UAV cluster. During the mission execution, the cluster dynamically adjusts its position according to target distribution, environmental changes, and mission requirements.

[0037] Initialize history file A:

[0038] Put the position of a random drone cluster in the initialized drone cluster into the historical file A:

[0039] A=A∪x i ,i∈[1,N]

[0040] Historical archive A is used to save the individual positions of drones that are eliminated due to failure in target strength comparison during the iterative position update process during the execution of the drone cluster target search mission.

[0041] The dynamic environment perception and obstacle information acquisition includes the following steps:

[0042] Collect sensor data:

[0043] The sensors equipped by the drones are used to collect real-time information about changes in the surrounding environment, including other drones, static and dynamic obstacles. The sensor perception range of each drone is set to r sensor , then the perception area of ​​each UAV i is expressed as:

[0044]

[0045] Among them, S i represents the area perceived by UAV i, (x j ,y j ) is the position of the jth object, and r sensor is the maximum sensing radius of the sensor;

[0046] Classification and recognition of sensor data:

[0047] According to the sensed data, different types of obstacles and target objects are identified and classified. The drone uses infrared sensors and visual sensors to identify static obstacles. static and dynamic obstacles O dynamic , the target object is represented by G, and the boundary between each obstacle and the target in the environment is approximated by a circular model. The detection formula for static and dynamic obstacles is:

[0048]

[0049] If d i,j ≤r static , d i,j ≤r dynamic , it is considered that UAV i detects static or dynamic obstacle j, where rstatic and r dynamic are the safety distances of static obstacles and dynamic obstacles, d i,j is the distance between the drone and the obstacle;

[0050] Data fusion technology is used to integrate the input data of multiple sensors. UAV i is equipped with K sensors for data collection. The output of sensor K is D k , then the environmental information D after sensor data fusion fusion , calculated by weighted average method,

[0051]

[0052] Among them, w k is the weight of sensor k;

[0053] Build a real-time environment map:

[0054] The fused data is used to build a real-time environment map, and each UAV i is located according to its current position P i =(x i ,y i ) and the perceived obstacle information D fusion Updated environment map M env , the environment map is represented as a matrix, where each element M i,j Indicates position (x i ,y i ) at the environmental state, the update rules are as follows:

[0055]

[0056] Among them, M env (x i ,y i )=0 means there is no object at this location, M env (x i ,y i )=1 indicates a static obstacle, M env (x i ,y i )=2 means there are dynamic obstacles, M env (x i ,y i )=3 means that there is a drone individual in the drone cluster at this location, M env (x i ,y i )=4 means that the position is the target object position.

[0057] The multi-subgroup collaborative optimization target search of UAVs includes the following steps:

[0058] The entire drone swarm is divided into three drone sub-groups:

[0059] Initially, the cluster is divided into three sub-clusters: exploration drone sub-cluster P1, development drone sub-cluster P2, and archive drone sub-cluster P3.

[0060] The size of the initial drone sub-cluster satisfies the following conditions: N1+N2+N3=N. The initial division of drones within the sub-cluster is completed by random selection:

[0061]

[0062] Among them, N1, N2, and N3 represent the number of drones in the subgroups P1, P2, and P3 respectively. N1, N2, and N3 are positive integers. In the later stage of target search, the size of a single sub-drone cluster is allowed to be 0. k Represents the subgroup P k The included drone collection;

[0063] Set the subgroup target search strategy:

[0064] Explore drone sub-clusters P1: Explore drone sub-clusters for global search, generating the next generation of positions near the current drone cluster’s random drone positions, Indicates the j-th dimension coordinate value of the i-th drone in the current t-th iteration time slice exploring the drone sub-cluster P1, and exploring the position of each drone in the drone sub-cluster P1 The update strategy is:

[0065]

[0066] Randomly select the j-th dimension coordinate value p of the position of an individual drone in the drone cluster in the current t-th iteration time slice rj As the reference individual position, the current position is updated based on its position and random disturbance, rand is a random number in [0,1], and r is a dynamically decreasing random factor;

[0067] Develop drone sub-cluster P2: Develop drone sub-cluster responsible for local search, generating the next generation of positions around the current drone cluster's best signal strength position, Indicates the j-th dimension coordinate value of the i-th drone in the current t-th iteration time slice of the drone sub-cluster P2, and the position of each drone The update formula is:

[0068]

[0069]

[0070] v rj=2·rand-1

[0071] Among them, d rj Indicates the j-dimension coordinate value of the current UAV individual position and the j-dimension coordinate value of the global optimal position g j distance, add random direction v rj To enhance diversity, update the position in g j Randomly generate nearby and gradually converge to the optimal solution area;

[0072] Archive drone sub-cluster P3: The archive drone sub-cluster uses the historical optimal solution to search and generates the next generation of positions near the positions of individual drones eliminated in the early search process. It represents the j-th dimension coordinate value of the i-th drone in the drone subcluster P3 of the current t-th iteration time slice archive. Its position update formula is:

[0073]

[0074] Among them, a rj The historical optimal solution is randomly selected from the historical archive A. The historical archive provides a stable and diverse reference by storing solutions with improved fitness over generations;

[0075] Information sharing and dynamic adjustment of subgroup size:

[0076] After obtaining the individual positions of the next generation of drones based on the different strategies of the three sub-swarms, the drones exchange information with each other. Each drone compares its current position information with the position information of the previous time slice to determine whether it is closer to the target. If successful, it will be incorporated into the winner drone cluster of this generation. If unsuccessful, the current position will be stored in the position history archive.

[0077] For the location set X of all drone clusters in the tth time slice t ={x1 t ,x2 t ,…,x N t}, updated winner drone cluster S t The relationship with location history file A is expressed as:

[0078] S t ={x i t ∈X t |f(x i t )≥f(x i t-1 )}

[0079] A=A∪{x i t |xi t ∈X t and f(x i t ) <f(x i t-1 )}

[0080] Among them, f(x i t ) represents the target signal strength value of the i-th UAV at the t-th time slice;

[0081] When placing individual locations into historical archive A, if the number of individual locations exceeds the size of the historical archive A h When , first randomly delete an individual position, and then perform the add operation:

[0082] x remove ∈A

[0083] A=(A\{x remove})∪{x new}

[0084] Among them, x remove represents a random individual position in the historical archive A, x new Indicates the individual to be placed in historical archive A.

[0085] Then, the size of the sub-UAV cluster is adjusted based on the optimization performance of the sub-UAV cluster, that is, the proportion of successful optimization of individual drones in the sub-UAV cluster:

[0086]

[0087] in, For subgroup P k The number of successfully optimized drones, R k is the optimization performance of the kth UAV subgroup in this iteration. According to the optimization performance R k , dynamically adjust the t+1 generation sub-UAV cluster P k t+1 The number of drone individuals N k t+1 , the total number of drones N remains unchanged, To dynamically adjust the t-generation sub-UAV cluster P k t The number of drone individuals in R i is the optimization performance of the k-th drone subgroup in this iteration.

[0088] The UAV cluster path planning and obstacle avoidance process includes the following steps:

[0089] After determining the next generation of detection locations for the drone swarm, dispatch the drones to the corresponding locations and determine the movement path of each drone:

[0090] The goal of minimizing the UAV's moving distance and the time to complete the task is to plan the path, and multiple cost functions in the path planning are combined and optimized in a weighted manner to optimize the objective function F. d as follows:

[0091] F d =min{a·γ+(1-a)·Δ}

[0092] Among them, a represents the weighting coefficient, γ represents the total moving distance of the drone cluster after completing the entire task, and Δ represents the maximum moving time of the drone cluster after completing the task. The specific variables are defined as follows:

[0093]

[0094]

[0095] in, represents the moving distance of the i-th UAV in the t-th generation, Z(t) represents the total moving distance of all UAVs in the t-th generation, represents the moving time of the i-th UAV in the t-th generation, T span (t) represents the maximum movement time of all drones in the tth generation, and N is the number of drones;

[0096] In the path planning process, the cost matrix F(i,j) is used to represent the i-th UAV’s movement from its current position S i (t) to the target position X j The comprehensive cost of (t+1) is calculated as follows:

[0097] F(i,j)=a·μ ij +(1-a)·θ ij

[0098]

[0099] Among them, μ ij is the normalized distance, indicating the length of the path, γ max Indicates the maximum moving distance of the previous generation of drones, θ ij is the normalized time cost, which represents the time required to complete the path, Δ max Indicates the maximum moving time of the previous generation of drones, v i represents the flight speed of the i-th UAV, D(S i (t),X j(t+1 represents the distance from the current position Sit to the target position Xjt+1 of the i-th UAV in the t-th generation;

[0100] Based on the generated cost matrix F(i,j), the optimal movement path of each drone in the drone cluster in this time slice is obtained;

[0101] During the movement of the drone, a real-time obstacle avoidance mechanism based on the force field model is used, and the path is adjusted according to the information fed back by the sensors during flight:

[0102] When the distance between a drone and other drones or obstacles is less than the set safety radius d, a virtual repulsive field is generated to make the drone deviate from the dangerous area. The repulsive force model U rep,ij for:

[0103]

[0104] Among them, d ij is the distance between the i-th UAV and the j-th obstacle or the j-th obstacle,

[0105] The total repulsive force on a drone is the vector sum of its repulsive forces with all other drones and obstacles:

[0106]

[0107] Among them, F i (t) is the total repulsive force of the i-th UAV in the t-th generation, r ij and r io are the unit vectors between the UAV and other UAVs and obstacles,

[0108] When considering collision avoidance between drones, a fitness value priority mechanism is adopted, and the formula is expressed as:

[0109]

[0110] Among them, f(x i ) represents the signal strength of the individual position of the i-th UAV, and the fraction represents the degree of avoidance of the UAV with poor fitness to the UAV with good fitness, U rep,ij It is a repulsion model that adopts the fitness value priority mechanism.

[0111] The updating of the UAV dispatching position includes the following steps:

[0112] Dynamic scheduling based on the current state of the cluster:

[0113] Assume that the current position of the i-th drone is P i (t), the task completion degree is C i(t), the obstacle information is O(t), and the scheduling system dynamically calculates the next position through the following process;

[0114] Calculate the scheduling weight W of each drone i (t), the expression is:

[0115] W i (t) = α·f pos (p i (t))+β·f task (C i (t))+γ·f obs (O(t))

[0116] Among them, α, β, γ are weight coefficients, f pos 、f task , and f obs Respectively represent the influence functions of position, task completion and obstacle information on UAV scheduling;

[0117] f pos (P i (t))=∥P i (t)-P target (t)∥,f task (C i (t))=1-C i (t),

[0118]

[0119] Among them, P target (t) represents the current target position, P o Represents the position of obstacles. Through this weight function, the scheduling system comprehensively considers different mission requirements and dynamically adjusts the position of the UAV so that it always moves in the optimal direction for the mission.

[0120] Continue the UAV swarm collaborative target search steps until the search mission is successfully completed:

[0121] Assuming that the current time is t and the state of target search is S(t), the target search process of the drone cluster can be expressed as:

[0122]

[0123] Among them, v i (t) is the velocity vector of the i-th UAV at time t. When the mission reaches the predetermined target state S goal When the target search task is completed, S(t)→S goal .

[0124] A computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for collaborative control of a drone cluster for target search in a dynamic obstacle environment as described in any one of claims 1 to 6 is implemented.

[0125] A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When executed by the processor, the method for collaborative control of a drone cluster for target search in a dynamic obstacle environment as described in any one of claims 1 to 6 is implemented.

[0126] Beneficial effects

[0127] Compared with the existing technology, the collaborative control method of a drone cluster for target search in a dynamic obstacle environment of the present invention combines real-time environmental perception, data fusion technology and an obstacle avoidance mechanism based on force field, ensuring the safety and stability of the drone cluster in complex environments, avoiding possible collisions and path planning errors of drones in dynamic obstacle environments; the path optimization method based on the cost matrix effectively reduces the total time and flight distance for task completion, further improving the efficiency of task execution.

[0128] The present invention fully considers the changes of obstacles and targets in dynamic environments through a multi-subgroup collaborative optimization strategy. It can optimize the target search efficiency of drone clusters based on real-time perception and flexible decision-making, avoiding the limitations of existing methods that rely on static maps and preset obstacle positions.

[0129] This invention utilizes dynamic subgrouping and collaborative control mechanisms to ensure efficient division of labor and cooperation among drone swarms in complex environments. It also uses information sharing to timely adjust the size and task allocation of each subgroup, thereby achieving collaborative cooperation and global optimization within the swarm. Compared to traditional path planning methods, this invention can flexibly address the impact of dynamic obstacles and rapidly update flight paths and mission strategies, improving the success rate and efficiency of search missions. BRIEF DESCRIPTION OF THE DRAWINGS

[0130] Figure 1 is a method sequence diagram of the present invention;

[0131] Figure 2 Schematic diagram of dynamic obstacles for forest fire rescue target search mission;

[0132] Figure 3 Schematic diagram of the environment model for forest fire rescue target search mission;

[0133] Figure 4 Provides a real-time environment map for target search missions;

[0134] Figure 5 Schematic diagram of the UAV's moving path planning;

[0135] Figure 6 This is a schematic diagram of the force field model for collision and obstacle avoidance of UAV swarms;

[0136] Figure 7a A distribution map of the first generation of positions of a UAV cluster using the method of the present invention;

[0137] Figure 7b A distribution map of the fifth generation positions of the drone cluster using the method of the present invention;

[0138] Figure 7c This is a distribution map of the tenth generation of positions of a UAV cluster using the method of the present invention. DETAILED DESCRIPTION

[0139] In order to provide a further understanding and appreciation of the structural features and effects achieved by the present invention, a detailed description is provided with reference to preferred embodiments and accompanying drawings as follows:

[0140] like Figure 1 As shown, the UAV cluster collaborative control method for target search in a dynamic obstacle environment described in the present invention includes the following steps:

[0141] The first step is to initialize the drone cluster: initialize the search area according to the mission requirements, set the initial flight parameters of the drone cluster, and establish a cluster communication network. The initialization of the drone cluster is the basis for mission execution. It is necessary to model the search area according to the complex environment, set flight parameters, and build a dynamic communication network to ensure that the cluster can collaborate efficiently, adapt to the dynamic environment, and complete the mission safely and reliably.

[0142] (1) According to the relative positions between the UAV clusters, an initial plane coordinate system is established in the target working area. When setting the initial coordinate system C in the target working area, it is assumed that the target area R is:

[0143] R=[x min ,x max ]×[y min ,y max ],

[0144] Among them, x max 、x min and represent the upper and lower boundaries of the x-axis, y max 、y min and represent the upper and lower boundaries of the y-axis respectively.

[0145] (2) Set the initial flight parameters of each UAV, including: the number of sub-UAV clusters N sIt represents the number of neutron drone clusters in the target search of multi-sub-swarm collaborative optimization of drones, and the historical archive size A h Indicates the size of the archive storing historical environmental information and the initial speed v0 of the UAV. These parameters affect the UAV's flight strategy, path planning, and swarm collaboration.

[0146] When setting the initial speed v0 of the drone, the area of ​​the target area, environmental conditions, and the urgency of the swarm mission are taken into consideration. The formula for setting the initial speed and flight mode is:

[0147] v0=MinSpeed+(MaxSpeed-MinSpeed)·β(β∈[0,1]),

[0148] Among them, MinSpeed ​​and MaxSpeed ​​are the minimum and maximum speeds of the drone respectively, and β is a random factor between [0,1], which is used to simulate the randomness of speed.

[0149] (3) The UAV cluster communication network is implemented through the wireless communication protocol Wi-Fi, and its topology is represented by graph theory:

[0150] The construction of cluster communication network is the basis for ensuring the coordinated operation of UAV clusters. In order to ensure that UAVs can coordinate with each other and exchange information, the network topology needs to be dynamically adjusted according to the relative position of each UAV. Each UAV is regarded as a node in the graph. If the distance d between two UAVs is less than 0. ij Less than the communication threshold d th , then there is an edge between the two drones. The formula for constructing the topological structure is as follows:

[0151] G=(V,E),V={1,2,…,N},E={(i,j)∣∣d ij ≤d th}

[0152] Among them, G is the graph structure of the communication network, v is all nodes in the drone cluster, E is the connected edge set, d ij is the distance between UAV i and UAV j, and d th is the set communication distance threshold.

[0153] (4) According to the UAV cluster target search task in a dynamic obstacle environment, a fitness value function f(x) is specified to represent the signal strength of the search target.

[0154] like Figure 2 This is a schematic diagram of dynamic obstacles in the forest fire rescue target search mission. Figure 3A schematic diagram of the forest fire rescue target search task environment model. A fitness value function f(x) is specified based on the forest fire rescue target search environment model.

[0155] x represents the coordinate position of the region. This function is evaluated based on the target signal strength. The stronger the target signal strength, the greater the fitness value of the region. Assuming that the target signal strength is calculated based on the target position x and the environmental condition ∈, the fitness function is expressed as:

[0156]

[0157] Where S(x) is the signal strength of the target at position x, d(x,x goal ) is from the current position x to the target position x goal distance, ∈(x) represents the degree of environmental obstacles or interference at that location, α is the attenuation coefficient, which determines the degree of environmental impact, and x represents the regional coordinate position.

[0158] The fitness value f(x) is used to guide the adjustment of the search strategy. The stronger the target signal strength, the greater the fitness, indicating that the search priority in this area is higher.

[0159] (5) Initialize the location set of the drone cluster:

[0160] Initialize the position set of the UAV cluster. The position initialization of the UAV cluster is allocated according to the mission area and target location. Set the initial position set P init When the location is selected according to the regional division and the distribution of the target, it is assumed that the UAV cluster is in the target area R = [x min ,x max ]×[y min ,y max ] distribution, the initialization position P of each UAV i It is set by the following formula:

[0161] P init ={P1,P2,…,P N},P i =(x i ,y i ),x i ∈[x min ,x max ],y i ∈[y min ,y max ]

[0162] Where N is the number of drones in the cluster, P i is the initial position of the i-th UAV, P initThe position set is the initial state of the UAV cluster. During the mission execution, the cluster dynamically adjusts its position according to target distribution, environmental changes and mission requirements.

[0163] (6) Initialize historical file A:

[0164] Put the position of a random drone cluster in the initialized drone cluster into the historical file A:

[0165] A=A∪x i ,i∈[1,N]

[0166] Historical archive A is used to save the individual positions of drones that are eliminated due to failure in target strength comparison during the iterative position update process during the execution of the drone cluster target search mission.

[0167] The second step is dynamic environment perception and obstacle information acquisition: using drone sensors to collect surrounding environment information in real time, generating real-time environment maps through data fusion technology, and promptly feeding back environmental data to each drone. Dynamic environment perception and obstacle information acquisition are the core links of autonomous flight of drone clusters. They can ensure the safe flight of drones in complex and changing environments, improve the accuracy of path planning and task execution, and provide high-quality environmental data support for cluster collaboration. It is necessary to solve the real-time and accuracy problems of sensor data, handle the fusion and error correction of multiple sensor data, and efficiently generate dynamic environment maps with limited computing resources.

[0168] (1) Collect sensor data:

[0169] The sensors equipped by the drones are used to collect real-time information about changes in the surrounding environment, including other drones, static and dynamic obstacles. The sensor perception range of each drone is set to r sensor , then the perception area of ​​each UAV i is expressed as:

[0170]

[0171] Among them, S i represents the area perceived by UAV i, (x j ,y j ) is the position of the jth object, and r sensor is the maximum sensing radius of the sensor.

[0172] (2) Classification and identification of sensor data:

[0173] According to the perceived data, different types of obstacles and target objects are identified and classified. The drone uses infrared sensors and visual sensors to identify static obstacles. static and dynamic obstacles O dynamic, the target object is represented by G, and the boundary between each obstacle and the target in the environment is approximated by a circular model. The detection formula for static and dynamic obstacles is:

[0174]

[0175] If d i,j ≤r static , d i,j ≤r dynamic , it is considered that UAV i detects static or dynamic obstacle j, where r static and r dynamic are the safety distances of static obstacles and dynamic obstacles, d i,j is the distance between the drone and the obstacle.

[0176] (3) Data fusion technology is used to integrate the input data of multiple sensors. UAV i is equipped with K sensors for data collection. The output of sensor K is D k , then the environmental information D after sensor data fusion fusion , calculated by weighted average method,

[0177]

[0178] Among them, w k is the weight of sensor k.

[0179] (4) Build a real-time environment map:

[0180] The fused data is used to build a real-time environment map, and each UAV i is located according to its current position P i =(x i ,y i ) and the perceived obstacle information D fusion Updated environment map M env , the environment map is represented as a matrix, where each element M i,j Indicates position (x i ,y i ) at the environmental state, the update rules are as follows:

[0181]

[0182] Among them, M env (x i ,y i )=0 means there is no object at this location, M env (x i ,y i )=1 indicates a static obstacle, M env (x i ,y i)=2 means there is a dynamic obstacle, M env (x i ,y i )=3 means that there is a drone individual in the drone cluster at this location, M env (x i ,y i )=4 means that the position is the target object position, Figure 4 It is a real-time environment schematic map for target search tasks.

[0183] The third step is to collaboratively optimize target search among multiple subgroups of drones: the drone cluster is divided into multiple subgroups, each of which is responsible for target search tasks in different areas. The flight missions and path planning of the subgroups are dynamically adjusted according to the target distribution and environmental information. The collaborative optimization of multiple subgroups of drones can improve the efficiency of task execution, achieve comprehensive coverage of the target area through task division and dynamic adjustment, reduce resource waste, and enhance the flexibility and adaptability of drone clusters in complex environments. It is necessary to solve the global optimization problem of subgroup division and task allocation, balance the task load between subgroups, ensure real-time adjustment in a dynamic environment, and achieve efficient collaboration.

[0184] (1) Divide the entire drone cluster into three drone subgroups:

[0185] Initially, the cluster is divided into three sub-clusters: exploration drone sub-cluster P1, development drone sub-cluster P2, and archive drone sub-cluster P3.

[0186] The size of the initial drone sub-cluster satisfies the following conditions: N1+N2+N3=N. The initial division of drones within the sub-cluster is completed by random selection:

[0187]

[0188] Among them, N1, N2, and N3 represent the number of drones in the subgroups P1, P2, and P3 respectively. N1, N2, and N3 are positive integers. In the later stage of target search, the size of a single sub-drone cluster is allowed to be 0. k Represents the subgroup P k Collection of drones included.

[0189] (2) Setting subgroup target search strategy:

[0190] The flight mission and path planning of each sub-group will be dynamically adjusted according to the target distribution and environmental information to ensure the optimality of the overall search strategy of the cluster.

[0191] Explore drone sub-clusters P1: Explore drone sub-clusters for global search, generating the next generation of positions near the positions of random drones in the current drone cluster, Indicates the j-th dimension coordinate value of the i-th drone in the current t-th iteration time slice exploring the drone sub-cluster P1, and exploring the position of each drone in the drone sub-cluster P1 The update strategy is:

[0192]

[0193] Randomly select the j-th dimension coordinate value p of the position of an individual drone in the drone cluster in the current t-th iteration time slice rj As the reference individual position, the current position is updated based on its position and random disturbance, rand is a random number in [0,1], and r is a dynamically decreasing random factor;

[0194] Develop drone sub-cluster P2: Develop drone sub-cluster responsible for local search, generating the next generation of positions around the current drone cluster's best signal strength position, Indicates the j-th dimension coordinate value of the i-th drone in the current t-th iteration time slice of the drone sub-cluster P2, and the position of each drone The update formula is:

[0195]

[0196] Among them, d rj Indicates the j-dimension coordinate value of the current UAV individual position and the j-dimension coordinate value of the global optimal position g j distance, add random direction v rj To enhance diversity, update the position in g j Randomly generate nearby and gradually converge to the optimal solution area;

[0197] Archive drone sub-cluster P3: The archive drone sub-cluster uses the historical optimal solution to search and generates the next generation of positions near the positions of individual drones eliminated in the early search process. It represents the j-th dimension coordinate value of the i-th drone in the drone subcluster P3 of the current t-th iteration time slice archive. Its position update formula is:

[0198]

[0199] Among them, a rj The historical optimal solution is randomly selected from the historical archive A. The historical archive provides a stable and diverse reference by storing solutions with improved fitness over generations.

[0200] (3) Information sharing and dynamic adjustment of subgroup size:

[0201] After obtaining the individual positions of the next generation of drones based on the different strategies of the three sub-swarms, the drones exchange information with each other. Each drone compares its current position information with the position information of the previous time slice to determine whether it is closer to the target. If successful, it will be incorporated into the winner drone cluster of this generation. If unsuccessful, the current position will be stored in the position history archive.

[0202] For the location set X of all drone clusters in the tth time slice t ={x1 t ,x2 t ,…,x N t}, updated winner drone cluster S t The relationship with location history file A is expressed as:

[0203] S t ={x i t ∈X t |f(x i t )≥f(x i t-1 )}

[0204] A=A∪{x i t |x i t ∈X t and f(x i t ) <f(x i t-1 )}

[0205] Among them, f(x i t ) represents the target signal strength value of the i-th UAV at the t-th time slice;

[0206] When placing individual locations into historical archive A, if the number of individual locations exceeds the size of the historical archive A h When , first randomly delete an individual position, and then perform the add operation:

[0207] x remove ∈A

[0208] A=(A\{x remove})∪{x new}

[0209] Among them, x remove represents a random individual position in the historical archive A, x new Indicates the individual to be placed in historical archive A.

[0210] Then, the size of the sub-UAV cluster is adjusted based on the optimization performance of the sub-UAV cluster, that is, the proportion of successful optimization of individual drones in the sub-UAV cluster:

[0211]

[0212] in, For subgroup P k The number of successfully optimized drones, R k is the optimization performance of the kth UAV subgroup in this iteration. According to the optimization performance R k , dynamically adjust the t+1 generation sub-UAV cluster P k t+1 The number of drone individuals N k t+1 , the total number of drones N remains unchanged, To dynamically adjust the t-generation sub-UAV cluster P k t The number of drone individuals in R i is the optimization performance of the k-th drone subgroup in this iteration.

[0213] The fourth step is drone cluster path planning and obstacle avoidance: calculate the optimal movement path for each drone, adopt a real-time obstacle avoidance mechanism, and adjust the path according to the information fed back by the sensors during flight. Drone cluster path planning and obstacle avoidance are key links to ensure the smooth execution of the mission. Through optimal path calculation and real-time obstacle avoidance, flight efficiency can be improved, energy consumption can be reduced, and the safety and mission execution capabilities of drones in complex and dynamic environments can be guaranteed. It is necessary to cope with the high computational complexity of real-time path updates in dynamic environments and solve path conflicts and coordination problems between drones.

[0214] (1) After determining the next generation of detection locations for the drone cluster, dispatch the drones to the corresponding locations and determine the movement path of each drone:

[0215] The planning objectives mainly include the following two aspects: minimizing the UAV's moving distance, which is directly related to energy consumption and equipment wear, and minimizing the time to complete the task, which is particularly important for scenarios with high mission urgency. The goal planning of minimizing the UAV's moving distance and minimizing the time to complete the task is to combine and optimize multiple cost functions in path planning in a weighted manner to optimize the objective function F. d as follows:

[0216] F d =min{a·γ+(1-a)·Δ}

[0217] Among them, a represents the weighting coefficient, γ represents the total moving distance of the drone cluster after completing the entire task, and Δ represents the maximum moving time of the drone cluster after completing the task. The specific variables are defined as follows:

[0218]

[0219] in, represents the moving distance of the i-th UAV in the t-th generation, Z(t) represents the total moving distance of all UAVs in the t-th generation, represents the moving time of the i-th UAV in the t-th generation, T span (t) represents the maximum movement time of all drones in the tth generation, and N is the number of drones;

[0220] In the path planning process, the cost matrix F(i,j) is used to represent the i-th UAV’s movement from its current position S i (t) to the target position X j The comprehensive cost of (t+1) is calculated as follows:

[0221] F(i,j)=a·μ ij +(1-a)·θ ij

[0222]

[0223] Among them, μ ij is the normalized distance, indicating the length of the path, γ max Indicates the maximum moving distance of the previous generation of drones, θ ij is the normalized time cost, which represents the time required to complete the path, Δ max Indicates the maximum moving time of the previous generation of drones, v i represents the flight speed of the i-th UAV, D(S i (t),X j (t+1 represents the distance from the current position Sit to the target position Xjt+1 of the i-th UAV in the t-th generation;

[0224] Based on the generated cost matrix F(i,j), the optimal moving path of each drone in the drone cluster in this time slice is obtained. The schematic diagram of drone moving path planning is as follows: Figure 5 shown.

[0225] (2) A real-time obstacle avoidance mechanism based on the force field model is used during the movement of the UAV, and the path is adjusted according to the information fed back by the sensors during the flight:

[0226] When the distance between a drone and other drones or obstacles is less than the set safety radius d, a virtual repulsive field is generated to make the drone deviate from the dangerous area. The repulsive force model Urep,ij for:

[0227]

[0228] Among them, d ij is the distance between the i-th UAV and the j-th obstacle or the j-th obstacle, and the UAV cluster collision avoidance force field model is shown in the figure below: Figure 6 shown.

[0229] The total repulsive force on a drone is the vector sum of its repulsive forces with all other drones and obstacles:

[0230]

[0231] Among them, F i (t) is the total repulsive force of the i-th UAV in the t-th generation, r ij and r io are the unit vectors between the UAV and other UAVs and obstacles,

[0232] When considering collision avoidance between drones, a fitness value priority mechanism is adopted, and the formula is expressed as:

[0233]

[0234] Among them, f(x i ) represents the signal strength of the individual position of the i-th UAV, and the fraction represents the degree of avoidance of the UAV with poor fitness to the UAV with good fitness, U rep,ij It is a repulsion model that adopts the fitness value priority mechanism.

[0235] Step 5: Update the UAV's dispatch position: The dispatch system dynamically dispatches the drone based on the current state of the cluster. After the drone is dispatched to the new position, the drone cluster continues the collaborative target search process until the target search mission is successfully completed. Updating the drone's dispatch position is the core link in achieving dynamic cluster response and optimizing resource utilization. It can adjust the drone distribution based on real-time status to avoid resource waste and missed mission areas, improving the comprehensiveness and efficiency of target search and the accuracy of mission completion. It also requires solving the real-time optimization problem of UAV position scheduling in a dynamic environment, balancing mission priority, UAV coverage efficiency, and obstacle avoidance, while also addressing communication delays, state feedback uncertainty, and collaborative conflicts between multiple UAVs.

[0236] (1) Dynamic scheduling based on the current state of the cluster:

[0237] The state of the cluster includes factors such as the current position of the drone, the completion of the task, and information about obstacles. The scheduling system calculates the optimal scheduling order for each drone in real time to ensure the efficiency and safety of the cluster when performing tasks. Assume that the current position of the i-th drone is P i (t), the task completion degree is C i (t), the obstacle information is O(t), and the scheduling system dynamically calculates the next position through the following process;

[0238] Calculate the scheduling weight W of each drone i (t), the expression is:

[0239] W i (t) = α·f pos (P i (t))+β·f task (C i (t))+γ·f obs (O(t))

[0240] Among them, α, β, γ are weight coefficients, f pos 、f task , and f obs Respectively represent the influence functions of position, task completion and obstacle information on UAV scheduling;

[0241] f pos (P i (t)0=∥P i (t)-P target (t)∥,f task (C i (t))=1-C i (t),

[0242]

[0243] Among them, P target (t) represents the current target position, P o Represents the position of the obstacle. Through this weight function, the scheduling system comprehensively considers different task requirements and dynamically adjusts the position of the UAV so that it always moves in the optimal direction for the task.

[0244] like Figure 7a 、 7b 7c and 7d show the distribution of individual drones at three different moments during the target search mission. The first generation of drones starts the target search and determines the initial position of each drone. Then, the drones begin to search for targets and update their positions based on multi-subgroup collaborative optimization. Figure 7c As shown, the drone has locked onto the target area.

[0245] (2) Continue the UAV cluster collaborative target search steps until the search mission is successfully completed:

[0246] Assuming that the current time is t and the state of target search is S(t), the target search process of the drone cluster can be expressed as:

[0247]

[0248] Among them, v i (t) is the velocity vector of the i-th UAV at time t. When the mission reaches the predetermined target state S goal When the target search task is completed, S(t)→S goal .

[0249] Also provided herein is a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, a method for cooperatively controlling a swarm of drones for target search in a dynamic obstacle environment is implemented. Also provided herein is a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, a method for cooperatively controlling a swarm of drones for target search in a dynamic obstacle environment is implemented.

[0250] To efficiently address the problem of target search and collaborative control of drone swarms in dynamic obstacle environments, this paper proposes a "target search and collaborative control method for drone swarms in dynamic obstacle environments." This method, through a multi-subgroup collaborative optimization strategy and a real-time dynamic adjustment mechanism, fully accounts for the dynamic changes in targets and obstacles in the environment, enabling drone swarms to perform target search and obstacle avoidance in complex environments. Combined with dynamic subgroup partitioning and information sharing techniques, it can flexibly adjust task allocation based on environmental changes, ensuring that drone swarms can continue to efficiently execute their tasks in the face of emergencies.

[0251] By integrating environmental perception, data fusion, and force field obstacle avoidance, this approach overcomes the limitations of traditional methods that rely on static models, enabling drone swarms to safely and stably perform target search and path planning in dynamic environments. Combined with cost matrix path optimization, this approach further improves mission execution efficiency, reduces flight distance and time, and enhances the overall collaborative capabilities of drone swarms.

[0252] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for cooperative control of a drone swarm for target search in a dynamic obstacle environment, characterized in that: The following steps are involved: 11) Initialization of the UAV swarm: Initialize the search area according to the mission requirements, set the initial flight parameters of the UAV swarm, and establish the swarm communication network; 12) Dynamic environment perception and obstacle information acquisition: Utilize drone sensors to collect real-time environmental information, generate real-time environmental maps through data fusion technology, and promptly feed back environmental data to each drone; 13) UAV multi-sub-swarm collaborative optimization of target search: The UAV cluster is divided into multiple sub-swarms, each responsible for target search tasks in different areas, and the flight mission and path planning of the sub-swarm are dynamically adjusted based on target distribution and environmental information; 14) UAV swarm path planning and obstacle avoidance: Calculate the optimal movement path for each drone, use a real-time obstacle avoidance mechanism, and adjust the path during flight based on sensor feedback; The UAV cluster path planning and obstacle avoidance process includes the following steps: 141) After determining the next generation of detection locations for the drone swarm, dispatch the drones to the corresponding locations and determine the movement path of each drone: The goal of minimizing the UAV's moving distance and the time to complete the task is to plan the path, and multiple cost functions in the path planning are combined and optimized in a weighted manner to optimize the objective function F. d as follows: F d =min{a·γ+(1-a)·Δ} Among them, a represents the weighting coefficient, γ represents the total moving distance of the drone cluster after completing the entire task, and Δ represents the maximum moving time of the drone cluster after completing the task. The specific variables are defined as follows: in, represents the moving distance of the i-th UAV in the t-th generation, Z(t) represents the total moving distance of all UAVs in the t-th generation, represents the moving time of the i-th UAV in the t-th generation, T span (t) represents the maximum movement time of all drones in the tth generation, and N is the number of drones; In the path planning process, the cost matrix F(i,j) is used to represent the i-th UAV’s movement from its current position S i (t) to the target position X j The comprehensive cost of (t+1) is calculated as follows: F(i,j)=a·μ ij +(1-a)·θ ij Among them, μ ij is the normalized distance, indicating the length of the path, γ max Indicates the maximum moving distance of the previous generation of drones, θ jj is the normalized time cost, which represents the time required to complete the path, Δ max Indicates the maximum moving time of the previous generation of drones, v i represents the flight speed of the i-th UAV, D(S i (t),X j (t+1)) indicates that the i-th UAV in the t-th generation moves from its current position S i (t) to the target position X j The distance to (t+1); Based on the generated cost matrix F(i,j), the optimal movement path of each drone in the drone cluster in this time slice is obtained; 142) A real-time obstacle avoidance mechanism based on a force field model is used during the movement of the drone, and the path is adjusted according to the information fed back by the sensors during flight: When the distance between a drone and other drones or obstacles is less than the set safety radius d, a virtual repulsive field is generated to make the drone deviate from the dangerous area. The repulsive force model U rep,ij for: Among them, d ij is the distance between the i-th UAV and the j-th obstacle or the j-th obstacle, The total repulsive force on a drone is the vector sum of its repulsive forces with all other drones and obstacles: Among them, F i (t) is the total repulsive force of the i-th UAV in the j-th generation, r ij and r io are the unit vectors between the UAV and other UAVs and obstacles, When considering collision avoidance between drones, a fitness value priority mechanism is adopted, and the formula is expressed as: Among them, f(x i ) represents the signal strength of the individual position of the i-th UAV, and the fraction represents the degree of avoidance of the UAV with poor fitness to the UAV with good fitness, U rep,ij It is a repulsion model that adopts the fitness value priority mechanism; 15) Update of UAV dispatch position: The dispatch system performs dynamic dispatch based on the current state of the cluster. After the UAV is dispatched to the new position, the UAV cluster collaborative target search step continues until the target search mission is successfully completed. Assuming that the current time is t and the state of target search is S(t), the target search process of the drone cluster can be expressed as: Among them, v i (t) is the velocity vector of the i-th UAV at time t. When the mission reaches the predetermined target state S goal When the target search task is completed, S(t)→S goal .

2. The method for cooperative control of a drone swarm for target search in a dynamic obstacle environment according to claim 1, characterized in that: The initialization of the drone cluster includes the following steps: 21) Based on the relative positions between the UAV clusters, an initial plane coordinate system is established in the target working area. When setting the initial coordinate system C in the target working area, it is assumed that the target area R is: R=[x min ,x max ]×[y min ,y max ], Among them, x max 、x min and represent the upper and lower boundaries of the x-axis, y max 、y min and represent the upper and lower boundaries of the y-axis respectively; 22) Set the initial flight parameters for each drone, including: the number of sub-drone clusters N s It represents the number of neutron drone clusters in the target search of multi-sub-swarm collaborative optimization of drones, and the historical archive size A h Indicates the size of the archive storing historical environmental information, and the initial speed v0 of the drone; The setting formula for initial speed and flight mode is: v0=MinSpeed+(MaxSpeed-MinSpeed)·β(β∈[0,1]), Among them, MinSpeed ​​and MaxSpeed ​​are the minimum and maximum speeds of the drone respectively, and β is a random factor between [0,1], which is used to simulate the randomness of speed; 23) Assume that the drone cluster communication network is implemented through the wireless communication protocol Wi-Fi, and its topology is represented by graph theory: Each drone is considered as a node in the graph. If the distance d between two drones ij Less than the communication threshold d th , then there is an edge between the two drones. The formula for constructing the topological structure is as follows: G=(V,E),V={1,2,…,N},E={(i,j)∣∣d ij ≤d th } Among them, G is the graph structure of the communication network, V is all the nodes in the drone cluster, E is the connected edge set, d ij is the distance between UAV i and UAV j, and d th is the set communication distance threshold; 24) According to the UAV cluster target search task in a dynamic obstacle environment, a fitness value function f(x) is specified to represent the signal strength of the search target: Assuming that the target signal strength is calculated based on the target position x and environmental conditions ∈, the fitness function is expressed as: Where β(x) is the signal strength of the target at position x, d(x,x goal ) is from the current position x to the target position x goal distance, ∈(x) represents the degree of environmental obstacles or interference at that location, α is the attenuation coefficient, which determines the degree of environmental impact, and x represents the regional coordinate position. The fitness value f(x) is used to guide the adjustment of the search strategy. The stronger the target signal strength, the greater the fitness, indicating that the search priority in this area is higher; 25) Initialize the location set of the drone cluster: Set the initial position set P init When the location is selected according to the regional division and the distribution of the target, it is assumed that the UAV cluster is in the target area R = [x min ,x max ]×[y min ,y max ] distribution, the initialization position P of each UAV i It is set by the following formula: P init ={P1,P2,…,P N },P i =(x i ,y i ),x i ∈[x min ,x max ],y i ∈ [and min ,and max ] Where N is the number of drones in the cluster, P i is the initial position of the i-th UAV, P init The position set is the initial state of the UAV cluster. During the mission execution, the cluster dynamically adjusts its position according to target distribution, environmental changes, and mission requirements. 26) Initialize history file A: Put the position of a random drone cluster in the initialized drone cluster into the historical file A: A=A∪x i ,i∈[1,N] Historical archive A is used to save the individual positions of drones that are eliminated due to failure in target strength comparison during the iterative position update process during the execution of the drone cluster target search mission.

3. The method for cooperative control of a drone swarm for target search in a dynamic obstacle environment according to claim 1, characterized in that: The dynamic environment perception and obstacle information acquisition includes the following steps: 31) Collect sensor data: The sensors equipped by the drones are used to collect real-time information about changes in the surrounding environment, including other drones, static and dynamic obstacles. The sensor perception range of each drone is set to r sensor , then the perception area of ​​each UAV i is expressed as: Among them, S i represents the area perceived by UAV i, (x j ,y j ) is the position of the jth object, and r sensor is the maximum sensing radius of the sensor; 32) Classification and identification of sensor data: According to the sensed data, different types of obstacles and target objects are identified and classified. The drone uses infrared sensors and visual sensors to identify static obstacles. static and dynamic obstacles O dynamic , the target object is represented by G, and the boundary between each obstacle and the target in the environment is approximated by a circular model. The detection formula for static and dynamic obstacles is: If d i,j ≤r static , d i,j ≤r dynamic , it is considered that UAV i detects static or dynamic obstacle j, where r static and r dynamic are the safety distances of static obstacles and dynamic obstacles, d i,j is the distance between the drone and the obstacle; 33) Data fusion technology is used to integrate the input data of multiple sensors. UAV i is equipped with K sensors for data collection. The output of sensor K is D k , then the environmental information D after sensor data fusion fusion , calculated by weighted average method, Among them, w k is the weight of sensor k; 34) Build real-time environment map: The fused data is used to build a real-time environment map, and each UAV i is located according to its current position P i =(x i ,y i ) and the perceived obstacle information D fusion Updated environment map M env , the environment map is represented as a matrix, where each element M i,j Indicates position (x i ,y i ) at the environmental state, the update rules are as follows: Among them, M env (x i ,y i )=0 means there is no object at this location, M env (x i ,y i )=1 indicates a static obstacle, M env (x i ,y i )=2 means there are dynamic obstacles, M env (x i ,y i )=3 means that there is a drone individual in the drone cluster at this location, M env (x i ,y i )=4 means that the position is the target object position.

4. The method for cooperative control of a drone swarm for target search in a dynamic obstacle environment according to claim 1, characterized in that: The multi-subgroup collaborative optimization target search of UAVs includes the following steps: 41) Divide the entire drone cluster into three drone sub-groups: Initially, the cluster is divided into three sub-clusters: exploration drone sub-cluster P1, development drone sub-cluster P2, and archive drone sub-cluster P3. The size of the initial drone sub-cluster satisfies the following conditions: N1+N2+N3=N. The initial division of drones within the sub-cluster is completed by random selection: Among them, N1, N2, and N3 represent the number of drones in the subgroups P1, P2, and P3 respectively. N1, N2, and N3 are positive integers. In the later stage of target search, the size of a single sub-drone cluster is allowed to be 0. k Represents the subgroup P k The included drone collection; 42) Set subgroup target search strategy: Explore drone sub-clusters P1: Explore drone sub-clusters for global search, generating the next generation of positions near the positions of random drones in the current drone cluster, Indicates the j-th dimension coordinate value of the i-th drone in the current t-th iteration time slice exploring the drone sub-cluster P1, and exploring the position of each drone in the drone sub-cluster P1 The update strategy is: Randomly select the j-th dimension coordinate value p of the position of an individual drone in the drone cluster in the current t-th iteration time slice rj As the reference individual position, the current position is updated based on its position and random disturbance, rand is a random number in [0,1], and r is a dynamically decreasing random factor; Develop drone sub-cluster P2: Develop drone sub-cluster responsible for local search, generating the next generation of positions around the current drone cluster's best signal strength position, Indicates the j-th dimension coordinate value of the i-th drone in the current t-th iteration time slice of the drone sub-cluster P2, and the position of each drone The update formula is: V rj =2·row-1 Among them, d rj Indicates the j-dimension coordinate value of the current UAV individual position and the j-dimension coordinate value of the global optimal position g j distance, add random direction v rj To enhance diversity, update the position in g j Randomly generate nearby and gradually converge to the optimal solution area; Archive drone sub-cluster P3: The archive drone sub-cluster uses the historical optimal solution to search and generates the next generation of positions near the positions of individual drones eliminated in the early search process. It represents the j-th dimension coordinate value of the i-th drone in the drone subcluster P3 of the current t-th iteration time slice archive. Its position update formula is: Among them, a rj The historical optimal solution is randomly selected from the historical archive A. The historical archive provides a stable and diverse reference by storing solutions with improved fitness over generations; 43) Information sharing and dynamic adjustment of subgroup size: After obtaining the individual positions of the next generation of drones based on the different strategies of the three sub-swarms, the drones exchange information with each other. Each drone compares its current position information with the position information of the previous time slice to determine whether it is closer to the target. If successful, it will be incorporated into the winner drone cluster of this generation. If unsuccessful, the current position will be stored in the position history archive. For the location set X of all drone clusters in the tth time slice t ={x1 t ,x2 t ,…,x N t }, updated winner drone cluster S t The relationship with location history file A is expressed as: S t ={x i t ∈X t |f(x i t )≥f(x i t-1 )} A=A∪{x i t |x i t ∈X t and f(x i t )<f(x i t-1 )} Among them, f(x i t ) represents the target signal strength value of the i-th UAV at the t-th time slice; When placing individual locations into historical archive A, if the number of individual locations exceeds the size of the historical archive A h When , first randomly delete an individual position, and then perform the add operation: x remove ∈A A=(A\{x remove })∪{x new } Among them, x remove represents a random individual position in the historical archive A, x new Indicates the individual to be placed in historical archive A. Then, the size of the sub-UAV cluster is adjusted based on the optimization performance of the sub-UAV cluster, that is, the proportion of successful optimization of individual drones in the sub-UAV cluster: in, For subgroup P k The number of successfully optimized drones, R k is the optimization performance of the kth UAV subgroup in this iteration. According to the optimization performance R k , dynamically adjust the t+1 generation sub-UAV cluster P k t+1 The number of drone individuals N k t+1 , the total number of drones N remains unchanged, To dynamically adjust the t-generation sub-UAV cluster P k t The number of drone individuals in R i is the optimization performance of the k-th drone subgroup in this iteration.

5. The method for cooperative control of a drone swarm for target search in a dynamic obstacle environment according to claim 1, characterized in that: The updating of the UAV dispatching position includes the following steps: 51) Dynamic scheduling based on the current state of the cluster: Assume that the current position of the i-th drone is P i (t), the task completion degree is C i (t), the obstacle information is O(t), and the scheduling system dynamically calculates the next position through the following process; Calculate the scheduling weight W of each drone i (t), the expression is: W i (t)=α·f pos (P i (t))+β·f task (C i (t))+γ·f obs (O(t)) Among them, α, β, γ are weight coefficients, f pos 、f task , and f obs Respectively represent the influence functions of position, task completion and obstacle information on UAV scheduling; f pos (P i (t))=||P i (t)-P target (t)||,f task (C i (t))=1-C i (t), Among them, P target (t) represents the current target position, P o Represents the position of obstacles. Through this weight function, the scheduling system comprehensively considers different mission requirements and dynamically adjusts the position of the UAV so that it always moves in the optimal direction for the mission. 52) Continue the UAV cluster collaborative target search steps until the search mission is successfully completed.

6. A computer-readable storage medium, characterized in that: A computer program is stored on the storage medium. When the computer program is executed by the processor, the drone cluster collaborative control method for target search in a dynamic obstacle environment as described in any one of claims 1 to 5 can be implemented.

7. A computer device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, a drone cluster collaborative control method for target search in a dynamic obstacle environment as described in any one of claims 1 to 5 can be implemented.

Images