Multi-unmanned aerial vehicle cooperative search method based on dynamic region division and multi-constraint virtual force field

Abstract

The invention discloses a multi-unmanned aerial vehicle cooperative search method based on dynamic region division and a multi-constraint virtual force field, and aims to solve the problems of low search efficiency, unbalanced load, difficulty in guaranteeing the motion safety of multiple unmanned aerial vehicles and the like in multi-unmanned aerial vehicle coverage search. Firstly, a task area, an unmanned aerial vehicle kinematics model and a sensor sensing model are constructed; then, dynamic region division is carried out based on a weighted Voronoi diagram, the search progress of the unmanned aerial vehicle, the responsibility region area and the obstacle distribution condition are comprehensively considered, and self-adaptive balancing of task loads is achieved; and selecting a target point based on a utility function, and finally designing a multi-constraint virtual force field control method consisting of target attraction, obstacle avoidance force, regional barrier force and exploration force, generating an expected speed meeting speed constraint and obstacle avoidance requirements, and realizing stable and efficient coverage search. The result shows that the multi-machine collaborative search efficiency can be effectively improved, and the method is suitable for disaster search, routing inspection monitoring, autonomous detection in a complex environment and other scenes.

Description

Technical Field

[0001] This invention belongs to the field of autonomous cooperative control and intelligent perception of unmanned systems, especially the direction of multi-UAV cooperative search, and particularly relates to a multi-UAV cooperative search method with dynamic region division and multi-constraint virtual force field. Background Technology

[0002] With the widespread use of drones in disaster relief, security patrols, forest monitoring, agricultural assessment, and environmental exploration, multi-drone collaborative coverage search has become a key technology for improving mission efficiency and reliability. In complex environments, to achieve full perception of a large area within a limited time, multiple drones need to achieve spatial division of labor, continuous coverage, and global collaboration while avoiding mutual interference. However, existing technologies still have significant shortcomings in mission area division, dynamic load balancing, multi-drone collaborative motion control, and ensuring coverage uniformity.

[0003] Existing coverage search methods mainly include rule-based path planning, potential field-based methods, and region partitioning methods. Rule-based path planning methods, such as spiral routes and matrix scanning, are effective in single-drone scenarios, but in multi-drone collaboration, they cannot adaptively adjust paths according to region coverage progress, leading to severe overlapping coverage and low global coverage efficiency. Potential field methods generate paths through target attraction and obstacle repulsion, but traditional methods struggle to handle dynamic obstacle avoidance between multiple drones, and the potential field design lacks consideration for coverage, often resulting in coverage gaps or localized over-coverage. Region partitioning methods can establish independent responsibility zones for multiple drones, but existing solutions often use fixed weights or geometric distances as the partitioning criteria, failing to achieve dynamic load balancing based on task progress, leading to coverage lag in some areas.

[0004] The paper "Multi-UAV Cooperative Search and Experimental Verification under Limited Information Conditions" addresses the cooperative search problem of multiple UAVs under limited information conditions, studying a cooperative search mechanism based on local information interaction to achieve coverage search of the target area by multiple UAVs. This method reduces dependence on global information to some extent and improves the distributed characteristics of the system. However, its search area division uses fixed weights, resulting in fixed sub-regions. It lacks an explicit adjustment mechanism for the real-time search progress and regional load of UAVs, easily leading to problems such as local area search lag or unbalanced task load during the search process. The paper "Multi-UAV Cooperative Spiral Search Optimization Algorithm Based on Minimum Circle Coverage" takes an environmental modeling and trajectory planning perspective, using minimum circle coverage and regular hexagonal splicing methods to discretize the search area, and combining a spiral search strategy and energy consumption model to achieve path planning and task allocation for multiple UAVs. This method has good coverage efficiency in regular or convex polygonal areas, but its core idea still belongs to the category of static area modeling and discrete trajectory planning. Task division is mainly based on the equal distribution of overall range or energy consumption, making it difficult to dynamically adjust according to the real-time status of UAVs, coverage conditions, and environmental changes during the search process. Furthermore, this method does not adequately consider obstacle avoidance, maintaining safe distances, and multi-force constraint coordination during continuous UAV movement, leaving room for improvement in motion safety and robustness in complex environments. Patent CN115857543A, "A Multi-UAV Cooperative Search Method Based on Prior Information," proposes a multi-UAV search method that integrates prior probability information. This method discretizes the search area into a pseudo-discrete grid, utilizes prior probability information of target appearance, employs a clustering algorithm to initially divide the search area into tasks, and resolves region conflicts through an auction mechanism. It further generates search paths within each task area based on a probability map. This method can improve search targeting in scenarios where target probability information is relatively reliable; however, its task allocation method remains static, and the region division results are difficult to continuously and dynamically adjust based on the real-time search progress of the UAVs and environmental complexity during the search process. Therefore, the efficiency of multi-UAV cooperative search still has room for improvement. Summary of the Invention

[0005] To address the problems of existing technologies, this invention provides a multi-UAV cooperative search method based on dynamic region partitioning and multi-constraint virtual force fields. The aim is to solve problems such as low search efficiency, uneven load, and difficulty in ensuring the safety of multi-UAV movement in multi-UAV coverage search. First, a task area, UAV kinematic model, and sensor perception model are constructed. Then, dynamic region partitioning is performed based on a weighted Voronoi diagram, comprehensively considering UAV search progress, area of ​​responsibility, and obstacle distribution to achieve adaptive load balancing. Next, target points are selected based on utility functions. Finally, a multi-constraint virtual force field control method is designed, consisting of target attraction, obstacle avoidance force, area barrier force, and exploration force, to generate a desired speed that meets speed constraints and obstacle avoidance requirements, achieving stable and efficient coverage search.

[0006] To achieve the above objectives, the present invention employs the following specific technical solutions:

[0007] A multi-UAV cooperative search method with dynamic region partitioning and multi-constraint virtual force fields includes the following steps:

[0008] S1: Construct a model covering the search task region, which includes the following sub-steps:

[0009] S1.1: Define the UAV coverage search mission area as a two-dimensional closed region Ω = [0, L x ]×[0,L y ];

[0010] S1.2: Set K static, impassable obstacles within the area. The obstacles are represented by cylinders, and the center point of the bottom surface of each obstacle is o. k Radius r k All heights are randomly assigned, with the height assumed to be infinite.

[0011] S1.3: Set coverage target: Within a finite task time T, ensure that the cumulative coverage of any point q within the region meets the threshold requirement.

[0012]

[0013] In the formula, C(q,T) represents the cumulative coverage of region point q, C tar It is a set threshold; only when C(q,T) is greater than or equal to C tar Only when the time is right is it determined that point q in the region has been scanned;

[0014] S2: Constructing the UAV kinematic model and setting constraints, which includes the following sub-steps:

[0015] S2.1: To characterize the motion state of the UAV, the kinematic model of the UAV is constructed as follows. Assume the system contains N UAVs, each numbered i = 1, 2, ..., N. The position and velocity of each UAV at time t are as follows:

[0016] p i (t)=[x i (t),y i (t),z i (t)] T (2)

[0017] v i (t)=[v i,x (t),v i,y (t),v i,z (t)] T (3)

[0018] In the formula, x i (t), y i (t), z i (t) represents the x-axis, y-axis, and z-axis coordinates of the UAV at time t, respectively. i,x (t), v i,y (t), v i,z (t) represents the x-axis velocity, y-axis velocity, and z-axis velocity of the UAV at time t;

[0019] S2.2: To ensure flight safety and feasibility, the following constraints are set:

[0020]

[0021] In the formula, a max It is the maximum acceleration of the drone, z min and z max These are the minimum and maximum altitudes of the drone during operation, v max (t) represents the maximum speed of the UAV during flight, and an adaptive adjustment strategy is adopted based on the full coverage H. global (t) Dynamically adjust the upper limit of speed, specifically as follows:

[0022]

[0023] In the formula, This is the base maximum speed, in the early stages of the search, when the global coverage H... global When (t) approaches 0, the drone uses a higher speed to quickly explore unknown areas; in the later stages of the search, when H... global When (t) approaches 1, the drone reduces its flight speed to achieve fine coverage and improve perception quality;

[0024] S3: Constructing the sensor perception model. The top-down visual sensor perception model on the UAV comprehensively considers the effects of distance attenuation and altitude attenuation. Its perception range can be represented as a circular area that varies with altitude, with a radius of:

[0025] R i (t)=z i (t)tanα (6)

[0026] In the formula, α is the sensor's half-field of view angle. For any point q∈Ω on the ground, the instantaneous perception intensity of the i-th UAV is defined as:

[0027]

[0028] In the formula, Φ(ρ) is the distance attenuation function, which is defined as:

[0029]

[0030] The distance decay function describes the characteristic that the sensing intensity decreases twice with the increase of the normalized distance ρ, satisfying the boundary conditions Φ(0)=1 and Φ(1)=0, ensuring the physical characteristic that the sensing intensity is maximum directly below the sensor and zero at the sensing boundary; Ψ(z i ) is the height decay function, which is defined as:

[0031]

[0032] The altitude attenuation function quantifies how perceived quality changes with flight altitude; higher altitudes offer a wider field of view but lower resolution, while lower altitudes offer a narrower field of view but higher resolution. i (t) is the adaptive sensing gain coefficient, which is dynamically adjusted according to the current mission state of the UAV, and is defined as follows:

[0033] M i (t)=M base (1+η(1-P i (t))) (10)

[0034] In the formula, M base It is the basic sensing gain coefficient, P i (t) represents the coverage rate of the drone's area of ​​responsibility, and η is the adjustment coefficient;

[0035] S4: The coverage metric defines a method for quantifying the cumulative sensed information of each point within the task area, which includes the following sub-steps;

[0036] S4.1: The cumulative coverage of region point q at time t is defined as:

[0037]

[0038] In the formula, λ>0 is the coverage forgetting factor, which makes the coverage of areas that have not been visited for a long time gradually decrease. This cumulative coverage model with forgetting mechanism simulates the aging process of environmental information, ensuring that the cumulative coverage of areas that have not been explored for a long time will naturally decrease over time, thus becoming a target that attracts drones again, effectively avoiding the algorithm from getting stuck in local optima.

[0039] S4.2: Overall system coverage progress is measured by global coverage rate H. global (t) quantization, which is defined as:

[0040]

[0041] The condition for determining the completion of the coverage search task is H. global (t)≥H tar H tar ∈(0,1] is a preset threshold for determining the completion of the search task;

[0042] S4.3: To evaluate the load balancing of a multi-UAV system, the coverage area of ​​responsibility for the i-th UAV is defined:

[0043]

[0044] In the formula, Ω i (t) represents the responsibility sub-region of the i-th UAV at time t. The coverage uniformity is defined based on the coefficient of variation of the coverage rate of each UAV's responsibility region as follows:

[0045]

[0046] In the formula, σ P (t) and μ P (t) represents all P i The standard deviation and mean of U(t) are used to monitor whether the load is balanced. The smaller the value of U(t), the more balanced the progress of each UAV mission is and the more reasonable the system load distribution is.

[0047] S5: The task area is adaptively divided using a weighted Voronoi diagram-based method to achieve adaptive load balancing for multiple UAVs. This includes the following sub-steps:

[0048] S5.1: For any point q within the mission area Ω, its responsible UAV i is determined according to the following rules:

[0049]

[0050] In the formula, q j (t) represents the planar position of the j-th UAV at time t, W j (t) represents its corresponding dynamic weight;

[0051] S5.2: Dynamic Weights W j (t) is constructed by combining factors such as drone coverage progress, regional area distribution, and obstacle density, and is defined as follows:

[0052]

[0053] In the formula, P j (t) represents the coverage of the current responsibility area of ​​UAV j, ξ1(t) is the global coverage adjustment factor, |Ω j (t)| represents the area of ​​its responsible region, |Ω avg |=|Ω| / N represents the average area of ​​the mission region, N is the total number of drones participating in the mission, and ξ2 is the region area adjustment factor. N represents the obstacle density within this region. obs ξ3 represents the number of obstacles in the area, and ξ3 is the obstacle density adjustment factor.

[0054] S5.3: To enable the region partitioning strategy to be dynamically adjusted according to the task phase, a global coverage adjustment factor is further introduced, which is defined as:

[0055]

[0056] In the formula, As a basic global coverage adjustment factor, this design enables the system to maintain the stability of region division in the early stages of the mission and enhance the ability to centrally compensate for under-covered areas in the later stages of the mission.

[0057] S6: After completing the dynamic division of responsibility areas for each UAV, in order to further guide the UAVs to perform efficient and balanced coverage search tasks within their respective responsibility areas, a target point selection step based on a utility function is performed within the responsibility area of ​​each UAV, which specifically includes the following sub-steps:

[0058] S6.1: The drone is at its current position p i Construct a region with radius r centered at (t). N The local neighborhood N(p) i (t)), this neighborhood constitutes the candidate space for the target point;

[0059] S6.2: After the candidate space for target points is constructed, construct the comprehensive utility function S for any potential location q within the responsibility area. i (q,t) is used to quantify the merits of this location as a true target point, and its utility function is defined as:

[0060]

[0061] In the formula, It is a distance efficiency weighting factor, adaptively adjusted based on global coverage, σ d It is a distance scale parameter used to control the distance decay rate. The missing weight factor is also adaptively adjusted based on global coverage, ω e It is the forgetting compensation weighting factor, t last (q) is the most recent access time, and τ is a time scale parameter that reflects the rate of forgetting.

[0062] S6.3: After completing the utility evaluation of candidate locations, the UAV selects the candidate location with the largest utility function value as the current search target point:

[0063]

[0064] Selected target point As a guiding input for the subsequent motion control stage;

[0065] S7: After determining the current target point of each UAV in step S6, a multi-constraint virtual force field continuous motion control method is adopted to convert the high-level target point decision into a continuous speed command for the UAV that satisfies kinematic and safety constraints. Specifically, it includes the following sub-steps:

[0066] S7.1: Target attraction is used to guide the UAV to move towards the target point determined in step S6, and is the main driving force for the UAV to complete the coverage task; for the i-th UAV, its target attraction is defined as:

[0067]

[0068] In the formula, k g As the target attraction gain coefficient, this force field forms a continuous guiding direction pointing towards the target point in space, enabling the UAV to approach the target position smoothly without external interference;

[0069] S7.2: Construct an obstacle avoidance force field for obstacles in the environment; when the drone approaches an obstacle, this force field rapidly strengthens to avoid a collision. The obstacle avoidance force is defined as:

[0070]

[0071] In the formula, d ik (t)=||p i (t)-o k ||-r k o k and r k Let d represent the center position and radius of the k-th obstacle, respectively. saf k is the safe distance threshold. oIt is the gain coefficient of obstacle avoidance force;

[0072] S7.3: Introduce a region barrier force in the control layer to flexibly constrain the movement range of the UAV. The region barrier force is defined as:

[0073]

[0074] In the formula, This indicates the minimum distance from the drone to the boundary of its area of ​​responsibility. For the corresponding nearest boundary point, d r This serves as a buffer distance for the region.

[0075] S7.4: To further guide drones to actively enter areas with insufficient coverage, an exploration force field based on coverage distribution is introduced; the exploration force reflects the spatial trend of coverage gaps in the area surrounding the drone, and is defined as:

[0076]

[0077] In the formula, N(p) i (t) represents the local neighborhood centered on the current location of the drone. This force field generates a resultant force in the direction of the insufficiently covered area, guiding the drone to move towards the area that has not yet been fully searched, thereby improving the overall coverage uniformity.

[0078] S7.5: After the above virtual force fields are constructed, they are linearly superimposed to obtain the total virtual force on the i-th UAV at time t:

[0079] F i (t)=F i attr (t)+F i obs (t)+F i region (t)+F i exp (t) (24)

[0080] Based on this total virtual force, the desired velocity command is generated through a continuous linear mapping:

[0081]

[0082] To meet the maximum speed constraint of the drone, the desired speed is continuously limited:

[0083]

[0084] By employing the aforementioned control methods, multiple UAVs can achieve safe, stable, and efficient collaborative coverage and search of a target area while satisfying the kinematic and safety constraints of the UAVs.

[0085] The present invention has the following beneficial effects:

[0086] 1. This invention employs a dynamic region division method based on a weighted Voronoi diagram to adjust the search area in real time according to the UAV's search progress, the area of ​​responsibility, and the distribution of obstacles. This avoids the task load imbalance problem that easily occurs during the search process in traditional fixed region division methods. By dynamically adjusting the area of ​​responsibility, each UAV can maintain a reasonable task allocation during the search process, which helps to reduce local area search lag and duplicate coverage, thereby improving the efficiency of multi-UAV collaborative coverage search.

[0087] 2. This invention introduces a coverage function and a forgetting mechanism in the target point selection process, and adaptively adjusts the utility weight in combination with the global coverage rate, so that the UAV can dynamically balance movement efficiency and coverage integrity in different search stages. In the later stages of the search, the system can actively guide the UAV to move to areas with insufficient coverage or long-term unvisited areas, thereby effectively reducing the generation of coverage holes and improving the integrity of search results and the uniformity of regional coverage.

[0088] 3. This invention adopts a continuous motion control method based on the superposition of multiple constraint virtual force fields. Through the combined effect of target attraction, obstacle avoidance force, regional barrier force and exploration force, the UAV can maintain a smooth and stable flight trajectory during the execution of search tasks, effectively avoiding collisions with obstacles and other UAVs, thereby improving the safety and reliability of the system in performing collaborative coverage search tasks in complex environments.

[0089] 4. The results show that, under the same area size, obstacle distribution, and number of UAVs, the comparative method requires approximately 223 seconds to complete the coverage search task, while the proposed method requires only 173 seconds to complete the same coverage requirement. The coverage search completion time is reduced by approximately 22.4%. This demonstrates that the proposed method can accelerate the overall progress of multi-UAV collaborative coverage search tasks and improve the system's search efficiency while ensuring flight safety and coverage quality. Attached Figure Description

[0090] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0091] Figure 1 A flowchart of a multi-UAV cooperative search method with dynamic region partitioning and multi-constraint virtual force fields;

[0092] Figure 2 A schematic diagram of the task area and obstacles;

[0093] Figure 3 A schematic diagram of a drone sensor perception model;

[0094] Figure 4 Diagram of a multi-UAV collaborative search simulation platform;

[0095] Figure 5 The search task completion status and region division map at different times;

[0096] Figure 6 A collaborative search track map for multiple drones;

[0097] Figure 7 To compare the completion status of the search task at different times;

[0098] Figure 8 To compare the multi-UAV cooperative search track maps of the algorithms;

[0099] Figure 9 This is a comparison chart of the search efficiency of our method and the comparison algorithm. Detailed Implementation

[0100] To make the technical solution of the present invention clearer and more complete, the present invention will be described in detail below with reference to the accompanying drawings and experimental examples. The overall flowchart is as follows. Figure 1 As shown, the main steps include:

[0101] S1: Construct a model covering the search task region, such as Figure 2 As shown, it specifically includes the following sub-steps:

[0102] S1.1: Define the UAV coverage search mission area as a two-dimensional closed region Ω = [0, L x ]×[0,L y In the embodiment, L x Take 100 meters, L y Take 100 meters;

[0103] S1.2: Set K static, impassable obstacles within the area. The obstacles are represented by cylinders, and the center point of the bottom surface of each obstacle is o. k Radius r k All heights are randomly assigned, with the height assumed to be infinite.

[0104] S1.3: Set coverage target: Within a finite task time T, ensure that the cumulative coverage of any point q within the region meets the threshold requirement.

[0105]

[0106] In the formula, C(q,T) represents the cumulative coverage of region point q, C tar It is a set threshold; only when C(q,T) is greater than or equal to C tar Only when the time is right is it determined that point q in the region has been scanned;

[0107] S2: Constructing the UAV kinematic model and setting constraints, which includes the following sub-steps:

[0108] S2.1: To characterize the motion state of the UAV, the kinematic model of the UAV is constructed as follows. Assume the system contains N UAVs, each numbered i = 1, 2, ..., N. The position and velocity of each UAV at time t are as follows:

[0109] p i (t)=[x i (t),y i (t),z i (t)] T (2)

[0110] v i (t)=[v i,x (t),v i,y (t),v i,z (t)] T (3)

[0111] In the formula, x i (t), y i (t), z i (t) represents the x-axis, y-axis, and z-axis coordinates of the UAV at time t, respectively. i,x (t), v i,y (t), v i,z (t) represents the x-axis velocity, y-axis velocity, and z-axis velocity of the UAV at time t;

[0112] S2.2: To ensure flight safety and feasibility, the following constraints are set:

[0113]

[0114] In the formula, a max It is the maximum acceleration of the drone, z min and z max These are the minimum and maximum altitudes of the drone during operation, v max (t) represents the maximum speed of the UAV during flight, and an adaptive adjustment strategy is adopted based on the full coverage H. global (t) Dynamically adjust the upper limit of speed, specifically as follows:

[0115]

[0116] In the formula, This is the base maximum speed, in the early stages of the search, when the global coverage H... global When (t) approaches 0, the drone uses a higher speed to quickly explore unknown areas; in the later stages of the search, when H... global When (t) approaches 1, the UAV reduces its flight speed to achieve fine coverage and improve perception quality. This adaptive speed adjustment mechanism enables the system to optimize the balance between flight efficiency and coverage accuracy at different mission stages.

[0117] S3: Construct a sensor perception model, such as Figure 3 As shown, the top-down visual sensor perception model onboard the UAV comprehensively considers both distance attenuation and altitude attenuation effects. Its perception range can be represented as a circular area that varies with altitude, with a radius of:

[0118] R i (t)=z i (t)tanα (6)

[0119] In the formula, α is the sensor's half-field of view angle. For any point q∈Ω on the ground, the instantaneous perception intensity of the i-th UAV is defined as:

[0120]

[0121] In the formula, Φ(ρ) is the distance attenuation function, which is defined as:

[0122]

[0123] The distance decay function describes the characteristic that the sensing intensity decreases twice with the increase of the normalized distance ρ, satisfying the boundary conditions Φ(0)=1 and Φ(1)=0, ensuring the physical characteristic that the sensing intensity is maximum directly below the sensor and zero at the sensing boundary; Ψ(z i ) is the height decay function, which is defined as:

[0124]

[0125] The altitude attenuation function quantifies how perceived quality changes with flight altitude; higher altitudes offer a wider field of view but lower resolution, while lower altitudes offer a narrower field of view but higher resolution. i (t) is the adaptive sensing gain coefficient, which is dynamically adjusted according to the current mission state of the UAV, and is defined as follows:

[0126] M i (t)=M base (1+η(1-P i (t))) (10)

[0127] In the formula, M base It is the basic sensing gain coefficient, Pi (t) is the coverage rate of the UAV's area of ​​responsibility, and η is the adjustment coefficient. When the coverage rate of the UAV's area of ​​responsibility is low, the system enhances its perception gain to accelerate the coverage process of that area; conversely, when the coverage rate is high, the perception gain is appropriately reduced to avoid resource waste caused by excessive search. This adaptive mechanism realizes closed-loop adjustment from task allocation to perception execution efficiency.

[0128] S4: The coverage metric defines a method for quantifying the cumulative sensed information of each point within the task area, which includes the following sub-steps;

[0129] S4.1: The cumulative coverage of region point q at time t is defined as:

[0130]

[0131] In the formula, λ>0 is the coverage forgetting factor, which makes the coverage of areas that have not been visited for a long time gradually decrease. This cumulative coverage model with forgetting mechanism simulates the aging process of environmental information, ensuring that the cumulative coverage of areas that have not been explored for a long time will naturally decrease over time, thus becoming a target that attracts drones again, effectively avoiding the algorithm from getting stuck in local optima.

[0132] S4.2: Overall system coverage progress is measured by global coverage rate H. global (t) quantization, which is defined as:

[0133]

[0134] The condition for determining the completion of the coverage search task is H. global (t)≥H tar H tar ∈(0,1] is a preset threshold for determining the completion of the search task;

[0135] S4.3: To evaluate the load balancing of a multi-UAV system, the coverage area of ​​responsibility for the i-th UAV is defined:

[0136]

[0137] In the formula, Ω i (t) represents the responsibility sub-region of the i-th UAV at time t. The coverage uniformity is defined based on the coefficient of variation of the coverage rate of each UAV's responsibility region as follows:

[0138]

[0139] In the formula, σ P (t) and μ P (t) represents all P iThe standard deviation and mean of U(t) are used to monitor whether the load is balanced. The smaller the value of U(t), the more balanced the progress of each UAV mission is and the more reasonable the system load distribution is.

[0140] S5: Dynamically divide the search area for each UAV, and use a weighted Voronoi diagram-based method to adaptively divide the task area to achieve adaptive load balancing for multiple UAVs. This includes the following sub-steps:

[0141] S5.1: For any point q within the mission area Ω, its responsible UAV i is determined according to the following rules:

[0142]

[0143] In the formula, q j (t) represents the planar position of the j-th UAV at time t, W j (t) represents its corresponding dynamic weight;

[0144] S5.2: Dynamic Weights W j (t) is constructed by combining factors such as drone coverage progress, regional area distribution, and obstacle density, and is defined as follows:

[0145]

[0146] In the formula, P j (t) represents the coverage of the current responsibility area of ​​UAV j, ξ1(t) is the global coverage adjustment factor, |Ω j (t)| represents the area of ​​its responsible region, |Ω avg |=|Ω| / N represents the average area of ​​the mission region, N is the total number of drones participating in the mission, and ξ2 is the region area adjustment factor. N represents the obstacle density within this region. obs ξ3 represents the number of obstacles in the area, and ξ3 is the obstacle density adjustment factor.

[0147] S5.3: To enable the region partitioning strategy to be dynamically adjusted according to the task phase, a global coverage adjustment factor is further introduced, which is defined as:

[0148]

[0149] In the formula, As a basic global coverage adjustment factor, this design enables the system to maintain the stability of region division in the early stages of the mission and enhance the ability to centrally compensate for under-covered areas in the later stages of the mission.

[0150] S6: After completing the dynamic division of responsibility areas for each UAV, in order to further guide the UAVs to perform efficient and balanced coverage search tasks within their respective responsibility areas, a target point selection step based on a utility function is performed within the responsibility area of ​​each UAV, which specifically includes the following sub-steps:

[0151] S6.1: The drone is at its current position p i Construct a region with radius r centered at (t). N The local neighborhood N(p) i (t)), this neighborhood constitutes the candidate space for the target point;

[0152] S6.2: After the candidate space for target points is constructed, construct the comprehensive utility function S for any potential location q within the responsibility area. i (q,t) is used to quantify the merits of this location as a true target point, and its utility function is defined as:

[0153]

[0154] In the formula, It is a distance efficiency weighting factor, adaptively adjusted based on global coverage, σ d It is a distance scale parameter used to control the distance decay rate. The missing weight factor is also adaptively adjusted based on global coverage, ω e It is the forgetting compensation weighting factor, t last (q) represents the most recent visit time, τ is a time scale parameter reflecting the rate of forgetting, and the first term of the utility function is the distance efficiency term, used to measure the distance from the drone's current position p. i (t) The cost of moving to candidate position q. The second term of the utility function is the coverage missing term, which is used to characterize the current coverage completion of the candidate position. The third term of the utility function is the forgetting compensation term, which is used to describe the time interval since the candidate position was last perceived. When a certain area has not been visited for a long time, this term increases to prevent the area from being ignored for a long time.

[0155] S6.3: After completing the utility evaluation of candidate locations, the UAV selects the candidate location with the largest utility function value as the current search target point:

[0156]

[0157] Selected target point As a guiding input for the subsequent motion control stage;

[0158] S7: After determining the current target point of each UAV in step S6, a multi-constraint virtual force field continuous motion control method is adopted to convert the high-level target point decision into a continuous speed command for the UAV that satisfies kinematic and safety constraints. Specifically, it includes the following sub-steps:

[0159] S7.1: Target attraction is used to guide the UAV to move towards the target point determined in step S6. It is the main driving force for the UAV to complete the coverage task. For the i-th UAV, its target attraction is defined as:

[0160]

[0161] In the formula, k g As the target attraction gain coefficient, this force field forms a continuous guiding direction pointing towards the target point in space, enabling the UAV to approach the target position smoothly without external interference;

[0162] S7.2: To ensure the flight safety of UAVs in complex environments, an obstacle avoidance force field is constructed for obstacles in the environment; when the UAV approaches an obstacle, this force field rapidly strengthens to avoid collision. The obstacle avoidance force is defined as:

[0163]

[0164] In the formula, d ik (t)=||p i (t)-o k ||-r k o k and r k Let d represent the center position and radius of the k-th obstacle, respectively. saf k is the safe distance threshold. o It is the gain coefficient of obstacle avoidance force. This method makes the UAV almost unaffected when it is far away from the obstacle, but produces a significant repulsive effect when it is close to a safe distance, thereby achieving continuous and stable obstacle avoidance behavior.

[0165] S7.3: To reduce the frequency of UAVs crossing the boundaries of their responsibility areas and avoid mutual interference, a regional barrier force is introduced at the control layer to flexibly constrain the movement range of the UAVs. The regional barrier force is defined as follows:

[0166]

[0167] In the formula, This indicates the minimum distance from the drone to the boundary of its area of ​​responsibility. For the corresponding nearest boundary point, d rAs a buffer zone, the force field gradually strengthens as the drone approaches the boundary of its area of ​​responsibility, ensuring the stability of the division of labor within the area while avoiding rigid restrictions on normal movement.

[0168] S7.4: To further guide drones to actively enter areas with insufficient coverage, an exploration force field based on coverage distribution is introduced; the exploration force reflects the spatial trend of coverage gaps in the area surrounding the drone, and is defined as:

[0169]

[0170] In the formula, N(p) i (t) represents the local neighborhood centered on the current location of the drone. This force field generates a resultant force in the direction of the insufficiently covered area, guiding the drone to move towards the area that has not yet been fully searched, thereby improving the overall coverage uniformity.

[0171] S7.5: After the above virtual force fields are constructed, they are linearly superimposed to obtain the total virtual force on the i-th UAV at time t:

[0172] F i (t)=F i attr (t)+F i obs (t)+F i region (t)+F i exp (t) (24)

[0173] Based on this total virtual force, the desired velocity command is generated through a continuous linear mapping:

[0174]

[0175] To meet the maximum speed constraint of the drone, the desired speed is continuously limited:

[0176]

[0177] By employing the aforementioned control methods, multiple UAVs can achieve safe, stable, and efficient collaborative coverage and search of a target area while satisfying the kinematic and safety constraints of the UAVs.

[0178] The method of this invention was verified through experiments. To enhance the engineering realism and visualization effect of the experimental verification, in this embodiment, a multi-UAV cooperative search simulation platform based on the combination of algorithm simulation and three-dimensional physical environment was constructed, such as... Figure 4As shown, the simulation platform includes an algorithm calculation module and a 3D simulation display module. The algorithm calculation module runs the multi-UAV cooperative search method proposed in this invention, calculating the search decision results and motion control commands of each UAV online based on a preset task area, obstacle distribution, and initial UAV state. The 3D simulation display module constructs a 3D simulation environment containing multiple UAV models and obstacles to visually demonstrate the motion process of UAVs performing cooperative search tasks in complex environments. The experimental scenario is a 100m × 100m closed area with eight obstacles of different sizes randomly placed within it. Under the same initial conditions, four UAVs are launched to perform a multi-UAV cooperative coverage search task to verify the applicability and effectiveness of the method in complex environments. Figure 5 These are the search task completion status and region division results at different times, among which... Figure 5 (a)~ Figure 5 (c) These represent the states at different points in the search process. Figure 5 (d) is the state at the end of the search mission. The yellow area in the figure represents the area that the drone has completed the coverage search, and the remaining areas represented by blue dots of different shades represent the areas that have not yet been covered. As can be seen from the figure, as the search mission progresses, the uncovered areas gradually decrease, and the overall coverage rate eventually reaches 99%, meeting the preset coverage requirements. At the same time, it can be seen from the figure that the division of the search area is not static and fixed, but dynamically adjusted according to factors such as the current search progress of each drone and the distribution of obstacles in the area of ​​responsibility. This dynamic area division mechanism can continuously balance the task load of each drone during the search process, which is conducive to improving the overall collaborative search efficiency. Figure 6 It is a multi-UAV collaborative search track map, in which Figure 6 (a)~ Figure 6 (c) shows the flight path at different times during the search process. Figure 6 (d) shows the flight path at the end of the mission. As can be seen from the figure, after the UAV took off from its initial position and entered the predetermined search altitude range, it began to perform the search mission. Its flight path did not come into contact with environmental obstacles throughout the entire process, and no collisions occurred between the UAVs. This indicates that the multi-constraint virtual force field continuous control method proposed in step S7 can effectively ensure the flight safety and motion continuity of the UAV in complex environments. To further verify the effectiveness and advancement of this invention, a comparative experiment was conducted with the method in the literature "A Multi-UAV Adaptive Cooperative Coverage Search Method Based on Regional Dynamic Perception." The comparative method mainly relies on the UAV's online perception information of the environment for cooperative search, without explicit division of responsibility for the task area. During the search process, it guides the UAV's movement through local information to complete the area coverage. Figure 7 It compares the search task completion status of the algorithm at different times, where Figure 7 (a)~ Figure 7 (c) These represent the states at different points in the search process. Figure 7 (d) shows the state at the end of the search task. As can be seen from the figure, the comparison algorithm did not perform sub-region division during the search process. The four drones jointly searched the entire search area until the search task was completed. Due to the lack of responsibility area constraints, there was a certain degree of duplicate searching by each drone during the search process, which affected the overall search efficiency. Figure 8 It is a comparison of multi-UAV cooperative search track maps of algorithms, in which Figure 8 (a)~ Figure 8 (d) shows the flight path at different times during the search process. As can be seen from the figure, at the end of the search mission, some UAV flight paths exhibited spatial clustering, indicating that there is still room for improvement in terms of cooperative avoidance and motion coordination. Figure 9 The results show the search efficiency comparison between the proposed method and the comparison algorithm. As the search task progresses, the coverage of the proposed method is consistently higher than that of the comparison algorithm. When the coverage reaches the preset threshold, the comparison algorithm takes 223 seconds, while the proposed method takes 173 seconds, a reduction of approximately 22.4% in time. Under the same experimental conditions, the proposed method demonstrates higher collaborative search efficiency. This shows that the proposed multi-UAV collaborative coverage search method based on dynamic region partitioning is effective and significantly improves the efficiency of multi-UAV collaborative coverage search tasks.

[0179] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the technical solutions of the present invention have been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the protection scope of the present invention.

Claims

1. A multi-UAV cooperative search method with dynamic region partitioning and multi-constraint virtual force fields, characterized in that, Includes the following steps: S1: Construct a model covering the search task region, which includes the following sub-steps: S1.1: Define the UAV coverage search mission area as a two-dimensional closed region Ω = [0, L x ]×[0,L y ]; S1.2: Set K static, impassable obstacles within the area. The obstacles are represented by cylinders, and the center point of the bottom surface of each obstacle is o. k Radius r k All heights are randomly assigned, with the height assumed to be infinite. S1.3: Set coverage target: Within a finite task time T, ensure that the cumulative coverage of any point q within the region meets the threshold requirement. In the formula, C(q,T) represents the cumulative coverage of region point q, C tar It is a set threshold; only when C(q,T) is greater than or equal to C tar Only when the time is right is it determined that point q in the region has been scanned; S2: Constructing the UAV kinematic model and setting constraints, which includes the following sub-steps: S2.1: To characterize the motion state of the UAV, the UAV kinematic model is constructed as follows. Assume the system contains N UAVs, each numbered i = 1, 2, ..., N, and the position p of each UAV at time t is... i (t) and velocity v i (t) are respectively: p i (t)=[x i (t),y i (t),z i (t)] T (2) v i (t)=[v i,x (t),v i,y (t),v i,z (t)] T (3) In the formula, x i (t), y i (t), z i (t) represents the x-axis, y-axis, and z-axis coordinates of the UAV at time t, respectively. i,x (t), v i,y (t), v i,z (t) represents the x-axis velocity, y-axis velocity, and z-axis velocity of the UAV at time t; S2.2: To ensure flight safety and feasibility, the following constraints are set: In the formula, a max It is the maximum acceleration of the drone, z min and z max These are the minimum and maximum altitudes of the drone during operation, v max (t) represents the maximum speed of the UAV during flight, and an adaptive adjustment strategy is adopted based on the full coverage H. global (t) Dynamically adjust the upper limit of speed, specifically as follows: In the formula, This is the base maximum speed, in the early stages of the search, when the global coverage H... global When (t) approaches 0, the drone uses a higher speed to quickly explore unknown areas; In the later stages of the search, when H global When (t) approaches 1, the drone reduces its flight speed to achieve fine coverage and improve perception quality; S3: Construct a sensor perception model. The visual sensor perception model on the UAV comprehensively considers the effects of distance attenuation and altitude attenuation. Its perception range can be represented as a circular area that varies with altitude, with a radius of: R i (t)=z i (t)tanα (6) In the formula, α is the sensor's half-field of view angle. For any point q∈Ω on the ground, the instantaneous perception intensity of the i-th UAV is defined as: In the formula, Φ(ρ) is the distance attenuation function, which is defined as: The distance decay function describes the characteristic that the sensing intensity decreases twice with the increase of the normalized distance ρ, satisfying the boundary conditions Φ(0)=1 and Φ(1)=0, ensuring the physical characteristic that the sensing intensity is maximum directly below the sensor and zero at the sensing boundary; Ψ(z i ) is the height decay function, which is defined as: The altitude attenuation function quantifies how perceived quality changes with flight altitude; higher altitudes offer a wider field of view but lower resolution, while lower altitudes offer a narrower field of view but higher resolution. i (t) is the adaptive sensing gain coefficient, which is dynamically adjusted according to the current mission state of the UAV, and is defined as follows: M i (t)=M base (1+η(1-P i (t))) (10) In the formula, M base It is the basic sensing gain coefficient, P i (t) represents the coverage rate of the drone's area of ​​responsibility, and η is the adjustment coefficient; S4: The coverage metric defines a method for quantifying the cumulative sensed information of each point within the task area, which includes the following sub-steps; S4.1: The cumulative coverage of region point q at time t is defined as: In the formula, λ>0 is the coverage forgetting factor. This cumulative coverage model with forgetting mechanism simulates the aging process of environmental information, ensuring that the cumulative coverage of long-undetected areas will naturally decrease over time, thus becoming a target that attracts drones again, effectively avoiding the algorithm from getting stuck in local optima. S4.2: Overall system coverage progress is measured by global coverage rate H. global (t) quantization, which is defined as: The condition for determining the completion of the coverage search task is H. global (t)≥H tar H tar ∈(0,1] is a preset threshold for determining the completion of the search task; S4.3: To evaluate the load balancing of a multi-UAV system, the coverage area of ​​responsibility for the i-th UAV is defined: In the formula, Ω i (t) represents the responsibility sub-region of the i-th UAV at time t. The coverage uniformity is defined based on the coefficient of variation of the coverage rate of each UAV's responsibility region as follows: In the formula, σ P (t) and μ P (t) represents all P i The standard deviation and mean of (t); S5: The task area is adaptively divided using a weighted Voronoi diagram-based method to achieve adaptive load balancing for multiple UAVs. This includes the following sub-steps: S5.1: For any point q within the mission area Ω, its responsible UAV i is determined according to the following rules: In the formula, q j (t) represents the planar position of the j-th UAV at time t, W j (t) represents its corresponding dynamic weight; S5.2: Dynamic Weights W j (t) is constructed by combining factors such as drone coverage progress, regional area distribution, and obstacle density, and is defined as follows: In the formula, P j (t) represents the coverage of the current responsibility area of ​​UAV j, ξ1(t) is the global coverage adjustment factor, |Ω j (t)| represents the area of ​​its responsible region, |Ω avg |=|Ω| / N represents the average area of ​​the mission region, N is the total number of drones participating in the mission, and ξ2 is the region area adjustment factor. N represents the obstacle density within this region. obs ξ3 represents the number of obstacles in the area, and ξ3 is the obstacle density adjustment factor. S5.3: To enable the region partitioning strategy to be dynamically adjusted according to the task phase, a global coverage adjustment factor is further introduced, which is defined as: In the formula, As a basic global coverage adjustment factor, this design enables the system to maintain the stability of region division in the early stages of the mission and enhance the ability to centrally compensate for under-covered areas in the later stages of the mission. S6: After completing the dynamic division of responsibility areas for each UAV, in order to further guide the UAVs to perform efficient and balanced coverage search tasks within their respective responsibility areas, a target point selection step based on a utility function is performed within the responsibility area of ​​each UAV, which specifically includes the following sub-steps: S6.1: The drone is at its current position p i Construct a region with radius r centered at (t). N The local neighborhood N(p) i (t)), this neighborhood constitutes the candidate space for the target point; S6.2: After the candidate space for target points is constructed, construct the comprehensive utility function S for any potential location q within the responsibility area. i (q,t) is used to quantify the merits of this location as a true target point, and its utility function is defined as: In the formula, It is a distance efficiency weighting factor, adaptively adjusted based on global coverage, σ d It is a distance scale parameter used to control the distance decay rate. The missing weight factor is also adaptively adjusted based on global coverage, ω e It is the forgetting compensation weighting factor, t last (q) is the most recent access time, and τ is a time scale parameter that reflects the rate of forgetting. S6.3: After completing the utility evaluation of candidate locations, the UAV selects the candidate location with the largest utility function value as the current search target point: Selected target point As a guiding input for the subsequent motion control stage; S7: After determining the current target point of each UAV in step S6, a multi-constraint virtual force field continuous motion control method is adopted to convert the high-level target point decision into a continuous speed command for the UAV that satisfies kinematic and safety constraints. Specifically, it includes the following sub-steps: S7.1: Target attraction is used to guide the UAV to move towards the target point determined in step S6, and is the main driving force for the UAV to complete the coverage task; for the i-th UAV, its target attraction is defined as: In the formula, k g As the target attraction gain coefficient, this force field forms a continuous guiding direction pointing towards the target point in space, enabling the UAV to approach the target position smoothly without external interference; S7.2: Construct an obstacle avoidance force field for obstacles in the environment; when the drone approaches an obstacle, this force field rapidly strengthens to avoid a collision. The obstacle avoidance force is defined as: In the formula, d ik (t)=||p i (t)-o k ||-r k o k and r k These represent the center position and radius of the k-th obstacle, respectively. d saf k is the safe distance threshold. o It is the gain coefficient of obstacle avoidance force; S7.3: Introduce a region barrier force in the control layer to flexibly constrain the movement range of the UAV. The region barrier force is defined as: In the formula, This indicates the minimum distance from the drone to the boundary of its area of ​​responsibility. For the corresponding nearest boundary point, d r This serves as a buffer distance for the region. S7.4: To further guide drones to actively enter areas with insufficient coverage, an exploration force field based on coverage distribution is introduced; the exploration force reflects the spatial trend of coverage gaps in the area surrounding the drone, and is defined as: In the formula, N(p) i (t) represents the local neighborhood centered on the current location of the drone. This force field generates a resultant force in the direction of the insufficiently covered area, guiding the drone to move towards the area that has not yet been fully searched, thereby improving the overall coverage uniformity. S7.5: After the above virtual force fields are constructed, they are linearly superimposed to obtain the total virtual force on the i-th UAV at time t: F i (t)=F i attr (t)+F i obs (t)+F i region (t)+F i exp (t) (24) Based on this total virtual force, the desired velocity command is generated through a continuous linear mapping: To meet the maximum speed constraint of the drone, the desired speed is continuously limited: By employing the aforementioned control methods, multiple UAVs can achieve safe, stable, and efficient collaborative coverage and search of a target area while satisfying the kinematic and safety constraints of the UAVs.

Images