Multi-unmanned aerial vehicle cooperative hunting method simulating Harris eagle group hunting

Abstract

The invention relates to a multi-unmanned aerial vehicle cooperative hunting method simulating Harris eagle group hunting. The multi-unmanned aerial vehicle cooperative hunting method comprises the implementation steps that 1, a hunting environment is initialized; 2, selecting an unmanned aerial vehicle leader and generating a surrounding formation; step 3, unmanned aerial vehicle speed synthesis control; 4, updating the expected surrounding distance; 5, attack unmanned aerial vehicle selection and formation reconstruction based on the optimal energy are carried out; and 6, outputting a hunting result. According to the method, aiming at the difficulty in capturing a dynamic target in an unknown environment, a cooperative control mechanism among multiple unmanned aerial vehicles is designed by referring to strategies such as energy balance scheduling, role division cooperative surrounding, relay strike and formation reconstruction embodied by Harris eagle in a hunting process. According to the method, efficient hunting and intelligent tracking of the dynamic target can be realized, and the cluster combat ability of the multi-unmanned aerial vehicle system in a complex environment is improved.

Description

Technical Field

[0001] This invention is a multi-drone collaborative encirclement and capture method that mimics the hunting of Harris eagle flocks, belonging to the field of drone autonomous control. Background Technology

[0002] Encirclement and control technology, as a key research direction in multi-agent systems, has attracted widespread attention in recent years in fields such as military, security, and disaster response. Firstly, existing methods mostly employ static or rule-based formation planning, lacking dynamic adaptability when facing high-speed maneuvering targets. Formation adjustments are often lagging, making effective encirclement difficult. Secondly, most systems do not clearly define the roles of individuals, leading to uneven task allocation and some drones exhausting their energy too quickly, affecting the overall sustainability of the encirclement and control. Based on this, this invention proposes an encirclement and control method that mimics the behavior of the Harris Eagle, aiming to overcome the limitations of existing technologies in terms of adaptability, energy balance, and role coordination.

[0003] During the hunt, the Harris eagles demonstrated exceptional teamwork. Through precise role division and efficient information exchange, they executed tactical encirclement and hunting maneuvers in the air. Some individuals acted as "drivers," forcing the prey to move in specific directions, while "interceptors" lay in ambush or flank from the sides, creating a closed-off hunting formation. This collective intelligence exhibited remarkable efficiency in performing complex tasks, providing valuable biomimetic references for swarm hunting mechanisms in multi-UAV systems.

[0004] During the hunt, the Harris eagles dynamically adjust their formation. Once prey is forced into the encirclement, the eagles take turns attacking. After the attacker completes their task, another eagle quickly fills the gap, ensuring the stability and continuity of the encirclement formation. This ability to reconfigure formation when switching between attack and encirclement modes is crucial for achieving sustained siege and suppression.

[0005] In highly uncertain aerial combat environments, mapping the coordinated hunting behavior of the Harris Hawk swarm to integrated reconnaissance-strike missions performed by multiple UAVs could significantly improve the system's performance in encircling and capturing dynamic targets. Multiple UAV swarms share information and coordinate missions through communication networks, making the encirclement and capture of highly maneuverable targets in complex environments more efficient and precise.

[0006] In this context, the goal of swarm-based collaborative encirclement technology is to confine the target within an encirclement (e.g., a convex hull) formed by multiple drones, achieving spatial blockade and dynamic suppression of the target. This technology holds significant application potential in fields such as military reconnaissance and target interception. Based on the hunting behavior of the Harris Eagle, we can summarize the optimization objectives of multi-drone encirclement systems into the following three aspects:

[0007] 1. Energy Balance Scheduling: Mimicking the mechanism by which eagles achieve energy balance through reasonable division of labor during hunting, a reasonable task allocation strategy is designed to ensure that the energy consumption of each drone remains balanced during the encirclement mission, thereby extending the system's combat time.

[0008] 2. Role-based coordinated encirclement: Referring to the role division of "driveers" and "interceptors" in a flock of eagles, guide drones to encircle the target from multiple directions. This can effectively compress the target's activity space and reduce the probability of the drones being detected and evaded by the target.

[0009] 3. Relay strikes and formation reconfiguration: Drawing on the strategy of eagle flocks rotating attackers during encirclement, the system enables role switching and dynamic formation adjustment between drones, ensuring the continuity and effectiveness of the encirclement mission over a long period of time.

[0010] When facing highly mobile targets, the Harris Hawks also rotate leaders. When the original leader is unable to continue guiding the encirclement due to limited visibility, other members will quickly take over, ensuring the continuity of the hunt. Introducing this mechanism into drone swarms, considering that modern communication systems can broadcast target information via data links, the change of leaders is no longer passively triggered by information gaps, but rather proactively serves the needs of formation reconfiguration and sustained strikes.

[0011] To fully verify the energy balance scheduling capability of this invention in encirclement-strike missions, detailed modeling of energy consumption in each behavioral stage is required. Energy consumption can be broadly divided into two categories: one is the speed consumption caused by maneuvering during the encirclement process, and the other is the strike consumption during the actual strike. By calculating the remaining energy of each encirclement drone in real time, the individual with the highest energy reserve is selected as the attacker, responsible for executing the high-energy-consuming strike mission after the encirclement is successfully completed. The remaining drones dynamically adjust their leader roles and reorganize their formations based on their remaining energy. After the strike mission is completed, the attack drones rejoin the formation, restore the encirclement formation, and continue to participate in subsequent encirclement operations. Summary of the Invention

[0012] 1. Purpose of the invention:

[0013] This invention aims to propose a multi-UAV cooperative encirclement method that mimics the hunting behavior of Harris eagles. Addressing the challenge of capturing dynamic targets in unknown environments, it draws upon the Harris eagle's hunting strategies, such as energy balancing, role-based cooperative encirclement, relay attacks, and formation reconfiguration, to design a cooperative control mechanism among multiple UAVs. This method enables efficient encirclement and intelligent tracking of dynamic targets, enhancing the swarm combat capabilities of multi-UAV systems in complex environments.

[0014] 2. Technical Solution:

[0015] This invention is a multi-drone collaborative encirclement and capture method mimicking the hunting of Harris eagle flocks, and its implementation process is as follows: Figure 2 As shown, the specific implementation steps are as follows:

[0016] Step 1: Initialize the enclosure environment

[0017] S11. Initialize the trapping environment and obstacle description matrix.

[0018] The simulation environment is a two-dimensional planar region with its origin located at the lower left corner. During initialization, its size is set to X×Y. In the encirclement mission, to ensure that the multi-UAV system can reasonably avoid obstacles and assess the feasibility of path planning during mission execution, it is necessary to model and structurally represent the obstacles in the environment. Therefore, an obstacle description matrix is ​​typically used in the initialization phase to uniformly record the geometric information and spatial position of all obstacles in the environment. All obstacles are simplified to circles; irregular obstacles can be formed by superimposing multiple circles. The obstacle description matrix is ​​denoted as:

[0019]

[0020] In this matrix, each row corresponds to an obstacle, and the three elements describe the obstacle's position and radius, respectively.

[0021] S12. Initialization of UAVs and Dynamic Targets

[0022] Assuming a drone swarm has N individuals, it can be represented as a set. Each individual must have a hierarchical label indicating its current identity as an attacker, leader, or ordinary individual. Each initialized individual must also have a sub-initial energy E. i Combining the individual's most basic state position p during movement i and speed v i The overall state structure of individual i can be represented by the following tuple:

[0023] agent i = <p i ,v i ,hier i E i > (2)

[0024] Similarly, combining the target's most basic states of position and velocity during motion, the overall state structure of target j can be represented by the following tuple:

[0025] target j = <p j ,v j diffj > (3)

[0026] In the initial stage of the group, all individuals are at the same level and are considered ordinary individuals. The initial position of the i-th drone is randomly set to (x... i ,y i Furthermore, during setup, it should be determined whether there is overlap; if the initial positions of the drones overlap, they should be regenerated. The target's position is randomly set to (x... j ,y j Considering the characteristics of dynamic targets, the maximum movement speed of dynamic targets is set to v. j The difficulty of the encirclement and capture is determined by the target's mobility as diff. j .

[0027] Step Two: Selection of Drone Leader and Generation of Encirclement Formation

[0028] S21, Selection of Drone Leaders

[0029] The encirclement and capture of a drone swarm can be abstracted as a formation encirclement of an escaping target point led by a lead drone. The drone swarm encirclement strategy is inspired by the cooperative hunting behavior of eagle flocks. In an eagle flock, the individual with the most advantageous position and closest to the prey's rear often assumes a leadership role, guiding the other members to encircle the target. Inspired by this, the lead drone should possess the following characteristics: always remain behind the target, i.e., in the opposite direction of the target's velocity vector; minimize frequent switching, as the individual already the lead drone has an advantage in assessment; and possess good communication capabilities, able to promptly broadcast the target's status to other drones.

[0030] Based on these three characteristics, the drones in the formation are evaluated, and the evaluation function is given:

[0031]

[0032] The evaluation function J i Advantages from the rear Leadership continuity items and communication centrality items These three parts constitute the equation. Here, w is the proportional gain of each term, boolean. leader (i) is a Boolean function used to determine whether the drone is the lead drone, while θ t For the target's heading, θ i Let n be the heading of the drones, and n be the total number of drones in the swarm.

[0033] The selection of the lead aircraft is based on the above allocation results. By default, the first aircraft in the ranking is selected as the leader. If the target's movement changes drastically or the current leader cannot maintain the ideal position, the program will automatically trigger the "leader switching mechanism", that is, select a new first aircraft in the ranking as the new lead aircraft. This mechanism effectively simulates the leadership rotation behavior in the eagle flock.

[0034] S22, Drone swarm encirclement formation generation

[0035] The lead drone senses the target's location and movement, maintaining a position behind it and sharing this information with other drones. Slave drones automatically generate their angle with the target based on their position in the formation, and determine their distance from the target by a distance r. aim The desired target point. When the target point changes direction at a low rate, the lead aircraft can track it stably, and from the target coordinate system, the encirclement formation remains a uniform circle and does not change. However, when the target point's position changes drastically, making tracking impossible, the lead aircraft's authority will be promptly transferred, which is consistent with the leadership rotation mechanism of the eagle flock described earlier.

[0036]

[0037] Where, θ leader k is the angle between the leader and the goal. i Let i be the relative position of the i-th drone and the lead drone. Here, we need to calculate the angle between each drone and the target and sort them. Based on the sorting result, we calculate the difference in sequence number between the drone and the lead drone, with clockwise as positive. N is the total number of drones.

[0038] Step 3: Drone speed synthesis control

[0039] S31. Calculation of Target Radial and Tangential Attractiveness

[0040] In swarm encirclement and control, velocity synthesis control is used. To achieve the effect of reaching the desired point without entering the target's reconnaissance range, a UAV body control coordinate system ∑ is established. b This makes its y-axis lie in the opposite direction of the line connecting the UAV and the target. Here, the gravitational force of the desired point on the UAV is a combination of radial and tangential attraction.

[0041] First, calculate the radial and tangential deviations between the drone and the desired point:

[0042] Δr=r aim -||P i -P target || (9)

[0043]

[0044] Δl=Δθ||P i -P target || (11)

[0045] Where Δr is the radial distance difference, Δl is the tangential distance difference, and r aim and θ aim P represents the expected radius and expected angle of the encirclement, respectively. i P represents the position of the i-th drone. target This represents the location of the target. To round down, the angle difference is kept within the range (-π, π). The gravity in the UAV coordinate system can be obtained from the calculated distance difference and then transferred to the world coordinate system Σ using a rotation matrix:

[0046]

[0047] Where R transforms the local coordinate system into the global coordinate system, F att To calculate the target attraction, a combination of tangential angle correction (first term) and radial distance correction (second term) is applied. The tangential attraction is square-rooted so that when the distance is far, the UAV prioritizes reducing the distance to the target point.

[0048] S32. Calculation of obstacle repulsion force and inter-machine repulsion force

[0049] Let R be the radius of the drone's perception of obstacles. obs Obstacles within this radius will exert a repulsive force on the drone. The direction of the repulsive force is from the obstacle towards the drone, and its magnitude decreases inversely with distance. For the i-th drone, the repulsive force from the k-th obstacle can be expressed as:

[0050]

[0051] Where, d ik The closest distance between the drone and the obstacle. η is the location of the obstacle edge closest to the drone. o This is the obstacle repulsion coefficient.

[0052] To avoid collisions between drones, an inter-drone repulsion mechanism is introduced. When the distance between two drones is less than a set safety radius R... safe At this time, a repulsive force will be generated, with the direction from the i'th neighboring UAV towards the main body, as shown in the following formula:

[0053] The total obstacle avoidance control force F is obtained by superimposing the repulsive forces of all neighboring drones and all obstacles. obs as follows:

[0054]

[0055] S33, Speed ​​Synthesis and Status Update

[0056] Regarding the interaction methods with interactive objects, the motion equations of the individual drone are first given to update the position and velocity of the individual drone:

[0057]

[0058] By calculating the original control force F orin For target attraction F att Neighbor repulsion force F rep and obstacle avoidance repulsion force F obs The combined force of the forces is then limited to obtain the final result F.

[0059] F orin =F att +F rep +F obs (18)

[0060]

[0061] Step 4: Update Expected Encirclement Distance

[0062] This paper uses DR and DE values ​​to describe the completion status of cluster encirclement and updates the expected encirclement distance in real time based on the encirclement status. DR represents the distance from the cluster center to the target location. Its calculation formula is as follows:

[0063]

[0064] DR (Rapid Approach) quantifies the overall proximity of the swarm to the target. A smaller DR value indicates that the swarm has approached the target and is a direct indicator of a successful encirclement; a larger DR value requires adjustments to the swarm's movement strategy to accelerate the approach. For example, in the early stages of an encirclement, DR is used to assess whether an overall acceleration in pursuit is necessary; during the encirclement phase, DR needs to approach 0 to achieve capture.

[0065] DE represents the variance of the distances from an individual to its nearest neighbor in a cluster, reflecting the uniformity of the distribution. Calculating the DE value requires first calculating the minimum distance from each individual i to its nearest neighbor, and then determining the variance of that minimum distance.

[0066]

[0067] DE = Var({d1,d2,...,d... N}) (twenty two) DE measures the tightness and uniformity of the encirclement. A DE value close to 0 indicates uniform spacing between individuals and a tight encirclement, preventing the target from escaping through gaps; a larger DE value indicates sparse or uneven distribution, requiring adjustments to the formation (such as reducing spacing or redistributing positions). In dynamic encirclement, a low DE is crucial for maintaining the encirclement posture, especially when the target is maneuvering and trying to escape, requiring real-time optimization.

[0068] Building upon the aforementioned encirclement control, the cluster dynamically adjusts the encirclement radius using the two parameters mentioned above to achieve consistent encirclement during the process. The encirclement radius gradually decreases from its maximum to its minimum as the parameter value shrinks to 0.

[0069] r aim =r max -(r max -r min )e aDR+βDE (twenty three)

[0070] Where, r aim α represents the expected capture radius, and β are constant coefficients.

[0071] Step 5: Selection and Formation Reconstruction of Energy-Optimal Attack Drones

[0072] S51, Individual Energy Consumption Calculation

[0073] In encirclement missions, drones need to continuously adjust their positions to maintain formation and encirclement structure, and the energy consumed in this process mainly comes from flight control. Assuming an ideal, undisturbed environment (i.e., without considering external factors such as wind speed and airflow disturbances), the energy consumption of a quadcopter drone during flight is determined by its hovering base power P. hover Speed-related power P speed Power P related to acceleration acc The system consists of several components. The hovering power is used to overcome gravity, allowing the aircraft to hover. The speed-related power primarily accounts for the work done by air resistance. The acceleration-related power is the power used to change the drone's speed during maneuvers. The specific power consumption can be expressed as follows:

[0074] P = P hover +P speed +P acc (twenty four)

[0075] The specific calculation methods for the power of each part are as follows:

[0076]

[0077] Where ρ is the air density, A represents the area of ​​a single propeller disk, η is the aerodynamic efficiency of the propeller, and N is the air density. p It is the number of propellers, CD The drag coefficient is related to the fuselage shape; a streamlined fuselage will have a lower drag coefficient. ref This is the reference cross-sectional area of ​​the aircraft, m is the weight of the drone, g represents the acceleration due to gravity, v is the velocity of the drone, and a is the acceleration of the drone.

[0078] Since this invention does not limit the specific strike method (e.g., ballistic, laser, jamming, etc.), the energy consumption of the strike behavior is simplified in the energy modeling. Considering that the maneuverability of different targets directly affects the complexity and duration of the strike, the strike energy consumption can be modeled as a function related to the target's maneuverability:

[0079] E strike =k s ||v target || (28)

[0080] Among them, v target The speed of the target represents its maneuverability, k s This is an empirical strike energy consumption coefficient, which can be adjusted according to the strike method and system capabilities. The more maneuverable the target, the more difficult it is to strike successfully, and the higher the energy required. Combining strike energy consumption and flight energy consumption, the remaining energy of each UAV is calculated as follows:

[0081]

[0082] S52, attack drone selection

[0083] Once the encirclement formation is formed, one of the drones will be selected to capture the target. In order to maintain the energy balance of the formation as much as possible, it is necessary to assign high-energy individuals in the encirclement formation to carry out the strike mission. Based on the individual energy consumption model given in S51, the remaining energy of the drones is calculated.

[0084] By sorting the remaining energy of each drone, the individual with the highest remaining energy is selected as the attacker. At this point, the drone leaves the group, reducing the number of drones to be captured, necessitating formation restructuring. When increasing or decreasing the number of drones in the formation, a leader election and reordering process needs to be performed on the group according to the method in S21 to redetermine the position of each drone.

[0085] Step Six: Output the capture results

[0086] After completing the encirclement mission, the system visualizes and performs performance evaluation analysis on the entire process to verify the effectiveness of the algorithm and the synergy of the encirclement strategy. First, the system dynamically displays the real-time movement trajectories of multiple drones and the movement paths of the target point during the encirclement process in animation, intuitively showing how the drones gradually form an encirclement and ultimately achieve target blockade. Furthermore, to assess the impact of the encirclement mission on the system's energy, the system records and outputs the energy consumption curves of each drone during the mission in real time. These curves allow observation of the energy consumption differences between individual drones and the balance of the overall scheduling strategy. To further aid understanding and verification, the system generates a series of accompanying visualizations.

[0087] This invention employs a multi-UAV collaborative encirclement method, mimicking the Harris Eagle hunting strategy, to enable multiple UAVs to rapidly locate, flexibly surround, and efficiently intercept targets in dynamic and unknown environments. This method not only improves the overall capture success rate of the system but also enhances the intelligence of the swarm in continuous target tracking, energy utilization, and ad-hoc coordination. It exhibits good robustness and adaptability, making it suitable for dynamic target control tasks in various scenarios such as border patrol, urban security, and wilderness search and rescue. Attached Figure Description

[0088] Figure 1 This is a demonstration image of a capture scenario.

[0089] Figure 2 A flowchart illustrating the multi-drone collaborative encirclement and capture process for a Harris eagle flock hunting model.

[0090] Figure 3 This is a diagram of cooperative encirclement and capture of static targets under unobstructed conditions.

[0091] Figure 4 This is a diagram of a coordinated encirclement under conditions with obstacles.

[0092] Figure 5 The energy decay curves for each drone are shown.

[0093] Figure 6 Reconstructing the encirclement formation. Detailed Implementation

[0094] The effectiveness of the multi-drone cooperative encirclement method for mimicking Harris eagle flock hunting proposed in this invention will be verified through specific examples below. According to... Figure 2 A simulation of the multi-drone coordinated encirclement method, mimicking the Harris eagle flock hunting approach, was conducted to verify the rationality and effectiveness of the proposed method. The simulation environment was configured with an Intel i7-12700 processor (3.40GHz), 16GB of RAM, and the following dependencies:

[0095] Operating System: Windows 10

[0096] Development environment: MATLAB R2023a

[0097] Programming language: MATLAB

[0098] The specific steps of this method are as follows:

[0099] Step 1: Initialize the enclosure environment

[0100] S11. Initialize the trapping environment and obstacle description matrix.

[0101] The simulation environment is a two-dimensional planar region, with the origin located at the lower left corner. During initialization, its size is set to 150×150. The obstacle description matrix is ​​initialized as follows:

[0102]

[0103] S12. Initialization of UAVs and Dynamic Targets

[0104] Assuming a drone swarm has 5 individuals, it can be represented as a set. All drones start as ordinary individuals, and each initialized individual must also have a sub-initial energy E. i =2e4 joules, and the initial position is randomly placed in the lower right corner of the map, with the initial velocity set to 0.

[0105] The target's location is randomly set to the center of the map. Considering the characteristics of dynamic targets, the maximum movement speed of dynamic targets is set to 2, and the difficulty of encirclement is determined to be 10 based on the target's mobility.

[0106] Step Two: Selection of Drone Leader and Generation of Encirclement Formation

[0107] S21, Selection of Drone Leaders

[0108] Based on the given evaluation function, calculate the leadership value of the 5 drones, and select the drone with the highest score, number 3, as the leader.

[0109] When the target's movement changes drastically and the current leader cannot maintain the ideal position, the program will automatically trigger the "leader switching mechanism," which selects a new leader as the new primary aircraft. This mechanism effectively simulates the leadership rotation behavior in a flock of eagles.

[0110] S22, Drone swarm encirclement formation generation

[0111] The lead drone senses the target's location and movement, maintaining a position behind the target and sharing its status information with other drones. Slave drones automatically generate their angles relative to the target based on their position in the formation; the current relative angle between the lead drone and the target is... At this point, if unit 1 is positioned to the left of the lead unit, then the desired angle needs to be subtracted from the lead unit's angle. Set as Step 3: Drone speed synthesis control

[0112] S31. Calculation of Target Radial and Tangential Attractiveness

[0113] In swarm encirclement and control, velocity synthesis control is used. To achieve the effect of reaching the desired point without entering the target's reconnaissance range, a UAV body control coordinate system ∑ is established. b This makes its y-axis lie in the opposite direction of the line connecting the UAV and the target. Here, the gravitational force of the desired point on the UAV is a combination of radial and tangential attraction.

[0114] First, calculate the radial deviation Δl between a drone and the desired point, which is 100 meters, and the tangential deviation Δr, which is... The gravity in the UAV coordinate system can be obtained from the calculated distance difference, and then transferred to the world coordinate system Σ using a rotation matrix:

[0115]

[0116] S32. Calculation of obstacle repulsion force and inter-machine repulsion force

[0117] The drone's obstacle detection radius is set to 20 meters. Obstacles within this radius will exert a repulsive force on the drone. The direction of the repulsive force is from the obstacle towards the drone, and its magnitude decreases inversely with distance. For the first drone, it experiences a repulsive force [-5,6] from the first obstacle, a repulsive force [-4,3] from the second obstacle, and a repulsive force [1,0.8] from its neighbor, drone number 2, resulting in a total obstacle avoidance control force F. obs = [-8, 9.8].

[0118] S33, Speed ​​Synthesis and Status Update

[0119] By calculating the original control force F orin The amplitude is then limited to obtain the final result F = [2.45, -1.24]. The individual's position and velocity are updated to V = [0.245, -0.124], and at this moment, the position moves by Δp = [0.0245, -0.0124].

[0120] Step 4: Update Expected Encirclement Distance

[0121] The DR and DE values ​​are calculated to describe the completion status of the swarm encirclement and update the expected encirclement distance. Based on the previous encirclement control, the swarm dynamically adjusts the encirclement radius using these two parameters to achieve consistent encirclement during the process. The encirclement radius gradually decreases from a maximum of 20 meters to a minimum of 10 meters as the parameter values ​​shrink to 0.

[0122] Step 5: Selection and Formation Reconstruction of Energy-Optimal Attack Drones

[0123] To maintain a balanced energy distribution within the formation, high-energy drones need to be assigned to strike missions. For example, after sorting, the third drone, with the highest remaining energy, is designated as the striker. The remaining four drones are then regrouped, with the formation interval angle decreasing from [previous angle]. Become The lead drone changed from No. 3 to No. 4. After the attack ended, the energy decreased by 200 joules, and No. 3 drone rejoined the formation, and the encirclement formation was restructured (as shown). Figure 6 (As shown).

[0124] Step Six: Output the capture results

[0125] The encirclement process is dynamically displayed, outputting real-time trajectory changes of the drone swarm and target point during the encirclement process, as well as the energy consumption curves of each drone during the encirclement process (e.g., ...). Figure 5 (As shown). Figure 1 The demonstration shows the overall encirclement scenario, including the initial formation, target location, and final encirclement result; Figure 3 and Figure 4 The static target capture process was demonstrated under both unobstructed and obstructed conditions, reflecting the system's adaptability to different environments. Figure 6 This demonstrates the formation reconfiguration mechanism during the encirclement process, highlighting the ability to dynamically adjust.

Claims

1. A multi-drone collaborative encirclement and capture method simulating the hunting of Harris eagle flocks, characterized in that: The specific implementation steps are as follows: Step 1: Initialize the encirclement environment, including: initializing the encirclement environment and obstacle description matrix, and initializing the UAV and dynamic targets; Step Two: Selection of Drone Leader and Generation of Encirclement Formation, including: The selection of the S21 drone leader abstracts the encirclement and capture of drone swarms into a form of encirclement of an escaping target point led by a primary drone. The primary drone, i.e. the leader, should meet the following characteristics: always stay behind the target, reduce frequent switching, and have good communication capabilities. The S22 encirclement formation is generated. The lead drone senses the target's location and movement, maintaining a position behind the target and sharing its status information with other drones. Slave drones automatically generate their angle with the target based on their position within the formation, and determine their distance from the target by a distance r. aim The expected point; when the target point changes direction at a low rate, the lead aircraft can track stably, and from the perspective of the target coordinate system, the encirclement formation remains a uniform circle and does not change; when the position of the target point changes drastically and makes tracking impossible, the lead aircraft's authority will be switched in a timely manner, which is the leadership rotation mechanism of the eagle flock. Step 3: UAV velocity synthesis control, including: S31, calculation of target radial and tangential attractive forces; S32, calculation of obstacle repulsive force and inter-UAV repulsive force; S33, velocity synthesis and state update; Step 4: Expected Encirclement Distance Update. Specifically, the DR and DE values ​​are used to describe the completion status of the cluster encirclement, and the expected encirclement distance is updated in real time based on the encirclement status. DR represents the distance from the center of the cluster to the target location, and DE represents the variance of the distance from the individual in the cluster to its nearest neighbor, reflecting the uniformity of distribution. Step 5: Selecting and reconstructing the energy-optimal attack drones, including: S51, Individual Energy Consumption Calculation Assuming an ideal, undisturbed environment, the energy consumption of a quadcopter drone during flight is determined by its hovering base power P. hover Speed-related power P speed Power P related to acceleration acc The components include: hovering base power, which overcomes gravity and allows the aircraft to hover; speed-related power, which is the work done by air resistance; and acceleration-related power, which is the power used to change the speed of the drone during maneuvers. The specific energy consumption power is expressed as follows: P = P0 hover +P speed +P acc ; S52, attack drone selection Once the encirclement formation is formed, one of the drones will be selected to capture the target. High-energy individuals in the encirclement formation need to be assigned to carry out the strike mission. The remaining energy of the drones is calculated based on the energy consumption model given in S51. By sorting the remaining energy of each drone, the individual with the highest remaining energy is selected as the attacker. At this point, the drone leaves the formation, reducing the number of drones to be captured, and formation reconstruction is required. When increasing or decreasing the number of drones in the formation, a leader election and reordering are required to redetermine the position of each drone. Step 6: Output the capture results.

2. The multi-drone collaborative encirclement and capture method for simulating Harris eagle flock hunting as described in claim 1, characterized in that: Based on the characteristics satisfied by the lead drone in step S21, the drones in the formation are evaluated, and an evaluation function is given: Evaluation function J i Advantages from the rear Leadership continuity items and communication centrality items These three parts constitute the equation; where w is the proportional gain of each term, bool leader (i) is a Boolean function used to determine whether the drone is the lead drone, while θ t For the target's heading, θ i Let n be the heading of the drones, and n be the total number of drones in the swarm. The selection of the leader is based on the above allocation results, with the first-ranked machine in the sorting being selected by default.

3. The multi-drone collaborative encirclement and capture method for simulating Harris eagle flock hunting as described in claim 1, characterized in that: According to the leadership rotation mechanism of the eagle flock described in step S22, it includes: Where, θ leader k is the angle between the leader and the goal. i Let i be the relative position of the i-th drone and the lead drone. Here, we need to calculate the angle between each drone and the target and sort them. Based on the sorting result, we calculate the difference in sequence number between the drone and the lead drone, with clockwise as positive. N is the total number of drones.

4. The multi-drone collaborative encirclement and capture method for simulating Harris eagle flock hunting as described in claim 1, characterized in that: The specific process of step S31 is as follows: Establish the UAV body control coordinate system ∑ b This makes the y-axis lie in the opposite direction of the line connecting the UAV and the target, and the gravitational force of the desired point on the UAV is a combination of radial and tangential attraction. First, calculate the radial and tangential deviations between the drone and the desired point: Δr=r aim -||P i -P target || Δl=Δθ||P i -P target || Where Δr is the radial distance difference, Δl is the tangential distance difference, and r aim and θ aim P represents the expected radius and expected angle of the encirclement, respectively. i P represents the position of the i-th drone. target This represents the location of the target; To round down, the angle difference is kept within the range of (-π, π). The gravitational force in the UAV coordinate system is obtained from the calculated distance difference and then transferred to the world coordinate system Σ using a rotation matrix. Where R transforms the local coordinate system into the global coordinate system, F att To calculate the target attraction, a combination of tangential angle correction and radial distance correction is applied. The tangential attraction is square-rooted so that when the distance is far, the UAV prioritizes reducing the distance to the target point.

5. The multi-drone collaborative encirclement and capture method for simulating Harris eagle flock hunting as described in claim 1, characterized in that: Step S32, calculation of obstacle repulsion force and inter-machine repulsion force; Let R be the radius of the drone's perception of obstacles. obs Within this radius, obstacles will generate a repulsive force; the direction of the repulsive force is from the obstacle towards the drone, and its magnitude decreases inversely with the distance; for the i-th drone, the repulsive force from the k-th obstacle is expressed as: Where, d ik The closest distance between the drone and the obstacle. η is the location of the obstacle edge closest to the drone. o The obstacle repulsion coefficient; When the distance between the two drones is less than the set safety radius R safe At this time, a repulsive force will be generated, with the direction from the i'th neighboring UAV to the main body, as shown in the following formula: The total obstacle avoidance control force F is obtained by superimposing the repulsive forces of all neighboring drones and all obstacles. obs as follows:

6. The multi-drone collaborative encirclement and capture method for simulating Harris eagle flock hunting as described in claim 1, characterized in that: The specific process of step S33, velocity synthesis and state update, is as follows: Regarding the interaction methods with interactive objects, the motion equations of the individual drone are first given to update the position and velocity of the individual drone: By calculating the original control force F orin For target attraction F att Neighbor repulsion force F rep and obstacle avoidance repulsion force F obs The combined force, after being limited, yields the final result F; F orin =F att +F rep +F obs 7. A multi-drone cooperative encirclement and capture method for mimicking Harris eagle flock hunting as described in claim 1, characterized in that: The DR value mentioned in step four represents the distance from the cluster center to the target location, and the calculation formula is as follows: DR quantifies the overall proximity of the cluster to the target; The DE mentioned in step four represents the variance of the distance from an individual in the cluster to its nearest neighbor, reflecting the uniformity of the distribution; Calculating the DE value requires first calculating the minimum distance from each individual i to its nearest neighbor, and then finding the variance of the minimum distance: DE=Var({d1,d2,...,d N }) DE measures the tightness and uniformity of the enclosing area.

8. A multi-drone collaborative encirclement and capture method for simulating Harris eagle flock hunting as described in claim 1, characterized in that: Step S51 further includes: The energy consumption of an attack is modeled as a function related to the target's mobility: E strike =k s ||v target || Among them, v target The speed of the target, representing maneuverability, k s This is an empirical strike energy consumption coefficient, adjusted according to the strike method and system capabilities; the more maneuverable the target, the more difficult it is to strike successfully, and the higher the energy required; combining strike energy consumption and flight energy consumption, the remaining energy of each UAV is calculated as follows:

Images