A method for deploying a UAV formation for a may beetle cooperative defense task

Abstract

The application discloses a kind of unmanned aerial vehicle formation deployment methods for honey beetle cooperative defense task, belong to unmanned aerial vehicle formation cooperative control technical field, including fusion multi-source sensing information, threat assessment model is established, constructs the interception probability model based on Monte Carlo simulation, with the goal of maximizing interception benefit, establish mixed integer nonlinear programming model, and adopt improved genetic algorithm solution, obtain optimal unmanned aerial vehicle formation deployment scheme;The application adopts the unmanned aerial vehicle formation deployment method for honey beetle cooperative defense task described above, through fusion multi-objective optimization, antagonistic constraint and Monte Carlo damage simulation, effectively improve the accuracy and actual combat reliability of unmanned aerial vehicle formation defense deployment.

Description

Technical Field

[0001] This invention relates to the field of drone formation cooperative control technology, and in particular to a drone formation deployment method for drone cooperative defense missions. Background Technology

[0002] In recent years, unmanned aerial vehicles (UAVs), especially loitering UAVs with long-duration cruise capabilities, have posed a severe challenge to the protection of important ground platforms such as armored vehicles. Traditional protection methods for armored vehicles mainly rely on physical camouflage, maneuver evasion, or close-range interception. These methods generally suffer from passive response and insufficient interception accuracy, making them ineffective against low-cost, high-precision UAV threats. While equipping armored vehicles with dedicated air defense systems or specialized defense units can improve interception capabilities, the high cost and system complexity result in a poor overall cost-effectiveness ratio. Therefore, there is an urgent need to equip armored vehicles with a new, low-cost, precise, and efficient air defense method to enhance their survivability in asymmetric threat environments.

[0003] The development of drone swarm technology has provided a new technological path for collaborative defense. Through the coordinated networking and information exchange between armored vehicles and drone swarms, an integrated perception and interception system can be formed, enhancing overall situational awareness and collaborative response capabilities. Against this backdrop, how to scientifically and rationally deploy drone swarms to maximize their interception effectiveness in collaborative defense missions has become a critical technical issue that urgently needs to be addressed.

[0004] Existing research on drone deployment largely focuses on civilian applications, such as communication relay in urban environments, disaster search and rescue, logistics delivery, and agricultural plant protection. Deployment methods in these scenarios typically optimize static indicators such as communication quality, coverage efficiency, and energy economy. Their constraints are also primarily focused on non-adversarial factors like the physical environment, battery life, and communication links, making them difficult to directly apply to collaborative defense tasks with highly dynamic and adversarial characteristics. Specifically, existing technologies have the following limitations: (1) The optimization target is singular and does not consider the dynamic coupling relationship between interception value, threat assessment and overall defense effectiveness; (2) The constraints lack adversarial modeling and do not cover survivability requirements such as accidental injury avoidance and maintaining a safe distance; Therefore, existing drone deployment methods cannot meet the needs of refined and adversarial planning for cluster deployment locations in collaborative defense missions, especially when dealing with multiple waves and multiple directions of incoming targets, lacking a cluster deployment scheme aimed at maximizing overall interception benefits. Summary of the Invention

[0005] The purpose of this invention is to provide a method for deploying UAV swarms for collaborative defense missions, in order to fill the gap in research on swarm deployment for collaborative defense purposes, and to empower the air defense capabilities of armored vehicles through the new equipment concept of the "swarm-armor integrated" system.

[0006] To achieve the above objectives, this invention provides a method for deploying drone swarms for collaborative defense missions involving drones, comprising the following steps: S1. Obtain multi-source perception information through the network communication module, fuse radar perception information and airborne camera perception information, construct a threat assessment model of incoming targets, and quantify the value of intercepting the target based on the distance, size and confidence level of perception and identification of the incoming target. S2. The interception process of an abstract UAV swarm formation meeting an incoming target is constructed based on the Monte Carlo simulation method, combined with the relative positional relationship between the UAV and the incoming target and the characteristics of the UAV's warhead, to quantify the interception probability of the UAV swarm against the incoming target. S3. Using the product of the value of the intercepted target and the interception probability as the interception benefit, and taking maximizing the overall interception benefit as the optimization objective, a mixed integer nonlinear programming model is constructed by selecting the interception allocation relationship matrix between the UAV and the incoming target and the UAV position set as decision variables. S4. Design an improved genetic algorithm to solve the mixed-integer nonlinear programming model. By improving the genetic operators and adjusting the Monte Carlo simulation parameters, obtain the drone formation deployment scheme with the optimal interception benefit.

[0007] Preferably, the specific steps of S1 are as follows: S11. Obtain radar perception information of incoming targets through the radar of armored vehicles, treat individual incoming targets as flat targets, and estimate the radar cross-section. For the first Group attack cluster target Calculate the total radar cross-section The radar echo power received by the radar cluster is calculated using radar equations. The relationship between radar echo power and cluster size and target distance is obtained to assess distance and scale information. The calculation formula is as follows: (1) (2) (3) (4) in, The area of ​​the flat plate for the target individual. For radar wavelength, The number of target individuals within the cluster. For transmission power, For the antenna gain of a single-station radar, The distance between the radar and the incoming cluster of targets. The location of the cluster target, The size of the target cluster; S12. Obtain visual perception information of the incoming target through the drone's onboard camera, and determine the pixel size of the target drone in the image. and recognition accuracy Calculate the positive detection rate To determine the target type, the calculation process is as follows: (5) (6) (7) (8) (9) in, For camera focal length, For the target drone size, The physical size of a single pixel in the sensor. The distance between the drone and the incoming swarm target. The smallest pixel size that the camera can recognize. The attenuation coefficient is... As an occlusion factor, For the imaging size of the cluster target, The image size of an individual within the target cluster. Sensitivity coefficient This indicates whether the cluster target is a loitering munition cluster. A value of 1 indicates that the cluster target is a loitering munition cluster, while a value of 0 indicates that the cluster target is a non-enemy target. For the first The distance between the cluster target and the drone equipped with an airborne camera; S13. Define the threat level of an incoming target to the protected platform based on radar echo power and positive detection rate. The calculation formula is as follows: (10) in, For armored vehicles.

[0008] Preferably, the recognition accuracy in S12 is... Modeled as a Sigmoid function, with a decay coefficient The value range is 5~10, and the sensitivity coefficient is... The value range is 0.1 to 0.5.

[0009] Preferably, the specific steps of S2 are as follows: S21. For interception of a single target by a single UAV, calculate the damage probability based on the number and distribution of fragment hits within the effective kill radius; S22. For the interception of a cluster of drones by a formation, the fragmentation process after the formation detonates is simulated in a virtual coordinate system. The probability of damage to the entire cluster of targets is taken as the overall damage probability, and the probability of the cluster of targets being intercepted is calculated based on the damage probability.

[0010] Preferably, the specific steps of S21 are as follows: S211. Let the position of the individual rotary-wing UAV be... The location of the single incoming target is The distance between the two ; S212. Set the effective kill radius of the drone. For targets outside this range, the probability of damage is determined to be zero; S213, For targets within the range Set a baseline damage probability for a single fragment, and assume that the fragment hits follow a Poisson distribution and are uniformly distributed within the target area. The calculation formula is as follows: (11) S214. Establish a spherical coordinate system centered on the UAV to simulate the azimuth angle of all fragments scattered during a single test. and polar angle Evenly distributed on the surface, any fragment The direction vector is ; S215, the incoming target is equivalent to a radius of... The sphere, with its direction vector relative to the individual rotor is By judging the fragments Is the angle between the direction of the scattering and the direction relative to the incoming target less than its equivalent half-angle? To determine whether the fragment hit the target, that is... Consider fragments Hit; S216. Through numerous independent Monte Carlo experiments, the hit results are statistically analyzed, and the final probability of damage to the target is calculated by approximating the frequency of hits. The calculation formula is as follows: (12) in, For the first Number of hits in the Monte Carlo test The probability of damage to the incoming target. The number of Monte Carlo trials.

[0011] Preferably, the specific steps of S22 are as follows: S221. Establish a virtual coordinate system that coincides with the coordinate system of the protected armored vehicle, and translate the cluster targets while maintaining the relative relationships between the nodes to a position where the center of the cluster is a distance from the centroid of the armored vehicle. The distance; S222. In the virtual combat situation formed by the virtual coordinate system, the Monte Carlo fragment hit simulation process of one-to-one interception is reused to simulate the fragment scattering after all UAVs in the formation are detonated at the same time. S223. In each Monte Carlo test, count the total number of times each individual target in the attacking swarm is hit by fragments from all drones, and calculate its individual damage probability accordingly. The calculation process is as follows: individual rotors Hit cluster individuals Number of fragments for: (13) Cluster Individual Number of fragments hit for: (14) in, For rotor formation; Cluster Individual The probability of being hit and causing damage The calculation formula is: (15) S224. The criterion for successful interception of the cluster in this experiment is that all individual targets within the cluster are destroyed. Finally, based on the statistical results of multiple experiments, the formation's ability to intercept the cluster target is calculated. Overall interception probability The calculation formula is as follows: (16) in, The size of the cluster target.

[0012] Preferably, the specific steps of S3 are as follows: S31. Define decision variables. Decision variables include a set of variables composed of... Rotor cluster Intercept Group cluster target The distribution relationship constitutes matrix and hive location ; (17) Among them, elements Indicates the first Are the rotors deployed to intercept the first Group cluster objectives; S32. Set constraints, including allocation constraints and position constraints. The allocation constraints ensure that each UAV is assigned to only one cluster target, and each cluster target is assigned at least one UAV. The position constraints ensure that all UAVs are deployed within the effective kill radius of the friendly warhead and meet the safe distance requirements to avoid friendly fire. S33. Establish optimization goals and define the "Honeycomb Armor Integration" system's response to all incoming attacks. The interception benefits of grouped targets are: (18) in, As for the level of threat, This represents the probability of interception. S34. Define key parameters, including whether the cluster target is an enemy unit. And fragment hit simulation Monte Carlo test number of times ; S35. Establish a formation planning and deployment model for UAVs, and optimize the allocation matrix. and hive location set This maximizes the total interception revenue.

[0013] Preferably, the specific formulas for the position constraints and assignment constraints in S32 are as follows: Assignment constraints: (19) (20) (twenty one) (twenty two) Position constraints: (twenty three) (twenty four) in, The effective kill radius of the fragmentation warhead. To determine the location and distance of the target in the intercept cluster.

[0014] Preferably, the specific formula for establishing the model in S35 is as follows: (25) (26) in, This indicates a constraint condition.

[0015] Preferably, the specific steps of S4 are as follows: S41. Define the total interception reward of the enemy's incoming target as the fitness function of the genetic algorithm, design the encoding, and assign the solution variable matrix. Encoded in binary order according to row and column sequence An array of 0s and 1s of length, representing the location of the bee colony. Position them according to rotor number Coordinate values ​​are encoded as real numbers. An array of real numbers; (27) (28) S42. Design a selection operator and adopt a roulette wheel selection strategy. The selection probability is allocated proportionally according to the individual fitness value to select parent individuals from the current population. S43. Design a crossover operator. For the allocation matrix encoding, point crossover is used, and the position of the crossover point must meet the allocation constraint. For the position encoding, uniform crossover is used, and the corresponding coordinate values ​​are exchanged according to a set probability. S44. Design a mutation operator. Use exchange mutation for the allocation matrix encoding, and its mutation interval must satisfy the allocation constraint. Use real number mutation for the position encoding, and randomly generate new coordinates in the feasible region according to a set probability. S45. During the algorithm iteration process, the number of Monte Carlo simulation trials is dynamically adjusted according to the current generation. A larger number of trials is used in the early stage of evolution to ensure evaluation accuracy, and a smaller number of trials is used in the later stage of evolution to accelerate convergence. When the algorithm meets the termination condition, the allocation matrix and position set corresponding to the individual with the highest fitness are output as the optimal UAV formation deployment scheme.

[0016] Therefore, the above-mentioned method for deploying drone swarms for collaborative defense missions using bee-like drones has the following beneficial effects: (1) Taking into account multiple optimization objectives: The optimization objectives of the formation deployment planning model integrate multi-source perception results of enemy targets, threat assessment indicators, and quantitative indicators related to air defense, such as the predicted interception probability of friendly swarms. (2) The constraints are adversarial: The constraints of the defensive deployment problem are highly adversarial and survival-oriented. For example, the deployment of interception positions must ensure that the armored vehicles are outside the effective damage range of the enemy and avoid friendly fragmentation. (3) Considering the damage process: Monte Carlo simulation was performed on the damage process of the UAV detonation to intercept the incoming target, and the fragment hit prediction number was obtained, which was used to model the interception probability of the swarm against the enemy cluster target.

[0017] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0018] Figure 1 This is a flowchart of a drone swarm deployment method for a bee-armor cooperative defense mission according to the present invention; Figure 2 This is a schematic diagram illustrating the descriptive system for the swarm deployment problem under the vehicle system of the present invention, which is a method for deploying drone swarms for collaborative defense missions. Figure 3 This is a schematic diagram illustrating the single-to-single fragment hit determination of a drone swarm deployment method for a bee-armor cooperative defense mission according to the present invention. Figure 4 This is a schematic diagram of the interception and destruction process under virtual system consideration when missiles and targets meet, which is a method for deploying UAV formations for collaborative defense missions with bee-armor drones according to the present invention. Figure 5 This is a schematic diagram of the genetic algorithm flow for a drone formation deployment method for a bee-armor cooperative defense mission according to the present invention. Detailed Implementation

[0019] The following detailed description of embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0020] Example like Figure 1 As shown, this invention provides a method for deploying UAV swarms for collaborative defense missions involving armored vehicles and drones. The "armored vehicle integrated" system includes armored vehicles, a drone swarm, and a network communication module. The armored vehicles are equipped with radar for sensing incoming targets. The drone swarm consists of multiple rotor-wing UAVs, each equipped with an airborne camera for auxiliary sensing. In this scenario, the system uses the state truth values ​​of the incoming swarm targets to simulate the localization and sensing results after multi-source fusion. A threat assessment model is constructed based on the positional relationship relative to the armored vehicles. An interception prediction model for the swarm is constructed by combining the damage characteristics analysis of individual rotor-wing UAVs. The threat assessment model and the interception prediction model provide the value and probability of intercepting incoming targets, thereby constructing a swarm deployment planning model. This model is used as the fitness function of an improved genetic algorithm. Genetic operators are designed for the deployment planning process to solve for the optimal deployment scheme.

[0021] Establish a descriptive framework for the "Hybrid Bee" system's defense against incoming enemy swarm targets. On the ground, the "Hybrid Bee" system can obtain the position and orientation angle of armored vehicles and the position and attitude of rotary-wing UAVs via communication networks. It can also simulate the perception results using the position of enemy loitering munitions. For example... Figure 2 As shown, the above information is used to describe the swarm deployment problem in the context of the vehicle system.

[0022] The "Bee and Armor Integrated" system includes Rotary-wing drones attacked from different directions. Group cluster objectives.

[0023] (29) (30) make Rotor cluster Intercept Group cluster target The allocation relationship is a matrix ,element It is a Boolean value, representing the first... Are the rotors deployed to intercept the first Group cluster objectives.

[0024] Intercept The rotor formation is Its number of rotors is .

[0025] (31) The specific steps are as follows: S1. Obtain multi-source perception information through the network communication module, fuse the perception information from radar and airborne cameras, construct a threat assessment model for incoming targets, and quantify the value of intercepting the target based on the distance, size, and confidence level of perception and identification of the incoming target.

[0026] For a certain cluster target Use its state truth value (position) and scale This was used to simulate the perception results of the armored vehicle's radar and rotor camera, and to quantify the threat level of this swarm of targets to the "swarm armored vehicle integrated" system.

[0027] S11. Radar Sensing Simulation: This simulation uses the radar of armored vehicles to acquire radar sensing information of incoming targets, treating individual incoming targets as flat targets with radar cross-section (RCS). It can be roughly estimated as follows: (1) in, The area of ​​the flat plate for the target individual. The radar wavelength; According to the quantitative description of the radar equation, the power of the cluster echo received by the radar can reflect the relationship between the radar detection result and the range and the number of clusters. Assuming this is an incoherent superposition process, the average RCS of the individual target is... The total RCS of the cluster The calculation can be simplified as follows: (2) Calculate the radar echo power received by the radar cluster using radar equations. The relationship between radar echo power and cluster size and target distance is obtained to assess distance and scale information. The calculation formula is as follows: (3) (4) in, The number of target individuals within the cluster. For transmission power, For the antenna gain of a single-station radar, The distance between the radar and the incoming cluster of targets. The location of the cluster target, The size of the target cluster; The above formula can be used to simulate the radar echo power detected by the armored vehicle radar based on the true state values ​​of different incoming cluster targets. This quantitative result shows differences at different distances from the armored vehicle and varies depending on the size of the cluster targets.

[0028] S12, Visual Perception Simulation: Visual perception information of the incoming target is acquired using the drone's onboard camera. The relationship between the drone's pixel size and distance in the image is obtained as follows: (5) in, For camera focal length, For the target drone size, The physical size of a single pixel in the sensor. The distance between the drone and the incoming swarm of targets; There exists a , If it cannot be identified at the moment, it can be deduced that for a single node of the cluster target, there is a maximum distance that satisfies image recognition: .

[0029] This distance serves as the boundary for whether the rotor camera provides positive detection rate information for enemy aircraft. When the swarm target is farther than this value, the positive detection rate is zero. Beyond this value, the rotor provides a positive detection rate value to the "Honeycomb Armor Integrated" system by detecting camera images, which describes how likely the swarm target is an enemy swarm target.

[0030] An empirical model based on target pixel size shows that recognition accuracy decreases as pixel size decreases, which can be modeled as a sigmoid function. This is the attenuation coefficient.

[0031] (6) When the distance between drones is less than the imaging resolution, target overlap leads to a decrease in recognition rate. An occlusion factor is defined. , For the number of drones, and For the imaging size of clusters and individuals, The sensitivity coefficient is denoted as .

[0032] (7) Overall accuracy, combining the two formulas above, is related to the target distance and the number of clusters.

[0033] (8) Use symbols There are only two states to indicate whether the target of the cluster is a loitering munition cluster. .when When the target of a cluster is a cluster of loitering munitions, and identifying it as a cluster of loitering munitions is a correct identification result, then the positive detection rate is the target identification accuracy rate. ;when When the cluster target is not an enemy target, identifying it as a loitering munition cluster would be a false positive. Therefore, the higher the accuracy of cluster target identification, the lower the positive detection rate of identifying it as a loitering munition cluster. .

[0034] In summary, the actual type of the cluster target is introduced as a new variable. Based on the above definition of recognition accuracy, and substituting it into the cluster target... State truth value, positive detection rate It can be represented as: (9) in, For the first The distance between the cluster target and the drone equipped with an airborne camera.

[0035] The above formula can be used to simulate the positive detection rate of different incoming cluster targets as perceived by the rotor camera based on the true state values. This quantitative result shows differences at different distances from the armored vehicle and varies depending on the size of the cluster targets.

[0036] S13. Define the threat level of an incoming target to the protected platform based on radar echo power and positive detection rate. The calculation formula is as follows: (10) in, For armored vehicles.

[0037] The closer the target group is and the larger its number, the greater the radar echo power. If the target is indeed an enemy aircraft, the positive detection rate will also increase, thus indicating that the threat level of the target group to armored vehicles has increased.

[0038] S2. The interception process of an abstract UAV swarm formation meeting an incoming target is constructed based on the Monte Carlo simulation method, combining the relative positional relationship between the UAVs and the incoming target and the characteristics of the UAV warhead, to quantify the interception probability of the UAV swarm against the incoming target.

[0039] S21. For interception of a single target by a single UAV, calculate the damage probability based on the number and distribution of fragment hits within the effective kill radius; S211. Let the position of the individual rotary-wing UAV be... The location of the single incoming target is The distance between the two ; S212. Effective kill radius, centered on the geometric center of the individual rotor. Even if an enemy unit outside the range is hit, it will not be damaged due to insufficient kinetic energy of the fragments; the probability of damage is 0. S213. When an enemy individual is within range, the damage probability of a single fragment hit is represented by the ratio of the projected area of ​​the vulnerable part to the overall projected area. The attitude of the enemy individual is not considered, and this value is assumed to be a constant. Assuming the fragment hit probability follows a Poisson distribution, the events of fragments hitting the target are independent, and the fragments are uniformly distributed in the hit area, then the damage probability... The calculation formula is as follows: (11) S214, Number of fragments hit by rotor detonation Simulate using the Monte Carlo method, and set the maximum number of trials for the Monte Carlo method. Based on the configuration of rotary-wing UAVs carrying omnidirectional fragmentation warheads with explosive charges similar to the Type 82 hand grenade, the effective kill radius of the rotor... Rice, number of fragments Establish a spherical coordinate system with the detonation point as the origin, with the polar axis aligned with the positive x-axis. The direction of fragment dispersion will be determined by the azimuth angle. and polar angle Evenly distributed on top ; Any fragment The direction vector is ; S215, the incoming target is equivalent to a radius of... The sphere, with its direction vector relative to the individual rotor is By judging the fragments Is the angle between the direction of the scattering and the direction relative to the incoming target less than its equivalent half-angle? To determine whether the fragment hit the target, that is... Consider fragments If it hits, like Figure 3 As shown; S216. Assuming the above hit detection process is followed, statistically, the first... The result of the Monte Carlo test was a hit. The probability of destroying the enemy target is After M parallel experiments, the Monte Carlo simulation results for rotor damage to enemy targets are considered to be: (12) The above process can be used to simulate the probability of a single rotor blade damaging an incoming enemy unit. .

[0040] S22. For the interception of a cluster of drones by a formation, the fragmentation process after the formation detonates is simulated in a virtual coordinate system. The probability of damage to the entire cluster of targets is taken as the overall damage probability, and the probability of the cluster of targets being intercepted is calculated based on the damage probability.

[0041] S221, such as Figure 4 As shown, a virtual coordinate system is established, coinciding with the vehicle system. The relative relationships between the cluster targets are maintained and the system is translated to a position where the cluster center is a distance from the armored vehicle's center of mass. The distance is used to simulate the impact of fragments on the cluster target after the rotor formation detonates, under the virtual positional relationship of this virtual coordinate system using the Monte Carlo method. For cluster target The total probability of destruction is considered as the predicted interception probability of the group against the group.

[0042] S222. In the virtual combat situation formed by the virtual coordinate system, the Monte Carlo fragment hit simulation process of one-to-one interception is reused to simulate the fragment scattering after all UAVs in the formation are detonated at the same time. S223. In each Monte Carlo test, count the total number of times each individual target in the attacking swarm is hit by fragments from all drones, and calculate its individual damage probability accordingly. The calculation process is as follows: individual rotors Hit cluster individuals Number of fragments for: (13) Cluster Individual Number of fragments hit for: (14) in, For rotor formation; Cluster Individual The probability of being hit and causing damage The calculation formula is: (15) S224. The criterion for successful interception of the cluster in this experiment is that all individual targets within the cluster are destroyed. Finally, based on the statistical results of multiple experiments, the formation's ability to intercept the cluster target is calculated. Overall interception probability The calculation formula is as follows: (16) in, The size of the cluster target.

[0043] The number of Monte Carlo simulations affects the simulation accuracy and solution time of rotor formation interception probability, and can be used as a parameter for subsequent design improvements to the genetic algorithm.

[0044] S3. Using the product of the value of the intercepted target and the interception probability as the interception benefit, and taking maximizing the overall interception benefit as the optimization objective, a mixed integer nonlinear programming model is constructed by selecting the interception allocation relationship matrix between the UAV and the incoming target and the UAV position set as decision variables.

[0045] S31. Define decision variables. Decision variables include a set of variables composed of... Rotor cluster Intercept Group cluster target The distribution relationship constitutes matrix and hive location ; (17) Among them, elements Indicates the first Are the rotors deployed to intercept the first Group cluster objectives; S32. Set constraints, including allocation constraints and position constraints. Allocation constraints ensure that each UAV is assigned to only one cluster target, and each cluster target is assigned at least one UAV. The allocation matrix should satisfy: each rotor is assigned to only one cluster target, meaning there is no overlap between rotor formations; each cluster target is assigned at least one rotor, meaning the rotor cluster is the sum of the formations. (19) (20) (twenty one) (twenty two) In addition, the location of the bee swarm must be constrained. This is based on the effective kill radius of the fragmentation warhead it carries. Distance from intercepting enemy targets The position of rotary-wing UAVs is constrained: the rotors should be intercepted within the effective kill range of the fragments and avoid being affected by each other. The position constraints ensure that all UAVs are deployed within the effective kill radius of their own warheads and meet the safe distance requirements to avoid mutual friendly fire. (twenty three) (twenty four) in, The effective kill radius of the fragmentation warhead. To determine the location and distance of the target in the intercept cluster.

[0046] S33. Establish optimization goals and define the "Honeycomb Armor Integration" system's response to all incoming attacks. The interception benefits of grouped targets are: (18) in, As for the level of threat, This represents the probability of interception. S34. Define key parameters: whether the cluster target is an enemy unit. It can be used to set up test groups; fragment hit simulation Monte Carlo test number of times This affects the accuracy of the interception probability and the solution time, and can be used to design and improve genetic algorithms; S35. Establish a formation planning and deployment model for UAVs, and optimize the allocation matrix. and hive location set This maximizes the total interception revenue.

[0047] The specific formula is as follows: (25) (26) S4. Design an improved genetic algorithm to solve the mixed-integer nonlinear programming model. The process of the genetic algorithm is as follows: Figure 5 As shown, by improving the genetic operator and adjusting the Monte Carlo simulation parameters, the drone formation deployment scheme with the optimal interception benefit is obtained.

[0048] Genetic algorithms simulate the evolutionary process of biological populations in nature. Through the genetic processes of selection, crossover, and mutation, they accumulate dominant traits and ultimately achieve optimal selection and solution.

[0049] S41. Define the total interception reward of the enemy's incoming target as the fitness function of the genetic algorithm, design the encoding, and assign the solution variable matrix. Encoded in binary order according to row and column sequence An array of 0s and 1s of length, representing the location of the bee colony. Position them according to rotor number Coordinate values ​​are encoded as real numbers. A real number array is used to simplify operations and to impose restrictions on subsequent genetic operations based on constraints. (27) (28) S42. Design a selection operator and adopt a roulette wheel strategy. The selection interval is allocated between 0 and 1 based on the fitness function value of the current generation sample. The sample in the selected interval is generated by random number and used as the parent of the next generation. The number of parents is equal to the population size, and the selected parents do not need to remove the current generation sample. S43. Design a crossover operator for the distribution matrix. Encoded genome ,use Point intersection, based on the allocation matrix Constraints on The position of the intersection point is restricted, and the intersection point is located at the... Allogeneic and the first Between alleles, It should meet the following requirements: (32) (33) Location of the bee colony Encoded genome Use uniform crossover and set the crossover probability. For each gene, we iterate through it and determine whether to exchange that gene from the parent sample based on the crossover probability, ensuring that the genes that are crossed over have the same physical meaning in the deployment problem. S44. Design mutation operators for the distribution matrix. Encoded genome Employing crossover mutation, where the first crossover is performed during mutation. Allogeneic and the first The entire gene segment between alleles. The above conditions should be met by formulas (32) and (33); Location of the bee colony Encoded genome Use real number mutations and set the mutation probability. For each gene, iterate through it and determine whether to randomly mutate the real number based on the mutation probability. It should be ensured that the gene satisfies the relevant position constraints in the deployment problem. S45. To optimize computation time, during algorithm iteration, analogous to the simulated annealing algorithm, the number of fragment hits in simulated Monte Carlo trials is adjusted according to the evolutionary algebra. Using a larger number of Monte Carlo trials in the fitness function of smaller algebras To ensure the accuracy of fragment hit simulation in the early stages of optimization, a smaller number of Monte Carlo trials are used in the fitness function of large algebras. This accelerates the computation speed in the later stages; when the algorithm meets the termination condition, it outputs the allocation matrix and position set corresponding to the individual with the highest fitness, which serves as the optimal UAV formation deployment scheme.

[0050] Therefore, the present invention adopts the above-mentioned method for drone formation deployment for collaborative defense missions, which effectively improves the accuracy and combat reliability of drone formation defense deployment by integrating multi-target optimization, adversarial constraints and Monte Carlo damage simulation.

[0051] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for deploying a drone swarm for a swarm-based collaborative defense mission, wherein the swarm-based integrated system comprises an armored vehicle, a drone swarm, and a networking communication module, wherein the armored vehicle is equipped with radar for detecting incoming targets, and the drone swarm consists of multiple rotary-wing drones, each equipped with an airborne camera for auxiliary detection, characterized in that... Includes the following steps: S1. Obtain multi-source perception information through the network communication module, fuse radar perception information and airborne camera perception information, construct a threat assessment model of incoming targets, and quantify the value of intercepting the target based on the distance, size and confidence level of perception and identification of the incoming target. S11. Obtain radar perception information of incoming targets through the radar of armored vehicles, treat individual incoming targets as flat targets, and estimate the radar cross-section. For the first Group attack cluster target Calculate the total radar cross-section The radar echo power received by the radar cluster is calculated using radar equations. The relationship between radar echo power and cluster size and target distance is obtained to assess distance and scale information. The calculation formula is as follows: （1） （2） （3） （4） in, The area of ​​the flat plate for the target individual. For radar wavelength, The number of target individuals within the cluster. For transmission power, For the antenna gain of a single-station radar, The distance between the radar and the incoming cluster of targets. The location of the cluster target, The size of the cluster target; S12. Obtain visual perception information of the incoming target through the drone's onboard camera, and determine the pixel size of the target drone in the image. and recognition accuracy Calculate the positive detection rate To determine the target type, the calculation process is as follows: （5） （6） （7） （8） （9） in, For camera focal length, For the target drone size, The physical size of a single pixel in the sensor. The distance between the drone and the incoming swarm target. The smallest pixel size that the camera can recognize. The attenuation coefficient is... As an occlusion factor, For the imaging size of the cluster target, The image size of an individual within the target cluster. Sensitivity coefficient This indicates whether the cluster target is a loitering munition cluster. A value of 1 indicates that the cluster target is a loitering munition cluster, while a value of 0 indicates that the cluster target is a non-enemy target. For the first The distance between the cluster target and the drone equipped with an airborne camera; S13. Define the threat level of an incoming target to the protected platform based on radar echo power and positive detection rate. The calculation formula is as follows: （10） in, For armored vehicles; S2. The interception process of an abstract UAV swarm formation meeting an incoming target is constructed based on the Monte Carlo simulation method, combined with the relative positional relationship between the UAV and the incoming target and the characteristics of the UAV's warhead, to quantify the interception probability of the UAV swarm against the incoming target. S21. For interception of a single target by a single UAV, calculate the damage probability based on the number and distribution of fragment hits within the effective kill radius; S22. For the interception of a cluster of drones by a formation, the fragmentation process after the formation is detonated is simulated in a virtual coordinate system. The probability of damage to the cluster target as a whole is taken as the damage probability, and the probability of the cluster target being intercepted is calculated based on the damage probability. S3. Using the product of the value of the intercepted target and the interception probability as the interception benefit, and taking maximizing the overall interception benefit as the optimization objective, a mixed integer nonlinear programming model is constructed by selecting the interception allocation relationship matrix between the UAV and the incoming target and the UAV position set as decision variables. S4. Design an improved genetic algorithm to solve the mixed-integer nonlinear programming model. By improving the genetic operators and adjusting the Monte Carlo simulation parameters, obtain the drone formation deployment scheme with the optimal interception benefit.

2. The method for deploying UAV swarms for collaborative defense missions against bee-like drones according to claim 1, characterized in that: S12 recognition accuracy Modeled as a Sigmoid function, with a decay coefficient The value range is 5~10, and the sensitivity coefficient is... The value range is 0.1 to 0.

5.

3. The method for deploying UAV swarms for collaborative defense missions using bee-like drones according to claim 1, characterized in that, The specific steps of S21 are as follows: S211, Let the position of the individual rotorcraft UAV be... The location of the single incoming target is The distance between the two ; S212. Set the effective kill radius of the drone. For targets outside this range, the probability of damage is determined to be zero; S213. For targets within the range, the baseline damage probability of a single fragment is set as follows: And assuming that the fragment hits follow a Poisson distribution and are uniformly distributed within the target area, the damage probability is... The calculation formula is as follows: （11） S214. Establish a spherical coordinate system centered on the UAV to simulate the azimuth angle of all fragments scattered during a single test. and polar angle Evenly distributed on the surface, any fragment The direction vector is ; S215, the incoming target is equivalent to a radius of... The sphere, with its direction vector relative to the individual rotor is By judging the fragments Is the angle between the direction of the scattering and the direction relative to the incoming target less than its equivalent half-angle? To determine whether the fragment hit the target, that is... Consider fragments Hit; S216. Through numerous independent Monte Carlo experiments, the hit results are statistically analyzed, and the final probability of damage to the target is calculated by approximating the frequency of hits. The calculation formula is as follows: （12） in, For the first Number of hits in the Monte Carlo test The probability of damage to the incoming target. The number of Monte Carlo trials.

4. The method of claim 3, wherein, The specific steps of S22 are as follows: S221. Establish a virtual coordinate system that coincides with the coordinate system of the protected armored vehicle, and translate the cluster targets while maintaining the relative relationships between the nodes to a position where the center of the cluster is a distance from the centroid of the armored vehicle. The distance; S222. In the virtual combat situation formed by the virtual coordinate system, the Monte Carlo fragment hit simulation process of one-to-one interception is reused to simulate the fragment scattering after all UAVs in the formation are detonated at the same time. S223. In each Monte Carlo test, count the total number of times each individual target in the attacking swarm is hit by fragments from all drones, and calculate its individual damage probability accordingly. The calculation process is as follows: individual rotors Hit cluster individuals Number of fragments for: （13） Cluster Individual Number of fragments hit for: （14） wherein, a rotor formation; Cluster individual Hit probability of damage The formula is: （15） S224. The criterion for successful interception of the cluster in this experiment is that all individual targets within the cluster are destroyed. Finally, based on the statistical results of multiple experiments, the formation's ability to intercept the cluster target is calculated. Overall interception probability The calculation formula is as follows: （16） wherein, is the size of the cluster target.

5. The method of claim 1, wherein, The specific steps for S3 are as follows: S31. Define decision variables. Decision variables include a set of variables composed of... Rotor cluster Intercept Group cluster target The distribution relationship constitutes matrix and hive location ; （17） wherein the elements represent the first mounting the rotor to intercept the second group of cluster targets; S32. Set constraints, including allocation constraints and position constraints. The allocation constraints ensure that each UAV is assigned to only one cluster target, and each cluster target is assigned at least one UAV. The position constraints ensure that all UAVs are deployed within the effective kill radius of the friendly warhead and meet the safe distance requirements to avoid friendly fire. S33. Establish optimization goals and define the bee-armor integrated system's response to all incoming attacks. The interception benefits of grouped targets are: （18） in, As for the level of threat, This represents the probability of interception. S34. Define key parameters, including whether the cluster target is an enemy unit. And fragment hit simulation Monte Carlo test number of times ; S35. Establish a formation planning and deployment model for UAVs, and optimize the allocation matrix. and hive location set This maximizes the total interception revenue.

6. A method for deploying UAV swarms for collaborative defense missions using bee-like drones, as described in claim 5, is characterized in that... The specific formulas for position constraints and assignment constraints in S32 are as follows: Assignment constraints: （19） （20） （21） （22） Position constraints: （23） （24） in, The effective kill radius of the fragmentation warhead. To determine the location and distance of the target in the intercept cluster.

7. The method of claim 6, wherein, The specific formula for building the S35 model is as follows: （25） （26） in, This indicates a constraint condition.

8. A method for deploying UAV swarms for collaborative defense missions against bee-like drones, as described in claim 5, is characterized in that... The specific steps for S4 are as follows: S41. Define the total interception reward of the enemy's incoming target as the fitness function of the genetic algorithm, design the encoding, and assign the solution variable matrix. Encoded in binary order according to row and column sequence An array of 0s and 1s of length, representing the location of the bee colony. Position them according to rotor number Coordinate values ​​are encoded as real numbers. An array of real numbers; （27） （28） S42. Design a selection operator and adopt a roulette wheel selection strategy. The selection probability is allocated proportionally according to the individual fitness value to select parent individuals from the current population. S43. Design a crossover operator to encode the allocation matrix using point crossover, and the position of the crossover point must satisfy the allocation constraint conditions. The location coding uses uniform crossover, and the corresponding coordinate values ​​are swapped according to a set probability. S44. Design a mutation operator to use exchange mutation for the encoding of the allocation matrix. The mutation interval must satisfy the allocation constraint. The location encoding uses real number mutation, and new coordinates are randomly generated within the feasible region according to a set probability; S45. During the algorithm iteration process, the number of Monte Carlo simulation trials is dynamically adjusted according to the current generation. A larger number of trials is used in the early stage of evolution to ensure evaluation accuracy, and a smaller number of trials is used in the later stage of evolution to accelerate convergence. When the algorithm meets the termination condition, the allocation matrix and position set corresponding to the individual with the highest fitness are output as the optimal UAV formation deployment scheme.

Images