A Multi-Objective UAV Dynamic Allocation and Synchronized Encirclement Method

Abstract

The present invention discloses a method for dynamic allocation and synchronous encirclement of unmanned aerial vehicles (UAVs) for multiple targets. Based on the AP98 wind-induced drift model and combined with the self-power of the targets, the positions of the targets during UAV encirclement are predicted. Then, through an improved algorithm, accurate target segmentation and reasonable resource allocation are achieved. Additionally, through a conflict resolution mechanism and a task allocation optimization link, the tasks can be dynamically adjusted in real time to effectively handle complex and changeable situations, greatly enhancing the flexibility and effectiveness of task allocation. For the problem of multi-target synchronous encirclement, the present invention adopts an improved two-stage algorithm for autonomous UAV swarm guidance. In the first stage, a consensus algorithm is used to reach an agreement on the initial conditions for the second-stage guidance. In the second stage, an improved proportional guidance algorithm is applied to dynamically adjust the acceleration direction to ensure that the UAVs approach the targets along the optimal path and complete the encirclement task.

Description

Technical Field

[0001] The present invention relates to the field of UAV control, and particularly to a method for dynamic allocation and synchronous encirclement of UAVs for multiple targets. Background Art

[0002] Due to advantages such as low cost, small size, and flexibility, UAVs are widely used in tasks such as reconnaissance, strike, search and early warning. The UAV swarm system completes more complex and cumbersome tasks that are difficult for a single UAV to achieve through collaborative cooperation. Among them, the encirclement method of UAV swarms is an important research direction in the field of UAV swarm technology, playing an important role in fields such as target reconnaissance and encirclement strikes, and having important theoretical research and practical exploration value.

[0003] Most current UAV encirclement studies are aimed at single targets, and the presence of multiple targets will affect the final execution effect of the task. In the scenario of UAV swarm encirclement of multiple targets, the environment is complex and dynamically changing, and the interference of wind is particularly prominent. Traditional prediction methods mostly rely on basic motion models and positioning, without considering wind-induced drift, resulting in large prediction deviations and poor encirclement effects. In UAV encirclement tasks, traditional task allocation methods often lack fine consideration of target dynamic changes and resource balance. In a multi-target scenario, some targets are easily ignored or resources are unevenly distributed, and it is difficult to flexibly adjust according to the task execution situation. In addition, during the UAV encirclement process, the position is constantly changing, and it is easy to be sensed by the target, resulting in the escape of the target. Therefore, how to dynamically allocate encirclement tasks according to the current real-time positions of UAVs and targets, and plan an optimal encirclement path is the key problem in current UAV multi-target encirclement research. Summary of the Invention

[0004] Object of the Invention: The present invention aims to provide a method for dynamic allocation and synchronous encirclement of UAVs for multiple targets that reasonably divides multiple targets and uses a conflict resolution mechanism and task allocation optimization to adjust tasks in real time and dynamically.

[0005] Technical Solution: The method for dynamic allocation and synchronous encirclement of UAVs for multiple targets according to the present invention includes the following steps:

[0006] (1) Based on the UAVs encircling the targets, construct a kinematic model of multiple UAVs;

[0007] (2) Based on the wind-induced drift model AP98 and the self-powered velocity component of the target, determine the target velocity and update the target position;

[0008] (3) According to the UAV positions and target positions, use the K-means clustering algorithm, with the target positions as the initial cluster centers, to perform task allocation;

[0009] (4) Using the PI performance impact algorithm to determine the RPI removal performance impact matrix and the IPI inclusion performance impact matrix of the roundup mission, we can obtain the UAV task allocation matrix;

[0010] (5) Perform conflict detection. If there is a conflict, adjust the task allocation. After resolving the conflict, update the UAV task allocation matrix, RPI matrix, and IPI matrix, and proceed to step (6);

[0011] (6) Considering the task completion time, task execution cost and task completion benefit, construct the objective function, determine the constraints, and optimize the task allocation;

[0012] (7) Construct a multi-objective information matrix based on the kinematic model of multiple UAVs. The multi-objective information matrix includes the state information matrix and the state information derivative matrix of the UAVs relative to the targets. Auxiliary variables are introduced for each target to coordinate the multi-objective parameters. The virtual control variables are determined based on the derivatives of the auxiliary variables to obtain the actual control signals of the UAVs.

[0013] (8) Use proportional guidance method to dynamically adjust the acceleration direction of the UAV, optimize the UAV path, and perform the capture mission;

[0014] (9) Determine the capture effect. If the target escapes or the distance between the drone and the target continues to increase, mark the corresponding task as an unassigned task and return to step (4) to reassign the drone task.

[0015] Furthermore, in step (2), the target velocity is determined based on the wind drift model AP98 and the target's own dynamic velocity component. j speed, target j exist x Axis speed for

[0016] ;

[0017] Target j exist y Axis speed for

[0018] ;

[0019] in, and Respectively represent the target j exist x Axis and y The velocity component of the own power on the shaft, is the downwind drift vector, is the angle between the wind drift direction and the reference coordinate axis, is the vertical wind pressure vector;

[0020] In step (2), the target j The updated position is

[0021] ;

[0022] ;

[0023] In the formula, is the moment t The target j The position coordinates of.

[0024] Furthermore, step (3) is specifically as follows:

[0025] Define the UAV position data set and the target position data set , where N is the total number of UAVs, M is the total number of targets, and ;

[0026] Adopt the K-means clustering algorithm to calculate the distance between the UAV and the target, use the target position as the initial cluster center for clustering, and perform task allocation. The number of UAVs corresponding to each task is q ;

[0027] Judge q Whether it satisfies , if the number of UAVs corresponding to a certain task q is greater than , then calculate the average center of all UAVs in this task, and reallocate the UAV farthest from the average center to the task corresponding to the target closest to it.

[0028] Furthermore, in step (5), there is a conflict, and the task allocation adjustment is specifically as follows:

[0029] If the number of UAVs corresponding to the task q is greater than , then calculate the average center of all UAVs in this task, and determine the UAV farthest from the average center;

[0030] Calculate the encirclement state index of other tasks SI , and allocate the UAV to the task corresponding to the minimum encirclement state index minSI .

[0031] Furthermore, before entering step (7), consider the task completion time, task execution cost, and task completion benefit to construct an objective function, and set constraint conditions to solve the optimal task allocation plan.

[0032] Furthermore, consider the task completion time, task execution cost, and task completion benefit to construct the objective function J For

[0033] ;

[0034] Among them, , , are weight coefficients, ; is the unmanned aerial vehicle i to complete the task j required time, is the unmanned aerial vehicle task list, is the cost required for the unmanned aerial vehicle i to complete the task j, is the unmanned aerial vehicle i to complete the task j the obtained benefit.

[0035] Furthermore, in step (7), for each objective, introduce auxiliary variables to coordinate multi-objective parameters. The auxiliary variables and are as follows:

[0036] ;

[0037] ;

[0038] ;

[0039] Among them, is the distance from the unmanned aerial vehicle i to the target j , V is the Euclidean norm of the unmanned aerial vehicle speed vector, is the angle between the unmanned aerial vehicle speed vector and the line of sight LOS, is the deviation angle between the unmanned aerial vehicle speed vector and the line of sight in the vertical plane, is the deviation angle between the unmanned aerial vehicle speed vector and the line of sight in the horizontal plane.

[0040] Furthermore, in step (7), the virtual control variable is

[0041] ;

[0042] ;

[0043] ;

[0044] The actual control signal of the unmanned aerial vehicle is

[0045] ;

[0046] ;

[0047] Among them, and respectively represent the control signals in the horizontal and vertical directions, and are the first-order derivatives of the auxiliary variables.

[0048] Furthermore, in step (8), the proportional guidance method includes a first guidance stage and a second guidance stage. In the first guidance stage, a consensus algorithm is used to perform parameter preprocessing on the auxiliary variables and to make the distance parameter and angular velocity parameter of the UAV tend to be consistent; in the second guidance stage, the acceleration is corrected by introducing the angular position to dynamically adjust the acceleration direction of the UAV. The acceleration direction of the UAV is as follows:

[0049] ;

[0050] ;

[0051] Among them, and respectively represent the control signals in the horizontal and vertical directions, is the navigation constant for target j, and respectively represent the line-of-sight rotation speed components in the horizontal and vertical directions.

[0052] Furthermore, the consensus algorithm in the first guidance stage of step (8) is specifically as follows:

[0053] ;

[0054] ;

[0055] ;

[0056] Among them, and respectively represent the state information of the UAV i relative to the target j and its derivative, is the control signal used to achieve the consistency of the state variables among UAVs, represents the i th UAV and the k th UAV can exchange information about the target j ; represents the i th UAV and the kIndividual drones cannot exchange information about the target j and and is the gain coefficient .

[0057] Advantageous effects: Compared with the prior art, the significant advantages of the present invention are as follows: 1. The present invention uses the wind-induced drift model AP98 for the dynamic allocation and synchronous encirclement of drones for multiple targets. By combining the target's own power, wind speed, wind direction, and correction coefficient, the drift speed is calculated in detail, and then the target position is accurately predicted through the discrete time step formula. It can effectively cope with the dynamic changes of targets in complex meteorological environments, greatly improving the accuracy and reliability of target position prediction, and laying a good foundation for subsequent encirclement; 2. The present invention uses an improved K-means algorithm for target segmentation to achieve the reasonable allocation of encircling drones, ensuring resource balance. When assigning tasks, multiple key factors are comprehensively considered to calculate rewards and metrics, improving the scientific nature of the allocation. Through the conflict resolution mechanism and the task allocation optimization link, the tasks can be dynamically adjusted in real time to effectively cope with complex and changeable situations, greatly enhancing the flexibility and effectiveness of task allocation; 3. The present invention uses an improved two-stage algorithm for autonomous drone swarm guidance. In the first stage, a consensus algorithm is used to reach an agreement on the initial conditions for the second-stage guidance; in the second stage, an improved proportional guidance algorithm is applied to dynamically adjust the acceleration direction through the proportional guidance method to ensure that the drones approach the target along the optimal path and complete the encirclement task. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 is a flowchart of the present invention;

[0059] Figure 2 is a wind-induced drift vector diagram of the wind-induced drift model AP98;

[0060] Figure 3 is a schematic diagram of the initial position distribution of the encircling drones and the target;

[0061] Figure 4 is a comparison diagram of the convergence curves of the objective functions;

[0062] Figure 5 is a change diagram of the distance from the encircling drones to the target. DETAILED DESCRIPTION OF THE INVENTION

[0063] The following further describes the present invention with reference to the accompanying drawings.

[0064] The method for dynamic allocation and synchronous encirclement of drones for multiple targets according to the present invention includes the following steps:

[0065] (1) Based on the encirclement of the target by drones, a kinematic model of multiple drones is constructed.

[0066] For the mathematical description of the UAV encirclement problem, for each target j, j = 1, 2, …, M, where M is the number of targets, the present invention adopts a fifth-order kinematic model, and the distance change rate formula is

[0067] ;

[0068] In the formula, V represents the Euclidean norm of the UAV velocity vector; represents the distance from UAV i to target j; LOS is the line of sight, i.e., the straight line from the UAV to target j; is the deviation angle between the UAV velocity vector and the line of sight in the vertical plane. is the deviation angle between the UAV velocity vector and the line of sight in the horizontal plane. It reflects the dynamic change of the distance when the UAV moves towards different targets;

[0069] The formula for the rotational speed component of the line of sight is

[0070] ;

[0071] ;

[0072] In the formula, represents the rotational speed of the line of sight in the horizontal direction, represents the rotational speed of the line of sight in the vertical direction.

[0073] The formula for the rate of change of the angle deviation is

[0074] ;

[0075] reflects the complex relationship between the change of the angle deviation of the UAV relative to target j and the acceleration and other parameters in a multi-target scenario. Formulas and are the inclination angles of the line of sight, i.e., the straight line from the UAV to target j (LOS), in the vertical and horizontal planes respectively.

[0076] ;

[0077] In the formula, is the acceleration component of the UAV acceleration in the horizontal direction, the acceleration component of the UAV acceleration in the vertical direction, which is used as the control signal. By adjusting these two acceleration components, the UAV can flexibly track the target in a complex environment and meet the requirements of the multi-target encirclement task. When the target is moving, the model adjusts the LOS direction and the UAV velocity vector in real time to ensure that the UAV can efficiently track the target in a complex environment. By controlling the acceleration signals and , the dynamic allocation and encirclement actions of the UAV are realized.

[0078] (2) Determine the target velocity based on the wind-induced drift model AP98 and the self-powered velocity component of the target, and update the target position.

[0079] The present invention improves the traditional AP98 model, focuses on the impact of wind-induced drift on the target position, and proposes a multi-target position prediction method while considering the target's own dynamic factors.

[0080] Calculate the wind-induced drift parameters:

[0081] Let the velocity components of the self-power of target j be and .

[0082] Based on the principle of the AP98 model, the actual wind pressure drift vector (Leeway, L) can be decomposed into two parts: the downwind drift vector (Downwind Leeway, DWL) and the crosswind pressure vector (Crosswind Leeway, CWL). For each target j, according to the wind speed at a height of 10 meters Calculate the wind-induced drift vector.

[0083] The downwind drift vector :

[0084] ;

[0085] The crosswind pressure vector is divided into the right-biased vertical direction and the left-biased vertical direction :

[0086] ;

[0087] ;

[0088] Among them, and are the fitting parameters of target j in the downwind direction, is the regression deviation, and the same applies to the other parameters.

[0089] Calculate the total velocity of target j:

[0090] ;

[0091] ;

[0092] Among them, is the angle between the wind-induced drift direction and the reference coordinate axis.

[0093] After time After time After that, the position coordinates of target j are updated to

[0094] ;

[0095] ;

[0096] In the formula, is the time t target j position coordinates.

[0097] Through the above steps, the AP98 model is used to predict and update the multi-target positions.

[0098] (3) According to the UAV position and the target position, the K-means clustering algorithm is adopted, and the target position is used as the initial cluster center for task allocation.

[0099] Considering a multi-agent system that includes N UAVs for encirclement and M targets. To ensure that all targets are successfully captured, it must be satisfied that . Ensure that each target corresponds to a subsystem, and the number q of UAVs for encirclement included in each subsystem needs to meet the prerequisite: . To make reasonable use of the encirclement ability of the multi-agent system, the number of UAVs for encirclement corresponding to each target should be evenly distributed. Therefore, the number q should also satisfy: .

[0100] (31) Define a data set containing the positions of N UAVs for encirclement as . The position of the target is set as the initial M cluster centers, defined as , The distance between is calculated by the following formula.

[0101] ;

[0102] According to the calculation results, is planned into the data cluster represented by the nearest target vector.

[0103] (32) Through step (31), M subsystems are initially formed. Then, calculate the number of UAVs for encirclement in the subsystem, denoted as q. Judge whether q satisfies .

[0104] (33) For those with more than A subsystem h for capturing drones. Calculate the average center of subsystem h, find the capturing drone farthest from the average center, and divide this capturing drone into the nearest other subsystem. To prevent the same capturing drone from moving back and forth between the two nearest subsystems and causing an algorithm loop, it is necessary to ensure that the mobilized capturing drone cannot be divided into the previous subsystem.

[0105] (34) Repeat step (33) and step (34) until all elements in the subsystem meet the requirements.

[0106] (4) Use the PI performance impact algorithm to determine the RPI removal performance impact matrix and IPI inclusion performance impact matrix for the capturing task, and obtain the UAV task allocation matrix.

[0107] According to the target segmentation result of the improved K - means algorithm, regard each subsystem containing the target as an overall task.

[0108] Regard the capturing task of each subsystem as an independent task . Calculate the task execution reward

[0109] ;

[0110] Among them, is the weight of target j, used to measure the importance or priority of target j. is the Euclidean distance from UAV i to target j, is the maximum distance constant, is the remaining energy of UAV i, is the total energy of UAV i.

[0111] In the PI algorithm, both the benefits of the multi - UAV system and the impact of the task on other tasks are considered. Among them, RPI is called , which can illustrate the impact of task on the i - th UAV during the task allocation process. The impact of task on the i - th UAV can be expressed as the equation

[0112] ;

[0113] Among them, is the task list of UAV i, recording the task assignment situation of this UAV. indicates that task is removed from . l is the possible insertion position of task in . is in Execution reward of the task inserted at index w. Indicates the reward of the i-th drone when the task is included in the i-th drone's reward when there is no task is in the i-th drone.

[0114] To measure the local inclusion performance of the task the inclusion performance impact IPI is used. In addition, the IPI of the task for the i-th drone can be defined as the reward of the task plus the maximum difference in rewards when performing the task after the task. Calculate the IPI value of the task :

[0115] ;

[0116] where indicates that the task is added to at position l. And determine the insertion position l, add the task to the appropriate drone task list while ensuring that the number of drones q in the subsystem after addition satisfies . In the PI algorithm, an RPI table and a drone list are stored on each drone, defining a single matrix representing the RPI list matrix to store the allocation between drones and tasks. If and are to be merged into one matrix, then has a size of . Additionally, the element in represents the RPI value of the estimated task for drone i. Update the matrix to reflect the association between the task

[0117] (5) Perform conflict detection. If there is a conflict, adjust the task allocation. After resolving the conflict, update the drone task allocation matrix, RPI matrix, and IPI matrix, and proceed to step (6).

[0118] Drones regularly exchange matrices , timestamps , target position changes and the encirclement state (such as distance to the target, integrity of the encirclement, etc.).

[0119] Check the number of UAVs in each subsystem If there is a number of UAVs in subsystem h Calculate the average center of the UAVs in this subsystem Find the pursuit UAV farthest from the average center .

[0120] For other subsystems k, calculate their current pursuit state indicators :

[0121] ;

[0122] Among them, and are weight coefficients, is the distance from UAV i in subsystem k to the target, is the variance of the distance.

[0123] Assign to the subsystem k with the smallest (i.e., the worst pursuit state), but ensure that to prevent loops.

[0124] Update according to the UAV adjustment the matrix and the task execution order For the adjusted UAV recalculate the RPI value estimate for its new subsystem task and update ; update the task list of subsystem k that receives the UAV and the relevant matrix elements.

[0125] (6) Consider the task completion time, task execution cost, and task completion benefit to construct an objective function, determine the constraint conditions, and perform task allocation optimization.

[0126] Consider the task completion time, task execution cost, and task completion benefit to construct an objective function. Let be the time required for UAV i to complete task j, be the UAV task list, then the partial objective function of the task completion time is

[0127] ;

[0128] Let be the cost required for UAV i to complete task j, then the partial objective function of the task execution cost is

[0129] ;

[0130] Let The revenue obtained by the UAVs after completing the tasks. Then, a partial objective function of the task completion revenue is

[0131] ;

[0132] Set weight coefficients 、 、 to balance the importance of the three factors of task completion time, task execution cost, and task completion revenue. Then, the comprehensive objective function J is

[0133] ;

[0134] The aim is to minimize the total time for all UAVs to complete their respective tasks, improve the task execution efficiency, effectively reduce the resource consumption during task execution, and enhance the resource utilization rate.

[0135] Determine the constraint conditions and define the binary decision variable such that when the i-th encirclement UAV is assigned to the j-th target, the value of is 1; when the i-th encirclement UAV is not assigned to the j-th target, the value of is 0.

[0136] Target quantity constraint: , indicating that all targets must be fully assigned to ensure that each target has a corresponding UAV to perform the task and avoid the situation where a target is not assigned.

[0137] Multi-UAV cooperation constraint: , meaning that one UAV can be assigned to at most one target to ensure the reasonable utilization of UAV resources and avoid resource dispersion and task conflicts caused by one UAV performing multiple target tasks simultaneously.

[0138] Task time constraint: , that is, the time for the UAV to execute the task should be within its maximum flight time range.

[0139] After determining the objective function and constraint conditions constructed by minimizing the task completion time, minimizing the task execution cost, and maximizing the task completion revenue, the existing improved multi-objective particle swarm optimization algorithm is used to optimize the task allocation results. Find the optimal or sub-optimal task allocation scheme that achieves a balance among multiple objectives and satisfies the constraint conditions to realize the efficiency and rationality of task allocation.

[0140] (7) Construct a multi-target information matrix according to the multi-UAV kinematic model. The multi-target information matrix includes the state information matrix and the state information derivative matrix of the UAVs relative to the targets. Introduce auxiliary variables for each target to coordinate the multi-target parameters, and determine the virtual control variables according to the derivatives of the auxiliary variables to obtain the actual control signals of the UAVs.

[0141] After the optimization of the encirclement mission assignment in step (6) is completed, each mission UAV group executes the mission of encircling the target. The consensus algorithm can coordinate the motion of the UAVs and adjust the flight parameters of each UAV. Ensure that the flight parameters of each UAV are consistent to prepare for the subsequent guidance phase.

[0142] Construct a multi-target information matrix and , where and represent the state information of UAV i relative to target j and its derivative respectively, i = 1, 2, …, N, and N is the number of UAVs; for example can be the distance-related information from UAV i to target j, is the rate-of-change information of it.

[0143] Define the comprehensive adjacency matrix , and the determination rule of its elements is as follows: if UAV i and UAV k can exchange information about target j, then , otherwise it is 0. Describe the connection relationship between UAVs in multi-target information sharing through this matrix.

[0144] The formula of the consensus algorithm is as follows:

[0145] ;

[0146] ;

[0147] where, , and are gain coefficients, is used to adjust the consensus between UAVs, and realize the collaborative calculation between UAVs under multi-target information through the comprehensive adjacency matrix, so that the state variables of UAVs on different targets asymptotically reach consensus, and promote the coordinated action of the encircling UAV swarm in multi-target tasks.

[0148] Perform multi-target parameter coordination. For each target j, introduce auxiliary variables and , where represents the distance from UAV i to target j. is the angle between the velocity vector of the UAV and the LOS, and the determination method is as follows:

[0149] ;

[0150] Considering the fifth-order kinematic model, the above two formulas are differentiated:

[0151] ;

[0152] ;

[0153] Virtual control According to the expression of, the virtual control is introduced as

[0154] ;

[0155] The fifth-order kinematic model is transformed into an equivalent linear model. The complex non-linear control problem is linearized. Considering the auxiliary variable and the virtual control , the above two equations can be rewritten as

[0156] ;

[0157] ;

[0158] The actual control signals and of the UAV can be calculated by , and their expressions are

[0159] ;

[0160] ;

[0161] Through these control signals, the UAV adjusts its flight parameters in a multi-target scenario, achieving parameter coordination of each UAV on different targets. Making the guidance initial conditions of the multi-target proportional guidance algorithm consistent.

[0162] (8) Adopt the proportional guidance method to dynamically adjust the acceleration direction of the UAV, optimize the UAV path, and perform the encirclement mission.

[0163] The improved multi - target proportional guidance algorithm is used to aim at the targets of each UAV, ensuring that the UAV can approach the target along the optimal path and finally complete the encirclement mission. It generates a UAV acceleration control signal proportional to the LOS angular velocity. However, before directly using this algorithm to encircle the target, it is necessary to coordinate the specified parameters, namely the relative distance to the target and the angle between the UAV velocity vector and the target line of sight. Therefore, the guidance process can be divided into two stages. In the first stage, after using the consensus algorithm in step 7 to make the guidance initial conditions consistent, in the second stage, the improved multi - target proportional guidance algorithm is used to aim at the targets of each UAV.

[0164] (81) The first guidance stage

[0165] The purpose of the first guidance stage is to achieve the approaching conditions for each UAV at the beginning of the next stage. To achieve this, the results obtained in step 7 are used, that is, the guidance initial conditions are made consistent through the consensus algorithm. Using the control values obtained in step 7, ensure that the parameters of each UAV at the end of this stage and are approaching:

[0166] ;

[0167] ;

[0168] where is the UAV number. and are the allowable ranges of accuracy. Ensure that the distance parameters and angular velocity parameters of each UAV gradually tend to be consistent.

[0169] For a single UAV, within a limited time interval, entering the next - stage guidance means that the difference between its alignment parameters and the similar parameters of all other UAVs is less than a small value representing the coordination accuracy.

[0170] When the parameters of each UAV are adjusted so that the difference between all UAVs is less than the set threshold, it marks the completion of the first stage, that is, it can enter the next stage for more precise guidance.

[0171] (82)The second guidance stage

[0172] The improved multi - target proportional guidance algorithm is adopted, and the angle position information is introduced to correct the acceleration, making the capture of the target by the UAV more accurate.

[0173] The control signal formula is

[0174] ;

[0175] ;

[0176] in, The navigation constant for target task j is dynamically adjusted according to the target characteristics and the real-time relative position between the UAV and the target. The acceleration direction is dynamically adjusted through the proportional guidance method to ensure that the UAV approaches the target in the optimal path. The improved multi-target proportional guidance algorithm is used to guide each UAV to the capture point around the target to complete the capture task.

[0177] (9) Determine the capture effect. If the target escapes or the distance between the drone and the target continues to increase, mark the corresponding task as an unassigned task and return to step (4) to reassign the drone task.

[0178] When the subsystem capture effect is not good (such as the target escapes, the distance between the drone and the target continues to increase, etc.), the capture task of this subsystem is Mark the task as unassigned. , return to step 4 and recalculate the RPI and IPI values ​​for each drone , try to include other suitable drone mission lists , and the number of drones in the subsystem is limited.

[0179] In order to verify the effectiveness and rationality of the method proposed in the present invention, the following examples are given:

[0180] In the MATLAB environment, a simulation experiment is conducted to verify the proposed multi-target UAV dynamic allocation and synchronous capture method. In order to simplify the calculation, the capture UAVs and targets are regarded as particle models in the MATLAB platform to verify the capture effectiveness and convergence of the proposed method.

[0181] Set 9 capture drones to capture 3 moving targets, where the positions of the targets and capture drones are randomly generated and distributed in a square area of ​​800 m × 800 m. The simulation scene settings are shown in Table 1.

[0182] Table 1 Simulation scenario settings

[0183] ;

[0184] A scene at a certain moment is selected to verify the effectiveness of the algorithm. The relative distribution positions of multiple capture drones and targets are as follows: Figure 3 As shown, a circle with the target as the center is formed around the target, and the captured drones are evenly distributed on the circle.

[0185] The multi-target UAV dynamic allocation and synchronous capture method proposed in the present invention is used to carry out multi-UAV collaborative capture.

[0186] Set weight coefficients , then the comprehensive objective function J is

[0187] ;

[0188] To verify the algorithm performance, a comparative experiment is carried out between the multi-objective pursuit method in this paper and the multi-objective particle swarm algorithm and the multi-objective wolf pack algorithm. The comparison of the objective function convergence curves after 100 iterations of the objective function is as Figure 4 shown. After 100 iterations, the multi-objective pursuit method proposed in the present invention shows better convergence effect compared with the multi-objective particle swarm algorithm and the multi-objective wolf pack algorithm. The optimal solution can be quickly reached during the iteration process.

[0189] Taking a subsystem in the multi-objective pursuit task as an example, three pursuit UAVs, namely UAV1, UAV2, and UAV3, are used in the subsystem to pursue the target Target. The simulation step size is set to 0.1, and the pursuit radius is set to 10m. As Figure 5 shown, the distances of each UAV relative to the target during the pursuit process can be seen to converge to the same value at 42 seconds, and the distances are all 10 meters, completing the pursuit of the target, meeting the expected effect.

Claims

1. A multi-target UAV dynamic allocation and synchronous capture method, characterized in that: The following steps are involved: (1) Construct a multi-UAV kinematic model based on the UAV encirclement target; (2) Determine the target speed and update the target position based on the wind drift model AP98 and the target’s own dynamic velocity component; (3) Based on the UAV position and target position, the K-means clustering algorithm is used to assign tasks with the target position as the initial cluster center; (4) Using the PI performance impact algorithm to determine the RPI removal performance impact matrix and the IPI inclusion performance impact matrix of the roundup mission, we can obtain the UAV task allocation matrix; (5) Perform conflict detection. If there is a conflict, adjust the task allocation. After resolving the conflict, update the UAV task allocation matrix, the RPI removal performance impact matrix, and the IPI inclusion performance impact matrix, and proceed to step (6). (6) Considering the task completion time, task execution cost and task completion benefit, construct the objective function, determine the constraints, and optimize the task allocation; (7) Construct a multi-objective information matrix based on the kinematic model of multiple UAVs. The multi-objective information matrix includes the state information matrix and the state information derivative matrix of the UAVs relative to the targets. Auxiliary variables are introduced for each target to coordinate the multi-objective parameters. The virtual control variables are determined based on the derivatives of the auxiliary variables to obtain the actual control signals of the UAVs. (8) Use proportional guidance method to dynamically adjust the acceleration direction of the UAV, optimize the UAV path, and perform the capture mission; (9) Determine the capture effect. If the target escapes or the distance between the drone and the target continues to increase, mark the corresponding task as an unassigned task and return to step (4) to reassign the drone task. In step (7), auxiliary variables are introduced for each target to coordinate multi-target parameters. and as follows: ； ； ； in, For drones i To the target j The distance V is the Euclidean norm of the UAV velocity vector, is the angle between the UAV’s velocity vector and the line of sight LOS, is the deviation angle between the UAV velocity vector and the sighting line in the vertical plane, is the deviation angle between the UAV velocity vector and the sighting line in the horizontal plane; The dummy control variable is ； ； ； The actual control signal of the drone is ； ； in, and Represent the control signals in the horizontal and vertical directions respectively, and is the first-order derivative of the auxiliary variable; In step (8), the proportional guidance method includes a first guidance stage and a second guidance stage. In the first guidance stage, the auxiliary variable and Parameter preprocessing is performed to make the distance parameters and angular velocity parameters of the UAV consistent; in the second guidance stage, the acceleration is corrected by introducing the angle position and the acceleration direction of the UAV is dynamically adjusted. The acceleration direction of the UAV is as follows: ； ； in, is the navigation constant for target j, and Represent the line of sight rotation velocity components in the horizontal and vertical directions respectively.

2. The multi-target UAV dynamic allocation and synchronous capture method according to claim 1 is characterized in that: In step (2), the target is determined based on the wind drift model AP98 and the target's own dynamic velocity component. j speed, target j exist x Axis speed for ； Target j exist y Axis speed for ； in, and Respectively represent the target j exist x Axis and y The velocity component of the own power on the shaft, is the downwind drift vector, is the angle between the wind drift direction and the reference coordinate axis, is the vertical wind pressure vector; In step (2), the target j Updated location for ； ； In the formula, For the moment t Target j The location coordinates of .

3. The multi-target UAV dynamic allocation and synchronous capture method according to claim 2 is characterized in that: Step (3) is as follows: Defining the drone location dataset and target location dataset , where N is the total number of drones, M is the total number of targets, and ; The K-means clustering algorithm is used to calculate the distance between the drone and the target, and the target position is used as the initial cluster center for clustering and task allocation. The number of drones corresponding to each task is q ; judge q Is it satisfied? , if the number of drones corresponding to a certain task q Greater than , then calculate the average center of all UAVs in the task, and reallocate the UAV farthest from the average center to the task corresponding to its nearest target.

4. The multi-target UAV dynamic allocation and synchronous capture method according to claim 3 is characterized in that: In step (5), there is a conflict, and the task allocation adjustment is as follows: If the number of drones corresponding to the task q Greater than , then calculate the average center of all drones in the mission and determine the drone farthest from the average center ; Calculate the roundup status indicators for other tasks SI , the drone Assigned to the minimum roundup status indicator minSI in the corresponding task.

5. The multi-target UAV dynamic allocation and synchronous capture method according to claim 4 is characterized in that: Before entering step (7), consider the task completion time, task execution cost and task completion benefit to construct the objective function, set constraints and solve the optimal task allocation plan.

6. The multi-target UAV dynamic allocation and synchronous capture method according to claim 4 is characterized in that: Construct the objective function by considering the task completion time, task execution cost and task completion benefit J for ； in, , , is the weight coefficient, ; For drones i Complete the task j Time required, For the drone mission list, The cost required for UAV i to complete task j, For drones i Complete the task j The income obtained.

7. The multi-target UAV dynamic allocation and synchronous capture method according to claim 1 is characterized in that: The consistency algorithm of the first guidance phase in step (8) is as follows: ； ； ； in, and Respectively represent drones i Relative to target j The state information and its derivatives, is a control signal used to achieve consistency of state variables between drones. Indicates i The drone and k The drones can exchange information about the target j information; Indicates i The drone and k The drones cannot exchange information about the target j Information, and is the gain coefficient, .

Images