Construction of comprehensive task complexity index system of UAV cluster and dynamic evaluation method

Abstract

The application relates to the technical field of intelligent unmanned aerial vehicle manufacturing, and discloses a method for constructing and dynamically evaluating an unmanned aerial vehicle cluster comprehensive task complexity index system, which comprises the following steps: a target state parameter and a task-equipment matching function are proposed and improved; the unmanned aerial vehicle cluster task allocation is converted into a multi-target optimization problem; an improved NSGA-II algorithm is used for task allocation under the condition of multi-target optimization; an index system for evaluating task complexity is established by comprehensively considering the index composition of the task target dimension and the task allocation dimension; and multi-dimensional task complexity comprehensive evaluation is carried out. According to the application, personnel can quickly and accurately understand the change of a task situation and make relatively correct judgments accordingly. The application considers the dynamic change of a target state and a task, and can provide guidance for unmanned aerial vehicle task execution.

Description

Technical Field

[0001] This invention relates to the field of intelligent unmanned aerial vehicle manufacturing technology, specifically to intelligent unmanned system mission planning, and proposes a method for constructing and dynamically evaluating a comprehensive mission complexity index system for unmanned aerial vehicle swarms. Background Technology

[0002] Drone swarms are widely used in various mission scenarios. When performing various tasks, factors such as task allocation, redistribution, and even mission risks need to be considered, all of which place higher demands on drone performance. Faced with complex and ever-changing environmental situations, it is necessary to accurately determine whether a drone swarm can perform a mission. This requires modeling and quantifying dynamically changing tasks and establishing a corresponding indicator system.

[0003] Current research largely relies on static or semi-dynamic environment modeling, which fails to effectively address real-time changes such as target movement and new threats. First, traditional task allocation models (such as TSP and VRP) suffer from high computational complexity in dynamic environments, making it difficult to adjust task allocation strategies in real time, leading to decreased task execution efficiency and limitations in multi-task dynamic scenarios. When the number of UAVs exceeds four, communication interference and flight path overlap significantly increase the collision rate in dynamic environments. Second, research on the quantitative relationship between task type and equipment performance is also limited; existing task allocation models do not fully consider the matching relationship between UAV heterogeneity and task type. Task allocation for heterogeneous UAV swarms needs to consider differences in payload capacity and sensor performance, but existing models such as MILP and CMTAP struggle to quantify these coupling relationships. Furthermore, reassignment algorithms for sudden events such as equipment failure and threat changes have insufficient response speed. Traditional heuristic algorithms such as genetic algorithms and particle swarm optimization experience exponentially increasing computation time during dynamic reassignment, failing to meet real-time requirements. Finally, traditional indicator systems ignore the morphological characteristics of targets and the impact of equipment confrontation situations. Existing evaluation indicators mostly focus on single dimensions such as time and energy consumption, without integrating multiple factors such as environment and mission. There is a gap in the quantitative evaluation system that links multi-dimensional mission models with dynamic risks. Summary of the Invention

[0004] To address the shortcomings of the existing technologies, this invention proposes a method for constructing and dynamically evaluating a comprehensive task complexity index system for unmanned aerial vehicle (UAV) swarms, comprising the following steps:

[0005] Step 1: Determine the target state parameters of the target point, which include the spatial distribution density of the target point, the shape of the target point, and the dynamic change frequency of the target point;

[0006] Step 2: Construct a matching function between the task type matrix and equipment performance parameters, and calculate the task-equipment matching degree to verify whether the equipment used is suitable for performing the target task.

[0007] Step 3: Transform the drone swarm task allocation into a multi-objective optimization problem. Take the target point target state parameters as the task objective and the task-equipment matching degree as the constraint of the multi-objective optimization problem. Use the improved NSGA-II algorithm to perform task allocation under the multi-objective optimization condition.

[0008] Step 4: Based on the indicators of task objectives and task allocation, establish a task complexity assessment indicator system and conduct a comprehensive assessment of multi-dimensional task complexity.

[0009] Furthermore, in step one, determining the target point's state parameters includes:

[0010] Construct the Voronoi diagram of the target point and calculate the area A of each Thiessen polygon. i Then, calculate the ratio of the standard deviation to the mean of the area of ​​each Thiessen polygon to obtain the spatial distribution density of the target point. Where, σ A μ represents the area standard deviation. A The mean;

[0011] The shape of a target point is represented by its target contour. The target contour is then parametrically encoded, and its shape characteristics are... N is the total number of contour points, z k The complex coordinates of the k-th contour point are represented by n, where n is the order of the Fourier coefficients.

[0012] Collect the target point's moving trajectory x(t), and calculate the maximum Lyapunov exponent λ:

[0013]

[0014] When λ≤0, the dynamic change frequency C of the target point env2 =0.2λ norm When λ > 0, the dynamic change frequency C of the target point env2 = 0.5 + 0.5·sigmoid(λ).

[0015] Furthermore, in step two, the task type matrix includes reconnaissance task M1, throwing task M2, and evaluation task M3. The equipment performance parameters for reconnaissance task M1 include detection range D, identification accuracy P, and the number of targets tracked simultaneously R. t The equipment performance parameters for the M2 throwing mission include throwing height S, accuracy H, number of throws F, and the number of targets simultaneously guided for throwing R. b The equipment performance parameters for evaluating mission M3 include communication bandwidth and data processing speed; the matching function for constructing the mission type matrix and equipment performance parameters includes:

[0016] Quantify task requirements;

[0017] The equipment performance parameters and mission requirements are normalized, including the forward equipment performance parameters. The normalization formula is:

[0018]

[0019] negative equipment performance parameters The normalization formula is:

[0020]

[0021] In the formula, Let j be the value of the performance parameter of equipment i. These are the maximum and minimum values ​​of the j-th performance parameter in the cluster, respectively.

[0022] Calculate the task-equipment matching degree based on the task type. Where, ω j Let be the weight of the j-th performance parameter.

[0023] Furthermore, step three, which transforms the drone swarm task allocation into a multi-objective optimization problem, includes:

[0024] The task allocation is modeled mathematically, and the decision variables are defined as the allocation matrix. In the formula, x ij ∈{0,1}, indicating whether drone i performs task j, M a N represents the total number of drones. m Total number of tasks;

[0025] The objective function for optimizing task allocation is:

[0026]

[0027] In the formula, f1 represents maximizing task benefits; f2 represents minimizing overall risk; f3 represents minimizing energy consumption costs; R j It is the basic benefit of task j; η ij ∈[0,1], representing the matching degree of drone i performing task j; T j Threat level of the mission; ρ ij ∈[0,1], representing the resilience of drone i to task j; E ij This represents the energy consumption of drone i performing task j;

[0028] The constraints include:

[0029] A maximum of three drones can be assigned to each mission:

[0030] Mission payload L j The carrying capacity C of the drone i shall not exceed i :

[0031] The time d of task j j Must be within the time window:

[0032] The communication topology graph G must maintain strong connectivity: algebraic connectivity λ2(G) ≥ 0.5;

[0033] Quest - Equipment Matching M ms ≥0.6.

[0034] Furthermore, the task allocation under multi-objective optimization conditions using the improved NSGA-II algorithm described in step three includes:

[0035] For M a 100 drones, totaling N m Population initialization is performed on task scenario data. Two-dimensional matrix encoding is used between drones and tasks. Each individual represents an allocation scheme X. Individuals that do not meet the constraints are repaired.

[0036] Perform a non-dominated sort, define the dominance relation, and find that X1 dominates X2 if and only if:

[0037]

[0038] The population is divided into multiple frontier levels by rapid non-dominated sorting;

[0039] Crowding is calculated, and the crowding distance is calculated for individuals within the same frontal layer to maintain solution set diversity.

[0040]

[0041] In the formula, f k (i) represents the k-th target value for individual i; These represent the maximum and minimum values ​​of the k-th target in the current frontier layer.

[0042] Randomly select K individuals from the population, retain the best one, and then select the parent individuals of the selected individuals. Inter-partitioning by task type;

[0043] Perform mutation operations.

[0044] Furthermore, step three also includes redistributing tasks according to the conditions for triggering redistribution. The conditions for triggering redistribution include hard threshold triggering and soft threshold triggering. Hard threshold triggering is a forced redistribution that calls on backup drones to fill the task gaps. Soft threshold triggering is a flexible redistribution that dynamically adjusts task priorities and drone formations.

[0045] Furthermore, the task complexity evaluation index system described in step four includes a target layer, a criterion layer, and an index layer. The target layer is the comprehensive task complexity, and the criterion layer is divided into the task target dimension and the task allocation dimension. The indicators of the task target dimension include the spatial distribution density of target points and the dynamic change frequency of target points. The indicators of the task allocation dimension include cross-domain task switching cost, resource conflict degree, and collaborative constraint strength.

[0046] Furthermore, step four, which involves performing a multi-dimensional task complexity assessment, includes:

[0047] The range method is used to normalize the indices of the task objective dimension and the task allocation dimension to the interval [0, 1].

[0048]

[0049] In the formula, C′ is the normalized value; C is the current index value; C min and C max These are the minimum and maximum values ​​of the normalized index, respectively;

[0050] Design membership functions for each indicator in the task objective dimension, and calculate the membership degree of the task objective dimension using the trapezoidal membership function:

[0051]

[0052] Design membership functions for each indicator in the task allocation dimension, and calculate the membership degree of the task allocation dimension using the Gaussian membership function:

[0053]

[0054] In the formula, μ is the central value of each indicator in the task allocation dimension, and σ is the standard deviation;

[0055] The centroid method is used for deblurring, converting the fuzzy output into precise values. The overall task complexity is then calculated.

[0056]

[0057] In the formula, μ i Membership degree for the task objective dimension and the task assignment dimension; ω i Assign weights to the membership of the target dimension and the task dimension.

[0058] This invention can properly handle comprehensive evaluation models that combine multiple indicators and require quantitative and qualitative analysis, and takes into account the dynamic changes of target status and tasks, thus providing guidance for UAV mission execution. Attached Figure Description

[0059] Figure 1 This is a flowchart of the present invention;

[0060] Figure 2 This is a structural diagram of the comprehensive task complexity evaluation index system. Detailed Implementation

[0061] The present invention will be described in detail below with reference to the accompanying drawings and preferred embodiments. The purpose and effects of the present invention will become clearer. It should be understood that the embodiments are merely used to explain the present invention and are not intended to limit the present invention.

[0062] The present invention specifically includes the following steps:

[0063] Step 1: Determine the target point state parameters, including the spatial distribution density of the target points, the shape of the target points, and the frequency of dynamic changes of the target points.

[0064] The spatial distribution density of target points is obtained through Vorono i The target region is divided and calculated; the shape of the target point is represented by the target contour, so Fourier descriptors are used to parameterize the target point contour; the dynamic change frequency of the target point is obtained by dynamically predicting the target movement trajectory based on the Markov chain model.

[0065] The spatial distribution density of the target points is used to quantify the degree of concentration of target points in the spatial distribution within the task area. The target point distribution density is calculated as the standard deviation of the Thiessen polygon area divided by the average Thiessen polygon area. By constructing a Voronoi diagram of the target points, the area A of each Thiessen polygon is first calculated. i Then calculate the ratio of the standard deviation to the mean of the area of ​​each Thiessen polygon. It reflects the non-uniformity of distribution, where σ A The area standard deviation; μ A The mean is C; env1 A value greater than 0.3 indicates high space complexity.

[0066] The target point contour is parametrically encoded to generate a shape complexity index and contour shape features. In the formula, N is the total number of contour points, z k The complex coordinates of the k-th contour point are represented by n, where n is the order of the Fourier coefficients.

[0067] The frequency of dynamic change of the target point is calculated by quantifying the degree of chaos in the dynamic system through the calculation of the maximum Lyapunov exponent (MLE) of the target point's trajectory. When the MLE is positive, the system is sensitive to initial conditions and exhibits drastic dynamic changes.

[0068] Collect the target's movement trajectory x(t), and calculate the maximum Lyapunov exponent λ:

[0069]

[0070] When λ≤0, the target moves in a stable state, and the dynamic change frequency C of the target point is... env2 =0.2λ norm When λ > 0, the target's motion changes rapidly, and the dynamic change frequency C of the target point is high. env2 = 0.5 + 0.5·sigmoid(λ).

[0071] Step 2: Construct a matching function between the task type matrix and equipment performance parameters, and calculate the task-equipment matching degree to verify whether the equipment used is suitable for performing the target task.

[0072] In one specific embodiment of the present invention, a swarm of drones performs firefighting missions, including fire reconnaissance, water bomb extinguishing, and firefighting effectiveness assessment. The mission type matrix includes reconnaissance, water bomb extinguishing, and assessment missions. The equipment performance parameters for the reconnaissance mission include detection range D, identification accuracy h, and the number of targets simultaneously tracked R. t The equipment performance parameters for throwing missions include throwing height S, accuracy H, number of throws F, and the number of targets simultaneously guided for throwing R. b The equipment performance parameters for evaluating the mission include communication bandwidth and data processing speed.

[0073] To establish a mission-equipment matching model, the mission requirements must first be quantified. For example, M is a reconnaissance mission, with a detection range of ≥0.5km and a target number of ≥2 targets to be tracked; M2 is a throwing mission, with a water bullet launch range of ≥0.5km and a hit accuracy of ≥90%; M3 is an assessment mission, with a communication bandwidth of ≥20Mbps and a processing speed of ≥1GHz.

[0074] To eliminate dimensional differences, the equipment performance parameters and mission requirements are normalized:

[0075] Forward equipment performance parameters such as sensor resolution The normalization formula is:

[0076]

[0077] Negative equipment performance parameters such as communication delay The normalization formula is:

[0078]

[0079] In the formula, This characterizes the j-th performance parameter value of equipment i. These represent the maximum and minimum values ​​of the j-th performance parameter in the cluster, respectively.

[0080] Calculate the equipment's compatibility with the corresponding task based on the task type: Where, ω j Let be the weight of the j-th performance parameter.

[0081] Step 3: Transform the drone swarm task allocation into a multi-objective optimization problem. Use the target point state parameters calculated in the previous step as the task objective and the task-equipment matching degree as the constraint of the multi-objective optimization problem. Use the improved NSGA-II algorithm to perform task allocation under the multi-objective optimization condition.

[0082] In a specific embodiment of the present invention, step three includes:

[0083] Step 3.1: Transform the drone swarm task allocation into a multi-objective optimization problem.

[0084] The task allocation is modeled mathematically, and the decision variables are defined as the allocation matrix.

[0085] In the formula, x ij ∈{0,1}, indicating whether drone i performs task j, M a N represents the total number of drones. m This represents the total number of tasks.

[0086] The objective function for task allocation optimization is:

[0087]

[0088] In the formula, f1 represents maximizing task benefits; f2 represents minimizing overall risk; f3 represents minimizing energy consumption costs; R j It is the basic benefit of task j; η ij ∈[0,1], representing the matching degree of drone i performing task j; T j Threat level of the mission; ρ ij ∈[0,1], representing the resilience of UAV i to mission j, which is related to equipment performance; E ij This represents the energy consumption of UAV i performing mission j, which is related to range and payload.

[0089] The constraints include:

[0090] A maximum of three drones can be assigned to each mission:

[0091] Mission payload L j The carrying capacity C of the drone i shall not exceed i :

[0092] The time d of task j j Must be within the time window:

[0093] The communication topology graph G must maintain strong connectivity: algebraic connectivity λ2(G) ≥ 0.5;

[0094] Quest - Equipment Matching M ms ≥0.6.

[0095] Step 3.2: Perform task allocation under multi-objective optimization conditions using the improved NSGA-II algorithm.

[0096] After completing the problem modeling, the improved NSGA-II algorithm is used to solve the above multi-objective optimization problem. The algorithm flow is as follows:

[0097] First, for M a 100 drones, totaling N m Population initialization is performed on task scenario data. A two-dimensional matrix encoding is used to encode the relationship between drones and tasks, with each individual representing an allocation scheme X, and individuals that do not meet the constraints are repaired.

[0098] Secondly, perform a non-dominated sort, define the dominance relation, and solve for X1 dominating X2 if and only if:

[0099]

[0100] The population is divided into multiple front levels (Front 1 is the optimal solution set) by fast non-dominated sorting.

[0101] Next, crowding is calculated, and the crowding distance is calculated for individuals within the same frontier layer to maintain solution set diversity.

[0102]

[0103] In the formula, f k (i) represents the k-th target value for individual i; These represent the maximum and minimum values ​​of the k-th target in the current frontier layer.

[0104] Fourth, randomly select K individuals from the population, retain the best (K=2), and then... Partitioning by task type and cross-partitioning.

[0105] Finally, perform the mutation operation.

[0106] The mutation operation is divided into single-point mutation and block mutation. Single-point mutation is performed with probability p. m =0.1 Flip a single gene locus; Block mutations affect high-risk task columns (T) j (>0.6) Perform a complete rearrangement to improve responsiveness to threats. Implement an elite retention strategy, merging parent and offspring populations to retain the top N best individuals, ensuring convergence.

[0107] Step 3.3: Reassign tasks.

[0108] Because task requirements can change at any time, it is also necessary to be able to adjust task deployment and reallocate tasks in a timely manner. In a specific embodiment of the present invention, the conditions for triggering reallocation are determined based on real-time monitoring of task status and threat changes, and are divided into hard threshold triggering and soft threshold triggering.

[0109] Hard threshold triggering forces a reallocation, prioritizing the use of backup drones to fill mission gaps:

[0110]

[0111] In the formula, the planned progress is the ratio of the current time to the total task time. If the deviation rate is greater than 15%, the task is considered to be behind schedule.

[0112] Soft threshold triggering enables elastic reallocation, dynamically adjusting task priorities and drone formations:

[0113]

[0114] In the formula, T h The average threat level over the past 5 minutes; T c This represents the current instantaneous threat level. If the threat increase is greater than 20% and lasts for more than 30 seconds, it is considered a sudden risk.

[0115] The objective function for task redistribution is:

[0116] F = α·f1 - β·f2 - γ·f3

[0117] In the formula, f1 is the task benefit; f2 is the comprehensive risk; f3 is the path conflict cost; α, β, and γ are weighting coefficients.

[0118] After determining the threshold for triggering task redistribution, rapid task redistribution can be achieved based on the parallel particle swarm optimization algorithm. The velocity update formula of the parallel particle swarm optimization algorithm is used to adjust the movement direction of particles (candidate solutions) in real time, combining individual experience with group collaboration to adapt to dynamic changes in the task. The velocity update formula is:

[0119]

[0120] In the formula, c1 and c2 represent the step size of the adjustment particle moving towards its historical best position and the step size of the adjustment particle moving towards the global best position, respectively; r1 and r2 represent random numbers uniformly distributed in the interval [0, 1]; p best G represents the individual's historical best position. best Indicates the globally optimal position. Indicates the current speed. Indicates the current location.

[0121] The location update formula is used to generate new allocation schemes, shortening response time through parallel computing. The location update formula is:

[0122]

[0123] In the formula, Indicates the updated speed. This indicates the updated position.

[0124] The iteration terminates when the optimal fitness changes by less than 1% for ten consecutive times and the response time does not exceed 5% of the remaining task time, thus obtaining the optimal solution.

[0125] Step 4: Based on the indicators of task objectives and task allocation, establish a task complexity assessment indicator system and conduct a comprehensive assessment of multi-dimensional task complexity.

[0126] In a specific embodiment of the present invention, step four includes:

[0127] Step 4.1: Establish a task complexity evaluation index system.

[0128] The task complexity assessment index system comprises an objective layer, a criterion layer, and an indicator layer. The objective layer represents the overall task complexity, while the criterion layer is divided into task objective dimension and task allocation dimension. Indicators for the task objective dimension include the spatial distribution density of target points and the frequency of dynamic changes in target points. The task allocation dimension reflects the difficulty of resource coordination when UAV swarms execute tasks, including cross-domain task switching costs, resource conflict degree, and the strength of coordination constraints.

[0129] Among them, the cross-domain task switching cost C task1 This is used to measure the difference in energy consumption when switching between different types of tasks such as reconnaissance, throwing, and assessment. The higher the switching cost, the greater the difficulty in coordinating resources for task allocation. In the objective function of step 3.1 above, the energy consumption term... Cross-domain task switching cost

[0130] In the formula, ΔE k The energy difference for switching from task type k to the next task; E base The baseline energy consumption is the power consumption during hovering; αK The weights for mission types are 0.3, 0.5, and 0.2, respectively, for reconnaissance, throwing, and assessment.

[0131] Resource conflict level C task2 To quantify the degree of competition for resources such as UAV payload and communication by the mission, the constraints in step 3.1 above require that the mission payload L... j The carrying capacity C of the drone i shall not exceed i ,Right now Therefore, resource conflict degree The larger the ratio, the more intense the competition for resources and the higher the complexity of the task.

[0132] Cooperative constraint strength C task3 To reflect the ability of a drone swarm to maintain coordinated action, the constraints in step 3.1 require that the communication topology graph G maintain strong connectivity, i.e., the algebraic connectivity λ2(G) ≥ 0.5. Therefore, the strength of the coordination constraint is quantified by the algebraic connectivity of the communication topology graph. The smaller the value, the weaker the network connectivity and the higher the difficulty of task coordination. Thus, a Laplacian matrix L of the communication topology graph G is constructed, and the second smallest eigenvalue λ2 of L is calculated. The coordination constraint strength C is then... task3 =λ2, where λ2≥0.5 indicates a strong collaborative constraint.

[0133] Step 4.2: Design the membership function and solve for the overall task complexity.

[0134] The range method is used to normalize the indices of the task objective dimension and the task allocation dimension to the interval [0, 1].

[0135]

[0136] In the formula, C′ is the normalized value; C is the current index value; C min and C max These are the minimum and maximum values ​​of the normalized index, respectively.

[0137] The membership functions for the task objective dimension and task allocation dimension are designed, and the centroid method is used for defuzzification to calculate the comprehensive task complexity. The membership function is a core concept in fuzzy logic, used to describe the degree to which an element belongs to a certain fuzzy set (such as "high complexity", "medium complexity", "low complexity"), and its value range is [0, 1].

[0138] Design membership functions for each indicator of the task objective dimension to reflect whether the density of the objective distribution belongs to the "high complexity" interval. The spatial distribution density and dynamic change frequency of objective points have phased characteristics, which are suitable for using trapezoidal functions to divide clear intervals. Therefore, trapezoidal membership functions are used to calculate the membership values ​​of the task objective dimension.

[0139]

[0140] In the formula, μ env (C′) is the membership degree of the task objective dimension calculated by the trapezoidal membership function based on the index C′ after normalization of the task objective dimension.

[0141] The membership functions for each indicator in the task allocation dimension are designed to reflect whether the task coordination difficulty is close to the typical state of "medium complexity". The energy consumption difference and communication topology connectivity of cross-domain task switching usually show a distribution characteristic of high in the middle and low at both ends. Therefore, Gaussian membership functions are used to calculate the membership values ​​of the task allocation dimension:

[0142]

[0143] In the formula, μ represents the central value of each indicator in the task allocation dimension, σ represents the standard deviation, and μ task (C′) is the membership degree of the task allocation dimension calculated by the Gaussian membership function based on the index C′ after normalization of the task allocation dimension.

[0144] Finally, the centroid method is used for deblurring, converting the fuzzy output into precise values, and the overall task complexity is calculated:

[0145]

[0146] In the formula, μ i Membership degree for the task objective dimension and the task assignment dimension; ω i Assign weights to the membership of the target dimension and the task dimension.

[0147] This invention first proposes mission objective state parameters and a mission-equipment matching degree function, and performs dynamic risk assessment on the objective. The mission objective state parameters are used as the mission objective, and the mission-equipment matching degree is used as a constraint in a multi-objective optimization problem, transforming UAV swarm mission allocation into a multi-objective optimization problem using an improved NSGA-II algorithm. Considering the abrupt changes in mission situations, the conditions for triggering mission reallocation are determined. Finally, an evaluation index system is established to assess the overall mission reliability of UAV swarms under the influence of both the mission objective dimension and the mission allocation dimension, and the solution methods for each index in the index system are provided.

[0148] This invention can properly handle comprehensive evaluation models that combine multiple indicators and require quantitative and qualitative analysis, and takes into account the dynamic changes of target state and task, providing guidance for UAV mission execution.

[0149] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for constructing and dynamically evaluating a comprehensive task complexity index system for unmanned aerial vehicle (UAV) swarms, characterized in that, Includes the following steps: Step 1: Determine the target point's state parameters. These parameters include the target point's spatial distribution density, shape, and dynamic change frequency. The determination of the target point's state parameters includes: Construct target points The diagram shows the area of ​​each Thiessen polygon. Then, calculate the ratio of the standard deviation to the mean of the area of ​​each Thiessen polygon to obtain the spatial distribution density of the target point. ,in, The standard deviation of the area. The mean; The shape of a target point is represented by its target contour. The target contour is then parametrically encoded, and its shape features are... , This represents the total number of contour points. Indicates the first Complex coordinates of a contour point The order of the Fourier coefficients; Collect target point movement trajectory Calculate the maximum index : when At that time, the frequency of dynamic change of the target point ;when At that time, the frequency of dynamic change of the target point ; Step 2: Construct a matching function between the task type matrix and equipment performance parameters, and calculate the task-equipment matching degree to verify whether the equipment used is suitable for performing the target task. Step 3: Transform the drone swarm task allocation into a multi-objective optimization problem. Use the target point state parameters as the task objective and the task-equipment matching degree as the constraint of the multi-objective optimization problem, and then use an improved... The algorithm performs task allocation under multi-objective optimization conditions; Tasks are reassigned based on the triggering conditions, which include hard threshold triggering and soft threshold triggering. Hard threshold triggering is a forced reassignment that calls upon a backup drone to fill the task gap. The deviation rate for hard threshold triggering is: In the formula, the planned progress is the ratio of the current time to the total task time; if the deviation rate is >15%, the task is considered to be lagging behind. The soft threshold trigger is a flexible reallocation mechanism that dynamically adjusts task priorities and drone formations, resulting in the following threat amplification: Threat increase = In the formula, The average threat level over the past 5 minutes. This represents the current instantaneous threat level. If the threat increase is greater than 20% and lasts for more than 30 seconds, it is considered a sudden risk. The objective function for task redistribution is: In the formula, For task rewards; For comprehensive risks; This is the cost of path conflict; These are the weighting coefficients; Step 4: Based on the indicators of the task objective dimension and task allocation dimension, establish a task complexity evaluation indicator system. This system includes an objective layer, a criterion layer, and an indicator layer. The objective layer represents the overall task complexity. The criterion layer is divided into task objective dimension and task allocation dimension. Indicators for the task objective dimension include the spatial distribution density of target points and the dynamic change frequency of target points. Indicators for the task allocation dimension include cross-domain task switching cost, resource conflict degree, and collaborative constraint strength. The multi-dimensional task complexity evaluation includes: The range method is used to normalize the indices of the task objective dimension and the task allocation dimension to the [0,1] interval: In the formula, C' is the normalized value; C is the current index value; and These are the minimum and maximum values ​​of the normalized index, respectively. Design membership functions for each indicator in the task objective dimension, and calculate the membership degree of the task objective dimension using the trapezoidal membership function: ； Design membership functions for each indicator in the task allocation dimension, and calculate the membership degree of the task allocation dimension using the Gaussian membership function: ； In the formula, μ Assign central values ​​to each metric across the task dimensions. σ Standard deviation; The centroid method is used for deblurring, converting the fuzzy output into precise values. The overall task complexity is then calculated. In the formula, Membership degrees for the task objective dimension and the task assignment dimension; Assign weights to the membership of the target dimension and the task dimension.

2. The method according to claim 1, characterized in that, In step two, the task type matrix includes reconnaissance tasks. Throwing mission and evaluation tasks Among them, reconnaissance mission The equipment performance parameters include detection range Recognition accuracy and the number of targets tracked simultaneously Throwing mission The equipment performance parameters include throwing height. Accuracy Number of throws and the number of targets to be guided at the same time Assessment Task The equipment performance parameters include communication bandwidth and data processing speed; the matching function between the constructed task type matrix and the equipment performance parameters includes: Quantify task requirements; The equipment performance parameters and mission requirements are normalized, including the forward equipment performance parameters. The normalization formula is: ， negative equipment performance parameters The normalization formula is: ， In the formula, For equipment The Performance parameter values, , The first in the cluster The maximum and minimum values ​​of the performance parameters; Calculate the task-equipment matching degree based on the task type. ,in, For the first j The weight of each performance parameter.

3. The method according to claim 1, characterized in that, Step three, which transforms the drone swarm task allocation into a multi-objective optimization problem, includes: The task allocation is modeled mathematically, and the decision variables are defined as the allocation matrix. In the formula, , indicating drone Should the task be executed? , The total number of drones, Total number of tasks; The objective function for optimizing task allocation is: In the formula, This indicates maximizing task rewards; This indicates minimizing overall risk; This represents minimizing energy consumption costs; It is a task Basic returns; , indicating drone Execute the task The degree of matching; Threat level of the mission; , indicating drone For the task The ability to withstand risks; Indicates drone Execute the task Energy consumption; The constraints include: A maximum of three drones can be assigned to each mission: ; mission payload No more than drones load-bearing capacity : ; Task j Time consumption Must be within the time window: ; Communication topology diagram Strong connectivity must be maintained: Algebraic connectivity ; Quest-Equipment Matching .

4. The method according to claim 1, characterized in that, The improvement described in step three The algorithm performs task allocation under multi-objective optimization conditions, including: for A total of 100 drones The population is initialized using task scenario data. A two-dimensional matrix encoding is used to link the drones and the tasks, with each individual representing an allocation scheme. Repair individuals that do not meet the constraints; Perform a non-dominated sort, define the dominance relation, and solve... Dominate If and only if: The population is divided into multiple frontier levels by rapid non-dominated sorting; Crowding is calculated, and the crowding distance is calculated for individuals within the same frontal layer to maintain solution set diversity. In the formula, Represents an individual The One target value; , The first in the current frontier layer The maximum and minimum values ​​of each target; Randomly selected from the population For each individual, retain the best one, and then select the parent individual. Partitioning by task type; Perform mutation operations.

Images