Intelligent unmanned aerial vehicle swarm performance test and evaluation method

Abstract

The invention relates to an intelligent unmanned aerial vehicle swarm performance test and evaluation method, belongs to the technical field of unmanned aerial vehicle swarms, and solves the problem of insufficient real-time performance and adaptability of unmanned aerial vehicle swarm performance evaluation in the prior art. The method comprises the steps that cooperative operation data, environment disturbance data and task execution data are collected in real time through an airborne sensor; calculating a real-time collaboration index based on the collaboration operation data, calculating a real-time environment index based on the environment disturbance data, and calculating a real-time task index based on the task execution data; constructing a support vector regression model, constructing a training sample set based on historical task data, and optimizing kernel function parameters of the support vector regression model by using an improved particle swarm algorithm based on the training sample set to obtain an efficiency prediction model; inputting a real-time cooperation index, a real-time environment index and a real-time task index into the efficiency prediction model, and outputting a real-time prediction efficiency value; and determining the efficiency level of the unmanned aerial vehicle swarm based on the real-time prediction efficiency value and the dynamic efficiency threshold.

Description

Technical Field

[0001] The present invention relates to the field of drone swarm technology, and in particular to a method for testing and evaluating the effectiveness of an intelligent drone swarm. Background Art

[0002] With the rapid development of drone technology, drone swarm systems have been widely used in military reconnaissance, disaster relief, environmental monitoring and other fields. By working together, drone swarms can efficiently complete complex tasks, improving mission success rates and resource utilization.

[0003] Current methods for evaluating drone swarm effectiveness primarily rely on offline analysis of historical data, using statistical analysis and empirical models to assess swarm effectiveness. These methods typically analyze collaborative operation data, environmental disturbance data, and mission execution data separately, but lack the ability to comprehensively process and evaluate multi-dimensional data in real time.

[0004] In existing technologies, performance evaluation is often performed offline, failing to reflect the actual operational status of drone swarms in complex, dynamic environments in real time. This lag prevents evaluation results from providing timely guidance for swarm adjustment and optimization. Existing technologies fail to fully account for the performance differences of drone swarms at different mission stages and lack dynamic performance thresholds for each stage, making it impossible to accurately reflect swarm performance at different mission stages. Existing performance evaluation models also lack kernel function parameter optimization, making it difficult to effectively process complex, multi-dimensional input features. This results in low prediction accuracy and an inability to meet practical application requirements. Summary of the Invention

[0005] In view of the above analysis, an embodiment of the present invention aims to provide an intelligent UAV swarm effectiveness testing and evaluation method to solve the problems of insufficient real-time performance and accuracy of existing UAV swarm effectiveness evaluation.

[0006] The present invention provides a method for testing and evaluating the effectiveness of an intelligent unmanned aerial vehicle swarm, comprising the following steps:

[0007] The system collects collaborative operation data, environmental disturbance data, and mission execution data in real time through onboard sensors, and extracts historical mission data from the historical database. The historical mission data includes historical collaborative indicators, historical environmental indicators, historical mission indicators, and pre-labeled historical performance tags corresponding to historical missions.

[0008] Calculating a real-time collaborative index based on the collaborative operation data, calculating a real-time environmental index based on the environmental disturbance data, and calculating a real-time task index based on the task execution data;

[0009] A support vector regression model was constructed, and a training sample set was constructed based on historical task data. Based on the training sample set, an improved particle swarm algorithm was used to optimize the kernel function parameters of the support vector regression model to obtain an efficiency prediction model.

[0010] Inputting the real-time coordination index, the real-time environment index and the real-time task index into the performance prediction model and outputting the real-time predicted performance value;

[0011] The drone swarm efficiency level is determined based on the real-time predicted efficiency value and dynamic efficiency threshold.

[0012] As a further improvement of this application, a training sample set is constructed based on historical task data. Based on the training sample set, an improved particle swarm algorithm is used to optimize the kernel function parameters of the support vector regression model to obtain an efficiency prediction model, including:

[0013] A three-dimensional input feature vector is constructed based on historical collaboration indicators, historical environment indicators, and historical task indicators, and a training sample set is constructed based on the three-dimensional input feature vector and historical performance labels; the three-dimensional input feature vector includes a historical collaboration sub-vector, a historical environment sub-vector, and a historical task sub-vector;

[0014] Calculate the mean vector of all historical collaboration subvectors in the training sample set as the collaboration indicator support subvector, calculate the mean vector of all historical environment subvectors in the training sample set as the environment indicator support subvector, and calculate the mean vector of all historical task subvectors in the training sample set as the task indicator support subvector;

[0015] Construct the support vector of the kernel function based on the collaborative indicator support subvector, the environmental indicator support subvector and the task indicator support subvector;

[0016] Define the search space of the improved particle swarm algorithm, which is the value range of the kernel function parameters;

[0017] Based on the training sample set, iteratively updating the kernel function parameters of the support vector regression model by the improved particle swarm algorithm so as to minimize the mean square error between the predicted output of the support vector regression model and the historical performance label;

[0018] Generate an efficacy prediction model based on the updated kernel function parameters.

[0019] As a further improvement of the present application, the dynamic performance threshold includes a first performance threshold, a second performance threshold, and a third performance threshold;

[0020] Determining the drone swarm effectiveness level based on real-time predicted effectiveness value and dynamic effectiveness threshold includes:

[0021] Compare the real-time predicted performance value and the dynamic performance threshold when the drone swarm is in the reconnaissance phase, operation phase, and return phase, including:

[0022] When the drone swarm is in the reconnaissance phase, the real-time predicted performance value is compared with the first performance threshold to obtain a first performance result;

[0023] When the drone swarm is in the operation phase, the real-time predicted performance value is compared with the second performance threshold to obtain a second performance result;

[0024] When the drone swarm is in the return phase, the real-time predicted performance value is compared with the third performance threshold to obtain a third performance result;

[0025] The drone swarm performance level is determined based on the first performance result, the second performance result, and the third performance result.

[0026] As a further improvement of the present application, determining the drone swarm performance level based on the first performance result, the second performance result, and the third performance result includes:

[0027] If the first performance result, the second performance result, and the third performance result are all greater than the corresponding dynamic performance thresholds, the drone swarm performance level is determined to be the first performance level;

[0028] If two of the first performance result, the second performance result, and the third performance result are greater than the corresponding dynamic performance threshold, the drone swarm performance level is determined to be the second performance level;

[0029] If the first performance result, the second performance result, and the third performance result are all smaller than the corresponding dynamic performance thresholds, the drone swarm performance level is determined to be the third performance level.

[0030] As a further improvement of the present application, the collaborative operation data includes a communication adjacency matrix, actual formation position data, and drone positioning data; and the calculation of real-time collaborative indicators includes:

[0031] The communication stability coefficient is calculated by the algebraic connectivity of the communication adjacency matrix;

[0032] The Hausdorff distance between the actual formation position data and the ideal formation position is calculated to obtain the formation retention rate;

[0033] The root mean square error of the relative positions between drones is calculated based on the drone positioning data to determine the collaborative positioning accuracy;

[0034] The communication stability coefficient, formation retention rate and collaborative positioning accuracy are weightedly integrated to obtain the real-time collaborative index.

[0035] As a further improvement of the present application, the environmental disturbance data includes three-dimensional wind speed vector data, obstacle distribution grid data, and electromagnetic spectrum data; and calculating the real-time environmental index based on the environmental disturbance data includes:

[0036] Calculate the modulus of the gradient vector of the three-dimensional wind speed vector data to determine the wind speed disturbance intensity;

[0037] Calculate the density of each grid cell based on the obstacle distribution grid data;

[0038] Performing energy integration on the interference frequency band of electromagnetic spectrum data to obtain electromagnetic interference intensity;

[0039] The wind speed disturbance intensity, grid cell density and electromagnetic interference intensity are harmonically averaged to obtain the real-time environmental index.

[0040] As a further improvement of this application, the mission execution data includes the coordinates of the drone coverage area, the actual path length data and the mission completion time; the calculation of real-time mission indicators includes:

[0041] Determine the area of the drone coverage area according to the coordinates of the drone coverage area, calculate the intersection area of the drone coverage area and the target area, and calculate the target coverage rate based on the ratio of the intersection area to the target area;

[0042] Calculate the ratio of the theoretical optimal path length to the actual path length data to determine the path selection reliability;

[0043] The ratio of the total planned time to the task completion time is calculated to calculate the task completion efficiency;

[0044] Calculate real-time task indicators based on target coverage, path selection reliability, and task completion efficiency.

[0045] As a further improvement of the present application, the efficacy prediction model is shown in the following formula:

[0046]

[0047] Among them, K σ is the dynamic kernel function, b is the bias term, α, α * is a pair of Lagrange multipliers, x is a support vector, x i is the three-dimensional input feature vector of the i-th sample, and N is the total number of training samples.

[0048] As a further improvement of the present application, the dynamic kernel function is shown in the following formula:

[0049]

[0050] Among them, x cis the synergy indicator support subvector, x e is the environmental indicator support subvector, x t is the task indicator support subvector, is the synergy indicator subvector, is the environmental indicator subvector, is the task indicator subvector, σ c is the kernel function parameter corresponding to the collaborative index sub-vector, σ e is the kernel function parameter corresponding to the environmental indicator subvector, σ t is the kernel function parameter corresponding to the collaborative indicator sub-vector.

[0051] As a further improvement of the present application, the objective function for iteratively updating the kernel function parameters by the improved particle swarm algorithm is:

[0052]

[0053] Among them, N is the total number of training samples, y n is the historical performance label corresponding to the nth training sample, f' n (x) is the predicted output of the support vector regression model.

[0054] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0055] 1. The present invention uses onboard sensors to collect collaborative operation data, environmental disturbance data, and mission execution data in real time, and combines historical mission data to train an efficiency prediction model. This allows for rapid response to dynamic changes in drone swarms in complex environments, solving the problem of delayed efficiency evaluation in existing technologies and achieving real-time efficiency evaluation.

[0056] 2. The present invention introduces a dynamic performance threshold based on the performance differences of drone swarms in different mission stages of reconnaissance, operation and return. By setting performance evaluation standards for the reconnaissance, operation and return stages respectively, it solves the problem in the existing technology that the performance of the swarm in different mission stages cannot be accurately reflected, making the performance evaluation results more in line with actual mission requirements and improving the accuracy of swarm drone performance evaluation.

[0057] In the present invention, the above-mentioned technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of the present invention will be described in the following description, and some advantages will become apparent from the description or be learned through practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The accompanying drawings are only used for the purpose of illustrating specific embodiments and are not to be considered as limiting the present invention. Throughout the drawings, the same reference symbols denote the same components.

[0059] Figure 1 A flowchart of a method for testing and evaluating the effectiveness of an intelligent drone swarm is provided in accordance with one embodiment of the present invention. DETAILED DESCRIPTION

[0060] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, and are not used to limit the scope of the present invention.

[0061] A specific embodiment of the present invention discloses a method for testing and evaluating the effectiveness of an intelligent drone swarm. Figure 1 A method for testing and evaluating the effectiveness of an intelligent drone swarm includes the following steps:

[0062] Step 101, collect collaborative operation data, environmental disturbance data and task execution data in real time through onboard sensors, and extract historical task data from the historical database. The historical task data includes historical collaborative indicators, historical environmental indicators and historical task indicators corresponding to historical tasks, as well as pre-labeled historical performance tags.

[0063] It should be noted that onboard sensors refer to various sensors installed on drones, used to obtain various data from drones in real time during flight. Collaborative operation data refers to data generated when drones collaborate with each other, such as communication status and formation information. Environmental disturbance data refers to data on external environmental factors affecting drones during flight, such as wind speed and electromagnetic interference. Mission execution data refers to data generated when drones complete missions, such as coverage area and mission completion time. Historical mission data is extracted from a historical database and includes historical collaboration indicators, historical environmental indicators, historical mission indicators corresponding to historical missions, as well as pre-labeled historical performance labels. Historical collaboration indicators, historical environmental indicators, and historical mission indicators are calculated based on historical mission data and are used to reflect the collaboration, environment, and mission execution of the drone swarm in historical missions. Pre-labeled historical performance labels are performance level labels pre-labeled based on the completion of historical missions and are used for model training.

[0064] Step 102 : Calculate a real-time collaborative index based on the collaborative operation data, calculate a real-time environmental index based on the environmental disturbance data, and calculate a real-time task index based on the task execution data.

[0065] Specifically, collaborative operation data includes the communication adjacency matrix, actual formation position data, and drone positioning data. The communication adjacency matrix describes the communication connection status between drones, the actual formation position data reflects the actual position of the drones in the formation, and the drone positioning data provides the precise location of the drones.

[0066] Environmental disturbance data includes three-dimensional wind speed vector data, obstacle distribution grid data, and electromagnetic spectrum data. The three-dimensional wind speed vector data describes the magnitude and direction of wind speed, the obstacle distribution grid data represents the spatial distribution of obstacles, and the electromagnetic spectrum data reflects the interference of the electromagnetic environment.

[0067] Mission execution data includes the coordinates of the drone's coverage area, actual path length, and mission completion time. The coordinates of the drone's coverage area determine the drone's coverage range, the actual path length reflects the actual length of the drone's flight path, and the mission completion time is the time it takes the drone to complete the mission.

[0068] Furthermore, the calculation of real-time collaboration indicators includes:

[0069] Step 10211, calculating the communication stability coefficient by the algebraic connectivity of the communication adjacency matrix.

[0070] The communication adjacency matrix is a two-dimensional matrix whose elements indicate whether a communication connection exists between drones. A value of 1 in the matrix indicates normal communication between the two drones, while a value of 0 indicates a communication interruption. The connectivity of each drone is calculated using the communication adjacency matrix. The average connectivity of the entire swarm is then calculated to serve as the communication stability coefficient.

[0071] Step 10212, calculate the Hausdorff distance between the actual formation position data and the ideal formation position to obtain the formation maintenance rate. The actual formation position data records the actual position coordinates of the UAV in the formation and can be expressed in the form of two-dimensional or three-dimensional coordinates. The ideal formation position refers to the position where the UAV should be in the formation. The formation maintenance rate is calculated by calculating the Hausdorff distance between the actual formation position data and the ideal formation position. The smaller the Hausdorff distance, the higher the formation maintenance rate. Specifically, the actual formation position data and the ideal formation position can be represented as point sets respectively, and the formation maintenance status can be quantified by calculating the Hausdorff distance between the two point sets.

[0072] Step 10213: Calculate the root mean square error (RMS) of the relative positions between UAVs based on the UAV positioning data to determine the collaborative positioning accuracy. The UAV positioning data can be used to determine the real-time position of the UAVs using satellite positioning or other positioning methods. The RMS error is calculated by calculating the RMS value of the deviation between the actual relative positions of all UAVs and their ideal relative positions, thereby determining the collaborative positioning accuracy.

[0073] In step 10214, the communication stability coefficient, formation retention rate, and collaborative positioning accuracy are weighted and fused to obtain a real-time collaborative index. During the weighted fusion process, the weights can be set based on mission requirements and experience.

[0074] Furthermore, the calculation of real-time environmental indicators includes:

[0075] Step 10221: Calculate the modulus of the gradient vector of the three-dimensional wind speed vector data to determine the wind speed disturbance intensity. Three-dimensional wind speed vector data refers to the wind speed magnitude and direction information of the environment in which the drone is located during flight, typically represented as a three-dimensional vector. The wind speed disturbance intensity is determined by calculating the modulus of the gradient vector of the three-dimensional wind speed vector data. The wind speed disturbance intensity reflects the severity of wind speed changes. When calculating the wind speed disturbance intensity, a time series analysis can be performed on the three-dimensional wind speed vector data to extract the time derivative of the wind speed change, i.e., the rate at which the wind speed changes over time. The modulus of the gradient vector of the time derivative is then calculated to quantify the intensity of the wind speed disturbance.

[0076] Step 10222: Calculate the density of each grid cell based on the obstacle distribution raster data. Obstacle distribution raster data refers to the spatial distribution of obstacles in the environment, as acquired by sensors. It is typically represented in a rasterized form, with each grid cell containing obstacle density information. The density of each grid cell is calculated based on the obstacle distribution raster data. For obstacle distribution raster data, different grid resolutions can be set based on task requirements. For example, a higher resolution grid may be used in complex terrain areas to more accurately reflect the distribution of obstacles.

[0077] Step 10223: Energy integration is performed on the interference frequency bands of the electromagnetic spectrum data to obtain the electromagnetic interference intensity. Electromagnetic spectrum data refers to the electromagnetic signal strength and frequency distribution in the drone's environment, reflecting the intensity and frequency band of the electromagnetic interference. Energy integration of the interference frequency bands of the electromagnetic spectrum data to obtain the electromagnetic interference intensity reflects the impact of the electromagnetic environment on the drone's communication and navigation systems. When processing electromagnetic spectrum data, electromagnetic interference in different frequency bands can be classified. For example, the interference frequency bands can be divided into low-frequency, medium-frequency, and high-frequency intervals, and the energy integrals of these intervals are calculated. The interference intensity of different frequency bands can then be assigned different weights based on mission requirements.

[0078] In step 10224, the wind speed disturbance intensity, grid cell density, and electromagnetic interference intensity are harmonized and averaged to obtain a real-time environmental index. By harmonizing and averaging the wind speed disturbance intensity, obstacle density, and electromagnetic interference intensity, a comprehensive real-time environmental index is obtained. This real-time environmental index can more comprehensively reflect the impact of the environment on the execution of the drone swarm mission.

[0079] Furthermore, the calculation of real-time task indicators includes:

[0080] Step 10231 determines the area of the drone's coverage area based on the drone's coverage area coordinates. The intersection of the drone's coverage area and the target area is calculated, and the target coverage rate is calculated based on the ratio of the intersection to the target area. The drone's coverage area coordinates refer to the geographic area covered by the drone during the mission, expressed in coordinate form. When calculating the target coverage rate, the drone's coverage area coordinates can be rasterized, dividing the target area into multiple small grid cells. The number of grid cells covered by the drone is then counted to calculate the target coverage rate.

[0081] Step 10232: Calculate the ratio of the theoretical optimal path length to the actual path length to determine path selection reliability. Path selection reliability measures the degree of closeness between the actual path and the theoretical optimal path. When calculating path selection reliability, a higher path selection reliability indicates that the actual path is closer to the theoretical optimal path and that the path selection is more reasonable.

[0082] Step 10233 calculates the mission completion efficiency by calculating the ratio of the planned total time to the mission completion time. Mission completion time refers to the total time it takes the UAV to actually complete the mission. Typically, the mission completion time is greater than the planned total time. When calculating mission completion efficiency, the planned total time can be dynamically adjusted based on the urgency and priority of the task. For high-priority tasks, the planned total time can be appropriately shortened. Mission completion time is typically greater than the planned total time. A higher mission completion efficiency value indicates a shorter mission completion time and higher mission execution efficiency.

[0083] Step 10234, calculate the real-time task index based on the target coverage, path selection reliability and task completion efficiency. The real-time task index can be obtained based on the weighted sum of the target coverage, path selection reliability and task completion efficiency.

[0084] Step 103 : construct a support vector regression model, construct a training sample set based on historical task data, and optimize the kernel function parameters of the support vector regression model using an improved particle swarm optimization algorithm based on the training sample set to obtain a performance prediction model.

[0085] When constructing a support vector regression model, first construct a training sample set based on historical task data. Specifically, a three-dimensional input feature vector is constructed based on historical coordination indicators, historical environmental indicators, and historical task indicators, and a training sample set is constructed based on the three-dimensional input feature vector and historical performance labels. Then, the search space of the improved particle swarm algorithm is defined, and the search space is the value range of the kernel function parameters. The kernel function parameters of the support vector regression model are iteratively updated by the improved particle swarm algorithm to minimize the mean square error between the predicted output of the support vector regression model and the historical performance labels. Finally, a performance prediction model is generated based on the updated kernel function parameters. Specifically, the predicted output of the model is calculated according to the current parameters at each iteration, and compared with the historical performance labels to calculate the mean square error. When the mean square error is minimized or meets the preset number of iterations, the iteration is stopped to obtain the optimized kernel function parameters.

[0086] Furthermore, a training sample set is constructed based on historical task data. Based on the training sample set, an improved particle swarm optimization algorithm is used to optimize the kernel function parameters of the support vector regression model to obtain an efficiency prediction model, including:

[0087] Step 1031: construct a three-dimensional input feature vector based on historical collaboration indicators, historical environment indicators, and historical task indicators, and construct a training sample set based on the three-dimensional input feature vector and historical performance labels; the three-dimensional input feature vector includes a historical collaboration sub-vector, a historical environment sub-vector, and a historical task sub-vector.

[0088] Historical mission data is data accumulated during previous drone swarm missions, including historical collaboration, environmental, and mission metrics, as well as corresponding pre-labeled historical performance tags. An improved particle swarm optimization algorithm is used to optimize the kernel function parameters in the support vector regression model to improve its predictive performance. Kernel function parameters determine the model's complexity and fit. The three-dimensional input feature vector consists of a historical collaboration subvector, a historical environmental subvector, and a historical mission subvector, corresponding to collaboration, environmental, and mission characteristics, respectively.

[0089] Step 1032: Calculate the mean vector of all historical collaboration sub-vectors in the training sample set as the collaboration indicator support sub-vector, calculate the mean vector of all historical environment sub-vectors in the training sample set as the environment indicator support sub-vector, and calculate the mean vector of all historical task sub-vectors in the training sample set as the task indicator support sub-vector.

[0090] Step 1033 : constructing a support vector of the kernel function based on the collaboration indicator support subvector, the environment indicator support subvector, and the task indicator support subvector.

[0091] Step 1034 , defining the search space of the improved particle swarm optimization algorithm, where the search space is the value range of the kernel function parameters.

[0092] Step 1035 : Based on the training sample set, iteratively update the kernel function parameters of the support vector regression model through the improved particle swarm optimization algorithm to minimize the mean square error between the predicted output of the support vector regression model and the historical performance label.

[0093] The kernel function parameters are iteratively updated using an improved particle swarm optimization algorithm. Each iteration calculates the predicted output of the support vector regression model based on the current kernel function parameters and calculates the mean squared error between the predicted output and the historical performance labels. When the mean squared error is minimized or the preset number of iterations is met, the iteration stops, resulting in the optimized kernel function parameters and the generation of a performance prediction model.

[0094] Step 1036: Generate an effectiveness prediction model based on the updated kernel function parameters.

[0095] The efficacy prediction model is shown in the following formula:

[0096]

[0097] Among them, K σ is the dynamic kernel function, b is the bias term, α, α * is a pair of Lagrange multipliers, x is a support vector, x i is the three-dimensional input feature vector of the i-th sample, and N is the total number of training samples.

[0098] The dynamic kernel function is used to map input data into a high-dimensional space, thereby improving the model's fitting ability. It can also dynamically adjust kernel function parameters based on differences in input feature vectors to better adapt to different input conditions. Lagrange multiplier pairs are parameters introduced during the SVR optimization process to balance model complexity and fitting error.

[0099] The dynamic kernel function is shown in the following formula:

[0100]

[0101] Among them, x c is the synergy indicator support subvector, x e is the environmental indicator support subvector, x t is the task indicator support subvector, is the collaborative index subvector of the i-th sample, is the environmental indicator subvector of the i-th sample, is the task indicator subvector of the i-th sample, σ c is the kernel function parameter corresponding to the collaborative index sub-vector, σe is the kernel function parameter corresponding to the environmental indicator subvector, σ t is the kernel function parameter corresponding to the collaborative indicator sub-vector.

[0102] The parameters of the dynamic kernel function include the kernel parameter σc corresponding to the synergy indicator subvector, the kernel parameter σe corresponding to the environmental indicator subvector, and the kernel parameter σt corresponding to the task indicator subvector. The kernel parameters σc, σe, and σt control the influence of the three subvectors in the kernel function. The dynamic kernel function dynamically adjusts its shape and width based on the specific characteristics of the input feature vector, thus better adapting to different input data.

[0103] The objective function of iteratively updating the kernel function parameters through the improved particle swarm algorithm is:

[0104]

[0105] Among them, N is the total number of training samples, y n is the historical performance label corresponding to the nth training sample, f' n (x) is the predicted output of the support vector regression model.

[0106] Step 104 : input the real-time collaboration index, the real-time environment index, and the real-time task index into the performance prediction model, and output a real-time predicted performance value.

[0107] Step 105 : Determine the drone swarm efficiency level based on the real-time predicted efficiency value and the dynamic efficiency threshold.

[0108] Dynamic performance thresholds are performance evaluation criteria adjusted based on the changes in the drone swarm's mission phases. These include the first, second, and third performance thresholds, corresponding to the reconnaissance, operation, and return phases, respectively. By comparing the real-time predicted performance values with these dynamic performance thresholds, we can obtain performance results for different phases and then comprehensively determine the overall performance level of the drone swarm. The performance level of a drone swarm determined based on the real-time predicted performance values and dynamic performance thresholds includes:

[0109] Step 1051 , respectively comparing the real-time predicted performance value and the dynamic performance threshold when the drone swarm is in the reconnaissance phase, the operation phase, and the return phase, includes:

[0110] Step 1052 , when the drone swarm is in the reconnaissance phase, compare the real-time predicted performance value with the first performance threshold to obtain a first performance result;

[0111] Step 1053 , when the drone swarm is in the operation phase, compare the real-time predicted performance value with the second performance threshold to obtain a second performance result;

[0112] Step 1054 , when the drone swarm is in the return phase, compare the real-time predicted performance value with the third performance threshold to obtain a third performance result;

[0113] Step 1055 , determining the drone swarm performance level based on the first performance result, the second performance result, and the third performance result.

[0114] Specifically, the first performance threshold is set as the performance evaluation standard for the drone swarm during the reconnaissance phase, used to measure whether the reconnaissance mission is completed as expected; the second performance threshold is set during the operational phase, used to evaluate the drone swarm's performance in executing its primary mission; and the third performance threshold is set during the return phase, used to determine the swarm's performance during the return process. The real-time predicted performance value is calculated using a performance prediction model and reflects the drone swarm's performance level at the current moment. During the reconnaissance phase, the real-time predicted performance value is compared with the first performance threshold to obtain the first performance result; during the operational phase, it is compared with the second performance threshold to obtain the second performance result; and during the return phase, it is compared with the third performance threshold to obtain the third performance result. These performance results are typically expressed in numerical form: a value greater than the threshold indicates satisfactory or excellent performance, while a value less than the threshold indicates substandard performance.

[0115] Dynamic performance thresholds can be adjusted based on the swarm's mission requirements and historical performance data. The first performance threshold can be set based on the complexity and importance of the reconnaissance mission, the second based on the mission's difficulty and objectives, and the third based on the safety and efficiency requirements of the return phase.

[0116] Furthermore, determining the drone swarm effectiveness level based on the first effectiveness result, the second effectiveness result, and the third effectiveness result includes:

[0117] If the first performance result, the second performance result, and the third performance result are all greater than the corresponding dynamic performance thresholds, the drone swarm performance level is determined to be the first performance level;

[0118] If two of the first performance result, the second performance result, and the third performance result are greater than the corresponding dynamic performance threshold, the drone swarm performance level is determined to be the second performance level;

[0119] If the first performance result, the second performance result, and the third performance result are all smaller than the corresponding dynamic performance thresholds, the drone swarm performance level is determined to be the third performance level.

[0120] If the performance results of the three stages are all greater than the corresponding dynamic performance threshold, the overall performance level of the drone swarm is the first performance level, indicating excellent overall performance; if the performance results of two stages are greater than the corresponding dynamic performance threshold, it is the second performance level, indicating good performance; if the performance results of the three stages are all less than the corresponding dynamic performance threshold, it is the third performance level, indicating that the overall performance does not meet the standard.

[0121] The above-mentioned embodiments of the present invention include at least the following technical effects: The present invention uses onboard sensors to collect collaborative operation data, environmental disturbance data, and mission execution data in real time, and combines historical mission data to train a performance prediction model. This allows for rapid response to the dynamic changes of drone swarms in complex environments, resolving the problem of delayed performance evaluation in the prior art and achieving real-time performance evaluation. The present invention introduces dynamic performance thresholds to address the performance differences of drone swarms during the different mission stages of reconnaissance, operation, and return. By setting performance evaluation standards for the reconnaissance, operation, and return stages separately, the present invention addresses the lack of dynamic adaptability in the prior art, making the performance evaluation results more aligned with actual mission requirements and improving applicability.

[0122] Those skilled in the art will appreciate that all or part of the process steps of the above-described embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, such as a magnetic disk, an optical disk, a read-only memory, or a random access memory.

[0123] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with this technical field within the technical scope disclosed by the present invention should be covered by the scope of protection of the present invention.

Claims

1. A method for testing and evaluating the effectiveness of intelligent drone swarms, characterized in that: The following steps are involved: The system collects collaborative operation data, environmental disturbance data, and mission execution data in real time through onboard sensors, and extracts historical mission data from the historical database. The historical mission data includes historical collaborative indicators, historical environmental indicators, historical mission indicators, and pre-labeled historical performance tags corresponding to historical missions. Calculating a real-time collaborative index based on the collaborative operation data, calculating a real-time environmental index based on the environmental disturbance data, and calculating a real-time task index based on the task execution data; A support vector regression model was constructed, and a training sample set was constructed based on historical task data. Based on the training sample set, an improved particle swarm algorithm was used to optimize the kernel function parameters of the support vector regression model to obtain an efficiency prediction model. Inputting the real-time coordination index, the real-time environment index and the real-time task index into the performance prediction model and outputting the real-time predicted performance value; The drone swarm efficiency level is determined based on the real-time predicted efficiency value and dynamic efficiency threshold.

2. The method according to claim 1, characterized in that A training sample set is constructed based on historical task data. Based on the training sample set, an improved particle swarm optimization algorithm is used to optimize the kernel function parameters of the support vector regression model to obtain an efficiency prediction model, including: A three-dimensional input feature vector is constructed based on historical collaboration indicators, historical environment indicators, and historical task indicators, and a training sample set is constructed based on the three-dimensional input feature vector and historical performance labels; the three-dimensional input feature vector includes a historical collaboration sub-vector, a historical environment sub-vector, and a historical task sub-vector; Calculate the mean vector of all historical collaboration subvectors in the training sample set as the collaboration indicator support subvector, calculate the mean vector of all historical environment subvectors in the training sample set as the environment indicator support subvector, and calculate the mean vector of all historical task subvectors in the training sample set as the task indicator support subvector; Construct the support vector of the kernel function based on the collaborative indicator support subvector, the environmental indicator support subvector and the task indicator support subvector; Define the search space of the improved particle swarm algorithm, which is the value range of the kernel function parameters; Based on the training sample set, iteratively updating the kernel function parameters of the support vector regression model by the improved particle swarm algorithm so as to minimize the mean square error between the predicted output of the support vector regression model and the historical performance label; Generate an efficacy prediction model based on the updated kernel function parameters.

3. The method according to claim 1, characterized in that The dynamic performance threshold includes a first performance threshold, a second performance threshold, and a third performance threshold; Determining the drone swarm effectiveness level based on real-time predicted effectiveness value and dynamic effectiveness threshold includes: Compare the real-time predicted performance value and the dynamic performance threshold when the drone swarm is in the reconnaissance phase, operation phase, and return phase, including: When the drone swarm is in the reconnaissance phase, the real-time predicted performance value is compared with the first performance threshold to obtain a first performance result; When the drone swarm is in the operation phase, the real-time predicted performance value is compared with the second performance threshold to obtain a second performance result; When the drone swarm is in the return phase, the real-time predicted performance value is compared with the third performance threshold to obtain a third performance result; The drone swarm performance level is determined based on the first performance result, the second performance result, and the third performance result.

4. The method according to claim 3, characterized in that Determining the drone swarm effectiveness level based on the first effectiveness result, the second effectiveness result, and the third effectiveness result includes: If the first performance result, the second performance result, and the third performance result are all greater than the corresponding dynamic performance thresholds, the drone swarm performance level is determined to be the first performance level; If two of the first performance result, the second performance result, and the third performance result are greater than the corresponding dynamic performance threshold, the drone swarm performance level is determined to be the second performance level; If the first performance result, the second performance result, and the third performance result are all smaller than the corresponding dynamic performance thresholds, the drone swarm performance level is determined to be the third performance level.

5. The method according to claim 1, characterized in that The collaborative operation data includes a communication adjacency matrix, actual formation position data, and UAV positioning data; Calculation of real-time collaborative indicators includes: The communication stability coefficient is calculated by the algebraic connectivity of the communication adjacency matrix; The Hausdorff distance between the actual formation position data and the ideal formation position is calculated to obtain the formation retention rate; The root mean square error of the relative positions between drones is calculated based on the drone positioning data to determine the collaborative positioning accuracy; The communication stability coefficient, formation retention rate and collaborative positioning accuracy are weightedly integrated to obtain the real-time collaborative index.

6. The method according to claim 1, characterized in that The environmental disturbance data includes three-dimensional wind speed vector data, obstacle distribution grid data, and electromagnetic spectrum data; and the calculation of real-time environmental indicators based on the environmental disturbance data includes: Calculate the modulus of the gradient vector of the three-dimensional wind speed vector data to determine the wind speed disturbance intensity; Calculate the density of each grid cell based on the obstacle distribution grid data; Performing energy integration on the interference frequency band of electromagnetic spectrum data to obtain electromagnetic interference intensity; The wind speed disturbance intensity, grid cell density and electromagnetic interference intensity are harmonically averaged to obtain the real-time environmental index.

7. The method according to claim 1, characterized in that The mission execution data includes the coordinates of the drone coverage area, actual path length data and mission completion time; Calculation of real-time task indicators includes: Determine the area of the drone coverage area according to the coordinates of the drone coverage area, calculate the intersection area of the drone coverage area and the target area, and calculate the target coverage rate based on the ratio of the intersection area to the target area; Calculate the ratio of the theoretical optimal path length to the actual path length data to determine the path selection reliability; The ratio of the total planned time to the task completion time is calculated to calculate the task completion efficiency; Calculate real-time task indicators based on target coverage, path selection reliability, and task completion efficiency.

8. The method according to claim 2, characterized in that The efficacy prediction model is shown in the following formula: Among them, K σ is the dynamic kernel function, b is the bias term, α, α * is a pair of Lagrange multipliers, x is a support vector, x i is the three-dimensional input feature vector of the i-th sample, and N is the total number of training samples.

9. The method according to claim 8, characterized in that The dynamic kernel function is shown in the following formula: Among them, x c is the synergy indicator support subvector, x e is the environmental indicator support subvector, x t is the task indicator support subvector, is the synergy indicator subvector, is the environmental indicator subvector, is the task indicator subvector, σ c is the kernel function parameter corresponding to the collaborative index sub-vector, σ e is the kernel function parameter corresponding to the environmental indicator subvector, σ t is the kernel function parameter corresponding to the collaborative indicator sub-vector.

10. The method according to claim 9, characterized in that The objective function of iteratively updating the kernel function parameters by the improved particle swarm algorithm is: Among them, N is the total number of training samples, y n is the historical performance label corresponding to the nth training sample, f' n (x) is the predicted output of the support vector regression model.

Images