Unmanned aerial vehicle cluster task planning method for area coverage reconnaissance

Abstract

The invention discloses an unmanned aerial vehicle cluster task planning method for area coverage reconnaissance, and relates to the technical field of unmanned aerial vehicle cluster cooperative reconnaissance. The method comprises the steps of determining a target area; rasterizing the target area to obtain a plurality of grid areas; constructing a global fitness function, simulating a hunting strategy of the wolf pack by adopting an improved grey wolf optimization algorithm, and evaluating the advantages and disadvantages of each grey wolf through the fitness function; the global fitness function is constructed as the weighted sum of the area coverage rate and the target point coverage rate, the area coverage rate is determined according to the number of grid areas covered by the sensor, and the target point coverage rate is calculated by counting the number of times that each target point is covered by the sensor and performing normalization. According to the method, the area coverage rate of the unmanned aerial vehicle cluster area coverage reconnaissance task can be improved, reliability and expandability are achieved, and further development and application of the unmanned aerial vehicle cluster area reconnaissance coverage technology can be promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle (UAV) cluster collaborative reconnaissance, and more specifically, to a UAV cluster mission planning method for regional coverage reconnaissance. Background Art

[0002] Drone swarm coordinated operations refer to the coordinated cooperation of multiple drones to achieve reconnaissance, monitoring, tracking and fire strikes on important enemy targets, which has important military application value. In the early stage of a war, drone swarms can conduct reconnaissance and search of the combat area to obtain important intelligence information such as the detailed location, defense composition, and combat power of the enemy targets to be attacked; at the end of a war, they can be used to check the battlefield and eliminate hidden dangers.

[0003] How to efficiently and quickly cover and reconnaissance the target area is one of the key issues in its combat application. With the development of drone technology, drone swarm collaborative reconnaissance should have intelligent features such as highly autonomous decision-making evaluation, task division, trajectory planning, and motion control. At present, most studies model the mapping relationship between drone swarms and the area to be reconnaissance as a task allocation model. By constructing constraints and indicator functions that need to be optimized, centralized or distributed allocation algorithms are designed to make the overall reconnaissance efficiency as optimal as possible under the constraints of indicator functions and spatiotemporal conditions. One of the challenges facing the problem of drone swarm regional coverage reconnaissance mission planning is its computational complexity. The drone swarm collaborative reconnaissance mission planning problem is modeled as a multi-objective combinatorial optimization problem. The computational complexity of the problem mainly comes from the scale of the problem, including the number of reconnaissance targets, the number of drones, and the scope of the area to be reconnaissance. As the scale of the problem increases, its search space increases dramatically, and it is difficult or even impossible to find the exact optimal solution to the problem. Therefore, when solving such problems, we should focus on finding an approximate optimal solution or a satisfactory solution, and heuristic algorithms are one of the best tools for finding such a satisfactory solution. In addition, the characteristics of heuristic algorithms based on population calculation make them a powerful tool for solving multi-objective optimization problems. Heuristic multi-objective optimization has become a hot issue in multi-objective optimization research in recent years. Grey Wolf Optimizer (GWO) is a heuristic algorithm based on the social hierarchy and hunting behavior of gray wolves, which is used to solve optimization problems. The algorithm simulates the hierarchy and hunting strategy of gray wolf packs to find the global optimal solution. In each iteration, α, β and δ wolves guide the population update to avoid local optimality. Despite the problems of population diversity and slow convergence in the later stage, GWO has been applied in many fields due to its simplicity and robustness, and some scholars have proposed improvement strategies to enhance its performance.

[0004] The mission scenario considered by the present invention is: a reconnaissance group composed of multiple rotor drones carrying optoelectronic payloads enters the target area, knowing only the geographical location information and boundary information of the target area, as well as the locations of several key targets in the area. The mission mission is to cover the entire area as much as possible under the premise that at least one drone continuously conducts reconnaissance of key targets. At present, there are still some limitations in using the Gray Wolf Optimization Algorithm to solve the planning of drone cluster area coverage reconnaissance missions, which are mainly reflected in:

[0005] (1) Easy to fall into local optimal solution: Due to the limitations of the algorithm, the gray wolf optimization algorithm leads to a fixed social hierarchy, which limits the diversity and adaptability of the wolf pack in the search process. Once the position update of α, β, and δ wolves is not enough to guide the group to develop towards the global optimal solution, the entire wolf pack may be guided to the local optimal solution, resulting in a decrease in coverage.

[0006] (2) Slow convergence speed: The linearly decreasing convergence factor of the Gray Wolf Optimization Algorithm is not applicable to the optimization process of all problems. It may cause the algorithm to focus on local areas in the early stage, thus affecting the overall convergence speed.

[0007] (3) Uneven population initialization: The quality and uniformity of population initialization have an important impact on the global search capability of the algorithm. If the initial population is unevenly distributed, it may cause the algorithm to concentrate on a local area in the early stage, thus affecting the overall convergence accuracy.

[0008] (4) Search capability needs to be improved: The Gray Wolf Optimization Algorithm does not adequately balance global search and local search, resulting in insufficient global exploration capability in the early stages and difficulty in effectively covering the entire search space. In the later stages of local search, there is a lack of effective search strategies, making it easy to fall into local optimality. Summary of the invention

[0009] In order to solve the above technical problems, the present invention provides a UAV cluster task planning method for regional coverage reconnaissance, aiming to improve the shortcomings of the Gray Wolf optimization algorithm in the field of UAV cluster regional coverage reconnaissance, and achieve more efficient and comprehensive regional coverage reconnaissance effects.

[0010] The present invention provides a method for planning a UAV cluster mission for regional coverage reconnaissance, the method comprising:

[0011] Determine a target area; wherein the target area is an area to be surveyed by the drone cluster;

[0012] Performing rasterization processing on the target area to obtain a plurality of grid areas;

[0013] Construct a global fitness function, adopt an improved gray wolf optimization algorithm, simulate the hunting strategy of wolves, and evaluate the quality of each gray wolf through the fitness function; wherein, the constructed global fitness function is a weighted sum of the area coverage rate and the target point coverage rate, the area coverage rate is determined according to the number of grid areas covered by the sensor, the target point coverage rate is calculated by counting the number of times each target point is covered by the sensor and normalizing it, and the target point is the location point of the key target in the target area;

[0014] Optimize the initial position of the wolf pack, initialize the position of the gray wolf by using Latin hypercube sampling, and amplify the random sampling data proportionally to obtain the sample position that adapts to the specific search space;

[0015] Introducing a nonlinear convergence factor; the nonlinear convergence factor is used to adjust the search range according to the number of iterations;

[0016] Based on the dynamic alpha wolf selection strategy, the alpha wolf is dynamically selected; wherein the alpha wolf includes alpha wolf, beta wolf and delta wolf, the alpha wolf is selected from the wolves with the highest fitness value, the beta wolf is randomly selected from the top 10% of the wolves with the highest fitness value, and the delta wolf is randomly selected from the top 10% to 20% of the wolves; the fitness value is calculated according to the global fitness function;

[0017] Based on the wolf pack position update strategy, the positions of all gray wolves are updated, and the fitness value of the gray wolf pack is recalculated based on the global fitness function, and the dynamic leader selection strategy wolf pack position update strategy is executed until the set termination condition is reached, and the optimal gray wolf position and corresponding fitness value are output to obtain the drone cluster area coverage reconnaissance mission planning scheme.

[0018] Furthermore, based on the dynamic leader selection strategy, the leader is dynamically selected, including:

[0019] Randomly generate a group of gray wolves, each gray wolf represents a potential solution x, and calculate its fitness value;

[0020] Define α wolf as the individual with the highest fitness, and select α wolf by the following formula:

[0021] α=argmaxf(x i ) (1)

[0022] Among them, f(x i ) represents the i-th gray wolf x i The fitness value function of ; argmax represents;

[0023] The beta wolf is selected by the following formula:

[0024] β=RandomSelect({x i |f(xi )∈Top10%\α}) (2)

[0025] Among them, RandomSelect means random selection within the range, TOP10%\α means the wolf pack with the top 10% fitness value after excluding α wolf;

[0026] The delta wolf is selected by the following formula:

[0027] δ=RandomSelect({x i |f(x i )∈Top10%to20%}) (3)

[0028] Among them, Top10% to 20% represents the wolf packs with the top 10% to 20% fitness values.

[0029] Furthermore, the wolf pack position update strategy includes a common wolf pack update strategy, a delta wolf update strategy, a beta wolf update strategy and an alpha wolf update strategy.

[0030] Furthermore, the common wolf pack update strategy includes:

[0031] In the gray wolf optimization algorithm, based on the positions of the three wolves α, β, and δ, the wolf pack disperses and searches, introducing random values. When the random value is greater than 1, the wolf pack moves away from the prey for a global search. The modeling formula for this process is as follows:

[0032] D=|C·X P (t)-X i (t)| (4)

[0033] X i (t+1)=X P (t)-A·D (5)

[0034] Among them, t represents the current iteration number, A and C are coefficients, and X P is the position of the prey, X i (t) is the position of the gray wolf individual at the tth iteration; X i (t+1) is the position of the gray wolf individual at the t+1th iteration; X p (t) represents the location of the prey; D represents the distance between the leader and other individuals;

[0035] The coefficients A and C are calculated by the following formula:

[0036] A=2a·r1-a (6)

[0037] C=2·r2 (7)

[0038] Where r1 and r2 are random values ​​in [0, 1], a is the convergence factor, and A is a random value in the interval [-a, a], where a decreases from 2 to 0 during the iteration process;

[0039] The hunting is carried out by α wolf, β wolf and δ wolf. The mathematical model of the gray wolf tracking the location of the prey is described as follows:

[0040] D n =|C1·X n -X i | (8)

[0041] Among them, D n represents the distance between the nth wolf leader and other wolves; X n represents the current position of the nth wolf leader; C1 is a random number, X i is the current location of the individual gray wolf;

[0042] X1 is calculated by the following formula:

[0043] X1=X α -A1·D α (9)

[0044] X2=X β -A2·D β (10)

[0045] X3=X δ -A3·D δ (11)

[0046] Among them, X1, X2, and X3 represent the positions of the wolves adjusted by the influence of α wolf, β wolf, and δ wolf respectively; D α , D β , D δ Respectively represent the distances between α wolf, β wolf, δ wolf and other wolf packs; A1, A2, A3 represent the control factors of the adjustment range of the wolf pack to the position of α wolf, β wolf, δ wolf respectively;

[0047] Update the position of the individual gray wolf using the following formula:

[0048]

[0049] Among them, W α represents the influence weight of wolf α, W β represents the influence weight of β wolf, W δ represents the influence weight of δ wolf, f α represents the fitness value of wolf α, f β represents the fitness value of β wolf, f δ represents the fitness value of δ wolf, X i(t+1) is the position of the gray wolf at the t+1th iteration; X1, X2 and X3 are the positions of the wolf pack after adjustment under the influence of α wolf, β wolf and δ wolf respectively.

[0050] Furthermore, the δ wolf updating strategy includes: updating the position of δ wolf by the following formula:

[0051]

[0052] Among them, X δ (t+1) is the updated position of wolf δ; X1 and X2 are the adjusted positions of the wolves under the influence of wolf α and wolf β respectively, W α represents the influence weight of wolf α, W β represents the influence weight of β wolf, f α represents the fitness value of wolf α, f β Represents the fitness value of β wolf.

[0053] Furthermore, the β wolf update strategy includes: updating the position of the β wolf by the following formula:

[0054] X β (t+1)=X1(1+Ν(0,1)cos(2πC1)) (17)

[0055] Among them, X β (t+1) is the updated position of wolf β, Ν(0,1) is a random value drawn from the standard normal function, C1 is a random number, and X1 represents the adjusted position of the wolf pack under the influence of wolf α;

[0056] The α wolf update strategy includes: updating the position of the α wolf by the following formula:

[0057]

[0058] u=σΝ(0,1) (19)

[0059] v=Ν(0,1) (20)

[0060]

[0061] Where Γ represents the gamma function, Ν(0,1) represents a random value drawn from a standard normal distribution, γ is the Mantegna index, σ is the standard deviation of the Levy distribution, u and v are random variables from a standard normal distribution, and X α (t+1) and X α (t) represent the old and new positions of wolf α respectively.

[0062] Furthermore, the nonlinear convergence factor is expressed as:

[0063]

[0064] Where a represents the convergence factor, t max represents the maximum number of iterations, e represents a natural constant, and t represents the current number of iterations.

[0065] The present invention has at least the following beneficial effects:

[0066] (1) Improve target coverage: The present invention optimizes the deployment of drone clusters using the improved Gray Wolf Optimization Algorithm, which greatly improves the coverage of target points in a specified area. By using a nonlinear convergence factor, the global search range of the algorithm is increased in the early stage, while the speed of local convergence in the later stage of the algorithm is increased, so that the algorithm approaches the optimal solution faster and improves target coverage.

[0067] (2) Increase reliability: In order to overcome the problem that the gray wolf optimization algorithm is prone to falling into local optimal solutions, the present invention innovatively uses a dynamic alpha wolf selection strategy, which aims to ensure that the selection of the alpha wolf is as diverse as possible in each iteration, thereby enhancing the adaptability and exploration ability of the population, thereby reducing the probability of the algorithm falling into local optimal solutions and improving the reliability of target reconnaissance.

[0068] (3) Improve scalability: The global function fitness value in the present invention can adapt to a variety of different reconnaissance needs. Whether it is the coverage of the target area or the monitoring of the key points of the target, different reconnaissance needs can be met by changing different fitness value calculation modes, and the reconnaissance mode can be adjusted in real time according to environmental changes and battlefield requirements.

[0069] (4) Enhance the collaborative search capability of the wolf pack: By optimizing the position update strategy of the wolf pack, the present invention strengthens the information sharing and synergy within the wolf pack. In particular, by giving the alpha wolf greater influence and introducing an independent exploration strategy for the alpha wolf, the algorithm can more effectively integrate the group wisdom and improve the synergy and overall performance of the search. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 A schematic diagram of an application scenario according to the prior art is shown.

[0071] Figure 2 A flowchart of a method for planning a UAV cluster mission for regional coverage reconnaissance according to an embodiment of the present invention is shown.

[0072] Figure 3 A flow chart of a dynamic alpha wolf selection strategy according to an embodiment of the present invention is shown.

[0073] Figure 4The diagram shows the layout of drones according to an embodiment of the present invention after the traditional grey wolf optimization algorithm and the method proposed by the present invention; wherein, (a) is the traditional grey wolf optimization algorithm; and (b) is the method proposed by the present invention.

[0074] Figure 5 A schematic diagram showing a comparison of convergence speeds of two convergence factors according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0075] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings and specific embodiments, but are not intended to limit the present invention. For the various steps described herein, if there is no necessity for a causal relationship between each other, the order in which they are described as examples herein should not be regarded as a limitation, and those skilled in the art should know that they can be adjusted in order, as long as the logic between them is not destroyed, resulting in the inability to implement the entire process.

[0076] The embodiment of the present invention provides a method for planning a UAV cluster mission for regional coverage reconnaissance. Figure 2 As shown, the method includes steps S1 to S7, which are described in detail as follows.

[0077] S1, determine the target area; wherein the target area is the area to be surveyed by the drone cluster.

[0078] For example, this embodiment provides a task scenario, such as Figure 1 As shown in the figure, the mission scenario is: a reconnaissance group consisting of multiple rotary-wing drones carrying optoelectronic payloads enters the target area, knowing only the geographical location information and boundary information of the target area, as well as the locations of several key targets in the area. The mission mission is to cover the entire area as much as possible while at least one drone continuously conducts reconnaissance of key targets.

[0079] Based on the above mission scenarios, a task allocation model for drone clusters can be constructed, and systematic modeling can be performed considering factors such as geographic information, drone sensors, and target point locations, thereby determining the scope of the target area and the key targets in the target area. It should be noted that the construction of the task allocation model for drone clusters and the method of systematic modeling to obtain the target area can adopt existing technologies. At the same time, this embodiment only illustrates a method for determining the target area. In actual implementation, the target area is not necessarily selected in the manner exemplified above.

[0080] S2, performing rasterization processing on the target area to obtain a plurality of grid areas.

[0081] In this embodiment, the target area is gridded to simplify the complex environment into manageable units, achieving effective regional coverage and key target monitoring, providing clear spatial division for subsequent task allocation, and optimizing resource allocation.

[0082] S3, constructing a global fitness function, using an improved gray wolf optimization algorithm, simulating the hunting strategy of the wolf pack, and evaluating the pros and cons of each gray wolf through the fitness function; wherein, the constructed global fitness function is a weighted sum of the area coverage rate and the target point coverage rate, the area coverage rate is determined according to the number of grid areas covered by the sensor, the target point coverage rate is calculated by counting the number of times each target point is covered by the sensor and normalizing it, and the target point is the location point of the key target in the target area.

[0083] In this embodiment, the global fitness function combines the area coverage rate and the target point coverage rate, and achieves different requirements through adjustable weights. The area coverage rate is determined by dividing the area into small grids and calculating the number of grids covered by the sensor, while the target point coverage rate is calculated by counting the number of times each target point is covered by the sensor and normalizing it. Ultimately, the weighted sum of these two coverage rates constitutes the global fitness function, which is used to guide the mission planning of the drone swarm, ensure that key targets are continuously monitored, and maximize the area coverage to improve the efficiency of mission execution.

[0084] S4, optimize the initial position of the wolf pack, initialize the position of the gray wolf by using Latin hypercube sampling, and amplify the random sampling data proportionally to obtain the sample position that adapts to the specific search space.

[0085] In this embodiment, Latin hypercube sampling (LHS) is used to initialize the gray wolf position, and the random sampling data can be amplified in proportion, so that the sample position suitable for the specific search space can be obtained, thereby improving the global search capability of the algorithm.

[0086] S5, introducing a nonlinear convergence factor; the nonlinear convergence factor is used to adjust the search range according to the number of iterations.

[0087] In some embodiments, a convergence factor based on a sigmoid function is used as a nonlinear convergence factor. By adopting the nonlinear convergence factor of the sigmoid function, the factor adjusts the range according to the number of iterations: when the absolute value of the convergence factor is greater than 1, the wolf pack will expand the search range for a global search; as the iteration proceeds, when the absolute value of the convergence factor is less than 1, the wolf pack will narrow the search range and focus on local search.

[0088] Specifically, based on the classic gray wolf optimization algorithm, when |A|>1, the gray wolf group will expand the search range to find prey, that is, global search, and the convergence speed is fast; when |A|<1, the gray wolf group will shrink the search range to attack the prey, that is, local search, and the convergence speed is slow. Therefore, the size of A has a great relationship with the global search and local search capabilities of the gray wolf optimization algorithm. A changes with the change of the nonlinear convergence factor a. The nonlinear convergence factor a decreases linearly from 2 to 0 with the number of iterations, but the algorithm is not linear in the process of continuous convergence, so the nonlinear convergence factor a cannot fully reflect the actual optimization search process. Therefore, this embodiment adopts a convergence factor based on the sigmoid function, and its expression is:

[0089]

[0090] By adjusting the function (22) to conform to the usage scenario of this patent, the updated convergence factor is as follows:

[0091]

[0092] Where a represents the convergence factor, t max represents the maximum number of iterations, e represents a natural constant, and t represents the current number of iterations.

[0093] Figure 5 This is a comparison chart of the convergence speed of the original linear convergence factor and the convergence factor of the improved grey wolf optimization algorithm. Figure 5 It can be seen that the nonlinear convergence factor used in this application can increase the wolf pack's global search ability in the early stage and the local convergence speed in the later stage compared to the traditional linear convergence factor.

[0094] S6, dynamically select the alpha wolf based on the dynamic alpha wolf selection strategy; wherein the alpha wolf includes α wolf, β wolf and δ wolf, the α wolf is selected with the highest fitness value, the β wolf is randomly selected from the top 10% of wolves in terms of fitness value, and the δ wolf is randomly selected from the top 10% to 20% of wolves; the fitness value is calculated according to the global fitness function.

[0095] In this embodiment, the dynamic leader selection strategy can solve the problem of fixed wolf pack levels causing the algorithm to fall into a local optimal solution, and can also increase the adaptability and exploration ability of the population, effectively improving the performance of the algorithm in complex optimization problems.

[0096] In some embodiments, Figure 3 As shown in the figure, the specific steps of the dynamic wolf pack selection strategy are as follows:

[0097] S61, first initialize and randomly generate a group of gray wolves, each gray wolf represents a potential solution x, and then calculate its fitness value.

[0098] S62, selection of α wolf, where α wolf is defined as the individual with the highest fitness.

[0099] α=argmaxf(x i ) (1)

[0100] Among them, f(x i ) represents the i-th gray wolf x i Fitness value function; argmax representation.

[0101] S63, while the beta wolf is randomly selected from the top 10% of the fitness ranking (excluding the first place)

[0102] β=RandomSelect({x i |f(x i )∈Top10%\α}) (2)

[0103] Among them, RandomSelect means random selection within the range, and TOP10% means the wolf packs with the top 10% fitness values ​​except wolf α.

[0104] S64, delta wolves are randomly selected from the top 10% to 20% in fitness ranking.

[0105] δ=RandomSelect({x i |f(x i )∈Top10%to20%}) (3)

[0106] Among them, Top10% to 20% indicates the wolf packs with the top 20% fitness values.

[0107] The dynamic wolf selection strategy proposed in this embodiment dynamically adjusts the selection of the alpha wolf to ensure that the selection of β and δ wolves is different in each iteration, achieving diversity as much as possible. This dynamic selection mechanism enhances the adaptability and exploration ability of the population, allowing each iteration to explore the solution space from different angles and ranges, thereby significantly improving the performance and search efficiency of the algorithm in complex optimization problems. By continuously updating the alpha wolf sequence, the algorithm can continuously introduce new search directions and information, avoiding the limitations brought by the fixed search mode and increasing the probability of finding the global optimal solution.

[0108] S7, based on the wolf pack position update strategy, updates the positions of all gray wolves, and recalculates the fitness value of the gray wolf pack based on the global fitness function and executes the dynamic leader selection strategy wolf pack position update strategy until the set termination condition is reached, outputs the optimal gray wolf position and the corresponding fitness value, and obtains the UAV cluster area coverage reconnaissance mission planning plan.

[0109] In this embodiment, the position update strategy of the wolf pack is optimized. Except for the alpha wolf, the position update of the ordinary gray wolf is only affected by the alpha wolf: α, β, δ wolf, and by adding a dynamic weight strategy, the influence of the alpha wolf with a larger fitness value on the wolf pack update is increased; the update of the δ wolf is only affected by the β wolf and the α wolf; the update of the β wolf is only affected by the α wolf; the α wolf has the highest level and status and should not be affected by any wolf pack, so consider adding a Levy-based independent exploration strategy to the α wolf. By adding this method, the α wolf can perform more flexible exploration in the search space, improving the performance of the algorithm in complex optimization problems.

[0110] In some embodiments, the overall wolf pack update strategy is as follows:

[0111] Ordinary wolf pack update strategy: In the gray wolf optimization algorithm, the update strategy of the ordinary wolf pack depends on the positions of the three wolves α, β, and δ. The wolf pack disperses the search and then concentrates on attacking the prey. In order to enhance the exploration ability, the algorithm introduces random values ​​to prompt the wolf pack to deviate from the current prey position to find a better solution. This divergent model is implemented by random values. When the random value is greater than 1, the wolf pack will move away from the prey for a global search. The process modeling formula is as follows:

[0112] D=|C·X P (t)-X i (t)| (4)

[0113] X i (t+1)=X P (t)-A·D (5)

[0114] Among them, t represents the current iteration number, A and C are coefficients, and X P is the position of the prey, X i (t) is the position of the gray wolf individual at the tth iteration; X i (t+1) is the position of the gray wolf individual at the t+1th iteration; X p (t) represents the location of the prey; D represents the distance between the leader and other individuals;

[0115] The calculation methods of A and C are as follows:

[0116] A=2a·r1-a (6)

[0117] C=2·r2 (7)

[0118] Where r1, r2 are random values ​​in [0, 1], and a is a convergence factor. To simulate approaching prey, A is a random value in the interval [-a, a], where a decreases from 2 to 0 during the iteration.

[0119] Gray wolves have the ability to identify the location of prey and hunt them in groups. The group hunts with alpha, beta, and delta wolves. The mathematical model of how individual gray wolves track the location of their prey is described as follows:

[0120] D n =|C1·X i -X| (8)

[0121] Among them, D n represents the distance between the nth wolf leader and other wolves; X n represents the current position of the nth wolf leader; C1 is a random number, X i is the current location of the individual gray wolf.

[0122] X1=X α -A1·D α (9)

[0123] X2=X β -A2·D β (10)

[0124] X3=X δ -A3·D δ (11)

[0125] Among them, X1, X2, and X3 represent the positions of the wolves adjusted by the influence of α wolf, β wolf, and δ wolf respectively. α , D β , D δ They respectively represent the distances between α wolf, β wolf, δ wolf and other wolf packs; A1, A2, A3 respectively represent the control factors of the adjustment amplitude of the wolf pack towards the position of α wolf, β wolf, δ wolf.

[0126] X1, X2 and X3 are the positions of the wolf pack ω after being affected by α wolf, β wolf and δ wolf. The dynamic weight formula is as follows:

[0127]

[0128]

[0129] Among them, W α represents the influence weight of wolf α, W β represents the influence weight of β wolf, W δ represents the influence weight of δ wolf, f α represents the fitness value of wolf α, f β represents the fitness value of β wolf, f δ represents the fitness value of δ wolf, X i (t+1) is the position of the gray wolf at the t+1th iteration; X1, X2 and X3 are the positions of the wolf pack after adjustment under the influence of α wolf, β wolf and δ wolf respectively.

[0130] The weights of the three wolves are calculated and assigned according to their fitness values, so that wolf α has a greater weight in the hunting process, followed by wolves β and δ.

[0131] δ wolf update strategy: The update of this wolf is only affected by β and α wolves, and its update method is as follows:

[0132]

[0133] Among them, X δ (t+1) is the updated position of wolf δ; X1 and X2 are the adjusted positions of the wolves under the influence of wolf α and wolf β respectively, W α represents the influence weight of wolf α, W β represents the influence weight of β wolf, f α represents the fitness value of wolf α, f β Represents the fitness value of β wolf.

[0134] β wolf update strategy: Similar to δ, this wolf is more influenced by α wolf, and its update method is as follows:

[0135] X β (t+1)=X1(1+Ν(0,1)cos(2πC1)) (17)

[0136] Among them, X β (t+1) is the updated position of β wolf, Ν(0,1) is the random value drawn from the standard normal function, C1 is a random number, and X1 represents the adjusted position of the wolf pack under the influence of α wolf. The introduction of random numbers is used to increase the exploration ability of β wolf.

[0137] α wolf update strategy: α wolf has the highest level and should not be affected by other wolf packs, so the update strategy of this wolf is random exploration. Levy flight has the characteristics of long jumps and short jumps. By introducing Levy flight in the individual update process, the α wolf pack can explore more flexibly in the search space and improve the performance of the algorithm in complex optimization problems. The formula is as follows:

[0138]

[0139] u=σΝ(0,1) (19)

[0140] v=Ν(0,1) (20)

[0141]

[0142] Where Γ represents the gamma function, Ν(0,1) represents a random value drawn from a standard normal distribution, γ is the Mantegna index (here 1.5), σ is the standard deviation of the Levy distribution, u and v are random variables from a standard normal distribution, and X α(t+1) and X α (t) represent the old and new positions of wolf α respectively.

[0143] This embodiment is also based on the same application scenario, that is, Figure 1 In the application scenario shown in the figure, the traditional gray wolf optimization algorithm and the method proposed in the present invention are used to simultaneously perform task planning. The performance comparison is shown in Table 1. The UAV layout diagram after the traditional gray wolf optimization algorithm and the method proposed in the present invention is shown in Figure 4 shown.

[0144] Table 1 Performance comparison between the traditional grey wolf optimization algorithm and the method proposed in this invention

[0145]

[0146] From Table 1 and Figure 4 It can be seen that the method proposed in the present invention is based on the traditional gray wolf optimization algorithm, which effectively improves the efficiency and reliability of drone swarm mission planning. The introduction of nonlinear convergence factors optimizes the search process and enhances the target detection coverage. The dynamic leader selection mechanism improves the adaptability and exploration ability of the population and reduces the risk of local optimality. The flexibility of the global fitness function allows the algorithm to adjust its strategy in real time according to reconnaissance needs. In addition, by optimizing the wolf pack position update strategy, the utilization of group intelligence is enhanced, and the search coordination and overall performance are improved. These improvements enable the present invention to show stronger adaptability and robustness in complex environments.

[0147] In addition, although exemplary embodiments have been described herein, the scope includes any and all embodiments based on the present invention with equivalent elements, modifications, omissions, combinations (e.g., various embodiments intersecting schemes), adaptations or changes. The elements in the claims will be interpreted broadly based on the language adopted in the claims, and are not limited to the examples described in this specification or during the implementation of this application, and the examples will be interpreted as non-exclusive. Therefore, this specification and examples are intended to be considered as examples only, and the true scope and spirit are indicated by the following claims and the full scope of their equivalents.

[0148] The above description is intended to be illustrative rather than restrictive. For example, the above examples (or one or more of them) can be used in combination with each other. For example, those of ordinary skill in the art can use other embodiments when reading the above description. In addition, in the above-mentioned specific embodiments, various features can be grouped together to simplify the present invention. This should not be interpreted as a feature of an invention that is not claimed for protection being necessary for any claim. On the contrary, the subject matter of the present invention may be less than all the features of the embodiments of a specific invention. Thus, the following claims are incorporated into the specific embodiments as examples or embodiments, wherein each claim is independently used as a separate embodiment, and it is considered that these embodiments can be combined with each other in various combinations or arrangements. The scope of the present invention should be determined with reference to the full scope of the equivalent forms of the attached claims and these claims.

Claims

1. A method for planning a UAV swarm mission for regional coverage reconnaissance, characterized in that: The method comprises: Determine a target area; wherein the target area is an area to be surveyed by the drone cluster; Performing rasterization processing on the target area to obtain a plurality of grid areas; Construct a global fitness function, adopt an improved gray wolf optimization algorithm, simulate the hunting strategy of wolves, and evaluate the quality of each gray wolf through the fitness function; wherein, the constructed global fitness function is a weighted sum of the area coverage rate and the target point coverage rate, the area coverage rate is determined according to the number of grid areas covered by the sensor, the target point coverage rate is calculated by counting the number of times each target point is covered by the sensor and normalizing it, and the target point is the location point of the key target in the target area; Optimize the initial position of the wolf pack, initialize the position of the gray wolf by using Latin hypercube sampling, and amplify the random sampling data proportionally to obtain the sample position that adapts to the specific search space; Introducing a nonlinear convergence factor; the nonlinear convergence factor is used to adjust the search range according to the number of iterations; Based on the dynamic alpha wolf selection strategy, the alpha wolf is dynamically selected; wherein the alpha wolf includes alpha wolf, beta wolf and delta wolf, the alpha wolf is selected from the wolves with the highest fitness value, the beta wolf is randomly selected from the top 10% of the wolves with the highest fitness value, and the delta wolf is randomly selected from the top 10% to 20% of the wolves; the fitness value is calculated according to the global fitness function; Based on the wolf pack position update strategy, the positions of all gray wolves are updated, and the fitness value of the gray wolf pack is recalculated based on the global fitness function, and the dynamic leader selection strategy wolf pack position update strategy is executed until the set termination condition is reached, and the optimal gray wolf position and corresponding fitness value are output to obtain the drone cluster area coverage reconnaissance mission planning scheme.

2. The method according to claim 1, characterized in that Based on the dynamic alpha selection strategy, the alpha is selected dynamically, including: Randomly generate a group of gray wolves, each gray wolf represents a potential solution x, and calculate its fitness value; Define α wolf as the individual with the highest fitness, and select α wolf by the following formula: α=arg max f(x i ) (1) Among them, f(x i ) represents the i-th gray wolf x i The fitness value function of ; argmax represents; The beta wolf is selected by the following formula: β=RandomSelect({x i |f(x i )∈Top10%\α}) (2) Among them, RandomSelect means random selection within the range, TOP10%\α means the wolf pack with the top 10% fitness value excluding wolf α; The delta wolf is selected by the following formula: δ=RandomSelect({x i |f(x i )∈Top10%to20%}) (3) Among them, Top10% to 20% represents the wolf packs with the top 10% to 20% fitness values.

3. The method according to claim 1, characterized in that The wolf pack position update strategy includes a common wolf pack update strategy, a delta wolf update strategy, a beta wolf update strategy and an alpha wolf update strategy.

4. The method according to claim 3, characterized in that The common wolf pack update strategy includes: In the gray wolf optimization algorithm, based on the positions of the three wolves α, β, and δ, the wolf pack disperses and searches, introducing random values. When the random value is greater than 1, the wolf pack moves away from the prey for a global search. The modeling formula for this process is as follows: D=|C·X P (t)-X i (t)| (4) X i (t+1)=X P (t)-A·D (5) Among them, t represents the current iteration number, A and C are coefficients, and X P is the position of the prey, X i (t) is the position of the gray wolf individual at the tth iteration; X i (t+1) is the position of the gray wolf individual at the t+1th iteration; X p (t) represents the location of the prey; D represents the distance between the leader and other individuals; The coefficients A and C are calculated by the following formula: A=2a·r1-a (6) C＝2·r2 (7) Where r1 and r2 are random values ​​in [0, 1], a is the convergence factor, and A is a random value in the interval [-a, a], where a decreases from 2 to 0 during the iteration process; The hunting is carried out by α wolf, β wolf and δ wolf. The mathematical model of the gray wolf tracking the location of the prey is described as follows: D n =|C1·X n -X i | (8) Among them, D n represents the distance between the nth wolf leader and other wolves; X n represents the current position of the nth wolf leader; C1 is a random number, X i is the current location of the individual gray wolf; X1, X2, and X3 are calculated using the following formula: X1=X α -A1·D α (9) X2=X β -A2·D β (10) X3=X δ -A3·D δ (11) Among them, X1, X2, and X3 represent the positions of the wolves adjusted by the influence of α wolf, β wolf, and δ wolf respectively; X α , X β , X δ Respectively represent the current positions of α wolf, β wolf, and δ wolf; D α , D β , D δ Respectively represent the distances between α wolf, β wolf, δ wolf and other wolf packs; A1, A2, A3 represent the control factors of the adjustment range of the wolf pack to the position of α wolf, β wolf, δ wolf respectively; Update the position of the individual gray wolf using the following formula: Among them, W α represents the influence weight of wolf α, W β represents the influence weight of β wolf, W δ represents the influence weight of δ wolf, f α represents the fitness value of wolf α, f β represents the fitness value of β wolf, f δ represents the fitness value of δ wolf, X i (t+1) is the position of the gray wolf individual in the t+1th iteration.

5. The method according to claim 3, characterized in that: The δ wolf update strategy includes: updating the position of the δ wolf by the following formula: Among them, X δ (t+1) is the updated position of wolf δ; X1 and X2 are the adjusted positions of the wolves under the influence of wolf α and wolf β respectively, W α represents the influence weight of wolf α, W β represents the influence weight of β wolf, f α represents the fitness value of wolf α, f β Represents the fitness value of β wolf.

6. The method according to claim 3, characterized in that The beta wolf update strategy includes: updating the position of the beta wolf by the following formula: X β (t+1)=X1(1+Ν(0,1)cos(2πC1)) (17) Among them, X β (t+1) is the updated position of wolf β, Ν(0,1) is a random value drawn from the standard normal function, C1 is a random number, and X1 represents the adjusted position of the wolf pack under the influence of wolf α.

7. The method according to claim 3, characterized in that The α wolf update strategy includes: updating the position of the α wolf by the following formula: u=σΝ(0,1) (19) v=Ν(0,1) (20) Where Γ represents the gamma function, Ν(0,1) represents a random value drawn from a standard normal distribution, γ is the Mantegna index, σ is the standard deviation of the Levy distribution, u and v are random variables from a standard normal distribution, and X α (t+1) and X α (t) represent the old and new positions of wolf α respectively.

8. The method according to claim 1, characterized in that The nonlinear convergence factor is expressed as: Where a represents the convergence factor, t max represents the maximum number of iterations, e represents a natural constant, and t represents the current number of iterations.

Images