A flapping-wing unmanned aerial vehicle bird repelling cruise route planning method based on WSN idea

Abstract

The application discloses a flapping-wing unmanned aerial vehicle bird-repelling cruise route planning method based on a WSN idea and belongs to the technical field of unmanned aerial vehicles. In view of problems of low deterrence, short time efficiency, small coverage range and the like of traditional bird-repelling means, the application generates sensor nodes by improving a particle swarm algorithm, realizes efficient coverage of a target area, adopts an improved k-means algorithm to cluster and partition the sensor nodes, divides detection areas for each unmanned aerial vehicle, and uses a genetic algorithm to optimize a cruise path, so that the unmanned aerial vehicle can guarantee biological deterrence effect, reduce labor cost, and improve inspection efficiency and coverage range. The method is suitable for bird damage prevention and treatment of broad areas such as orchards, power transmission and distribution lines, high-voltage substations and airports, and has remarkable economic and social benefits.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) technology, specifically a method for planning bird-scaring cruise routes for flapping-wing UAVs based on the WSN concept. Background Technology

[0002] Bird damage has always been an issue that cannot be ignored in my country's production development. In agriculture and forestry, the reduced grain yield and lower fruit quality caused by bird damage seriously restrict the development of the agricultural and forestry economy. Meanwhile, in the power system, the losses caused by bird damage leading to power transmission and distribution line tripping and short circuits are also increasing year by year. In the transportation sector, bird damage at airports often causes huge economic losses to airlines and seriously threatens the lives of airport staff and passengers.

[0003] Unmanned aerial vehicles (UAVs) are unmanned aircraft controlled by radio remote control equipment and their own program control devices, or operated autonomously, either completely or intermittently, by an onboard computer. With the continuous advancement of UAV technology and the development of related industries, UAVs have been widely used in various fields such as agricultural and forestry monitoring, geological exploration, power and communication inspection, safety monitoring, and medical rescue.

[0004] Traditional bird control methods, such as manual inspections and noise scare, have drawbacks including low deterrence, short effectiveness, and limited coverage. In contrast, using drones for bird control significantly reduces labor costs while offering higher inspection efficiency and wider coverage. Furthermore, compared to traditional rotary-wing drones, flapping-wing drones have an eagle-like shape, and their movement is closer to that of natural birds of prey, exerting a biological deterrent effect on birds and achieving better bird control results.

[0005] Using flapping-wing drones for bird control requires the drones to cruise within the mission area. Their flight paths should ideally ensure they traverse every point within the mission area using scanning sensors during the mission, thereby monitoring for bird activity. This involves issues related to full-area coverage path planning.

[0006] The coverage problem of Wireless Sensor Networks (WSNs) refers to ensuring that every target area in the network is covered by at least one sensor node to achieve effective monitoring and data collection. This aligns perfectly with the goal of bird deterrence patrols. Applying the WSN concept to the path planning of a flapping-wing robot's bird deterrence patrol, assuming that sensor nodes already cover the target area as much as possible, the flapping-wing robot only needs to traverse these nodes along a specific path to achieve full-area coverage patrol, transforming the full-area coverage path planning problem into a traveling salesman problem.

[0007] Meanwhile, key areas for bird control include agricultural and forestry orchards, power transmission and distribution lines, high-voltage substations, and airports. These areas cover a wide range, and using only a single flapping-wing drone for patrol and detection would be relatively time-consuming, requiring a high battery capacity. Furthermore, excessively long single patrol times would diminish the drone's biological deterrent effect. Therefore, when using flapping-wing drones for bird control in agricultural and forestry orchards, power transmission and distribution lines, high-voltage substations, and airports, multiple devices should be deployed for coordinated patrols to improve efficiency and deterrent effect. This involves the issue of coordinated patrols by multiple robots. Summary of the Invention

[0008] To address the above problems, this invention provides a bird-scaring cruise route planning method for flapping-wing UAVs based on the WSN concept, which solves the problems of low coverage efficiency and insufficient optimization of cruise path planning in existing flapping-wing UAV bird-scaring cruise route planning methods.

[0009] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0010] A method for bird-scaring cruise route planning for flapping-wing UAVs based on the concept of Wireless Sensor Networks (WSN) includes the following steps:

[0011] a. Sensor nodes are generated in the target area using an improved particle swarm optimization algorithm. Each particle represents the center coordinates of a potential coverage circle. The optimal position of the coverage circle is determined through iterative optimization. Inter-particle influence variables are introduced to enhance the mutual learning ability between particles.

[0012] b. The generated sensor nodes are clustered and partitioned using an improved k-means clustering algorithm. By optimizing the selection of the initial cluster centers, the nodes are divided into several clusters, and each cluster corresponds to the detection area of ​​a UAV.

[0013] c. A genetic algorithm is used to generate traversal paths for sensor nodes within each cluster, simulating the natural evolution process. The path length and turning angle are optimized through crossover and mutation operations, taking into account the flight dynamics of the UAV.

[0014] d. Conduct simulation verification, set parameters such as target area, number of UAVs, and detection radius, and verify the feasibility and effectiveness of the proposed method through simulation experiments;

[0015] e. Compare and analyze the effects of the proposed method with the traditional ox-plowing method for coverage cruise, and evaluate indicators such as coverage, cruise distance, endurance requirements, and single cruise time to determine the applicability of the method in specific mission scenarios.

[0016] The beneficial effects of the above scheme are: it realizes an efficient method for planning bird-scare patrol routes for flapping-wing UAVs. This method can ensure that UAVs can carry out effective bird-scare operations within the target area, while improving the coverage and biological deterrence effect of the patrol, and reducing manpower and time costs.

[0017] The detailed working principle of the above scheme is as follows: by comprehensively applying an improved particle swarm optimization algorithm, k-means clustering algorithm, and genetic algorithm, sensor nodes are first generated to determine the optimal position of the coverage circle. Then, these nodes are clustered and partitioned. Finally, an optimal traversal path is generated for each node within a cluster. This process involves the transformation of full-area coverage path planning and the solution of multi-robot cooperative cruising problems.

[0018] As a further improvement to the above scheme, the particle velocity update formula in the improved particle swarm optimization algorithm is as follows:

[0019]

[0020] in, and These represent the particle velocities in the t-th and t+1-th iterations, respectively. Let be the inertia weight of the algorithm at the t-th iteration; and These represent the learning factors that influence the particle's individual optimal position and the group's optimal position, respectively. Represents a random number between 0 and 1. :particle In time Location; and These are the individual optimal location storage variables and the group optimal location storage variables, respectively.

[0021] The beneficial effects of the above improvements are: they improve the efficiency and accuracy of the particle swarm optimization algorithm in searching for the global optimum, and enhance the algorithm's ability to search for both global and local optima by dynamically updating particle velocities.

[0022] The detailed working principle of the above improvement is as follows: the particle velocity update formula takes into account the particle's current velocity, the individual optimal position, the group optimal position, and random perturbation terms. This update mechanism allows particles to explore the solution space more deeply and extensively.

[0023] As a further improvement to the above scheme, the improved particle swarm optimization algorithm includes inertia weights. Update formula:

[0024]

[0025] in, Let be the inertia weight of the algorithm at the t-th iteration; It is the maximum value of the inertia weight. It is the minimum value of the inertia weight. It is the maximum number of iterations. This is the current iteration number.

[0026] The beneficial effect of the above improvements is that by adaptively adjusting the inertia weights, the performance of the particle swarm optimization algorithm is balanced between global and local search, thereby avoiding premature convergence to a suboptimal solution.

[0027] The detailed working principle of the above improvement is as follows: the inertia weight decreases as the number of iterations increases. This linear decreasing method can maintain a large global search capability in the early stage of the algorithm and gradually enhance the search capability of the neighborhood of the optimal solution in the later stage, that is, the local search capability.

[0028] As a further improvement to the above scheme, the particle swarm optimization algorithm introduces inter-particle influence variables. The formula for enhancing the mutual learning ability between particles is:

[0029]

[0030] in The best historical position of another particle is randomly selected; For particles In time Location; For the first The historical best fitness function value of each particle; Let be the historical best fitness function value of the j-th particle.

[0031] The beneficial effects of the above improvements are: enhanced information sharing among particles in the particle swarm optimization algorithm, and improved learning ability and convergence speed of the entire swarm by introducing inter-particle influence variables.

[0032] The detailed working principle of the above improvement is as follows: the inter-particle influence variable takes into account the difference between the particle and its historical best position. In this way, each particle considers not only its individual experience but also the successful experience of other particles when updating its position, thereby strengthening the synergistic effect of swarm intelligence.

[0033] As a further improvement to the above scheme, in the sensor node generation step, the coverage rate of the coverage circle is... The calculation formula is:

[0034]

[0035] in, The total number of covered circles; For all covering circles From 1 to Summation; : The set of points within the area to be patrolled; : Boolean variable, if point Located in the covering circle Within its coverage area, then If it is 1; otherwise, it is 0. For all points within the area to be scanned of Perform summation; where Point The number of times a device is repeatedly covered by multiple covering circles; Area of ​​the area to be patrolled; : Dispersion of the region.

[0036] The beneficial effects of the above improvements are: they provide an accurate method for calculating coverage, which can quantitatively evaluate the coverage effect of each coverage circle on the target area, and provide an important basis for path planning decisions.

[0037] The detailed working principle of the above improvement is as follows: by calculating the relationship between the coverage circle and each point in the area to be patrolled, and taking into account the coverage of points in the area that does not need to be detected, the coverage effect of the coverage circle is comprehensively evaluated, and the coverage rate of the entire system is calculated accordingly.

[0038] As a further improvement to the above scheme, the improved k-means clustering algorithm includes an evaluation function for selecting cluster centers. The calculation formula is as follows:

[0039]

[0040] in, The number of clusters; The total number of sensor nodes; Sensor Node Location; It is the first Cluster centers, Represents a node To the cluster center The distance.

[0041] The beneficial effects of the above improvements are: improved clustering performance of the k-means clustering algorithm, and improved clustering efficiency and quality by optimizing the selection of cluster centers.

[0042] The detailed working principle of the above improvement is as follows: the selection of initial cluster centers is optimized through an evaluation function based on the sum of the distances from each node to the potential cluster centers, thereby ensuring the representativeness of the cluster centers and the compactness of the clusters.

[0043] As a further improvement to the above scheme, the fitness function of the particle swarm optimization algorithm... The calculation formula is:

[0044]

[0045] in, Point Is it within the covering circle? Within its coverage area, Point The number of times the covered circle is decreased by one; : Represents the penalty value; : The area of ​​the target region where the battery needs to be extended; : Dispersion of the region.

[0046] The beneficial effects of the above improvements are: ensuring that the traversal path generated by the genetic algorithm is both the shortest and adapted to the flight characteristics of the UAV, thus optimizing the UAV's cruise efficiency and flight safety.

[0047] The detailed working principle of the above improvement is as follows: the fitness function takes into account both the path length and the cosine value of the turning angle, guiding the genetic algorithm to ensure that the path is as short as possible while avoiding excessive path curvature, thereby adapting to the flight dynamics characteristics of the UAV.

[0048] As a further improvement to the above scheme, the detection radius of the UAV in the simulation verification step... The calculation formula is:

[0049]

[0050] in, It is the flight altitude. It refers to the angle of the sensor's field of view.

[0051] The beneficial effects of the above improvements are: providing drones with accurate detection range, ensuring effective sensor coverage and bird deterrence.

[0052] The detailed working principle of the above improvement is as follows: based on the drone's flight altitude and the sensor's field of view angle, the drone's detection radius is calculated, thereby determining the actual range that the drone can monitor at the center point of each coverage circle.

[0053] As a further improvement to the above scheme, the genetic algorithm further includes a mechanism for adaptively adjusting the crossover probability and mutation probability to optimize the search performance of the algorithm; this mechanism includes:

[0054] a. Define the initial value of the crossover probability as 0.9, and dynamically adjust the crossover probability according to the iteration progress of the algorithm and the quality of the solution;

[0055] b. Define the initial value of the mutation probability as 0.05, and dynamically adjust the mutation probability according to the changes in population diversity during the algorithm search process;

[0056] c. Design an adaptive adjustment strategy to appropriately increase the mutation probability to maintain the diversity of solutions when the population shows signs of convergence;

[0057] d. The adaptive adjustment strategy also includes appropriately increasing the crossover probability during the local search phase to promote the generation of excellent solutions.

[0058] The beneficial effects of the above improvements are: by adaptively adjusting the crossover and mutation probabilities, the search performance of the genetic algorithm is optimized, and the efficiency and quality of the algorithm in solving problems are improved.

[0059] The detailed working principle of the above improvement is as follows: the crossover probability and mutation probability are dynamically adjusted according to the iteration progress of the algorithm and the diversity of the population, so as to balance global exploration and local development during the search process, avoid premature convergence, and promote the generation of excellent solutions.

[0060] As a further improvement to the above scheme, the simulation verification step further includes the following steps:

[0061] a. Define the target area as a rectangular area of ​​specific size, and divide the area into discrete grids, with each grid serving as a discrete point, to simulate the drone's cruise and monitoring;

[0062] b. Set the detection radius of the UAV so that the UAV can cover a certain range of grid around each discrete point;

[0063] c. Based on the number of drones and the layout of the target area, assign a patrol grid area to each drone to achieve coordinated patrol;

[0064] d. Simulate the drone cruising above the grid along the generated traversal path using simulation software, collect coverage data, and verify the effectiveness of the path planning;

[0065] e. Simulation verification also includes evaluating key indicators such as the coverage of the drone's cruise route, cruise distance, and battery life to ensure the practicality and efficiency of the planned path.

[0066] The beneficial effects of the above improvements are: simulation verification ensures the practicality and effectiveness of the proposed bird deterrence patrol route planning method, providing a reliable basis for practical applications.

[0067] The detailed working principle of the above improvement is as follows: simulate the drone cruising along the planned path in a simulation environment, evaluate the performance of the path planning by setting parameters and collecting data, including key indicators such as coverage, cruising distance and battery life, and verify the feasibility and efficiency of the method.

[0068] Beneficial effects of this invention:

[0069] This invention proposes a bird-scaring patrol route planning method for flapping-wing drones based on the WSN (Wild Bird Network) concept, designed to solve the path planning problem for multi-robot collaborative patrols covering the entire area. Compared to traditional rotary-wing drones, flapping-wing drones, with their eagle-like shape and more natural raptor-like movements, have a biological deterrent effect on birds, achieving better bird-scaring results. Attached Figure Description

[0070] Figure 1 This is a flowchart for clustering sensor nodes.

[0071] Figure 2 This is the final result diagram of the node generation.

[0072] Figure 3 This is a visual representation of the clustering results.

[0073] Figure 4 This is a schematic diagram of the results obtained by the genetic algorithm.

[0074] Figure 5 This is a graph showing the change in the shortest path for path 1 obtained by the genetic algorithm.

[0075] Figure 6 This is a graph showing the change in fitness values ​​for path 1 obtained by the genetic algorithm.

[0076] Figure 7 This is a graph showing the change in the shortest path for path 2 obtained by the genetic algorithm.

[0077] Figure 8 This is a graph showing the change in fitness values ​​for path 2 obtained by the genetic algorithm.

[0078] Figure 9 This is a graph showing the change of the shortest path for path 3 obtained by the genetic algorithm.

[0079] Figure 10 This is a graph showing the change in fitness values ​​for path 3 obtained by the genetic algorithm.

[0080] Figure 11 This is a graph showing the change in the shortest path (path 4) obtained by the genetic algorithm.

[0081] Figure 12 This is a graph showing the fitness value changes of path 4 obtained by the genetic algorithm.

[0082] Figure 13 This is a schematic diagram of the overall path. Detailed Implementation

[0083] To enable those skilled in the art to better understand the technical solution, the present invention will be described in detail below with reference to embodiments. The description in this part is only exemplary and explanatory, and should not be used to limit the scope of protection of the present invention in any way.

[0084] This invention provides a comprehensive path planning method based on the WSN concept, which improves the particle swarm optimization algorithm for generating sensor nodes, improves the kmene algorithm for clustering and partitioning, and uses a genetic algorithm for modeling and encoding to determine the node connection order. Specific technical solutions are as follows:

[0085] An improved particle swarm optimization algorithm is used to generate sensor nodes in the target area, ensuring that the sensing range covers the target area as much as possible.

[0086] An improved k-means algorithm is used to cluster and partition the sensor nodes, with each partition corresponding to the detection area of ​​a single drone.

[0087] Genetic algorithms are used to generate the most suitable traversal path for each sensor node in each cluster.

[0088] The following technical effects can be achieved:

[0089] This improved the efficiency of bird deterrence patrols and the effectiveness of biological deterrence.

[0090] Simulation verification demonstrated the feasibility and efficiency of the method.

[0091] Compared with the traditional ox-plowing method, although the coverage is reduced, the cruising distance is significantly shortened, the requirements for drone endurance are lower, and the single cruising time is shorter.

[0092] The implementation method is as follows:

[0093] The target area was constructed using simulation software, and the detection radius and dispersion of the area were set by the UAV.

[0094] The covering circle is generated by improving the particle swarm optimization algorithm, and clustering is performed by improving the kmeans algorithm.

[0095] A genetic algorithm was used to generate a cruise path, which was then verified through simulation.

[0096] Specifically, such as Figure 1-13 As shown, the technical solution provided by this invention is as follows: a method for planning bird-scaring cruise routes for flapping-wing UAVs based on the concept of Wireless Sensor Networks (WSN), the method comprising the following steps:

[0097] a. Sensor nodes are generated in the target area using an improved particle swarm optimization algorithm. Each particle represents the center coordinates of a potential coverage circle. The optimal position of the coverage circle is determined through iterative optimization. Inter-particle influence variables are introduced to enhance the mutual learning ability between particles.

[0098] b. The generated sensor nodes are clustered and partitioned using an improved k-means clustering algorithm. By optimizing the selection of the initial cluster centers, the nodes are divided into several clusters, and each cluster corresponds to the detection area of ​​a UAV.

[0099] c. A genetic algorithm is used to generate traversal paths for sensor nodes within each cluster, simulating the natural evolution process. The path length and turning angle are optimized through crossover and mutation operations, taking into account the flight dynamics of the UAV.

[0100] d. Conduct simulation verification, set parameters such as target area, number of UAVs, and detection radius, and verify the feasibility and effectiveness of the proposed method through simulation experiments;

[0101] e. Compare and analyze the effects of the proposed method with the traditional ox-plowing method for coverage cruise, and evaluate indicators such as coverage, cruise distance, endurance requirements, and single cruise time to determine the applicability of the method in specific mission scenarios.

[0102] As a preferred embodiment of the above, the particle velocity update formula in the improved particle swarm optimization algorithm is:

[0103]

[0104] in, and These represent the particle velocities in the t-th and t+1-th iterations, respectively. Let be the inertia weight of the algorithm at the t-th iteration; and These represent the learning factors that influence the particle's individual optimal position and the group's optimal position, respectively. Represents a random number between 0 and 1. :particle In time Location; and These are the individual optimal location storage variables and the group optimal location storage variables, respectively.

[0105] As a preferred embodiment of the above embodiments, the improved particle swarm optimization algorithm includes inertia weights. Update formula:

[0106]

[0107] in, In time Inertial weight; It is the maximum value of the inertia weight. It is the minimum value of the inertia weight. It is the maximum number of iterations. This is the current iteration number.

[0108] As a preferred embodiment of the above embodiments, the particle swarm optimization algorithm introduces inter-particle influence variables. The calculation formula is:

[0109]

[0110] in The best historical position of another particle is randomly selected; For particles In time Location; For the first The historical best fitness function value of each particle; Let be the historical best fitness function value of the j-th particle.

[0111] As a preferred embodiment of the above embodiments, in the sensor node generation step, the coverage rate of the coverage circle is... The calculation formula is:

[0112]

[0113] in, The total number of covered circles; : The set of points within the area to be patrolled; : Indicates the set of points in the target area that do not need to be detected; Point Is it within the covering circle? Within its coverage area, Point The number of times the covered circle is decreased by one; Area of ​​the area to be patrolled; : Dispersion of the region.

[0114] As a preferred embodiment of the above, the improved k-means clustering algorithm includes an evaluation function for selecting cluster centers. The calculation formula is as follows:

[0115]

[0116] in, The number of clusters; The total number of sensor nodes; Sensor Node Location; It is the first Cluster centers, Represents a node To the cluster center The distance.

[0117] As a preferred embodiment of the above embodiments, the fitness function of the particle swarm optimization algorithm... The calculation formula is:

[0118]

[0119] in, Point Is it within the covering circle? Within its coverage area, Point The number of times the covered circle is decreased by one; : Represents the penalty value; : The area of ​​the target region where the battery needs to be extended; : Dispersion of the region.

[0120] As a preferred embodiment of the above embodiments, the drone detection radius in the simulation verification step... The calculation formula is:

[0121]

[0122] in, It is the flight altitude. It refers to the angle of the sensor's field of view.

[0123] As a preferred embodiment of the above, the genetic algorithm further includes a mechanism for adaptively adjusting the crossover probability and mutation probability to optimize the search performance of the algorithm; this mechanism includes:

[0124] a. Define the initial value of the crossover probability as 0.9, and dynamically adjust the crossover probability according to the iteration progress of the algorithm and the quality of the solution;

[0125] b. Define the initial value of the mutation probability as 0.05, and dynamically adjust the mutation probability according to the changes in population diversity during the algorithm search process;

[0126] c. Design an adaptive adjustment strategy to appropriately increase the mutation probability to maintain the diversity of solutions when the population shows signs of convergence;

[0127] d. The adaptive adjustment strategy also includes appropriately increasing the crossover probability during the local search phase to promote the generation of excellent solutions.

[0128] As a preferred embodiment of the above, the simulation verification step further includes the following steps:

[0129] a. Define the target area as a rectangular area of ​​specific size, and divide the area into discrete grids, with each grid serving as a discrete point, to simulate the drone's cruise and monitoring;

[0130] b. Set the detection radius of the UAV so that the UAV can cover a certain range of grid around each discrete point;

[0131] c. Based on the number of drones and the layout of the target area, assign a patrol grid area to each drone to achieve coordinated patrol;

[0132] d. Simulate the drone cruising above the grid along the generated traversal path using simulation software, collect coverage data, and verify the effectiveness of the path planning;

[0133] e. Simulation verification also includes evaluating key indicators such as the coverage of the drone's cruise route, cruise distance, and battery life to ensure the practicality and efficiency of the planned path.

[0134] The principle of this invention is as follows:

[0135] 1.1 Improved Particle Swarm Optimization Algorithm:

[0136] Particle Swarm Optimization (PSO) is a swarm intelligence optimization algorithm based on the foraging behavior of bird flocks. Similar to genetic algorithms, the algorithm initializes a set of solutions and iteratively optimizes the value of these solutions until it obtains the optimal solution for the problem. As a metaheuristic optimization algorithm, PSO has advantages such as fast convergence speed and ease of adjustment. This paper uses PSO to determine the optimal position of a covering circle.

[0137] For the basic particle swarm optimization algorithm, particles adjust their velocity and position during iteration based on their individual optimal position and the group's optimal position, as shown in the following formula:

[0138]

[0139]

[0140] in, and These represent the particle velocities in the t-th and t+1-th iterations, respectively. and represents the position of the particle at the t-th iteration and the (t+1)-th iteration, respectively; w is the inertial weight of the particle's motion; and These represent the learning factors that influence the particle's individual optimal position and the group's optimal position, respectively. Represents a random number between 0 and 1; and These are the individual optimal location storage variables and the group optimal location storage variables, respectively.

[0141] For the inertia weight w, the larger its value, the stronger the algorithm's global search capability, but the weaker its local optimization capability. To ensure the global search capability in the early stage of the algorithm's iteration, and at the same time to ensure more accurate local optimization capability in the later stage of the iteration, w should decrease according to a certain rule during the iteration process. Therefore, w should be updated according to the following formula:

[0142]

[0143] in, Let be the inertia weight of the algorithm at the t-th iteration; It is the maximum value of the inertia weight. It is the minimum value of the inertia weight. It is the maximum number of iterations. This is the current iteration number.

[0144] To address the problem that traditional particle swarm optimization algorithms are prone to getting trapped in local optima, the following improvement can be made: Introduce inter-particle influence variables. This enhances the mutual learning ability between particles, among which... The calculation formula is as follows:

[0145]

[0146] in, The best historical position of another particle is randomly selected; For the first The historical best fitness function value of each particle.

[0147] In summary, we can derive the improved particle swarm optimization algorithm formula as follows:

[0148]

[0149]

[0150] 1.2 Establishment of the Coverage Model:

[0151] Assume the area of ​​the target monitoring region is It consists of two parts: the area to be patrolled and the area not to be patrolled. Deployment is carried out within the target area. A covering circle, The estimated values ​​are as follows:

[0152]

[0153] in, This is a compensation coefficient, related to the intersection area of ​​the covering circles; The area of ​​the target region where the battery needs to continue operating. This represents the area of ​​the covered circle. For Let the flight altitude of the flapping-wing UAV be... The field of view angle of its installed detection sensor is The detection radius of the flapping-wing UAV can be obtained. Therefore, the area of ​​the covering circle is set. .

[0154] Establish a Cartesian coordinate system, and discretize the points within the target area into... Let there be a set of discrete points, with a dispersion degree of D = S / mn. The set of points represents the set of all discrete points within the target region. The set of points is the set of centers of the circles covering the target region. It is the set of points in the target region that does not need to be detected. Let be the set of points in the target region to be scanned. It is easy to see that... , , ', and for any point within the target area Its coordinates are .

[0155] In particle swarm optimization (PSO), each particle represents a set of random solutions to the problem. For the covering circle generation problem, the position information of each particle is defined as the center coordinates of the generated covering circle, indicating that the particle dimension is n. The covering circle needs to cover as much of the area to be scanned in the target region as possible. Here, a Boolean variable is introduced. The formula is as follows:

[0156]

[0157] It can be seen that, Representing points Is ∈E' within the covering circle? Within its coverage area. It is known that... The larger the value, the more points the covering circle covers within the target area. The total number of points covered by the covering circle is:

[0158]

[0159] Meanwhile, considering that two covering circles may simultaneously cover the same point pi, another variable gi is introduced, and the formula is as follows:

[0160]

[0161] At this point, the total number of points covered by the covering circle is:

[0162]

[0163] Therefore, the coverage rate C of the covering circle can be obtained, and its formula is as follows:

[0164]

[0165] However, to avoid the covering circle covering areas that do not require detection, thus wasting observation resources, it should be avoided to cover points pi∈E. Therefore, a variable P is introduced as a penalty value, expressed in the following formula: The compensation coefficient is:

[0166]

[0167] In summary, the fitness function of the particle swarm optimization algorithm can be obtained as follows:

[0168]

[0169] 2. Cluster the sensor nodes:

[0170] Assuming to use If a flapping-wing robot is to patrol and monitor a target area, the generated coverage circle needs to be divided into m regions. This paper divides these regions by clustering the nodes. Each of the major regions.

[0171] 2.1 k-means clustering:

[0172] k-means clustering is an unsupervised learning algorithm widely used in data mining and machine learning. Its basic idea is to divide the samples in the dataset into k clusters, so that the distance between samples within the same cluster is as small as possible, while the distance between samples in different clusters is as large as possible.

[0173] The standard k-means clustering method is as follows: First, randomly select k initial cluster centers. Calculate the Euclidean distance from all points to the cluster centers and assign each point to the cluster with the smallest Euclidean distance. Calculate the average distance of all points within each cluster to obtain k new cluster centers. Calculate the Euclidean distance from all points to the cluster centers again and assign all points to the nearest clusters again. Repeat this process until the selected cluster centers no longer change. The specific process is as follows: Figure 1 As shown.

[0174] a) Improved k-means clustering method:

[0175] Standard k-means clustering is highly sensitive to the choice of initial cluster centers, which significantly impacts the convergence of subsequent algorithms. Therefore, improvements to k-means clustering are necessary. This can be achieved by introducing an evaluation function. The method optimizes the selection of initial cluster centers. The expression is as follows:

[0176]

[0177] In the formula, Used to represent each cluster center in each cluster; Represents a node To the cluster center The distance is ensured through algorithm optimization. The result with the smallest value is used as the initial cluster center, which can improve the clustering effect of the k-means algorithm.

[0178] 3. Generation of the cruising path:

[0179] 3.1 Genetic Algorithm Modeling:

[0180] Genetic Algorithm (GA) is an algorithm inspired by the evolutionary patterns of organisms in nature. It is a method for searching for optimal solutions by simulating the natural evolutionary process. This algorithm uses mathematical methods and computer simulations to transform the problem-solving process into processes similar to the crossover and mutation of chromosomes and genes in biological evolution. Genetic algorithms have advantages such as strong global search capabilities, wide applicability, and high flexibility. Based on these advantages, this paper uses a genetic algorithm to solve the problem of sequential connection of nodes within each cluster.

[0181] To efficiently perform genetic operations, the chromosomes in the algorithm must first be encoded. This is done by establishing a sequence of node connections. In the algorithm, each node is considered a gene unit, like a chromosome. For example, a sequence... = Indicates from node The connection order is used as the chromosome in the genetic algorithm to perform subsequent genetic operations.

[0182] The crossover operator in a genetic algorithm is represented as follows: Given two parents, P1 and P2, first, two crossover points are randomly selected from the parents, retaining both parents. Two offspring, F1 and F2, are obtained from the middle segment between these two crossover points. At this point, apart from the nodes in these two middle segments, the other nodes in the offspring are not yet determined. For the undetermined cities in F1, first, the cities in P2 are sequentially extracted from the second crossover point according to their relative order. Then, the cities appearing in the middle segment of F1 in this sequence are deleted. Finally, the remaining subsequence of P2 is filled into the corresponding positions of F1 from the second crossover point. F2 is obtained in the same way.

[0183] The mutation operator in a genetic algorithm is represented by randomly selecting two nodes in a sequence and swapping their positions.

[0184] Furthermore, considering the flight dynamics of the UAV, its cruise path should be constrained by its turning angle. To ensure the flight safety of the UAV, the curvature of its flight path should be limited. At the same time, considering the energy consumption of the UAV, the length of the generated path should be as small as possible. Therefore, the fitness function of the genetic algorithm should comprehensively consider the path length and the turning problem.

[0185] The fitness function is derived from the quotient of the path length and the cosine of the turning angle, as follows:

[0186]

[0187] This is used to encode and solve the connection order problem of the covering circle, which in turn allows us to find the shortest path length of the cruise cluster region.

[0188] Because standard genetic algorithms are prone to premature convergence and getting stuck in local convergence, improvements are made to genetic algorithms by introducing multi-population competition to increase population diversity and thus optimize the local convergence problem.

[0189] 4. Simulation verification:

[0190] 4.1 Set up a scenario for verification:

[0191] To verify the feasibility of the method, simulations were performed. The target area was assumed to be a 100x100 rectangular region, containing an irregular quadrilateral region that did not require scanning. Four drones were launched from the center of the target area to perform a cruise scan of the entire region. The drones' detection radius was set to 8.92, and the region's dispersion was set to 1.

[0192] An improved particle swarm optimization algorithm was used to generate a covering circle for the above region. The number of particles was set to 100, the number of iterations to 100, and the weight w was limited to the range of 0.4 to 0.8. The learning factor was... , A uniform value of 1.5 is used, and the velocity v is limited to the range of -2 to 2. The compensation coefficient is... Take 2.5; the final node generation result is as follows Figure 1-2 As shown.

[0193] The coverage rate calculated based on the node generation results is 90.76%, which shows that the node coverage rate generated by the algorithm is relatively high and meets the basic coverage requirements.

[0194] The improved k-means algorithm uses 4 cluster centers. The clustering results are visualized as follows: Figure 3 As shown, points of the same color represent the centers of circles belonging to the same cluster. The cluster profile coefficient is 0.65107, indicating that the clustering effect is quite ideal.

[0195] In the genetic algorithm, the crossover probability is 0.9, the mutation probability is 0.05, and the number of iterations is 1000. The result obtained is as follows. Figure 4-13 As shown.

[0196] 4.2 Comparison with traditional ox-plowing methods and cover-based cruise:

[0197] Compared with the traditional ox-plowing method, the method proposed in this paper has a lower coverage area but a significantly shorter cruising distance. This method has lower requirements for the drone's battery life and a shorter single cruising time, making it more suitable for mission scenarios that require high-frequency drone cruising, such as bird control and inspection.

[0198] 5. Summary:

[0199] This paper proposes a full-area coverage path planning method based on the concept of wireless sensor networks. By improving the particle swarm optimization algorithm, sensor nodes with detection ranges that cover the target area as much as possible are generated. The nodes are then clustered and partitioned by an improved k-means algorithm. Genetic algorithm is used to model and encode the nodes to determine the connection order. This generates the cruise paths of multiple UAVs. The method is then verified by simulation based on a specific scenario. Compared with the traditional ox-plowing inspection method, this method is more suitable for high-frequency cruise missions such as bird-scaring inspections. Specific implementation examples:

[0201] 1. Implementation environment setup:

[0202] The target monitoring area was set as a rectangular area of ​​100×100 square meters, including an irregular quadrilateral area that did not require scanning. Four drones were launched to conduct a cruise scan of the entire area. The drones flew at an altitude of [missing information]. Meters, sensor field of view angle .

[0203] 2. Applications of improved particle swarm optimization algorithm:

[0204] Initialize particle swarm: Generate 100 particles, each representing a potential cruising path within the target area. The position of each particle... and speed They are all randomly initialized.

[0205] Determine the parameters:

[0206] Inertia weight initial value final value .

[0207] Learning factor .

[0208] The speed range is limited to -2 to 2.

[0209] Iterative process: Perform 100 iterations to optimize the position and velocity of the particles.

[0210] Update inertia weights: in each iteration In the middle, the inertia weight decreases linearly according to the following formula. :

[0211]

[0212] Update particle velocity and position: for each particle In each iteration In the middle, its speed is updated using the following formula. :

[0213]

[0214] in, It is a random number in the interval [0,1]. It is a particle The optimal position of an individual It is the first The global optimal position in the next iteration.

[0215] Introducing inter-particle influence: updating particles Location When, add the inter-particle influence term :

[0216]

[0217] in, It is based on other particles to particles The vector obtained by calculating the influence.

[0218] Calculate inter-particle effects: for each particle Randomly select another particle Calculate according to the following formula :

[0219]

[0220] in The best historical position of another particle is randomly selected; For particles In time Location; For the first The historical best fitness function value of each particle; Let be the historical best fitness function value of the j-th particle.

[0221] Fitness evaluation: The position of each particle is evaluated using a fitness function that takes into account the coverage, efficiency and other relevant factors of the cruise path.

[0222] Update individual and global optimum: if particles In iteration The position in the middle If an individual has better fitness, then update its optimal value. and global optimal .

[0223] Termination condition: The algorithm terminates and outputs the global optimal solution when the maximum number of iterations is reached or other termination conditions are met. As the cruise path for drones.

[0224] Through this improvement, the particle swarm optimization algorithm not only considers the search history and global information of individual particles, but also increases the diversity and adaptability of the search through the mutual influence between particles, which helps to find a better cruising path.

[0225] 3. Inertia weight update:

[0226] In iteration At that time, inertial weight The calculation is as follows:

[0227]

[0228]

[0229] 4. Calculation of inter-particle influence variables:

[0230] Suppose a randomly selected optimal position in the particle's history. For (5,5), the particle Current location For (3,4), performance indicators , ,but:

[0231]

[0232] 5. Calculation of coverage rate of the cover circle:

[0233] 1) Determine the area to be patrolled. :

[0234] Total area of ​​target monitoring area square meters.

[0235] Assuming the non-scannable area accounts for 4% of the total area, then the area to be patrolled is... square meters.

[0236] 2) Calculate the area of ​​a single covering circle. :

[0237] Detection radius of drones ,in rice, .

[0238] rice.

[0239] Area of ​​a single covering circle square meters.

[0240] 3) Calculate the total number of circles to be covered. :

[0241] To achieve 96% coverage, we set A covering circle, making near .

[0242] .

[0243] 4) Number of circles covered by the rounding function:

[0244] Since the number of covered circles must be an integer, we take... A coverage circle is used to ensure that the coverage area exceeds 96%.

[0245] 5) Calculate coverage :

[0246] Assuming each covering circle can cover independently Area, with no overlap or omission.

[0247] Coverage The calculation formula is:

[0248]

[0249] because (Each point represents 1 square meter), simplified to:

[0250]

[0251] Substitution and Value:

[0252]

[0253] 6. Applications of improved k-means clustering algorithm:

[0254] Let the number of clusters be Total number of sensor nodes Each node To the cluster center average distance Then the evaluation function for:

[0255]

[0256] 7. Application of Genetic Algorithms: Genetic Algorithm Parameter Settings:

[0257] Crossover probability: 0.9

[0258] Mutation probability: 0.05

[0259] Number of iterations: 1000

[0260] fitness function Calculation, assumption , (No duplicate coverage), penalty value This is a value determined based on the number of points covering invalid regions or other rules; it is tentatively set here as follows. ,but:

[0261]

[0262] 8. Simulation verification:

[0263] UAV detection radius calculate:

[0264]

[0265] Data tables and technical effect demonstrations

[0266]

[0267] 10. Results Analysis:

[0268] Simulation results demonstrate that this invention achieves a coverage rate of 96% and a detection radius of 17.32 meters. With a simulated cruising distance of 500 meters and a drone cruising speed of 5 meters per second, the cruising time is 100 seconds. Compared to traditional methods, this invention offers significant advantages in improving coverage efficiency, reducing cruising time, and enhancing drone endurance.

[0269] It should be noted that, in this document, the terms "comprising," "including," and any other variations are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Specific examples have been used in this document to illustrate the principles and implementation methods of the present invention. These examples are merely for the purpose of helping to understand the method and core ideas of the present invention. The above descriptions are only preferred embodiments of the present invention. It should be pointed out that, due to the limitations of written expression and the objective existence of infinite specific structures, those skilled in the art can make several improvements, modifications, or variations without departing from the principles of the present invention, and can also combine the above technical features in an appropriate manner. These improvements, modifications, variations, or combinations, or the direct application of the concept and technical solution of the present invention to other situations without modification, should all be considered within the scope of protection of the present invention.

Claims

1. A method for planning bird-scaring cruising routes for flapping-wing UAVs based on WSN concepts, characterized in that, The method includes the following steps: a. Sensor nodes are generated in the target area using an improved particle swarm optimization algorithm. Each particle represents the center coordinates of a potential coverage circle. The optimal position of the coverage circle is determined through iterative optimization. Inter-particle influence variables are introduced to enhance the mutual learning ability between particles. b. The generated sensor nodes are clustered and partitioned using an improved k-means clustering algorithm. By optimizing the selection of the initial cluster centers, the nodes are divided into several clusters, and each cluster corresponds to the detection area of ​​a UAV. c. A genetic algorithm is used to generate traversal paths for sensor nodes within each cluster, simulating the natural evolution process. The path length and turning angle are optimized through crossover and mutation operations, taking into account the flight dynamics of the UAV. d. Conduct simulation verification, set the target area, number of UAVs, and detection radius parameters, and verify the feasibility and effectiveness of the proposed method through simulation experiments; e. Compare and analyze the effects of the proposed method and the traditional ox-plowing method on coverage cruise, and evaluate the coverage rate, cruise distance, endurance requirements and single cruise time indicators to determine the applicability of the method in specific mission scenarios; In the sensor node generation step, the coverage rate of the coverage circle The calculation formula is: in, The total number of covered circles; For all covering circles From 1 to Summation; : The set of points within the area to be patrolled; : Boolean variable, if point Located in the covering circle Within its coverage area, then If it is 1; otherwise, it is 0. For all points within the area to be scanned of Perform summation; where Point The number of times a device is repeatedly covered by multiple covering circles; Area of ​​the area to be patrolled; The dispersion of the region; The improved k-means clustering algorithm includes an evaluation function for cluster center selection. The calculation formula is as follows: in, The number of clusters; The total number of sensor nodes; Sensor Node Location; It is the first Cluster centers, Represents a node To the cluster center The distance.

2. The method according to claim 1, characterized in that, The particle velocity update formula in the improved particle swarm optimization algorithm is as follows: in, and These represent the particle velocities in the t-th and t+1-th iterations, respectively. Let be the inertia weight of the algorithm at the t-th iteration; and These represent the learning factors that influence the particle's individual optimal position and the group's optimal position, respectively. Represents a random number between 0 and 1. :particle In time Location; and These are the individual optimal location storage variables and the group optimal location storage variables, respectively.

3. The method according to claim 1, characterized in that, The improved particle swarm optimization algorithm includes inertia weights. Update formula: in, Let be the inertia weight of the algorithm at the t-th iteration; It is the maximum value of the inertia weight. It is the minimum value of the inertia weight. It is the maximum number of iterations. This is the current iteration number.

4. The method according to claim 1, characterized in that, Introducing inter-particle influence variables in particle swarm optimization Enhance the mutual learning ability between particles. The calculation formula is: in The best historical position of another particle is randomly selected; For particles In time Location; For the first The historical best fitness function value of each particle; Let be the historical best fitness function value of the j-th particle.

5. The method according to claim 1, characterized in that, Fitness function of particle swarm optimization algorithm The calculation formula is: in, Point Is it within the covering circle? Within its coverage area, Point The number of times the covered circle is decreased by one; : Represents the penalty value; : The area of ​​the target region where the battery needs to be extended; : Dispersion of the region.

6. The method according to claim 1, characterized in that, The detection radius of the UAV in the simulation verification step The calculation formula is: in, It is the flight altitude. It refers to the angle of the sensor's field of view.

7. The method according to claim 1, characterized in that, The genetic algorithm further includes a mechanism for adaptively adjusting the crossover and mutation probabilities to optimize the algorithm's search performance; this mechanism includes: a. Define the initial value of the crossover probability as 0.9, and dynamically adjust the crossover probability according to the iteration progress of the algorithm and the quality of the solution; b. Define the initial value of the mutation probability as 0.05, and dynamically adjust the mutation probability according to the changes in population diversity during the algorithm search process; c. Design an adaptive adjustment strategy to appropriately increase the mutation probability to maintain the diversity of solutions when the population shows signs of convergence; d. The adaptive adjustment strategy also includes appropriately increasing the crossover probability during the local search phase to promote the generation of excellent solutions.

8. The method according to claim 1, characterized in that, The simulation verification process further includes the following steps: a. Define the target area as a rectangular area of ​​specific size, and divide the area into discrete grids, with each grid serving as a discrete point, to simulate the drone's cruise and monitoring; b. Set the detection radius of the UAV so that the UAV can cover a certain range of grid around each discrete point; c. Based on the number of drones and the layout of the target area, assign a patrol grid area to each drone to achieve coordinated patrol; d. Simulate the drone cruising above the grid along the generated traversal path using simulation software, collect coverage data, and verify the effectiveness of the path planning; e. Simulation verification also includes evaluating key indicators such as coverage, cruising distance, and battery life of the drone's cruise route to ensure the practicality and efficiency of the planned path.

Images