Double-layer improved particle swarm unmanned aerial vehicle path planning method, system, device and medium

Abstract

The application belongs to the field of unmanned aerial vehicle path planning, and provides a double-layer improved particle swarm unmanned aerial vehicle path planning method, system, device and medium, which comprises the following steps: obtaining the starting point coordinates and the ending point coordinates of the unmanned aerial vehicle path to perform coordinate conversion, determining the path maximum boundary under a new coordinate system, and dividing the path maximum boundary into groups of random boundaries; creating a path particle swarm archive library within each group of random boundaries and iteratively performing path optimization; after a preset number of iterations, performing boundary optimization once, then iteratively performing path optimization within each group of processed random boundaries and then performing boundary optimization again until a stop condition is met, and obtaining an optimal boundary; based on the result of the last boundary optimization, constructing a combined path particle swarm and a combined path particle swarm archive library, and iteratively performing path optimization based on the combined path particle swarm until a maximum number of iterations is reached, and obtaining an optimal path.

Description

Technical Field

[0001] This invention belongs to the field of UAV path planning technology, specifically relating to a two-layer improved particle swarm UAV path planning method, system, equipment, and medium. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] In existing UAV 3D path planning technologies, particle swarm optimization (PSO) algorithms are widely used because they can avoid the curse of dimensionality. The combination of parameters such as inertia weight, individual learning factor, and social learning factor in PSO algorithms has a significant impact on the optimization effect. However, existing technologies mostly adopt fixed or single parameter adjustment strategies, which cannot dynamically adapt to path optimization requirements according to the iterative process, thus limiting the algorithm's global exploration and local development capabilities. Moreover, because traditional PSO algorithms lack effective constraints on the iteration range, particles tend to converge prematurely in non-optimal regions during 3D path planning. When particles get stuck in local optima, traditional PSO algorithms do not have a targeted guidance mechanism, causing the generated paths to frequently deviate from the potential optimal trajectory. This makes it difficult to guarantee the optimality of each planning result, seriously affecting the UAV's navigation efficiency and safety.

[0004] To address these shortcomings, existing population optimization mechanisms often achieve combinatorial optimization by introducing different algorithms or combining them with other algorithms. For example, the gravity search algorithm can be used to enhance local exploration capabilities. However, this approach does not consider the correlation logic between nodes on different paths when calculating gravity. Instead, it calculates gravity using the Euclidean distance between paths, which leads to the averaging of node guidance and an inability to adapt to the differentiated needs of different nodes. Summary of the Invention

[0005] To address the aforementioned problems, this invention proposes a two-layer improved particle swarm unmanned aerial vehicle (UAV) path planning method, system, device, and medium. This invention can solve the problem that traditional particle swarm algorithms frequently get trapped in local optima in UAV path planning, while improving the efficiency of path planning.

[0006] According to some embodiments, the first aspect of the present invention provides a two-layer improved particle swarm unmanned aerial vehicle path planning method, which adopts the following technical solution:

[0007] A two-layer improved particle swarm unmanned aerial vehicle (UAV) path planning method includes:

[0008] Obtain the start and end coordinates of the UAV path, perform coordinate transformation, determine the maximum boundary of the path in the new coordinate system, and then divide the maximum boundary of the path. Group random boundary;

[0009] Within each set of random boundaries, a path particle swarm archive is created and path optimization is iteratively performed. After a set number of iterations, a boundary optimization is performed to check whether the path particles within each set of optimized random boundaries exceed the boundary. Path particles that exceed the boundary are handled for out-of-bounds behavior. The path particles within each set of processed random boundaries are evaluated and the path particle swarm archive is updated. Then, path optimization is iteratively performed within each set of processed random boundaries, and boundary optimization is performed again until the stopping condition is met, and the optimal boundary is obtained.

[0010] Based on the result of the last boundary optimization, a combined path particle swarm and a combined path particle swarm archive are constructed. Path optimization is performed iteratively based on the combined path particle swarm until the maximum number of iterations is reached to obtain the optimal path.

[0011] Furthermore, the starting and ending coordinates of the obtained UAV path are transformed to determine the maximum boundary of the path in the new coordinate system, and the maximum boundary of the path is then divided. Group random boundaries include:

[0012] Load the map, obtain the starting and ending coordinates of the drone path in the map coordinate system, take the line segment connecting the starting and ending points in the two-dimensional plane as the x-axis, the perpendicular bisector of the line segment in the two-dimensional plane as the y-axis, and keep the z-axis unchanged to obtain the new coordinate system and the starting and ending coordinates of the drone path in the new coordinate system.

[0013] Transform the coordinates of the four vertices of the top view of the map in the map coordinate system to obtain the maximum boundary of the path in the new coordinates.

[0014] In the new coordinate system, the maximum path boundary is divided into... A group of random boundaries is used as a swarm of boundary particles, with each random boundary being a boundary particle.

[0015] Further, the process of creating a path particle swarm archive within each set of random boundaries and iteratively optimizing paths, after a set number of iterations, performing boundary optimization, checking whether path particles within each optimized set of random boundaries exceed the boundaries, handling path particles exceeding the boundaries, evaluating path particles within each set of processed random boundaries, and updating the path particle swarm archive, includes:

[0016] Within each set of random boundaries, initialize the positions of all path particles in the new coordinate system to obtain the initial path particles, and initialize m sets of path particle groups and the corresponding path particle group archive.

[0017] Evaluate all initial path particles within each set of random boundaries, determine the optimal set of path particles at the current moment, and add it to the corresponding path particle swarm archive to obtain the optimized path particle swarm. Then, classify all path particles in the optimized path particle swarm into levels to obtain the cost evaluation result of the path optimization.

[0018] Based on the cost evaluation results after each path optimization, a dynamic strategy for updating the velocity weight parameters of path particles is selected, and the updated velocity weight parameters are used to update the velocity of path particles within each set of random boundaries.

[0019] After each velocity and position update, all path particles and their corresponding levels must be evaluated.

[0020] The above process is iteratively repeated for all path particle swarms within each initial random boundary. After a set number of iterations, boundary optimization is performed to obtain the optimized random boundary. If any particles within the boundary exceed the boundary, they are processed. The cost of internal path particles is re-evaluated, and the corresponding path particle swarm archive within each boundary is updated.

[0021] Furthermore, a dynamic speed weight parameter update strategy is selected based on the cost evaluation results after each path optimization, specifically as follows:

[0022] The path particle swarm within each random boundary randomly selects a policy from the policy pool. After each iteration, if the number of Pareto optimal path particles newly obtained by each path particle swarm is higher than that of the previous generation, the policy is retained and the iteration continues. If the number of Pareto optimal path particles newly obtained after the iteration is less than that of the previous generation, other policies are selected from the policy pool and the iteration continues.

[0023] Furthermore, the step of checking whether the path particles within each optimized random boundary exceed the boundary, performing boundary overflow processing on path particles exceeding the boundary, evaluating the path particles within each processed random boundary, and updating the path particle swarm archive includes:

[0024] Check whether the path particles within the optimized random boundary of each group exceed the boundary.

[0025] If the Y value of a path particle that exceeds the boundary range is close to the left boundary value of the boundary particle, then its Y value is set to that left boundary value. If the Y value of a path particle that exceeds the boundary range is close to the right boundary value of the boundary particle, then its Y value is set to that right boundary value.

[0026] Re-evaluate the cost of all path particles within the processed random boundary and update the path particle swarm archive.

[0027] Furthermore, the construction of a combined path particle swarm and a combined path particle swarm archive based on the result of the last boundary optimization, and the iterative path optimization based on the combined path particle swarm until the maximum number of iterations is reached to obtain the optimal path, includes:

[0028] Based on the results of the last boundary optimization, construct a combined path particle swarm, a comparative path particle swarm, and a corresponding path particle swarm archive.

[0029] Path optimization is performed iteratively using combined path particle swarm optimization and comparative path particle swarm optimization until the maximum number of iterations is reached, at which point the optimal path is obtained.

[0030] Furthermore, based on the results of the final boundary optimization, a combined path particle swarm and a comparative path particle swarm, along with their corresponding path particle swarm archives, are constructed, including:

[0031] Based on the results of the final boundary optimization, the top three boundary particles with the highest fitness values ​​(from lowest to highest) are selected, and their boundaries are then used to determine the boundaries of each particle. The top 30% of path particles are defined as the initial optimal path particles;

[0032] It also determines whether all initial optimal path particles exceed the optimal boundary range, performs out-of-bounds processing on initial optimal path particles that exceed the optimal boundary range, and obtains the processed initial optimal path particle swarm.

[0033] The generation of some new path particles in the optimal boundary is defined as the initial random particle swarm.

[0034] The processed initial optimal path particle swarm and the initial random particle swarm are used to form a combined path particle swarm with a population size of Np = NpB. ;

[0035] By replicating the combined path particle swarm, we obtain the comparison path particle swarm. ;

[0036] Create separate path particle swarm archives for the combined path particle swarm and the comparative path particle swarm, respectively, to store path particles with PraetoRank=1 in the two sets of path particle swarms.

[0037] According to some embodiments, the second aspect of the present invention provides a two-layer improved particle swarm unmanned aerial vehicle path planning system, which adopts the following technical solution:

[0038] A two-layer improved particle swarm unmanned aerial vehicle (UAV) path planning system includes:

[0039] The path boundary delineation module is configured to acquire the start and end coordinates of the UAV path, perform coordinate transformation, determine the maximum path boundary in the new coordinate system, and delineate the maximum path boundary. Group random boundary;

[0040] The boundary optimization module is configured to create a path particle swarm archive within each set of random boundaries and iteratively perform path optimization. After a set number of iterations, a boundary optimization is performed to check whether the path particles within each set of optimized random boundaries exceed the boundary. Path particles that exceed the boundary are handled for out-of-bounds behavior. The path particles within each set of processed random boundaries are evaluated and the path particle swarm archive is updated. Then, path optimization is iteratively performed within each set of processed random boundaries, and boundary optimization is performed again until the stopping condition is met to obtain the optimal boundary.

[0041] The path optimization module is configured to construct a combined path particle swarm and a combined path particle swarm archive based on the result of the last boundary optimization, and perform path optimization iteratively based on the combined path particle swarm until the maximum number of iterations is reached to obtain the optimal path.

[0042] According to some embodiments, a third aspect of the present invention provides a computer-readable storage medium.

[0043] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the two-layer improved particle swarm unmanned aerial vehicle path planning method as described in the first embodiment above.

[0044] According to some embodiments, a fourth aspect of the present invention provides a computer device.

[0045] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the two-layer improved particle swarm unmanned aerial vehicle path planning method as described in the first embodiment above.

[0046] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0047] This invention employs a two-layer particle swarm optimization framework (outer optimization boundary and inner optimization boundary path). By introducing dynamic boundary constraints, it restricts the particles to iterate within a finite space containing potential optimal solutions, thus solving the problem of trajectory deviation from expectations.

[0048] This invention designs a competitive elimination mechanism that divides the population into subpopulations and assigns different parameter combinations. By reducing the magnitude of the cost function, the optimal selection is dynamically achieved, thus solving the problem of rigid parameter tuning.

[0049] This invention calculates the gravitational value of corresponding nodes along a path by independently calculating the distance between them. The overall gravitational force of the path is then obtained by combining these gravitational values. This achieves a pairwise calculation of node distance and gravitational force, aligning with the ordered nature of path nodes. Furthermore, it introduces the escape mechanism of the Kepler optimization algorithm, improving the ability of path particles to escape local optima. Attached Figure Description

[0050] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0051] Figure 1 This is a flowchart of selecting the optimal boundary in an embodiment of the present invention;

[0052] Figure 2 This is a schematic diagram of the lower boundary division of the map top view in an embodiment of the present invention;

[0053] Figure 3 This is a comparison of the path under a map top view between the method described in this embodiment of the invention and other unconstrained intelligent optimization algorithms. Detailed Implementation

[0054] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0055] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0056] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0057] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0058] Example 1

[0059] This embodiment provides a two-layer improved particle swarm unmanned aerial vehicle (UAV) path planning method. This embodiment uses the application of this method to a server as an example for illustration. It is understood that this method can also be applied to terminals, and can also be applied to systems including terminals, servers, and other components, and implemented through interaction between the terminal and the server. The server can be an independent physical server, a server cluster composed of multiple physical servers, or a distributed system. It can also be a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network servers, cloud communication, middleware services, domain name services, CDN security services, and big data and artificial intelligence platforms. The terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, etc., but is not limited to these. The terminal and server can be directly or indirectly connected via wired or wireless communication, which is not limited herein. In this embodiment, the method includes the following steps:

[0060] Step S1: Obtain the starting and ending coordinates of the UAV path, perform coordinate transformation, determine the maximum boundary of the path in the new coordinate system, and divide the maximum boundary of the path. Group random boundary;

[0061] Step S2: Create a path particle swarm archive within each set of random boundaries and iteratively perform path optimization. After a set number of iterations, perform boundary optimization once to check whether the path particles within each set of optimized random boundaries exceed the boundary. Perform out-of-bounds processing on path particles that exceed the boundary. Evaluate the path particles within each set of processed random boundaries and update the path particle swarm archive. Then iteratively perform path optimization within each set of processed random boundaries and perform boundary optimization again until the stopping condition is met to obtain the optimal boundary.

[0062] Step S3: Based on the result of the last boundary optimization, construct a combined path particle swarm and a combined path particle swarm archive. Iterate the path optimization based on the combined path particle swarm until the maximum number of iterations is reached to obtain the optimal path.

[0063] Step S1: Obtain the starting and ending coordinates of the UAV path, perform coordinate transformation, determine the maximum boundary of the path in the new coordinate system, and divide the maximum boundary of the path. Group random boundaries include:

[0064] Step S1.1: Load the map, set the obstacles as a series of cylinders; obtain the starting coordinates of the drone's path in the map coordinate system. and endpoint coordinates Using the line segment connecting the start and end points in the two-dimensional plane as the x-axis, the perpendicular bisector of the line segment as the y-axis, and keeping the z-axis unchanged, we obtain a new coordinate system and the start and end coordinates of the UAV path under the new coordinate system.

[0065] Specifically, the coordinate transformation formula from the map coordinate system to the new coordinate system is as follows:

[0066] (1);

[0067] in, θ The rotation angle is... , , , , These are the coordinates in the world coordinate system.

[0068] Step S1.2: Transform the coordinates of the four vertices of the top view of the map in the map coordinate system to obtain the maximum path boundary in the new coordinate system, including:

[0069] The coordinates of the four vertices in the top view of the map Perform a coordinate transformation to obtain the coordinates of the four vertices in the new coordinate system. ;

[0070] By comparing the y-coordinate values ​​of the four vertices, we can obtain the range of y-coordinate values. ;

[0071] The range of ordinate values ​​is used as the maximum boundary of the path in the new coordinate system, and the point with the largest ordinate value is used as the upper limit of the maximum boundary of the path. The point with the smallest y-axis coordinate is used as the lower bound of the maximum path boundary. ;

[0072] Step S1.3: In the new coordinate system, divide the maximum path boundary into... A group of random boundaries is used as a boundary particle swarm, with each random boundary as a boundary particle, as follows:

[0073] (2);

[0074] (3);

[0075] in, Indicates the first Group of uniform random boundaries, total A uniform random boundary; Indicates the first Group of non-uniform random boundaries, total Group of uniform random boundaries, boundary partitioning results are as follows Figure 2 As shown.

[0076] Step S2: Create a path particle swarm archive within each set of random boundaries and iteratively perform path optimization. After a set number of iterations, perform boundary optimization once, checking whether the path particles within each optimized random boundary exceed the boundary. Path particles exceeding the boundary are handled for out-of-bounds errors. Evaluate the path particles within each processed random boundary and update the path particle swarm archive. Then, iteratively perform path optimization within each processed random boundary and perform boundary optimization again until the stopping condition is met, obtaining the optimal boundary, such as... Figure 1 As shown, it includes:

[0077] Step S2.1: Initialize the positions of all path particles in the new coordinate system within each set of random boundaries to obtain the initial path particles, and initialize m sets of path particle swarms and their corresponding path particle swarm archives, including:

[0078] The path particle swarm archive is used to store path particles with a Paretorank of 1. After each path particle cost is evaluated and Paretorank is determined within each set of random boundaries, the path particle swarm archive within each set of random boundaries is updated. If the boundary is updated after the path particle swarm is updated, the particle cost is evaluated, and the Paretorank is determined, then the path particle swarm archive does not need to be updated. The archive of the path particle swarm within each set of boundaries is updated after the boundary is updated and the cost of all path particles within it is evaluated.

[0079] The update operation is as follows: store the path particles with the current Paretorank of 1 into the path particle swarm archive. After adding them, perform non-dominated sorting with the path particles in the original path particle swarm archive. Remove path particles in the path particle swarm archive that are dominated by other path particles. If the number of particles in the path particle swarm archive exceeds the preset capacity, sort them according to the crowding distance from smallest to largest, and remove the path particles ranked in the top Se position.

[0080] (4);

[0081] in: Indicates the number of objective functions. Represents path particles in the archive. Crowded distance, Indicates the first The path particle Fitness values ​​for each objective function Indicates the first The path particle Fitness values ​​for each objective function Indicates the current number The maximum value of each objective function. Indicates the current number The minimum value of an objective function.

[0082] Within each set of random boundaries, a path particle swarm is set. The iterative update path has a population size of . The population initialization of the path particle swarm is performed by the following formula:

[0083] (5);

[0084] in, , For the left and right boundaries of each set of boundaries, The number of nodes for each path particle. The minimum safe flight altitude and the maximum safe flight altitude, The mapping factor is specifically:

[0085] (6);

[0086] During the initialization of the path particle swarm, the path particles are obtained proportionally from the three different forms of mapping factors mentioned above. Indicates the number of mappings. Indicates the first The function value of the second-order tentative mapping is calculated using the following formula:

[0087] (7);

[0088] By setting different initial values This forms multiple sets of chaotic sequences.

[0089] Step S2.2: Evaluate all initial path particles within each set of random boundaries, determine the optimal set of path particles at the current time, and add it to the corresponding path particle swarm archive to obtain the optimized path particle swarm. Then, classify all path particles in the optimized path particle swarm into different levels to obtain the cost evaluation results after path optimization, including:

[0090] The position coordinates of the initial path particles within each random boundary are converted to the map coordinate system for evaluation. The cost of each initial path particle is evaluated. The set of Pareto optimal solutions of all path particles in the current time path particle group is selected and added to the corresponding path particle group archive. The path particle group archive is checked to see if it exceeds the preset capacity. Path particles that exceed the preset capacity in the archive are removed according to formula (4).

[0091] The following cost functions are used to evaluate the path solution:

[0092] The path length cost function is as follows:

[0093] (8);

[0094] in, This represents the minimum distance between particle nodes along the path. For the path particle's first The path node and the first The distance between path nodes is calculated using the following formula:

[0095] (9);

[0096] in, This represents the number of nodes that make up the path particles.

[0097] The path collision cost function is as follows:

[0098] (10);

[0099] in, The number of obstacles, Path node With path nodes The collision discrimination formula between the constituent vector and the obstacle is expressed as follows:

[0100] (11);

[0101] in, To maintain a safe distance between the drone and obstacles, Let the radius be the obstacle. Path node With path nodes Composition vector and obstacle center The distance between them.

[0102] The path height cost function is as follows:

[0103] (12);

[0104] in, and for and Convert to values ​​in a map coordinate system.

[0105] The path corner cost function is as follows:

[0106] (13);

[0107] Among them, path turning angle The calculation formula is:

[0108] (14);

[0109] in, Path particles Path nodes With path nodes Path segment vector, Path particles Path nodes With path nodes Path segment vector.

[0110] The cost evaluation function described above is used to evaluate all initial path particles in the path particle swarm to obtain the optimized path particle swarm. All path particles in the optimized path particle swarm are then classified into different levels to obtain the cost evaluation results after path optimization.

[0111] The division method is as follows:

[0112] The Pareto optimal path particles in the optimized path particle swarm are taken as non-dominated solutions, and the remaining path particles are taken as dominated solutions.

[0113] All non-dominated solutions in the optimized path particle swarm Set to 1, then re-compare non-dominated solutions from all dominant solutions and set their parteorank to 2, and so on until all path particles are assigned to a single rank.

[0114] Step S2.3: Dynamically select the velocity weight parameter update strategy for path particles based on the cost evaluation results after each path optimization, and update the velocity of path particles within each set of random boundaries using the updated velocity weight parameters, including:

[0115] The speed weight parameter update strategy is dynamically selected based on the cost evaluation results after each path optimization. In other words, an evaluation mechanism is used to dynamically select the speed weight parameter update strategy. Specifically, for each group of path particle swarms within the random boundary, a strategy is randomly selected from the strategy pool. After each iteration, if the number of Pareto optimal path particles newly obtained by each group of path particle swarms is higher than that of the previous generation, the strategy is retained and the iteration continues. If the number of Pareto optimal path particles newly obtained after the iteration is less than that of the previous generation, other strategies are selected from the strategy pool and the iteration continues.

[0116] For the weight parameters of the speed update formula , , The selection of the parameters and the setting of the speed weight parameters for updating the policy pool are shown in Table 1.

[0117] Table 1. Policy Pool for Updating Velocity Weight Parameters of Path Particles

[0118]

[0119] in, This represents the number of iterations for the current boundary particle swarm or path particle swarm. Given its maximum number of iterations, , Inertia factors Initial and final values, , These represent the initial and final values ​​of the individual learning factor. , Let the initial and final values ​​of the individual social learning factor be given, and let the particle swarm optimization be applied to each group of boundary paths. , , The strategies are respectively: ;

[0120] The velocity weight parameters are updated based on a defined velocity weight parameter update strategy. The updated velocity weight parameters are then used to update the velocity of path particles within each group of random boundaries. The velocity update formula for each group of path particles within random boundaries is as follows:

[0121] (15);

[0122] in, For individual inertia weights, For the first The path particle in the first... Speed ​​at the next iteration , As a random factor, It is the first The path particle Acceleration during the next iteration It is the acceleration weight of the particle. It is the social weight of the path particles. Posbest is the guiding particle selected from the archive by roulette wheel. If there are no collision-free particles after the boundary is updated, the corresponding Posbest within the boundary is selected from the archive of the path particle group within the current optimal boundary.

[0123] Path particles The acceleration at the current iteration number is expressed as:

[0124] (16);

[0125] in, To constitute path particles The The acceleration of each path node is calculated using the following formula:

[0126] (17);

[0127] in: Path particles quality Path particles No. The path node is affected by particles from other paths. The gravitational force of each path node is calculated using the following formulas:

[0128] (18);

[0129] (19);

[0130] in, For path particle swarm Gravity coefficient at the next iteration For the first Path particles in the next iteration and No. The distance between each path node To avoid the denominator being 0 for the minus term, Np represents the population size. and Path particles , The unnormalized mass at the current moment is calculated using the following formula:

[0131] (20);

[0132] in, The Pareto order of path particles in the path particle swarm. It is a mass discrimination function, based on path particles. crowded distance The formula for crowding distance is as follows:

[0133] (twenty one);

[0134] in: Indicates the number of objective functions. Represents path particles Crowded distance, Indicates the first The path particle The fitness value of each objective function. Indicates the first The path particle The fitness value of each objective function. Indicates the current number The maximum value of each objective function. Indicates the current number The minimum value of an objective function.

[0135] In calculation At that time, based on the crowding distance of path particles, the mass discrimination function value of the path particle with the largest crowding distance among path particles with ParetoRank=1 is calculated. =1, the mass discrimination function value of the path particle with the second largest crowding distance. =2, if u is the path particle with the smallest crowding distance in this Pareto level, and the mass discrimination function value is 2. =u, then among the particles with ParetoRank=2, the mass discrimination function value of the path particle with the largest crowding distance is... =u+1, and so on, until all particles... A value was assigned to it.

[0136] Posbest's selection probability is related to its crowding distance in the archive, and the selection probability formula is:

[0137] (twenty two);

[0138] in, For the first in the archive The probability of choosing a path particle. It is the corresponding crowding distance. Set the preset capacity for the archive.

[0139] Step S2.4: The formula for updating the position of path particles within each group of random boundaries is as follows:

[0140] (twenty three);

[0141] in, For path particle swarm Path particles in the next iteration Location, Path particles No. Position at the next iteration As an escape indicator, if the mass of the particle swarm along each random boundary path is [value missing] in the early stages of iteration... If the smallest path particle remains unchanged for three consecutive generations, then... A value of -1 causes the path particle to escape from the local optimum; otherwise... The value is 1.

[0142] After each velocity and position update, all path particles and their corresponding levels must be evaluated.

[0143] Step S2.5: Iterate through steps S2.1-S2.4 for all path particle swarms within each initial random boundary. After a set number of iterations, perform boundary optimization to obtain the optimized random boundary. Check if any path particles within the optimized random boundary have exceeded the boundary. Handle any out-of-bounds path particles, re-evaluate the cost of internal path particles, and update the path particle swarm archive corresponding to each group of processed random boundaries, including:

[0144] After the path particle swarm iteration has been performed a set number of times, another boundary optimization is performed to determine the optimal random boundary solution at the current moment.

[0145] The cost results after boundary optimization are obtained by evaluating each set of random boundaries using the following cost function:

[0146] (twenty four);

[0147] in, These are the weighting coefficients. , For the boundary width, For each set of random boundaries, the mass The weighted average of all cost functions for the path particle with the minimum cost is given by the following formula:

[0148] (25);

[0149] in, These are different weighting coefficients.

[0150] Based on the cost evaluation results after boundary optimization, a dynamic velocity weight parameter update strategy is selected. The updated velocity weight parameters are then used to update the velocities of boundary particles in the boundary particle swarm, including:

[0151] The velocity weight parameter update strategy is dynamically selected based on the cost evaluation results after each boundary optimization. In other words, an evaluation mechanism is used to dynamically select the velocity weight parameter update strategy. Specifically, the boundary particle swarm randomly selects a velocity weight parameter update strategy from the strategy pool. After each iteration, if the width of the optimal boundary particle in the boundary particle swarm is less than or equal to the width of the optimal boundary particle in the previous generation, the velocity weight parameter update strategy is retained and the iteration continues. Otherwise, another velocity weight parameter update strategy is selected from the strategy pool and the iteration continues.

[0152] The velocity weight parameters are updated based on a defined velocity weight parameter update strategy. The updated velocity weight parameters are then used to update the velocities of boundary particles in the boundary particle swarm. The velocity update formula is as follows:

[0153] (26);

[0154] in: This represents the current boundary particle swarm iteration number. For the individual history fitness of boundary particles The position with the smallest value The historical fitness in the boundary particle swarm The position of the particle with the smallest value. When the globally optimal boundary particle remains unchanged for three consecutive generations or reaches the preset number of iterations, the globally optimal boundary particle determined at this point is taken as the final determined boundary.

[0155] The position update formula for the boundary particle swarm is as follows:

[0156] (27);

[0157] in, It is the first boundary particles Location, It is the first boundary particles Location, It is the first boundary particles The speed.

[0158] Before updating the position of the boundary particle, obtain the mass of all collision-free path particles inside it. The minimum path particle's average X-axis distance to its boundary particles is calculated using the following formula:

[0159] (28);

[0160] in, Finding the best boundary The average X-axis distance corresponding to each set of boundaries at that time. The path particle with the smallest mass and no collisions within the boundary particle. The Y-coordinate of each path node, if there are no collision-free particles within the boundary particle. and The average X-axis distance of the boundary particle remains consistent throughout the iteration. If there are no collision-free path particles within a boundary particle obtained in the initial initialization, the average X-axis distance of this boundary particle is set to infinity. If the average X-axis distance of the previous generation of a boundary particle is smaller than that of the current generation during the iteration process, the boundary particle does not undergo position update in this iteration. Instead, the position of the boundary particle is obtained through the mutation formula, which is:

[0161] (29);

[0162] in, For the boundary particle to first obtain this smaller average X-axis distance, the path particle with the smallest mass and no collisions inside is the first... The Y-coordinate of each path node. To ensure a safety margin, a certain search space still exists after the boundary particles mutate.

[0163] Step 2.6: After updating the boundary particle positions, check whether any internal path particles exceed the boundary range. For path particles exceeding the boundary range, perform out-of-bounds processing, specifically:

[0164] If the Y value of a path particle that exceeds the boundary range is close to the left boundary value of the boundary particle, then its Y value is set to that left boundary value. If the Y value of a path particle that exceeds the boundary range is close to the right boundary value of the boundary particle, then its Y value is set to that right boundary value.

[0165] Re-evaluate the cost of all path particles within the processed random boundary and update the path particle swarm archive.

[0166] Step S2.7: Iterate the path optimization once within each optimized boundary, then perform boundary optimization again, until the stopping condition is met to obtain the optimal boundary, such as... Figure 3 As shown, the red line represents the path update within the boundary; while the yellow line and the path line represent the path update results in the other two cases without boundaries.

[0167] Step S3: Based on the results of the last boundary optimization, construct a combined path particle swarm optimization (PSO) system and a combined path PSO archive. Iterate through the PSO system to find the optimal path until the maximum number of iterations is reached, obtaining the optimal path, including:

[0168] Step S3.1: Based on the results of the last boundary optimization, construct the combined path particle swarm and the comparative path particle swarm, as well as the corresponding path particle swarm archive, including:

[0169] Based on the results of the final boundary optimization, the top three boundary particles with the highest fitness values ​​(from lowest to highest) are selected, and their boundaries are then used to select the particles within the boundary. The top 30% of path particles are defined as the initial optimal path particles;

[0170] It also determines whether all initial optimal path particles exceed the optimal boundary range, performs out-of-bounds processing on initial optimal path particles that exceed the optimal boundary range, and obtains the processed initial optimal path particle swarm.

[0171] The generation of some new path particles in the optimal boundary is defined as the initial random particle swarm.

[0172] The processed initial optimal path particle swarm and the initial random particle swarm are used to form a combined path particle swarm with a population size of Np = NpB. ;

[0173] By replicating the combined path particle swarm, we obtain the comparison path particle swarm. ;

[0174] Create separate path particle swarm archives for the combined path particle swarm and the comparative path particle swarm, respectively, to store path particles with PraetoRank=1 in the two sets of path particle swarms.

[0175] Step S3.2: Iteratively optimize the path using the combined path particle swarm optimization and the comparative path particle swarm optimization until the maximum number of iterations is reached, obtaining the optimal path, including:

[0176] The costs of the combined path particle swarm and the comparative path particle swarm are calculated based on the path particle cost function. The Pareto optimal solution set at the current time is obtained, and the two sets of path particles are sorted to obtain their ParteoRank and update their corresponding archives.

[0177] For the path particles in the combined path particle swarm and the comparison path particle swarm respectively, update the velocity and position. The velocity is updated according to formula (15), and the formula for updating the position is as follows:

[0178] (30);

[0179] in, To finally determine the combined path particle swarm and the contrasting path particle swarm within the boundary... Path particles in the next iteration Location, Path particles exist The position of the next iteration. To ultimately determine the combined path particle swarm and the contrast path particle swarm within the boundary, the total number of iterations for iteratively updating the UAV flight path is as follows: As an escape indicator, if the combined path particle swarm and the contrast path particle swarm are in the early stages of iteration, the mass... If the smallest path particle remains unchanged for three consecutive generations, then... =-1, triggering the escape mechanism, guiding the path particles to seek optimization in other regions; other cases... =1.

[0180] For the path particle velocity update formula , , The selection is also made from the velocity weight parameter update strategy pool. Specifically, the combined path particle swarm and the contrast path particle swarm within the optimal boundary are selected. , , The strategies are respectively:

[0181] The combined path particle swarm parameters are: ;

[0182] The particle swarm parameters for the comparison path are: ;

[0183] The initial parameters of both the combined path particle swarm and the contrasting path particle swarm are randomly obtained from the policy pool, and the parameter selection of the combined path particle swarm and the contrasting path particle swarm is dynamically adjusted through a pre-set competition mechanism. The competition mechanism is as follows:

[0184] Compare the combined path particle swarms under the current parameters one by one. Comparison of path particle swarms middle For the same solution, the comparison function is:

[0185] (31);

[0186] in, These are different weighting coefficients.

[0187] Calculate the same path particle swarm in both the combined path particle swarm and the contrasting path particle swarm. particles value, Particles with larger path values ​​are affected by particles from another path swarm. of If the total number of replaced path particles in a certain group of path particles is greater than that in another group of path particles, then the parameter selection of that group of path particles is changed, and it should not be the same as the parameter selection strategy of the current group of path particles, and the next iteration update is performed.

[0188] The iteration continues until it terminates or the optimal path solution that meets the requirements is obtained.

[0189] The method for outputting the optimal path solution that meets the requirements is as follows:

[0190] Unless otherwise specified, the archives of the two particle swarms are merged and the path solution with the largest crowding distance is directly output based on the crowding distance sorted.

[0191] If there are special requirements, the archives of the two particle swarms are merged, and the four cost functions of each path particle are weighted according to the required weights. The formula is then used to output the path solution with the smallest value.

[0192] This embodiment improves path quality stability, reduces the impact of local optima, and introduces a two-layer iterative framework to restrict particle iteration to a finite space containing potential optimal solutions. This fundamentally avoids the problem of particles mistakenly entering non-optimal regions due to unbounded exploration. In practice, the path is always updated within the potential optimal solution space, significantly reducing the probability of trajectory deviation caused by local optima and improving the stability of path quality (such as smoothness and optimality).

[0193] Dynamically optimize parameters to balance global exploration and local development capabilities; design a competitive elimination mechanism to divide the population into subpopulations and randomly allocate strategies from the pool. , , The optimal combination is dynamically selected by comparing its reduction in the population cost function. Compared with fixed-parameter strategies, the optimization efficiency is improved (theoretically by 20%-30%), and the global exploration and local exploitation capabilities are more balanced.

[0194] To address the shortcomings of traditional gravity search algorithms that treat the entire path as a single particle and calculate the total gravity using only a single global vector distance, leading to averaged node guidance and an inability to adapt to the differentiated needs of different nodes, this invention proposes a node-level independent gravity calculation mechanism. By calculating the independent distances to corresponding nodes along the path group and solving for gravity node by node, each path node receives precise gravitational drive.

[0195] To enhance global optimization capabilities and increase the probability of finding the optimal solution, an escape mechanism from the Kepler optimization algorithm is introduced. By simulating the directional reversibility of planetary motion, when a particle gets trapped in a local optimum (e.g., the optimal fitness of the population remains unchanged in the early stages of iteration), a directional flag is used to adjust the particle's motion direction, guiding it out of the non-optimal region. In practice, the success rate of particles escaping from local optima is improved, and the probability of finding the global optimum is significantly increased compared to traditional methods, making it particularly suitable for path planning in complex urban environments with multiple obstacles.

[0196] In summary, this embodiment, through mechanism innovation, optimizes existing methods in terms of path quality stability, computational efficiency, parameter adaptability, resource consumption, and global optimization capability, making it more suitable for the actual needs of UAV 3D path planning in urban environments.

[0197] Experimental instructions

[0198] Simulation experiments were conducted under the same randomized simulation map model, running 10 times independently. The traditional particle swarm optimization algorithm and the method described in this embodiment were selected for path planning. The best results of the traditional particle swarm optimization algorithm and the method described in this embodiment were selected for comparative analysis, and the corresponding cost function convergence table was obtained, as shown in Table 2.

[0199] Table 2. Comparison of the optimal path cost function for each generation between this embodiment and the standard particle swarm optimization algorithm.

[0200]

[0201] Based on the data in Table 2, it can be seen that:

[0202] By dynamically adjusting the parameter weights in the velocity update formulas of the two sets of path particle swarms in the final determined boundary through a designed competition mechanism, the algorithm described in this embodiment can effectively explore the potential optimal path solution space. The boundary output by the designed boundary optimization module can effectively compress the path solution space in the initial map and reduce the probability of poor local optima. It can be seen that the method described in this embodiment has improved convergence compared with the traditional particle swarm algorithm in the later stage of iteration.

[0203] A simulation experiment was conducted using a randomized simulated map model in an environment with denser obstacles. The simulation was run 10 times independently. The best results of the traditional particle swarm optimization algorithm and the method described in this embodiment were compared and analyzed to obtain the corresponding cost function convergence table, as shown in Table 3. It can be seen that the algorithm described in this embodiment has improved convergence compared with the traditional particle swarm optimization algorithm in the later stages of iteration.

[0204] Table 3. Comparison of the optimal path cost function for each generation between the method in this embodiment and the standard particle swarm optimization algorithm in dense environments.

[0205]

[0206] In environments with dense obstacles, the path planning results and optimal path cost function convergence results obtained by running the method described in this embodiment and the standard particle swarm optimization algorithm are respectively tested. Under the constraints of the boundary, the method described in this embodiment can effectively get rid of the local optimum trap and fully explore the potential optimal solution region.

[0207] Example 2

[0208] This embodiment provides a two-layer improved particle swarm unmanned aerial vehicle (UAV) path planning system, including:

[0209] The path boundary delineation module is configured to acquire the start and end coordinates of the UAV path, perform coordinate transformation, determine the maximum path boundary in the new coordinate system, and delineate the maximum path boundary. Group random boundary;

[0210] The boundary optimization module is configured to create a path particle swarm archive within each set of random boundaries and iteratively perform path optimization. After a set number of iterations, a boundary optimization is performed to check whether the path particles within each set of optimized random boundaries exceed the boundary. Path particles that exceed the boundary are handled for out-of-bounds behavior. The path particles within each set of processed random boundaries are evaluated and the path particle swarm archive is updated. Then, path optimization is iteratively performed within each set of processed random boundaries, and boundary optimization is performed again until the stopping condition is met to obtain the optimal boundary.

[0211] The path optimization module is configured to construct a combined path particle swarm and a combined path particle swarm archive based on the result of the last boundary optimization, and perform path optimization iteratively based on the combined path particle swarm until the maximum number of iterations is reached to obtain the optimal path.

[0212] The examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1 above. It should be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer-executable instructions.

[0213] The descriptions of each embodiment in the above embodiments have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0214] The proposed system can be implemented in other ways. For example, the system embodiments described above are merely illustrative, and the division of modules described above is only a logical functional division. In actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed.

[0215] Example 3

[0216] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the two-layer improved particle swarm unmanned aerial vehicle path planning method described in Embodiment 1 above.

[0217] Example 4

[0218] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the two-layer improved particle swarm unmanned aerial vehicle path planning method as described in Embodiment 1 above.

[0219] Those skilled in the art will understand that embodiments of the present invention can provide methods, systems, or computer program products. Therefore, the present invention can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0220] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0221] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0222] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0223] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0224] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A two-layer improved particle swarm unmanned aerial vehicle (UAV) path planning method, characterized in that, include: Obtain the start and end coordinates of the UAV path, perform coordinate transformation, determine the maximum boundary of the path in the new coordinate system, and then divide the maximum boundary of the path. Group random boundary; Within each set of random boundaries, a path particle swarm archive is created and path optimization is iteratively performed. After a set number of iterations, boundary optimization is performed once to check whether the path particles within each set of optimized random boundaries exceed the boundaries. Path particles that exceed the boundaries are handled for out-of-bounds behavior. The path particles within each set of processed random boundaries are evaluated, and the path particle swarm archive is updated, including: Within each set of random boundaries, initialize the positions of all path particles in the new coordinate system to obtain the initial path particles, and initialize m sets of path particle groups and the corresponding path particle group archive. Evaluate all initial path particles within each set of random boundaries, determine the optimal set of path particles at the current moment, and add it to the corresponding path particle swarm archive to obtain the optimized path particle swarm. Then, classify all path particles in the optimized path particle swarm into levels to obtain the cost evaluation result of the path optimization. Based on the cost evaluation results after each path optimization, a dynamic strategy for updating the velocity weight parameters of path particles is selected, and the updated velocity weight parameters are used to update the velocity of path particles within each set of random boundaries. After each velocity and position update, all path particles and their corresponding levels must be evaluated. The above process is repeated iteratively for all path particle swarms within each initial random boundary. After a set number of iterations, boundary optimization is performed to obtain the optimized random boundary. If any particles in the boundary exceed the boundary, they are processed. The cost of the internal path particles is re-evaluated, and the corresponding path particle swarm archive within each boundary is updated. Then, path optimization is iteratively performed within the random boundary after each group of processing, followed by boundary optimization, until the stopping condition is met and the optimal boundary is obtained. Based on the result of the last boundary optimization, a combined path particle swarm and a combined path particle swarm archive are constructed. Path optimization is performed iteratively based on the combined path particle swarm until the maximum number of iterations is reached to obtain the optimal path.

2. The dual-layer improved particle swarm unmanned aerial vehicle path planning method as described in claim 1, characterized in that, The process involves obtaining the start and end coordinates of the UAV path, performing coordinate transformation, determining the maximum path boundary in the new coordinate system, and then dividing the maximum path boundary. Group random boundaries include: Load the map, obtain the starting and ending coordinates of the drone path in the map coordinate system, take the line segment connecting the starting and ending points in the two-dimensional plane as the x-axis, the perpendicular bisector of the line segment in the two-dimensional plane as the y-axis, and keep the z-axis unchanged to obtain the new coordinate system and the starting and ending coordinates of the drone path in the new coordinate system. Transform the coordinates of the four vertices of the top view of the map in the map coordinate system to obtain the maximum boundary of the path in the new coordinates. In the new coordinate system, the maximum path boundary is divided into... A group of random boundaries is used as a swarm of boundary particles, with each random boundary being a boundary particle.

3. The dual-layer improved particle swarm unmanned aerial vehicle path planning method as described in claim 1, characterized in that, The speed weight parameter update strategy is dynamically selected based on the cost evaluation results after each path optimization, specifically: The path particle swarm within each random boundary randomly selects a policy from the policy pool. After each iteration, if the number of Pareto optimal path particles newly obtained by each path particle swarm is higher than that of the previous generation, the policy is retained and the iteration continues. If the number of Pareto optimal path particles newly obtained after the iteration is less than that of the previous generation, other policies are selected from the policy pool and the iteration continues.

4. The dual-layer improved particle swarm unmanned aerial vehicle path planning method as described in claim 1, characterized in that, The process of checking whether path particles within each optimized random boundary exceed the boundary, handling path particles that exceed the boundary, evaluating path particles within each processed random boundary, and updating the path particle swarm archive includes: Check whether the path particles within the optimized random boundary of each group exceed the boundary. If the Y value of a path particle that exceeds the boundary range is close to the left boundary value of the boundary particle, then its Y value is set to that left boundary value. If the Y value of a path particle that exceeds the boundary range is close to the right boundary value of the boundary particle, then its Y value is set to that right boundary value. Re-evaluate the cost of all path particles within the processed random boundary and update the path particle swarm archive.

5. The dual-layer improved particle swarm unmanned aerial vehicle path planning method as described in claim 1, characterized in that, The process involves constructing a combined path particle swarm and a combined path particle swarm archive based on the result of the last boundary optimization. Path optimization is then performed iteratively based on the combined path particle swarm until the maximum number of iterations is reached, yielding the optimal path. This includes: Based on the results of the last boundary optimization, construct a combined path particle swarm, a comparative path particle swarm, and a corresponding path particle swarm archive. Path optimization is performed iteratively using combined path particle swarm optimization and comparative path particle swarm optimization until the maximum number of iterations is reached, at which point the optimal path is obtained.

6. The dual-layer improved particle swarm unmanned aerial vehicle path planning method as described in claim 5, characterized in that, Based on the results of the final boundary optimization, a combined path particle swarm and a comparative path particle swarm, along with their corresponding path particle swarm archives, are constructed, including: Based on the results of the final boundary optimization, the top three boundary particles with the highest fitness values ​​(from lowest to highest) are selected, and their boundaries are then used to select the particles within the boundary. The top 30% of path particles are defined as the initial optimal path particles; It also determines whether all initial optimal path particles exceed the optimal boundary range, performs out-of-bounds processing on initial optimal path particles that exceed the optimal boundary range, and obtains the processed initial optimal path particle swarm. The generation of some new path particles in the optimal boundary is defined as the initial random particle swarm. The processed initial optimal path particle swarm and the initial random particle swarm are used to form a combined path particle swarm with a population size of Np = NpB. ; By replicating the combined path particle swarm, we obtain the comparison path particle swarm. ; Create separate path particle swarm archives for the combined path particle swarm and the comparative path particle swarm, respectively, to store path particles with PraetoRank=1 in the two sets of path particle swarms.

7. A two-layer improved particle swarm unmanned aerial vehicle path planning system, characterized in that, include: The path boundary delineation module is configured to acquire the start and end coordinates of the UAV path, perform coordinate transformation, determine the maximum path boundary in the new coordinate system, and delineate the maximum path boundary. Group random boundary; The boundary optimization module is configured to create a path particle swarm archive within each set of random boundaries and iteratively optimize paths. After a set number of iterations, a boundary optimization is performed to check whether path particles within each optimized set of random boundaries exceed the boundary. Path particles exceeding the boundary are handled for out-of-bounds errors. The module evaluates the path particles within each set of processed random boundaries and updates the path particle swarm archive, including: Within each set of random boundaries, initialize the positions of all path particles in the new coordinate system to obtain the initial path particles, and initialize m sets of path particle groups and the corresponding path particle group archive. Evaluate all initial path particles within each set of random boundaries, determine the optimal set of path particles at the current moment, and add it to the corresponding path particle swarm archive to obtain the optimized path particle swarm. Then, classify all path particles in the optimized path particle swarm into levels to obtain the cost evaluation result of the path optimization. Based on the cost evaluation results after each path optimization, a dynamic strategy for updating the velocity weight parameters of path particles is selected, and the updated velocity weight parameters are used to update the velocity of path particles within each set of random boundaries. After each velocity and position update, all path particles and their corresponding levels must be evaluated. The above process is repeated iteratively for all path particle swarms within each initial random boundary. After a set number of iterations, boundary optimization is performed to obtain the optimized random boundary. If any particles in the boundary exceed the boundary, they are processed. The cost of the internal path particles is re-evaluated, and the corresponding path particle swarm archive within each boundary is updated. Then, path optimization is iteratively performed within the random boundary after each group of processing, followed by boundary optimization, until the stopping condition is met and the optimal boundary is obtained. The path optimization module is configured to construct a combined path particle swarm and a combined path particle swarm archive based on the result of the last boundary optimization, and perform path optimization iteratively based on the combined path particle swarm until the maximum number of iterations is reached to obtain the optimal path.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the two-layer improved particle swarm unmanned aerial vehicle path planning method as described in any one of claims 1-6.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the two-layer improved particle swarm unmanned aerial vehicle path planning method as described in any one of claims 1-6.

Images