A three-dimensional path planning method for low-altitude unmanned aerial vehicles

Abstract

The present invention discloses a three-dimensional path planning method for a low-altitude UAV, which belongs to the technical field of UAV path planning, and includes the following steps: constructing a three-dimensional flight environment model; performing global path search and path cost optimization on the three-dimensional flight environment model according to a particle swarm optimization method, and obtaining the optimized path cost and position corresponding to each particle in the particle swarm; performing repeated iterations of selection, crossover and mutation operations on the particles through a genetic algorithm according to the path cost and position of each particle in the particle swarm, and obtaining a global optimal path solution based on the optimal path cost of each iteration; converting the optimal path in the target coordinate system according to the global optimal path solution, and obtaining the optimal path in the three-dimensional flight environment model by drawing a drawing function. The present invention solves the problems of low efficiency and insufficient reliability in UAV path planning.

Description

Technical Field

[0001] The invention belongs to the technical field of unmanned aerial vehicle path planning, and in particular relates to a three-dimensional path planning method for a low-altitude unmanned aerial vehicle. Background Art

[0002] With the development of the Internet of Things and artificial intelligence technologies, drone technology has seen rapid growth in the use of logistics and tactical missions due to its advantages in speed, flexibility, and cost. However, in the face of complex flight environments, drone path planning also faces many challenges.

[0003] UAV path planning methods include artificial potential field method, traditional particle swarm algorithm and other methods. The artificial potential field method guides the movement of UAV by constructing virtual gravitational field and repulsive field, generates gravitational force at the target point to attract UAV to approach, and generates repulsive force at the obstacle to make UAV avoid. However, the essential defect of this method is that it is easy to fall into the local minimum point, making the UAV seem to be in a stable position, but in fact it is not the global optimal path; and in the later stage of the algorithm operation of the traditional particle swarm method, the search range of particles gradually narrows, and the speed and position update of particles mainly depend on the current optimal solution and the individual historical optimal solution. When there are multiple local optimal solutions in the search space and the distribution is relatively scattered, the particles can easily converge to a local optimal solution prematurely. Moreover, in complex environments, the stability of the algorithm is crucial for UAV path planning. The original particle swarm algorithm and others will show a large standard deviation in multiple tests, which means that the output results of the algorithm have a large degree of discreteness in different test runs. This will not only affect the reliability and safety of UAV flight, but also lead to uncertainty in the execution of UAV tasks, and may even cause collision risks due to the instability of the path, affecting the efficiency and reliability of the UAV task execution system. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides a three-dimensional path planning method for a low-altitude UAV, which solves the problems of low efficiency and insufficient reliability in UAV path planning.

[0005] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0006] The present invention provides a three-dimensional path planning method for a low-altitude UAV, comprising the following steps:

[0007] S1. Construct a three-dimensional flight environment model;

[0008] S2. According to the particle swarm optimization method, a global path search and path cost optimization are performed on the three-dimensional flight environment model to obtain the optimized path cost and position corresponding to each particle in the particle swarm;

[0009] S3, according to the path cost and position of each particle in the particle swarm, the genetic algorithm is used to repeatedly iterate the particle selection, crossover and mutation operations, and the global optimal path solution is obtained based on the optimal path cost of each iteration;

[0010] S4. According to the global optimal path solution, the optimal path in the target coordinate system is converted and obtained, and the optimal path in the three-dimensional flight environment model is obtained by drawing through a drawing function.

[0011] The beneficial effects of the present invention are as follows: a three-dimensional path planning method for a low-altitude UAV provided by the present invention continuously adjusts the position and speed of particles in a three-dimensional flight environment model through a particle swarm algorithm, so that the path solution corresponding to the particles approaches the global optimal solution to ensure the initial optimality of the path, and then simulates the natural selection process through a genetic algorithm to perform operations such as selection, crossover and mutation on particles corresponding to the optimized path cost, thereby optimizing the local details of the three-dimensional path of the UAV, thereby obtaining the global optimal path solution and improving the smoothness of the UAV path and the flight safety; the scheme of the present invention combines the kinematic constraints of the position and speed of the UAV, obstacle detection and path cost evaluation, to ensure that the UAV completes the task using the shortest route planned efficiently and stably while avoiding collisions with obstacles.

[0012] Furthermore, S2 comprises the following steps:

[0013] S21, randomly generating a number of particles to form a particle group, and correspondingly using each particle as a path solution in the three-dimensional flight environment model, wherein each generated particle has its initial position and initial velocity;

[0014] S22, constructing a position and velocity update model for each particle;

[0015] S23, calculating the path cost of each particle according to the position of each particle;

[0016] S24. According to the path cost of each particle, based on the position speed update model and the path cost position update model, iteratively update the speed, historical optimal position and path cost of each particle until the path cost of the particle converges to a preset threshold or the number of iterative updates reaches a maximum iteration number threshold, stop the iterative update, and obtain the optimized path cost and position corresponding to each particle.

[0017] The beneficial effect of adopting the above further scheme is: the present invention continuously adjusts the position and speed of particles in the three-dimensional flight environment model through the particle swarm algorithm to reduce the path cost of the particles, so that the path solution corresponding to the particles is close to the global optimal solution, and the optimized path cost and position corresponding to each particle are obtained, which provides a basis for further processing through the genetic algorithm to obtain the global optimal path solution.

[0018] Further, the S22 comprises the following steps:

[0019] S221, constructing an inertia weight adjustment model according to Chebyshev mapping to adjust the inertia weight during iteration according to the number of iterations of particle position and velocity;

[0020] The calculation expression of the inertia weight adjustment model is as follows:

[0021] ,

[0022] ,

[0023] in, represents the inertia weight, represents the Chebyshev dynamic adjustment value at the kth iteration, represents the maximum inertia weight, represents the minimum inertia weight, Indicates k +1 Chebyshev dynamic adjustment value at iteration 1, represents the initial Chebyshev adjustment value, k Indicates the number of iterations;

[0024] S222, constructing a learning factor update model to adjust the individual learning factor and social learning factor during iteration according to the number of iterations of particle position and velocity;

[0025] The calculation expression of the learning factor update model is as follows:

[0026] ,

[0027] ,

[0028] in, represents the maximum value of the individual learning factor, represents the exponential decay factor of the individual learning factor, represents the minimum value of the individual learning factor, represents the maximum number of iterations, represents the maximum value of the social learning factor, represents the exponential decay factor of the social learning factor, represents the minimum value of the social learning factor;

[0029] S223, constructing a position and velocity update model for each particle according to the inertia weight, the individual learning factor and the social learning factor;

[0030] The calculation expression of the position and velocity update model of each particle is as follows:

[0031] ,

[0032] ,

[0033] in, represents the position of the i-th particle at the k+1-th iteration, represents the position of the i-th particle at the k-th iteration, represents the velocity of the i-th particle at the k+1-th iteration, represents the inertia weight, represents the velocity of the i-th particle at the k-th iteration, represents the individual learning factor, represents the first random number, represents the historical optimal position of the i-th particle, represents the social learning factor, represents the second random number, Indicates i The global historical optimal position of particles, where i and k are both positive integers.

[0034] The beneficial effects of adopting the above further scheme are as follows: the present invention avoids the problem of falling into local optimum and limited global search ability in the later stage of exploration caused by improper setting of inertia weight in traditional particle swarm algorithm by setting inertia weight based on Chebyshev mapping. The nonlinear inertia weight can enable particles to have strong exploration ability in the whole search process, and can search in a large range in the early stage, and can balance local and global search in the later stage, which greatly improves the probability of finding the optimal path cost, and also enhances the stability and convergence speed in complex path planning scenarios; the present invention overcomes the shortcomings of fixed or simple changes of learning factors by setting nonlinear asynchronous learning factors, namely individual learning factors and social learning factors, and avoids the defects of lack of particle diversity and premature convergence to local optimum caused by this. Through individual learning factors and social learning factors, particles rely more on self-exploration in the early stage of exploration, increase diversity, and avoid premature aggregation, thereby realizing a smooth transition from individual exploration to group collaboration in the whole search process, so that the scheme can better adapt to different stages and complex environments, effectively avoid premature convergence, and significantly improve the search ability and the possibility of finding the global optimal solution.

[0035] Further,

[0036] The calculation expression of the path cost of each particle is as follows:

[0037] ,

[0038] in, represents the current path cost of the i-th particle, represents the path cost objective function value corresponding to the current position of the i-th particle, represents the distance from the current position of the i-th particle to the current position of the i+1-th particle, Indicates the degree of threat from the current position of the i-th particle to the current position of the i+1-th particle, represents the vector from the current position of the ith particle to the current position of the i+1th particle, It means to find the vector distance. It indicates the degree of threat from the current position of the i-th particle to the current position of the i+1-th particle by the o-th obstacle, O represents the total number of obstacles, and n represents the total number of particles.

[0039] The beneficial effect of adopting the above further scheme is that the path cost objective function provided in the present invention combines the path cost direction of the particle position and the results of the flight safety cost analysis to achieve the measurement of the quality of the path represented by the current position of the particle, that is, the current path cost of the particle.

[0040] Furthermore, the calculation expression of the path cost position update model is as follows:

[0041] ,

[0042] in, represents the historical optimal position of the i-th particle after update, Represents the path cost of the i-th particle at the historical optimal position.

[0043] Furthermore, S3 comprises the following steps:

[0044] S31, according to the optimized path cost corresponding to each particle in the particle swarm, select a number of particles as parent particles in turn based on the particle selection model;

[0045] S32, according to the uniform crossover operation, the path solutions corresponding to any two parent particles are combined and exchanged to generate new child particles, and the positions of the child particles are obtained;

[0046] S33, changing the positions of the daughter particles in each dimension through random mutation operation to obtain mutated daughter particles;

[0047] S34, if the path cost of the mutated offspring particle is less than that of the parent particle, the mutated offspring particle replaces the parent particle, and after the replacement is completed, the lowest path cost is selected as the optimal path cost of this iteration;

[0048] S35, repeat S32-S34 until the number of repetitions reaches a preset repetition number threshold, and select the particle corresponding to the lowest path cost from the optimal path cost of each iteration, and use the path solution corresponding to the particle as the global optimal path solution.

[0049] The beneficial effects of adopting the above further scheme are as follows: the hybrid iterative cycle of the particle swarm algorithm and the genetic algorithm is utilized in the present invention, which fully combines the advantages of both and solves the problems of insufficient global detection capability, low local search accuracy, easy premature convergence and slow convergence speed in the traditional particle swarm algorithm or genetic algorithm. In the hybrid iterative process, the initial parameter setting of the particle swarm algorithm and the genetic algorithm operation jointly expand the exploration range of the particles and enhance the global perception capability. The later parameter changes of the particle swarm algorithm and the genetic algorithm operation improve the local search accuracy for particles with excellent path costs. The characteristics of the two complement each other, continuously explore new areas and path solutions, effectively avoid premature convergence, and the synergistic effect of the two accelerates the convergence speed, reduces the number of iterations, and improves the efficiency and stability of the present scheme in implementing drone path planning.

[0050] Furthermore, the calculation expression of the particle selection model is as follows:

[0051] ,

[0052] in, represents the selected particle, Indicated in When particles belong to particle group T, find the particle that makes the optimized path cost reach the minimum value.

[0053] The beneficial effect of adopting the above further scheme is: the present invention provides a calculation method for selecting parent particles according to the optimized path cost using a particle selection model, and selects particles that are more likely to be close to the optimal solution as parent particles according to the optimized path cost corresponding to the particles, providing a basic population for subsequent operations.

[0054] Furthermore, the calculation expression of the position of the descendant particle is as follows:

[0055] ,

[0056] ,

[0057] in, Indicates The positions of the daughter particles, represents the cross coefficient, Indicates The position of the parent particle, Indicates The position of the parent particle, Indicates The position of the descendant particles.

[0058] The beneficial effect of adopting the above further scheme is: the present invention provides a calculation method for obtaining the position of child particles based on the crossover operation of parent particles, by performing a crossover operation on the selected particles in pairs, exchanging part of the position information in each dimension, generating new child particles, increasing the population diversity, and facilitating the exploration of more potential optimal path combinations.

[0059] Furthermore, the calculation expression of the position of the mutated offspring particle is as follows:

[0060] ,

[0061] ,

[0062] ,

[0063] in, Indicates the radial distance dimension position of the offspring particle after particle mutation, Represents the radial distance dimension position of the descendant particles, Indicates the polar angular position of the offspring particle after particle mutation. represents the polar angular position of the descendant particles, Indicates the azimuth dimension position of the offspring particle after particle mutation. Represents the azimuth dimension position of the descendant particle, Gaussian noise representing the radial distance dimension position, Gaussian noise representing the polar angle position, Gaussian noise representing the position in the azimuth dimension.

[0064] The beneficial effects of adopting the above-mentioned further scheme are as follows: the present invention adopts a spherical coordinate system to replace the traditional Cartesian coordinate system, thereby overcoming the drawbacks of the Cartesian coordinate system in scenarios such as drones that require frequent attitude adjustment, including unreasonable coordinate representation, difficulty in generating feasible solutions, and difficulty in accurately describing motion trajectories and attitude changes. The present invention performs variability adjustment on various dimensions of particle positions based on the spherical coordinate system, thereby improving the accuracy and feasibility of path planning, and making it easier to generate high-quality solutions. Especially in fixed-speed flight and situations involving drone attitude adjustment, the present invention can excellently control and describe the motion state, effectively reduce unreasonable path phenomena caused by coordinate problems, and make the planned path more in line with actual flight conditions.

[0065] Other advantages of the present invention will be analyzed in more detail in subsequent embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0067] Figure 1 The present invention is a flowchart of the steps of a three-dimensional path planning method for a low-altitude UAV in an embodiment of the present invention.

[0068] FIG. 2 (a) is a top view of the UAV planning path implemented by the traditional particle swarm algorithm in an embodiment of the present invention.

[0069] FIG. 2( b ) is a top view of the planned path of the UAV implemented by this solution in an embodiment of the present invention.

[0070] FIG. 3 (a) is a three-dimensional diagram of the UAV planning path implemented by the traditional particle swarm algorithm in an embodiment of the present invention.

[0071] FIG3( b ) is a three-dimensional diagram of the planned path of the UAV implemented by this solution in an embodiment of the present invention. DETAILED DESCRIPTION

[0072] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0073] like Figure 1 As shown, in one embodiment of the present invention, the present invention provides a three-dimensional path planning method for a low-altitude UAV, comprising the following steps:

[0074] S1. Construct a three-dimensional flight environment model;

[0075] The S1 comprises the following steps:

[0076] S11, obtaining terrain data within the flight area;

[0077] S12. Construct a three-dimensional flight environment model based on the terrain data in the flight area through the elevation map and obstacle data. In this solution, the three-dimensional flight environment data is used to provide the terrain undulations, obstacle distribution and dynamically changing environment that the aircraft may face in the flight area. In this embodiment, a three-dimensional simulation scene based on irregular mountain features is used as the three-dimensional flight environment model. By highly restoring the complex impact of the mountain environment on the flight of the UAV, the terrain undulations and dynamic changes in the real flight scene can be simulated.

[0078] S2. According to the particle swarm optimization method, a global path search and path cost optimization are performed on the three-dimensional flight environment model to obtain the optimized path cost and position corresponding to each particle in the particle swarm;

[0079] The S2 comprises the following steps:

[0080] S21. Randomly generate a number of particles to form a particle swarm, and use each particle as a path solution in the three-dimensional flight environment model, wherein each generated particle has its initial position and initial velocity. In this solution, the initial position of each particle is randomly generated through a preset range of position variables, and the initial velocity of the particle is randomly generated through a preset range of particle velocities in each dimension. The initial position of the particle provides an initial flight node of the flight path, and its value is limited to the space of the three-dimensional flight environment model to provide a starting point for subsequent exploration. The initial velocity of the particle provides the initial movement direction and velocity information of the particle. The path solutions corresponding to each particle are respectively composed of a series of flight nodes. The position and velocity of the particle are used to represent the path coordinates and the optimization direction, respectively, wherein the path coordinates refer to the positions of each flight node, and the optimization direction refers to the movement trend of the particle in the exploration space.

[0081] In this real-time example, the size of the generated particle swarm is set to 500 to ensure that the algorithm has sufficient exploration capabilities in the complex mountainous three-dimensional space while maintaining computational efficiency. In order to avoid the wrong selection of the UAV flight path due to too fast speed, or the search efficiency of the three-dimensional flight environment model due to too slow speed, the maximum and minimum values ​​of the particle speed are reasonably set.

[0082] S22, constructing a position and velocity update model for each particle;

[0083] The S22 comprises the following steps:

[0084] S221, constructing an inertia weight adjustment model according to Chebyshev mapping to adjust the inertia weight during iteration according to the number of iterations of particle position and velocity;

[0085] The calculation expression of the inertia weight adjustment model is as follows:

[0086] ,

[0087] ,

[0088] in, represents the inertia weight, represents the Chebyshev dynamic adjustment value at the kth iteration, represents the maximum inertia weight, represents the minimum inertia weight, Indicates k+ Chebyshev dynamic adjustment value at 1 iteration, represents the initial Chebyshev adjustment value, k represents the number of iterations; in this embodiment, the inertia weight gradually decreases with the number of iterations, ensuring greater exploration in the early stage and increasing convergence in the later stage. In the initial stage, the calculated inertia weight value is relatively large, so that the particle is more dependent on the previous speed during the speed update process, so that it can be explored in a wider search space, which is helpful to find potential global optimal solution areas, just like conducting a wide "casting net" search in a wide solution space. As the iterations proceed, the weight gradually changes, and the particle will be more inclined to refer to the optimal position of the current particle itself and the global optimal position when updating the speed, and then conduct a more refined local search in the discovered better area, thereby effectively improving the possibility and accuracy of converging to the global optimal solution or close to the global optimal solution.

[0089] In this scheme, the inertia weight generated by the Chebyshev mapping characteristics changes according to a specific nonlinear pattern during iteration. Different from the traditional method, it effectively improves the global search ability of particles, avoids falling into local optimality too early, and optimizes the search guidance process, which is an important guarantee for the efficient operation of the algorithm of this scheme.

[0090] S222, constructing a learning factor update model to adjust the individual learning factor and social learning factor during iteration according to the number of iterations of particle position and velocity;

[0091] The calculation expression of the learning factor update model is as follows:

[0092] ,

[0093] ,

[0094] in, represents the maximum value of the individual learning factor, represents the exponential decay factor of the individual learning factor, represents the minimum value of the individual learning factor, represents the maximum number of iterations, represents the maximum value of the social learning factor, represents the exponential decay factor of the social learning factor, represents the minimum value of the social learning factor;

[0095] The learning factor implements dynamic adjustment based on the number of iterations, aiming to optimize the algorithm's search behavior at different stages to better balance the global search and local search capabilities, thereby improving the algorithm's overall performance and convergence efficiency. In the initial stage of the algorithm, since the value of the individual learning factor is relatively large and the value of the social learning factor is relatively small, the particles will focus more on referring to their own historical optimal positions when updating the speed, that is, they will rely more on their own search experience and conduct diversified exploration in the vast search space to avoid falling into the local optimal solution too early, which is conducive to global search. As the number of iterations increases, the value of the individual learning factor gradually decreases, and the value of the social learning factor gradually increases. This makes the particles gradually increase the reference weight to the global optimal position when updating the speed, and learn more from the excellent experience of other particles in the group, thereby accelerating the convergence speed, enabling the algorithm to perform more efficient local searches in the discovered better areas, and finally achieving an effective balance between the algorithm's search ability and convergence performance, and improving the probability and efficiency of finding the global optimal solution.

[0096] S223, constructing a position and velocity update model for each particle according to the inertia weight, the individual learning factor and the social learning factor;

[0097] The calculation expression of the position and velocity update model of each particle is as follows:

[0098] ,

[0099] ,

[0100] in, represents the position of the i-th particle at the k+1th iteration, represents the position of the i-th particle at the k-th iteration, represents the velocity of the i-th particle at the k+1-th iteration, represents the inertia weight, represents the velocity of the i-th particle at the k-th iteration, represents the individual learning factor, represents the first random number, represents the historical optimal position of the i-th particle, represents the social learning factor, represents the second random number, Indicates i The global optimal historical position of particles, where i and k are both positive integers.

[0101] The inertia weight is used to control the influence of the current speed of the particle on the speed in the next iteration; the individual learning factor and the social learning factor are used to control the step size of the particle's own historical optimal position and the global historical optimal position respectively; the first random number and the second random number are used to increase the randomness of the search; according to the particle swarm optimization method, in the next iteration, the i-th particle will move to a new position according to its new speed and current position. By adjusting the individual learning factor and the social learning factor, the particles can better balance individual exploration and group experience utilization during the search process to cope with the complex environment in mountain flight.

[0102] S23, calculating the path cost of each particle according to the position of each particle;

[0103] The calculation expression of the path cost of each particle is as follows:

[0104] ,

[0105] in, represents the current path cost of the i-th particle, represents the path cost objective function value corresponding to the current position of the i-th particle, represents the distance from the current position of the i-th particle to the current position of the i+1-th particle, Indicates the degree of threat from the current position of the i-th particle to the current position of the i+1-th particle, represents the vector from the current position of the ith particle to the current position of the i+1th particle, It means to find the vector distance. It indicates the degree of threat from the current position of the i-th particle to the current position of the i+1-th particle by the o-th obstacle, O represents the total number of obstacles, and n represents the total number of particles.

[0106] In this scheme, the path cost objective function value corresponding to the particle position is calculated based on the path cost objective function. The path cost objective function combines the path cost direction of the particle position and the results of the flight safety cost analysis to measure the quality of the path represented by the particle's current position, that is, the particle's current path cost.

[0107] S24. According to the path cost of each particle, based on the position speed update model and the path cost position update model, iteratively update the speed, historical optimal position and path cost of each particle until the path cost of the particle converges to a preset threshold or the number of iterative updates reaches a maximum iteration number threshold, stop the iterative update, and obtain the optimized path cost and position corresponding to each particle.

[0108] In this scheme, if the current path cost of a particle is better than the path cost of the particle at its historical optimal position, the particle's historical optimal position is updated so that the particle's historical optimal position is the particle's current position. Otherwise, the particle's historical optimal position remains unchanged, and the optimized path cost corresponding to each particle is the path cost corresponding to the particle's historical optimal position after the iterative update is completed.

[0109] The calculation expression of the path cost position update model is as follows:

[0110] ,

[0111] in, represents the historical optimal position of the i-th particle after update, Represents the path cost of the i-th particle at the historical optimal position.

[0112] S3, according to the path cost and position of each particle in the particle swarm, the genetic algorithm is used to repeatedly iterate the particle selection, crossover and mutation operations, and the global optimal path solution is obtained based on the optimal path cost of each iteration;

[0113] The S3 comprises the following steps:

[0114] S31, according to the optimized path cost corresponding to each particle in the particle swarm, select a number of particles as parent particles in turn based on the particle selection model;

[0115] The calculation expression of the particle selection model is as follows:

[0116] ,

[0117] in, represents the selected particle, Indicated in When particles belong to particle group T, find the particle that makes the optimized path cost reach the minimum value.

[0118] S32, according to the uniform crossover operation, the path solutions corresponding to any two parent particles are combined and exchanged to generate new child particles, and the positions of the child particles are obtained;

[0119] The calculation expression of the position of the descendant particle is as follows:

[0120] ,

[0121] ,

[0122] in, Indicates The positions of the daughter particles, represents the cross coefficient, Indicates The position of the parent particle, Indicates The position of the parent particle, Indicates The position of the descendant particles.

[0123] S33, changing the positions of the daughter particles in each dimension through random mutation operation to obtain the positions of the daughter particles after mutation;

[0124] The calculation expression of the position of the mutated offspring particle is as follows:

[0125] ,

[0126] ,

[0127] ,

[0128] in, Indicates the radial distance dimension position of the offspring particle after particle mutation, Represents the radial distance dimension position of the descendant particles, Indicates the polar angular position of the offspring particle after particle mutation. represents the polar angular position of the descendant particles, Indicates the azimuth dimension position of the offspring particle after particle mutation. Represents the azimuth dimension position of the descendant particle, Gaussian noise representing the radial distance dimension position, Gaussian noise representing the polar angle dimension position, Indicates the Gaussian noise of the azimuth dimension position. In this solution, the mean of the Gaussian noise of the radial distance dimension position, the Gaussian noise of the polar angle dimension position, and the Gaussian noise of the azimuth dimension position are all 0, and their standard deviation , and Used to control the intensity of mutation; adding small changes to the position of each dimension of the particle within a reasonable range while satisfying the mutation probability can avoid falling into the local optimal solution, thereby expanding the search space to explore a better path.

[0129] S34, if the path cost of the mutated offspring particle is less than that of the parent particle, the mutated offspring particle replaces the parent particle, and after the replacement is completed, the lowest path cost is selected as the optimal path cost of this iteration;

[0130] S35, repeat S32-S34 until the number of repetitions reaches a preset repetition number threshold, and select the particle corresponding to the lowest path cost from the optimal path cost of each iteration, and use the path solution corresponding to the particle as the global optimal path solution.

[0131] In this embodiment, the stability and convergence of each iterative optimization are evaluated by analyzing the standard deviation and the fluctuation coefficient of the standard deviation of the optimal path cost updated in each iteration;

[0132] The calculation expressions of the standard deviation of the optimal path cost and the fluctuation coefficient of the standard deviation for each iterative update are as follows:

[0133] ,

[0134] ,

[0135] in, represents the standard deviation of the iteratively updated optimal path cost, represents the average value of the optimal path cost updated in each iteration, The coefficient of fluctuation that represents the standard deviation of the optimal path cost for each iteration update, represents the standard deviation of the average value of the optimal path cost updated in each iteration, It represents the average value of the optimal path cost of each iteration update. This scheme ensures that the path does not change drastically during the actual flight process by evaluating the stability of the planned path under different flight conditions.

[0136] S4. According to the global optimal path solution, the optimal path in the target coordinate system is converted and obtained, and the optimal path in the three-dimensional flight environment model is obtained by drawing through a drawing function.

[0137] In a specific embodiment of this scheme, the optimal path cost, the standard deviation of the optimal path cost, and the fluctuation coefficient of the standard deviation of the optimal path cost after each iteration update of this scheme and the traditional particle swarm algorithm are compared. Among them, the optimal path cost, the standard deviation of the optimal path cost, and the fluctuation coefficient of the standard deviation of the optimal path cost after each iteration update under the traditional particle swarm algorithm and this scheme are shown in Table 1 and Table 2 respectively:

[0138]

[0139]

[0140] In addition, the comparison of the path length after each iteration between this scheme and the traditional particle swarm algorithm is shown in Table 3:

[0141]

[0142] As shown in Figure 2 (a), Figure 2 (b), Figure 3 (a) and Figure 3 (b), this embodiment provides a top view of the UAV planning path implemented by the traditional particle swarm algorithm and the UAV planning path implemented by this solution, as well as a three-dimensional diagram converted to a spatial rectangular coordinate system. It can be found that this solution has obvious advantages over the UAV path planned by the traditional particle swarm algorithm in terms of both the stability of the algorithm and the final planned path. This solution can effectively handle more path constraints and optimization goals, and will not significantly reduce performance due to the increase in the scale of the problem.

[0143] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A three-dimensional path planning method for a low-altitude UAV, characterized in that: The steps include: S1. Construct a three-dimensional flight environment model; S2. According to the particle swarm optimization method, a global path search and path cost optimization are performed on the three-dimensional flight environment model to obtain the optimized path cost and position corresponding to each particle in the particle swarm; The S2 comprises the following steps: S22, constructing a position and velocity update model for each particle; The S22 comprises the following steps: S221, constructing an inertia weight adjustment model according to Chebyshev mapping to adjust the inertia weight during iteration according to the number of iterations of particle position and velocity; The calculation expression of the inertia weight adjustment model is as follows: ， ， in, represents the inertia weight, Indicates k The Chebyshev dynamic adjustment value at the iteration, represents the maximum inertia weight, represents the minimum inertia weight, Indicates k+ Chebyshev dynamic adjustment value at 1 iteration, represents the initial Chebyshev adjustment value, k Indicates the number of iterations; S222, constructing a learning factor update model to adjust the individual learning factor and social learning factor during iteration according to the number of iterations of particle position and velocity; The calculation expression of the learning factor update model is as follows: ， ， in, represents the maximum value of the individual learning factor, represents the exponential decay factor of the individual learning factor, represents the minimum value of the individual learning factor, represents the maximum number of iterations, represents the maximum value of the social learning factor, represents the exponential decay factor of the social learning factor, represents the minimum value of the social learning factor; S3, according to the path cost and position of each particle in the particle swarm, the genetic algorithm is used to repeatedly iterate the particle selection, crossover and mutation operations, and the global optimal path solution is obtained based on the optimal path cost of each iteration; The S3 comprises the following steps: S31, according to the optimized path cost corresponding to each particle in the particle swarm, select a number of particles as parent particles in turn based on the particle selection model; S32, according to the uniform crossover operation, the path solutions corresponding to any two parent particles are combined and exchanged to generate new child particles, and the positions of the child particles are obtained; S33, changing the positions of the daughter particles in each dimension through random mutation operation to obtain the positions of the daughter particles after mutation; The calculation expression of the position of the mutated offspring particle is as follows: ， ， ， in, Indicates the radial distance dimension position of the mutated offspring particles, Represents the radial distance dimension position of the descendant particles, Indicates the polar angular position of the mutated offspring particles. represents the polar angular position of the descendant particles, Indicates the azimuth dimension position of the mutated offspring particle, Represents the azimuth dimension position of the descendant particle, Gaussian noise representing the radial distance dimension position, Gaussian noise representing the polar angle dimension position, Gaussian noise representing the position in the azimuth dimension; S34, if the path cost of the mutated offspring particle is less than that of the parent particle, the mutated offspring particle replaces the parent particle, and after the replacement is completed, the lowest path cost is selected as the optimal path cost of this iteration; S35, repeat S32-S34 until the number of repetitions reaches a preset repetition number threshold, and select the particle corresponding to the lowest path cost from the optimal path costs of each iteration, and use the path solution corresponding to the particle as the global optimal path solution; S4. According to the global optimal path solution, the optimal path in the target coordinate system is converted and obtained, and the optimal path in the three-dimensional flight environment model is obtained by drawing through a drawing function.

2. The three-dimensional path planning method for a low-altitude UAV according to claim 1 is characterized in that: The S2 further comprises the following steps: S21, randomly generating a number of particles to form a particle group, and correspondingly using each particle as a path solution in the three-dimensional flight environment model, wherein each generated particle has its initial position and initial velocity; S23, calculating the path cost of each particle according to the position of each particle; S24. According to the path cost of each particle, based on the position speed update model and the path cost position update model, iteratively update the speed, historical optimal position and path cost of each particle until the path cost of the particle converges to a preset threshold or the number of iterative updates reaches a maximum iteration number threshold, stop the iterative update, and obtain the optimized path cost and position corresponding to each particle.

3. The three-dimensional path planning method for a low-altitude UAV according to claim 2 is characterized in that: The S22 further comprises the following steps: S223, constructing a position and velocity update model for each particle according to the inertia weight, the individual learning factor and the social learning factor; The calculation expression of the position and velocity update model of each particle is as follows: ， ， in, Indicates i The particle in k The position at +1 iteration, Indicates i The particle in k The position at the iteration, Indicates i The particle in k +1 iteration speed, represents the inertia weight, Indicates i The particle in k The speed at the iteration, represents the individual learning factor, represents the first random number, Indicates i The historical optimal position of a particle, represents the social learning factor, represents the second random number, Indicates i The global historical optimal position of particles, where i and k All are positive integers.

4. The three-dimensional path planning method for a low-altitude UAV according to claim 3 is characterized in that: The calculation expression of the path cost of each particle is as follows: ， in, Indicates i The current path cost of a particle, Indicates i The path cost objective function value corresponding to the current position of the particle, Indicates i The current position of the particle to the i +1 distance from the particle's current position, Indicates i The current position of the particle to the i +1 particle's threat level at the current location, Indicates that from i The current position of the particle to the i +1 vector of the particle's current position, It means to find the vector distance. Indicates i The current position of the particle to the i +1 The current position of the particle is affected by o The threat level of an obstacle O represents the total number of obstacles, n Represents the total number of particles.

5. The three-dimensional path planning method for a low-altitude UAV according to claim 4 is characterized in that: The calculation expression of the path cost position update model is as follows: ， in, Indicates i The historical optimal position of the particle after update, Indicates i The path cost of a particle at the historical optimal position.

6. The three-dimensional path planning method for a low-altitude UAV according to claim 1, characterized in that: The calculation expression of the particle selection model is as follows: ， in, represents the selected particle, Indicated in Particles belong to a particle swarm T Find the particle that minimizes the cost of the optimized path.

7. The three-dimensional path planning method for a low-altitude UAV according to claim 1, characterized in that: The calculation expression of the position of the descendant particle is as follows: ， ， in, Indicates The positions of the daughter particles, represents the cross coefficient, Indicates The position of the parent particle, Indicates The position of the parent particle, Indicates The position of the descendant particles.

Images