Unmanned aerial vehicle path planning method based on multi-target adaptive particle swarm optimization

Abstract

The invention relates to the technical field of unmanned aerial vehicles, in particular to an unmanned aerial vehicle path planning method based on multi-target adaptive particle swarm optimization, which constructs a three-dimensional map of a real environment through mathematical modeling, initializes a population and a weight vector, and combines a population movement density estimation mechanism to realize path planning of the unmanned aerial vehicle. And calculating a population movement density estimation value of each particle in the population according to a population movement density estimation mechanism, screening out a plurality of small population movement density estimation values, and finally selecting an unmanned aerial vehicle flight path corresponding to the minimum value of the target function. According to the method, the path planning efficiency and safety of the unmanned aerial vehicle in a complex environment are improved, and the defects of a traditional algorithm in the aspects of calculation efficiency and robustness are overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicles (UAVs), and more specifically, to a UAV path planning method based on multi-objective adaptive particle swarm optimization. Background Art

[0002] With the rapid development of unmanned aerial vehicle (UAV) technology, UAV path planning has become a key research hotspot and is widely used in various fields, including terrain mapping, disaster search and rescue, and surveillance and reconnaissance missions. At present, the research on this technology is mainly divided into three categories: traditional algorithms, machine learning algorithms, and meta-heuristic algorithms. For traditional algorithms, such as Voronoi diagram algorithm, A* algorithm, rapid exploration random tree algorithm, etc., they provide a key framework for the field of UAV path planning and lay a solid foundation for the development of subsequent algorithms. However, they usually do not fully consider the kinematic and dynamic constraints of UAVs, which limits their application in actual situations. In addition, the reinforcement learning framework shows great potential in processing the maneuverability of UAVs, but its high memory requirements and long processing time restrict its application areas.

[0003] Metaheuristic algorithms are favored by researchers because of their flexibility, simplicity, and low computational complexity compared to other types of algorithms. The goals of drone path planning usually include multiple aspects, such as reducing flight time, reducing the number of collisions, and shortening path length. When dealing with these multi-objective problems, single-objective optimization or multi-objective optimization can be used. When single-objective optimization is used, the algorithm comprehensively considers multiple goals by introducing weight factors. Specifically, these goals can be weighted and summed in a certain proportion according to their importance to form a single optimization goal, where the weight factor needs to be adjusted according to the specific flight scenario and mission requirements. When single-objective optimization is used to solve the drone path planning problem, the solution process can be simplified, but a large number of experiments are required to determine the appropriate weight factor. Then, external factors may make it impossible to determine the most appropriate weight factor, resulting in the final solution not being the global optimal solution. The drone path planning problem is essentially a constrained multi-objective optimization problem because it involves multiple conflicting goals. For example, pursuing the shortest flight path may increase the risk of collision with obstacles. On the contrary, in order to reduce the number of collisions, it may be necessary to bypass obstacles, which will result in a longer flight path. In addition, some constraints should be considered to ensure the practicality and feasibility of the planned path, such as flight altitude, pitch angle and yaw angle, etc. However, single-objective optimization methods often find it difficult to achieve a balance between objectives when dealing with multi-objective problems. Summary of the invention

[0004] In order to overcome at least one of the defects of the prior art described above, the present invention provides a UAV path planning method based on multi-objective adaptive particle swarm optimization.

[0005] The present invention aims to solve the above technical problems at least to a certain extent.

[0006] In order to solve the above technical problems, the technical solution of the present invention is as follows:

[0007] A UAV path planning method based on multi-objective adaptive particle swarm optimization includes the following steps:

[0008] S1: constructing a mathematical model for the three-dimensional map, wherein the mathematical model simulates obstacles in the three-dimensional map;

[0009] S2: setting the objective function and constraint function according to the mathematical model;

[0010] S3: Initialize the parameters of the particle swarm optimization algorithm and divide the particle swarm into multiple sub-populations;

[0011] S4: Iterate the current multiple sub-populations through the particle swarm optimization algorithm to obtain the iterated sub-populations;

[0012] S5: calculating the population movement density estimation values ​​of all the iterative sub-populations according to the population movement density estimation mechanism, and selecting multiple iterative sub-populations according to the population movement density estimation values;

[0013] S6: If there are multiple filtered sub-populations that satisfy the constraint function, then the UAV flight path set generated by the filtered sub-populations is output, and the UAV flight path corresponding to the minimum value of the objective function is selected from the UAV flight path set. If there are sub-populations that do not satisfy the constraint function, the sub-populations that do not satisfy the constraint function are repaired and the process returns to step S4.

[0014] Further, in step S1, the mathematical model is constructed for the three-dimensional map, and the mathematical model simulates obstacles in the three-dimensional map, including:

[0015]

[0016] In the formula, is the three-dimensional coordinate of the kth mountain, is the horizontal center of the kth mountain, h k is the first design parameter, x′ k is the second design parameter, y′ k is the third design parameter.

[0017] Further, in step S3, the parameters of the particle swarm optimization algorithm are initialized, including:

[0018] Set the maximum inertia weight w of the particle swarm algorithm parameters max and the minimum value w min , total number of iterations t max , maximum learning factor c max and the minimum value c min .

[0019] Furthermore, the objective function includes calculating the flight path length and the number of collisions between the drone and obstacles, including:

[0020]

[0021] Where F(x) is the objective function, η is the preset constraint violation threshold, and f obj1 (x) is the first objective function, used to calculate the flight path length, f obj2 (x) is the second objective function, which is used to calculate the number of collisions between the drone and obstacles, f con1 (x) is the first constraint function, f con2 (x) is the second constraint function.

[0022]

[0023] Where, d k,i,j is the distance of the k-segment path between the ith path point and the jth radar position, R max is the maximum detection radius of the radar, A i,j is the penalty value for the collision between the i-th path point and the j-th radar, N p is the total number of path points, B i,q is the penalty value for the collision between the i-th path point and the q-th mountain peak height, n is the number of segments between the current path point and the next path point, is the distance of the k-segment path between the ith path point and the qth peak height, z i,q is the distance between the i-th path point and the q-th peak height;

[0024]

[0025] (x i-1 ,y i-1 ,z i-1 ) and (x i ,y i ,z i ) are the positions of the i-1th path point and the i-th path point, respectively. p is the total number of path points.

[0026] Further, in step S6, the constraint function includes:

[0027] The i-th pitch angle β i The calculation formula is:

[0028]

[0029] x i+1 -x i is the distance between the i+1th path point and the ith path point on the x-axis, y i+1 -y i is the distance between the i+1th path point and the ith path point on the y axis, z i+1 -z i is the distance between the i+1th path point and the ith path point on the z-axis;

[0030] The calculation formula of the pitch angle constraint function is:

[0031]

[0032] Among them, β min is the minimum pitch angle, β max is the maximum pitch angle, and C i is the penalty value for violating the pitch angle constraint of the i-th path point;

[0033] The i-th yaw angle α i The calculation formula is:

[0034]

[0035] The i-th yaw angle α i The constraint function is:

[0036]

[0037] In the formula, α max is the maximum yaw angle, α min is the minimum yaw angle, D i is the penalty value for violating the yaw angle constraint of the i-th path point.

[0038] Further, in step S2, the particle group is divided into a plurality of sub-populations, including:

[0039] The particle swarm is divided into multiple sub-populations by calculating the distance of the initialized weight vector between each particle in the particle swarm.

[0040] Further, in step S4, the current multiple sub-populations are iterated by the particle swarm optimization algorithm to obtain the iterated sub-populations, including:

[0041] S4.1: Taking the bifurcation factor and the number of iterations as the input of the particle swarm optimization algorithm, the chaos factor is generated;

[0042]

[0043] Where ρ(t) is the chaos factor of the t-th iteration, μ is the bifurcation factor, and t is the t-th iteration;

[0044] S4.2: Calculate the first learning factor c1 and the first learning factor c2 according to the tth iteration number and the total iteration number and the chaotic factor obtained in step S4.1;

[0045]

[0046] S4.3: Adjust inertia weights based on population renewal rate;

[0047]

[0048] In the formula, Ratio is the population update rate, success_num is the total number of times the optimal solution of the population particles is updated, N is the total population size, and w is the inertia weight;

[0049] S4.4: Calculate the population particle velocity and current position according to the inertia weight, and obtain the population particles of the t+1th iteration and the tth iteration;

[0050] v i (t+1)=wv i (t)+c1r1(pbest i (t)-x i (t))+c2r2(gbest(t)-x i (t));

[0051] x i (t+1)=x i (t)+v i (t+1);

[0052] Where w is the inertia weight, pbest i (t) is the historical optimal position of the i-th population particle, gbest(t) is the global optimal position, v i (t+1) is the moving speed of the population particles in the tth iteration, x i (t+1) is the position of the population particle at the t+1th iteration, r1 and r2 are random numbers;

[0053] S4.5: Merge the population particles of the t+1th iteration and the tth iteration to obtain a synthetic population with a population size of 2N;

[0054] Further, in step S5, the population movement density estimation values ​​of all the iterative sub-populations are calculated according to the population movement density estimation mechanism, and a plurality of iterative sub-populations are selected according to the population movement density estimation values, including:

[0055] Calculate the population movement density estimation values ​​of all the iterated sub-populations according to the population movement density estimation mechanism, sort each population particle from high to low according to the population movement density estimation value, retain the first N population particles, and eliminate the last N population particles;

[0056] D′(p,Pop)=D(dist(p,q′1 ) , dist(p, q′2),..., dist(p, q′ N -1))

[0057] D′(p,Pop) is the density estimate of the population particle p in the population Pop, dist() is the similarity between the two population particles, p is the pth population particle in the population Pop, q′ N-1 is the population particle after p displacement.

[0058] Further, in step S6, if there are multiple screened sub-populations that satisfy the constraint function, the drone flight path set generated by the screened sub-populations is output, and the drone flight path corresponding to the minimum value of the objective function is selected from the drone flight path set. If there are sub-populations that do not satisfy the constraint function, the sub-populations that do not satisfy the constraint function are repaired, and the process returns to step S4, including:

[0059] If the screened sub-population meets the constraints, the UAV flight path corresponding to the minimum value of the first objective function or the minimum value of the second objective function is output from the sub-population that meets the constraints. If there is a screened sub-population that does not meet the constraints, the sub-population that does not meet the constraints is repaired and returns to step S4.4.

[0060] Furthermore, the repairing of the sub-population that does not satisfy the constraint condition includes:

[0061]

[0062]

[0063] Δx=x(i+1)-x(i)

[0064] Δy=y(i+1)-y(i)

[0065]

[0066] Where, X min , X max, Y min , Y max , Z min , Z max is the boundary value of the UAV flight space, Dis is the distance between adjacent path points on the horizontal plane, r is a random number, Δx is the distance between adjacent path points on the x-axis, Δy is the distance between adjacent path points on the y-axis, β new,i is the pitch angle of the i-th repaired population particle, α new,i is the yaw angle of the i-th repaired population particle.

[0067] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0068] The present invention constructs a three-dimensional map of the real environment through mathematical modeling, and randomly initializes the population and weight vector. By using the idea of ​​decomposing multi-objective optimization, the multi-objective optimization problem is transformed into optimizing a series of single-objective optimization problems. The particles with good convergence and uniform density distribution characteristics are screened out through the nonlinear adjustment particle swarm algorithm, and the optimal path is finally output as the navigation solution of the UAV. This method improves the path planning efficiency and safety of UAVs in complex environments, and solves the shortcomings of traditional algorithms in terms of computational efficiency and robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 This is a flow chart of the UAV path planning method according to the present invention;

[0070] Figure 2 This is a flow chart of the UAV path planning method described in this embodiment;

[0071] Figure 3 A top view of the optimal path planning of each algorithm described in this embodiment after 500 iterations;

[0072] Figure 4 This is a main view of the optimal path planning of each algorithm described in this embodiment after 500 iterations;

[0073] Figure 5 It is a side view of the optimal path planning of each algorithm described in this embodiment after 500 iterations. DETAILED DESCRIPTION

[0074] The drawings are for illustrative purposes only and should not be construed as limiting the present patent;

[0075] In order to better illustrate the present embodiment, some parts in the drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product;

[0076] It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0077] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0078] Example 1

[0079] A UAV path planning method based on multi-objective adaptive particle swarm optimization, such as Figure 1 As shown, the following steps are included:

[0080] S1: constructing a mathematical model for the three-dimensional map, wherein the mathematical model simulates obstacles in the three-dimensional map;

[0081] S2: setting the objective function and constraint function according to the mathematical model;

[0082] S3: Initialize the parameters of the particle swarm optimization algorithm and divide the particle swarm into multiple sub-populations;

[0083] S4: Iterate the current multiple sub-populations through the particle swarm optimization algorithm to obtain the iterated sub-populations;

[0084] S5: calculating the population movement density estimation values ​​of all the iterative sub-populations according to the population movement density estimation mechanism, and selecting multiple iterative sub-populations according to the population movement density estimation values;

[0085] S6: If there are multiple filtered sub-populations that satisfy the constraint function, then the UAV flight path set generated by the filtered sub-populations is output, and the UAV flight path corresponding to the minimum value of the objective function is selected from the UAV flight path set. If there are sub-populations that do not satisfy the constraint function, the sub-populations that do not satisfy the constraint function are repaired and the process returns to step S4.

[0086] In the specific implementation process, Figure 2 As shown, in a static environment, mountains and radars are the main obstacles that simulate the real environment.

[0087] Step 1: Use mathematical modeling to build a 3D map of the real environment. In this process, mountains and radar obstacles are simulated in the form of a 3D grid;

[0088] Step 2: Randomly initialize the population and weight vector, and calculate the objective function value of the current population;

[0089] Step 3: Divide the weight vectors into K weight vector subsets according to the distances between each pair, that is, divide the population into K sub-populations according to the weight vectors;

[0090] Step 4: nonlinearly adjust the learning factor of the particle swarm algorithm according to the total number of iterations and the chaos factor, and nonlinearly adjust the inertia factor of the particle swarm algorithm according to the ratio of updated particles to unupdated particles in the population and the chaos factor;

[0091] Step 5: Execute the particle swarm optimization algorithm to use the parent population of N particles to generate the offspring population of N particles, and merge them to obtain a temporary population with a population size of 2N;

[0092] Step 6: Calculate the population movement density estimation value of each particle in the temporary population according to the population movement density estimation mechanism, and select the first N smaller population movement density estimation values, which represent the selected particles with superior convergence and density distribution;

[0093] Step 7: Repeat the above steps until the preset number of iterations is met or other termination conditions are reached, such as the quality of the solution reaches a predetermined threshold, and output the optimal path as the navigation solution for the UAV.

[0094] The present invention uses a particle swarm optimization algorithm to plan a flight path that best meets the needs for the UAV. The goal is to plan the shortest possible flight path for the UAV under the premise of minimizing collisions with obstacles. The population movement density estimation value of each particle in the temporary population is calculated according to the population movement density estimation mechanism, and the first N smaller population movement density estimation values ​​are screened out. Finally, a set of optimal solutions is obtained, and a solution that meets the needs is selected, that is, an optimal UAV flight path.

[0095] Example 2

[0096] This embodiment further discloses the following contents based on the embodiment 1:

[0097] Step S1, constructing a mathematical model for the three-dimensional map, wherein the mathematical model simulates obstacles in the three-dimensional map, includes:

[0098]

[0099] In the formula, is the three-dimensional coordinate of the kth mountain, is the horizontal center of the kth mountain, h k is the first design parameter, x′ k is the second design parameter, y′ k is the third design parameter.

[0100] In the specific implementation process, in order to avoid collision with obstacles, the path of the drone should be as far away from mountains and radar obstacles as possible. Mountains and radars are regarded as areas where drones cannot pass. Drones cannot collide with them and need to take a detour. The data grid method is used to construct a simulated three-dimensional map of the real environment, such as Figure 3 , Figure 4 and Figure 5As shown, the simulation environment is at a scale of 60m*60m*60m. The eight peaks in the figure simulate mountain obstacles and radar obstacles in the real environment.

[0101] Example 3

[0102] This embodiment further discloses the following contents based on Embodiments 1 and 2:

[0103] In step S3, the parameters of the particle swarm optimization algorithm are initialized, including:

[0104] Set the maximum inertia weight w of the particle swarm algorithm parameters max and the minimum value w min , total number of iterations t max , maximum learning factor c max and the minimum value c min .

[0105] In the specific implementation process, the population and weight vector size are set to N = 100, and the total number of iterations is t max =500, the maximum number of times the population can be replaced by offspring in each iteration is nr, and the number of sub-populations is K.

[0106] Set the initial value of the particle swarm algorithm parameter, ω max =0.9,ω min =0.4,c max =2.5,c min =1.5.

[0107] Example 4

[0108] This embodiment further discloses the following contents based on Embodiments 1, 2 and 3:

[0109] The objective function includes calculating the flight path length and the number of collisions between the drone and obstacles, including:

[0110]

[0111] Where F(x) is the objective function, η is the preset constraint violation threshold, and f obj1 (x) is the first objective function, used to calculate the flight path length, f obj2 (x) is the second objective function, which is used to calculate the number of collisions between the drone and obstacles, f con1 (x) is the first constraint function, f con2 (x) is the second constraint function.

[0112]

[0113] Where, d k,i,jis the distance of the k-segment path between the ith path point and the jth radar position, R max is the maximum detection radius of the radar, A i,j is the penalty value for the collision between the i-th path point and the j-th radar, N p is the total number of path points, B i,q is the penalty value for the collision between the i-th path point and the q-th mountain peak height, n is the number of segments between the current path point and the next path point, is the distance of the k-segment path between the ith path point and the qth peak height, z i,q is the distance between the i-th path point and the q-th peak height;

[0114]

[0115] (x i-1 ,y i-1 ,z i-1 ) and (x i ,y i ,z i ) are the positions of the i-1th path point and the i-th path point, respectively. p is the total number of path points.

[0116] Further, in step S6, the constraint function includes:

[0117] The i-th pitch angle β i The calculation formula is:

[0118]

[0119] x i+1 -x i is the distance between the i+1th path point and the ith path point on the x-axis, y i+1 -y i is the distance between the i+1th path point and the ith path point on the y axis, z i+1 -z i is the distance between the i+1th path point and the ith path point on the z-axis;

[0120] The present invention adopts a dual-objective optimization model that comprehensively considers the number of collisions and the length of the flight path, and considers the two constraints of yaw angle and pitch angle at the same time. The proposed algorithm is committed to planning a safe and efficient flight path. The present invention takes into account that in a real environment, when a drone encounters an emergency, the flight trajectory may deviate from the path set by the kinematic principle.

[0121] The calculation formula of the pitch angle constraint function is:

[0122]

[0123] Among them, β min is the minimum pitch angle, β max is the maximum pitch angle, and C i is the penalty value for violating the pitch angle constraint of the i-th path point;

[0124] The i-th yaw angle α i The calculation formula is:

[0125]

[0126] The i-th yaw angle α i The constraint function is:

[0127]

[0128] In the formula, α max is the maximum yaw angle, α min is the minimum yaw angle, D i is the penalty value for violating the yaw angle constraint of the i-th path point.

[0129] In step S2, the particle group is divided into a plurality of sub-populations, including:

[0130] The particle swarm is divided into multiple sub-populations by calculating the distance of the initialized weight vector between each particle in the particle swarm.

[0131] In step S4, the current multiple sub-populations are iterated by the particle swarm optimization algorithm to obtain the iterated sub-populations, including:

[0132] S4.1: Taking the bifurcation factor and the number of iterations as the input of the particle swarm optimization algorithm, the chaos factor is generated;

[0133]

[0134] Where ρ(t) is the chaos factor of the t-th iteration, μ is the bifurcation factor, and t is the t-th iteration;

[0135] S4.2: Calculate the first learning factor c1 and the first learning factor c2 according to the tth iteration number and the total iteration number and the chaotic factor obtained in step S4.1;

[0136]

[0137] S4.3: Adjust inertia weights based on population renewal rate;

[0138]

[0139] In the formula, Ratio is the population update rate, success_num is the total number of times the optimal solution of the population particles is updated, N is the total population size, and w is the inertia weight;

[0140] S4.4: Calculate the population particle velocity and current position according to the inertia weight, and obtain the population particles of the t+1th iteration and the tth iteration;

[0141] v i (t+1)=wv i (t)+c1r1(pbest i (t)-x i (t))+c2r2(gbest(t)-x i (t));

[0142] x i (t+1)=x i (t)+v i (t+1);

[0143] Where w is the inertia weight, pbest i (t) is the historical optimal position of the i-th population particle, gbest(t) is the global optimal position, v i (t+1) is the moving speed of the population particles in the tth iteration, x i (t+1) is the position of the population particle at the t+1th iteration, r1 and r2 are random numbers;

[0144] S4.5: Merge the population particles of the t+1th iteration and the tth iteration to obtain a synthetic population with a population size of 2N;

[0145] In the specific implementation process, step A: chaotic factors are generated by a chaotic mapping method;

[0146] Step B: Nonlinearly adjust the learning factors c1 and c2 according to the current iteration number, total iteration number and chaos factor

[0147] Step C: The population update rate Ratio is defined as the ratio of the total number of updates of the last optimal solution of the swarm particles to the total swarm size. The larger the Ratio is, the more promising the current movement direction of the swarm particles is. In this case, the value of ω should be increased. If the algorithm is the first iteration, set w = w max .

[0148] Example 5

[0149] Based on Embodiments 1, 2, 3 and 4, this embodiment further discloses the following contents: In step S5, the population movement density estimation values ​​of all the iterative sub-populations are calculated according to the population movement density estimation mechanism, and multiple iterative sub-populations are selected according to the size of the population movement density estimation value, including:

[0150] Calculate the population movement density estimation values ​​of all the iterated sub-populations according to the population movement density estimation mechanism, sort each population particle from high to low according to the population movement density estimation value, retain the first N population particles, and eliminate the last N population particles;

[0151] D′(p,Pop)=D(dist(p,q′1), dist(p,q′2),...,dist(p,q′ N-1 ))

[0152] D′(p,Pop) is the density estimate of the population particle p in the population Pop, dist() is the similarity between the two population particles, p is the pth population particle in the population Pop, q′ N-1 is the population particle after p displacement.

[0153] In the specific implementation process, the population movement density estimation mechanism calculates the population movement density estimation value of each individual in the synthetic population, thereby screening out the first N individuals with smaller population movement density estimation values, which means that these individuals have good convergence and uniform density distribution characteristics. The individuals with larger population movement density estimation values ​​are directly discarded, which means that these individuals have poor convergence or poor density distribution.

[0154] In step S6, if there are multiple filtered sub-populations that satisfy the constraint function, the drone flight path set generated by the filtered sub-populations is output, and the drone flight path corresponding to the minimum value of the objective function is selected from the drone flight path set. If there are sub-populations that do not satisfy the constraint function, the sub-populations that do not satisfy the constraint function are repaired, and the process returns to step S4, including:

[0155] If the screened sub-population meets the constraints, the UAV flight path corresponding to the minimum value of the first objective function or the minimum value of the second objective function is output from the sub-population that meets the constraints. If there is a screened sub-population that does not meet the constraints, the sub-population that does not meet the constraints is repaired and returns to step S4.4.

[0156] The repairing of the sub-population that does not satisfy the constraint condition includes:

[0157]

[0158]

[0159] Δx=x(i+1)-x(i)

[0160] Δy=y(i+1)-y(i)

[0161]

[0162] In the formula, X min ,X max ,Y min ,Y max ,Z min ,Z max is the boundary value of the UAV flight space, Dis is the distance between adjacent path points on the horizontal plane, r is a random number, Δx is the distance between adjacent path points on the x-axis, Δy is the distance between adjacent path points on the y-axis, β new ,i is the pitch angle of the i-th repaired population particle, α new , i is the yaw angle of the i-th repaired particle in the population. The constraint repair technology in step S5 is used to repair the particles that violate the constraints to within the constraint range, so that all particles in the population are feasible solutions. In order to reduce the computational burden, constraint repair is performed every m iterations.

[0163] In the specific implementation process, it is determined whether the algorithm meets the iteration conditions. If not, the population evolves according to step S4.4, and the particle swarm algorithm continues to optimize the search process. Otherwise, a set of optimal solutions is output (the individuals in the population whose number of constraint violations is less than the preset constraint violation threshold when the algorithm reaches the termination condition), and the optimal solution that best meets the requirements is selected from them (if the focus is on pursuing the shorter path length, the first objective function min (f obj1 (x)); if the focus is on pursuing the least number of collisions, then the second objective function min(f obj2 (x)), that is, to plan a feasible optimal path for the UAV from the starting point to the end point.

[0164] Example 6

[0165] This embodiment, based on Embodiments 1, 2, 3, 4 and 5, continues to disclose the following contents:

[0166] The simulation experiment was carried out in Matlab2023a, and representative similar algorithms MOEADDAE, CCMO, MCCMO, ICMA, ToP, MOEA2DE and MOEADTF were selected for comparison. The basic information is shown in Table 1. In addition, the parameters of the experimental setting are: N is 100, that is, the population size (number of individuals), t maxThe value of is 500, which is the maximum number of iterations of the population. m is the number of iterations between the execution of the repair mechanism, T is the size of the sub-population, and the number of sub-populations K = N / T. Each algorithm is independently iterated 30 times in each case. Figure 5 The visualization comparison of 30 best path planning results of different algorithms in 500 iterations is shown. Tables 2 and 3 respectively give the average ranking of the maximum value (max), minimum value (min), average value (mean) and standard deviation (std) performance indicators of each algorithm in 30 independent runs of the two objectives of number of collisions and path length. According to the simulation experimental results, the optimal solution found by the method proposed in the present invention has surpassed other existing methods in the simulation.

[0167] Table 1 shows the basic information of each algorithm.

[0168]

[0169] Table 2

[0170] Performance Indicators MOEADDAE CCMO MCCMO ICMA ToP MOEA2DE MOEADTF Proposed max 4 2 3 6 7 5 8 1 min 6 4 1 5 7 3 8 2 mean 4 2 3 5 7 6 8 1 std 4 2 3 6 7 5 8 1

[0171] Table 3

[0172]

[0173]

[0174] The same or similar reference numerals correspond to the same or similar components;

[0175] The terms used in the drawings to describe positional relationships are only used for illustrative purposes and should not be construed as limiting this patent;

[0176] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A UAV path planning method based on multi-objective adaptive particle swarm optimization, characterized in that: The following steps are involved: S1: constructing a mathematical model for the three-dimensional map, wherein the mathematical model simulates obstacles in the three-dimensional map; S2: setting the objective function and constraint function according to the mathematical model; S3: Initialize the parameters of the particle swarm optimization algorithm and divide the particle swarm into multiple sub-populations; S4: Iterate the current multiple sub-populations through the particle swarm optimization algorithm to obtain the iterated sub-populations; S5: calculating the population movement density estimation values ​​of all the iterative sub-populations according to the population movement density estimation mechanism, and selecting multiple iterative sub-populations according to the population movement density estimation values; S6: If there are multiple filtered sub-populations that satisfy the constraint function, then the UAV flight path set generated by the filtered sub-populations is output, and the UAV flight path corresponding to the minimum value of the objective function is selected from the UAV flight path set. If there are sub-populations that do not satisfy the constraint function, the sub-populations that do not satisfy the constraint function are repaired and the process returns to step S4.

2. The multi-objective adaptive particle swarm optimization UAV path planning method according to claim 1 is characterized in that: Step S1, constructing a mathematical model for the three-dimensional map, wherein the mathematical model simulates obstacles in the three-dimensional map, includes: In the formula, is the three-dimensional coordinate of the kth mountain, is the horizontal center of the kth mountain, h k is the first design parameter, x′ k is the second design parameter, y′ k is the third design parameter.

3. The multi-objective adaptive particle swarm optimization method for unmanned aerial vehicle path planning according to claim 1 is characterized in that: In step S3, the parameters of the particle swarm optimization algorithm are initialized, including: Set the maximum inertia weight w of the particle swarm algorithm parameters max and the minimum value w min , total number of iterations t max , maximum learning factor c max and the minimum value c min .

4. The multi-objective adaptive particle swarm optimization UAV path planning method according to claim 3 is characterized in that: The objective function includes calculating the flight path length and the number of collisions between the drone and obstacles, including: Where F(x) is the objective function, η is the preset constraint violation threshold, and f obj1 (x) is the first objective function, used to calculate the flight path length, f obj2 (x) is the second objective function, which is used to calculate the number of collisions between the drone and obstacles, f con1 (x) is the first constraint function, f con2 (x) is the second constraint function. Where, d k,i,j is the distance of the k-segment path between the ith path point and the jth radar position, R max is the maximum detection radius of the radar, A i,j is the penalty value for the collision between the i-th path point and the j-th radar, N p is the total number of path points, B i,q is the penalty value for the collision between the i-th path point and the q-th mountain peak height, n is the number of segments between the current path point and the next path point, is the distance of the k-segment path between the ith path point and the qth peak height, z i,q is the distance between the i-th path point and the q-th peak height; (x i-1 ,y i-1 ,z i-1 ) and (x i ,y i ,z i ) are the positions of the i-1th path point and the i-th path point, respectively. p is the total number of path points.

5. The multi-objective adaptive particle swarm optimization UAV path planning method according to claim 4 is characterized in that: In step S6, the constraint conditions of the objective function include: The i-th pitch angle β i The calculation formula is: x i+1 -x i is the distance between the i+1th path point and the ith path point on the x-axis, y i+1 -y i is the distance between the i+1th path point and the ith path point on the y axis, z i+1 -z i is the distance between the i+1th path point and the ith path point on the z-axis; The calculation formula of the pitch angle constraint function is: Among them, β min is the minimum pitch angle, β max is the maximum pitch angle, and C i is the penalty value for violating the pitch angle constraint of the i-th path point; The i-th yaw angle α i The calculation formula is: The i-th yaw angle α i The constraint function is: In the formula, α max is the maximum yaw angle, α min is the minimum yaw angle, D i is the penalty value for violating the yaw angle constraint of the i-th path point.

6. The multi-objective adaptive particle swarm optimization UAV path planning method according to claim 1 is characterized in that: In step S3, the particle group is divided into a plurality of sub-populations, including: The particle swarm is divided into multiple sub-populations by calculating the distance of the initialized weight vector between each particle in the particle swarm.

7. The multi-objective adaptive particle swarm optimization UAV path planning method according to claim 3 is characterized in that: In step S4, the current multiple sub-populations are iterated by the particle swarm optimization algorithm to obtain the iterated sub-populations, including: S4.1: Taking the bifurcation factor and the number of iterations as the input of the particle swarm optimization algorithm, the chaos factor is generated; Where ρ(t) is the chaos factor of the t-th iteration, μ is the bifurcation factor, and t is the t-th iteration; S4.2: Calculate the first learning factor c1 and the first learning factor c2 according to the tth iteration number and the total iteration number and the chaotic factor obtained in step S4.1; S4.3: Adjust inertia weights based on population renewal rate; Where Ratio is the population update rate, success-num is the total number of times the optimal solution of the population particles is updated, N is the total population size, and w is the inertia weight; S4.4: Calculate the population particle velocity and current position according to the inertia weight, and obtain the population particles of the t+1th iteration and the tth iteration; v i (t+1)=wv i (t)+c1r1(pbest i (t)-x i (t))+c2r2(gbest(t)-x i (t)); x i (t+1)=x i (t)+v i (t+1); Where w is the inertia weight, pbest i (t) is the historical optimal position of the i-th population particle, gbest(t) is the global optimal position, v i (t+1) is the moving speed of the population particles in the tth iteration, x i (t+1) is the position of the population particle at the t+1th iteration, r1 and r2 are random numbers; S4.5: Merge the population particles of the t+1th iteration and the tth iteration to obtain a synthetic population with a population size of 2N.

8. The multi-objective adaptive particle swarm optimization UAV path planning method according to claim 1 is characterized in that: In step S5, the population movement density estimation mechanism is used to calculate the population movement density estimation values ​​of all the iterated sub-populations, and multiple iterated sub-populations are selected according to the population movement density estimation values, including: Calculate the population movement density estimation values ​​of all the iterated sub-populations according to the population movement density estimation mechanism, sort each population particle from high to low according to the population movement density estimation value, retain the first N population particles, and eliminate the last N population particles; D′(p,Pop)=D(dist(p,q′1),dist(p,q′2),...,dist(p,q′ N-1 )) D′(p,Pop) is the density estimate of the population particle p in the population Pop, dist() is the similarity between the two population particles, p is the pth population particle in the population Pop, q′ N-1 is the population particle after p displacement.

9. The multi-objective adaptive particle swarm optimization UAV path planning method according to claim 4 is characterized in that: In step S6, if there are multiple filtered sub-populations that satisfy the constraint function, the drone flight path set generated by the filtered sub-populations is output, and the drone flight path corresponding to the minimum value of the objective function is selected from the drone flight path set. If there are sub-populations that do not satisfy the constraint function, the sub-populations that do not satisfy the constraint function are repaired, and the process returns to step S4, including: If the screened sub-population meets the constraints, the UAV flight path corresponding to the minimum value of the first objective function or the minimum value of the second objective function is output from the sub-population that meets the constraints. If there is a screened sub-population that does not meet the constraints, the sub-population that does not meet the constraints is repaired and returns to step S4.

4.

10. The multi-objective adaptive particle swarm optimization UAV path planning method according to claim 9 is characterized in that: The repairing of the sub-population that does not satisfy the constraint condition includes: Δx=x(i+1)-x(i) Δy=y(i+1)-y(i) Where, X min ,X max ,Y min ,Y max ,Z min ,Z max is the boundary value of the UAV flight space, Dis is the distance between adjacent path points on the horizontal plane, r is a random number, Δx is the distance between adjacent path points on the x-axis, Δy is the distance between adjacent path points on the y-axis, β new,i is the pitch angle of the i-th repaired population particle, α new,i is the yaw angle of the i-th repaired population particle.

Images