Flower pollination optimization method and implementation system based on nonlinear transgenerational differential evolution

Abstract

The present invention relates to a flower pollination optimization method and implementation system based on nonlinear intergenerational differential evolution, and belongs to the field of computer science. The method comprises the following steps: S1: nonlinear inertia weight setting; S2: intergenerational differential evolution; S3: parameter adaptive adjustment; S4: intergenerational roulette jump out of local optimal solution. The method can set nonlinear inertia weight to enable the flower pollination algorithm to enhance the search capability at a faster speed in the early stage of global search, thereby improving the search convergence speed of the algorithm. Adaptive adjustment realizes dynamic update of scaling factor and crossover probability, thereby improving population richness and reducing the number of local solutions, and then combining the intergenerational roulette method to reduce the probability of falling into the local optimal solution.

Description

Technical Field

[0001] The invention belongs to the field of computers and relates to a flower pollination optimization method and an implementation system based on nonlinear cross-generational differential evolution. Background Art

[0002] With the development and innovation of science and technology such as the Internet of Things and artificial intelligence, there are a large number of complex high-dimensional variable optimization problems in engineering and industrial applications that need to be solved, such as urban public facilities planning and construction, electronic networks and communications, unmanned vehicle intelligent path setting, drone intelligent inspection, etc. At present, many studies on these problems have adopted heuristic optimization algorithms, which are developed based on the characteristics of biological systems and physical theories, such as Darwin's theory of evolution, insect or animal group behavior, etc., and have the characteristics of high search accuracy, high efficiency and strong robustness.

[0003] Heuristic optimization algorithms can be roughly divided into three categories: 1) Algorithms that can balance local search and global search, such as the Artificial Bee Colony (ABC) algorithm and the Social Spider Optimization (SSO) algorithm. These algorithms can combine rapidity and global optimality well, but in the later stage of the search process, the algorithm performance will deteriorate. 2) Algorithms that have both global convergence and computational robustness, such as the Wolf Pack Algorithm (WPA) and the Whale Algorithm (WA). These algorithms can search from multiple points at the same time and have good convergence performance, but they have low search efficiency and insufficient population diversity. 3) Algorithms with good population richness, such as Particle Swarm Optimization (PSO) and Genetic Algorithm (GA). However, these algorithms have the problem of falling into local optimality due to numerical perturbations. In summary, traditional heuristic optimization algorithms have disadvantages such as premature convergence and numerical fluctuations, and their ability to explore optimal solutions needs to be improved.

[0004] In view of the above shortcomings of traditional heuristic optimization algorithms, Yang proposed a flower pollination algorithm (FPA) based on flowering plants, which has been widely used in multi-objective optimization. FPA can continuously complete the search tasks for multiple objects with high accuracy and good convergence performance, and maintain a good balance between local and global search processes. This algorithm is widely used in various optimization problems. For example, T. Nguyen used FPA to solve the problem of data oscillation in the optimization process, and KPPANDA used FPA to eliminate target harmonics in symmetrical and asymmetrical multilevel inverters. By analyzing related research, FPA also has three main problems:

[0005] 1) The local search process lacks search guidance and may fall into local optimality, resulting in reduced search efficiency and convergence.

[0006] 2) It is sensitive to parameters. The setting of parameters such as inertia weight will affect the accuracy and number of iterations of the global search for the optimal solution.

[0007] 3) The increase in the complexity of the optimization problem increases the number of local optimal solutions, which hinders FPA from obtaining the optimal solution. Summary of the invention

[0008] In view of this, the object of the present invention is to provide a flower pollination optimization method and implementation system based on nonlinear cross-generational differential evolution. In order to solve the shortcomings of insufficient optimization accuracy, poor convergence effect, poor stability, and easy to fall into local optimal solutions in the optimization process of heuristic optimization algorithms, a flower pollination algorithm based on nonlinear cross-generational differential evolution (FPA-NCDE) is proposed. The method uses cross-generational differential evolution to guide individuals to approach the optimal solution, so that the local search process of the algorithm has guidance. At the same time, the nonlinear inertia weight is set to enhance the search ability at a faster speed in the initial stage of the global search, and improve the search convergence speed of the algorithm. In addition, the scaling factor and the crossover probability are dynamically updated by performing parameter adaptive adjustment, thereby improving the population richness and reducing the number of local solutions, and combining the cross-generational roulette method to reduce the probability of falling into the local optimal solution.

[0009] In order to achieve the above object, the present invention provides the following technical solutions:

[0010] A flower pollination optimization method based on nonlinear transgenerational differential evolution, the method comprising the following steps:

[0011] S1: nonlinear inertia weight setting;

[0012] S2: Transgenerational differential evolution;

[0013] S3: adaptive adjustment of parameters;

[0014] S4: Cross-generational roulette jumps out of the local optimal solution.

[0015] Optionally, the S1 is specifically:

[0016] Based on the simulated annealing algorithm (SA) with the advantages of global exploration and local development, the nonlinear inertia weight is designed as follows:

[0017]

[0018] k1=1-0.5h+0.2s (2)

[0019]

[0020]

[0021] Where T max is the maximum number of iterations, t is the current number of iterations, is the minimum fitness value of the current population, is the minimum fitness value of the previous generation population; the change of weight k can guide the global search direction and reduce the cost of blind search; based on the nonlinear inertia weight, the iterative formula of pollen in global search is:

[0022]

[0023] in, and are the upper and lower limits of the pollen individual value in each dimension, Refers to the average value of the upper and lower limit dimensions of each individual, and rand is a random number distributed in [0,1].

[0024] Optionally, S2 is specifically:

[0025] Cross-generation differential evolution includes two mutation strategies, namely, neighborhood-based cross-generation strategy (NCG) and population-based cross-generation strategy (PCG). NCG and PCG guide the search direction to approach the optimal solution by analyzing the differences between consecutive generations, and combine the information of two generations of populations to reduce the numerical oscillation of search results.

[0026] In order to improve the search stability of FPA-NCDE in the local search process, a nonlinear arccosine acceleration factor c is designed to make the search enter the local search faster at the beginning and maintain a high search speed at the end without premature convergence; its definition is as follows:

[0027]

[0028] Among them, d num is the dimension of pollen individuals, t is the number of iterations, T max is the maximum number of iterations; let c max =1.0,c min =0.6; As the number of algorithm iterations increases, the decay rate slows down, improving the local search capability of the algorithm;

[0029] NCG calculates the Euclidean distance between the parent individual and other individuals in the population, selects the nearest T individuals to form a neighborhood pool, and selects the previous generation of T individuals as members of the neighborhood pool based on the Euclidean distance. Its expression is as follows:

[0030]

[0031] Among them, V i,g is the mutation vector representing the new individual produced after differential evolution, i is the index of the parent individual, g is the number of generations of differential mutation, is a randomly selected individual from the neighborhood pool of the current generation, is an individual randomly selected from the neighborhood pool of the previous generation parent; rn1 is an individual randomly selected from {I i,g,1 ,I i,g,2 ,...I i,g,T}A random integer selected, {I i,g,1 ,I i,g,2 ,...I i,g,T} records the index of individuals in the neighborhood pool, rn2 is from {I i,g-1,1 ,I i,g-1,2 ,...I i,g-1,T}A randomly selected integer, the size of T is 5% of the total number of individuals in the population;

[0032] PCG increases the richness of the population by randomly selecting individuals from the previous generation population of the parent individual and individuals from the contemporary population to participate in the mutation process. Its expression is as follows:

[0033]

[0034] Among them, x i,g For the parent individual, is a randomly selected individual from the current population, is an individual randomly selected from the previous generation population, rp1 and rp2 are random integers selected from {1,2,...N}, and N is the population size; in order to balance the convergence and search capabilities, the probability of using PCG and NCG in the mutation operation is the same.

[0035] Optionally, S3 is specifically:

[0036] During parameter initialization, the initial scaling factor F i,g and the crossover probability CR i Because it is from [F min ,F max ] and [CR min ,CR max ] is randomly generated, F min and F max are the upper and lower limits of the scaling factor F, CR min and CR max are the upper and lower limits of the crossover probability;

[0037] The expression for dynamic update of the scaling factor is as follows:

[0038]

[0039] Randomly select three individuals x from the neighborhood of the current individual p1 、x p2 、x p3 , whose fitness values ​​are f1, f2, f3 and f1<f2<f3; similarly, three individuals x are randomly selected from the neighborhood pool of the previous generation of parent individuals p4 ,x p5 ,x p6 The fitness values ​​are f4, f5, f6 and f4<f5<f6; the expression of dynamic update of crossover probability is as follows:

[0040]

[0041] where f ave is the average fitness of all individuals in the current population.

[0042] Optionally, the S4 is specifically:

[0043] After completing the search iterations for all individuals, select T individuals with the smallest fitness value, and then randomly select α individuals from the remaining individuals in the current population and β parent individuals from the previous generation. A total of 2T individuals form a roulette pool to participate in the roulette process; the probability of each pollen individual in the roulette pool being selected is as follows:

[0044]

[0045] Where i is the pollen individual, wig i is the mapping weight of the fitness value;

[0046] The process of intergenerational roulette is as follows:

[0047] S41: sort the pollen individuals in the parent pollen population in ascending order according to fitness values;

[0048] S42: Select the top T pollen individuals to form a subpopulation P t ;

[0049] S43: Randomly select β individuals from the remaining individuals in the parent population;

[0050] S44: randomly select α individuals from the contemporary population;

[0051] S45: The subpopulation and the selected individuals form a roulette pool W t ;

[0052] S46: According to formula (11), the individuals in the roulette pool participate in the roulette gambling process.

[0053] A flower pollination optimization system based on nonlinear transgenerational differential evolution comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor. When the processor executes the computer program, the method described above is implemented.

[0054] A computer-readable storage medium stores a computer program, which implements the method described above when executed by a processor.

[0055] The beneficial effect of the present invention is that the method can set nonlinear inertia weights to enable the flower pollination algorithm to enhance the search capability at a faster speed in the early stage of global search, thereby improving the search convergence speed of the algorithm. Adaptive adjustment is used to dynamically update the scaling factor and crossover probability, thereby increasing the population richness and reducing the number of local solutions, and combining the cross-generational roulette method to reduce the probability of falling into the local optimal solution.

[0056] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0058] Figure 1 Path planning model for UAV intelligent inspection;

[0059] Figure 2 It is a convergence curve diagram fitted by the cost function value under each algorithm iteration number;

[0060] Figure 3 It is a local enlarged diagram of the convergence curve of the cost function of each algorithm;

[0061] Figure 4 Plan a 3D map for the drone’s flight path;

[0062] Figure 5 A bird's-eye view of the flight path planned for the drone. DETAILED DESCRIPTION

[0063] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0064] Among them, the drawings are only used for illustrative explanations, and they only represent schematic diagrams rather than actual pictures, and should not be understood as limitations on the present invention. In order to better illustrate the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0065] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "front", "rear", etc. indicate the orientation or position relationship, they are based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the terms describing the position relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0066] Figure 1It is a path planning model for UAV intelligent inspection. The straight line segment between the flight start and end point is divided into D equal parts. A perpendicular line is drawn through each equal-division point on the line segment. Appropriate points are selected on the perpendicular line where each equal-division point is located as track points. The trajectory connecting these track points is the path planning route of the UAV flight. Assume that the position coordinates of obstacles and threat targets encountered by the UAV during flight in the new coordinate system are (x i ,y i ), the threat radius is r i , then the attack range of the threat target is:

[0067] (xx i ) 2 +(yy i ) 2 =r i 2 (12)

[0068] In the process of drone path planning, there are three constraints, namely flight length limit, turning angle limit, and flight altitude limit. The following will explain them in detail:

[0069] 1) Flight length limit

[0070] The first constraint is the flight length limit. Before each turn, there is a straight-line distance for error correction. The minimum flight length affects the trajectory of the drone. Assume that the starting point of the flight trajectory is S and the end point is T, and ST is segmented into D+1 segments. The length of each segment should be greater than the minimum flight length limit. The maximum total length of each segment follows the following conditions:

[0071]

[0072] where ||P i-1 P i || is the length of the segment, P0 is the starting point, P D+1 It's the end point, L max is the maximum value of the total flight length.

[0073] 2) Turning Angle Limit

[0074] The turning angle of the drone between two adjacent segments follows the following conditions:

[0075]

[0076] where θ max is the maximum steering angle.

[0077] 3) Flight altitude restrictions

[0078] The flight altitude h of the drone complies with the following conditions:

[0079] H min <h<H max (15)

[0080] H max is the maximum flight altitude, H min is the minimum flight altitude.

[0081] Based on the above constraints, energy consumption cost, turning cost, threat attack cost and flight altitude cost are considered as the performance evaluation criteria of the UAV path planning algorithm.

[0082] 1) Energy consumption cost

[0083] Energy cost Cost e It is related to the speed of the UAV and the total flight length, and its expression is as follows:

[0084]

[0085] Where η is the energy consumption factor, v is the speed of the drone, l i is the length of the i-th segment of the total flight track.

[0086] 2) Turning cost

[0087] The steering cost is related to the steering angle. Formula (17) gives the steering cost Cos t s Calculation

[0088]

[0089] where θ j is the steering angle, k is the steering coefficient, θ max is the maximum steering angle.

[0090] 3) Threat attack cost

[0091] The flight trajectory is more susceptible to threat targets such as communication radars, base stations and buildings, and the threat attack cost Cost needs to be considered. t The threat level of each track point is expressed as:

[0092]

[0093] Where kj is the threat level of the jth threat target, a and b are threat weights, and d xj It refers to the distance between the UAV at point x and the center of the jth threat target, R j In order to calculate the threat attack cost of each threat target on each segment path, the length of each segment is divided into 15 parts. The threat attack cost of each segment is the location The average cost of all threat attacks, such as Figure 1 As shown in Figure 2, the threat attack cost is Cost t for:

[0094]

[0095] l i is the length of the ith flight path segment, and N is the number of threat targets.

[0096] 4) Flight altitude cost

[0097] The flight altitude cost is related to the flight altitude of each segment. The flight altitude cost Cost h for:

[0098] Cost h =μ(h i -h best ) (20)

[0099] Where μ is the flight height factor, h i is the flight height of the UAV in the ith segment, h best The best flying altitude.

[0100] In summary, the cost function of UAV inspection path planning is:

[0101]

[0102] Among them, λ1, λ2, λ3, and λ4 represent Cost t 、Cos t e 、Cos t h 、Cos t s The weight of .

[0103] In order to better illustrate the effectiveness of the present invention, the present invention uses simulation experiments to illustrate it. Table 1 shows the simulation parameters.

[0104] Table 1 Simulation parameters

[0105]

[0106] In the simulation experiment, six algorithms were compared, namely, Flower Pollination Algorithm (FPA), Artificial Bee Colony (ABC), Spider Swarm Optimization (SSO), Particle Swarm Optimization (PSO), Genetic Algorithm (GA) and the algorithm proposed in the present invention. The average values ​​of each optimization algorithm independently run 20 times are shown in Table 2.

[0107] Table 2 The average value of each optimization algorithm running independently 20 times

[0108]

[0109] It can be seen that the time consumed by the FPA-NCDE algorithm designed by the present invention in the optimization process of the UAV path planning is better than that of FPA, ABC, PSO and SSO, and slightly worse than that of the GA algorithm. In terms of the cost consumption of UAV path planning, it is reduced by 65.7% compared with FPA, 2.8% compared with ABC, 1.7% compared with PSO, 1.7% compared with SSO, and 0.9% compared with GA. In terms of average variance, FPA-NCDE is reduced by 99.8% compared with FPA, 3.2% compared with ABC, 63.1% compared with PSO, 80.7% compared with SSO, and 89.8% compared with GA. It can be seen that compared with other comparative algorithms, FPA-NCDE can maintain good efficiency in UAV path planning while consuming the least flight cost, and the average error of the algorithm is the smallest, and the robustness of the algorithm is better.

[0110] Figure 2 and Figure 3 They are respectively the convergence curve fitted by the cost function value under each algorithm iteration number and its corresponding local enlarged diagram. Figure 2 It can be seen that the cost of planning the flight path by the FPA algorithm is much greater than that of other algorithms. Figure 3 It can be seen that the ABC algorithm found the optimal solution only after about 900 iterations, and ABC, PSO, and GA all have the disadvantage of early convergence and cannot reduce the flight cost. The number of algorithm iterations required for SSO to find the optimal solution is much greater than that of other algorithms. When the number of iterations of FPA-NCDE is 700, the cost function value no longer changes, and the cost consumed at this time is much less than that of other algorithms. When the algorithm starts running, the convergence speed of FPA-NCDE exceeds that of FPA, SSO, ABC, and GA, and when the number of iterations is 200, the convergence speed exceeds that of PSO.

[0111] Figure 4 and Figure 5 is the UAV flight path diagram and the corresponding top view, Figure 4 It can be seen that a path needs to be planned from the lower left corner to the upper right corner of the area, where the mountain is an obstacle encountered during the flight, and the icons of various colors represent threat targets. Figure 5 It can be seen from the top view that the SSO and PSO algorithms fluctuate violently during the optimization process, especially at the end of the optimization. FPA-NCDE has better stability, smoother curves, and less volatility during the optimization process, and its optimization effect is better than other algorithms. Figure 5 The flight path diagram shown can more intuitively show that FPA-NCDE has a better overall obstacle avoidance effect. It can avoid attacks from threatening targets without making large turns. Its flight trajectory is smoother and the overall flight effect is better.

[0112] Compared with other heuristic algorithms that search for the optimal solution, using cross-generational differential evolution to strengthen the local search for the optimal solution can guide the algorithm to continuously approach the global optimal solution and improve the accuracy of the search.

[0113] In order to verify the performance of the present invention, it is used to optimize the algorithm test function set. The test results in low dimension and high dimension are shown in Table 3 and Table 4 respectively.

[0114] Table 3 Low dimensional test results

[0115]

[0116]

[0117] Table 4 High dimensional test results

[0118]

[0119]

[0120] As can be seen from Table 3, compared with FPA, the FPA-NCDE designed by the present invention shows the best performance in 14 test functions, and its average error value has been significantly improved in 17 test functions, which is better than FPA. Compared with ABC, FPA-NCDE has better performance in 7 of the 17 test functions. The same optimal solution as ABC is obtained in the remaining 10 functions. For the average error value, FPA-NCDE is better than ABC in 11 functions, and the two are the same in 2 (F2, F5) functions. ABC is better than FPA-NCDE in the average error value of only 3 (F4, F6, F11) functions. In general, the search accuracy, average error value and stability of FPA-NCDE are better than ABC. Compared with PSO, the optimal solution obtained by FPA-NCDE in 5 function values ​​is better than PSO, and the optimal solution obtained in 12 functions is the same as PSO. For the average error value, FPA-NCDE is better in 8 functions, and the two are the same in 4 functions (F2, F3, F5 and F13). For functions F1, F4, F6, F11, F15 and F16, although PSO has a smaller average error, the difference between FPA-NCDE and PSO is small. Compared with GA, the optimization accuracy of FPA-NCDE is better than GA in 15 functions, and the optimization effects of the two are the same only in 2 functions. From the perspective of the average error value, FPA-NCDE is worse than GA only in F5, and the average error in other functions is smaller than GA. Compared with SSO, FPA-NCDE can provide better or the same optimal solution than SSO in 17 functions, and the average error is only slightly worse than GA in F8.

[0121] As can be seen from Table 4, compared with FPA, the optimal solution obtained by FPA-NCDE is better than FPA, PSO, GA, and SSO algorithms on all functions. In terms of average error, FPA-NCDE is worse than FPA only on 2 test functions, and only 1 function has an average error worse than PSO and SSO, while the average error on all functions is better than GA. Compared with ABC, FPA-NCDE has better optimal solutions on 12 test functions, 1 of which is the same, and 1 of which is better than ABC. In terms of average error, ABC is slightly better than FPA-NCDE on 3 functions. In addition, the average error of FPA-NCDE on 3 test functions is 0, and its stability is significantly better than other algorithms. Among the 14 test functions, compared with the optimization results of other algorithms, FPA-NCDE is at least 10 orders of magnitude higher on 5 functions (F19, F22, F27, F28, F32), at least 5 orders of magnitude higher on 2 functions (F29, F31), and can directly obtain the theoretical global optimal solution on 3 functions (F30, F26, F21). In summary, for low-dimensional test functions, FPA-NCDE has smaller numerical oscillations and higher stability, and greatly improves the reliability of optimization accuracy during the search process. For most high-dimensional test functions, FPA-NCDE has more significant optimization results than other optimization algorithms.

[0122] It should be appreciated that embodiments of the present invention may be implemented or enforced by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable memory. The method may be implemented in a computer program using standard programming techniques - including a non-transitory computer-readable storage medium configured with a computer program, wherein the storage medium so configured causes the computer to operate in a specific and predefined manner - according to the methods and drawings described in the specific embodiments. Each program may be implemented in an object-oriented programming language to interact with the computer system. However, if desired, the program may be implemented in an assembly or machine language. In any case, the language may be a compiled or interpreted language. In addition, the program may be run on a programmed ASIC for this purpose.

[0123] Furthermore, the operations of the processes described herein may be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by context. The processes described herein (or variations and / or combinations thereof) may be performed under the control of one or more computer systems configured with executable instructions, and may be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications) that is executed collectively on one or more processors, by hardware, or a combination thereof. The computer program includes a plurality of instructions that may be executed by one or more processors.

[0124] Further, the method can be implemented in any type of computing platform that is operably connected to a suitable computer, including but not limited to a personal computer, a minicomputer, a mainframe, a workstation, a network or distributed computing environment, a separate or integrated computer platform, etc. Aspects of the present invention can be implemented in a machine-readable code stored on a non-transitory storage medium or device, whether it is removable or integrated into a computing platform, such as a hard disk, an optical read and / or write storage medium, a RAM, a ROM, etc., so that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the process described herein. In addition, the machine-readable code, or part thereof, can be transmitted via a wired or wireless network. When such media includes instructions or programs that implement the steps described above in conjunction with a microprocessor or other data processor, the invention described herein includes these and other different types of non-transitory computer-readable storage media. When programming according to the flower pollination optimization method and technology based on nonlinear cross-generational differential evolution described in the present invention, the present invention also includes the computer itself.

[0125] The computer program can be applied to input data to perform the functions described herein, thereby converting the input data to generate output data stored in a non-volatile memory. The output information can also be applied to one or more output devices such as a display. In a preferred embodiment of the present invention, the converted data represents physical and tangible objects, including specific visual depictions of physical and tangible objects produced on the display.

[0126] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A flower pollination optimization method based on nonlinear transgenerational differential evolution, characterized in that: The method comprises the following steps: S1: nonlinear inertia weight setting; S2: Transgenerational differential evolution; S3: adaptive adjustment of parameters; S4: Cross-generational roulette jumps out of the local optimal solution; The S1 is specifically: Based on the simulated annealing algorithm SA with the advantages of global exploration and local development, the nonlinear inertia weight is designed as follows: Where T max is the maximum number of iterations, t is the current number of iterations, is the minimum fitness value of the current population, is the minimum fitness value of the previous generation population; the change of weight k can guide the global search direction and reduce the cost of blind search; based on nonlinear inertia weight, the global search iteration formula of flower pollination is: in, and are the upper and lower limits of the pollen individual value in each dimension, Refers to the average value of the upper and lower limit dimensions of each individual, and rand is a random number distributed in [0,1]; The S2 is specifically: Transgenerational differential evolution includes two mutation strategies, namely the neighborhood-based transgenerational evolution strategy NCG and the population-based transgenerational evolution strategy PCG; NCG and PCG guide the search direction to approach the optimal solution by analyzing the differences between consecutive generations, and combine the information of two generations of populations to reduce the numerical oscillation of search results; In order to improve the search stability of FPA-NCDE in the local search process, a nonlinear arccosine acceleration factor c is designed to make the search enter the local search faster at the beginning and maintain a high search speed at the end without premature convergence; its definition is as follows: Among them, d num is the dimension of pollen individuals, t is the number of iterations, T max is the maximum number of iterations; let c max =1.0,c min =0.6; As the number of algorithm iterations increases, the decay rate slows down, improving the local search capability of the algorithm; NCG calculates the Euclidean distance between the parent individual and other individuals in the population, selects the nearest T individuals to form a neighborhood pool, and selects the previous generation of T individuals as members of the neighborhood pool based on the Euclidean distance. Its expression is as follows: Among them, V i,g is a mutation vector, representing the new individual produced after differential evolution, i is the index of the parent individual, g is the number of generations of differential mutation, is a randomly selected individual from the neighborhood pool of the current generation, is an individual randomly selected from the neighborhood pool of the previous generation parent; rn1 is an individual randomly selected from {I i,g,1 ,I i,g,2 ,...I i,g,T }A random integer selected, {I i,g,1 ,I i,g,2 ,...I i,g,T } records the index of individuals in the neighborhood pool, rn2 is from {I i,g-1,1 ,I i,g-1,2 ,...I i,g-1,T }A randomly selected integer, the size of T is 5% of the total number of individuals in the population; PCG increases the richness of the population by randomly selecting individuals from the previous generation population of the parent individual and individuals from the contemporary population to participate in the mutation process. Its expression is as follows: Among them, x i,g For the parent individual, is a randomly selected individual from the current population, is an individual randomly selected from the previous generation population, rp1 and rp2 are random integers selected from {1,2,...N}, and N is the population size; in order to balance the convergence and search capabilities, the probability of using PCG and NCG in the mutation operation is the same; The S3 is specifically: During parameter initialization, the initial scaling factor F i,g and the crossover probability CR i It is from [F min ,F max ] and [CR min ,CR max ] is randomly generated, F min and F max are the upper and lower limits of the scaling factor F, CR min and CR max are the upper and lower limits of the crossover probability; The expression for dynamic update of the scaling factor is as follows: Randomly select three individuals x from the neighborhood of the current individual p1 、x p2 、x p3 , whose fitness values ​​are f1, f2, f3 and f1<f2<f3; similarly, three individuals x are randomly selected from the neighborhood pool of the previous generation of parent individuals p4 ,x p5 ,x p6 The fitness values ​​are f4, f5, f6 and f4<f5<f6; the expression of dynamic update of crossover probability is as follows: where f ave is the average fitness of all individuals in the current population; The S4 is specifically: After completing the search iterations for all individuals, select T individuals with the smallest fitness value, and then randomly select α individuals from the remaining individuals in the current population and β parent individuals from the previous generation. A total of 2T individuals form a roulette pool to participate in the roulette process; the probability of each pollen individual in the roulette pool being selected is as follows: Where i is the pollen individual, wig i is the mapping weight of the fitness value; The process of intergenerational roulette is as follows: S41: sort the pollen individuals in the parent pollen population in ascending order according to fitness values; S42: Select the top T pollen individuals to form a subpopulation P t ; S43: Randomly select β individuals from the remaining individuals in the parent population; S44: randomly select α individuals from the contemporary population; S45: The subpopulation and the selected individuals form a roulette pool W t ; S46: according to formula (11), the individuals in the roulette pool participate in the roulette gambling process; The straight line segment of the flight start and end point of the UAV intelligent inspection path planning is divided into D equal parts, and a perpendicular line is drawn through each equal-division point on the line segment. Appropriate points are selected on the perpendicular line where each equal-division point is located as track points. The trajectory route connecting these track points is the UAV flight path planning route; let the position coordinates of the obstacles and threat targets encountered by the UAV during flight in the new coordinate system be (x i ,y i ), the threat radius is r i , then the attack range of the threat target is: In the process of UAV path planning, there are three constraints, namely flight length limit, turning angle limit, and flight altitude limit; 1) Flight length limit Before each turn, there is a straight-line distance for error correction. The minimum flight length affects the trajectory of the drone. Let the starting point of the flight trajectory be S and the end point be T, and ST be segmented into D+1 segments. The length of each segment is greater than the minimum flight length limit. The maximum total length of each segment follows the following conditions: where ||P i-1 P i || is the length of the segment, P0 is the starting point, P D+1 It's the end point, L max is the maximum value of the total flight length; 2) Turning Angle Limit The turning angle of the drone between two adjacent segments follows the following conditions: where θ max is the maximum steering angle; 3) Flight altitude restrictions The flight altitude h of the drone complies with the following conditions: H min <h<H max (15) H max is the maximum flight altitude, H min is the minimum flight altitude; Based on the above constraints, energy consumption cost, steering cost, threat attack cost and flight altitude cost are considered as performance evaluation criteria of the UAV path planning algorithm; 1) Energy consumption cost Energy cost e It is related to the speed of the UAV and the total flight length, and its expression is as follows: Where η is the energy consumption factor, v is the speed of the drone, l i is the length of the ith segment of the total flight track; 2) Turning cost The steering cost is related to the steering angle. Formula (17) gives the steering cost Cost s Calculation: where θ j is the steering angle, k is the steering coefficient, θ max is the maximum steering angle; 3) Threat attack cost Consider the cost of threat attacks t , the threat level of each track point is expressed as: Where kj is the threat level of the jth threat target, a and b are threat weights, and d xj It refers to the distance between the UAV at point x and the center of the jth threat target, R j represents the radius of the threat target; in order to calculate the threat attack cost of each threat target on the drone in each segment path, the length of each segment is divided into 15 parts on average; the threat attack cost of each segment is the location The average cost of all threat attacks; then the threat attack cost Cost t for: l i is the length of the ith flight path segment, and N is the number of threat targets; 4) Flight altitude cost The flight altitude cost is related to the flight altitude of each segment. h for: Cost h =μ(h i -h best ) (20) Where μ is the flight height factor, h i is the flight height of the UAV in the ith segment, h best is the best flight altitude; The cost function of UAV inspection path planning is: Among them, λ1, λ2, λ3, and λ4 represent Cost t Cost e Cost h Cost s The weight of .

2. A flower pollination optimization implementation system based on nonlinear transgenerational differential evolution, characterized by: The system comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, and the method according to claim 1 is implemented when the processor executes the computer program.

3. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to claim 1 is implemented.

Images