Multi-unmanned aerial vehicle cooperative flight path planning method fusing co-evolution

Abstract

The invention discloses a multi-unmanned aerial vehicle cooperative flight path planning method fusing coevolution, and relates to the technical field of multi-unmanned aerial vehicle cooperative flight path planning. According to the method, a traditional genetic algorithm is improved, an effective and accurate fitness function is given in the improved genetic algorithm, compared with the traditional genetic algorithm, the improved genetic algorithm better avoids the problem of local optimum and is high in convergence speed, a coevolution mechanism is combined with the improved genetic algorithm (GA), and by means of the coevolution mechanism, the fitness function of the improved genetic algorithm is improved. Cooperation among the populations is fully considered, collision among multiple unmanned aerial vehicles is avoided, and each unmanned aerial vehicle can obtain an optimal or nearly optimal collision-free path.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-UAV collaborative trajectory planning, and in particular to a multi-UAV collaborative trajectory planning method integrating collaborative evolution. Background Art

[0002] A drone is an unmanned aircraft with no manned behavior during flight. The flight control of a drone needs to rely on computers, the Internet, and control terminals. A large number of running programs are involved during flight control. The flight control of a drone can be divided into three modules: the main body, the terminal system, and the load. At present, drones are mainly used in military and civilian applications. The flight control capability of drones is strong, and they can complete designated tasks in an efficient state, and they are competent for difficult tasks. Flight control enables drones to operate in a variety of environments. There will be no casualties during drone operations, which is an advantage of drone flight control. Multi-drone collaborative trajectory planning is a key issue in the intelligent decision-making and mission planning of unmanned systems. Collaborative trajectory planning refers to the optimal or near-optimal collision-free path of multiple drones from the starting position to the target position in an environment with obstacles. Most of the existing collaborative trajectory planning uses genetic algorithms (GAs) to obtain the optimal or near-optimal collision-free path. Genetic algorithms (GAs) are a powerful optimization problem because they can find the global optimum and have high parallelism. However, traditional genetic algorithms have shortcomings such as slow convergence, local optimality and neglect of cooperation between populations, which will make collaborative trajectory planning take a long time to converge to the optimal solution, especially in large-scale or complex trajectory planning problems. This may lead to the failure to meet real-time requirements, or fall into the local optimal solution, resulting in the inability to find the global optimal solution and the inability to find the most efficient route. Summary of the invention

[0003] In view of the shortcomings of the prior art, the present invention proposes a multi-UAV collaborative trajectory planning method integrating collaborative evolution to solve the problem that the use of genetic algorithms for multi-UAV collaborative trajectory planning is prone to fall into local optimality, resulting in poor UAV collaborative trajectory planning effect.

[0004] A multi-UAV collaborative trajectory planning method integrating collaborative evolution includes the following steps:

[0005] Step 1: Initialize several subpopulations according to the number of drones, and then obtain an initial population including several subpopulations;

[0006] Specific: According to the number of drones n robot , construct n robot subpopulations, each subpopulation has a size of μ, and the size is M = μ*n robotThe initial population of , where each subpopulation corresponds to a drone, each individual in the subpopulation represents a path of the corresponding drone, and the gene represents a node in the path;

[0007] Step 2: Set the maximum number of evolutions for each subpopulation;

[0008] Step 3: Establish individual fitness function;

[0009] The fitness function of the individual, that is, the fitness function of the path p of the drone is designed as:

[0010]

[0011] Among them, F(p) is the fitness function of the individual, C is a constant, and f f (p) is a feasible path evaluation function, where the feasible path is a path that does not collide with obstacles, and f c (p) is the infeasible path evaluation function, and the infeasible path is the path that collides with obstacles;

[0012] Assume that path p includes N nodes and N-1 line segments, then the feasible path evaluation function f of path p is f (p) is defined as:

[0013]

[0014] Where d(p) is the length of path p, s(p) is the safety factor of path p, h(p) is the smoothness of path p; w d 、w s and w k are the length, safety factor and smoothness weight of path p respectively;

[0015] The length of path p is:

[0016]

[0017] In the formula, (x i ,y i ) is the coordinate of the i-th node in path p, i is the node number;

[0018] The safety factor s(p) of path p is:

[0019]

[0020] Among them, l v is the vth line segment in path p, v is the line segment number, α(l v ) is the vth line segment l v The shortest distance to obstacles;

[0021] The smoothness h(p) of path p is:

[0022]

[0023] Among them, β(l v ,l v+1 ) represents line segment l v and line segment l v+1 The angle between

[0024] The infeasible path evaluation function f of path p c (p) is:

[0025]

[0026] Where l(p) is the proportion of infeasible line segments in path p, c(p) is the proportion of the length of path p that crosses obstacles; w l and w c represents the weights of l(p) and c(p);

[0027] The proportion l(p) of infeasible segments in path p is:

[0028]

[0029] Among them, L Infesible is the number of infeasible line segments in path p, where the infeasible line segments are line segments that collide with obstacles; L Total is the total number of line segments in path p;

[0030] The length ratio c(p) of path p that crosses obstacles is:

[0031]

[0032] In the formula, o(l v ) is the length of the vth line segment in path p that passes through the obstacle;

[0033] Step 4: Based on the established individual fitness function, the improved genetic algorithm is used to evolve each subpopulation separately to obtain the optimal path for each UAV, and then the result of multi-UAV trajectory planning is obtained;

[0034] Step 4.1: Determine the feasibility of the path represented by each individual in each subpopulation, and calculate the fitness function value of the individual according to the feasibility; the feasibility includes a feasible path and an infeasible path;

[0035] Step 4.2: Based on the fitness function value of the individual, first use the elite selection method to select the individual with the largest fitness function value in each subpopulation as the elite individual of the subpopulation;

[0036] Step 4.3: Use the roulette wheel selection method to select the individuals in each subpopulation except the elite individuals, and select 2 individuals from each subpopulation;

[0037] Step 4.4: For the two individuals selected from each subpopulation, use the two-point crossover method to perform a crossover operation to obtain offspring individuals, sort the nodes in the path represented by the offspring individuals in order, and if there are identical nodes in the path, delete the individual corresponding to the path and randomly initialize an individual as the offspring individual;

[0038] Step 4.5: Perform mutation operation on the offspring individuals obtained from each subpopulation, sort the nodes in the path represented by the mutated offspring individuals in order, and if there are identical nodes in the path, delete the offspring individuals corresponding to the path and randomly initialize an individual;

[0039] Step 4.6: Repeat steps 4.1 to 4.5 until n is obtained. robot A new subpopulation consisting of μ offspring individuals;

[0040] Step 4.7: For each current subpopulation, use the elite individuals selected from the previous generation of subpopulations to replace the individual with the smallest fitness function value in the current subpopulation;

[0041] Step 4.8: For a subpopulation X, select a subpopulation Y according to the island model, X≠Y;

[0042] Step 4.9: Select λ representative individuals from subpopulation X and subpopulation Y respectively; the representative individuals are the λ individuals with the largest fitness function value;

[0043] Step 4.10: Calculate the synergy function value between each representative individual in subpopulation X and each representative individual in subpopulation Y;

[0044] Step 4.10.1: Randomly select one representative individual from the representative individual of subpopulation X and one representative individual from subpopulation Y, and determine whether the paths represented by the two representative individuals intersect. If they do, the synergy function value between the two selected representative individuals is 0 and execute 4.11, otherwise execute 4.10.2;

[0045] Step 4.10.2: Determine the intersection of the paths represented by the two representative individuals and the number of intersections num cr ;

[0046] Step 4.10.3: Determine whether the paths represented by the two representative individuals conflict, and determine the collision coefficient of the paths represented by the two representative individuals according to whether they conflict;

[0047] The method for determining whether paths conflict is as follows:

[0048] Calculate the sum of the lengths of the line segments from the starting point of the kth path to the jth intersection point:

[0049]

[0050] Where k is the number of the path and k∈{1,2}, is the sum of the lengths of the line segments from the starting point of the kth path to the jth intersection point, dis(p j -p s ) is the sum of the lengths of the line segments from the starting point of the kth path to the jth intersection point, p s is the starting point, p j is the jth intersection, j is the number of the intersection and j=1,2……,num cr ;

[0051] If any of the paths represented by the two representative individuals exists equal The paths represented by the two representative individuals conflict with each other. equal Then the paths represented by the two representative individuals do not conflict;

[0052] The collision coefficient is:

[0053]

[0054] Among them, w b is the collision coefficient;

[0055] Step 4.10.4: Calculate the individual cooperative function value based on the number of intersections and the collision coefficient;

[0056] cf(p)=w b / num cr

[0057] Step 4.11: Traverse the λ representative individuals in subpopulation X, calculate the synergy function value of each individual in subpopulation X, and find the individual with the smallest synergy function value in subpopulation X as the elite individual;

[0058] Step 4.12: Repeat steps 4.8 to 4.11 to obtain the elite individuals of each subpopulation, and increase the current evolution times by 1;

[0059] Step 4.13: Determine whether the current number of iterations has reached the maximum number of iterations. If so, output the elite individuals of each subpopulation at present, and the paths represented by all elite individuals are used as the result of multi-UAV trajectory planning. Otherwise, execute step 4.14;

[0060] Step 4.14: Use the elite individuals of each subpopulation obtained in step 4.11 to replace the individual with the smallest fitness function value in the initial subpopulation as the current subpopulation and return to step 4.1.

[0061] Compared with the prior art, the beneficial effects of the present invention are:

[0062] The present invention improves the traditional genetic algorithm and provides an effective and accurate fitness function in the improved genetic algorithm. Compared with the traditional genetic algorithm, the improved genetic algorithm better avoids the local optimal problem and accelerates the convergence speed. The co-evolution mechanism is combined with the improved genetic algorithm (GA). The co-evolution mechanism is utilized to fully consider the cooperation between populations and avoid collisions between multiple UAVs, which is conducive to each UAV obtaining an optimal or near-optimal collision-free path. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 A flowchart of a multi-UAV collaborative trajectory planning method integrating collaborative evolution provided by an embodiment of the present invention;

[0064] Figure 2 A simulation diagram of the trajectory planning of the improved genetic algorithm provided by an embodiment of the present invention;

[0065] Figure 3 A CIGA simulation diagram of the global trajectory planning of multiple UAVs provided in an embodiment of the present invention;

[0066] Figure 4 The path evolution process of three drones provided in the embodiments of the present invention. DETAILED DESCRIPTION

[0067] The present invention is described in detail below in conjunction with the accompanying drawings and embodiments. The following embodiments are used to illustrate the present invention but are not used to limit the scope of the present invention.

[0068] Coevolution is the process of mutual adaptation between two or more populations, which reflects the fact that all species co-evolve simultaneously in a given physical environment. Therefore, the coevolution mechanism is a potential way to improve the shortcomings of traditional genetic algorithms.

[0069] A multi-UAV collaborative trajectory planning method integrating co-evolution, such as Figure 1 As shown, the following steps are included:

[0070] Step 1: Initialize several subpopulations according to the number of drones, and then obtain an initial population including several subpopulations;

[0071] Subpopulation initialization includes problem decomposition and parameter selection. The multi-UAV trajectory planning problem can be decomposed into the sub-problem of each UAV finding the optimal collision-free path. Then, each UAV is evolved separately using the improved genetic algorithm;

[0072] Specific: According to the number of drones n robot , construct n robot subpopulations, each subpopulation has a size of μ, and the size is M = μ*n robot The initial population of , where each subpopulation corresponds to a drone, each individual in the subpopulation represents a path of the corresponding drone, and the gene represents a node in the path;

[0073] In this embodiment, the number of drones n robot =3, the size of each initial subpopulation is μ=10, and the size of the initial population is M=μ*n robot =3*10; fixed-length integer coding is used for initialization, which is stable and globally convergent, and is more convenient for large working spaces.

[0074] Step 2: Set the maximum number of evolutions for each subpopulation;

[0075] In this implementation, the number of evolutions of each subpopulation is unified to 80;

[0076] Step 3: Establish individual fitness function;

[0077] The fitness function of the individual, that is, the fitness function of the path p of the drone is designed as:

[0078]

[0079] Among them, F(p) is the fitness function of the individual, C is a constant, and f f (p) is a feasible path evaluation function, where the feasible path is a path that does not collide with obstacles, and f c (p) is the infeasible path evaluation function, and the infeasible path is the path that collides with obstacles;

[0080] Assume that path p includes N nodes and N-1 line segments. The feasible path needs to consider three issues, namely path length, safety and stability. Then the feasible path evaluation function f of path p is f (p) is defined as:

[0081]

[0082] Where d(p) is the length of path p, s(p) is the safety factor of path p, h(p) is the smoothness of path p; w d 、w s and wk are the length, safety factor and smoothness weight of path p respectively;

[0083] The length of path p is:

[0084]

[0085] In the formula, (x i ,y i ) is the coordinate of the i-th node in path p, i is the node number;

[0086] The safety factor s(p) of path p is:

[0087]

[0088] Among them, l v is the vth line segment in path p, v is the line segment number, α(l v ) is the vth line segment l v The shortest distance to obstacles;

[0089] The smoothness h(p) of path p is:

[0090]

[0091] Among them, β(l v ,l v+1 ) represents line segment l v and line segment l v+1 The angle between

[0092] The infeasible path needs to consider three issues, namely the path length, the proportion of infeasible line segments and the proportion of length passing through obstacles. The infeasible path evaluation function f of path p is c (p) is:

[0093]

[0094] Where l(p) is the proportion of infeasible line segments in path p, c(p) is the proportion of the length of path p that crosses obstacles; w l and w c represents the weights of l(p) and c(p);

[0095] The proportion l(p) of infeasible segments in path p is:

[0096]

[0097] Among them, L Infesible is the number of infeasible line segments in path p, where the infeasible line segments are line segments that collide with obstacles; L Total is the total number of line segments in path p;

[0098] The length ratio c(p) of path p that crosses obstacles is:

[0099]

[0100] In the formula, o(l v ) is the length of the vth line segment in path p that passes through the obstacle;

[0101] Step 4: Based on the established individual fitness function, the improved genetic algorithm is used to evolve each subpopulation separately to obtain the optimal path for each UAV, and then the result of multi-UAV trajectory planning is obtained;

[0102] Step 4.1: Determine the feasibility of the path represented by each individual in each subpopulation, and calculate the fitness function value of the individual according to the feasibility; the feasibility includes a feasible path and an infeasible path;

[0103] Step 4.2: Based on the fitness function value of the individual, first use the elite selection method to select the individual with the largest fitness function value in each subpopulation as the elite individual of the subpopulation;

[0104] Elite selection is to prevent the optimal solution from disappearing in the next generation. The idea of ​​elite selection is to retain some of the best performing individuals and use them to replace the same number of the worst individuals in the next generation. That is, the offspring will evolve to be at least as good as their parents.

[0105] Step 4.3: Use the roulette wheel selection method to select the individuals in each subpopulation except the elite individuals, and select 2 individuals from each subpopulation;

[0106] The basic idea of ​​roulette wheel selection is that the probability of selecting an individual is proportional to the value of the individual’s fitness function;

[0107] Step 4.4: For the two individuals selected from each subpopulation, use the two-point crossover method to perform a crossover operation to obtain offspring individuals, sort the nodes in the path represented by the offspring individuals in order, and if there are identical nodes in the path, delete the individual corresponding to the path and randomly initialize an individual as the offspring individual;

[0108] The idea of ​​two-point crossover is to randomly select two crossover sites and then exchange the genes between the two sites in the two parental paths; for example, path 1 is the sequence (0,22,41,52,74,85,99) and path 2 is the sequence (0,41,44,75,78,83,99);

[0109] Step 4.5: Perform mutation operation on the offspring individuals obtained from each subpopulation, sort the nodes in the path represented by the mutated offspring individuals in order, and if there are identical nodes in the path, delete the offspring individuals corresponding to the path and randomly initialize an individual;

[0110] The idea of ​​the mutation operator is that, with a certain probability, in a way specific to each gene of each individual, the gene is replaced by a new gene that is not contained in the individual;

[0111] Step 4.6: Repeat steps 4.1 to 4.5 until n is obtained. robot A new subpopulation consisting of μ offspring individuals;

[0112] Step 4.7: For each current subpopulation, use the elite individuals selected from the previous generation of subpopulations to replace the individual with the smallest fitness function value in the current subpopulation;

[0113] Step 4.8: For a subpopulation X, select a subpopulation Y (X≠Y) based on the island model;

[0114] Step 4.9: Select λ representative individuals from subpopulation X and subpopulation Y respectively; the representative individuals are the λ individuals with the largest fitness function value;

[0115] Select λ representative individuals (x1, x2, ..., x λ ), (y1,y2,...,y λ ) from subpopulation Y;

[0116] Step 4.10: Calculate the synergy function value between each representative individual in subpopulation X and each representative individual in subpopulation Y;

[0117] The coordination function cf(p) takes into account the cross-interference of paths and the cooperation between UAVs for information exchange;

[0118] Step 4.10.1: Randomly select one representative individual from the representative individual of subpopulation X and one representative individual from subpopulation Y, and determine whether the paths represented by the two representative individuals intersect. If they do, the synergy function value between the two selected representative individuals is 0 and execute 4.11, otherwise execute 4.10.2;

[0119] Step 4.10.2: Determine the intersection of the paths represented by the two representative individuals and the number of intersections num cr ;

[0120] Step 4.10.3: Determine whether the paths represented by the two representative individuals conflict, and determine the collision coefficient of the paths represented by the two representative individuals according to whether they conflict;

[0121] The drone moves at a fixed speed, so the distance (sum of line segment lengths) from the initial point to each intersection can be calculated to estimate whether the two paths conflict;

[0122] The method for determining whether paths conflict is as follows:

[0123] Calculate the sum of the lengths of the line segments from the starting point of the kth path to the jth intersection point:

[0124]

[0125] Where k is the number of the path and k∈{1,2}, is the sum of the lengths of the line segments from the starting point of the kth path to the jth intersection point, dis(p j -p s ) is the sum of the lengths of the line segments from the starting point of the kth path to the jth intersection point, p s is the starting point, p j is the jth intersection, j is the number of the intersection and j=1,2……,num cr ;

[0126] If any of the paths represented by the two representative individuals exists equal The paths represented by the two representative individuals conflict with each other. equal Then the paths represented by the two representative individuals do not conflict;

[0127] The collision coefficient is:

[0128]

[0129] Among them, w b is the collision coefficient;

[0130] Step 4.10.4: Calculate the individual cooperative function value based on the number of intersections and the collision coefficient;

[0131] cf(p)=w b / num cr ;

[0132] Step 4.11: Traverse the λ representative individuals in subpopulation X, calculate the synergy function value of each individual in subpopulation X, and find the individual with the smallest synergy function value in subpopulation X as the elite individual;

[0133] Step 4.12: Repeat steps 4.8 to 4.11 to obtain the elite individuals of each subpopulation, and increase the current evolution times by 1;

[0134] Step 4.13: Determine whether the current number of iterations has reached the maximum number of iterations. If so, output the elite individuals of each subpopulation at present, and the paths represented by all elite individuals are used as the result of multi-UAV trajectory planning. Otherwise, execute step 4.14;

[0135] Step 4.14: Use the elite individuals of each subpopulation obtained in step 4.11 to replace the individual with the smallest fitness function value in the initial subpopulation as the current subpopulation and return to step 4.1;

[0136] The battlefield environment in this embodiment is represented by a two-dimensional grid; there are K threat sources in the environment, represented by circular areas, and the existence of dynamic threat sources is not considered; all drone information, target information and threat source information are known; drones are considered to be point-sized, with a fixed speed, occupying a grid. In this embodiment, a 20×20 grid is used as the battlefield; in this method, fixed-length integer coding is used, that is, each grid is numbered with a real number, and the route points are represented by a set of codes, because it is stable, has global convergence, and is more convenient for large workspaces.

[0137] The obstacles in the trajectory planning process of a single UAV using the improved genetic algorithm are set as three circular areas with a radius of 2 and coordinates of (4, 5), (11.5, 7), and (9.8, 14), respectively. Figure 2 As shown;

[0138] The obstacles in the global trajectory planning process of multi-UAVs using the cooperative evolution improved genetic algorithm are set as four circular areas with different radii, and their coordinates are (3, 16), (9.5, 3.5), (10, 11.5), and (17.3, 12), respectively. Figure 3 The path evolution process of the UAV is shown in Figure 4 shown.

[0139] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them. Although the present invention has been described in detail with reference to the above embodiments, a person skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some or all of the technical features thereof can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope defined by the claims of the present invention.

Claims

1. A multi-UAV collaborative trajectory planning method integrating collaborative evolution, characterized in that: The following steps are involved: Step 1: Initialize several subpopulations according to the number of drones, and then obtain an initial population including several subpopulations; Step 2: Set the maximum number of evolutions for each subpopulation; Step 3: Establish individual fitness function; Step 4: Based on the established individual fitness function, use the improved genetic algorithm to evolve each subpopulation separately to obtain the optimal path for each UAV, and then obtain the result of multi-UAV trajectory planning.

2. According to the method of claim 1, the method is characterized in that: The step 1 is specifically as follows: according to the number of drones n robot , construct n robot subpopulations, each subpopulation has a size of μ, and the size is M = μ*n robot The initial population is , where each subpopulation corresponds to a UAV, each individual in the subpopulation represents a path of the corresponding UAV, and the gene represents a node in the path.

3. The method for multi-UAV collaborative trajectory planning based on fusion collaborative evolution according to claim 1 is characterized in that: The fitness function of the individual, that is, the fitness function of the path p of the drone is designed as: Among them, F(p) is the fitness function of the individual, C is a constant, and f f (p) is a feasible path evaluation function, where the feasible path is a path that does not collide with obstacles, and f c (p) is the infeasible path evaluation function, and the infeasible path is the path that collides with the obstacle.

4. The method for multi-UAV collaborative trajectory planning based on fusion collaborative evolution according to claim 3 is characterized in that: Assuming that the path p includes N nodes and N-1 line segments, the feasible path evaluation function is: Where d(p) is the length of path p, s(p) is the safety factor of path p, h(p) is the smoothness of path p; w d 、w s and w k are the length, safety factor and smoothness weight of path p respectively; The length of path p is: In the formula, (x i ,y i ) is the coordinate of the i-th node in path p, i is the node number; The safety factor s(p) of path p is: Among them, l v is the vth line segment in path p, v is the line segment number, α(l v ) is the vth line segment l v The shortest distance to obstacles; The smoothness h(p) of path p is: Among them, β(l v ,l v+1 ) represents line segment l v and line segment l v+1 The angle between.

5. The method for multi-UAV collaborative trajectory planning based on fusion collaborative evolution according to claim 4 is characterized in that: The infeasible path evaluation function f c (p) is: Where l(p) is the proportion of infeasible line segments in path p, c(p) is the proportion of the length of path p that crosses obstacles; w l and w c represents the weights of l(p) and c(p); The proportion of infeasible line segments in path p, l(p), is: Among them, L Infesible is the number of infeasible line segments in path p, where the infeasible line segments are line segments that collide with obstacles; L Total is the total number of line segments in path p; The length ratio c(p) of path p that crosses obstacles is: In the formula, o(l v ) is the length of the vth line segment in path p that passes through the obstacle.

6. The method for multi-UAV collaborative trajectory planning based on fusion collaborative evolution according to claim 1 is characterized in that: The step 4 specifically includes: Step 4.1: Determine the feasibility of the path represented by each individual in each subpopulation, and calculate the fitness function value of the individual according to the feasibility; the feasibility includes a feasible path and an infeasible path; Step 4.2: Based on the fitness function value of the individual, first use the elite selection method to select the individual with the largest fitness function value in each subpopulation as the elite individual of the subpopulation; Step 4.3: Use the roulette wheel selection method to select the individuals in each subpopulation except the elite individuals, and select 2 individuals from each subpopulation; Step 4.4: For the two individuals selected from each subpopulation, use the two-point crossover method to perform a crossover operation to obtain offspring individuals, sort the nodes in the path represented by the offspring individuals in order, and if there are identical nodes in the path, delete the individual corresponding to the path and randomly initialize an individual as the offspring individual; Step 4.5: Perform mutation operation on the offspring individuals obtained from each subpopulation, sort the nodes in the path represented by the mutated offspring individuals in order, and if there are identical nodes in the path, delete the offspring individuals corresponding to the path and randomly initialize an individual; Step 4.6: Repeat steps 4.1 to 4.5 until n is obtained. robot A new subpopulation consisting of μ offspring individuals; Step 4.7: For each current subpopulation, use the elite individuals selected from the previous generation of subpopulations to replace the individual with the smallest fitness function value in the current subpopulation; Step 4.8: For a subpopulation X, select a subpopulation Y according to the island model, X≠Y; Step 4.9: Select λ representative individuals from subpopulation X and subpopulation Y respectively; the representative individuals are the λ individuals with the largest fitness function value; Step 4.10: Calculate the synergy function value between each representative individual in subpopulation X and each representative individual in subpopulation Y; Step 4.11: Traverse the λ representative individuals in subpopulation X, calculate the synergy function value of each individual in subpopulation X, and find the individual with the smallest synergy function value in subpopulation X as the elite individual; Step 4.12: Repeat steps 4.8 to 4.11 to obtain the elite individuals of each subpopulation, and increase the current evolution times by 1; Step 4.13: Determine whether the current number of iterations has reached the maximum number of iterations. If so, output the elite individuals of each subpopulation at present, and the paths represented by all elite individuals are used as the result of multi-UAV trajectory planning. Otherwise, execute step 4.14; Step 4.14: Use the elite individuals of each subpopulation obtained in step 4.11 to replace the individual with the smallest fitness function value in the initial subpopulation as the current subpopulation and return to step 4.

1.

7. The method for multi-UAV collaborative trajectory planning based on fusion collaborative evolution according to claim 6 is characterized in that: The step 4.10 specifically includes: Step 4.10.1: Randomly select one representative individual from the representative individual of subpopulation X and one representative individual from subpopulation Y, and determine whether the paths represented by the two representative individuals intersect. If they do, the synergy function value between the two selected representative individuals is 0 and execute 4.11, otherwise execute 4.10.2; Step 4.10.2: Determine the intersection of the paths represented by the two representative individuals and the number of intersections num cr ; Step 4.10.3: Determine whether the paths represented by the two representative individuals conflict, and determine the collision coefficient of the paths represented by the two representative individuals according to whether they conflict; The method for determining whether paths conflict is as follows: Calculate the sum of the lengths of the line segments from the starting point of the kth path to the jth intersection point: Where k is the number of the path and k∈{1,2}, is the sum of the lengths of the line segments from the starting point of the kth path to the jth intersection point, dis(p j -p s ) is the sum of the lengths of the line segments from the starting point of the kth path to the jth intersection point, p s is the starting point, p j is the jth intersection, j is the number of the intersection and j=1,2……,num cr ; If any of the paths represented by the two representative individuals exists equal The paths represented by the two representative individuals conflict with each other. equal Then the paths represented by the two representative individuals do not conflict; The collision coefficient is: Among them, w b is the collision coefficient; Step 4.10.4: Calculate the individual cooperative function value based on the number of intersections and the collision coefficient; cf(p)=w b / num cr 。

Images