Multi-unmanned aerial vehicle path planning method based on multi-population grey wolf optimization algorithm

Abstract

The invention belongs to the technical field of unmanned system intelligent control and path planning, and discloses a multi-unmanned aerial vehicle path planning method based on a multi-population grey wolf optimization algorithm, and the specific technical scheme comprises the steps: 1, building an environment model and constraint conditions of unmanned aerial vehicle three-dimensional path planning; 2, constructing a multi-target cost function including fuel consumption cost, flight height cost, threat area cost, time cooperation cost and space cooperation cost; 3, optimizing the multi-target cost function by adopting a multi-population grey wolf optimization algorithm to obtain an optimal flight path; according to the method, the global search capability is remarkably enhanced, and the multi-group clustering strategy and the periodic convergence factor act together, so that the algorithm can effectively jump out of local optimum, a path scheme closer to global optimum is found in a complex solution space, the convergence speed and the calculation precision are greatly improved, the collaborative planning effect is excellent, and the method is suitable for large-scale popularization and application. The method shows strong environmental adaptability and robustness, and has a very high practical application value.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent control and path planning of unmanned systems, and particularly relates to a method for cooperative path planning of multiple unmanned combat aircrafts in three-dimensional space based on an improved grey wolf optimization algorithm. BACKGROUND

[0002] With the wide application of unmanned aerial vehicles, especially unmanned combat aircrafts (UCAVs) in the fields of military reconnaissance, precision strike, cooperative combat, etc., the cooperative control technology of multiple unmanned systems has become a current research hotspot and frontier. Among them, the path planning technology is a core component of the multiple unmanned aerial vehicle system, and its goal is to plan an optimal or suboptimal flight path for the aircraft from the starting point to the target point under the premise of meeting various physical and task constraints. A reasonable path not only can effectively save fuel and shorten the task time, but also can significantly improve the survival probability of UCAVs in complex threat environments. Therefore, the cooperative path planning technology of multiple UCAVs is a key supporting technology for realizing the intelligence of unmanned aerial vehicle clusters and completing joint combat tasks.

[0003] At present, in the research of multiple unmanned aerial vehicle path planning at home and abroad, many scholars have proposed different solutions. For example, some studies use A* algorithm for real-time strike path planning of multiple unmanned aerial vehicles to meet the real-time requirements in complex battlefield environments; some studies combine pigeon optimization algorithm and artificial potential field method to solve the local path re-planning problem in the cooperative task environment of multiple unmanned aerial vehicles; in addition, there are multiple unmanned aerial vehicle cooperative formation and rendezvous methods based on pigeon optimization algorithm, and consistent formation and rendezvous control algorithms based on Dubins curve to solve the problem of rapid formation generation.

[0004] Among various optimization algorithms, the grey wolf optimization algorithm is introduced into the field of path planning due to its simple structure, few parameters, and easy implementation, and good results have been achieved. However, the standard GWO algorithm gradually exposes its inherent defects when dealing with high-dimensional, multi-constrained complex optimization problems. First, it uses a random way to initialize the population, which leads to the generated path points being chaotic and disorganized, poor population diversity, and low individual average fitness. Second, the convergence factor a, which controls the balance between global and local search in the algorithm, adopts a linear decreasing strategy. This simple change rule is difficult to adapt to the complex search process, which can easily lead to premature convergence and fall into local optimum, or slow convergence and insufficient solution accuracy due to improper search step in the later period.

[0005] To improve the above problems, a series of improvement strategies have been proposed in academia. For example, by hypercube sampling to initialize the population, introducing dynamic weight strategy, using reverse learning strategy to initialize individuals, improving convergence factor and introducing chaos search strategy, etc. These improvements have improved the performance of GWO algorithm to some extent. However, existing improvement researches mostly focus on single UAV path planning in two-dimensional or three-dimensional space, and the research on multi-UCAV cooperative path planning in complex three-dimensional environment is still insufficient. When GWO algorithm is directly applied to multi-UCAV three-dimensional cooperative path planning, the population diversity is further reduced, it is more likely to fall into local optimum, the convergence speed is slow, the solution accuracy is not high, and the coordination ability among the fleet is weak.

[0006] Although the grey wolf optimization algorithm and its improved algorithm have achieved certain application effect in path planning, when they are applied to the complex problem of multi-UCAV three-dimensional cooperative path planning, there are still a series of technical defects, which are analyzed as follows:

[0007] The single population initialization strategy leads to poor search starting point: the standard GWO algorithm uses a completely random method to initialize the population individuals (i.e. candidate paths) in the search space. This strategy will produce a large number of chaotic and unreasonable flight path point sequences in three-dimensional space. The direct consequence is that the overall quality of the initial population is low, and the average fitness of individuals is poor, which makes the algorithm need to spend more iteration times to evolve from these poor solutions, seriously slowing down the convergence speed and easily falling into local optimum.

[0008] The rigid convergence factor adjustment mechanism cannot balance the search ability: the key parameter a in the standard GWO algorithm that controls the global exploration and local exploitation behavior adopts the strategy of linearly decreasing from 2 to 0. This linear and monotonous change mode does not match the complex actual optimization search process. It leads to insufficient global search ability in the early stage and insufficient local search in the later stage.

[0009] Single population structure is easy to fall into local optimum: the standard GWO uses a single population for search, and all individuals share the same information and tend to the few leader wolves (a, b, d) of the current optimum. When solving complex multi-peak optimization problems, this structure is easy to cause rapid loss of population diversity, and the whole population converges to the same local optimum too early, and cannot explore other possible better areas in the solution space.

[0010] Lack of effective coordination planning mechanism, difficult to meet the multi-machine constraints: most of the existing GWO-based improved algorithm is for single-machine path planning, and its fitness function usually only considers single-machine indicators such as path length, obstacle threat and flight height. In the multi-machine system, there is a lack of time and space coordination constraints between the groups, which cannot ensure that multiple UCAVs can reach the target point at the same time or in a specific order within the specified time window. And lack of effective anti-collision mechanism, unable to guarantee that the path between any two UCAVs meets the requirement of minimum safe flight distance during planning. This makes the path planned by the existing method only a simple set of multiple single-machine paths, rather than an organic and coordinated group action plan, which cannot meet the task requirements of modern coordinated combat.

[0011] In summary, due to the above-mentioned inherent defects in population initialization, convergence factor adjustment, multi-machine coordination and population structure, the existing technology (including standard GWO and its improved versions) has the technical bottlenecks of slow convergence speed, low solution accuracy, easy to fall into local optimum, and weak coordination ability when solving the problem of three-dimensional coordinated path planning of multiple UCAVs. The present application is proposed to overcome these specific defects. SUMMARY

[0012] In order to solve the technical problems of poor initial solution quality, low population diversity, easy to fall into local optimum, imbalance between global exploration and local development caused by linear convergence factor, affecting convergence speed and solution accuracy, and inability of existing single-machine path planning methods to meet multiple constraints such as time coordination and space coordination in multi-UCAV coordination tasks caused by random initialization and single population structure, the present application provides a multi-UAV path planning method based on multi-population grey wolf optimization algorithm, which can quickly and stably plan a safe, energy-saving and coordinated-constraint-satisfied high-quality flight path for multiple UCAVs in a complex three-dimensional threat environment.

[0013] To achieve the above purpose, the technical scheme adopted by the present application is as follows: a multi-UAV path planning method based on multi-population grey wolf optimization algorithm, the specific steps are as follows:

[0014] Step one, establish the environment model and constraint conditions of UAV three-dimensional path planning;

[0015] Step two, construct a multi-objective cost function including fuel consumption cost, flight height cost, threat area cost, time coordination cost and space coordination cost;

[0016] Step three, use multi-population grey wolf optimization algorithm to optimize the multi-objective cost function and obtain the optimal flight path, which includes the following steps:

[0017] Step S31, a hybrid initialization method is introduced by combining the greedy idea and the random initialization, and an initial grey wolf population is generated by using the hybrid initialization strategy; when generating each successive path point of an individual (a path), the coordinates of the previous path point are taken as the lower limit (the first point takes the starting point as the lower limit), and the coordinates of the target point are taken as the upper limit, and each point is randomly generated according to the uniform distribution in the interval;

[0018] Step S32, the initial grey wolf population is divided into multiple subpopulations by using the minimum value clustering strategy, in each iteration, the path population is evaluated according to multiple indexes respectively, similar individuals are grouped into the same subpopulation by using the clustering algorithm, and excellent and potential individuals are selected out, each subpopulation independently performs optimization search, and the diversity of the search direction is retained. At the same time, part of the excellent individuals in each subpopulation are extracted to form a reserved population, which is used to store and share the global excellent genes. This mechanism effectively maintains the population diversity, avoids the premature convergence of all individuals, and greatly enhances the global exploration ability of the algorithm and prevents falling into local optimum.

[0019] Step S33, in each iteration period, the global search and local development ability of the nonlinear convergence factor control algorithm is adopted;

[0020] Step S34, the position of each grey wolf individual is updated through the leadership update mechanism;

[0021] Step S35, whether the termination condition is met is judged, if yes, the optimal flight path is output, and if not, the step S35 is returned to continue iteration.

[0022] The environment model established in step one includes:

[0023] The spherical obstacle model is used to simulate buildings and natural terrain;

[0024] The threat area model includes the meteorological threat area, the missile threat area and the radar threat area;

[0025] The constraint conditions in step one include:

[0026] The maximum range constraint: , wherein, is the range of the ith unmanned aerial vehicle;

[0027] The flight height constraint: , wherein, is the height of the kth path point, are the minimum height of the unmanned aerial vehicle, are the maximum height of the unmanned aerial vehicle;

[0028] The speed range constraint: , wherein, The flight speed of the ith UAV; Mach number, commonly used in the field of aviation, space, etc., represents the ratio of flight speed to local sound speed.

[0029] Angle constraints: , , The maximum yaw angle, The maximum pitch angle;

[0030] Safety distance constraints: , The distance between UAV i and j, The minimum safety distance.

[0031] In step two, to evaluate the pros and cons of multi-UCAV path generation, the multi-objective cost function is represented as:

[0032] ;

[0033] , The number of UCAVs, w1, w2, w3, w4, w5 are the weights of each cost item, and satisfy w1+ w2 + w3 + w4+ w5 = 1;

[0034] The fuel consumption cost of the ith UAV, the calculation formula is: ;

[0035] The actual range, the formula is: ;

[0036] The horizontal coordinate of the kth waypoint, The vertical coordinate of the kth waypoint, The height of the kth waypoint, The maximum allowable range;

[0037] The flight height cost of the ith UAV, the calculation formula is:

[0038] ;

[0039] The threat area cost of the ith UAV, the calculation formula is: ;

[0040] The distance between the UAV and the center of the threat area, The threat radius;

[0041] The time coordination cost of the ith UAV is calculated by the following formula:

[0042] ;

[0043] wherein, is the instruction arrival time, is the actual arrival time;

[0044] The space coordination cost of the ith UAV is calculated by the following formula: .

[0045] The hybrid initialization strategy in step S31 is specifically:

[0046] The random initialization method is used to generate the remaining individuals to maintain the population diversity, and the initialization method based on the greedy idea is used to generate part of the individuals:

[0047] ;

[0048] wherein, is the horizontal coordinate of the two adjacent waypoint coordinates in the individual; is the vertical coordinate of the two adjacent waypoint coordinates in the individual; is the horizontal and vertical coordinates of the target point, and U is a uniform distribution function;

[0049] The initial path points are roughly distributed towards the target point, forming a meaningful initial path, and the average fitness of the initial population is significantly improved.

[0050] The initialization method based on the greedy idea will reduce the diversity of the population to some extent, and the mutation strategy is added to form a hybrid initialization method through mutation threshold control. In order to avoid the decline in diversity that may be caused by the greedy idea, when the population initialization mutation parameter is less than the threshold value, the initialization of this individual uses the initialization method based on the greedy idea; otherwise, the individual uses the original random initialization method. This mechanism maintains the diversity of the population through controllable randomness while ensuring the quality of the initial solution.

[0051] The nonlinear convergence factor in step S33 is adjusted by using the cosine function, and the linearly decreasing convergence factor in the standard GWO is improved into a nonlinear periodic adjustment strategy based on the cosine function, and the specific calculation formula is:

[0052] ;

[0053] wherein, is the initial value of the convergence factor, t is the current iteration number, T is the maximum iteration number, the value of the convergence factor changes in a cosine waveform rather than monotonically decreases with the iteration process.

[0054] In the early stage of iteration, the value decreases rapidly but still maintains a high level, which is conducive to extensive global exploration; in the middle and later stages of iteration, the periodic fluctuation of the value enables the algorithm to periodically jump out of the current search area and re-enhance the exploration ability, thereby effectively avoiding falling into local optimum due to premature convergence, while also maintaining fine local development ability in the later stage.

[0055] The leader update mechanism in step S34 is specifically:

[0056] Each sub-population maintains three leader wolves, namely, α, β and δ, and updates the population position according to the population search formula:

[0057] ,

[0058] The individual position is updated by weighted average of the leader wolf positions:

[0059] .

[0060] wherein, is the weighted distance between the current wolf and , and is the weighted distance between the current wolf and . is the weighted distance between the current wolf and . respectively represent the position calculated based on the α leader wolf, respectively represent the position calculated based on the β leader wolf, respectively represent the position calculated based on the δ leader wolf, and the final updated position is

[0061] By adjusting the flight speed of each UAV, time coordination is realized to ensure that multiple UAVs reach the target position within the command arrival time; by monitoring the distance between UAVs in real time, the minimum safety distance requirement is ensured to realize space collision avoidance; based on dynamic threat information, the path is re-planned in real time to improve the system robustness.

[0062] ​Through simulation experiment verification, the technical scheme provided by the application brings the following remarkable beneficial effects: the global search ability is significantly enhanced: the multi-population clustering strategy and the periodic convergence factor jointly act, so that the algorithm can effectively jump out of the local optimum and find a path scheme closer to the global optimum in a complex solution space.

[0063] The convergence speed and the calculation precision are greatly improved: the high-quality mixed initialization and the more intelligent convergence factor adjustment mechanism enable the algorithm to more quickly approach a high-quality solution, and the comprehensive cost (fitness value) of the final solution (planned path) is obviously lower than that of the traditional algorithm.

[0064] The collaborative planning effect is excellent: based on the multi-objective collaborative fitness function, the application successfully plans a collaborative path that can meet strict space-time constraints. Experiments show that each UCAV can reach the target within the instruction time window, and always maintains a safe distance from each other during the entire flight process, without collision risk.

[0065] High stability in complex environment: even in a simulation setting with more threats and more complex environment, the path planned by the application can still be smooth and safe, showing strong environmental adaptability and robustness, and having high practical application value. BRIEF DESCRIPTION OF DRAWINGS

[0066] Figure 1 The system flowchart of the application.

[0067] Figure 2 The convergence factor simulation comparison chart, Figure 2 (a) is a curve change chart of different convergence factor values with the increase of iteration times, Figure 2 (b) is a convergence process analysis chart of the MP-GWO algorithm proposed in the application.

[0068] Figure 3 The multi-UCAV path planning simulation result chart based on the multi-population grey wolf optimization algorithm, Figure 3 (a) is a simulation result chart in a two-dimensional environment, Figure 3 (b) is a simulation result chart in a three-dimensional environment.

[0069] Figure 4 The convergence curve comparison chart of the multi-population grey wolf optimization algorithm and other algorithms, wherein the red curve is the multi-population grey wolf optimization algorithm (MP-GWO) proposed in the application, the blue curve is the standard grey wolf optimization algorithm (GWO), the green curve is the particle swarm optimization algorithm (PSO), and the purple curve is the evolutionary algorithm (EA).

[0070] Figure 5 The two-dimensional path planning chart in a complex environment, Figure 5(a) is a simulation result graph of the MP-GWO algorithm two-dimensional path planning proposed in the application, Figure 5 (b) is a simulation result graph of the GWO algorithm two-dimensional path planning, Figure 5 (c) is a simulation result graph of the PSO algorithm two-dimensional path planning, Figure 5 (d) is a simulation result graph of the EA algorithm two-dimensional path planning.

[0071] Figure 6 for three-dimensional path planning in a complex environment, Figure 6 (a) is a simulation result graph of the MP-GWO algorithm three-dimensional path planning proposed in the application, Figure 6 (b) is a simulation result graph of the GWO algorithm three-dimensional path planning, Figure 6 (c) is a simulation result graph of the PSO algorithm three-dimensional path planning, Figure 6 (d) is a simulation result graph of the EA algorithm three-dimensional path planning. DETAILED DESCRIPTION

[0072] In order to make the technical problems to be solved by the application, the technical solutions and beneficial effects clearer and more apparent, the application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the application and do not limit the application.

[0073] Example 1:

[0074] The multi-UAV path planning method based on the multi-population grey wolf optimization algorithm establishes a three-dimensional path planning environment model and sets the starting point and target point coordinates of three unmanned combat aircrafts (UCAVs). Among them, the starting point coordinates of UCAV1 are (0, 0), and the target point coordinates are (875, 875); the starting point coordinates of UCAV2 are (0, 100), and the target point coordinates are (800, 875); the starting point coordinates of UCAV3 are (300, 0), and the target point coordinates are (875, 800). Threat areas are set in the three-dimensional space, including radar threat areas and other (meteorological, missile, artillery) threat areas.

[0075] The UAV performance constraint parameters are set as follows: maximum range = 1.5 times the straight-line distance between the starting point and the target point, minimum range = 2km, minimum flight height = 20km, flight speed range [0.3Ma, 0.7Ma], maximum yaw angle = 60°, maximum pitch angle = 45°, command arrival time = 160s, safety distance = 25km.

[0076] MP-GWO algorithm parameters are set: population size 100, maximum iteration number 500, subpopulation number 5, convergence factor initial value =2. According to the combat environment, the fuel cost weight in the path planning evaluation function =0.05, height cost =0.05, threat cost =0.1, time cost =0.7, collision cost =0.1.

[0077] The MP-GWO algorithm is executed for path planning, specifically including:

[0078] An initial population is generated using a hybrid initialization strategy, combining a greedy idea and a random initialization method, wherein the selection interval of the greedy idea uses the coordinates of the previous waypoint as the lower bound and the coordinates of the target point as the upper bound;

[0079] The initial path population is analyzed by clustering using a minimum value clustering strategy, and the population is divided into multiple subpopulations; in each iteration cycle, the nonlinear convergence factor of the application is used to update the parameters;

[0080] The individual position is updated through a leadership update mechanism, and each subpopulation maintains alpha, beta, and delta leader wolves, and the individual position is updated by weighted average of the leader wolf position;

[0081] It is judged whether the maximum iteration number 500 is reached, and if so, the optimal path is output.

[0082] After 500 iterations, the optimal path planning result is obtained. As shown in the simulation results in the accompanying Figure 3 , three UCAVs can effectively avoid threat areas and reach the target destination in both two-dimensional and three-dimensional environments. As shown in Table 1, the flight time error of the path planned by each UCAV with the command time is within 1%, and no collision occurs between the UCAVs.

[0083] Table 1 UCAV simulation results

[0084]

[0085] Example Two:

[0086] Algorithm performance comparison experiment:

[0087] The superiority of the MP-GWO algorithm is verified through a comparison experiment, and the implementation is as follows:

[0088] The same environment configuration and parameter setting as in Example 1 are used;

[0089] Selecting the contrast algorithm, including the standard GWO algorithm, particle swarm optimization (PSO), evolutionary algorithm (EA);

[0090] Performing a contrast experiment, each algorithm is independently run 30 times, and is tested under the same three-dimensional cooperative path planning model;

[0091] Recording the experimental results and analyzing the results. Figure 4 The simulation results are shown in Table 2. Figure 4 From Table 2, it can be seen that the convergence speed and convergence accuracy of the MP-GWO algorithm are better than those of other algorithms, and the optimal function value can be reached under a proper number of iterations, and the path quality planned by the MP-GWO algorithm is optimal, and the optimization time is reduced by 12.4% compared with GWO, and the convergence value is improved by 17.4.

[0092] Table 2 Simulation results of different algorithms

[0093]

[0094] Example three

[0095] Path planning test in complex environment:

[0096] In this embodiment, on the basis of example 1, new threat areas are added to test the performance of the algorithm in a complex environment:

[0097] A new threat area is added to the basic environment to simulate a more realistic battlefield environment, and a complex environment path planning is performed, and the same MP-GWO algorithm parameters are used for path planning in the enhanced threat environment.

[0098] Evaluate the planning effect, as shown in the attached Figure 5 and Figure 6 From the two-dimensional and three-dimensional path planning diagrams, it can be seen that the path planned by the MP-GWO algorithm is the safest, the planned path is smoother, and the turning points are significantly less than those of the other three algorithms, the UCAV can still search for a low-cost trajectory and effectively avoid various threat areas, and good stability and adaptability are shown in a complex dynamic threat environment.

[0099] The above only describes the preferred embodiments of the present application and does not limit the present application, and any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall be included in the scope of the present application.

Claims

1. A multi-UAV path planning method based on a multi-population grey wolf optimization algorithm, characterized in that, The specific steps are as follows: Step one, establish the environment model and constraint conditions of the three-dimensional path planning of the unmanned aerial vehicle; Step two, construct a multi-objective cost function including fuel consumption cost, flight altitude cost, threat area cost, time coordination cost and space coordination cost; Step three, use the multi-population grey wolf optimization algorithm to optimize the multi-objective cost function to obtain the optimal flight path, which includes the following steps: Step S31, combine the greedy idea with random initialization, introduce a mutation strategy to form a hybrid initialization method, and use the hybrid initialization strategy to generate the initial grey wolf population; when generating each successive path point of the individual, take the coordinates of the previous path point as the lower limit and the coordinates of the target point as the upper limit, and generate them according to uniform distribution in this interval; Step S32, use the minimum value clustering strategy to divide the initial grey wolf population into multiple subgroups, in each iteration, evaluate the path population according to multiple indexes, and through clustering algorithm, similar individuals are classified into the same subgroup, and each subpopulation independently optimizes the search; Step S33, in each iteration period, the global search and local development ability of the nonlinear convergence factor control algorithm is adopted; Step S34, update the position of each grey wolf individual through the leadership update mechanism; Step S35, judge whether the termination condition is met, if yes, output the optimal flight path, if not, return to step S35 and continue iteration.

2. The multi-unmanned aerial vehicle path planning method based on the multi-population grey wolf optimization algorithm according to claim 1, characterized in that, The environment model established in step one includes: Spherical obstacle model, used to simulate buildings and natural terrain; Threat area model, including meteorological threat area, missile threat area and radar threat area; The constraint conditions in step one include: Maximum range constraint: wherein, is the range of the ith drone, is the minimum range distance of the drone, is the maximum allowed range of the drone; Flight height constraint: wherein, is the height of the kth waypoint, is the minimum height of the UAV; is the maximum height of the UAV; Speed range constraint: wherein, is the flight speed of the i-th drone; Angle constraints: , wherein, is the maximum yaw angle, is the maximum pitch angle; Safety distance constraint: wherein, is the distance between the drones i and j, is the minimum safety distance.

3. The multi-unmanned aerial vehicle path planning method based on the multi-population grey wolf optimization algorithm according to claim 2, characterized in that, The multi-objective cost function in step two is represented as: ； wherein, is the number of UCAVs, w1, w2, w3, w4, w5 are weight coefficients, and satisfy w1+ w2 + w3 + w4 + w5 =1; The fuel consumption cost for the i-th UAV is calculated by the formula: ; For actual distance, the formula is: ; is the x-coordinate of the kth waypoint, is the y-coordinate of the kth waypoint, is the altitude of the kth waypoint, is the maximum allowed range; The flight height cost of the ith UAV is calculated by the following formula: ； The threat area cost of the i-th UAV is calculated by the following formula: ; is the distance of the drone from the center of the threat area, is the threat radius; The time coordination cost of the ith UAV is calculated by the following formula: ； wherein, is the instruction arrival time, is the actual arrival time; The spatial coordination cost of the ith UAV is calculated by the following formula: .

4. The multi-unmanned aerial vehicle path planning method based on the multi-population grey wolf optimization algorithm according to claim 3, characterized in that, The hybrid initialization strategy in step S31 is as follows: The rest of the individuals are generated by using a random initialization method, and part of the individuals are generated by using an initialization method based on the greedy idea: ; wherein, is the x-coordinate of two adjacent waypoint coordinates in the individual; is the y-coordinate of two adjacent waypoint coordinates in the individual; is the x and y coordinates of the target point, and U is a uniform distribution function. Join the mutation strategy, through the mutation threshold control to form a hybrid initialization method, that is, when the population initialization mutation parameters less than the threshold, the individual initialization uses the initialization method based on the greedy idea; otherwise, the individual uses the original random initialization method.

5. The multi-unmanned aerial vehicle path planning method based on the multi-population grey wolf optimization algorithm according to claim 4, characterized in that, The nonlinear convergence factor in step S33 is adjusted by using the cosine function: ； wherein is the initial value of the convergence factor, t is the current iteration number, and T is the maximum number of iterations.

6. The multi-unmanned aerial vehicle path planning method based on the multi-population grey wolf optimization algorithm according to claim 5, characterized in that, The leadership update mechanism in step S34 is as follows: Each subpopulation maintains three leader wolves, alpha, beta and delta, and updates the population position according to the population search formula: ， ； The individual position is updated by weighted average of the leader wolf position: .； wherein, current wolf respectively, a weighted distance between, respectively represent the position calculated based on the alpha, beta, delta leaders wolf, is the final updated position.

Images