Unmanned aerial vehicle three-dimensional path planning method based on improved grey wolf algorithm

Abstract

The invention provides an unmanned aerial vehicle three-dimensional path planning method based on an improved grey wolf algorithm, and the method comprises the steps: firstly, carrying out the environment modeling of a three-dimensional threat region, and stipulating a total cost function of the flight of an unmanned aerial vehicle for a constraint condition; secondly, an opposite learning strategy is added in grey wolf initialization, the search range of the understanding space is increased, and the individual position is updated by improving a self-adaptive weight factor, so that the convergence speed is increased; and finally, in order to avoid falling into local optimum, a particle swarm algorithm is introduced to balance global development and local convergence. A simulation experiment result shows that compared with a traditional grey wolf path planning algorithm, the improved grey wolf algorithm can find a safe and feasible path, and has a relatively stable optimization capability.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional path planning for unmanned aerial vehicles, and in particular to a three-dimensional path planning method for unmanned aerial vehicles based on an improved grey wolf algorithm. Background Art

[0002] UAV path planning is a key area in the development of modern UAV technology, such as disaster relief, logistics distribution and agricultural monitoring. With the widespread application of UAVs, how to optimize the path from the starting point to the target point to improve efficiency and safety is becoming more and more important. Common path planning algorithms are divided into classical algorithms and intelligent bionic algorithms. Classical algorithms include A* algorithm, artificial potential field method and fast search random tree algorithm, and intelligent optimization algorithms include gray wolf algorithm, genetic algorithm, particle swarm algorithm, etc. Among them, the gray wolf algorithm is based on the hierarchy of wolves in the population. Compared with other heuristic algorithms, it has simple structure, easy implementation and strong global search ability in solving performance, and is easy to combine with other heuristic algorithms. Therefore, the gray wolf algorithm has been widely used in the field of path planning. However, the gray wolf algorithm also contains some disadvantages, such as slow convergence speed and easy to fall into local optimal solution. Summary of the invention

[0003] Purpose of the invention: The purpose of the present invention is to first add an adversarial learning strategy to the gray wolf initialization, increase the search range of the understanding space, and increase the diversity of the population; secondly, by improving the adaptive weight factor to update the individual position, thereby accelerating the convergence speed; finally, in order to fall into the local optimum, the particle search strategy in the particle swarm algorithm is introduced to balance the global development and local convergence. Then, a three-dimensional path planning method for UAVs based on the improved gray wolf algorithm is provided to ensure that the UAV obtains a more stable optimal path in a shorter time.

[0004] Technical solution: The present invention provides a three-dimensional path planning method for a UAV based on an improved gray wolf algorithm, which specifically includes the following steps:

[0005] (1): Use the free space method to model the map environment and set the positions of the starting point and the target point;

[0006] (2): Initialize and set the parameters of the gray wolf path planning method;

[0007] (3): Based on the current position of the UAV and the allowed flight area, an adversarial learning strategy is used to randomly generate the initial population in the search space.

[0008] (4): Calculate the fitness value of each individual and determine the fitness value based on the first three definitions in the sorting results;

[0009] (5): Calculate the coefficient vectors A, C and update the new individual position according to the adaptive weight factor;

[0010] (6): The particle search strategy calculation in the PSO algorithm is introduced to expand the search position area, and finally enter the optimal update and replacement stage to determine the position of the individual and recalculate the new fitness value;

[0011] (7): When the maximum number of iterations is reached, the iteration stops and the three-dimensional optimal path of the UAV is obtained.

[0012] Furthermore, the implementation process of step (1) is as follows:

[0013] The three-dimensional coordinates of the drone are expressed as (X, Y, H). The horizontal coordinate X and the vertical coordinate Y are taken in the order of 0~1000. The starting point and the target point are set to (0, 0, 0) and (1000, 1000, 1000) respectively. The units of the horizontal coordinate X and the vertical coordinate Y are meters, and the value range is 0~1000 meters. The unit of the height H is meters, and the range is 0~1000 meters. Set the upper and lower limits of the gray wolf's movement to form a three-dimensional area where the drone is allowed to move. In this area, it can move left and right, up and down to get new points.

[0014] Furthermore, the implementation process of step (2) is as follows:

[0015] In the process of initializing the parameters of the improved gray wolf path planning method, the corresponding parameters include the population size , maximum number of iterations , inertia weight , the cost coefficient of the objective function and dimensions , cognitive learning factors and social learning factors .

[0016] Furthermore, the implementation process of step (3) is as follows:

[0017] First, the adversarial learning strategy is integrated into the gray wolf algorithm through population initialization. is randomly generated in the search space, where It is the dimension of problem solving. is the population size. Then, this population is used to reproduce the next generation. The opposite position of the random search of adversarial learning (OBL) is generated by the following formula:

[0018] (1)

[0019] in, and They are The upper and lower bounds of the dimension; is the original value, is the opposite value, then, the initial population and opposing populations Combine and choose ,from The solution within the interval is selected as the optimal solution.

[0020] Furthermore, the implementation process of step (4) is as follows:

[0021] The fitness value of each individual is calculated, and the first three fitness values ​​are defined as .

[0022] Furthermore, the implementation process of step (5) is as follows:

[0023] In the gray wolf optimization algorithm, the gray wolf population is divided into and Four categories, The wolf's position is defined as the optimal solution. The wolf's position is defined as the suboptimal solution. The wolf's position is the third best solution, and the other candidate solutions are the remaining Wolf. When gray wolves hunt in groups, their main behaviors include surrounding, hunting, and attacking prey. The behavior of gray wolf populations around prey can be described by the following formula:

[0024] (2)

[0025] (3)

[0026] in, Indicates the distance between the individual and the target; is the current iteration number; It's the prey The position coordinates of the iteration can be expressed as ; is the position vector after the iteration A wolf can be represented as , =1,2… ; and is a coefficient vector, expressed as follows:

[0027] (4)

[0028] (5)

[0029] (6)

[0030] in, and is a random number in the range [0,1]; is the convergence factor; is the maximum number of iterations;

[0031] The complexity of the actual environment leads to The wolf is not necessarily the global optimal solution. Therefore, we also need to know the relationship between each individual and Wolf and How close is the wolf? Calculate the distance from each individual to , and The formula for the centrifugal distance is as follows:

[0032] (7)

[0033] (8)

[0034] in, Is an individual of dimensional position vector; Is an individual of dimensional position vector; this is the historical optimal position vector. The current centrifugal distance of each individual is relative to , and The rate of change can be expressed as:

[0035] (9)

[0036] (10)

[0037] (11)

[0038] in, Represents the gray wolf individual relative to Changes in the wolf's centrifugal distance; Represents the gray wolf individual relative to Changes in the wolf's centrifugal distance; Represents the gray wolf individual relative to The change of the wolf's centrifugal distance. Then, we can calculate the centrifugal distance of each individual according to its relative , and The following adaptive weight factors are designed based on the centrifugal distance change rate:

[0039] (12)

[0040] (13)

[0041] (14)

[0042] in, express the extent of the influence that wolves have on other wolves; express the extent of the influence that wolves have on other wolves; express The influence of wolves on other wolves. Combined with the designed adaptive weight factor, the new position update formula is as follows:

[0043] (15)

[0044] in, Is an individual With Wolves The distance between Is an individual With Wolves The distance between Is an individual With Wolves The distance between.

[0045] Furthermore, the implementation process of step (6) is as follows:

[0046] In the GWO algorithm, when When When , the gray wolf population tends to narrow the local search range, surround the prey from all directions and attack. From formula (15), we can see that the parameter As the number of iterations increases, it decreases linearly to 0. However, since the GWO algorithm is a nonlinear optimization process, the linear decrease It does not meet the actual requirements. Therefore, in order to avoid falling into the local optimum too early, this article combines the PSO algorithm with the GWO algorithm and uses the adaptive coefficient to optimize the search ability of the algorithm. The new update formula is as follows:

[0047] (16)

[0048] in, is a random vector coefficient between [0,1]; is the adaptability vector coefficient, and the specific formula is as follows:

[0049] (17)

[0050] in, They are The fitness value of the wolf. Finally, the latest position of the individual species is confirmed, which is expressed as follows:

[0051] (18)

[0052] Finally, the fitness values ​​of equation (15) and equation (16) are compared.

[0053] Furthermore, the implementation process of step (7) is as follows:

[0054] When the maximum number of iterations is reached, the iteration stops and the optimal three-dimensional path of the UAV is obtained.

[0055] Beneficial effects: Compared with the prior art, the present invention has the following beneficial effects: the present invention expands the diversity of the population by adding an adversarial learning strategy to the initialization of the traditional gray wolf algorithm; secondly, the individual position is updated by improving the adaptive weight factor, thereby accelerating the convergence speed; finally, in order to avoid falling into the local optimum, the particle swarm algorithm is introduced to balance the global development and local convergence. A three-dimensional path planning method for UAVs based on the improved gray wolf algorithm is provided; it is conducive to reducing the possibility of local optimum, improving the efficiency of search, shortening the search path, and thus ensuring that the UAV obtains the optimal path in a shorter time. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 Flowchart for improving the Grey Wolf Algorithm;

[0057] Figure 2 It is the optimal path of the improved grey wolf algorithm in the present invention;

[0058] Figure 3 The best individual fitness change trend of the improved grey wolf algorithm in the present invention;

[0059] Figure 4 is the optimal path of the traditional grey wolf algorithm;

[0060] Figure 5 The changing trend of the best individual fitness of the traditional grey wolf algorithm; DETAILED DESCRIPTION

[0061] The present invention is further described in detail below with reference to the accompanying drawings. The present invention provides a three-dimensional path planning method for a UAV based on an improved gray wolf algorithm, such as Figure 1 As shown, the following steps are included:

[0062] Step 1: The three-dimensional coordinates of the drone are expressed as (X, Y, H). The horizontal coordinate X and the vertical coordinate Y are taken in the order of 0~1000. The starting point and the target point are set to (0, 0, 0) and (1000, 1000, 1000) respectively. The units of the horizontal coordinate X and the vertical coordinate Y are meters, and the value range is 0~1000 meters. The unit of the height H is meters, and the range is 0~1000 meters. Set the upper and lower limits of the gray wolf's movement to form a three-dimensional area where the drone is allowed to move. In this area, it can move left and right, up and down to get new points.

[0063] Step 2: Initialize the parameters of the improved gray wolf path planning method. In the process of initializing the parameters of the improved gray wolf path planning method, the corresponding parameters include the population size , maximum number of iterations , inertia weight , the cost coefficient of the objective function and dimensions , cognitive learning factors and social learning factors .

[0064] Step 3: First, the adversarial learning strategy is integrated into the gray wolf algorithm through the population initialization method. The initial population is set to is randomly generated in the search space, where It is the dimension of problem solving. is the population size. Then, this population is used to reproduce the next generation. The opposite position of the random search of adversarial learning (OBL) is generated by the following formula:

[0065] (1)

[0066] in, and They are The upper and lower bounds of the dimension; is the original value, is the opposite value, then, the initial population and opposing populations Combine and choose ,from The solution within the interval is selected as the optimal solution.

[0067] Step 4: Calculate the fitness value of each individual and define it as .

[0068] Step 5: In the gray wolf optimization algorithm, the gray wolf population is divided into and Four categories, The wolf's position is defined as the optimal solution. The wolf's position is defined as the suboptimal solution. The wolf's position is the third best solution, and the other candidate solutions are the remaining Wolf. When gray wolves hunt in groups, their main behaviors include surrounding, hunting, and attacking prey. The behavior of gray wolf populations around prey can be described by the following formula:

[0069] (2)

[0070] (3)

[0071] in, Indicates the distance between the individual and the target; is the current iteration number; It's the prey The position coordinates of the iteration can be expressed as ; is the position vector after the iteration A wolf can be represented as , =1,2… ; and is a coefficient vector, expressed as follows:

[0072] (4)

[0073] (5)

[0074] (6)

[0075] in, and is a random number in the range [0,1]; is the convergence factor; is the maximum number of iterations;

[0076] The complexity of the actual environment leads to The wolf is not necessarily the global optimal solution. Therefore, we also need to know the relationship between each individual and Wolf and How close is the wolf? Calculate the distance from each individual to , and The formula for the centrifugal distance is as follows:

[0077] (7)

[0078] (8)

[0079] in, Is an individual of dimensional position vector; Is an individual of dimensional position vector; this is the historical optimal position vector. The current centrifugal distance of each individual is relative to , and The rate of change can be expressed as:

[0080] (9)

[0081] (10)

[0082] (11)

[0083] in, Represents the gray wolf individual relative to Changes in the wolf's centrifugal distance; Represents the gray wolf individual relative to Changes in the wolf's centrifugal distance; Represents the gray wolf individual relative to The change of the wolf's centrifugal distance. Then, we can calculate the centrifugal distance of each individual according to its relative , and The following adaptive weight factors are designed based on the centrifugal distance change rate:

[0084] (12)

[0085] (13)

[0086] (14)

[0087] in, express the extent of the influence that wolves have on other wolves; express the extent of the influence that wolves have on other wolves; express The influence of wolves on other wolves. Combined with the designed adaptive weight factor, the new position update formula is as follows:

[0088] (15)

[0089] in, Is an individual With Wolves The distance between Is an individual With Wolves The distance between Is an individual With Wolves The distance between.

[0090] Step 6: In the GWO algorithm, when When When , the gray wolf population tends to narrow the local search range, surround the prey from all directions and attack. From formula (15), we can see that the parameter As the number of iterations increases, it decreases linearly to 0. However, since the GWO algorithm is a nonlinear optimization process, the linear decrease It does not meet the actual requirements. Therefore, in order to avoid falling into the local optimum too early, this article combines the PSO algorithm with the GWO algorithm and uses the adaptive coefficient to optimize the search ability of the algorithm. The new update formula is as follows:

[0091] (16)

[0092] in, is a random vector coefficient between [0,1]; is the adaptability vector coefficient, and the specific formula is as follows:

[0093] (17)

[0094] in, They are The fitness value of the wolf, and the latest position of the last confirmed species individual, is expressed as follows:

[0095] (18)

[0096] Finally, the fitness values ​​of equation (15) and equation (16) are compared.

[0097] Step 7: When the maximum number of iterations is reached, the iteration stops and the three-dimensional optimal path of the UAV is obtained.

[0098] To verify the correctness and effectiveness of the present invention, this paper simulates and compares the traditional gray wolf algorithm and the improved gray wolf algorithm using Matlab 2023b simulation software under AMD R7-5600H CPU 3.30GHz Window11 operating system. The algorithm parameters are set as follows: the starting point of the drone flight is defined as (0cm, 0cm, 0cm), the final point is set to (1000cm, 1000cm, 1000cm), and the cost coefficient of the objective function is is 0.4, the maximum number of iterations is 500, dimension is 10, the population size is 50, where the PSO inertia weight is 0.9, cognitive learning factor and social learning factor is 2.

[0099] The optimal path of the simulation results of the improved gray wolf algorithm and the traditional gray wolf algorithm proposed in the present invention in the three-dimensional environment model is as follows: Figure 2 and Figure 4 As shown in Figure 3 and Figure 5 As shown. By analyzing and comparing the simulation result data, it can be seen that the standard deviation of the traditional gray wolf algorithm is 47.3 and the running time is 74, and the path has more turns, the path nodes are more dispersed, and the convergence speed is slow. The standard deviation of the improved algorithm proposed in the present invention is 26.42 and the running time is 50, and the path has fewer turns, better path smoothness, and faster convergence speed. Therefore, the path obtained by the improved gray wolf algorithm is closer to the actual path, which meets the actual needs of the path planning of drones in a three-dimensional environment. The planning results of the traditional gray wolf algorithm and the improved gray wolf algorithm are compared in the form of a table, as shown in Table 1:

[0100] Table 1 Comparison between the present invention and the traditional grey wolf algorithm

[0101] algorithm Minimum fitness value Maximum fitness value Standard Deviation Algorithm running time / s Traditional Grey Wolf Algorithm 803.425 905.422 47.3 74 Improved Grey Wolf Algorithm (the present invention) 704.521 799.425 26.42 50

[0102] Through comparative simulation, it can be concluded that the improved gray wolf algorithm proposed in the present invention can effectively improve the convergence speed of the algorithm, and the optimization effect is better than that of the traditional gray wolf algorithm.

Claims

1. A three-dimensional path planning method for unmanned aerial vehicles based on an improved gray wolf algorithm, characterized in that: The following steps are involved: (1): Use the free space method to model the map environment and set the positions of the starting point and the target point; (2): Initialize and set the parameters of the gray wolf path planning method; (3): Based on the current position of the UAV and the allowed flight area, the adversarial learning strategy is used to randomly generate the initial population in the search space; (4): Calculate the fitness value of each individual and determine the fitness value based on the first three definitions in the sorting results; (5): Calculate the coefficient vectors A, C and update the new individual position according to the adaptive weight factor; (6): The particle search strategy in the PSO algorithm is introduced to calculate and update the position of another individual, and the selection and comparison phase is carried out. Finally, the position of the individual is determined and the new fitness value is recalculated; (7): When the maximum number of iterations is reached, the iteration stops and the three-dimensional optimal path of the UAV is obtained.

2. The method for three-dimensional path planning of an unmanned aerial vehicle based on an improved gray wolf algorithm according to claim 1, characterized in that: The implementation process of step (1) is as follows: The three-dimensional coordinates of the drone are expressed as (X, Y, H). The horizontal coordinate X and the vertical coordinate Y are taken in the order of 0~1000. The starting point and the target point are set to (0, 0, 0) and (1000, 1000, 1000) respectively. Among them, the units of the horizontal coordinate X and the vertical coordinate Y are meters, and the value range is 0~1000 meters. The unit of the height H is meters, and the range is 0~1000 meters. The upper and lower limits of the gray wolf's movement are set to form a three-dimensional area where the drone is allowed to move. In this area, it can move left and right, up and down to get new points.

3. The method for three-dimensional path planning of an unmanned aerial vehicle based on an improved gray wolf algorithm according to claim 1, characterized in that: The implementation process of step (2) is as follows: In the process of initializing the parameters of the improved gray wolf path planning method, the corresponding parameters include the population size , maximum number of iterations , inertia weight , the cost coefficient of the objective function and dimensions , cognitive learning factors and social learning factors .

4. The method for three-dimensional path planning of an unmanned aerial vehicle based on an improved gray wolf algorithm according to claim 1, characterized in that: The implementation process of step (3) is as follows: First, the adversarial learning strategy is integrated into the gray wolf algorithm through population initialization. is randomly generated in the search space, where It is the dimension of problem solving. is the population size; this population is then used to reproduce the next generation, and the adversarial learning (OBL) randomly searches for the opposite position generated by the following formula: (1) in, and They are The upper and lower bounds of the dimension; is the original value, is the opposite value; then, the initial population and opposing populations Combine and choose ,from The solution within the interval is selected as the optimal solution.

5. According to the method for three-dimensional path planning of unmanned aerial vehicles based on the improved gray wolf algorithm of claim 1, the implementation process of step (4) is as follows: the fitness value of each individual is calculated, and the first three fitness values ​​are defined as .

6. The method for three-dimensional path planning of an unmanned aerial vehicle based on an improved gray wolf algorithm according to claim 1, characterized in that: The implementation process of step (5) is as follows: In the gray wolf optimization algorithm, the gray wolf population is divided into and Four categories, The wolf's position is defined as the optimal solution. The wolf's position is defined as the suboptimal solution. The wolf's position is the third best solution, and the other candidate solutions are the remaining Wolf; When gray wolves hunt in groups, their main behaviors include surrounding, hunting, and attacking prey. The behavior of gray wolf populations around prey can be described by the following formula: (2) (3) in, Indicates the distance between the individual and the target; is the current iteration number; It's the prey The position coordinates of the iteration can be expressed as ; is the position vector after the iteration A wolf can be represented as , =1,2… ; and is a coefficient vector, expressed as follows: (4) (5) (6) in, and is a random number in the range [0,1]; is the convergence factor; is the maximum number of iterations. The complexity of the actual environment leads to The wolf is not necessarily the global optimal solution, so we also need to know the relationship between each individual and Wolf and How close are the wolves? Calculate the distance between each individual and , and The formula for the centrifugal distance is as follows: (7) (8) in, Is an individual of dimensional position vector; Is an individual of dimensional position vector; this is the historical optimal position vector; the current centrifugal distance of each individual is relative to , and The rate of change can be expressed as: (9) (10) (11) in, Represents the gray wolf individual relative to Changes in the wolf's centrifugal distance; Represents the gray wolf individual relative to Changes in the wolf's centrifugal distance; Represents the gray wolf individual relative to The change of the wolf's centrifugal distance; then, each individual can be calculated based on the current , and The following adaptive weight factors are designed based on the centrifugal distance change rate: (12) (13) (14) in, express the extent of the influence that wolves have on other wolves; express the extent of the influence that wolves have on other wolves; express The influence of wolves on other wolves. Combined with the designed adaptive weight factor, the new position update formula is as follows: (15) in, Is an individual With Wolves The distance between Is an individual With Wolves The distance between Is an individual With Wolves The distance between.

7. The method for three-dimensional path planning of an unmanned aerial vehicle based on an improved gray wolf algorithm according to claim 1, characterized in that: The implementation process of step (6) is as follows: In the GWO algorithm, when When When the prey is in danger, the gray wolf population tends to narrow the local search range, surround the prey from all directions and attack; From formula (15), we can see that the parameter As the number of iterations increases, it decreases linearly to 0. However, since the GWO algorithm is a nonlinear optimization process, the linear decrease It does not meet the actual requirements. Therefore, in order to avoid falling into the local optimum too early, this article combines the PSO algorithm with the GWO algorithm and uses the adaptive coefficient to optimize the search ability of the algorithm. The new update formula is as follows: (16) in, is a random vector coefficient between [0,1]; is the adaptability vector coefficient, and the specific formula is as follows: (17) in, They are The fitness value of the wolf finally enters the optimal update and replacement stage. By comparing the fitness values ​​of formula (30) and formula (31), it is expressed as follows: (18) Finally, the latest position of the species individual is confirmed by equations (15) and (16).

8. The method for three-dimensional path planning of an unmanned aerial vehicle based on an improved gray wolf algorithm according to claim 1, characterized in that: The implementation process of step (7) is as follows: when the maximum number of iterations is reached, the iteration stops, and the three-dimensional optimal path of the UAV is obtained.

Images