Unmanned aerial vehicle three-dimensional path planning method based on dynamic adaptive differential bat algorithm

The dynamic adaptive differential bat algorithm addresses the limitations of existing UAV path planning by integrating differential mutation and artificial potential fields, improving three-dimensional path planning efficiency and safety.

CN119063724APending Publication Date: 2024-12-03GUANGZHOU UNIVERSITY
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202411028694.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-30
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

Current UAV path planning technologies are limited to two-dimensional environments and face high calculation complexity and local optimal solutions in three-dimensional scenarios, failing to effectively consider obstacle maps and resulting in unsatisfactory path planning.

Method used

A three-dimensional path planning method for UAVs using a dynamic adaptive differential bat algorithm, incorporating differential mutation operators, artificial potential fields, and adaptive parameter adjustments to enhance global optimization and obstacle avoidance.

Benefits of technology

The method improves global optimization capability, enabling UAVs to perceive and avoid obstacles in complex three-dimensional environments, enhancing path planning feasibility and safety.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119063724A_ABST
    Figure CN119063724A_ABST
Patent Text Reader

Abstract

The invention relates to an unmanned aerial vehicle three-dimensional path planning method based on a dynamic adaptive difference bat algorithm, which comprises the following steps: A, environment modeling: setting a three-dimensional geographical environment in a task space of a single unmanned aerial vehicle individual, and constructing a function formula of multiple peaks in the three-dimensional geographical environment; b, evaluating a path planning effect, designing corresponding cost functions for path length, obstacle avoidance, flight height, path smoothness and track point uniformity, and proposing a total cost function based on the functions; and C, a dynamic adaptive differential bat algorithm, which introduces a differential mutation operator and an artificial potential field based on the bat algorithm, and introduces an adaptive parameter adjustment mechanism based on a population fitness historical change trend to the bat algorithm. The invention aims to provide the three-dimensional path planning method for the unmanned aerial vehicle, and the method can improve the global optimization capability of the unmanned aerial vehicle in the three-dimensional path planning, and enables the unmanned aerial vehicle to have the capability of sensing and avoiding obstacles under a three-dimensional ring.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to unmanned aerial vehicle path planning technology, in particular to a three-dimensional path planning method for unmanned aerial vehicle. Background Art

[0002] Unmanned Aerial Vehicles (UAVs), also known as unmanned aircraft, are aircraft that are operated by radio remote control equipment and self-contained program control devices, or are fully or intermittently autonomously operated by onboard computers. In the past two decades, UAVs have made significant contributions in both military and civilian fields and have become a popular tool for widespread use. As one of the key technologies of current UAV technology, UAV path planning has great research value and significance, and is a hot research field to solve the problem of UAVs performing flight missions in complex geographical environments.

[0003] The research on path planning of unmanned aerial vehicles (UAVs) was originally carried out to solve the shortest path problem on a two-dimensional plane. The research in this field has a history of several decades. Internationally, American researchers began to explore the path planning problem in the late 1950s. The path planning problem has gradually become one of the focuses of global research.

[0004] The current UAV path planning has the following problems: 1. The existing technology is generally applicable to two-dimensional plane path planning. When dealing with complex three-dimensional path planning, the calculation complexity is high and it is easy to fall into the local optimal solution, which affects the quality of path planning. 2. The existing technology generally studies the improvement method of the swarm intelligence algorithm itself, without considering the impact of the obstacle map space contained in the path planning task on the behavior of the intelligent group. The optimization algorithm search effect is not good in complex three-dimensional environments, resulting in unsatisfactory path planning. In order to improve the quality of path planning, it is necessary to develop a path planning method to overcome the above problems. Summary of the invention

[0005] In view of the above-mentioned defects, the purpose of the present invention is to propose a three-dimensional path planning method for UAV, which can improve the global optimization capability of the UAV in three-dimensional path planning and enable the UAV to have the ability to perceive and avoid obstacles in a three-dimensional environment.

[0006] To achieve this object, the present invention adopts the following technical solutions:

[0007] The three-dimensional path planning method of UAV based on dynamic adaptive differential bat algorithm includes the following steps:

[0008] A. Environmental modeling: For a single UAV, the three-dimensional geographical environment in its mission space is set, and the function formula of multiple peaks in the three-dimensional geographical environment is constructed;

[0009] B. Evaluation of path planning effect: design corresponding cost functions for path length, obstacle avoidance, flight altitude, path smoothness and track point uniformity, and propose a total cost function based on the above functions;

[0010] C. Dynamic adaptive differential bat algorithm. This algorithm introduces differential mutation operators and artificial potential fields based on the bat algorithm, and also introduces an adaptive parameter adjustment mechanism based on the historical change trend of population fitness.

[0011] Preferably, the method further includes step D, which is a measurement step in which an actual flight test is performed on a UAV using the above method in a preset three-dimensional space.

[0012] Preferably, in step A:

[0013] For a single drone, its mission space is a three-dimensional geographical environment with multiple mountain peaks at a certain altitude. The function formula for each mountain peak is as follows:

[0014] z peak (x, y) = h·e Λ (-((xx c ) Λ 2+(yy c ) Λ 2) / (2·w Λ 2)) Formula (1)

[0015] Where h is the height of the mountain, x c ,y c is the center coordinate of the peak, w is the width of the peak;

[0016] On this basis, the above-mentioned mountains are combined with the digital elevation model to form a simulated map environment.

[0017] Preferably, in step B:

[0018] Five cost functions Path length cost J1(X i ), obstacle avoidance cost J2(X i ), flight altitude cost J3(X i ), path smoothing cost J4(X i ), average cost of track points J5(X i ) design;

[0019] In formula (2), the shorter the path length, the better the path planning effect. For each search node, that is, the track point contains the corresponding three-dimensional coordinates P ij =(x ij ,y ij ,z ij ), the path cost is obtained by calculating the total distance from the starting point to the end point through all the track points:

[0020]

[0021] in, is the Euclidean distance between two search nodes;

[0022] In formula (3), let K be the set of all threats. In the flight path of the drone, any two consecutive track points P ij and P i,j+1 The path segments between can be represented as vectors The drone and the center of each mountain obstacle c k The closest distance between k Indicates that this distance is used to determine whether the drone has a collision risk;

[0023] The collision criterion depends on the radius R of the mountain. k , the distance between the UAV radius D and the danger distance S;

[0024] If the distance between the drone path and the mountain is less than the sum of these three distances d k <S+D+R k , then the path is considered to have a collision risk, and the obstacle avoidance cost is defined as follows:

[0025]

[0026] In formula (4), it is assumed that the UAV needs to maintain a maximum altitude h when performing a flight mission. max With minimum height h min The flight altitude cost is defined as follows:

[0027]

[0028] Among them, h ij Indicates the relative ground height of the drone between the i-th track point and the j-th track point;

[0029] In equations (5), (6) and (7), during the flight of the UAV, the pitch ψ and yaw φ actions are the key to controlling the flight direction and attitude. In order to ensure the stability and comfort of the flight, it is essential to smooth the pitch angle and yaw angle between consecutive track points in the flight path. The yaw angle φ ij , pitch angle ψ ij and path smoothing cost J4(X i ) is defined as follows:

[0030]

[0031] Among them, a1 is the yaw angle smoothing parameter, and a2 is the pitch angle smoothing parameter;

[0032] Formula (8) and Formula (9) assume that the UAV needs to fully search the mission space and needs the track points to be evenly distributed. The uniform cost of the track points is defined as follows:

[0033]

[0034] Where N is the number of track points, d i is the distance between track points;

[0035] In summary, the total cost function is obtained:

[0036]

[0037] Among them, b k is the weight of the cost, J k It's the price, X i is location information;

[0038] Further, in step C:

[0039] The standard bat algorithm is as follows:

[0040] In the D-dimensional search space, suppose there is a virtual bat i whose position is x i , the speed is v i , the transmitting frequency is f i ; At time t, the position of bat i is defined as Speed The speed and position update formulas are as follows:

[0041]

[0042] Among them, f max and f min are the minimum and maximum values ​​of the frequency, respectively, and β is a random number in [0,1]; x * is the current global optimal solution, which is the current optimal position obtained by comparing all bats;

[0043] Then, a bat is randomly selected and a local search is performed based on the neighborhood of that bat:

[0044] x new =x old +∈A t (14)

[0045] Among them, ∈ is a random number in [-1,1], A t is the average loudness of all bats at time t;

[0046] Finally, the important parameters that determine the search performance of the bat algorithm are updated:

[0047]

[0048] Among them, α and γ are constants, controlling the rate of change of loudness and pulse rate, and 0<α<1,γ>0; as the number of iterations t tends to infinity (t→∞), the loudness will approach 0, and the pulse emission rate will approach

[0049] Step C1 introduces a differential mutation operator into the bat algorithm, which enhances the global optimization capability of the bat algorithm:

[0050] In the differential evolution algorithm, when the population iterates to the tth generation, for each parent individual x i,t Perform mutation operation to obtain the mutant individual v i,t+1 , the mutation formula is as follows:

[0051]

[0052] in, are three different parent individuals randomly selected from the current population, are called basis vectors, is the difference vector, F is the mutation factor, and F>0 controls the influence of the difference vector on the mutant individual;

[0053] For the local search part of the bat algorithm, in order to enhance the algorithm’s exploration capability and improve the diversity of the population, two different update strategies were implemented, depending on the comparison between the random number rand and the pulse emission rate r;

[0054] The local search formula is as follows:

[0055]

[0056] Among them, rand∈[0,1] is a random real number extracted from a uniform distribution, ∈ is a random number in [-1,1], F is the mutation weight factor, which controls the influence of the differential vector, and a1, a2, a3 are random integers between 1 and the population number NP that are different from each other;

[0057] Step C2, introducing an artificial potential field into the bat algorithm to improve the feasibility and safety of path planning;

[0058] The potential function U(q) of the robot at a certain point q can be expressed as the gravitational potential U att (q) and the repulsive potential U att (q) Addition, such as Figure 7 As shown, the formula is as follows:

[0059] U(q)=U att (q)+Urep (q) (19)

[0060]

[0061] Among them, ζ is the gravitational gain, d(q) is the distance from the current point q to the target point q goal D(q) is the distance between point q and its nearest obstacle; η is the repulsive gain, Q * is the range threshold of the obstacle repulsion force;

[0062] The gravitational force and repulsive force are respectively given by the negative gradients of the gravitational potential and the repulsive potential It means that the gravitational force F at point q att (q) and repulsive force F rep (q) The formula is as follows:

[0063]

[0064] in, Represents the direction vector of the repulsive force;

[0065] When the population iterates to the tth generation, repulsion is applied to the search nodes that enter the influence range of the obstacle, and their positions are fine-tuned to keep the overall path at a relatively safe distance from the obstacle;

[0066] The local position update formula of the artificial potential field is as follows:

[0067]

[0068] in, Indicates the search node position that enters the obstacle threshold range, F rep (k) represents the repulsive force exerted on this node;

[0069] Step C3, adaptive parameter adjustment mechanism based on the historical trend of population fitness;

[0070] When the population iterates to the tth generation, the trend of the historical optimal fitness value of the last N generations is calculated as the fitness trend vector Trend, and the update method of the bat algorithm loudness and pulse rate is selected through this trend vector Trend; among them, when this trend vector Trend<0, the fitness is generally on a downward trend, indicating that the population is evolving towards a better solution; otherwise, it means that the population is trapped in a local optimum and needs to adjust the loudness and pulse rate for global search to jump out of the local extreme value;

[0071] The specific update formula is as follows:

[0072]

[0073] Where W is the number of generations selected as the historical fitness value, J is the fitness value, and α and θ are constants that control the update rate of loudness and pulse rate;

[0074] C31. When only dynamically adaptively adjusting parameters to improve the bat algorithm is selected, the standard bat algorithm update method is used for update;

[0075] C32. When the dynamic adaptive adjustment parameters and differential mutation strategy are used to improve the bat algorithm, the update is performed according to the following formula:

[0076]

[0077] Among the above two points, when Trend is less than 0, the update method of loudness and pulse rate is updated according to the standard bat algorithm process, that is, it is updated when the new solution is better than the current individual optimal solution.

[0078] Furthermore, the same UAV is used in the same mission space to test different path planning methods, and the cost data obtained by testing various methods are collected, and multiple sets of cost data are compared.

[0079] One of the above technical solutions includes the following beneficial effects: the dynamic adaptive differential bat algorithm of the present invention integrates the mutation strategy of differential evolution into the bat algorithm to enhance the diversity of the algorithm and improve the global optimization capability, which is suitable for solving multi-constrained high-dimensional optimization problems, such as three-dimensional path planning; the swarm intelligent agent perceives and avoids obstacles through artificial potential fields, and can avoid obstacles and maintain a safe distance in a complex three-dimensional obstacle environment, thereby improving the feasibility and safety of path planning; the dynamic adaptive differential bat algorithm of the present invention utilizes historical fitness to adaptively adjust its own loudness and pulse rate parameters, thereby enhancing the dynamic adaptive capability of the population and improving the local optimal dilemma problem of the optimization algorithm to seek a better quality path. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 is a flow chart of an overall scheme of an embodiment of the present invention;

[0081] Figure 2 is a schematic diagram of a three-dimensional simulation map according to an embodiment of the present invention;

[0082] Figure 3 is a schematic diagram of the obstacle avoidance cost definition according to an embodiment of the present invention;

[0083] Figure 4 is a schematic diagram of the pitch angle and yaw angle of a drone according to an embodiment of the present invention;

[0084] Figure 5 is a flow chart of a standard bat algorithm according to an embodiment of the present invention;

[0085] Figure 6 is a schematic diagram of artificial potential field force analysis of an embodiment of the present invention;

[0086] Figure 7 It is a schematic diagram of the improved strategy of the bat algorithm of the present invention;

[0087] Figure 8 is a flow chart of a dynamic adaptive differential bat algorithm according to an embodiment of the present invention;

[0088] Fig. 9 is an optimal path cost iteration graph during an experiment of an embodiment of the present invention;

[0089] Fig.10 is a path curve diagram during an experiment of an embodiment of the present invention;

[0090] Fig.11 It is an optimal cost comparison diagram during the experiment of an embodiment of the present invention. DETAILED DESCRIPTION

[0091] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0092] like Figure 1 As shown, the three-dimensional path planning method for UAV based on dynamic adaptive differential bat algorithm includes the following steps:

[0093] A. Environmental modeling: For a single UAV, the three-dimensional geographical environment in its mission space is set, and the function formula of multiple peaks in the three-dimensional geographical environment is constructed;

[0094] B. Evaluation of path planning effect: design corresponding cost functions for path length, obstacle avoidance, flight altitude, path smoothness and track point uniformity, and propose a total cost function based on the above functions;

[0095] C. Dynamic adaptive differential bat algorithm. This algorithm introduces differential mutation operators and artificial potential fields based on the bat algorithm, and also introduces an adaptive parameter adjustment mechanism based on the historical change trend of population fitness.

[0096] In addition, the method also includes step D, which is a measurement step in which an actual flight test is performed on a UAV using the above method in a preset three-dimensional space.

[0097] The dynamic adaptive differential bat algorithm of the present invention integrates the mutation strategy of differential evolution into the bat algorithm to enhance the diversity of the algorithm and improve the global optimization capability, and is suitable for solving multi-constrained high-dimensional optimization problems, such as three-dimensional path planning; the swarm intelligent agent perceives and avoids obstacles through artificial potential fields, and can avoid obstacles and maintain a safe distance in a complex three-dimensional obstacle environment, thereby improving the feasibility and safety of path planning; the dynamic adaptive differential bat algorithm of the present invention utilizes historical fitness to adaptively adjust its own loudness and pulse rate parameters, enhances the dynamic adaptive capability of the population, and improves the local optimal dilemma problem of the optimization algorithm to seek a better quality path.

[0098] like Figure 2 As shown, in step A:

[0099] For a single drone, its mission space is a three-dimensional geographical environment with multiple mountain peaks at a certain altitude. The function formula for each mountain peak is as follows:

[0100] z peak (x, y) = h·e Λ (-((xx c ) Λ 2+(yy c ) Λ 2) / (2·w Λ 2)) Formula (1)

[0101] Where h is the height of the mountain, x c ,y c is the center coordinate of the peak, w is the width of the peak;

[0102] On this basis, the above-mentioned mountains are combined with the digital elevation model to form a simulated map environment.

[0103] A three-dimensional geographical environment model with multiple mountains was constructed, which provided a realistic simulation environment for the three-dimensional path planning of UAVs and enhanced the adaptability and feasibility of the path planning algorithm.

[0104] Wherein, in said step B:

[0105] Five cost functions Path length cost J1(X i ), obstacle avoidance cost J2(X i ), flight altitude cost J3(X i ), path smoothing cost J4(X i ), average cost of track points J5(X i ) design;

[0106] In formula (2), the shorter the path length, the better the path planning effect. For each search node, that is, the track point has the corresponding three-dimensional coordinate Pij =(xi j ,yi j ,zi j ), the path cost is obtained by calculating the total distance from the starting point to the end point through all the track points:

[0107]

[0108] in, is the Euclidean distance between two search nodes;

[0109] In formula (3), let K be the set of all threats. In the flight path of the drone, any two consecutive track points P ij and P i,j+1 The path segments between can be represented as vectors The drone and the center of each mountain obstacle c k The closest distance between k Indicates that this distance is used to determine whether the drone has a collision risk;

[0110] The collision criterion depends on the radius R of the mountain. k , the distance between the UAV radius D and the danger distance S;

[0111] If the distance between the drone path and the mountain is less than the sum of these three distances d k <S+D+R k , then the path is considered to have a collision risk, such as Figure 3 As shown. The obstacle avoidance cost is defined as follows:

[0112]

[0113] In formula (4), it is assumed that the UAV needs to maintain a maximum altitude h when performing a flight mission. max With minimum height h min The flight altitude cost is defined as follows:

[0114]

[0115] Among them, h ij Indicates the relative ground height of the drone between the i-th track point and the j-th track point;

[0116] In equations (5), (6) and (7), during the flight of the UAV, the pitch ψ and yaw φ actions are the key to controlling the flight direction and attitude. In order to ensure the stability and comfort of the flight, it is essential to smooth the pitch angle and yaw angle between consecutive track points in the flight path, as shown in Figure 4 As shown. Yaw angle φ ij , pitch angle ψ ijand path smoothing cost J4(X i ) is defined as follows:

[0117]

[0118] Among them, a1 is the yaw angle smoothing parameter, and a2 is the pitch angle smoothing parameter;

[0119] Formula (8) and Formula (9) assume that the UAV needs to fully search the mission space and needs the track points to be evenly distributed. The uniform cost of the track points is defined as follows:

[0120]

[0121] Where N is the number of track points, d i is the distance between track points;

[0122] In summary, the total cost function is obtained:

[0123]

[0124] Among them, b k is the weight of the cost, J k It's the price, X i It is location information.

[0125] A variety of cost functions have been designed, including path length, obstacle avoidance, flight altitude, path smoothness and uniform cost of track points. Various requirements for UAV flight have been comprehensively considered. The issues that UAVs need to pay attention to when performing flight missions are expressed in the form of cost functions as much as possible and the path planning effect is verified, thereby improving the quality and reliability of path planning.

[0126] In step C:

[0127] The standard bat algorithm is as follows:

[0128] In the D-dimensional search space, suppose there is a virtual bat i whose position is x i , the speed is v i , the transmitting frequency is f i ; At time t, the position of bat i is defined as Speed The speed and position update formulas are as follows:

[0129] f i =f min +(f max -f min )β (11)

[0130]

[0131] Among them, f max and fmin are the minimum and maximum values ​​of the frequency, respectively, and β is a random number in [0,1]; x * is the current global optimal solution, which is the current optimal position obtained by comparing all bats;

[0132] Then, a bat is randomly selected and a local search is performed based on the neighborhood of that bat:

[0133] x new =x old +∈A t (14)

[0134] Among them, ∈ is a random number in [-1,1], A t is the average loudness of all bats at time t;

[0135] Finally, the important parameters that determine the search performance of the bat algorithm are updated:

[0136]

[0137]

[0138] Among them, α and γ are constants, controlling the rate of change of loudness and pulse rate, and 0<α<1,γ>0; as the number of iterations t tends to infinity (t→∞), the loudness will tend to 0, and the pulse emission rate will tend to r i 0 ;

[0139] Step C1 introduces a differential mutation operator into the bat algorithm, which enhances the global optimization capability of the bat algorithm:

[0140] In the differential evolution algorithm, when the population iterates to the tth generation, for each parent individual x i,t Perform mutation operation to obtain the mutant individual v i,t+1 , the mutation formula is as follows:

[0141]

[0142] in, are three different parent individuals randomly selected from the current population, are called basis vectors, is the difference vector, F is the mutation factor, and F>0 controls the influence of the difference vector on the mutant individual;

[0143] For the local search part of the bat algorithm, in order to enhance the algorithm’s exploration capability and improve the diversity of the population, two different update strategies were implemented, depending on the comparison between the random number rand and the pulse emission rate r;

[0144] The local search formula is as follows:

[0145]

[0146] Among them, rand∈[0,1] is a random real number extracted from a uniform distribution, ∈ is a random number in [-1,1], F is the mutation weight factor, which controls the influence of the differential vector, and a1, a2, a3 are random integers between 1 and the population number NP that are different from each other;

[0147] Step C2, introducing an artificial potential field into the bat algorithm to improve the feasibility and safety of path planning;

[0148] The potential function U(q) of the robot at a certain point q can be expressed as the gravitational potential U att (q) and the repulsive potential U att (q) Addition, such as Figure 7 As shown, the formula is as follows:

[0149] U(q)=U att (q)+U rep (q) (19)

[0150]

[0151] Among them, ζ is the gravitational gain, d(q) is the distance from the current point q to the target point q goal D(q) is the distance between point q and its nearest obstacle; η is the repulsive gain, Q * is the range threshold of the obstacle repulsion force;

[0152] The gravitational force and repulsive force are respectively given by the negative gradients of the gravitational potential and the repulsive potential It means that the gravitational force F at point q att (q) and repulsive force F rep (q) The formula is as follows:

[0153]

[0154] in, Represents the direction vector of the repulsive force;

[0155] When the population iterates to the tth generation, repulsion is applied to the search nodes that enter the influence range of the obstacle, and their positions are fine-tuned to keep the overall path at a relatively safe distance from the obstacle;

[0156] The local position update formula of the artificial potential field is as follows:

[0157]

[0158] in, Indicates the search node position that enters the obstacle threshold range, F rep (k) represents the repulsive force exerted on this node;

[0159] Step C3, adaptive parameter adjustment mechanism based on the historical trend of population fitness;

[0160] When the population iterates to the tth generation, the trend of the historical optimal fitness value of the last N generations is calculated as the fitness trend vector Trend, and the update method of the bat algorithm loudness and pulse rate is selected through this trend vector Trend; among them, when this trend vector Trend<0, the fitness is generally on a downward trend, indicating that the population is evolving towards a better solution; otherwise, it means that the population is trapped in a local optimum and needs to adjust the loudness and pulse rate for global search to jump out of the local extreme value;

[0161] The specific update formula is as follows:

[0162]

[0163] Where W is the number of generations selected as the historical fitness value, J is the fitness value, and α and θ are constants that control the update rate of loudness and pulse rate;

[0164] C31. When only dynamically adaptively adjusting parameters to improve the bat algorithm is selected, the standard bat algorithm update method is used for update;

[0165] C32. When the dynamic adaptive adjustment parameters and differential mutation strategy are used to improve the bat algorithm, the update is performed according to the following formula:

[0166]

[0167] Among the above two points, when Trend is less than 0, the update method of loudness and pulse rate is updated according to the standard bat algorithm process, that is, it is updated when the new solution is better than the current individual optimal solution.

[0168] By introducing differential mutation operators, artificial potential fields and adaptive parameter adjustment mechanisms, the global optimization and obstacle avoidance capabilities of the bat algorithm are improved, and the ability of the bat algorithm to solve complex high-dimensional multi-constraint problems is improved, thereby enabling drones to plan paths more effectively in complex three-dimensional environments.

[0169] As a supplement to the above technologies, the same UAV is used in the same mission space to test different path planning methods, collect cost data obtained by testing various methods, and compare multiple sets of cost data.

[0170] By comparing multiple sets of cost data, the advantages of the dynamic adaptive differential bat algorithm over other path planning methods in the same task space are verified, which provides data support and enhances the credibility of this method.

[0171] In the experimental test, the starting position is set to [450, 50, 100], the end position is set to [50, 550, 120], and the weights of the indicators in the cost function are path length weight 2, obstacle avoidance weight 3, flight height weight 1, path smoothing weight 2, and track point uniform weight 2. The parameter settings of the bat algorithm are shown in Table 1, and the parameter settings of the mountain obstacle are shown in Table 2.

[0172] Table 1 Dynamic adaptive differential bat algorithm parameters

[0173]

[0174] Table 2 Mountain obstacle parameters

[0175]

[0176]

[0177] In this experiment, the dynamic adaptive differential bat algorithm (DADBA) combines the three improvement strategies designed in the previous article and is compared with the standard bat algorithm (BA), particle swarm algorithm (PSO) and artificial bee colony algorithm (ABC). Fig. 9 Comparison of the optimal path cost between the state adaptive differential bat algorithm (DADBA) and other algorithms. Fig.10 Its path curve diagram. Fig.11 This is the optimal cost comparison diagram. The number of iterations is 200 and the number of populations is 200.

[0178] The experimental results show that the dynamic adaptive differential bat algorithm (DADBA) achieves the best cost result of 1349.8 in the path planning of UAVs in complex three-dimensional environments, which is significantly better than the standard bat algorithm (2618.1) as well as the particle swarm algorithm (1886.7) and the artificial bee colony algorithm (1738.9). This result confirms the superiority of the DADBA algorithm, which uses differential evolution strategy, artificial potential field method and adaptive parameter adjustment, in terms of obstacle avoidance efficiency, safety and cost efficiency of UAV path planning.

[0179] The technical principle of the present invention is described above in conjunction with specific embodiments. These descriptions are only for explaining the principle of the present invention and cannot be interpreted as limiting the scope of protection of the present invention in any way. Based on the explanations herein, those skilled in the art can associate other specific implementations of the present invention without paying creative labor, and these methods will fall within the scope of protection of the present invention.

Claims

1. A three-dimensional path planning method for unmanned aerial vehicles based on a dynamic adaptive differential bat algorithm, characterized in that: The steps include: A. Environmental modeling: For a single UAV, the three-dimensional geographical environment in its mission space is set, and the function formula of multiple peaks in the three-dimensional geographical environment is constructed; B. Evaluation of path planning effect: design corresponding cost functions for path length, obstacle avoidance, flight altitude, path smoothness and track point uniformity, and propose a total cost function based on the above functions; C. Dynamic adaptive differential bat algorithm. This algorithm introduces differential mutation operators and artificial potential fields based on the bat algorithm, and also introduces an adaptive parameter adjustment mechanism based on the historical change trend of population fitness.

2. The method for three-dimensional path planning of unmanned aerial vehicle based on dynamic adaptive differential bat algorithm according to claim 1 is characterized in that: The method further includes step D, which is a measurement step in which an actual flight test is performed on a UAV using the above method in a preset three-dimensional space.

3. The method for three-dimensional path planning of unmanned aerial vehicle based on dynamic adaptive differential bat algorithm according to claim 1 is characterized in that: In step A: For a single drone, its mission space is a three-dimensional geographical environment with multiple mountain peaks at a certain altitude. The function formula for each mountain peak is as follows: z peak (x, y) = h·e Λ (-((x - x c ) Λ 2 + (y - y c ) Λ 2) / (2·w Λ 2)) Equation (1) Where h is the height of the mountain, x c ,y c is the center coordinate of the peak, w is the width of the peak; On this basis, the above-mentioned mountains are combined with the digital elevation model to form a simulated map environment.

4. The method for three-dimensional path planning of unmanned aerial vehicle based on dynamic adaptive differential bat algorithm according to claim 1 is characterized in that: In the step B: Five cost functions Path length cost J1(X i ), obstacle avoidance cost J2(X i ), flight altitude cost J3(X i ), path smoothing cost J4(X i ), average cost of track points J5(X i ) design; In formula (2), the shorter the path length, the better the path planning effect. For each search node, that is, the track point contains the corresponding three-dimensional coordinates P ij =(x ij ,y ij ,z ij ), the path cost is obtained by calculating the total distance from the starting point to the end point through all the track points: in, is the Euclidean distance between two search nodes; In formula (3), let K be the set of all threats. In the flight path of the drone, any two consecutive track points P ij and P i,j+1 The path segments between can be represented as vectors The drone and the center of each mountain obstacle c k The closest distance between k Indicates that this distance is used to determine whether the drone has a collision risk; The collision criterion depends on the radius R of the mountain. k , the distance between the UAV radius D and the danger distance S; If the distance between the drone path and the mountain is less than the sum of these three distances d k <S+D+R k , then the path is considered to have a collision risk, and the obstacle avoidance cost is defined as follows: In formula (4), it is assumed that the UAV needs to maintain a maximum altitude h when performing a flight mission. max With minimum height h min The flight altitude cost is defined as follows: Among them, h ij Indicates the relative ground height of the drone between the i-th track point and the j-th track point; In equations (5), (6) and (7), during the flight of the UAV, the pitch ψ and yaw φ actions are the key to controlling the flight direction and attitude. In order to ensure the stability and comfort of the flight, it is essential to smooth the pitch angle and yaw angle between consecutive track points in the flight path. The yaw angle φ ij , pitch angle ψ ij and path smoothing cost J4(X i ) is defined as follows: Among them, a1 is the yaw angle smoothing parameter, and a2 is the pitch angle smoothing parameter; Formula (8) and Formula (9) assume that the UAV needs to fully search the mission space and needs the track points to be evenly distributed. The uniform cost of the track points is defined as follows: Where N is the number of track points, d i is the distance between track points; In summary, the total cost function is obtained: Among them, b k is the weight of the cost, J k It's the price, X i It is location information.

5. The method for three-dimensional path planning of unmanned aerial vehicle based on dynamic adaptive differential bat algorithm according to claim 1, characterized in that: In step C: The standard bat algorithm is as follows: In the D-dimensional search space, suppose there is a virtual bat i whose position is x i , the speed is v i , the transmitting frequency is f i ; At time t, the position of bat i is defined as Speed The speed and position update formulas are as follows: f i =f min +(f max -f min )β (11) Among them, f max and f min are the minimum and maximum values ​​of the frequency, respectively, and β is a random number in [0,1]; x * is the current global optimal solution, which is the current optimal position obtained by comparing all bats; Then, a bat is randomly selected and a local search is performed based on the neighborhood of that bat: x new =x old +∈A t (14) Among them, ∈ is a random number in [-1,1], A t is the average loudness of all bats at time t; Finally, the important parameters that determine the search performance of the bat algorithm are updated: Among them, α and γ are constants, controlling the rate of change of loudness and pulse rate, and 0<α<1,γ>0; as the number of iterations t tends to infinity (t→∞), the loudness will tend to 0, and the pulse emission rate will tend to r i 0 ; Step C1 introduces a differential mutation operator into the bat algorithm, which enhances the global optimization capability of the bat algorithm: In the differential evolution algorithm, when the population iterates to the tth generation, for each parent individual x i,t Perform mutation operation to obtain the mutant individual v i,t+1 , the mutation formula is as follows: in, are three different parent individuals randomly selected from the current population, are called basis vectors, is the difference vector, F is the mutation factor, and F>0 controls the influence of the difference vector on the mutant individual; For the local search part of the bat algorithm, in order to enhance the algorithm’s exploration capability and improve the diversity of the population, two different update strategies were implemented, depending on the comparison between the random number rand and the pulse emission rate r; The local search formula is as follows: Among them, rand∈[0,1] is a random real number extracted from a uniform distribution, ∈ is a random number in [-1,1], F is the mutation weight factor, which controls the influence of the differential vector, and a1, a2, a3 are random integers between 1 and the population number NP that are different from each other; Step C2, introducing an artificial potential field into the bat algorithm to improve the feasibility and safety of path planning; The potential function U(q) of the robot at a certain point q can be expressed as the gravitational potential U att (q) and the repulsive potential U att (q) is added, and the formula is as follows: U(q)=U att (q)+U rep (q) (19) Among them, ζ is the gravitational gain, d(q) is the distance from the current point q to the target point q goal D(q) is the distance between point q and its nearest obstacle; η is the repulsive gain, Q * is the range threshold of the obstacle repulsion force; The gravitational force and repulsive force are respectively given by the negative gradients of the gravitational potential and the repulsive potential It means that the gravitational force F at point q att (q) and repulsive force F rep (q) The formula is as follows: in, Represents the direction vector of the repulsive force; When the population iterates to the tth generation, repulsion is applied to the search nodes that enter the influence range of the obstacle, and their positions are fine-tuned to keep the overall path at a relatively safe distance from the obstacle; The local position update formula of the artificial potential field is as follows: in, Indicates the search node position that enters the obstacle threshold range, F rep (k) represents the repulsive force exerted on this node; Step C3, adaptive parameter adjustment mechanism based on the historical trend of population fitness; When the population iterates to the tth generation, the trend of the historical optimal fitness value of the last N generations is calculated as the fitness trend vector Trend, and the update method of the bat algorithm loudness and pulse rate is selected through this trend vector Trend; among them, when this trend vector Trend<0, the fitness is generally on a downward trend, indicating that the population is evolving towards a better solution; otherwise, it means that the population is trapped in a local optimum and needs to adjust the loudness and pulse rate for global search to jump out of the local extreme value; The specific update formula is as follows: in, W is the number of generations chosen as the calculation of historical fitness values, J is the fitness value, α and θ are constants that control the update rate of loudness and pulse rate; C31. When only dynamically adaptively adjusting parameters to improve the bat algorithm is selected, the standard bat algorithm update method is used for update; C32. When the dynamic adaptive adjustment parameters and differential mutation strategy are used to improve the bat algorithm, the update is performed according to the following formula: Among the above two points, when Trend is less than 0, the update method of loudness and pulse rate is updated according to the standard bat algorithm process, that is, it is updated when the new solution is better than the current individual optimal solution.

6. The method for three-dimensional path planning of unmanned aerial vehicle based on dynamic adaptive differential bat algorithm according to claim 2 is characterized in that: Use the same drone in the same mission space and different path planning methods to test, collect cost data obtained from testing with various methods, and compare multiple sets of cost data.

Citation Information

Cited By

  • Flying car multi-sensor fusion path planning method, system and equipment and storage medium

    CN119737969A

  • Arc diameter measuring method

    CN122041737A

  • A method for measuring the diameter of a circular arc

    CN122041737B