Unmanned aerial vehicle path planning method based on global search and meta-heuristic optimization

Abstract

The invention discloses an unmanned aerial vehicle path planning method based on global search and meta-heuristic optimization, and the method comprises the steps: firstly building an unmanned aerial vehicle path planning model under complex constraints, determining a fitness function and a constraint condition thereof, setting algorithm initialization parameters, carrying out the chaotic mapping of an initialization population position, and carrying out the optimization of the algorithm; and performing loop iteration solution by using a meta-heuristic optimization algorithm, outputting a global optimal path, finally realizing the global optimal path on the unmanned aerial vehicle, and calculating a flight distance corresponding to the optimal path to complete unmanned aerial vehicle path planning. According to the method, a meta-heuristic algorithm, namely a crown porcupine algorithm, is combined with tabu search, the global search capability is enhanced, and meanwhile, population initialization is optimized in combination with chaotic mapping, so that the randomness and unpredictability are increased, the algorithm can converge more quickly, and the robustness of the algorithm is improved. The method has obvious advantages in solving the problem of path planning with complex environment and complex constraint, and solves the problem of path planning of the unmanned aerial vehicle in the three-dimensional complex mountain environment.

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned aerial vehicle mission planning, and in particular relates to a unmanned aerial vehicle path planning method based on global search and meta-heuristic optimization. Background Art

[0002] In recent years, drones have played an important role in environmental monitoring, disaster relief, transportation and other fields with their low cost, small size and high flexibility. Among them, path planning, as one of the most important technologies in drone engineering applications, has been widely concerned and studied. Drone path planning refers to the process of finding one or more flyable paths for drones in a given flight environment. The path minimizes the flight cost or maximizes the coverage while meeting the mission constraints. The effectiveness of path planning directly affects the efficiency and safety of drone mission execution.

[0003] At present, path planning methods can be divided into two categories: one is the path planning method based on traditional optimization algorithms, and the other is the path planning method based on intelligent optimization algorithms. Traditional optimization algorithms include sampling-based algorithms, such as A* algorithm, probabilistic roadmap (PRM) algorithm, rapidly expanding random tree (RRT) algorithm, etc. Sampling-based algorithms are usually difficult to guarantee to find the global optimum, and the sampled path is not smooth enough; in addition to sampling algorithms, there are also methods based on mathematical modeling, that is, considering multiple constraints and establishing mathematical models to solve problems. This method is difficult to establish models in complex environments and cannot handle fuzzy information. Intelligent optimization algorithms include neural networks, reinforcement learning and meta-heuristic algorithms. Such algorithms have the advantages of high adaptability, strong generalization ability, and no need for precise modeling, but they also have a series of defects. Among them, neural networks rely on a large number of learning samples and have slow convergence speed; reinforcement learning takes a long time to train, has poor interpretability, and is difficult to set reward functions. Meta-heuristic algorithms inspired by swarm intelligence (SI) simulate population behavior in nature and build optimizers to solve problems with lower computational complexity, so they have received extensive research attention. However, these algorithms currently face some challenges, including slow convergence and the problem of being easily trapped in local optimal solutions. In summary, the current algorithms for solving the UAV path planning problem still have the following shortcomings:

[0004] (1) The algorithm has poor global search capabilities. Existing sampling-based algorithms rely on sampling density and are difficult to find the global optimum. Swarm intelligence algorithms are prone to falling into local optimal solutions during iterations, especially under complex constraints. This problem is more likely to occur when the solution space is very complex and there are multiple local optimal solutions.

[0005] (2) The algorithm converges slowly. Methods based on neural networks and reinforcement learning have a long training time and often require a large number of iterations to find the optimal solution during the calculation process. Not only does it converge slowly, it also occupies a large amount of lithium acid resources.

[0006] (3) The path obtained by the algorithm is difficult to use for actual flight. The paths obtained by the existing sampling algorithms are not smooth enough and often have large curvatures, which exceed the maximum turning angle of the drone and do not conform to actual flight conditions. General path planning algorithms lack consideration of constraints and only focus on range constraints, while altitude, pitch angle, yaw angle, etc. are also constraints that need to be considered in actual flight. Summary of the invention

[0007] To solve the above technical problems, the present invention provides a UAV path planning method based on global search and meta-heuristic optimization, which solves the path planning problem of UAV in three-dimensional complex mountainous environments and has obvious advantages when facing problems with complex constraints and complex environments.

[0008] The technical solution adopted by the present invention is: a UAV path planning method based on global search and meta-heuristic optimization, the specific steps are as follows:

[0009] S1. Establish a UAV path planning model under complex constraints, that is, establish a three-dimensional mountain model based on function representation, and establish the fitness function and its constraints;

[0010] S2, set the algorithm initialization parameters and initialize the population position by chaos mapping;

[0011] The initialization parameters include: initial population size, minimum population size, maximum number of iterations, number of cyclic population cycles, convergence rate and balance coefficient.

[0012] S3, using a meta-heuristic optimization algorithm to perform cyclic iterative solution on the path planning model in step S1;

[0013] Set the problem-solving parameters and iteration end conditions; record the global optimal solutions in the solution space, that is, the optimal position; use the cyclic population reduction strategy to update the population size, and simulate the four different defensive behaviors of crested porcupines when facing predators in the natural environment to establish an optimization iteration method; update the global optimal solution after one cycle; after traversing the population, determine whether the global optimal solution is updated. If the global optimal solution is not updated, perform a taboo search in the neighborhood of the current position to find the optimal solution in the neighborhood and compare it with the global optimal value; until the iteration end conditions are met, end the optimization and output the global optimal path.

[0014] S4. Based on step S3, the global optimal path is implemented on the UAV, and the flight distance corresponding to the optimal path, that is, the fitness function value corresponding to the optimal path, is calculated to complete the UAV path planning.

[0015] Furthermore, the step S1 is specifically as follows:

[0016] S11, setting the UAV flight environment to a rectangular space in a Cartesian coordinate system;

[0017] S12, considering the mountain terrain environment and obstacle distribution, the environment is modeled by combining the reference terrain model and the exponential function, that is, a three-dimensional mountain model is established based on the function representation method;

[0018] The expression of the three-dimensional mountain model based on the function representation is as follows:

[0019]

[0020] Among them, Z(x,y) represents the height corresponding to the point (x,y), M represents the number of peaks, and H i represents the height control parameter, X si and Y si Represents the attenuation parameter in x,y coordinates, (x i ,y i ) represents the center coordinates of the i-th peak.

[0021] S13, determining the constraint conditions of the UAV path planning model;

[0022] Between the control points, multiple path points are generated by the Bezier curve interpolation method, and a set of control point sequences is set as {S,Q 1 ,Q 2 ,...,Q n-1 ,G}.

[0023] Among them, the control points are ordered three-dimensional coordinate points, indicating the basic direction of path search; S and G represent the starting point and the end point respectively, and Q 1 To Q n-1 Represents an intermediate node of a path.

[0024] Then use the above n+1 nodes to generate a smooth Bezier curve, the expression is as follows:

[0025]

[0026]

[0027] Among them, τ represents a one-dimensional sequence of equally spaced points from 0 to 1, B i,n(τ) represents the Bezier curve between the i-th control point and the i+1-th control point, P i Indicates {Q 1 ,Q 2 ,...,Q n-1 , a control point in G}, and P(τ) represents a path point.

[0028] The constraints of the UAV path planning model are as follows:

[0029] (1) In order to avoid collision with obstacles, the height of the drone should always be higher than the ground. The expression is as follows:

[0030]

[0031] Among them, z j represents the height of the drone at the jth path point, Z(x j ,y j ) represents the peak height of the jth path point.

[0032] (2) The drone can only fly within a certain area and cannot exceed the environmental boundary. The expression is as follows:

[0033] x min ≤x j ≤x max

[0034] Among them, x min and x max Indicates the minimum and maximum values ​​of the horizontal coordinate x in the flight space, and the same applies to y and z.

[0035] (3) The drone should fly within an appropriate flight altitude. Flying too high or too low will be subject to a height penalty, as shown in the following expression:

[0036]

[0037] Among them, f h Indicates high penalty, h min and h max represents the upper and lower bounds of reasonable flight altitude, w h Represents the height penalty factor.

[0038] (4) Actual flight requires the path to be as smooth as possible. Excessive yaw and pitch angles will be penalized. The expression is as follows:

[0039]

[0040]

[0041] Among them, f yaw and fpit denote the yaw angle penalty and pitch angle penalty respectively, ψ j and θ j denote the yaw angle and pitch angle formed by the j-th path point and the j+1-th path point, respectively, and ψ max and θ max Respectively represent the maximum yaw angle and the maximum pitch angle, w yaw and w pit Represent the penalty factors for yaw angle and pitch angle respectively.

[0042] S14, establishing a fitness function based on the constraints determined in step S13;

[0043] The fitness function f is expressed as follows:

[0044] f=min(f L +f yaw +f pit +f h )

[0045] Among them, f L Represents the UAV path length.

[0046] Furthermore, the step S2 is specifically as follows:

[0047] S21. Define the initial population size as N and the minimum population size as N min , the number of cyclic population cycles T, the maximum number of iterations t max , convergence rate α and balance coefficient T f ;

[0048] S22, chaos mapping initializes population position;

[0049] First, the population position is randomly generated according to the uniform distribution. The expression is as follows:

[0050]

[0051] in, Represents a population individual, a feasible solution in the solution space, that is, a feasible set of control points. A set of control point positions is taken as an individual, and the optimal solution is found through the optimization strategy as the final control point position set; and represents the lower and upper bounds of the solution space, represents a random vector uniformly distributed in the interval [0,1]. The initial population is represented as a matrix, that is,

[0052] For the initialized population Use the Singer chaotic mapping strategy to initialize the population again. The expression is as follows:

[0053]

[0054] Where a represents the chaotic mapping control parameter, and a∈(0.9,1.08), Represents the updated initial population position.

[0055] Furthermore, the step S3 is specifically as follows:

[0056] S31, updating the population size based on the cyclic population reduction strategy;

[0057] Calculate the current optimal position in the environment solution space when the current iteration number t is reached The expression is as follows:

[0058]

[0059] in, Represents the individual position that minimizes the objective function value at time t.

[0060] According to the current population size N t , based on the cyclic population reduction strategy, the population size is updated, and the update expression is as follows:

[0061]

[0062] S32, traversing each crested porcupine individual in the population, each individual randomly performs exploration or development, that is, calculating individual defense parameters and attack parameters;

[0063] Among them, exploration includes two behaviors: visual defense and sound defense, and development includes two behaviors: odor attack and physical attack.

[0064] Before exploration or exploitation, each individual parameter that needs to be updated includes: Odor diffusion rate Predator quality The average force of the predator on the crested porcupine is expressed as follows:

[0065]

[0066]

[0067]

[0068] in, represents the objective function value of the i-th individual at time t, ε 1 , ε 2 It means to avoid small values ​​with denominator of 0; represents a random vector from 0 to 1, represents the initial velocity of the ith individual at time t, represents the final velocity of the i-th individual at the next iteration t+1, Indicates the position of a crested porcupine individual randomly selected at the current iteration number, and assigns it to And 1≤r≤N t , t represents the number of current iterations.

[0069] Then generate two uniformly distributed random numbers in the interval (0,1), ξ 1 ,ξ 2 , when 1 <ξ 2 When ξ 1 ≥ξ 2 Execute development and enter step S34.

[0070] S33. Based on step S32, generate two uniformly distributed random numbers, ξ 3 ,ξ 4 , when 3 <ξ 4 When ξ 3 ≥ξ 4 When , the second defense strategy, namely sound defense, is executed, and the individual position is updated according to the defense strategy;

[0071] 1) Visual defense;

[0072] To simulate visual defense behavior, the position update method is expressed as follows:

[0073]

[0074] Among them, ξ 5 represents a random number based on normal distribution, ξ 6 Represents a random value in the interval [0,1]; represents the vector generated between the current crested porcupine and a randomly selected crested porcupine from the population, which is used to represent the position of the predator at time t.

[0075] 2) Sound defense;

[0076] Simulate sound defense behavior, and update the position as follows:

[0077]

[0078] Among them, ξ 7 represents a random value in the interval [0,1], r 1 and r 2 represents a random integer between [1, N], Represents a randomly generated binary vector of 0 and 1.

[0079] S34: Based on step S32, generate a random number ξ uniformly distributed between (0, 1) 8 , when 8 <T f When ξ 8 ≥T f When , the defense strategy 4, i.e., physical attack, is executed, and the individual position is updated according to the defense strategy;

[0080] 3) Odor attack;

[0081] To simulate the odor attack behavior, the position update method is expressed as follows:

[0082]

[0083] Among them, r 3 represents a random integer between [1, N], represents a parameter vector used to control the search direction, with equal probability of selecting all-one or all-negative-one vectors. represents the defense factor, ξ 9 Represents a random value in the interval [0,1].

[0084] 4) physical assault;

[0085] To simulate physical attack behavior, the position update method is expressed as follows:

[0086]

[0087] Among them, ξ 10 ,ξ 11 Represents a random value in the interval [0,1].

[0088] S35, based on the individual positions updated in steps S33-S34, determine whether the global optimal solution is updated, if not updated, proceed to step S36, if updated, proceed to step S37;

[0089] Update the global optimal solution if It means that the global optimal solution of this iteration has been updated. Go to step S36; if This means that the global optimal solution of this iteration has not been updated, and the process goes to step S37.

[0090] S36, performing a taboo search in the neighborhood of the individual's current position, finding the optimal solution in the neighborhood and comparing it with the global optimal solution;

[0091] Perform taboo search at the individual's current position. First, generate a neighborhood solution, the expression is as follows:

[0092]

[0093]

[0094] in, represents the neighborhood solution set, the subscript NB represents the number of neighborhood solutions, and w b represents the step size coefficient, A vector of random values ​​in the range [-1,1].

[0095] Then eliminate the solutions in the taboo table, update the global optimal solution and the taboo table, find the optimal solution among the remaining solutions, and compare it with the global optimal solution, that is, set the number of iterations t=t+1, and return to step S31.

[0096] S37. Based on the updated global optimal solution outputted in step S35, determine whether the maximum number of iterations is met. If not, set t=t+1 and return to step S31 until the maximum number of iterations is met. The iteration ends and the global optimal path is outputted.

[0097] Beneficial effects of the present invention: The present invention discloses a method for UAV path planning based on global search and metaheuristic optimization. First, a UAV path planning model under complex constraints is established, and the fitness function and its constraints are established. Then, the algorithm initialization parameters are set and the population position is initialized by chaotic mapping. The metaheuristic optimization algorithm is used for cyclic iteration to solve the problem, and the global optimal path is output. Finally, the global optimal path is implemented on the UAV, and the flight distance corresponding to the optimal path is calculated to complete the UAV path planning. The method of the present invention combines the metaheuristic algorithm-Crowned Porcupine Algorithm with taboo search to enhance the global search capability, and at the same time combines with chaotic mapping to optimize the population initialization, increase randomness and unpredictability, so that the algorithm can converge faster, and has obvious advantages when facing complex environment and complex constraint path planning problems, and solves the path planning problem of UAVs in three-dimensional complex mountainous environments. Compared with the existing methods, the method of the present invention introduces a newer and better-performing crested porcupine algorithm. When designing the fitness function, not only the path length is considered, but also the height penalty, yaw angle penalty and pitch angle penalty are added, which is more in line with the actual task requirements; the taboo search algorithm is introduced and integrated with the crested porcupine algorithm to generate neighborhood solutions, and the taboo table is maintained to explore new solutions, which effectively enhances the algorithm's global search capability. The taboo search performance depends on the initial value, so the chaotic mapping is introduced to optimize the population initialization, increase the randomness of the initial population, and speed up the algorithm convergence speed; the Bezier curve is introduced to replace the existing cubic spline interpolation curve. When facing a complex mountain environment, the selection of control points is more free, and the scope of the solution space is further expanded. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] Figure 1The present invention is a flow chart of a UAV path planning method based on global search and meta-heuristic optimization.

[0099] Figure 2 The flowchart of using the meta-heuristic optimization algorithm to perform cyclic iterative solution in an embodiment of the present invention.

[0100] Figure 3 It is a task scenario diagram in an embodiment of the present invention.

[0101] Figure 4 This is a diagram of the optimal path planning result in an embodiment of the present invention.

[0102] Figure 5 It is an iteration curve diagram of the algorithm fitness function value in an embodiment of the present invention. DETAILED DESCRIPTION

[0103] The method of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0104] like Figure 1 As shown, a UAV path planning method based on global search and meta-heuristic optimization of the present invention comprises the following steps:

[0105] S1. Establish a UAV path planning model under complex constraints, that is, establish a three-dimensional mountain model based on function representation, and establish the fitness function and its constraints;

[0106] S2, set the algorithm initialization parameters and initialize the population position by chaos mapping;

[0107] The initialization parameters include: initial population size, minimum population size, maximum number of iterations, number of cyclic population cycles, convergence rate and balance coefficient.

[0108] S3, using a meta-heuristic optimization algorithm (Crowned Porcupine Algorithm combined with Taboo Search) to perform cyclic iterative solution on the path planning model in step S1;

[0109] like Figure 2 As shown in the figure, the problem-solving parameters and iteration end conditions are set; the global optimal solutions in the solution space, that is, the optimal positions, are recorded; the population size is updated using a cyclic population reduction strategy, and the four different defensive behaviors of crested porcupines facing predators in a natural environment are simulated to establish an optimization iteration method; the global optimal solution is updated after one cycle; after traversing the population, it is determined whether the global optimal solution is updated. If the global optimal solution is not updated, a taboo search is performed in the neighborhood of the current position to find the optimal solution in the neighborhood and compare it with the global optimal value; until the iteration end condition is met, the optimization is terminated and the global optimal path is output.

[0110] S4. Based on step S3, the global optimal path is implemented on the UAV, and the flight distance corresponding to the optimal path, that is, the fitness function value corresponding to the optimal path, is calculated to complete the UAV path planning.

[0111] In this embodiment, the step S1 is specifically as follows:

[0112] S11, setting the UAV flight environment to a rectangular space in a Cartesian coordinate system;

[0113] S12, considering the mountain terrain environment and obstacle distribution, the environment is modeled by combining the reference terrain model and the exponential function, that is, a three-dimensional mountain model is established based on the function representation method;

[0114] The expression of the three-dimensional mountain model based on the function representation is as follows:

[0115]

[0116] Among them, Z(x,y) represents the height corresponding to the point (x,y), M represents the number of peaks, and H i represents the height control parameter, X si and Y si Represents the attenuation parameter in x,y coordinates, (x i ,y i ) represents the center coordinates of the i-th peak.

[0117] S13, determining the constraint conditions of the UAV path planning model;

[0118] Between the control points, multiple path points are generated by the Bezier curve interpolation method, and a set of control point sequences is set as {S,Q 1 ,Q 2 ,...,Q n-1 ,G}.

[0119] Among them, the control points are ordered three-dimensional coordinate points, indicating the basic direction of path search; S and G represent the starting point and the end point respectively, and Q 1 To Q n-1 Represents an intermediate node of a path.

[0120] Then use the above n+1 nodes to generate a smooth Bezier curve, the expression is as follows:

[0121]

[0122]

[0123] Among them, τ represents a one-dimensional sequence of equally spaced points from 0 to 1, B i,n(τ) represents the Bezier curve between the i-th control point and the i+1-th control point, P i Indicates {Q 1 ,Q 2 ,...,Q n-1 , a control point in G}, and P(τ) represents a path point.

[0124] The constraints of the UAV path planning model are as follows:

[0125] (1) In order to avoid collision with obstacles, the height of the drone should always be higher than the ground. The expression is as follows:

[0126] z j ≥Z(x j ,y j )

[0127] Among them, z j represents the height of the drone at the jth path point, Z(x j ,y j ) represents the peak height of the jth path point.

[0128] (2) The drone can only fly within a certain area and cannot exceed the environmental boundary. The expression is as follows:

[0129] x min ≤x j ≤x max

[0130] Among them, x min and x max Indicates the minimum and maximum values ​​of the horizontal coordinate x in the flight space, and the same applies to y and z.

[0131] (3) The drone should fly within an appropriate flight altitude. Flying too high or too low will be subject to a height penalty, as shown in the following expression:

[0132]

[0133] Among them, f h Indicates high penalty, h min and h max represents the upper and lower bounds of reasonable flight altitude, w h Represents the height penalty factor.

[0134] (4) Actual flight requires the path to be as smooth as possible. Excessive yaw and pitch angles will be penalized. The expression is as follows:

[0135]

[0136]

[0137] Among them, f yaw and f pit denote the yaw angle penalty and pitch angle penalty respectively, ψ j and θ j denote the yaw angle and pitch angle formed by the j-th path point and the j+1-th path point, respectively, and ψ max and θ max Respectively represent the maximum yaw angle and the maximum pitch angle, w yaw and w pit Represent the penalty factors for yaw angle and pitch angle respectively.

[0138] S14, establishing a fitness function based on the constraints determined in step S13;

[0139] The fitness function f is expressed as follows:

[0140] f=min(f L +f yaw +f pit +f h )

[0141] Among them, f L Represents the UAV path length.

[0142] In this embodiment, step S2 is specifically as follows:

[0143] S21. Define the initial population size as N and the minimum population size as N min , the number of cyclic population cycles T, the maximum number of iterations t max , convergence rate α and balance coefficient T f ;

[0144] S22, chaos mapping initializes population position;

[0145] First, the population position is randomly generated according to the uniform distribution. The expression is as follows:

[0146]

[0147] in, Represents a population individual, a feasible solution in the solution space, that is, a feasible set of control points. A set of control point positions is taken as an individual, and the optimal solution is found through the optimization strategy as the final control point position set; and express

[0148] →

[0149] The lower and upper bounds of the solution space, r represents a random vector uniformly distributed in the interval [0,1]. The initial population is represented in matrix form, that is,

[0150] For the initialized population Use the Singer chaotic mapping strategy to initialize the population again. The expression is as follows:

[0151]

[0152] Where a represents the chaotic mapping control parameter, and a∈(0.9,1.08), Represents the updated initial population position.

[0153] like Figure 2 As shown, in this embodiment, step S3 is specifically as follows:

[0154] S31, updating the population size based on the cyclic population reduction strategy;

[0155] Calculate the current optimal position in the environment solution space when the current iteration number t is reached The expression is as follows:

[0156]

[0157] in, Represents the individual position that minimizes the objective function value at time t.

[0158] According to the current population size N t , based on the cyclic population reduction strategy, the population size is updated, and the update expression is as follows:

[0159]

[0160] S32, traversing each crested porcupine individual in the population, each individual randomly performs exploration or development, that is, calculating individual defense parameters and attack parameters;

[0161] Among them, exploration includes two behaviors: visual defense and sound defense, and development includes two behaviors: odor attack and physical attack.

[0162] Before exploration or exploitation, each individual parameter that needs to be updated includes: Odor diffusion rate Predator quality The average force of the predator on the crested porcupine is expressed as follows:

[0163]

[0164]

[0165]

[0166] in, represents the objective function value of the i-th individual at time t, ε 1 , ε 2 It means to avoid small values ​​with denominator of 0; represents a random vector from 0 to 1, represents the initial velocity of the ith individual at time t, represents the final velocity of the i-th individual at the next iteration t+1, Indicates the position of a crested porcupine individual randomly selected at the current iteration number, and assigns it to And 1≤r≤N t , t represents the number of current iterations.

[0167] Then generate two uniformly distributed random numbers in the interval (0,1), ξ 1 ,ξ 2 , when 1 <ξ 2 When ξ 1 ≥ξ 2 Execute development and enter step S34.

[0168] S33. Based on step S32, generate two uniformly distributed random numbers, ξ 3 ,ξ 4 , when 3 <ξ 4 When ξ 3 ≥ξ 4 When , the second defense strategy, namely sound defense, is executed, and the individual position is updated according to the defense strategy;

[0169] 1) Visual defense;

[0170] When the crested porcupine is aware of a predator, it will raise its spikes to give the predator a greater impression, and the predator will either move toward it or away from it. To simulate this behavior, the position update expression is as follows:

[0171]

[0172] Among them, ξ 5 represents a random number based on normal distribution, ξ 6 Represents a random value in the interval [0,1]; represents the vector generated between the current crested porcupine and a randomly selected crested porcupine from the population, which is used to represent the position of the predator at time t.

[0173] 2) Sound defense;

[0174] Crested porcupines make noises to threaten predators, and their noises become louder as the predator approaches. To simulate the sound defense behavior, the position is updated as follows:

[0175]

[0176] Among them, ξ 7 represents a random value in the interval [0,1], r 1 and r 2 represents a random integer between [1, N], Represents a randomly generated binary vector of 0 and 1.

[0177] S34: Based on step S32, generate a random number ξ uniformly distributed between (0, 1) 8 , when 8 <T f When ξ 8 ≥T f When , the defense strategy 4, i.e., physical attack, is executed, and the individual position is updated according to the defense strategy;

[0178] 3) Odor attack;

[0179] The crested porcupine will secrete a foul smell that spreads around it to prevent predators from approaching it. To simulate this behavior, the position update method is expressed as follows:

[0180]

[0181] Among them, r 3 represents a random integer between [1, N], represents a parameter vector used to control the search direction, with equal probability of selecting all-one or all-negative-one vectors. represents the defense factor, ξ 9 Represents a random value in the interval [0,1].

[0182] 4) physical assault;

[0183] When the predator is very close, the crested porcupine uses its quills to physically attack, which is a one-dimensional inelastic collision. To simulate the physical attack behavior, the position update method is expressed as follows:

[0184]

[0185] Among them, ξ 10 ,ξ 11 Represents a random value in the interval [0,1].

[0186] S35, based on the individual positions updated in steps S33-S34, determine whether the global optimal solution is updated, if not updated, proceed to step S36, if updated, proceed to step S37;

[0187] Update the global optimal solution if It means that the global optimal solution of this iteration has been updated. Go to step S36; if This means that the global optimal solution of this iteration has not been updated, and the process goes to step S37.

[0188] S36, performing a taboo search in the neighborhood of the individual's current position, finding the optimal solution in the neighborhood and comparing it with the global optimal solution;

[0189] Perform taboo search at the individual's current position. First, generate a neighborhood solution, the expression is as follows:

[0190]

[0191]

[0192] in, represents the neighborhood solution set, the subscript NB represents the number of neighborhood solutions, and w b represents the step size coefficient, A vector of random values ​​in the range [-1,1].

[0193] Then eliminate the solutions in the taboo table, update the global optimal solution and the taboo table, find the optimal solution among the remaining solutions, and compare it with the global optimal solution, that is, set the number of iterations t=t+1, and return to step S31.

[0194] S37. Based on the updated global optimal solution outputted in step S35, determine whether the maximum number of iterations is met. If not, set t=t+1 and return to step S31 until the maximum number of iterations is met. The iteration ends and the global optimal path is outputted.

[0195] like Figure 3 As shown, in this embodiment, the mission scenario is a mountain space of 200km×200km×6km, the number of mountain peaks M is 30, the coordinates of the starting point and the end point are (20km, 165km, 2km), (170km, 26km, 3km), the number of control points is 4, the number of equally spaced sequences is 200, and the reasonable flight altitude h min and h max They are 2km and 3.5km respectively, and the height penalty factor is w h =0.4,ψ max =π / 3 and θ max =π / 6, the penalty factors for yaw and pitch angles are w yaw and w pit0.6 and 0.55 respectively.

[0196] In this embodiment, the initial population size N is set to 70, and the minimum population size N min =50, the number of cyclic population cycles T = 2, the maximum number of iterations t max = 200, convergence rate α = 0.02 and balance coefficient T f =0.8, the chaotic mapping parameter a is set to 1.05, the number of neighborhood solutions NB = 10, w b =10, the length of the taboo list is 50.

[0197] according to Figure 3 The task scenario shown and the above parameter settings, the optimal path result is calculated as follows Figure 4 As shown, the algorithm converges as follows Figure 5 As shown in the figure, it can be seen that the path obtained by the algorithm is a valid flight path, and the algorithm also converges to the global optimum. The simulation results confirm that the UAV path planning method based on global search and meta-heuristic optimization can solve the optimal path of the UAV path planning problem.

[0198] In summary, the method of the present invention is solved by establishing mountain terrain and environmental constraints, and then designing a global search and meta-heuristic optimization algorithm, combining the meta-heuristic algorithm-Crown Porcupine Algorithm with taboo search, enhancing the global search capability, and combining chaos mapping to optimize the population initialization, increasing randomness and unpredictability, so that the algorithm can converge faster. Therefore, the method of the present invention has obvious advantages when facing complex environments and complex constraint path planning problems. Compared with existing A* algorithms, ant colony algorithms, particle swarm algorithms and other methods, the method of the present invention introduces a relatively new and better-performing Crown Porcupine Algorithm. When designing the fitness function, not only the path length is considered, but also the height penalty, yaw angle penalty and pitch angle penalty are added, which is more in line with the actual task requirements. In view of the problem that the Crown Porcupine Algorithm itself has weak global search capability, the method of the present invention introduces the taboo search algorithm and the Crown Porcupine Algorithm to merge, and effectively enhance the algorithm global search capability by generating neighborhood solutions and maintaining the taboo table to explore new solutions. Taboo search performance depends on the initial value, so chaotic mapping is introduced to optimize population initialization, increase the randomness of the initial population, and accelerate the convergence of the algorithm. The path between control points is smooth, and the method of the present invention introduces Bezier curves to replace the existing cubic spline interpolation curves. When facing complex mountain environments, the selection of control points is more free, and the scope of the solution space is further expanded.

[0199] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A UAV path planning method based on global search and meta-heuristic optimization, the specific steps are as follows: S1. Establish a UAV path planning model under complex constraints, that is, establish a three-dimensional mountain model based on function representation, and establish the fitness function and its constraints; S2, set the algorithm initialization parameters and initialize the population position by chaos mapping; in, The initialization parameters include: initial population size, minimum population size, maximum number of iterations, number of cyclic population cycles, convergence rate and balance coefficient; S3, using a meta-heuristic optimization algorithm to perform cyclic iterative solution on the path planning model in step S1; Set the problem-solving parameters and iteration end conditions; record the global optimal solutions in the solution space, that is, the optimal position; use the cyclic population reduction strategy to update the population size, and simulate the four different defensive behaviors of crested porcupines facing predators in the natural environment to establish an optimization iteration method; update the global optimal solution after one cycle; after traversing the population, determine whether the global optimal solution is updated. If the global optimal solution is not updated, perform a taboo search in the neighborhood of the current position to find the optimal solution in the neighborhood and compare it with the global optimal value; until the iteration end condition is met, end the optimization and output the global optimal path; S4. Based on step S3, the global optimal path is implemented on the UAV, and the flight distance corresponding to the optimal path, that is, the fitness function value corresponding to the optimal path, is calculated to complete the UAV path planning.

2. The method for UAV path planning based on global search and meta-heuristic optimization according to claim 1, characterized in that: The step S1 is specifically as follows: S11, setting the UAV flight environment to a rectangular space in a Cartesian coordinate system; S12, considering the mountain terrain environment and obstacle distribution, the environment is modeled by combining the reference terrain model and the exponential function, that is, a three-dimensional mountain model is established based on the function representation method; The expression of the three-dimensional mountain model based on the function representation is as follows: Among them, Z(x,y) represents the height corresponding to the point (x,y), M represents the number of peaks, and H i represents the height control parameter, X si and Y si Represents the attenuation parameter in x,y coordinates, (x i ,y i ) represents the center coordinate of the i-th peak; S13, determining the constraint conditions of the UAV path planning model; Between the control points, multiple path points are generated by the Bezier curve interpolation method, and a set of control point sequences is set as {S, Q1, Q2, ..., Q n-1 ,G}; Among them, the control points are ordered three-dimensional coordinate points, indicating the basic direction of path search; S and G represent the starting point and the end point respectively, Q1 to Q n-1 Represents the intermediate node of the path; Then use the above n+1 nodes to generate a smooth Bezier curve, the expression is as follows: Among them, τ represents a one-dimensional sequence of equally spaced points from 0 to 1, B i,n (τ) represents the Bezier curve between the i-th control point and the i+1-th control point, P i represents {Q1,Q2,...,Q n-1 , a control point in G}, P(τ) represents a path point; The constraints of the UAV path planning model are as follows: (1) In order to avoid collision with obstacles, the height of the drone should always be higher than the ground. The expression is as follows: z j ≥Z(x j ,y j ) Among them, z j represents the height of the drone at the jth path point, Z(x j ,y j ) represents the peak height of the jth path point; (2) The drone can only fly within a certain area and cannot exceed the environmental boundary. The expression is as follows: x min ≤x j ≤x max Among them, x min and x max Indicates the minimum and maximum values ​​of the horizontal coordinate x in the flight space, and the same applies to y and z; (3) The drone should fly within an appropriate flight altitude. Flying too high or too low will be subject to a height penalty, as shown in the following expression: Among them, f h Indicates high penalty, h min and h max represents the upper and lower bounds of reasonable flight altitude, w h represents the height penalty factor; (4) Actual flight requires the path to be as smooth as possible. Excessive yaw and pitch angles will be penalized. The expression is as follows: Among them, f yaw and f pit denote the yaw angle penalty and pitch angle penalty respectively, ψ j and θ j denote the yaw angle and pitch angle formed by the j-th path point and the j+1-th path point, respectively, and ψ max and θ max Respectively represent the maximum yaw angle and the maximum pitch angle, w yaw and w pit Represent the penalty factors for yaw angle and pitch angle respectively; S14, establishing a fitness function based on the constraints determined in step S13; The fitness function f is expressed as follows: f=min(f L +f yaw +f pit +f h ) Among them, f L Represents the length of the drone path.

3. The method for UAV path planning based on global search and meta-heuristic optimization according to claim 1, characterized in that: The step S2 is specifically as follows: S21. Define the initial population size as N and the minimum population size as N min , the number of cyclic population cycles T, the maximum number of iterations t max , convergence rate α and balance coefficient T f ; S22, chaos mapping initializes population position; First, the population position is randomly generated according to uniform distribution, and the expression is as follows: in, Represents a population individual, a feasible solution in the solution space, that is, a feasible set of control points. A set of control point positions is taken as an individual, and the optimal solution is found through the optimization strategy as the final control point position set; and represents the lower and upper bounds of the solution space, represents a random vector uniformly distributed in the interval [0,1]; the initial population is represented in matrix form, that is, For the initialized population Use the Singer chaotic mapping strategy to initialize the population again. The expression is as follows: Where a represents the chaotic mapping control parameter, and a∈(0.9,1.08), Represents the updated initial population position.

4. The method for UAV path planning based on global search and meta-heuristic optimization according to claim 1, characterized in that: The step S3 is specifically as follows: S31, updating the population size based on the cyclic population reduction strategy; Calculate the current optimal position in the environment solution space when the current iteration number t is reached The expression is as follows: in, represents the individual position that minimizes the objective function value at time t; According to the current population size N t , based on the cyclic population reduction strategy, the population size is updated, and the update expression is as follows: S32, traversing each crested porcupine individual in the population, each individual randomly performs exploration or development, that is, calculating individual defense parameters and attack parameters; Among them, exploration includes two behaviors: visual defense and sound defense, and development includes two behaviors: odor attack and physical attack; Before exploration or exploitation, each individual parameter that needs to be updated includes: Odor diffusion rate Predator quality The average force of the predator on the crested porcupine is expressed as follows: in, represents the objective function value of the i-th individual at time t, ε1 and ε2 represent small values ​​that avoid the denominator being 0; represents a random vector from 0 to 1, represents the initial velocity of the ith individual at time t, represents the final velocity of the i-th individual at the next iteration t+1, Indicates the position of a crested porcupine individual randomly selected at the current iteration number, and assigns it to And 1≤r≤N t , t represents the number of current iterations; Then, two random numbers uniformly distributed in the interval (0,1) are generated, ξ1, ξ2. When ξ1<ξ2, exploration is performed and the process goes to step S33; when ξ1≥ξ2, development is performed and the process goes to step S34; S33, based on step S32, generate two uniformly distributed random numbers, ξ3, ξ4, when ξ3<ξ4, execute defense strategy one, i.e. visual defense, when ξ3≥ξ4, execute defense strategy two, i.e. sound defense, and update the individual position according to the defense strategy; 1) Visual defense; To simulate visual defense behavior, the position update method is expressed as follows: Among them, ξ5 represents a random number based on normal distribution, and ξ6 represents a random value in the interval [0,1]; represents the vector generated between the current crested porcupine and a randomly selected crested porcupine from the population, which is used to represent the position of the predator at time t; 2) Sound defense; Simulate sound defense behavior, and update the position as follows: Among them, ξ7 represents a random value in the interval [0,1], r1 and r2 represent random integers between [1,N], Represents a randomly generated binary vector of 0 and 1; S34, based on step S32, generate a random number ξ8 uniformly distributed between (0,1), when ξ8<T f When ξ8≥T f When , the defense strategy 4, i.e., physical attack, is executed, and the individual position is updated according to the defense strategy; 3) Odor attack; To simulate the odor attack behavior, the position update method is expressed as follows: Among them, r3 represents a random integer between [1, N], represents a parameter vector used to control the search direction, with equal probability of selecting all-one or all-negative-one vectors. represents the defense factor, ξ9 represents a random value in the interval [0,1]; 4) physical assault; To simulate physical attack behavior, the position update method is expressed as follows: Among them, ξ 10 ,ξ 11 Represents a random value in the interval [0,1]; S35, based on the individual positions updated in steps S33-S34, determine whether the global optimal solution is updated, if not updated, proceed to step S36, if updated, proceed to step S37; Update the global optimal solution if It means that the global optimal solution of this iteration has been updated. Go to step S36; if This means that the global optimal solution of this iteration has not been updated, and the process goes to step S37; S36, performing a taboo search in the neighborhood of the individual's current position, finding the optimal solution in the neighborhood and comparing it with the global optimal solution; Perform taboo search at the individual's current position; first generate a neighborhood solution, the expression is as follows: in, represents the neighborhood solution set, the subscript NB represents the number of neighborhood solutions, and w b represents the step size coefficient, Represents a vector of random values ​​in [-1,1]; Then, the solutions in the taboo table are eliminated, the global optimal solution and the taboo table are updated, the optimal solution is found among the remaining solutions, and then compared with the global optimal solution, that is, the number of iterations t=t+1, and the process returns to step S31; S37. Based on the updated global optimal solution outputted in step S35, determine whether the maximum number of iterations is met. If not, set t=t+1 and return to step S31 until the maximum number of iterations is met. The iteration ends and the global optimal path is outputted.

Images