A UAV 3D Path Planning Method Based on Multi-Strategy Improved Lemming Optimization Algorithm

Abstract

This invention discloses a UAV 3D path planning method based on a multi-strategy improved lemming optimization algorithm, comprising the following steps: S1, constructing a 3D task space model; S2, constructing a flight cost function; S3, initializing the population using Latin hypercube sampling; S4, iterative global exploration and local exploitation with adaptive parameters; S5, updating the population using a bidirectional population evolution strategy and a memory mechanism; S6, evaluating fitness and selecting the current optimal individual; S7, iterative judgment and outputting the optimal path. This invention improves upon the UAV 3D path planning problem in complex environments by employing multi-strategy collaborative improvement, using Latin hypercube sampling to initialize the population, designing an adaptive parameter adjustment strategy to balance exploration and exploitation, and introducing a bidirectional population evolution strategy and a memory mechanism to enhance optimization performance. Through boundary reflection processing and flight cost function evaluation, the optimal flight path is obtained from the population, thus effectively solving the UAV 3D path planning problem in complex environments.

Description

Technical Field

[0001] This invention relates to the field of path planning technology, and in particular to a UAV 3D path planning method based on a multi-strategy improved lemming optimization algorithm. Background Technology

[0002] Unmanned Aerial Vehicles (UAVs), as a new technology platform characterized by high mobility, flexible deployment, and low cost, have demonstrated broad application potential in the civilian sector and are widely used in scenarios such as logistics delivery, agricultural monitoring, emergency rescue, and facility inspection. Therefore, planning a safe and reliable flight path for UAVs in complex environments has become a key technological foundation for ensuring the successful execution of their missions.

[0003] Currently, common UAV path planning methods mainly include grid-based methods, free-space methods, and neural network methods. Among them, grid-based methods usually need to be combined with various intelligent optimization algorithms to search for feasible paths in a discrete grid space. In recent years, the use of metaheuristic optimization algorithms for path planning has become an important research direction. These algorithms simulate natural phenomena or swarm intelligence behavior to perform efficient searches in a continuous solution space, and can better handle multi-dimensional and multi-constrained path optimization problems.

[0004] The Lemming Optimization Algorithm (ALA) is a recently proposed metaheuristic algorithm that simulates the migration and foraging behaviors of lemming populations. In initial applications, the algorithm demonstrated good stability and local exploitation capabilities. However, when directly applied to the specific engineering problem of UAV path planning in complex three-dimensional space, the algorithm still exhibits several significant limitations, restricting its practical performance.

[0005] 1) Insufficient convergence speed and optimization efficiency: When dealing with high-dimensional planning problems with complex terrain and multiple obstacle constraints, the search mechanism of the basic lemming algorithm is difficult to quickly approach the potential solution region in the early stage of iteration, resulting in a slow convergence speed and an overall optimization efficiency that needs to be improved.

[0006] 2) Prone to getting trapped in local optima: The algorithm's balance mechanism between development and exploration is relatively fixed and lacks adaptive adjustment capabilities. In complex and multi-peaked solution spaces, it is easy to get trapped in local optima too early, making it difficult to achieve a full search of the global scope, thus limiting the quality of the final path.

[0007] 3) Limited universality and robustness: When faced with task scenarios of different scales and constraints, the algorithm's parameter settings and search strategies are not adaptable enough, resulting in large performance fluctuations and a need to strengthen its robustness.

[0008] Therefore, there is an urgent need to provide a UAV 3D path planning method based on a multi-strategy improved lemming optimization algorithm to address the shortcomings of the existing technologies. Summary of the Invention

[0009] To address the limitations of the lemming algorithm in UAV 3D path planning, including limited universality, insufficient convergence speed, and susceptibility to local optima, this invention provides a UAV 3D path planning method based on a multi-strategy improved lemming optimization algorithm. This method establishes a global path optimization function with the goal of minimizing flight cost using an improved lemming algorithm to solve for the optimal UAV flight path. This allows the improved lemming algorithm to balance optimization quality, robustness, and convergence speed, while ensuring the feasibility of UAV flight path planning.

[0010] The technical solution of this invention is implemented as follows: a UAV 3D path planning method based on a multi-strategy improved lemming optimization algorithm, comprising:

[0011] S1. Obtain real-world environmental data, construct a three-dimensional mission space model, and determine the starting and ending coordinates of the UAV's flight path;

[0012] S2. Construct the flight cost function for the flight mission based on the UAV parameters and the three-dimensional mission space model;

[0013] S3. Initialize the lemming population using Latin hypercube sampling;

[0014] S4. Based on the initialized lemming population, perform global exploration and local development iterations of adaptive parameters in sequence;

[0015] S5. Based on the iteratively updated lemming population, a two-way population evolution strategy and a memory mechanism are used to update the lemming population;

[0016] S6. Based on the UAV flight paths planned by all lemming individuals, calculate the corresponding flight cost using the flight cost function, obtain the fitness of all individuals in the population, and select the individual with the lowest fitness as the optimal lemming individual.

[0017] S7. Determine if the current iteration count has reached the maximum iteration count; if it has, end the path optimization and output the optimal path; if it has not, re-simulate the lemming population iteration and continue to optimize the path.

[0018] This invention constructs a task space model and a flight cost function, introduces Latin hypercube sampling to initialize the lemming population, and then sequentially performs global exploration and local development iterations. It also incorporates a bidirectional population evolution strategy and a memory mechanism to update the lemming population. Finally, based on the flight cost function, the optimal lemming individual is obtained from the lemming population to obtain the optimal flight path of the UAV, effectively solving the path planning problem of UAVs in complex environments and improving the efficiency of path planning.

[0019] As a further improvement to the above solution, the construction of the three-dimensional task space model includes:

[0020] Based on the UAV flight mission, a three-dimensional terrain map is drawn in a spatial coordinate system, and non-collision obstacles are set in areas in the three-dimensional space scene that may pose a threat to the UAV flight mission, thereby obtaining the three-dimensional mission space model.

[0021] By constructing a three-dimensional task space model, the threat area is clearly identified, providing an accurate and reliable environmental foundation for subsequent path planning.

[0022] As a further improvement to the above scheme, the flight cost function for constructing the flight mission includes:

[0023] Based on the total length of the flight mission path, construct the flight length cost function for the UAV;

[0024] Based on the distance between each obstacle and the flight path in the three-dimensional mission space model, a threat avoidance cost function for the UAV is constructed.

[0025] Based on the altitude of each waypoint in the flight path, construct the flight altitude cost function of the UAV;

[0026] Based on the changes in the UAV's turning and climb angles along its flight path, a path smoothing cost function for the UAV is constructed.

[0027] By constructing cost functions for the length of UAVs, threat avoidance, flight altitude, and path smoothing, key factors in actual flight are transformed into quantifiable comprehensive evaluation criteria, ensuring that the planned paths meet the requirements of economy, safety, and flyability.

[0028] As a further improvement to the above scheme, the initialization of the lemming population using Latin hypercube sampling includes:

[0029] Set the relevant parameters for the lemming optimization algorithm; among them, the relevant parameters include: lemming population size and maximum number of iterations;

[0030] The position of each lemming is initialized by Latin hypercube sampling. It is then checked whether the position of each lemming is within the map boundary of the task space model. The fitness value generated by each lemming at the current position is calculated, and the lemming with the best current fitness is selected as the best.

[0031] By introducing Latin hypercube sampling and performing boundary detection and initial fitness assessment, the convergence speed was effectively improved.

[0032] As a further improvement to the above scheme, the Latin hypercube sampling model is as follows:

[0033] ;

[0034] in, For the initial lemming individual, and Let the j-th dimension be the upper and lower bounds. For a random permutation of the range 1 to N belonging to the j-th dimension, For population size, In the interval Random numbers on the array.

[0035] By using the Latin hypercube sampling method to initialize the lemming population, a batch of uniformly distributed initial candidate flight paths are generated for the entire population, which significantly improves the quality of the initial population and enhances the global exploration capability of the algorithm.

[0036] As a further improvement to the above scheme, the global exploration iteration based on adaptive parameters includes:

[0037] During the lemming's path-finding process, the migration and burrowing behavior of the lemming are simulated to establish a mathematical model, and a global exploration is conducted to select the next path point for the lemming to move.

[0038] The global exploration mathematical model based on adaptive parameters is as follows:

[0039] ;

[0040] in, and These represent the positions of the i-th search agent at the (t+1)-th and (t)-th iterations, respectively. and Let a and b represent two random individuals in the population; a and b are the ranges belonging to Integer index; This indicates the current optimal position; F represents randomly selecting one of the two search directions, positive or negative. This indicates that the elements are uniformly distributed in the interval. Vectors in the text; random numbers ;

[0041] The mathematical model for the adaptive exploration factor is as follows:

[0042] ;

[0043] in, and Here, represents the upper and lower bounds of the development factor; p represents the iteration progress factor. Where t is the current iteration number and T is the total number of iterations;

[0044] The mathematical model for the adaptive development factor is as follows:

[0045] ;

[0046] in, and These are the upper and lower limits of the development factor, respectively;

[0047] The mathematical model for a random number vector representing Brownian motion is:

[0048] ;

[0049] L represents a random number related to the current iteration number, and the mathematical model is:

[0050] ;

[0051] Here, rand is a random number between 0 and 1.

[0052] By introducing a global exploration mathematical model that incorporates adaptive exploration factors, the search behavior of the algorithm at different iteration stages is balanced, thereby enhancing the global exploration capability and diversity of the population.

[0053] As a further improvement to the above scheme, the local development iteration based on adaptive parameters includes:

[0054] During the lemming's search path, a mathematical model is established to simulate the lemming's foraging and predator avoidance behaviors, and local development is carried out to select the next path point for the lemming to move.

[0055] The mathematical model for local development based on adaptive parameters is as follows:

[0056] ;

[0057] in, and These represent the positions of the i-th search agent at the (t+1)-th and (t)-th iterations, respectively. Indicates the current optimal position; F represents randomly selecting one of the two search directions, positive or negative; random number. ;

[0058] The mathematical model for the adaptive development factor is as follows:

[0059] ;

[0060] in, and Here, represents the upper and lower bounds of the development factor; p represents the iteration progress factor. Where t is the current iteration number and T is the total number of iterations;

[0061] The Euclidean distance between the current position and the optimal solution is represented by the following mathematical model:

[0062] ;

[0063] Represents the dimension of a given problem; This indicates the position of the i-th search agent in the j-th dimension during the t-th iteration; This indicates that the current optimal search agent is at the . Vi in the The position of the next iteration;

[0064] G is the escape coefficient, and the mathematical model is as follows: Where t is the current iteration number; T is the total number of iterations; For Lévy's flight function.

[0065] By introducing a local development model based on adaptive parameters, the convergence accuracy of the algorithm in the potential optimal region is enhanced, effectively improving the quality of the final flight path.

[0066] As a further improvement to the above scheme, the use of a two-way population evolution strategy and memory mechanism to update the lemming population includes:

[0067] A simulated bidirectional population evolution strategy and memory mechanism are added to the end of the lemming position update to optimize the lemming population. The population is divided into a high-quality population with a proportion of u and a low-quality population with a proportion of v. Different strategies are used to update the high-quality population and the low-quality population.

[0068] Among them, u+v=1, and 0<u<v<1;

[0069] The memory mechanism is as follows: after updating the population position and simulating a two-way population evolution strategy, the fitness of the entire population is evaluated, and individuals with lower fitness are retained as the new generation of lemming population.

[0070] By introducing a two-way population evolution and memory mechanism, the algorithm's local development and global exploration capabilities are effectively balanced, improving the algorithm's robustness in escaping local optima and the quality of the final solution.

[0071] As a further improvement to the above scheme, the mathematical model for high-quality population renewal using the bidirectional population evolution strategy is as follows:

[0072] ;

[0073] in, and This refers to two high-quality individuals in a high-quality population that are different from the current individual. Indicates the current optimal position; This indicates the position of the i-th search agent in the (t+1)-th iteration; Let represent a discrete random variable taking the value 1 or 2, obtained through a random number generation function; W represents a mutation factor based on a sine function, with the mathematical model as follows: ,in, Let t represent the dimension of the given problem, t be the current iteration number, and T be the total number of iterations.

[0074] The mathematical model for low-quality population renewal under a two-way population evolution strategy is as follows:

[0075] ;

[0076] in, and These represent the positions of the i-th search agent at the (t+1)-th and (t)-th iterations, respectively. and These represent the upper and lower bounds of the population variable, respectively. Represents the random direction factor; random number .

[0077] By defining mathematical models for population updates of different qualities, a balance is effectively struck between depth mining and breadth search, improving the algorithm's ability to escape local optima and its overall convergence efficiency.

[0078] Compared with the prior art, the present invention achieves the following beneficial effects:

[0079] (1) Improve the quality and efficiency of the initial search: By introducing Latin hypercube sampling for population initialization, uniform sampling is performed in all dimensions of the solution space, which improves the diversity and spatial coverage of the initial population and lays the foundation for subsequent optimization.

[0080] (2) Achieving a dynamic balance in the search process: By designing new adaptive development factor parameters and adaptive exploration factor parameters in the exploration and development stages of the algorithm, the algorithm’s development or exploration capabilities are emphasized at different stages of iteration, effectively balancing the development and exploration capabilities of the lemming algorithm, and effectively solving the problem that the traditional algorithm’s exploration and development capabilities are fixed and difficult to balance.

[0081] (3) Significantly enhances the ability to escape local optima: A two-way population evolution strategy and memory mechanism are introduced at the end of the position update. By implementing differentiated updates for populations of different quality and retaining better solutions from previous generations, the algorithm is forced to thoroughly explore the solution space, thereby enhancing its ability to escape local optima and approach the global optimal solution. Attached Figure Description

[0082] Figure 1 This is a flowchart of a UAV 3D path planning method based on a multi-strategy improved lemming optimization algorithm according to an embodiment of the present invention;

[0083] Figure 2 This is a three-dimensional image of the UAV path planning result in an embodiment of the present invention;

[0084] Figure 3 The image shown is a right view of the UAV path planning result according to an embodiment of the present invention.

[0085] Figure 4 The graph shows a comparison of the convergence of the improved lemming algorithm and the unimproved lemming algorithm under a composite test function, as provided in the embodiments of the present invention. Detailed Implementation

[0086] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0087] Example:

[0088] like Figure 1 As shown, this embodiment proposes a UAV 3D path planning method based on a multi-strategy improved lemming optimization algorithm, including:

[0089] S1. Acquire real-world environmental data and construct a 3D mission space model. Construct a mission space model based on the UAV flight mission scenario, and set the UAV's start and end coordinates within the mission space model. In this embodiment, the UAV flight path consists of a starting point and ten waypoints. The UAV's starting coordinates in the mission space are set to... The drone's endpoint coordinates are set to Define the effective range of the x and y axes in the coordinate space to ensure that the UAV path planning coordinates do not exceed the mission space boundaries. Set the map size and generate grid coordinates for terrain visualization. Simultaneously, set the number of threats to 6. Integrate key parameters such as start point, end point, map size, coordinate range, grid coordinates, waypoint coordinates, terrain data, and threats to generate a realistic 3D visualization map for path planning.

[0090] S2. Based on the UAV parameters and mission space environment, construct the UAV flight cost function. This function must comprehensively consider the UAV's flight range cost, threat avoidance cost, altitude constraint cost, and flight path smoothing cost. Calculate the flight cost of the planned path based on the flight cost function; a lower flight cost indicates better performance of the planned UAV flight path.

[0091] In this embodiment, the height constraint range of the threatening object is preferably... ;

[0092] The polar coordinate radius constraint is: ,in, The coordinates of the drone's starting point. The coordinates of the drone's destination; The number of waypoints is set to ten in this embodiment;

[0093] The preferred polar angle constraint range is... This ensures the smoothness of the segmented path and avoids excessive turning angles in the drone's flight path;

[0094] The azimuth constraint range is: This ensures that the overall direction of the drone's path is towards the destination;

[0095] In this embodiment, the flight cost cost function mainly consists of four parts: the cost of the UAV's flight range, the cost of avoiding threats, the cost of the UAV's flight altitude constraint, and the cost of smoothing the UAV's flight path.

[0096] The specific mathematical expression for the cost of drone flight range is:

[0097] ;

[0098] in, The number of points on the complete flight path, including the origin, waypoints, and destination. and These are the x-coordinates of the current path point and the next path point, respectively. and These are the y-coordinates of the current path point and the next path point, respectively. and These are the z-coordinates of the current path point and the next path point, respectively.

[0099] The specific mathematical expression for the cost of avoiding threats is:

[0100] ;

[0101] in, This represents the total number of threats, which is set to six in this embodiment. The total number of points along the complete flight path, including the origin, waypoints, and destination; the threat radius of the threat object is set to... , Represents the size of the drone. Represents the distance the danger zone extends. Threat center representing current threats The shortest distance to the current path segment is calculated using the following mathematical formula:

[0102] ;

[0103] Where t is the projection of the threat's center point onto the current path segment: ;when Time represents the distance between the threatening object and the endpoint of the path segment. Closer, when Time represents the distance between the threatening object and the endpoint of the path segment. Closer, when The time indicates that the threat is inside the path segment, and the shortest distance is the length of the perpendicular line segment from the threat to the path segment; The coordinates of the center of the threat object;

[0104] The specific mathematical expression for the cost constrained by the flight altitude of a drone is as follows:

[0105] ;

[0106] in, The number of points on the complete flight path, including the origin, waypoints, and destination. The z-coordinate of the current path point; and These represent the minimum / maximum relative height allowed by the terrain, respectively.

[0107] The specific mathematical expression for the cost of smoothing the flight path of a drone is:

[0108] ;

[0109] in, The number of points on the complete flight path, including the origin, waypoints, and destination. For the drone's turning angle, The drone's climb angle;

[0110] ;

[0111] ;

[0112] ;

[0113] ;

[0114] in, This is the horizontal projection vector of the first path segment in an adjacent path segment; This is the horizontal projection vector of the second path segment; These are the coordinates of the starting endpoint of the first path segment in an adjacent path segment. The coordinates of the end point of the first path segment are the same as the coordinates of the start point of the second path segment. The coordinates of the end point of the second path segment.

[0115] The cost function for drone flight is the weighted sum of the four cost functions mentioned above, and the mathematical formula is shown below:

[0116] ;

[0117] in, , , , These correspond to the weights of the four cost functions, which are set to 10, 100, 10, and 50 respectively in this embodiment.

[0118] S3. Set relevant parameters for the lemming optimization algorithm, including the lemming population size. Maximum number of iterations, etc.

[0119] The location of each lemming is initialized using Latin hypercube sampling. It is then checked whether each lemming is within the map boundaries, and the fitness value of each lemming at its current location is calculated. The lemming with the best current fitness is selected as the current best lemming. The mathematical formula for Latin hypercube sampling is shown below:

[0120] ;

[0121] in, For the initial lemming individual, and Let the j-th dimension be the upper and lower bounds. For a random permutation of the range 1 to N belonging to the j-th dimension, For population size, In the interval Random numbers;

[0122] The initial population position can be represented in the following matrix form:

[0123] ;

[0124] Where pop represents the lemming population size; The dimension of the given problem.

[0125] S4. Based on the initialized lemming population, global exploration and local development iterations of adaptive parameters are carried out in sequence. During the lemming search path process, mathematical models are established by simulating the lemming's migration, burrowing, foraging and predator avoidance behaviors. The next path point for the lemming is selected. After global exploration and local development, the current lemming fitness is calculated. If the fitness is better than before the move, the data before the move is replaced and retained.

[0126] The mathematical model for improved lemming migration and burrowing behavior based on adaptive parameters is as follows:

[0127] ;

[0128] in, and They represent the first The search agent in the 1st and the The position of the next iteration. and Let each represent a random individual in the population. and The scope belongs to Integer index, This represents the current optimal position, and F represents randomly selecting one of the two search directions, positive or negative. This indicates that the elements are uniformly distributed in the interval. Vectors in the array, random numbers .

[0129] The mathematical model for the adaptive exploration factor is as follows:

[0130] ;

[0131] and The upper and lower limits of the exploration factor are set to 0.1 and 1, respectively, and p is the iteration progress factor: t represents the current iteration number, and T represents the total number of iterations.

[0132] The mathematical model for the adaptive development factor is as follows:

[0133] ;

[0134] in, and These are the upper and lower limits of the development factor, which are preferably 0.5 and 2 in this embodiment.

[0135] The mathematical model for a random number vector representing Brownian motion is:

[0136] ;

[0137] The mathematical model for representing random numbers related to the current iteration number t is:

[0138] ;

[0139] Here, rand is a random number between 0 and 1.

[0140] The mathematical model for improving lemming foraging and predator avoidance behavior based on adaptive parameters is as follows:

[0141] ;

[0142] in, The Euclidean distance between the current position and the optimal solution is represented by the following mathematical model:

[0143] ;

[0144] in, Indicates the dimension of the data. Indicates the first The first search agent Vi in the The position of the next iteration. This indicates the position of the current optimal search agent in the j-th dimension during the t-th iteration;

[0145] Let be the escape coefficient, and the mathematical model be as follows: t is the current iteration number, and T is the total number of iterations; For Lévy's flight function.

[0146] S5. Simulated bidirectional population evolution strategy and memory mechanism are added to the end of the lemming position update to update and optimize the lemming population. In this embodiment, the population is divided into a high-quality population of 20% and a low-quality population of 80% and different strategies are used for updating.

[0147] Simulating a bidirectional population evolution strategy improves the coverage of the lemming population in the search space, significantly enhancing the algorithm's ability to escape local optima, and enabling the algorithm to thoroughly explore both local and global optima in the space.

[0148] The mathematical model for simulating a two-way population evolution strategy is as follows:

[0149] ;

[0150] in and This refers to two high-quality individuals that are different from the current individual in a high-quality population. This represents a discrete random variable that takes the value 1 or 2, obtained through a random number generation function. The mathematical model represents the mutation factor based on the sine function. ;

[0151] Where t is the current iteration number and T is the total number of iterations;

[0152] The mathematical model for low-quality population renewal in the bidirectional population evolution strategy is as follows:

[0153] ;

[0154] in, Represents the random direction factor, random number , and This represents the upper and lower bounds of a population variable.

[0155] A memory mechanism is added after the bidirectional population evolution strategy. Specifically, after the population position is iterated, the fitness of the entire population matrix is ​​evaluated, the fitness of the updated population is compared with that of the historical population, and the lemmings with better fitness are retained as the new generation population. The quality of the solution improves as the iteration progresses.

[0156] S6. After simulating migration, burrowing, foraging and predator avoidance, simulating two-way population evolution strategies and memory mechanism updates, the corresponding flight cost is calculated using the flight cost function to obtain the fitness of all individuals in the population, and the individual with the lowest fitness is selected as the optimal lemming individual.

[0157] S7. Determine whether the current iteration number t has reached the maximum iteration number T. If it has, end the path optimization and output the optimal path. If it has not, re-simulate the lemming population iteration and repeat steps S4-S7 to continue optimizing the path.

[0158] Figure 2 This is a 3D image of the optimal path planning result output in this embodiment. Figure 3This is a right-hand view of the optimal path planning result output in this embodiment.

[0159] To verify the effectiveness of the improved algorithm proposed in this embodiment, the improved lemming-optimized ALA algorithm and the unimproved lemming-optimized ALA algorithm in this embodiment are verified in a composite test function.

[0160] Figure 4 This is a comparison of the convergence of the improved lemming algorithm and the unimproved lemming algorithm provided in this embodiment of the invention under a composite test function. The improved lemming-optimized ALA algorithm significantly outperforms the unimproved lemming-optimized ALA algorithm in both convergence speed and accuracy.

[0161] This embodiment utilizes an improved lemming algorithm to establish a global path optimization function with the goal of minimizing flight cost, solving for the optimal flight path of the UAV. Latin hypercube sampling is introduced for population initialization, improving the quality of the initial population and increasing the search space coverage. Novel adaptive development and exploration factor parameters are designed during the algorithm's exploration and development phases, effectively balancing the lemming algorithm's development and exploration capabilities, emphasizing the development / exploration capabilities at different iteration stages. A bidirectional population evolution strategy and memory mechanism are added at the end of the lemming position update, significantly enhancing the algorithm's ability to escape local optima, enabling it to thoroughly explore both local and global optima in the search space. The combination of multiple strategies allows the improved lemming algorithm to balance optimization quality, robustness, and convergence speed, ensuring the feasibility of UAV flight path planning.

[0162] Based on the disclosure and teachings of the foregoing specification, those skilled in the art can make changes and modifications to the above embodiments. Therefore, the present invention is not limited to the specific embodiments disclosed and described above, and some modifications and changes to the invention should also fall within the protection scope of the claims of the present invention. Furthermore, although some specific terms are used in this specification, these terms are only for convenience of explanation and do not constitute any limitation on the present invention.

Claims

1. A UAV 3D path planning method based on a multi-strategy improved lemming optimization algorithm, characterized in that, include: S1. Obtain real-world environmental data, construct a three-dimensional mission space model, and determine the starting and ending coordinates of the UAV's flight path; S2. Construct the flight cost function for the flight mission based on the UAV parameters and the three-dimensional mission space model; S3. Initialize the lemming population using Latin hypercube sampling, including: Set the relevant parameters for the lemming optimization algorithm; among them, the relevant parameters include: lemming population size and maximum number of iterations; The position of each lemming is initialized by Latin hypercube sampling, and it is detected whether the position of each lemming is within the map boundary of the task space model. The fitness value generated by each lemming at the current position is calculated, and the lemming with the best current fitness is selected as the best. S4. Based on the initialized lemming population, perform global exploration and local development iterations of adaptive parameters in sequence; The global exploration iteration based on adaptive parameters includes: During the lemming's path-finding process, the migration and burrowing behavior of the lemming are simulated to establish a mathematical model, and a global exploration is conducted to select the next path point for the lemming to move. The global exploration mathematical model based on adaptive parameters is as follows: ； in, and These represent the positions of the i-th search agent at the (t+1)-th and (t)-th iterations, respectively. and Let a and b represent two random individuals in the population; a and b are the ranges belonging to Integer index; This indicates the current optimal position; F represents randomly selecting one of the two search directions, positive or negative. This indicates that the elements are uniformly distributed in the interval. Vectors in the text; random numbers ; The mathematical model for the adaptive exploration factor is as follows: ； in, and Here, represents the upper and lower bounds of the development factor; p represents the iteration progress factor. Where t is the current iteration number and T is the total number of iterations; The mathematical model for the adaptive development factor is as follows: ； in, and These are the upper and lower limits of the development factor, respectively; The mathematical model for a random number vector representing Brownian motion is: ； L represents a random number related to the current iteration number, and the mathematical model is: ； Where rand is a random number between 0 and 1; Local development iterations based on adaptive parameters include: During the lemming's search path, a mathematical model is established to simulate the lemming's foraging and predator avoidance behaviors, and local development is carried out to select the next path point for the lemming to move. The mathematical model for local development based on adaptive parameters is as follows: ； in, and These represent the positions of the i-th search agent at the (t+1)-th and (t)-th iterations, respectively. Indicates the current optimal position; F represents randomly selecting one of the two search directions, positive or negative; random number. ; The mathematical model for the adaptive development factor is as follows: ； in, and Here, represents the upper and lower bounds of the development factor; p represents the iteration progress factor. Where t is the current iteration number and T is the total number of iterations; The Euclidean distance between the current position and the optimal solution is represented by the following mathematical model: ； Represents the dimension of a given problem; This indicates the position of the i-th search agent in the j-th dimension during the t-th iteration; This indicates that the current optimal search agent is at the . Vi in the The position of the next iteration; G is the escape coefficient, and the mathematical model is as follows: Where t is the current iteration number; T is the total number of iterations; For the Lévy flight function; S5. Based on the iteratively updated lemming population, a two-way population evolution strategy and a memory mechanism are used to update the lemming population, including: The population is divided into a high-quality population with a proportion of u and a low-quality population with a proportion of v, and different strategies are used to update the high-quality population and the low-quality population. Among them, u+v=1, and 0<u<v<1; The memory mechanism is as follows: after updating the population position and simulating a two-way population evolution strategy, the fitness of the entire population is evaluated, and individuals with lower fitness are retained as the new generation of lemming population. S6. Based on the UAV flight paths planned by all lemming individuals, calculate the corresponding flight cost using the flight cost function, obtain the fitness of all individuals in the population, and select the individual with the lowest fitness as the optimal lemming individual. S7. Determine if the current iteration count has reached the maximum iteration count; if it has, end the path optimization and output the optimal path; if it has not, re-simulate the lemming population iteration and continue to optimize the path.

2. The UAV 3D path planning method based on the multi-strategy improved lemming optimization algorithm according to claim 1, characterized in that, Constructing the three-dimensional task space model includes: Based on the UAV flight mission, a three-dimensional terrain map is drawn in a spatial coordinate system, and non-collision obstacles are set in areas in the three-dimensional space scene that may pose a threat to the UAV flight mission, thereby obtaining the three-dimensional mission space model.

3. The UAV 3D path planning method based on the multi-strategy improved lemming optimization algorithm according to claim 1, characterized in that, The flight cost function for constructing the flight mission includes: Based on the total length of the flight mission path, construct the flight length cost function for the UAV; Based on the distance between each obstacle and the flight path in the three-dimensional mission space model, a threat avoidance cost function for the UAV is constructed. Based on the altitude of each waypoint in the flight path, construct the flight altitude cost function of the UAV; Based on the changes in the UAV's turning and climb angles along its flight path, a path smoothing cost function for the UAV is constructed.

4. The UAV 3D path planning method based on the multi-strategy improved lemming optimization algorithm according to claim 1, characterized in that, The Latin hypercube sampling model is as follows: ； in, For the initial lemming individual, and Let the j-th dimension be the upper and lower bounds. For a random permutation of the range 1 to N belonging to the j-th dimension, For population size, In the interval Random numbers on the array.

5. The UAV 3D path planning method based on the multi-strategy improved lemming optimization algorithm according to claim 1, characterized in that, The mathematical model for high-quality population renewal using a two-way population evolution strategy is as follows: ； in, and This refers to two high-quality individuals in a high-quality population that are different from the current individual. Indicates the current optimal position; This indicates the position of the i-th search agent in the (t+1)-th iteration; Let represent a discrete random variable taking the value 1 or 2, obtained through a random number generation function; W represents a mutation factor based on a sine function, with the mathematical model as follows: ,in, Let t represent the dimension of the given problem, t be the current iteration number, and T be the total number of iterations. The mathematical model for low-quality population renewal under a two-way population evolution strategy is as follows: ； in, and These represent the positions of the i-th search agent at the (t+1)-th and (t)-th iterations, respectively. and These represent the upper and lower bounds of the population variable, respectively. Represents the random direction factor; random number .

Images