A three-dimensional unmanned aerial vehicle path planning method based on an improved differential evolution algorithm

Abstract

The application provides a three-dimensional unmanned aerial vehicle path planning method based on an improved differential evolution algorithm, and comprises the following steps: parameterized modeling of a three-dimensional terrain and a flight environment of the unmanned aerial vehicle operation; construction of a total cost function comprising four dimensions of distance, safety, height and smoothness; minimization solving of the total cost function by using the improved differential evolution algorithm, determination of an optimal control point set, and generation of a smooth and continuous final three-dimensional flight path by a Catmull-Rom spline interpolation method based on the optimal control point set. The improved differential evolution algorithm is used to optimize the path control points, traditional random mutation and knowledge-guided mutation operations are dynamically fused in the evolution process, the knowledge model is updated by continuously collecting successful evolution samples, the mutation direction can be adaptively adjusted according to the search stage and the path feasibility state, so that the global exploration ability is ensured and the local search efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of UAV path planning, specifically a three-dimensional UAV path planning method based on an improved differential evolution algorithm. Background Technology

[0002] Traditional differential evolution algorithms are widely used in 3D UAV path planning problems due to their simple structure and strong global search capabilities. Their basic idea is to iteratively optimize path control points through a "mutation-crossover-selection" operation. However, in complex 3D flight environments, the mutation operation of this algorithm mainly relies on the differential information between randomly selected individuals in the current population, using only the current population state for searching. It lacks a systematic utilization of the large amount of effective path structure information and constraint satisfaction experience generated during the evolutionary process. When facing multi-obstacle, highly constrained 3D scenes, this characteristic easily leads to the algorithm generating a large number of infeasible paths that do not meet flight safety, altitude, or spatial constraints during the search process, resulting in a significant waste of computational resources, unstable convergence, and difficulty in supporting high-quality path planning requirements in complex environments.

[0003] Furthermore, while some existing improved differential evolution methods introduce historical information reuse mechanisms, the relevant knowledge is mostly limited to the simple utilization of numerical search directions or successful mutation vectors. They fail to combine the spatial structural characteristics and flight constraint semantics of the UAV path planning task itself for modeling, and the knowledge utilization methods usually adopt fixed strategies, lacking the ability to dynamically adjust with changes in the search stage and path feasibility status, making it difficult to play a stable role in actual engineering scenarios.

[0004] In the 3D UAV path planning problem, differential evolution algorithms need to maintain a dynamic balance between global exploration capability and local exploitation capability. On the one hand, in the early stages of the search, strong stochastic exploration capability is needed to cover the complex space to avoid getting trapped in local feasible regions. On the other hand, after gradually entering the feasible region, the path that satisfies the constraints needs to be finely adjusted to improve path smoothness and flight efficiency. However, existing technologies usually use a fixed mutation strategy or a single empirical rule applied to the entire evolutionary process, which cannot dynamically switch or adjust the search behavior according to the search stage, the evolutionary state of path feasibility, and the constraint satisfaction. This fixed strategy mode is prone to two types of problems in complex 3D environments: first, over-reliance on stochastic exploration leads to slow convergence speed; second, premature concentration on local exploitation leads to limited search space, thus getting trapped in local optima. Summary of the Invention

[0005] To address the aforementioned problems, this invention proposes a 3D UAV path planning method based on an improved differential evolution algorithm. This method achieves safe, compliant, optimal, and efficient convergence of UAV paths in complex 3D terrain. The core objective is to solve technical challenges such as multi-constraint fusion, reuse of historical experience, and dynamic parameter adaptation, generating an optimal flight path that satisfies multiple objectives including shortest distance, safety and controllability, altitude compliance, and trajectory smoothness. The technical solution provided by this invention is as follows:

[0006] A 3D UAV path planning method based on an improved differential evolution algorithm includes the following steps:

[0007] Step 1: Perform parametric modeling of the 3D terrain and flight environment for UAV operations;

[0008] Step 2: Construct a total cost function that includes four dimensions: distance, security, height, and smoothness;

[0009] Step 3: The total cost function is minimized using an improved differential evolution algorithm to determine the optimal control point set. Based on the optimal control point set, a smooth and continuous final three-dimensional flight path is generated using the Catmull-Rom spline interpolation method. The improved differential evolution algorithm introduces an adaptive knowledge mutation mechanism based on neural networks, which dynamically integrates random mutation and knowledge-guided mutation during the evolution process.

[0010] Preferably, the modeling process in step 1 is as follows:

[0011] The terrain elevation model is constructed using the Gaussian peak stacking method, and the calculation formula is as follows:

[0012]

[0013] in, Two-dimensional coordinates The topographic elevation value at that location, The center two-dimensional coordinates of the i-th Gaussian peak are generated by random and uniform sampling within a preset UAV operating area; Let be the height parameter of the i-th Gaussian peak, with a value range of . , This refers to the maximum flight altitude of the drone. and , respectively, are the attenuation coefficients of the i-th Gaussian peak along the x-axis and y-axis, and N is the total number of Gaussian peaks;

[0014] Flight space range constraints: The x-axis coordinate of the operational area is within Between, the y-axis coordinate is in Between, flight altitude z is in Between; among them, , , and Determined based on the geographical scope of the work scenario; This is the minimum safe flight altitude for drones; This is the maximum flight altitude of the drone.

[0015] Preferably, a safety height constraint is set: the height z of any point on the drone's flight path must be higher than the terrain height v corresponding to that point by at least a preset clearance height, and the terrain height v is obtained by bilinear interpolation.

[0016] Preferably, the cost function for step 2 is as follows:

[0017] The distance cost formula is as follows:

[0018]

[0019] Where M is the total number of sampling points, Let K be the three-dimensional coordinates of the k-th sampling point. The three-dimensional coordinates of the (k-1)th sampling point;

[0020] The formula for safety cost is as follows:

[0021]

[0022] in, Let k be the horizontal distance from the k-th sampling point to the center of the i-th peak. Let be the obstacle radius of the i-th peak. This is an approximation of the drone's diameter. The width of the safety buffer zone;

[0023] The formula for high cost is as follows:

[0024]

[0025] in, The terrain height corresponding to the k-th sampling point. The flight altitude is the kth sampling point, and clearance is the preset airspace clearance.

[0026] The formula for smoothness cost is as follows:

[0027]

[0028] in, The angle between the directions of adjacent trajectory segments;

[0029] The total cost function is:

[0030]

[0031] Where X is the decision variable. , , and These are the weighting coefficients for distance, safety, height, and smoothness costs, respectively.

[0032] Preferably, the improved differential evolution algorithm includes the following steps:

[0033] In each generation of evolution, a sequence of rules is generated for any given parent individual. The random number r; when When, knowledge-guided mutation operations are triggered; when At that time, the random mutation operation of traditional differential evolution is performed; among which, The probability of mutation is guided by knowledge;

[0034] When performing knowledge-guided mutation, the updated knowledge base model is invoked. This knowledge base model employs a feedforward neural network structure; the current position vector of the parent individual is then used. As input to the neural network, the corresponding evolutionary direction vector is output by the neural network. Based on the evolutionary direction vector, the knowledge-guided mutation vector is generated as follows:

[0035]

[0036] in, This is the direction scaling factor, used to adjust the step size of knowledge-guided mutations.

[0037] Preferably, the knowledge-guided mutation trigger probability The formula is adjusted as follows:

[0038]

[0039] in, This serves as the initial value for the probability of knowledge-guided mutation triggering. This represents the upper limit of the probability of knowledge-guided mutation triggering, where g is the current iteration number. This represents the maximum number of iterations.

[0040] Preferably, the specific process of generating the smooth and continuous final three-dimensional flight path in step 3 is as follows:

[0041] Set of control points Expand to ,in Starting from, The endpoint;

[0042] Based on the extended control point sequence, a smooth trajectory is generated using the Catmull-Rom spline interpolation method, as follows:

[0043] For the expanded control point sequence, trajectory points are generated using a piecewise interpolation method. For the i-th control point segment, the formula for calculating any interpolation point pt is:

[0044]

[0045] Where t is the interpolation parameter, controlling the position of the interpolation point in the current segment; a, b, c, and d are the spline basis function coefficients, respectively:

[0046]

[0047]

[0048]

[0049]

[0050] in, , , and These are the four adjacent control points after the expansion.

[0051] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0052] This invention employs an improved differential evolution algorithm to optimize path control points. During the evolution process, it dynamically integrates traditional random mutation and knowledge-guided mutation operations, and continuously updates the knowledge model by collecting successful evolution samples. This allows the mutation direction to adaptively adjust with the search stage and path feasibility status, thereby improving local search efficiency while ensuring global exploration capabilities. Attached Figure Description

[0053] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0054] Figure 1 This is the overall flowchart of the three-dimensional UAV path planning method of the present invention;

[0055] Figure 2 This is a diagram illustrating the specific steps of the improved differential evolution algorithm of this invention;

[0056] Figure 3 This is a schematic diagram of a specific scenario for the three-dimensional UAV path planning method of the present invention;

[0057] Figure 4 This invention provides a 3D path and mountain environment map for the 3D UAV path planning method. Detailed Implementation

[0058] 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 some embodiments of the present invention, and not all embodiments. 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.

[0059] To make the above-mentioned objectives, features and effects of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0060] Example 1: A 3D UAV path planning method based on an improved differential evolution algorithm, such as... Figure 1 As shown, it includes the following steps:

[0061] Step 1: Perform parametric modeling of the 3D terrain and flight environment for UAV operations;

[0062] Step 2: Construct a total cost function that includes four dimensions: distance, safety, height, and smoothness. Achieve multi-objective optimization through weighted summation, transforming the complex path planning problem into a single-objective minimization problem.

[0063] Step 3: The total cost function is minimized using an improved differential evolution algorithm to determine the optimal control point set. Based on this optimal control point set, a smooth and continuous final three-dimensional flight path is generated using the Catmull-Rom spline interpolation method. The improved differential evolution algorithm introduces an adaptive knowledge mutation mechanism based on neural networks, which dynamically integrates random mutation and knowledge-guided mutation during the evolution process to improve search efficiency and path feasibility.

[0064] In practice, the modeling process in step 1 is as follows:

[0065] A topographic elevation model is constructed using the Gaussian peak stacking method. This involves abstracting real mountains into N superimposed Gaussian peaks of varying heights and attenuation characteristics, thereby digitally reconstructing the terrain morphology and specifying the exact elevation. (two-dimensional coordinates) The formula for calculating the terrain elevation at a given location is as follows:

[0066]

[0067] in, The center two-dimensional coordinates of the i-th Gaussian peak are generated by random and uniform sampling within a preset UAV operating area; Let be the height parameter of the i-th Gaussian peak, with a value range of . , This is the preset maximum flight altitude for the drone; and Here, are the attenuation coefficients of the i-th Gaussian peak along the x-axis and y-axis, respectively, and N is the total number of Gaussian peaks, determined based on the terrain complexity of the UAV's operating area, preferably... .

[0068] Furthermore, to enable rapid terrain altitude lookup at any point along the UAV's flight path (supporting safe altitude constraint judgment), bilinear interpolation is used to interpolate the terrain mesh generated by Gaussian peak stacking to obtain the terrain altitude at any query point. The terrain height v is calculated in the following process:

[0069]

[0070]

[0071]

[0072]

[0073]

[0074]

[0075]

[0076]

[0077] in, and These are the start and end coordinates of the x-axis of the work area, respectively. and These are the start and end coordinates of the y-axis of the work area, respectively; and These represent the horizontal and vertical resolutions of the terrain grid, respectively. and The normalized coordinates of the query point; and A floating-point grid index for the query point; , , and An integer grid index for the query point; and For interpolation weights; v represents the terrain height of the grid vertex; v represents the final terrain height of the query point.

[0078] Furthermore, to limit the operational space boundaries of the UAV and prevent it from exceeding the controllable flight area, the flight space range constraint is clearly defined: the operational area's x-axis coordinate lies within... Between, the y-axis coordinate is in Between, flight altitude z is in Between. Among them, , , and The geographical scope is determined based on the actual operational scenario (such as mountain survey or border surveillance area). This is the minimum safe flight altitude for drones; This is the maximum flight altitude of the drone.

[0079] Furthermore, to avoid collisions between the drone and the terrain and ensure sufficient clearance between the flight path and the terrain, a safety height constraint is set: the height z of any point on the drone's flight path must be higher than the corresponding terrain height v by at least a preset clearance height. The clearance value is 3.0, which is determined based on the drone's rotor diameter and flight stability requirements, and can be adjusted according to the drone model; the terrain height v is obtained using the aforementioned bilinear interpolation method.

[0080] In practice, the cost function for step 2 is as follows:

[0081] Distance cost is defined as the sum of the Euclidean distances between all adjacent sampling points on the trajectory, as shown in the following formula:

[0082]

[0083] Where X is the decision variable, and M is the total number of sampling points. Let K be the three-dimensional coordinates of the k-th sampling point. The coordinates of the (k-1)th sampling point are given.

[0084] Safety cost is defined as the penalty imposed on a drone when it approaches a mountain obstacle, and the specific formula is as follows:

[0085]

[0086] in, Let k be the horizontal distance from the k-th sampling point to the center of the i-th peak. Let be the obstacle radius of the i-th peak. This is an approximation of the drone's diameter. This refers to the width of the safety buffer zone.

[0087] Altitude cost is defined as the penalty for a drone's flight altitude deviating from the compliant range, and the specific formula is as follows:

[0088]

[0089] in, The terrain height corresponding to the k-th sampling point. is the flight altitude of the kth sampling point, and clearance is the preset clearance altitude.

[0090] Smoothness cost is defined as the degree of trajectory bend and angle change, including two parts: the sum of the angles between adjacent trajectory segments and the sum of the angle differences. The specific formula is as follows:

[0091]

[0092]

[0093]

[0094]

[0095] in, The trajectory vector of the (k-1)th to kth sampling points; for , unit vector; The angle between the directions of adjacent trajectory segments; To smooth out costs, a smaller value indicates a smoother trajectory.

[0096] The total cost function is:

[0097]

[0098] Where X is the decision variable. , , and These are the weighting coefficients for distance, safety, height, and smoothness costs, respectively.

[0099] In practical implementation, the improved differential evolution algorithm in step 3 is as follows: Figure 2 As shown, the details are as follows:

[0100] To ensure sufficient diversity in the initial population and match the problem complexity, the population size and the dimension of the decision variables are set: the population size is set to NP=40 to achieve a balance between search diversity and computational efficiency; the dimension of the decision variables is set to... ,in, The total number of control points, including the start and end points, is given, and the x, y, and z three-dimensional spatial coordinates of each intermediate control point are optimized.

[0101] To ensure that decision variables always remain within the compliant flight space, boundary constraints are set for the decision variables: the lower bound vector of the decision variables is set as... Set the upper bound vector of the decision variable as ,in, The function is used to repeat boundary values ​​according to the control point dimension to ensure that the three-dimensional coordinates of each intermediate control point are within the preset space range.

[0102] To provide an initial baseline for iterative evolution, NP candidate solutions are uniformly and randomly generated within the boundary constraints to form an initial population. Each candidate solution is combined with a fixed start and end point to form a complete control point sequence, and its fitness value is calculated. The initial optimal solution, optimal fitness, and corresponding trajectory are recorded, and the number of function evaluations is initialized. .

[0103] In the differential evolution iteration process, in order to improve the algorithm's ability to utilize historical effective path adjustment experience and enhance its adaptive search performance in complex 3D UAV flight environments, this invention introduces a learning-based knowledge-guided mutation mechanism to dynamically fuse traditional random mutation operations during the evolution process, thereby achieving adaptive control of the search direction.

[0104] In each generation of evolution, for any parent individual Its corresponding position vector is denoted as Generate a range that conforms to the specified interval. The random number r; when When, knowledge-guided mutation operations are triggered; when At that time, the random mutation operation of traditional differential evolution is performed. Among them, This serves as a knowledge-guided mutation trigger probability, used to dynamically balance the ability of targeted search and random exploration based on historically valid experience during evolution, preventing search behavior from becoming prematurely fixed or relying entirely on random perturbations. The formula is adjusted as follows:

[0105]

[0106] in, This serves as the initial value for the probability of knowledge-guided mutation triggering. This represents the upper limit of the probability of knowledge-guided mutation triggering, where g is the current iteration number. To maximize the number of iterations, this rule aims to reduce the probability of knowledge guidance in the early stages of the search to enhance random exploration capabilities, and increase the probability of knowledge guidance in the later stages of the search to enhance local development capabilities, thereby refining the path optimization. This is used to dynamically balance the ability of directional search and random exploration based on historical effective experience during the evolution process, and to avoid the search behavior from becoming too fixed or relying entirely on random perturbations.

[0107] When performing a knowledge-guided mutation operation, an updated knowledge base model is invoked. This knowledge base model employs a feedforward neural network structure. Its function is not a fixed mapping obtained through offline training, but rather it is continuously updated during the evolutionary process to learn the mapping relationship between an individual's position state and its successful evolutionary displacement under the current path planning task and constraints. Specifically, the current position vector of the parent individual is... As input to the neural network, the corresponding evolutionary direction vector is output by the neural network. The evolutionary direction represents the adjustment direction that can effectively reduce the path cost function and satisfy the constraints within the current search stage. Based on the evolutionary direction vector, a knowledge-guided mutation vector is generated as follows:

[0108]

[0109] in, This is the direction scaling factor, used to adjust the step size of knowledge-guided mutations to adapt to the path adjustment accuracy requirements of different search stages.

[0110] When random number At that time, the traditional differential evolution random mutation operation is performed. Specifically, for the current parent individual... Three individuals with different indices are randomly selected from the population. The basic mutation vector is generated as follows:

[0111]

[0112] in, This is a mutation factor used to control the scaling of the difference vector, thereby maintaining the algorithm's global exploration capability and preventing the search process from getting trapped in local optima.

[0113] After the mutation operation is completed, a binary crossover strategy is used to recombine the genes of the parent individuals with the corresponding mutation vectors to generate experimental individuals. For each dimension Generate random numbers ,when or season Otherwise Where CR is the crossover probability. The dimension index is randomly selected to ensure that experimental individuals inherit variation information in at least one dimension.

[0114] Calculate the path fitness value for each generated experimental individual and compare it with the fitness value of its corresponding parent individual. When the total cost function... When the parent individual is replaced by a test individual, it enters the next generation of the population; otherwise, the parent individual is retained to ensure that the overall search performance of the population does not degrade.

[0115] When a test individual successfully replaces its corresponding parent individual, it is considered a valid evolutionary experience. The position vector of the parent individual and its actual evolutionary displacement vector form a set of online training samples, which are used to incrementally update and train the feedforward neural network, thereby enabling the knowledge base model to learn and adaptively evolve during the evolutionary process. By continuously accumulating samples derived only from successful evolution, interference from invalid or incorrect search directions is avoided.

[0116] Through the above-mentioned update mechanism, the neural network can gradually form a directional guidance capability that matches the current UAV path planning task and environmental constraints during the differential evolution process, and provide more targeted search direction prediction for knowledge-guided mutation in subsequent iterations.

[0117] After each generation iteration, the individual with the best fitness in the current population is selected as the optimal solution for that generation; when its fitness is better than the historical best fitness, the global optimal solution, the global optimal fitness, and the corresponding set of trajectory sampling points are updated.

[0118] To control computational complexity and avoid invalid iterations, an algorithm termination condition is set: the algorithm terminates when the number of iterations reaches the maximum number of iterations, the number of function evaluations reaches the maximum number of iterations, or the change in the global optimal fitness is less than a preset threshold for several consecutive generations.

[0119] Once the algorithm meets the termination condition, it outputs the global optimal control point set and reconstructs it into a three-dimensional coordinate sequence of several intermediate control points. Together with the preset start and end points, it forms a complete path control point sequence, thereby generating three-dimensional flight trajectory sampling points that the UAV can execute, providing specific path data for UAV path planning and flight control.

[0120] In practice, the specific process of generating the smooth and continuous final three-dimensional flight path in step 3 is as follows:

[0121] The original control point set Expand to ,in Starting from, The endpoint is the extended sequence, which ensures that adjacent extended control points can be called at the start and end points during interpolation calculations, avoiding abrupt changes in derivatives at the endpoints and ensuring the smoothness of the entire trajectory.

[0122] Based on the extended control point sequence, a smooth trajectory is generated using the Catmull-Rom spline interpolation method, as follows:

[0123] For the expanded control point sequence, trajectory points are generated using piecewise interpolation. For the i-th control point segment (corresponding to the original control point...) and The formula for calculating any interpolation point pt (between points) is:

[0124]

[0125] Where t is the interpolation parameter (range of values) ), controls the position of the interpolation point in the current segment; a, b, c, and d are the spline basis function coefficients, respectively:

[0126]

[0127]

[0128]

[0129]

[0130] in, , , and For the four adjacent control points after expansion, ensure trajectory continuity and derivative smoothness.

[0131] Combination Figure 3 This invention provides a specific scenario for a 3D UAV path planning method based on an improved differential evolution algorithm. The scenario is an operational environment with complex 3D mountainous terrain obstacles, occurring within a closed mountainous area of ​​several square kilometers, such as mountain reconnaissance or emergency material delivery. The operational requirements are for the UAV to plan a safe and efficient flight path in the 3D mountainous terrain, avoiding terrain obstacles and meeting constraints such as safe altitude and spatial range. The number of operational nodes (such as reconnaissance points and material delivery points) is dynamically adjusted according to the task, and the operation is characterized by suddenness (such as in emergency scenarios). The UAVs participating in the operation are multi-rotor UAVs adapted to complex terrain. A single UAV performs the path planning task, achieving autonomous path optimization through an improved differential evolution algorithm.

[0132] Although the terrain is complex, the scene has significant spatial constraints, including mountainous terrain obstacles modeled by superimposed Gaussian peaks (no-fly zones formed by multiple peaks), safe flight altitude constraints (the drone's flight altitude must be higher than the terrain altitude by at least a preset clearance altitude), and flight space range constraints (the boundaries of the operational area along the x, y, and z axes). Drones cannot fly into the lethal areas of the terrain obstacles and must strictly meet the safe altitude and space range restrictions.

[0133] Combination Figure 3In this invention, the 3D UAV path planning method optimizes the control points of a 3D mountainous environment modeled using an improved differential evolution algorithm to generate an optimal flight path that satisfies multiple constraints. First, Gaussian peak overlay modeling is performed on the mountainous terrain to determine the terrain elevation. Then, a multi-objective total cost function is constructed, including distance, safety, altitude, and smoothness. Finally, an improved differential evolution algorithm incorporating a knowledge learning mechanism is used to solve for the optimal set of control points, generating a smooth flight path (as shown by the dotted line in the figure, connecting the red starting point and the blue ending point, avoiding mountain obstacles and meeting the safety altitude requirement). This method efficiently plans UAV paths, balancing multiple objectives such as path length, safety, and smoothness, generating flight trajectories that meet the requirements of "safety and compliance, optimal path, and efficient convergence," making it suitable for practical operational needs such as mountain surveying and emergency logistics.

[0134] Combination Figure 4 This invention provides a 3D path and mountain environment map based on an improved differential evolution algorithm for 3D UAV path planning.

[0135] Figure 4 The map presents multiple undulating mountain peaks in a three-dimensional form. In the legend, the red curve represents the optimized drone flight path, the green dot is the starting point, and the red dot is the ending point. The path is planned in a complex three-dimensional mountain environment, avoiding terrain obstacles. It intuitively demonstrates the safe and efficient planning effect of the drone path based on the improved algorithm from the starting point to the ending point.

[0136] Example 2: The computer-readable storage medium of this example stores a computer program that, when executed by a processor, implements the steps in the three-dimensional UAV path planning method based on an improved differential evolution algorithm in Example 1.

[0137] The computer-readable storage medium in this embodiment can be an internal storage unit of the terminal, such as the terminal's hard disk or memory; the computer-readable storage medium in this embodiment can also be an external storage device of the terminal, such as a plug-in hard disk, smart memory card, secure digital card, flash memory card, etc. equipped on the terminal; furthermore, the computer-readable storage medium can include both the terminal's internal storage unit and external storage devices.

[0138] The computer-readable storage medium of this embodiment is used to store computer programs and other programs and data required by the terminal. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0139] Example 3: The computer device of this example includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the three-dimensional UAV path planning method based on an improved differential evolution algorithm in Example 1.

[0140] In this embodiment, the processor can be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits, off-the-shelf programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc. The memory can include read-only memory and random access memory, and provides instructions and data to the processor. A portion of the memory can also include non-volatile random access memory. For example, the memory can also store device type information.

[0141] Those skilled in the art will clearly understand that each implementation can be achieved using software plus the necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0142] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A three-dimensional UAV path planning method based on an improved differential evolution algorithm, characterized in that, Includes the following steps: Step 1: Perform parametric modeling of the 3D terrain and flight environment for UAV operations; Step 2: Construct a total cost function that includes four dimensions: distance, security, height, and smoothness; Step 3: The total cost function is minimized using an improved differential evolution algorithm to determine the optimal control point set. Based on the optimal control point set, a smooth and continuous final three-dimensional flight path is generated using the Catmull-Rom spline interpolation method. The improved differential evolution algorithm introduces an adaptive knowledge mutation mechanism based on neural networks, which dynamically integrates random mutation and knowledge-guided mutation during the evolution process.

2. The three-dimensional UAV path planning method based on an improved differential evolution algorithm according to claim 1, characterized in that, The modeling process in step 1 is as follows: The terrain elevation model is constructed using the Gaussian peak stacking method, and the calculation formula is as follows: ； in, Two-dimensional coordinates The topographic elevation value at that location, The center two-dimensional coordinates of the i-th Gaussian peak are generated by random and uniform sampling within a preset UAV operating area; Let be the height parameter of the i-th Gaussian peak, with a value range of . , This refers to the maximum flight altitude of the drone. and , respectively, are the attenuation coefficients of the i-th Gaussian peak along the x-axis and y-axis, and N is the total number of Gaussian peaks; Flight space range constraints: The x-axis coordinate of the operational area is within Between, the y-axis coordinate is in Between, flight altitude z is Between; among them, , , and Determined based on the geographical scope of the work scenario; This is the minimum safe flight altitude for drones; This is the maximum flight altitude of the drone.

3. The three-dimensional UAV path planning method based on an improved differential evolution algorithm according to claim 2, characterized in that, Set a safe altitude constraint: The altitude z of any point on the drone's flight path must be higher than the terrain altitude v corresponding to that point by at least a preset clearance altitude. The terrain altitude v is obtained by bilinear interpolation.

4. The three-dimensional UAV path planning method based on an improved differential evolution algorithm according to claim 2, characterized in that, The cost function for step 2 is as follows: The distance cost formula is as follows: ； Where M is the total number of sampling points, Let K be the three-dimensional coordinates of the k-th sampling point. The three-dimensional coordinates of the (k-1)th sampling point; The formula for safety cost is as follows: ； in, Let k be the horizontal distance from the k-th sampling point to the center of the i-th peak. Let be the obstacle radius of the i-th peak. This is an approximation of the drone's diameter. The width of the safety buffer zone; The formula for high cost is as follows: ； in, The terrain height corresponding to the kth sampling point. The flight altitude is the kth sampling point, and clearance is the preset airspace clearance. The formula for smoothness cost is as follows: ； in, The angle between the directions of adjacent trajectory segments; The total cost function is: ； Where X is the decision variable. , , and These are the weighting coefficients for distance, safety, height, and smoothness costs, respectively.

5. A three-dimensional UAV path planning method based on an improved differential evolution algorithm according to claim 4, characterized in that, The improved differential evolution algorithm includes the following steps: In each generation of evolution, a sequence of rules is generated for any given parent individual. The random number r; when When, knowledge-guided mutation operations are triggered; when At that time, the random mutation operation of traditional differential evolution is performed; among which, The probability of mutation is guided by knowledge; When performing knowledge-guided mutation, the updated knowledge base model is invoked. This knowledge base model employs a feedforward neural network structure; the current position vector of the parent individual is then used. As input to the neural network, the corresponding evolutionary direction vector is output by the neural network. Based on the evolutionary direction vector, the knowledge-guided mutation vector is generated as follows: ； in, This is the direction scaling factor, used to adjust the step size of knowledge-guided mutations.

6. A three-dimensional UAV path planning method based on an improved differential evolution algorithm according to claim 5, characterized in that, Knowledge guides mutation trigger probability The adjusted formula is as follows: ； in, This serves as the initial value for the probability of knowledge-guided mutation triggering. This represents the upper limit of the probability of knowledge-guided mutation triggering, where g is the current iteration number. This represents the maximum number of iterations.

7. A three-dimensional UAV path planning method based on an improved differential evolution algorithm according to claim 5, characterized in that, The specific process of generating the smooth and continuous final three-dimensional flight path in step 3 is as follows: Set of control points Expand to ,in Starting from, The endpoint; Based on the extended control point sequence, a smooth trajectory is generated using the Catmull-Rom spline interpolation method, as follows: For the expanded control point sequence, trajectory points are generated using a piecewise interpolation method. For the i-th control point segment, the formula for calculating any interpolation point pt is: ； Where t is the interpolation parameter, controlling the position of the interpolation point in the current segment; a, b, c, and d are the spline basis function coefficients, respectively: ； ； ； ； in, , , and These are the four adjacent control points after the expansion.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the three-dimensional UAV path planning method based on the improved differential evolution algorithm as described in any one of claims 1-7.

9. A computer device comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the three-dimensional UAV path planning method based on the improved differential evolution algorithm as described in any one of claims 1-7.

Images