Three-dimensional autonomous path planning method based on sunflower orientation positive

Abstract

The invention discloses a three-dimensional multi-unmanned aerial vehicle path planning method based on a sunflower phototaxis mechanism. According to the method, a three-dimensional hormone field for simulating hormone diffusion and attenuation and a memory avoidance field of a path are constructed, a plurality of agents are guided to plan the path under the combined action of a target point direction, a local hormone gradient and the memory avoidance field, and the path evolution direction is dynamically corrected in combination with an obstacle avoidance strategy; and strengthening the hormone field along the path of the intelligent agent and the condition of reaching the target, and iteratively forming a feasible path track leading to the target. Along with iteration, the intelligent agent gradually approaches the target point; and when a plurality of agents successfully arrive at the target, selecting an optimal path track according to a preset reward scoring function. The invention provides a natural heuristic path planning model which is suitable for autonomous navigation tasks in a complex three-dimensional space.

Description

Technical Field

[0001] This invention belongs to the field of UAV path planning technology, and specifically relates to a three-dimensional path planning method based on a nature-inspired mechanism to simulate the phototactic growth behavior of sunflower plants under light stimulation, thereby guiding the path evolution process of UAVs. Background Technology

[0002] With the widespread application of intelligent autonomous systems in military reconnaissance, urban patrol, disaster search and rescue, and agricultural spraying, unmanned aerial vehicles (UAVs), as an important component, have become one of the core equipment in modern autonomous missions due to their advantages such as high mobility, flexible deployment, and low operating costs. As mission scenarios become increasingly complex, UAVs, when performing autonomous navigation missions, often need to avoid obstacles in three-dimensional space, meet path continuity requirements, and quickly find feasible paths to the target. This places higher demands on the intelligence and environmental adaptability of their path planning algorithms.

[0003] In the field of path planning, existing mainstream methods can be broadly divided into two categories: one is deterministic methods, such as the A* algorithm, Dijkstra's algorithm, and Rapid Expanding Random Tree (RRT), which are characterized by graph structures or heuristic search strategies, relying on precise mathematical models or rule-driven approaches, and gradually approximating the optimal solution through iterative calculations. They are suitable for global path planning in static and structured environments. The other category is nondeterministic methods, namely various metaheuristic intelligent algorithms, such as Particle Swarm Optimization (PSO), Genetic Algorithm (GA), and Ant Colony Optimization (ACO), which demonstrate good global search capabilities in multi-objective optimization and nonlinear constraint problems by simulating the collective intelligence or evolutionary behavior in natural systems.

[0004] Although the above methods have achieved good results in different scenarios, existing methods still face the following limitations in path planning tasks facing complex, dynamic, and unstructured 3D environments: First, many traditional methods rely on rule-based modeling or heuristic function design, lacking a generative path growth mechanism, and the path configurations are mostly polylines or segmented structures, lacking natural evolutionary continuity; Second, some metaheuristic algorithms focus on the optimal selection of path points, ignoring the spatial coherence and biological rationality of the path generation process; Third, currently few path planning models can simultaneously integrate multiple physiological regulatory mechanisms such as "target attraction," "trajectory memory," and "local incentives," and there is still considerable room for expansion in simulating biological behavior.

[0005] Therefore, there is an urgent need to propose a new three-dimensional path planning method that can draw inspiration from natural systems and construct a path growth model with goal orientation, self-organizing evolution and spatial adaptability. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and propose a three-dimensional path planning method based on the phototaxis mechanism of sunflower plants. This method simulates the response behavior of organisms to target stimuli during growth and constructs a path generation system with multiple physiological inspiration elements such as direction guidance, path memory, and autonomous obstacle avoidance, which is suitable for autonomous navigation tasks in complex spaces.

[0007] The objective of this invention is achieved through the following technical solution: a three-dimensional autonomous path planning method based on sunflower helixity, comprising the following steps:

[0008] S1. Initialize the Sunflower algorithm parameters as follows: Model the UAV path planning task as a three-dimensional spatial path generation problem, and set the space as a cubic voxel network of size L×L×L, where each spatial unit represents a unit voxel;

[0009] After initialization, all drones start from the starting point S = (x s y s , z s Starting from point G, and avoiding all obstacles in the environment, plan a route to the target point G = (x). g y g , z g The path planning is considered complete when all drones successfully reach the target area. in Let be the latest position coordinates of the i-th drone, and ∈ be the path completion distance threshold;

[0010] The initial number of drones is N, and each drone has its current position, historical path point set, and activation flag φ. i Historical path point set P i Initialize to empty, containing only the starting position; activation flag φ i Setting it to true indicates that the evolution can continue.

[0011] Several static obstacles are placed in a three-dimensional space. Each obstacle is modeled as a sphere, and its parameters are represented by a quadruple (x...). j y j , z j r j ), where (x j y j , z j () represents the coordinates of the sphere's center, r j The radius is defined as the area within the obstacle, which is impassable and must be actively avoided during path evolution.

[0012] Two three-dimensional distribution fields are initialized: a hormone field H(x, y, z) and a memory field M(x, y, z); the hormone field is used to simulate the diffusion process of growth hormones in plant phototropism, and the memory field is used to simulate the self-avoidance phenomenon of non-repeatable growth of plant shoot tips during natural growth; both fields are initialized as zero matrices, consistent with the size of the environmental space.

[0013] S2. Construct a model for the diffusion and decay of hormones to simulate the process of hormone diffusion and decay; the hormone field is updated through a two-step process: exponential decay and spatial diffusion convolution.

[0014] S3. Based on the natural directional repulsion of plants, establish the attenuation and path avoidance mechanism of the memory field, and construct a path memory avoidance model;

[0015] S4. Based on the sunflower's perception and adaptation behavior to the external spatial structure during actual growth, construct the UAV path evolution process and obstacle avoidance strategy;

[0016] S5. Determine the target completion status. When the distance between the drone and the target point is less than the preset threshold, the target task is considered to be completed. If all drones have completed the task, proceed to step S6; otherwise, proceed to step S2.

[0017] S6. Construct a comprehensive reward scoring function that integrates multiple dimensions of indicators and select the optimal path.

[0018] The beneficial effects of this invention are as follows: This invention proposes a natural heuristic path evolution method that integrates a three-field coupling mechanism of "target attraction-memory avoidance-local repulsion". Compared with existing methods, it does not rely on fixed heuristic functions or graph search rules, but instead constructs a path growth modeling framework by simulating the distribution of phototropic hormones and self-organizing behavior in plants. This method is applicable to path planning tasks in unknown or semi-known environments and has good adaptability and scalability.

[0019] This invention establishes a natural heuristic path evolution model with "hormone diffusion attraction - memory avoidance inhibition - obstacle rejection correction" as its core mechanism by simulating hormone distribution and spatial response behavior in plant phototropism. It proposes a natural heuristic path planning framework that does not rely on fixed heuristic functions or graph search rules, but rather simulates the distribution and self-organizing behavior of hormones in plant phototropism. Compared with existing methods, this invention has the following beneficial effects and theoretical value:

[0020] (1) In terms of path generation mechanism, it shifts from point set optimization to continuous evolution path modeling, which has stronger biological rationality and trajectory coherence.

[0021] (2) By introducing a memory field, the spatial diversity of path distribution is enhanced, path oscillation and repeated traversal are avoided, and the path evolution area is made wider.

[0022] (3) The method structure has good scalability and versatility, and can be integrated with modules such as dynamic target modeling and multi-UAV collaborative mechanism, and has broad application prospects;

[0023] (4) It is suitable for autonomous navigation tasks in complex and unknown three-dimensional environments, especially in scenarios with incomplete path information, dense environmental obstacles, or unclear target areas, showing good potential adaptability. Attached Figure Description

[0024] Figure 1 This is a flowchart of a three-dimensional autonomous path planning method based on sunflower heliotropy according to the present invention;

[0025] Figure 2 This is a schematic diagram of the optimal UAV flight path scheme for path evolution in an embodiment of the present invention;

[0026] Figure 3 The above are schematic diagrams of the "diffused isosurface hormone blocks" and "maximum hormone heatmap" in the embodiments of the present invention.

[0027] Figure 4 This is a comparison chart of different path scores for the method of the present invention in an embodiment of the present invention. Detailed Implementation

[0028] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0029] like Figure 1 As shown, the present invention provides a three-dimensional autonomous path planning method based on sunflower heliotropy, comprising the following steps:

[0030] S1. Initialize the parameters of the Sunflower algorithm as follows: Model the UAV path planning task as a typical three-dimensional spatial path generation problem, set the space as a cubic voxel network of size L×L×L, each spatial unit represents a unit voxel, and construct a closed and controllable environment.

[0031] After initialization, all drones start from the starting point S = (x s y s , z s Starting from point G, and avoiding all obstacles in the environment, plan a route to the target point G = (x). g y g , z g The feasible path is determined; the path planning is considered complete when all drones successfully reach the target area. in Let be the latest position coordinates of the i-th drone, and ∈ be the path completion distance threshold;

[0032] The initial number of drones is N, and each drone has its current position, historical path point set, and activation flag φ. i Historical path point set P i Initialize to empty, containing only the starting position; activation flag φ i Setting it to true indicates that the evolution can continue.

[0033] During the drone initialization phase, the system assigns each drone a starting position in three-dimensional space, which is set at the starting point S = (x s y s , z s Within a finite neighborhood centered on P, stored i In this system, floating-point coordinates are supported for definition, and mapping to discrete spatial networks is performed through rounding and spatial boundary clipping operations; the specific operations are as follows:

[0034]

[0035] Where x, y, and z represent the initial coordinates of the i-th UAV after the cropping operation; x i y i , z i Let x0, y0, and z0 represent the initial random 3D coordinates generated by the i-th UAV within a finite neighborhood of the starting point S; let L represent the random offsets of the 3D coordinates; and let L represent the maximum boundary value in each dimension of the 3D space. In actual tasks, the initial position of the UAV can be configured according to the task scenario, such as uniform sampling within a specified area, clustering distribution near the target, or simulating the initial dispersion state of the group through random perturbation.

[0036] Several static obstacles are placed in a three-dimensional space to simulate path feasibility constraints in a complex environment; each obstacle is modeled as a sphere, and its parameters are represented by a quadruple (x...). j y j , z j r j ), where (x j y j , z j () represents the coordinates of the sphere's center, r j The radius is defined as the area within the obstacle, which is impassable and must be actively avoided during path evolution.

[0037] To simulate the growth mechanism in path evolution, the system initializes two three-dimensional distribution fields: a hormone field H(x,y,z) and a memory field M(x,y,z). The hormone field is used to simulate the diffusion process of growth hormones in plant phototropism, while the memory field is used to simulate the self-avoidance phenomenon of non-repeatable growth of plant shoot tips during natural growth. That is, individual plants have a natural directional repulsion during growth, avoiding their shoot tips from reversing the growth direction. Both fields are initially zero matrices with the same size as the environmental space, and diffusion is performed using three-dimensional convolution.

[0038] S2. Construct a hormone diffusion and decay model to simulate the hormone diffusion and decay process; within each iteration of the path evolution, the hormone field is updated through a two-step process: exponential decay and spatial diffusion convolution; simulate the natural distribution evolution of plant auxins in physiological space: the specific steps are as follows:

[0039] S21. The hormone field decays exponentially in a discrete form; in the hormone field H(x,y,z), the hormone intensity at each coordinate position is updated according to a discrete exponential decay model, expressed as:

[0040] H′ t (x,y,z)=(1-α)·H t-1 (x,y,z)

[0041] Among them, H′ t (x, y, z) represents the value of the hormone field at position (x, y, z) at time t after exponential decay; α∈(0,1) is the hormone decay rate, controlling the natural dissipation of hormones over time; H t-1 (x, y, z) represents the value of the hormone field at (x, y, z) at time t-1; this operation simulates the physiological phenomenon of hormone concentration in sunflower decreasing over time.

[0042] S22, Spatial diffusion convolution of hormone field voxels; a 3×3×3 three-dimensional mean diffusion kernel K(i,j,k) of size is used to perform spatial convolution processing on the decayed hormone field:

[0043]

[0044] Among them, H′ t (x+i, y+j, z+k) represents the exponentially decaying value of the hormone field at time t; H t (x,y,z) represents the new value of the hormone field at the current location after diffusion; this mean kernel ensures smoothness while maintaining the consistency of diffusion direction, and can effectively simulate the average diffusion behavior of plant auxin in local space.

[0045] Through the two-step update mechanism described above, the hormone field can not only simulate the natural diffusion of phototactic hormones, but also form a stable gradient over time, guiding the drone to "grow" towards the target area. Furthermore, this mechanism eliminates the need for manual path direction setting; instead, it achieves path induction based on a biological-like mechanism, which is a key component distinguishing this invention from traditional path optimization methods.

[0046] S23. Preliminary calculation of the path advancement direction; for each UAV, extract the hormone field gradient at the current position:

[0047]

[0048] Among them, H t (x, y, z) represents the new value of the hormone field at time t after diffusion at position (x, y, z); g h This represents the hormone field gradient value at the current location. Simultaneously, its direction vector relative to the target point is calculated:

[0049]

[0050] Where p represents the current position coordinates of the UAV; G represents the position coordinates of the target point;

[0051] Then they are merged into the initial path evolution direction:

[0052] dir h =ω1·d g +ω2·g h

[0053] Where ω1 is the weight of the target point term; ω2 is the weight of the hormone field term; dir h This represents the evolution direction of drone paths after being influenced by hormonal fields.

[0054] S3. Based on the natural directional repulsion of plants, a memory field decay and path avoidance mechanism is established to construct a path memory avoidance model. In the process of path evolution, in order to prevent UAVs from repeatedly entering the traversed area, based on the self-avoidance mechanism exhibited by plant stem tips in natural growth, the "historical penalty" effect in the path growth process is simulated. This invention designs a time decay type three-dimensional memory field structure to record the positions traversed by the UAV and guides the path evolution process to avoid the historical trajectory area through local gradient calculation.

[0055] The specific steps for S3 are as follows:

[0056] S31. The memory field decays exponentially in a discrete form; in each iteration, the intensity recorded by the memory field M(x,y,z) decays exponentially, and the update rule is as follows:

[0057] M t(x,y,z)=(1-β)·M t-1 (x, y, z)

[0058] Among them, M t (x, y, z) represents the value of the memory field at position (x, y, z) at time t, denoted as M. t β∈(0,1) is the memory decay coefficient, used to control the duration of the influence of historical trajectories in subsequent paths; M t-1 (x, y, z) represents the value of the memory field at position (x, y, z) at time (t-1) before decay;

[0059] S32. Perform memory field gradient calculation and path direction fusion correction; during the UAV path evolution process, obtain the memory gradient of the current position:

[0060]

[0061] This is then combined with the light-attracting direction as the reverse vector to update the path direction as follows:

[0062]

[0063] Among them, dilf inal This represents the path evolution direction of the UAV after correction by the memory field; dir h The direction of UAV path evolution under the influence of the hormone field is represented by γ; γ is the weight of the memory term; ∈ is a small positive number to prevent division by zero. This strategy can effectively suppress path looping and oscillation, and improve the spatial distribution diversity of paths.

[0064] S4. Based on the sunflower's perception and adaptation behavior to external spatial structures during actual growth, a UAV path evolution process and obstacle avoidance strategy are constructed. In this invention, the UAV path evolution process further simulates the sunflower's perception and adaptation behavior to external spatial structures during actual growth. When the plant stem tip encounters physical obstacles (such as rocks, branches, etc.) in the environment during growth, its end will deflect under mechanical stimulation, changing its original growth direction to avoid the resistance area, forming a spatial avoidance phenomenon.

[0065] The obstacle avoidance mechanism of the path proxy in this invention corresponds to this natural process: when the UAV approaches an obstacle area in three-dimensional space, it calculates the repulsion correction force based on its distance from the obstacle's center point, dynamically adjusts its propulsion direction, and simulates the flexible obstacle avoidance behavior of plants in complex spatial environments. The specific steps are as follows:

[0066] S4 1. Calculate the final path evolution direction; standardize the path direction after memory field correction to obtain the advancement direction:

[0067]

[0068] Where, dir final The path evolution direction of the UAV after memory field correction; ∈ is a small positive number to prevent division by zero;

[0069] S42. Proceed along the path; update the drone's position based on the step size:

[0070] p′ t+1 =p t +δ·v

[0071] Where, p′ t+1 The coordinates of the location the drone will arrive at in the next moment; p t Let t be the position coordinates of the UAV at time t; δ be the step size; v be the standardized UAV path forward direction.

[0072] S43. Implement obstacle avoidance mechanism; traverse all obstacles (c j r j If the distance between the current position of the drone and the center of the obstacle satisfies:

[0073] d j =||p t -c j ||<r j +R s

[0074] Where, p t c represents the position coordinates of the UAV at time t; j r represents the coordinates of the center point of the spherical obstacle. j R is the radius of the spherical obstacle; s To determine the safe radius of influence of the obstacle, a repulsive directional correction is applied:

[0075]

[0076] Where, p t+1 p′ represents the position coordinates of the UAV after propulsion, after repulsion correction. t+1 The coordinates of the drone's next position are given; λ is the weight of the obstacle repulsion term; r j R is the radius of the spherical obstacle; s The radius of influence of the obstacle;

[0077] S44. Perform boundary truncation; apply boundary constraints to each coordinate component to prevent it from positioning outside the constraint space, and update the current UAV position and path record structure:

[0078]

[0079] Where, p t The updated drone position coordinates; p t+1 The coordinates of the UAV after propulsion, after repulsion correction; L represents the maximum boundary value in each dimension of the three-dimensional space; P ti Let be the coordinates of all path points traversed by the i-th UAV at time t;

[0080] S45. Implement path status recording and hormone field / memory field updates; add the current position to the path set and strengthen the hormone and memory values ​​of that point:

[0081]

[0082] Where r is the amount of hormonal field reinforcement; η is the amount of memory field reinforcement, used to reinforce path history records and enhance subsequent avoidance effects.

[0083] S5. Determine the target completion status. When the distance between the drone and the target point is less than a preset threshold, the target task is considered completed. Iterate through all drones. If all drones have completed the task, proceed to step S6; otherwise, proceed to step S2. The specific determination method is: if the distance between the drone's current position and the target point meets the threshold... The drone is then marked as "complete," its subsequent path evolution is stopped, and the coordinate points along the entire path are reinforced.

[0084]

[0085] in, Let G be the current position coordinates of the i-th UAV; G be the position coordinates of the target point; ε be the path completion threshold; P be the path completion threshold. t This refers to the coordinates of all the path points traversed by the drone, i.e., the path recording structure. The amount of hormone enhancement at each pathway point; φ i This is the activation indicator for the drone.

[0086] S6. Construct a comprehensive reward scoring function integrating multi-dimensional indicators to select the optimal path. In the path selection stage, to simulate the natural selection mechanism of plants such as sunflowers during phototropic growth, this invention designs a comprehensive reward scoring function integrating multi-dimensional indicators. This scoring function takes the trajectory finally generated by the path surrogate as input and combines the goal orientation, structural continuity, hormone responsiveness, and environmental adaptability during the organism's growth process to comprehensively evaluate each path. The specific steps are as follows:

[0087] S61. Calculate the path length cost term; during the path evolution iteration process, for the i-th UAV (i∈[1,N]), the path record structure P has been obtained. ti as follows:

[0088]

[0089] Where p1, p2, ..., p t Let be the position coordinates of the i-th UAV at all times (1 to t);

[0090] The formula for calculating the total path length of the i-th UAV is:

[0091]

[0092] Where, p k+1 p represents the discrete coordinates of the (k+1)th point on the path. k Let k be the discrete coordinates of the k-th point on the path;

[0093] S62. Calculate the hormone response integral term; for the i-th UAV (i∈[1,N]), sample local values ​​in the hormone field along the route and accumulate them to simulate the preference for forming paths in areas with high hormone concentrations; the formula for calculating the hormone response integral term is as follows:

[0094]

[0095] Among them, (x ki y ki , z ki Let be the discrete coordinates of the l-th point on the path of the i-th UAV;

[0096] S63. Calculate the path smoothness term; naturally growing plant branches do not frequently and abruptly change direction, and their paths exhibit good directional continuity and flexible growth characteristics. In this invention, the angle between two adjacent direction vectors in the path of the i-th UAV (i∈[1,N]) is used as a measure of path smoothness, and the calculation formula is as follows:

[0097]

[0098] Where, p k+1 Let p be the discrete coordinates of the (k+1)th point on the path of the i-th UAV; k Let p be the discrete coordinates of the k-th point on the path of the i-th UAV; k-1 Let S be the discrete coordinates of the (k-1)th point on the path; ∈ is a small positive number to prevent division by zero; S i For path smoothness;

[0099] S64. Calculate the minimum obstacle distance term; During plant growth, plants actively avoid densely structured areas in space, and their stem tips deflect or stop growing when approaching obstacles. This invention uses the minimum distance between each point on the path of the i-th UAV (i∈[1,N]) and the boundaries of all obstacles as a metric for obstacle avoidance effectiveness. The calculation for all points on the path is as follows:

[0100] d kj =||p k -c j ||-r j

[0101] Where, p k c represents the discrete coordinates of the k-th point on the path of the i-th UAV; j r represents the coordinates of the center point of the spherical obstacle. j d is the radius of the spherical obstacle; kj Let be the distance between the k-th point on the path and the obstacle; take the minimum value as the minimum distance between the path and the obstacle:

[0102]

[0103] Where, d kj Let be the distance between the k-th point on the path of the i-th drone and the obstacle;

[0104] S65. For all paths that successfully reach the target point, calculate their overall path efficiency:

[0105] Score i =ω F ·F i +ω s ·S i +ω D ·D i +ω H ·H i , i∈[1,N]

[0106] Where N is the total number of drones; ω F It is the weight of the path length term; ω S It is the weight of the path smoothness term; ω D It is the weight of the distance term from the obstacle; ω H This is the weight of the hormone intensity term. Since the hormone field and memory field mechanism are introduced in this invention, there is no path that cannot reach the target point, and the value of i is i∈[1,N].

[0107] Select the path with the highest score from all completed paths as the final planned path output for the i-th drone.

[0108] In this embodiment, the task scenario is a 100×100×100 cube-shaped area. The center point coordinates of the starting point S are (10, 10, 10). There are 3 obstacle threats, all of which are spheres. Among them, sphere R... i The three quadruples of information are (40, 40, 40, 6), (50, 50, 50, 7), and (60, 60, 60, 8). The center point coordinates of the target point G are (90, 90, 90).

[0109] In this embodiment, the number of drones in the model is N=10, the path evolution step size is δ=1.5, the hormone field decay rate is α=0.005, the memory field decay rate is β=0.01, and the obstacle safety influence radius is R. S =4, path completion threshold ε = 2, hormone field enhancement amount r = 5, memory field enhancement amount η = 1, hormone enhancement amount for the complete path

[0110] Based on the above task scenario and parameter settings, the optimal path point, i.e., the optimal UAV flight path scheme, is calculated as follows: Figure 2 As shown. The "Schematic diagram of the isosurface hormone block of the hormone field diffusion" and the "maximum hormone heatmap" are as follows. Figure 3 As shown in the figure, the score comparison chart for different paths is as follows: Figure 4 As shown, the highest path score is 13. The simulation results confirm that the sunflower-based UAV path planning method can achieve autonomous 3D path planning for UAVs.

[0111] In summary, the 3D path planning algorithm based on the sunflower phototaxis mechanism proposed in this invention verifies the applicability and scalability of its core bio-inspired mechanism in various path generation models through modular mechanism design. First, the hormone field diffusion and decay module can serve as a novel path guidance mechanism, replacing traditional attraction-based modeling methods based on graph search or potential field functions. It achieves path growth control through a combination of goal-driven and local diffusion, making it suitable for navigation tasks. Second, the path memory avoidance mechanism, by constructing a time-decaying memory field, achieves repulsive avoidance control of traversed regions. This can be combined with historical backtracking mechanisms in models such as ant colony optimization and artificial potential field methods to improve path space diversity and planning efficiency. Furthermore, the hormone-memory dual-field synergistic mechanism, as a natural heuristic path evolution framework, can be embedded as a guidance mechanism in path decision-making models such as particle swarm optimization, multi-agent systems, and graph neural networks, strengthening the biological rationality and evolutionary continuity of path generation. It possesses strong integrability and theoretical extension value.

[0112] Experimental results show that the mechanism proposed in this invention can achieve a dynamic balance between global attraction and local avoidance through parameter adjustment, improve the rationality of path distribution while satisfying path accessibility, and has the potential to adapt to complex spatial constraints.

[0113] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A three-dimensional autonomous path planning method based on sunflower heliotropy, characterized in that, Includes the following steps: S1. Initialize the Sunflower algorithm parameters as follows: Model the UAV path planning task as a three-dimensional spatial path generation problem, and set the space as a cubic voxel network of size L×L×L, where each spatial unit represents a unit voxel; After initialization, all drones start from the starting point S = (x s y s , z s Starting from point G, and avoiding all obstacles in the environment, plan a route to the target point G = (x). g ,y g ,z g The path planning is considered complete when all drones successfully reach the target area. in Let be the latest position coordinates of the i-th drone, and ∈ be the path completion distance threshold; The initial number of drones is N, and each drone has its current position, historical path point set, and activation flag φ; the historical path point set P i Initialize to empty, containing only the starting position; activation flag φ i Setting it to true indicates that the evolution can continue. Several static obstacles are placed in a three-dimensional space. Each obstacle is modeled as a sphere, and its parameters are represented by a quadruple (x...). j y j , z j r j ), where (x j y j , z j () represents the coordinates of the sphere's center, r j The radius is defined as the area within the obstacle, which is impassable and must be actively avoided during path evolution. Two three-dimensional distribution fields are initialized: a hormone field H(x, y, z) and a memory field M(x, y, z); the hormone field is used to simulate the diffusion process of growth hormones in plant phototropism, and the memory field is used to simulate the self-avoidance phenomenon of non-repeatable growth of plant shoot tips during natural growth. Both fields are initialized as zero matrices, consistent with the size of the ambient space. S2. Construct a model for the diffusion and decay of hormones to simulate the process of hormone diffusion and decay; The hormone field is updated through a two-step process: exponential decay and spatial diffusion convolution. S3. Based on the natural directional repulsion of plants, establish the attenuation and path avoidance mechanism of the memory field, and construct a path memory avoidance model; S4. Based on the sunflower's perception and adaptation behavior to the external spatial structure during actual growth, construct the UAV path evolution process and obstacle avoidance strategy; S5. Determine the target completion status. When the distance between the drone and the target point is less than the preset threshold, the target task is considered to be completed. If all drones complete their tasks, proceed to step S6; otherwise, proceed to step S2. S6. Construct a comprehensive reward scoring function that integrates multiple dimensions of indicators and select the optimal path.

2. The three-dimensional autonomous path planning method according to claim 1, characterized in that, During the drone initialization phase, each drone is assigned a starting position in three-dimensional space, which is set at the starting point S = (x S y S , z S Within a finite neighborhood centered on P, stored i In this system, floating-point coordinates are supported for definition, and mapping to discrete spatial networks is performed through rounding and spatial boundary clipping operations; the specific operations are as follows: Where x, y, and z represent the initial coordinates of the i-th UAV after the cropping operation; x i y i , z i Let x0, y0, and z0 represent the initial random values ​​of the three-dimensional coordinates generated by the i-th UAV within a finite neighborhood of the starting point S; let L represent the random offsets of the three-dimensional coordinates; and let L represent the maximum boundary value of the three-dimensional space in each dimension.

3. The three-dimensional autonomous path planning method according to claim 1, characterized in that, The specific steps of S2 are as follows: S21. The hormone field decays exponentially in a discrete form; in the hormone field H(x, y, z), the hormone intensity at each coordinate position is updated according to a discrete exponential decay model, expressed as: H′ t (x,y,z)=(1−α)·H t-1 (x,y,z) Among them, H′ t (x, y, z) represents the value of the hormone field at position (x, y, z) at time t after exponential decay; α∈(0,1) is the hormone decay rate; H t-1 (x,y,z) represents the value of the hormone field at (x,y,z) at time t-1; S22, Spatial diffusion convolution of hormone field voxels; a 3×3×3 three-dimensional mean diffusion kernel is used to perform spatial convolution on the decayed hormone field to obtain... S22, Spatial diffusion convolution of hormone field voxels; A 3×3×3 three-dimensional mean diffusion kernel is used to perform spatial convolution on the decayed hormone field to obtain H. t (x, y, z); S23. Preliminary calculation of the path advancement direction; for each UAV, extract the hormone field gradient at the current position: At the same time, calculate its direction vector relative to the target point: Where p represents the current position coordinates of the UAV; G represents the position coordinates of the target point; Then, they are merged into the initial path evolution direction: dir h =ω1·d g +ω2·g h Where ω1 is the weight of the target point term; ω2 is the weight of the hormone field term.

4. The three-dimensional autonomous path planning method according to claim 1, characterized in that, The specific steps of S3 are as follows: S31. The memory field decays exponentially in a discrete form; in each iteration, the intensity recorded by the memory field M(x,y,z) decays exponentially, and the update rule is as follows: M t (x,y,z)=(1-β)·M t-1 (x,y,z) Among them, M t (x, y, z) represents the value of the memory field at position (x, y, z) at time t; β∈(0,1) is the memory decay coefficient; M t-1 (x,y,z) represents the value of the memory field at position (x,y,z) at time (t-1) before decay; S32. Perform memory field gradient calculation and path direction fusion correction; during the UAV path evolution process, obtain the memory gradient of the current position: This is then combined with the light-attracting direction as the reverse vector to update the path direction as follows: Where, dir final This represents the path evolution direction of the UAV after correction by the memory field; dir h γ represents the direction of drone path evolution under the influence of the hormone field; γ is the weight of the memory term; ∈ is a small positive number to prevent division by zero.

5. The three-dimensional autonomous path planning method according to claim 1, characterized in that, The specific steps of S4 are as follows: S41. Calculate the final path evolution direction; standardize the path direction after memory field correction to obtain the advancement direction: Where, dir final The path evolution direction of the UAV after memory field correction; ∈ is a small positive number to prevent division by zero; S42. Proceed along the path; update the drone's position based on the step size: p′ t+1 =p t +d·v Where, p′ t+1 The coordinates of the location the drone will arrive at in the next moment; p t Let t be the position coordinates of the UAV at time t; δ be the step size; v be the standardized UAV path forward direction. S43. Implement obstacle avoidance mechanism; traverse all obstacles (c j r j If the distance between the current position of the drone and the center of the obstacle satisfies: d j =||p t -c j ||<r j +R s Among them, c j R represents the coordinates of the center point of the spherical obstacle. s To determine the safe radius of influence of the obstacle, a repulsive directional correction is applied: Where, p t+1 The coordinates of the UAV after propulsion, after the rejection correction; λ is the weight of the obstacle rejection term; S44. Perform boundary truncation; apply boundary constraints to each coordinate component to prevent it from positioning outside the constraint space, and update the current UAV position and path record structure: Where, p t P represents the updated drone position coordinates. ti Let be the coordinates of all path points traversed by the i-th UAV at time t; S45. Implement path status recording and hormone field / memory field updates; add the current position to the path set and strengthen the hormone and memory values ​​of that point: Where r is the amount of hormonal field reinforcement; η is the amount of memory field reinforcement, used to reinforce path history records and enhance subsequent avoidance effects.

6. The three-dimensional autonomous path planning method according to claim 1, characterized in that, The specific steps of S6 are as follows: S61. Calculate the path length cost term; during the path evolution iteration process, for the i-th UAV, the path record structure P has already been obtained. ti as follows: Where p1, p2, ..., p t Let be the coordinates of the i-th drone's position at all times; The formula for calculating the total path length of the i-th UAV is: Where, p k+1 p represents the discrete coordinates of the (k+1)th point on the path. k Let k be the discrete coordinates of the k-th point on the path; S62. Calculate the hormone response integral term; for the i-th UAV, the formula for calculating the hormone response integral term is as follows: Among them, (x ki y ki , z ki Let be the discrete coordinates of the k-th point on the path of the i-th UAV; S63. Calculate the path smoothness term; the angle between two adjacent direction vectors in the path of the i-th UAV is used as a measure of path smoothness, and the calculation formula is as follows: Where, p k+1 Let p be the discrete coordinates of the (k+1)th point on the path of the i-th UAV; k Let p be the discrete coordinates of the k-th point on the path of the i-th UAV; k-1 Let S be the discrete coordinates of the (k-1)th point on the path; ∈ is a small positive number to prevent division by zero; S i For path smoothness; S64. Calculate the minimum obstacle spacing term; calculate for all points on the path as follows: d kj =||p k -c j ||-r j Among them, c j r represents the coordinates of the center point of the spherical obstacle. j d is the radius of the spherical obstacle; kj Let be the distance between the k-th point on the path and the obstacle; take the minimum value as the minimum distance between the path and the obstacle: S65. For all paths that successfully reach the target point, calculate their overall path efficiency: Score i =ω F ·F i +oh s ·S i +oh D ·D i +oh H ·H i ,i∈[1,N] Where N is the total number of drones; ω F It is the weight of the path length term; ω S It is the weight of the path smoothness term; ω D It is the weight of the distance term from the obstacle; ω H This refers to the weight of the hormone intensity item; Select the path with the highest score from all completed paths as the final planned path output for the i-th drone.

Images