Unmanned aerial vehicle three-dimensional path planning method based on improved IRRTstar

Abstract

The improved IRRTstar-based unmanned aerial vehicle three-dimensional path planning method relates to the technical field of unmanned aerial vehicle navigation and path optimization, and comprises the following steps: S1, a three-dimensional environment and wind field fusion model is constructed, including obstacle modeling and synthetic wind field modeling; S2, unmanned aerial vehicle flight constraint conditions are set, and voyage constraints and turning angle constraints are defined; S3, an initial path is generated based on the improved APF-IRRT* algorithm, and wind field APF sampling strategy, wind field adaptive ellipsoid region optimization strategy and energy-aware pruning strategy are fused; S4, the safety of the path is verified through wind field perception collision detection, and the optimal path is output.The present application is designed for the three-dimensional path planning requirement of unmanned aerial vehicles under wind field disturbance, can take into account the planning efficiency, path safety and energy consumption optimization, solves the technical problems of slow path planning and high flight energy consumption in actual operation, improves the task execution success rate; and the algorithm has high lightweight degree, small calculation cost, and meets the requirements of real-time performance and reliability in actual operation.

Description

Technical Field

[0001] This invention relates to the field of UAV navigation and path optimization technology, specifically to a UAV 3D path planning method based on an improved IRRTstar. Background Technology

[0002] With the rapid development of drone technology, its application in complex scenarios such as forest fire reconnaissance, mountain resource exploration, and canyon rescue is becoming increasingly widespread. These scenarios typically feature three-dimensional terrain heterogeneity, complex obstacle distribution (such as trees, fire sources, and no-fly zones), and are often affected by climate change, resulting in significant wind field disturbances, including multiple wind field components such as stable conventional winds, vertical wind shear, and random turbulence.

[0003] Among existing UAV 3D path planning algorithms, Rapidly-exploring Random Trees (RRT) ensures path optimality through random sampling and rewiring mechanisms, but suffers from slow convergence speed and low sampling efficiency in complex wind fields. Informed RRT* (IRRT*) reduces the search space and improves convergence efficiency through ellipsoidal constraints, but the standard ellipsoidal direction deviates from the actual optimal path direction in wind fields, leading to sampling redundancy. The Artificial Potential Field Fusion Rapidly-Exploring Random Tree (APF-RRT*) algorithm integrates the guiding characteristics of the Artificial Potential Field (APF) method, accelerating the exploration speed. However, traditional APF suffers from local minima traps, and the fixed gain coefficient cannot balance obstacle avoidance safety and path efficiency in strong wind environments, easily leading to path deviation or collision risks.

[0004] Meanwhile, existing algorithms generally lack energy optimization mechanisms for windy environments and fail to fully consider the impact of wind on UAV energy consumption. This results in planned paths that, while meeting obstacle avoidance requirements, suffer from excessive energy consumption and insufficient endurance. Furthermore, in scenarios with limited 3D height, the unreasonable allocation of sampling space further reduces planning efficiency. Therefore, there is an urgent need for a UAV 3D path planning method that can adapt to changes in wind fields, balance planning efficiency and safety, and optimize energy consumption. Summary of the Invention

[0005] This invention provides a UAV 3D path planning method based on an improved IRRTstar, aiming to solve the problems of slow convergence, low sampling efficiency, easy getting trapped in local minima, and insufficient energy consumption optimization of existing algorithms under wind field disturbances. The specific technical solution is as follows:

[0006] The UAV 3D path planning method based on the improved IRRTstar includes the following steps:

[0007] S1. Construct a 3D environment and wind field fusion model, including obstacle modeling and synthetic wind field modeling;

[0008] S2. Set flight constraints for the UAV, and clarify the range constraints and turning angle constraints;

[0009] S3. Generate the initial path based on the improved APF-IRRT* algorithm, and integrate the wind field APF sampling strategy, the wind field adaptive ellipsoid region optimization strategy and the energy-sensing pruning strategy.

[0010] S4. Verify path safety through wind field perception and collision detection, and output the optimal path;

[0011] The wind field adaptive APF sampling strategy includes: dynamically adjusting the gravity gain and repulsion gain according to the synthetic wind speed at the current location; generating deflection sampling points based on the adjusted potential field force; and fusing target sampling, in-ellipsoid Informed sampling, and APF deflection sampling with a preset probability to generate mixed sampling points.

[0012] The wind field adaptation ellipsoid region optimization strategy includes: introducing a height compression coefficient to compress the ellipsoid height half-axis to reduce height direction sampling redundancy; and correcting the ellipsoid principal axis direction according to the coordinates of the starting point, the target point, and the current synthetic wind speed vector to align the ellipsoid principal axis with the wind field reachable path.

[0013] The energy-sensing pruning strategy includes: constructing an energy cost model by integrating cruise energy consumption, take-off and landing energy consumption, and wind farm-assisted energy consumption; backtracking from the target node and pruning intermediate nodes of the path according to the energy cost threshold, while retaining the core path segment with the best energy.

[0014] Furthermore, in step S1, the obstacle modeling employs multiple types of geometric models to represent obstacles in complex scenes, including:

[0015] A spherical model, used to represent fire sources and small no-fly zones, satisfies the following formula:

[0016]

[0017] in, A collection of spherical obstacle regions; The current coordinates of the drone. Represents the components along the x, y, and z axes; The coordinates of the center of the spherical obstacle are: Represents the components along the x, y, and z axes; The radius of the spherical obstacle;

[0018] A conical model, used to represent forest trees, satisfies the following equation:

[0019]

[0020] in, A collection of conical obstacle regions; The current coordinates of the drone. Represents the components along the x, y, and z axes; The components of the center coordinates of the conical obstacle on the x, y, and z axes; The cone height represents the tree height. The radius of the cone's base;

[0021] A vertical barrier model, used to simulate the valley orientation, satisfies the following equation:

[0022]

[0023] in, This is a collection of vertical barrier regions; The current coordinates of the drone. Indicates the location of the drone. Components along the axial direction; The center point of the barrier is at Coordinate components along the axial direction; as a barrier The length of half a side along the axial direction, It represents the length of half a side on the x, y, and z axes.

[0024] Furthermore, in step S1, the synthetic wind field modeling constructs a dynamic wind field model by superimposing conventional wind field, wind shear, and turbulence, as shown below:

[0025] A typical wind field, simulating stable large-scale airflow, satisfies the following equation:

[0026]

[0027] in, This represents the wind speed vector in a typical wind field. The current spatial coordinates of the drone; For time; As a constant for the wind field, its direction is determined by the canyon's orientation. This represents the components of conventional wind on the x, y, and z axes;

[0028] The wind shear model describes the vertical variation of wind speed with altitude, satisfying the following equation:

[0029]

[0030] in, For wind shear in position Wind speed vector at the location; is the wind shear coefficient, which describes the rate at which wind speed increases with altitude; This is the current flight altitude of the drone; For surface roughness; It is a unit vector in the horizontal direction;

[0031] The turbulent disturbance model, simulating random wind field fluctuations, satisfies the following equation:

[0032]

[0033] in, For turbulent disturbances at position ,time Wind speed vector at the location; For the first The amplitude of each turbulent component; For the first The spatial frequency of the turbulent component; For the first The time frequency of each turbulent component; For the first The initial phase of each turbulent component; For the first The direction unit vector of each turbulent component;

[0034] In summary, the composite wind field satisfies the following equation:

[0035]

[0036] And correlate the drone's motion with the ground-to-air speed relationship:

[0037]

[0038] in, For the synthetic wind field at location ,time Wind speed vector at the location; This is the UAV's ground velocity vector; This is the airspeed vector of the UAV; To control the cycle.

[0039] Furthermore, in step S2, the range constraint is: , ,in, For the first Length of the path segment The minimum length of a single path segment. For the maximum permissible total voyage, This represents the total number of path segments;

[0040] The rotation constraint is: horizontal rotation angle. Vertical turning angle ,in, For the first Yaw angle of the segment path, For the maximum permissible yaw angle, For the first Pitch angle of the path segment The maximum permissible pitch angle;

[0041] The yaw angle satisfies: ,in, For the first The horizontal projection vector of the path segment. For the first The horizontal projection vector of the path segment;

[0042] Pitch angle satisfies: ,in, and These are two adjacent path segment vectors.

[0043] Furthermore, in step S3, the wind field APF sampling strategy is as follows:

[0044] Gain dynamic adjustment: based on wind speed normalization factor Adjusting the gravity gain With repulsive force gain ,in, The magnitude of the synthesized wind speed at the current location. The maximum flight speed of the drone, For the fundamental gain of gravity, For repulsive force base gain;

[0045] Generate deflection sampling points: , ,in, The coordinates of the deflection sampling points generated by the APF guide. These are the coordinates of the current random tree node. The step-size strength coefficient, Let the force vector be the force vector in the total potential field. For the target gravity vector, Let this be the repulsive force vector of the obstacle;

[0046] Generate hybrid sampling points: fuse target sampling, informed sampling, and APF sampling according to probability, and finally generate the coordinates of the sampling points. as follows:

[0047]

[0048] in, For uniformly distributed random variables, This represents the probability of directly obtaining the target point. The probability of Informed sampling within the ellipsoid. The coordinates of the target endpoint. These are the coordinates of the sampling point within the ellipsoid. These are the coordinates of the APF sampling points.

[0049] Furthermore, in step S3, the wind field adaptation ellipsoidal region optimization strategy is as follows:

[0050] Introducing a high compression factor Make the semi-axis of the ellipsoid height To reduce sampling redundancy in the height direction, among which, The horizontal minor axis;

[0051] Effective direction vector of the major axis of the ellipsoid Adjust the principal axis of the ellipsoid to align with the reachable path of the wind field, where... The wind field influence coefficient. The coordinates of the target endpoint. The coordinates of the starting point, This is the current synthesized wind speed vector.

[0052] Furthermore, in step S3, the energy cost model in the energy-aware pruning strategy is:

[0053]

[0054] in, The total energy cost of flying a certain distance; These are the weighting coefficients for each energy component; This provides the basic cruise power for the drone. The length of the current path segment; This refers to the airspeed of the drone. This represents the composite wind speed for the current path segment; For the quality of drones, It is the acceleration due to gravity; This represents the change in height of the path segment; The climbing efficiency coefficient; To reduce energy recovery efficiency;

[0055] The pruning process is as follows: backtracking from the target node, calculating the energy cost between the test node and the anchor point. ,like If there are no collisions, then prune intermediate nodes and retain the core path. The upper limit threshold for pruning decisions Based on the basic energy threshold, For test nodes The composite wind speed vector at that location. This is the maximum permissible wind speed.

[0056] Further, step S4 includes: uniformly sampling the path segment to generate several detection points, and simultaneously performing obstacle collision detection and wind speed over-limit detection on each detection point. If any detection point has a collision or the wind speed exceeds the preset wind speed safety threshold, the path is determined to have a collision risk and is replanned; otherwise, the safe and optimal path is output.

[0057] Furthermore, in step S4, the number of detection points is: ,in, The length of the path segment. For safe distance.

[0058] The beneficial effects of this invention are as follows:

[0059] 1. This invention uses a dynamic potential field gain adjustment and ellipsoidal direction correction mechanism to adapt path planning to wind field changes, reduce path deviation and improve obstacle avoidance success rate in strong wind environments. It can cope with various wind field types such as conventional wind, wind shear, and turbulence, as well as complex obstacles such as trees, fire sources, and valleys, and is suitable for multi-scenario operations.

[0060] 2. This invention reduces redundant sampling by using a hybrid sampling strategy and a highly compressed ellipsoid. Compared with traditional RRT* and IRRT* algorithms, it shortens the planning time and reduces the number of nodes, effectively improving planning efficiency. This invention also reduces the total energy consumption of the path through an energy-aware pruning strategy, significantly optimizing energy consumption and effectively improving the endurance of UAVs to meet the needs of long-distance operations.

[0061] 3. This invention retains the ability of random exploration through a hybrid sampling mechanism, enabling the algorithm to escape local minima in complex wind field environments, ensuring the probabilistic completeness of path planning, and ensuring that the algorithm can converge to a feasible path when a feasible solution exists.

[0062] 4. This invention can balance planning efficiency, path safety and energy consumption optimization, solve the technical problems of slow path planning and high flight energy consumption in actual operations, and improve the success rate of mission execution; moreover, the algorithm is highly lightweight, has low computational overhead, does not require high-performance hardware support, and is easy to deploy on existing UAV platforms.

[0063] 5. This invention is designed for the three-dimensional path planning needs of UAVs under wind disturbance. It features strong adaptability, high efficiency, and low energy consumption. It can be directly integrated into the navigation systems of equipment such as forest fire prevention UAVs, mountain inspection UAVs, and canyon rescue UAVs. It can also be extended to fields such as urban aerial delivery and mountain resource exploration. It does not rely on complex hardware upgrades. Performance can be improved simply through algorithm optimization. It is suitable for embedded platforms and edge computing devices and meets the requirements for real-time performance and reliability in actual operations. Attached Figure Description

[0064] Figure 1 This is a system framework diagram of the UAV three-dimensional path planning method of the present invention.

[0065] Figure 2 This is a schematic diagram of wind field environment and obstacle environment modeling in an embodiment of the present invention.

[0066] Figure 3 This is a schematic diagram illustrating the improved APF heuristic sampling principle of the UAV 3D path planning method of the present invention.

[0067] Figure 4 This is a schematic diagram of the generation of the hyperellipsoid region in this invention.

[0068] Figure 5 This is a schematic diagram illustrating the energy-sensing pruning principle of the present invention.

[0069] Figure 6 This is a three-dimensional diagram of the path planning experiment of the improved algorithm in a simulated wind field environment according to an embodiment of the present invention.

[0070] Figure 7 This is a top view of the path planning experiment of the improved algorithm in a simulated wind field environment according to an embodiment of the present invention. Detailed Implementation

[0071] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0072] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0073] Based on the improved IRRTstar UAV 3D path planning method, referring to Figure 1 As shown, it includes the following steps:

[0074] S1. Construct a 3D environment and wind field fusion model, including obstacle modeling and synthetic wind field modeling;

[0075] S2. Set flight constraints for the UAV, and clarify the range constraints and turning angle constraints;

[0076] S3. Generate the initial path based on the improved APF-IRRT* algorithm, and integrate the wind field adaptive APF sampling strategy, the wind field adapted ellipsoidal region optimization strategy and the energy-sensing pruning strategy.

[0077] S4. Verify path safety through wind field perception and collision detection, and output the optimal path.

[0078] The following is a detailed explanation of each step.

[0079] Step S1 is as follows.

[0080] I. Obstacle Modeling: Multiple types of geometric models are used to represent obstacles in complex scenes, including:

[0081] (a) A spherical model is used to represent fire sources and small no-fly zones, satisfying formula (1):

[0082]

[0083] in, A collection of spherical obstacle regions; The current coordinates of the drone. Represents the components along the x, y, and z axes; The coordinates of the center of the spherical obstacle are: Represents the components along the x, y, and z axes; Let be the radius of the spherical obstacle.

[0084] (ii) A conical model, used to represent forest trees, satisfies formula (2):

[0085]

[0086] in, A collection of conical obstacle regions; The current coordinates of the drone. Represents the components along the x, y, and z axes; The coordinates of the center (center of the base) of the conical obstacle are represented by the components of the x, y, and z axes. The height of the cone (i.e., the tree height); Let be the radius of the base of the cone.

[0087] (iii) A vertical barrier model is used to simulate the valley orientation, satisfying formula (3):

[0088]

[0089] in, This is a collection of vertical barrier regions; The current coordinates of the drone. Indicates the location of the drone. Components along the axial direction; The center point of the barrier is at Coordinate components along the axial direction; as a barrier The length of half a side along the axial direction, It represents the length of half a side on the x, y, and z axes.

[0090] II. Synthetic Wind Field Modeling: By superimposing conventional wind field, wind shear, and turbulence, a dynamic wind field model is constructed, as follows:

[0091] (a) Conventional wind field, simulating stable large-scale airflow, satisfies formula (4):

[0092]

[0093] in, This represents the wind speed vector in a typical wind field. The current spatial coordinates of the drone; For time; As a constant for the wind field, its direction is determined by the canyon's orientation. This represents the components of conventional wind on the x, y, and z axes.

[0094] (ii) Wind shear model, which describes the vertical variation of wind speed with altitude, satisfies formula (5):

[0095]

[0096] in, For wind shear in position Wind speed vector at the location; is the wind shear coefficient, which describes the rate at which wind speed increases with altitude; The current flight altitude of the drone (i.e. (z-axis component); For surface roughness, take ; It is a unit vector in the horizontal direction.

[0097] (iii) Turbulent disturbance model, simulating random wind field fluctuations, satisfies formula (6):

[0098]

[0099] in, For turbulent disturbances at position ,time Wind speed vector at the location; For the first The amplitude of each turbulent component; For the first The spatial frequency of the turbulent component; For the first The time frequency of each turbulent component; For the first The initial phase of each turbulent component; For the first The direction unit vector of each turbulent component.

[0100] (iv) In summary, the composite wind field satisfies formula (7):

[0101]

[0102] And by relating the ground-air velocity relationship to the UAV motion, it satisfies formula (8):

[0103]

[0104] in, For the synthetic wind field at location ,time Wind speed vector at the location; This is the UAV's ground velocity vector; This is the airspeed vector of the UAV; To control the cycle, take .

[0105] Step S2 is as follows.

[0106] I. Setting range constraints: , ,in, For the first Length of the path segment The minimum length of a single path segment. For the maximum permissible total voyage, This represents the total number of path segments.

[0107] II. Setting Corner Constraints: Horizontal Corner Vertical turning angle ,in, For the first Yaw angle of the segment path, For the maximum permissible yaw angle, For the first Pitch angle of the path segment This is the maximum permissible pitch angle.

[0108] The yaw angle satisfies formula (9):

[0109]

[0110] in, For the first The horizontal projection vector of the path segment. For the first The horizontal projection vector of the path segment;

[0111] The pitch angle satisfies formula (10):

[0112]

[0113] in, and These are two adjacent path segment vectors.

[0114] Step S3 is as follows.

[0115] I. Wind field adaptive APF sampling strategy, including:

[0116] (i) Dynamic gain adjustment: based on wind speed normalization factor Adjusting the gravity gain With repulsive force gain ,in, The magnitude of the synthesized wind speed at the current location. The maximum flight speed of the drone, For the fundamental gain of gravity, This is the basic gain of the repulsive force.

[0117] (ii) Generating deflection sampling points: , ,in, The coordinates of the deflection sampling points generated by the APF guide. These are the coordinates of the current random tree node. Let be the step-size strength coefficient, taken as... , Let the force vector be the force vector in the total potential field. For the target gravity vector, This is the repulsive force vector of the obstacle.

[0118] (III) Generating Hybrid Sampling Points: Fusion of target sampling, informed sampling, and APF sampling according to probability, resulting in the coordinates of the final sampling points. as follows:

[0119]

[0120] in, For uniformly distributed random variables, This represents the probability of directly obtaining the target point. The probability of Informed sampling within the ellipsoid. The coordinates of the target endpoint. These are the coordinates of the sampling point within the ellipsoid. These are the coordinates of the APF sampling points.

[0121] Figure 3The diagram illustrates the improved APF heuristic sampling principle, demonstrating the sampling point generation mechanism. Here, prve is the current node, X_random is the generated random point, and X_new is the new node; f_rep is the repulsive force of obstacles within the range, f_att is the gravitational force of the target point on the current point, and f_total is the vector sum of the gravitational and repulsive forces; f_1 is a vector with the same direction as f_total, and f_2 is the gravitational force of the random point on the current point; the combined force of both points towards the new node.

[0122] II. Wind field adaptation ellipsoid region optimization strategy, including:

[0123] (a) Three-dimensional height compression: Introducing a height compression coefficient Make the semi-axis of the ellipsoid height To reduce sampling redundancy in the height direction, among which, It is the horizontal minor axis.

[0124] (ii) Ellipsoidal orientation correction: the effective direction vector through the major axis of the ellipsoid Adjust the principal axis of the ellipsoid to align with the reachable path of the wind field, where... Let be the wind field influence coefficient, taken as... , The coordinates of the target endpoint. The coordinates of the starting point, This is the current synthesized wind speed vector.

[0125] In step S3, the generation of the wind field adaptation ellipsoid region is as follows: Figure 4 As shown. The initial direction of the major axis of the ellipsoid is determined by the direction from the starting point to the target point. Half the distance between the two points is 'a', and half the length of the minor axis is 'b'. Multiplying by (taking 0.28) yields the polar half-axis c.

[0126] III. Energy-sensing pruning strategies, including:

[0127] (i) Taking into account cruise energy consumption, take-off and landing energy consumption, and wind farm auxiliary energy consumption, an energy cost model is constructed:

[0128]

[0129] in, The total energy cost of flying a certain distance; These are the weighting coefficients for each energy component; This provides the basic cruise power for the drone. The length of the current path segment; This refers to the airspeed of the drone. This represents the composite wind speed for the current path segment; For the quality of drones, It is the acceleration due to gravity; This represents the change in height of the path segment; The climbing efficiency coefficient; To reduce energy recovery efficiency.

[0130] (ii) Pruning process: Backtracking from the target node, calculate the energy cost between the test node and the anchor point. ,like If there are no collisions, then prune intermediate nodes and retain the core path. The upper limit threshold for pruning decisions Based on the basic energy threshold, For test nodes The composite wind speed vector at that location. This is the maximum permissible wind speed.

[0131] Figure 5 The diagram shows the principle of energy-sensing pruning. The blue path represents the initial path, the red dashed path represents the path segment where pruning failed, and the red solid path represents the path after successful pruning.

[0132] Step S4 is as follows.

[0133] Wind field sensing collision detection:

[0134] For path segments Perform uniform sampling to generate There are several detection points, if any one of the detection points satisfies... or If the path is not found to be at risk of collision, it is determined that the path is replanned; otherwise, the safest and optimal path is output. and These are the start and end nodes of the path segment (used for collision detection). The length of the path segment. For a safe distance, For collision detection function, the detection point Is it inside any obstacle? The wind speed at the detection point. This is the wind speed safety threshold.

[0135] Example:

[0136] The implementation process will be explained in detail below with reference to a specific experimental scenario.

[0137] I. Experimental Environment Configuration

[0138] Hardware platform: CPU is Intel i5-9300H 2.40GHz, simulation software is PyCharm2025.

[0139] 3D scene parameters: Canyon forest terrain, starting coordinates End point coordinates ,Include Tall trees (conical model) , ), One fire source (spherical model) ), A valley barrier (vertical barrier model) , , (and several no-fly zones).

[0140] Wind field parameters: conventional wind Wind shear coefficient turbulent amplitude spatial frequency Time frequency .

[0141] UAV parameters: Maximum pitch angle Maximum yaw angle Minimum path segment length Maximum range .

[0142] Figure 2 The diagram shows a model of the wind field environment and obstacle environment. Obstacles and their types are labeled, arrows represent wind field vectors, and arrow color indicates wind speed.

[0143] II. Specific Implementation Steps

[0144] (I) Model initialization (corresponding to steps S1-S2)

[0145] An obstacle model is constructed based on scene parameters, and synthetic wind field data is generated using formulas (4)-(8); constraint parameters are set: , , , Collision detection safety distance Wind field threshold .

[0146] (II) Improve the APF-IRRT algorithm operation (corresponding to step S3)

[0147] 1. Initial Iteration Phase:

[0148] Random tree initialization, starting from... The target point is ;

[0149] Sampling points are generated using a hybrid sampling mechanism in the initial stage. Prioritize Informed ellipsoid sampling to quickly explore feasible regions.

[0150] 2. Path optimization stage:

[0151] After finding the initial solution, calculate the optimal cost. Update the ellipsoid parameters and apply height compression and wind direction correction;

[0152] Perform energy-sensing pruning and set... Backtrack and prune redundant nodes, retaining the energy-optimal path segment.

[0153] 3. Dynamic Adjustment Phase:

[0154] Real-time wind field data is acquired, and the APF gain coefficient and ellipsoid orientation are adjusted. If a sudden change in wind speed is detected ( ), reduce the repulsion gain and increase the sampling density.

[0155] Figures 6-7 The image shows a visualization of the improved algorithm's path planning experiment in a simulated wind field environment. The thin red line represents the original path, the thick blue line represents the pruned path, and the thin green line represents the branches generated by the algorithm.

[0156] (III) Path verification and output (corresponding to step S4)

[0157] Perform wind field sensing collision detection on the optimized path to generate 20 detection points / path segments, and verify that there are no collisions and the wind speed does not exceed the limit;

[0158] Output the key parameters of the final path: path length, planning time, number of nodes, and total energy consumption, and compare them with traditional RRT*, IRRT*, and APF-RRT* algorithms.

[0159] Thirty experiments were conducted, and the planning time, path length, number of nodes, path standard deviation, and sampling success rate were compared. To minimize errors, the experimental environment was kept consistent. The comparison results are shown in the table below.

[0160]

[0161] III. Comparison of Experimental Results

[0162] Compared to traditional algorithms, the APF-IRRT* algorithm in this embodiment not only shortens the planning time and reduces the number of nodes, but also shortens the path length, reduces the total energy consumption, improves the obstacle avoidance success rate, and effectively controls the path offset.

[0163] Experimental results show that the algorithm of this invention outperforms existing algorithms in terms of planning efficiency, path quality, energy consumption control and security, and is suitable for complex three-dimensional path planning tasks in wind farm environments.

[0164] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for three-dimensional path planning of unmanned aerial vehicles based on improved IRRTstar, characterized in that, Includes the following steps: S1. Construct a 3D environment and wind field fusion model, including obstacle modeling and synthetic wind field modeling; S2. Set flight constraints for the UAV, and clarify the range constraints and turning angle constraints; S3. Generate the initial path based on the improved APF-IRRT* algorithm, and integrate the wind field APF sampling strategy, the wind field adaptive ellipsoid region optimization strategy and the energy-sensing pruning strategy. S4. Verify path safety through wind field perception and collision detection, and output the optimal path; The wind field adaptive APF sampling strategy includes: dynamically adjusting the gravity gain and repulsion gain according to the synthetic wind speed at the current location; generating deflection sampling points based on the adjusted potential field force; and fusing target sampling, in-ellipsoid Informed sampling, and APF deflection sampling with a preset probability to generate mixed sampling points. The wind field adaptation ellipsoid region optimization strategy includes: introducing a height compression coefficient to compress the ellipsoid height half-axis to reduce height direction sampling redundancy; and correcting the ellipsoid principal axis direction according to the coordinates of the starting point, the target point, and the current synthetic wind speed vector to align the ellipsoid principal axis with the wind field reachable path. The energy-sensing pruning strategy includes: constructing an energy cost model by integrating cruise energy consumption, take-off and landing energy consumption, and wind farm-assisted energy consumption; backtracking from the target node and pruning intermediate nodes of the path according to the energy cost threshold, while retaining the core path segment with the best energy.

2. The UAV 3D path planning method based on the improved IRRTstar according to claim 1, characterized in that, In step S1, the obstacle modeling employs multiple types of geometric models to represent obstacles in complex scenes, including: A spherical model, used to represent fire sources and small no-fly zones, satisfies the following formula: in, A collection of spherical obstacle regions; The current coordinates of the drone. Represents the components along the x, y, and z axes; The coordinates of the center of the spherical obstacle are: Represents the components along the x, y, and z axes; The radius of the spherical obstacle; A conical model, used to represent forest trees, satisfies the following equation: in, A collection of conical obstacle regions; The current coordinates of the drone. Represents the components along the x, y, and z axes; The components of the center coordinates of the conical obstacle on the x, y, and z axes; The cone height represents the tree height. The radius of the cone's base; A vertical barrier model, used to simulate the valley orientation, satisfies the following equation: in, This is a collection of vertical barrier regions; The current coordinates of the drone. Indicates the location of the drone. Components along the axial direction; The center point of the barrier is at Coordinate components along the axial direction; as a barrier The length of half a side along the axial direction, It represents the length of half a side on the x, y, and z axes.

3. The UAV 3D path planning method based on the improved IRRTstar according to claim 1, characterized in that, In step S1, the synthetic wind field modeling constructs a dynamic wind field model by superimposing conventional wind field, wind shear, and turbulence, as shown below: A typical wind field, simulating stable large-scale airflow, satisfies the following equation: in, This represents the wind speed vector in a typical wind field. The current spatial coordinates of the drone; For time; As a constant for the wind field, its direction is determined by the canyon's orientation. This represents the components of conventional wind on the x, y, and z axes; The wind shear model describes the vertical variation of wind speed with altitude, satisfying the following equation: in, For wind shear in position Wind speed vector at the location; is the wind shear coefficient, which describes the rate at which wind speed increases with altitude; This is the current flight altitude of the drone; For surface roughness; It is a unit vector in the horizontal direction; The turbulent disturbance model, simulating random wind field fluctuations, satisfies the following equation: in, For turbulent disturbances at position ,time Wind speed vector at the location; For the first The amplitude of each turbulent component; For the first The spatial frequency of the turbulent component; For the first The time frequency of each turbulent component; For the first The initial phase of each turbulent component; For the first The direction unit vector of each turbulent component; In summary, the composite wind field satisfies the following equation: And correlate the drone's motion with the ground-to-air speed relationship: in, For the synthetic wind field at location ,time Wind speed vector at the location; This is the UAV's ground velocity vector; This is the airspeed vector of the UAV; To control the cycle.

4. The UAV 3D path planning method based on the improved IRRTstar according to claim 1, characterized in that, In step S2, the range constraint is: , ,in, For the first Length of the path segment The minimum length of a single path segment. For the maximum permissible total voyage, This represents the total number of path segments; The rotation constraint is: horizontal rotation angle. Vertical turning angle ,in, For the first Yaw angle of the segment path, For the maximum permissible yaw angle, For the first Pitch angle of the path segment Maximum permissible pitch angle; The yaw angle satisfies: ,in, For the first The horizontal projection vector of the path segment. For the first The horizontal projection vector of the path segment; Pitch angle satisfies: ,in, and These are two adjacent path segment vectors.

5. The UAV 3D path planning method based on the improved IRRTstar according to claim 1, characterized in that, In step S3, the wind field APF sampling strategy is as follows: Gain dynamic adjustment: based on wind speed normalization factor Adjusting the gravity gain With repulsive force gain ,in, The magnitude of the synthesized wind speed at the current location. The maximum flight speed of the drone, For the fundamental gain of gravity, For repulsive force base gain; Generate deflection sampling points: , ,in, The coordinates of the deflection sampling points generated by the APF guide. These are the coordinates of the current random tree node. The step-size strength coefficient, Let the force vector be the force vector in the total potential field. For the target gravity vector, Let this be the repulsive force vector of the obstacle; Generate hybrid sampling points: fuse target sampling, informed sampling, and APF sampling according to probability, and finally generate the coordinates of the sampling points. as follows: in, For uniformly distributed random variables, This represents the probability of directly obtaining the target point. The probability of Informed sampling within the ellipsoid. The coordinates of the target endpoint. These are the coordinates of the sampling point within the ellipsoid. These are the coordinates of the APF sampling points.

6. The UAV 3D path planning method based on the improved IRRTstar according to claim 1, characterized in that, In step S3, the wind field adaptation ellipsoidal region optimization strategy is as follows: Introducing a high compression factor Make the semi-axis of the ellipsoid height To reduce sampling redundancy in the height direction, among which, The horizontal minor axis; Effective direction vector of the major axis of the ellipsoid Adjust the principal axis of the ellipsoid to align with the reachable path of the wind field, where... The wind field influence coefficient. The coordinates of the target endpoint. The coordinates of the starting point, This is the current synthesized wind speed vector.

7. The UAV 3D path planning method based on the improved IRRTstar according to claim 1, characterized in that, In step S3, the energy cost model in the energy-aware pruning strategy is: in, The total energy cost of flying a certain distance; These are the weighting coefficients for each energy component; This provides the basic cruise power for the drone. The length of the current path segment; This refers to the airspeed of the drone. This represents the composite wind speed for the current path segment; For the quality of drones, It is the acceleration due to gravity; This represents the change in height of the path segment; The climbing efficiency coefficient; To reduce energy recovery efficiency; The pruning process is as follows: backtracking from the target node, calculating the energy cost between the test node and the anchor point. ,like If there are no collisions, then prune intermediate nodes and retain the core path. The upper limit threshold for pruning decisions Based on the basic energy threshold, For test nodes The composite wind speed vector at that location. This is the maximum permissible wind speed.

8. The UAV 3D path planning method based on the improved IRRTstar according to claim 1, characterized in that, Step S4 includes: uniformly sampling the path segment to generate several detection points, and simultaneously performing obstacle collision detection and wind speed over-limit detection on each detection point. If any detection point has a collision or the wind speed exceeds the preset wind speed safety threshold, the path is determined to have a collision risk and is replanned; otherwise, the safe and optimal path is output.

9. The UAV 3D path planning method based on the improved IRRTstar according to claim 8, characterized in that, In step S4, the number of detection points is: ,in, The length of the path segment. For a safe distance.

Images