Layered cooperative three-dimensional path planning method for unmanned aerial vehicle inspection of converter valve hall

Abstract

This application belongs to the field of UAV path planning technology and provides a hierarchical collaborative 3D path planning method for UAVs in converter valve hall inspection. The method includes: first, equating the valve tower to a square column and the casing to a cylinder, constructing a 3D occupying grid and expanding it to obtain a feasible flight domain; within the domain, constructing a total cost function including path length, safety, altitude window, altitude difference, angle smoothing, and electromagnetic penalty; then, running the A* algorithm on a low-resolution grid to generate a global coarse path and construct a guiding corridor; under corridor constraints, using the RRT* algorithm for bias sampling and reconnection optimization to obtain a continuous and smooth reference path; finally, using DWA in velocity space for dynamic feasibility screening and safety assessment to achieve local tracking and online obstacle avoidance. This application effectively reduces the total path cost and planning time fluctuations, improves the success rate and safety distance, reduces the maximum turning angle, and enhances planning stability and executability in the narrow environment of converter valve halls.

Description

Technical Field

[0001] This application belongs to the field of UAV path planning technology, and in particular relates to a hierarchical collaborative three-dimensional path planning method, device, terminal equipment and storage medium for UAVs for converter valve hall inspection. Background Technology

[0002] As a key hub in the high-voltage direct current (HVDC) transmission system, the converter station's converter valve hall is densely packed with equipment and has a complex structure. The arrayed arrangement of converter valve towers and valve-side bushings creates numerous narrow passages and obstructed areas, resulting in an inspection environment characterized by "limited space, numerous obstacles, and narrow passable areas." Currently, commonly used three-dimensional path planning methods mainly include global planning algorithms based on graph search, randomized planning algorithms based on sampling, and dynamic planning methods for local obstacle avoidance and tracking control.

[0003] However, current methods suffer from the problem that a single algorithm cannot simultaneously take into account global reachability, path quality, and real-time obstacle avoidance capabilities, resulting in low planning efficiency, large fluctuations in solution quality, and difficulty in making the trajectory executable online. This fails to meet the comprehensive requirements of UAV inspection for safety, planning efficiency, and trajectory stability. Summary of the Invention

[0004] This application provides a hierarchical collaborative 3D path planning method, device, terminal equipment, and storage medium for UAVs in converter valve hall inspection. It can solve the problems of current methods, where a single algorithm cannot simultaneously take into account global reachability, path quality, and real-time obstacle avoidance capabilities, resulting in low planning efficiency, large fluctuations in solution quality, and difficulty in online trajectory execution, thus failing to meet the comprehensive requirements of UAV inspection for safety, planning efficiency, and trajectory stability.

[0005] In a first aspect, embodiments of this application provide a hierarchical collaborative 3D path planning method for UAVs in converter valve hall inspection, comprising: S1, establishing a 3D environment model of the converter valve hall, equating the converter valve tower of the converter valve hall to a cuboid obstacle, equating the valve side sleeve of the converter valve hall to a cylindrical obstacle, and constructing a 3D occupancy grid; S2, performing safety gap expansion on the 3D occupancy grid to obtain a feasible flight domain that satisfies the minimum safety gap; S3, constructing a total cost function within the feasible flight domain, and transforming path planning into a total cost minimization problem under safety gap and altitude window constraints; S4, within the feasible flight domain, based on the total cost function, adopting... S5. Using the A* algorithm, generate a first global reference path from the start point to the end point; S6. Guided by the first global path and with the total cost function as the optimization objective, use the RRT* algorithm for bias sampling, expansion, and reconnection optimization to obtain a continuous and smooth second global reference path; S7. Use the Dynamic Window Method (DWA) to dynamically screen and assess the safety of candidate velocities in the velocity space, perform local tracking of the global reference path and achieve online obstacle avoidance, and output an executable inspection track; S8. When at least one preset condition is detected, trigger closed-loop replanning, update the 3D occupancy grid, and repeat steps S4 to S6 until the UAV reaches the end point.

[0006] Optionally, in another possible implementation of the first aspect, S3 above involves constructing a total cost function within the feasible flight domain and transforming path planning into a total cost minimization problem under safety clearance and altitude window constraints, specifically including:

[0007] S301. Calculate the path length cost by summing the Euclidean distances between adjacent discrete waypoints:

[0008]

[0009] In the formula, For path length cost, For the first on the path discrete location points and N is the number of waypoints;

[0010] S302. Calculate the safety cost by calculating the minimum distance between any path point and the set of obstacles, and based on the minimum safety gap. With buffer bandwidth Construct a segmented penalty function to obtain the full path safety cost:

[0011] set up Given a set of obstacles, for any path segment To the obstacle The minimum distance is denoted as:

[0012]

[0013] Define minimum safety clearance and buffer bandwidth Minimum safety clearance Composed of the equivalent radius of the drone and the safety margin, the single-segment safety penalty is defined as a piecewise function:

[0014]

[0015] based on The safety cost obtained from a single-segment safety penalty is:

[0016]

[0017] S303. Calculate the cost of height constraints, including absolute height range constraints and height difference constraints between adjacent path points:

[0018] Let the permissible flight altitude range be For each path point height Define and sum the height constraint costs to obtain the absolute height range constraints. :

[0019]

[0020]

[0021] Based on each waypoint height Thus, the height difference constraint is obtained:

[0022]

[0023] In the formula, The height window penalty is applied to the i-th path point;

[0024] S304, Calculate the cost of angle smoothing:

[0025] Define the adjacent path segment vector as:

[0026] The steering angle is defined as:

[0027]

[0028] In the formula, A constant greater than 0;

[0029] The climb angle is defined as:

[0030]

[0031] The cost of angle smoothing is:

[0032]

[0033] In the formula, , Preset weights;

[0034] S305, Calculate the electromagnetic penalty cost:

[0035] The spatial electromagnetic risk field strength index is defined as follows:

[0036]

[0037] In the formula, , Positions The electric and magnetic field strengths at that location , This is a normalized reference value. , Preset weighting coefficients;

[0038] Set two threshold levels: safety threshold No-fly threshold The electromagnetic environment penalty term is then defined as a piecewise function:

[0039]

[0040] In the formula, The penalty index;

[0041] For path points The cumulative electromagnetic penalty results in the following electromagnetic penalty cost:

[0042]

[0043] S306. Construct the total cost function as follows:

[0044]

[0045] In the formula, The preset cost weighting coefficient.

[0046] Optionally, in another possible implementation of the first aspect, S4 above, within the feasible flight domain, uses the A* algorithm based on the total cost function to generate a first global reference path from the start point to the end point, as follows:

[0047] S401. Within the feasible flight domain, using the starting point as the initial node and the ending point as the target node, node expansion is performed using an evaluation function that includes heuristics:

[0048]

[0049] in, The cost is the path length from the starting point to the current node;

[0050] S402, to To extend this further, we introduce a composite cost function and consider multidimensional constraints. Defined as:

[0051]

[0052] In the formula, The threat cost represents the nearest grid distance from a node to the threat region. This represents the altitude difference cost between nodes and their neighbors, used to limit drastic changes in flight altitude. Threat distance cost weighting, Threat distance cost weighting;

[0053] S403 The heuristic distance estimate from the current node to the target node is calculated using three-dimensional Euclidean distance:

[0054]

[0055] In the formula, , , They are nodes of coordinate, coordinates and coordinate; , , These are the target nodes. coordinate, coordinates and coordinate;

[0056] S404. The node expansion is guided by the evaluation function until the target node is found, and the first global reference path composed of discrete nodes is obtained by backtracking.

[0057] Optionally, in another possible implementation of the first aspect, S5 above, guided by the first global path and with the total cost function as the optimization objective, employs the RRT* algorithm for bias sampling, expansion, and reconnection optimization to obtain a continuous and smooth second global reference path, specifically including:

[0058] S501. Construct a global guidance corridor with a preset width, centered on the first global path;

[0059] S502. Under the constraint of the global guidance corridor, bias sampling is performed: the sampling points fall into the guidance corridor with a higher probability than uniform sampling, and the remaining sampling points are randomly generated in the feasible flight domain to guide the random tree to grow in the target direction.

[0060] S503. Based on the sampling points generated by bias sampling, perform nearest neighbor node search and new node expansion, generate new nodes and add them to the random tree;

[0061] S504. Using the total cost function as the optimization objective, traverse the nearest neighbor nodes within a preset radius near the new node, and determine whether connecting through the new node can reduce the total path cost of the nearest neighbor nodes. If so, update the parent node connection relationship of the nearest neighbor nodes.

[0062] S505. Repeat steps S502 to S504 until the distance between a node in the random tree and the target node is less than a first preset threshold. Backtrack to obtain a discrete point sequence composed of continuous path points, which serves as the second global reference path; wherein, the discrete point sequence is represented as:

[0063]

[0064] In the formula, and All of these are the current poses of the drone, i.e. , For each local target point in the sequence It represents a location in three-dimensional space.

[0065] Optionally, in another possible implementation of the first aspect, S6 above employs the dynamic window method DWA to dynamically screen and assess the feasibility of candidate velocities in the velocity space, performs local tracking of the global reference path and implements online obstacle avoidance, and outputs an executable inspection track, specifically including:

[0066] S601, Select distance from current pose point As a local guiding point, that is:

[0067]

[0068] S602. Based on the dynamic constraints of the UAV, generate a candidate velocity set in the velocity space. And perform forward simulation on each candidate velocity to simulate the trajectory of the UAV within a preset time window;

[0069] S603. Perform dynamic feasibility screening on the candidate trajectories obtained from the forward simulation, eliminating those that exceed the dynamic limits of the UAV, exceed the height constraints, or are combined with obstacles. Candidate trajectories whose safe distance is less than the second preset threshold;

[0070] S604. Construct an evaluation function for the candidate trajectories that have passed the feasibility screening. :

[0071]

[0072] In the formula, For candidate velocity unit vectors, The unit direction pointing to the RRT* guide point. For each candidate velocity Uniform forward simulation, directional consistency weights Safety distance weight and speed weight satisfy and ;

[0073] S605. Select the candidate velocity with the best evaluation function value as the current control variable, perform local trajectory tracking, and output the executable inspection track segment.

[0074] S606. Repeat steps S601 to S605 in each control cycle until the target node is reached.

[0075] Optionally, in another possible implementation of the first aspect, the preset conditions in S7 above include: dynamic obstacles entering a preset warning distance, the deviation between the local trajectory and the global reference path exceeding a threshold, and the dynamic window method DWA failing to generate a feasible local trajectory that satisfies the safety distance constraint within a preset time window.

[0076] Secondly, this application provides a hierarchical collaborative 3D path planning device for UAVs used in converter valve hall inspection, comprising: an environment modeling module for establishing a 3D environment model of the converter valve hall, equating the converter valve tower of the converter valve hall to a prismatic obstacle and the valve side sleeve of the converter valve hall to a cylindrical obstacle, and constructing a 3D occupancy grid; a feasible domain construction module for expanding the safety gap of the 3D occupancy grid to obtain a feasible flight domain that satisfies the minimum safety gap; a total cost function construction module for constructing a total cost function within the feasible flight domain and transforming path planning into a total cost minimization problem under the constraints of safety gap and altitude window; and a first path generation module for generating a path within the feasible flight domain based on the total cost function using A... The algorithm generates a first global reference path from the start point to the end point; a second path generation module, guided by the first global path and with the total cost function as the optimization objective, uses the RRT* algorithm for bias sampling, expansion, and reconnection optimization to obtain a continuous and smooth second global reference path; a dynamic obstacle avoidance module, using the Dynamic Window Method (DWA) to dynamically screen and assess the safety of candidate velocities in the velocity space, performs local tracking of the global reference path and achieves online obstacle avoidance, and outputs an executable inspection track; an adaptive replanning module, when at least one preset condition is detected, triggers closed-loop replanning, updates the 3D occupancy grid, and repeatedly executes the steps corresponding to the first path generation module to the dynamic obstacle avoidance module until the UAV reaches the end point.

[0077] Thirdly, embodiments of this application provide a terminal device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the aforementioned hierarchical collaborative three-dimensional path planning method for UAV inspection of converter valve hall.

[0078] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the aforementioned hierarchical collaborative three-dimensional path planning method for UAV inspection of converter valve halls.

[0079] Beneficial Effects: This application's solution incorporates complex inspection constraints into an optimization framework by constructing a total cost function that includes path length, safety distance, altitude window, altitude difference, angle smoothing, and electromagnetic penalty. This effectively balances flight efficiency, safety, and controllability. Secondly, the A* algorithm is used to quickly generate a globally discrete coarse path on a low-resolution grid and construct a guiding corridor, providing a clear boundary for subsequent continuous spatial searches and significantly compressing invalid sampling areas. Based on this, the RRT* algorithm performs bias sampling and reconnection optimization within the guiding corridor, significantly improving convergence speed and path quality in narrow corridor scenarios and effectively suppressing detours and suboptimal paths caused by random expansion. Finally, dynamic... The window-based DWA performs dynamic feasibility screening and safety assessment in the velocity space, achieving smooth tracking and online obstacle avoidance of the global reference path. This ensures both global optimality and meets real-time response requirements. In narrow, highly constrained valve hall environments, this method can stably maintain sufficient safety margins, with smooth path turning, strong controllability, high success rate, short planning time, and small fluctuations. The overall path quality and stability are significantly better than single algorithms or conventional metaheuristic methods. In addition, through an adaptive replanning mechanism, this method can trigger local repair or global replanning when dynamic obstacles appear, forming a closed-loop control framework, further improving the robustness and reliability of inspection under complex working conditions. Attached Figure Description

[0080] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0081] Figure 1 This is a flowchart illustrating a hierarchical collaborative three-dimensional path planning method for UAV inspection of converter valve halls, provided in one embodiment of this application.

[0082] Figure 2 This is a three-dimensional view of the converter station valve hall provided in one embodiment of this application;

[0083] Figure 3 This is a flowchart of the overall path planning provided in one embodiment of this application;

[0084] Figure 4 This is a composite radar image provided in one embodiment of this application;

[0085] Figure 5 This is a stacked bar chart of success rate provided in one embodiment of this application;

[0086] Figure 6 This is a box plot of the minimum safety clearance provided in one embodiment of this application;

[0087] Figure 7 This is a box gating diagram of the maximum rotation angle provided in one embodiment of this application;

[0088] Figure 8 This is a schematic diagram of the CDF curve for the safety clearance provided in an embodiment of this application;

[0089] Figure 9 This is a schematic diagram of the structure of a UAV hierarchical collaborative three-dimensional path planning device for inspecting converter valve halls, provided in one embodiment of this application;

[0090] Figure 10 This is a schematic diagram of the structure of the terminal device provided in the embodiments of this application. Detailed Implementation

[0091] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0092] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0093] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0094] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0095] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0096] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0097] The following description, with reference to the accompanying drawings, details a hierarchical collaborative three-dimensional path planning method, device, terminal equipment, and storage medium for UAV inspection of converter valve halls provided in this application.

[0098] Figure 1 The diagram shows a flowchart of a UAV hierarchical collaborative three-dimensional path planning method for converter valve hall inspection provided in an embodiment of this application.

[0099] like Figure 1 As shown, the UAV hierarchical collaborative 3D path planning method for converter valve hall inspection includes the following steps:

[0100] S1. Establish a three-dimensional environment model of the converter valve hall, and treat the converter valve tower in the converter valve hall as a square column obstacle and the valve side sleeve in the converter valve hall as a cylindrical obstacle, and construct a three-dimensional occupation grid.

[0101] In one embodiment, such as Figure 2 As shown, the converter station valve hall is set as a cubic space of 200*100*50 (m), where the gray cylinders are converter valves and the red cylinders are converter transformer valve side bushings. Their spatial data in the converter station valve hall and their own data are shown in Tables 1 and 2.

[0102] Table 1

[0103] Table 2

[0104] S2. Expand the safety gap of the three-dimensional occupied grid to obtain a feasible flight domain that meets the minimum safety gap.

[0105] For example, the minimum safety clearance can be 10 meters, but this application does not limit it.

[0106] S3. Construct the total cost function within the feasible flight domain and transform the path planning into a total cost minimization problem under the constraints of safety distance and altitude window;

[0107] Furthermore, in this embodiment of the application, step S3 includes:

[0108] S301. Calculate the path length cost by summing the Euclidean distances between adjacent discrete waypoints:

[0109]

[0110] In the formula, For path length cost, For the first on the path discrete location points and N is the number of waypoints;

[0111] In this embodiment, path length directly affects flight time and energy consumption in UAV inspection missions. Under the premise of satisfying obstacle avoidance and flightability constraints, a shorter path generally means higher inspection efficiency. Therefore, this application uses path length as a basic cost term, calculating the path length cost by accumulating the Euclidean distances between adjacent discrete waypoints.

[0112] S302. Calculate the safety cost by calculating the minimum distance between any path point and the set of obstacles, and based on the minimum safety gap. With buffer bandwidth Construct a segmented penalty function to obtain the full path safety cost:

[0113] set up Given a set of obstacles, for any path segment To the obstacle The minimum distance is denoted as:

[0114]

[0115] Define minimum safety clearance and buffer bandwidth Minimum safety clearance Composed of the equivalent radius of the drone and the safety margin, the single-segment safety penalty is defined as a piecewise function:

[0116]

[0117] based on The safety cost obtained from a single-segment safety penalty is:

[0118]

[0119] In this embodiment of the application, the safety cost is a core constraint that must be met in the UAV power inspection. Its goal is to avoid collisions between the UAV and power equipment such as converter valves and converter transformer side bushings during the entire inspection process, so as to prevent equipment damage.

[0120] S303. Calculate the cost of height constraints, including absolute height range constraints and height difference constraints between adjacent path points:

[0121] Let the permissible flight altitude range be For each path point height Define and sum the height constraint costs to obtain the absolute height range constraints. :

[0122]

[0123]

[0124] Based on each waypoint height Thus, the height difference constraint is obtained:

[0125]

[0126] In the formula, The height window penalty is applied to the i-th path point;

[0127] In this embodiment, the vertical position of the UAV during inspection is strictly limited to ensure safety and mission efficiency. On the one hand, flying too low an altitude may lead to a risk of collision with the ground / low-lying structures; on the other hand, flying too high may be subject to airspace restrictions or mission requirements. Furthermore, even if the altitude is within the allowable range, if the altitude changes of adjacent waypoints too drastically, it may cause difficulties in controller tracking, increased energy consumption, or even induce attitude instability. Therefore, this application imposes constraints on two levels: absolute altitude range constraints and altitude difference constraints between adjacent waypoints.

[0128] S304, Calculate the cost of angle smoothing:

[0129] Define the adjacent path segment vector as:

[0130] The steering angle is defined as:

[0131]

[0132] In the formula, A constant greater than 0;

[0133] It should be noted that, It is a very small constant greater than 0, used to avoid zero denominators and improve computational stability.

[0134] The climb angle is defined as:

[0135]

[0136] The cost of angle smoothing is:

[0137]

[0138] In the formula, , Preset weights;

[0139] In this embodiment, during valve hall inspection, narrow passages and densely populated equipment areas can easily cause frequent turning and vertical fluctuations in the path. Excessive trajectory curvature and sharp turns significantly increase the difficulty of DWA tracking by the local controller and reduce actual flight capability. Therefore, this application applies smooth constraints to horizontal turning and planar climb.

[0140] The valve hall is a typical high-voltage, strong electromagnetic environment. When an unmanned aerial vehicle (UAV) platform is near the converter valve, it is highly susceptible to electromagnetic radiation and discharge noise interference. To account for the potential interference and partial discharge noise risks posed by the high-voltage, strong electromagnetic environment of the converter valve hall to the UAV platform, the following constraints are imposed:

[0141] S305, Calculate the electromagnetic penalty cost:

[0142] The spatial electromagnetic risk field strength index is defined as follows:

[0143]

[0144] In the formula, , Positions The electric and magnetic field strengths at that location , This is a normalized reference value. , Preset weighting coefficients;

[0145] Set two threshold levels: safety threshold No-fly threshold The electromagnetic environment penalty term is then defined as a piecewise function:

[0146]

[0147] In the formula, The penalty index;

[0148] For path points The cumulative electromagnetic penalty results in the following electromagnetic penalty cost:

[0149]

[0150] S306. Construct the total cost function as follows:

[0151]

[0152] In the formula, The preset cost weighting coefficient.

[0153] To address the inspection path planning problem of UAVs in the complex environment of converter valve halls, this application proposes a hierarchical hybrid planning framework with three levels of collaboration: A*, RRT*, and DWA. This framework integrates the advantages of A*, RRT*, and DWA algorithms, decomposing the overall path planning task into three stages.

[0154] Phase 1: Run the A* algorithm within the feasible flight domain to quickly generate a discretized coarse path that avoids known static obstacles, ensuring that a reachable path from the starting point to the destination exists;

[0155] Phase 2: Based on the coarse path generated by A*, perform RRT* optimization to regenerate a smooth reference trajectory that satisfies kinematic constraints;

[0156] Phase 3: Simultaneously, the DWA algorithm is used to track the optimized reference trajectory and dynamically avoid unmodeled obstacles and environmental disturbances in the converter valve hall to ensure the safety of the system under dynamic uncertainty.

[0157] This architecture effectively ensures the reliability of drone inspections through multi-level task decomposition and algorithm collaboration. The overall flowchart is as follows: Figure 3 As shown. The specific steps are as follows:

[0158] S4. Within the feasible flight domain, the first global reference path from the origin to the destination is generated using the A* algorithm based on the total cost function.

[0159] The A* algorithm is an improved search method based on the D algorithm (Dijkstra's shortest path algorithm) by incorporating heuristic information. The D algorithm expands nodes gradually from the starting point in ascending order of cost, essentially an unbiased global search that guarantees the shortest path. However, on large-scale maps or with a large number of nodes, it requires traversing many nodes unrelated to the goal, resulting in significant computational overhead. The A* algorithm, by adding a heuristic term to the cost function, uses goal direction information to guide the search, making the expansion process more focused on potentially optimal regions, thus reducing ineffective exploration. When the heuristic function satisfies the acceptability condition, A* also maintains optimality while significantly outperforming the D algorithm in planning efficiency.

[0160] Furthermore, in this embodiment of the application, step S4 includes:

[0161] S401. Within the feasible flight domain, using the starting point as the initial node and the ending point as the target node, node expansion is performed using an evaluation function that includes heuristics:

[0162]

[0163] in, The cost is the path length from the starting point to the current node;

[0164] S402, to To extend this further, we introduce a composite cost function and consider multidimensional constraints. Defined as:

[0165]

[0166] In the formula, The threat cost represents the nearest grid distance from a node to the threat region. This represents the altitude difference cost between nodes and their neighbors, used to limit drastic changes in flight altitude. Threat distance cost weighting, Threat distance cost weighting;

[0167] S403 The heuristic distance estimate from the current node to the target node is calculated using three-dimensional Euclidean distance:

[0168]

[0169] In the formula, , , They are nodes of coordinate, coordinates and coordinate; , , These are the target nodes. coordinate, coordinates and coordinate;

[0170] S404. The node expansion is guided by the evaluation function until the target node is found, and the first global reference path composed of discrete nodes is obtained by backtracking.

[0171] S5. Guided by the first global path and with the total cost function as the optimization objective, the RRT* algorithm is used to perform bias sampling, expansion and reconnection optimization to obtain a continuous and smooth second global reference path.

[0172] Traditional RRT* algorithms sample uniformly and randomly throughout the state space, resulting in a large number of nodes being generated in regions unrelated to the target. This leads to extremely slow convergence, especially in high-dimensional spaces or with narrow passages, making it difficult to find a feasible solution. In contrast, the RRT* algorithm proposed in this application utilizes a globally coarse path (the first global reference path) quickly generated by A* in a grid map as guidance, performing high-probability biased sampling near this path. This significantly reduces invalid sampling and effectively improves planning speed. In power grid inspection path planning tasks using UAVs, a rapid response to sudden obstacles is required. However, due to the high dynamism of UAV flight and the high uncertainty of the search space, the time window for obstacle avoidance decisions is often very limited. Therefore, it is necessary to combine the RRT* algorithm with the DWA algorithm to better improve the real-time performance of obstacle avoidance during inspections.

[0173] Furthermore, in this embodiment of the application, step S5 includes:

[0174] S501. Construct a global guidance corridor with a preset width, centered on the first global path;

[0175] In one embodiment, the aforementioned global guidance corridor is a strip-shaped feasible area constructed with the first global reference path as the center line, and its preset width is 3 to 8 m, preferably 5 m; the width is determined according to the valve hall passage size, the safe distance of the UAV and the subsequent continuous spatial sampling requirements.

[0176] S502. Under the constraint of the global guidance corridor, bias sampling is performed: the sampling points fall into the guidance corridor with a higher probability than uniform sampling, and the remaining sampling points are randomly generated in the feasible flight domain to guide the random tree to grow in the target direction.

[0177] S503. Based on the sampling points generated by bias sampling, perform nearest neighbor node search and new node expansion, generate new nodes and add them to the random tree;

[0178] S504. Using the total cost function as the optimization objective, traverse the nearest neighbor nodes within a preset radius near the new node, and determine whether connecting through the new node can reduce the total path cost of the nearest neighbor nodes. If so, update the parent node connection relationship of the nearest neighbor nodes.

[0179] S505. Repeat steps S502 to S504 until the distance between a node in the random tree and the target node is less than a first preset threshold. Backtrack to obtain a discrete point sequence composed of continuous path points, which serves as the second global reference path; wherein, the discrete point sequence is represented as:

[0180]

[0181] In the formula, and All of these are the current poses of the drone, i.e. , For each local target point in the sequence It represents a location in three-dimensional space.

[0182] S6. The dynamic window method DWA is used to perform dynamic feasibility screening and safety assessment of candidate velocities in the velocity space, perform local tracking of the global reference path and realize online obstacle avoidance, and output an executable inspection track.

[0183] Furthermore, in this embodiment of the application, step S6 includes:

[0184] S601, Select distance from current pose point As a local guiding point, that is:

[0185]

[0186] S602. Based on the dynamic constraints of the UAV, generate a candidate velocity set in the velocity space. And perform forward simulation on each candidate velocity to simulate the trajectory of the UAV within a preset time window;

[0187] S603. Perform dynamic feasibility screening on the candidate trajectories obtained from the forward simulation, eliminating those that exceed the dynamic limits of the UAV, exceed the height constraints, or are combined with obstacles. Candidate trajectories whose safe distance is less than the second preset threshold;

[0188] S604. Construct an evaluation function for the candidate trajectories that have passed the feasibility screening. :

[0189]

[0190] In the formula, For candidate velocity unit vectors, The unit direction pointing to the RRT* guide point. For each candidate velocity Uniform forward simulation, directional consistency weights Safety distance weight and speed weight satisfy and ;

[0191] S605. Select the candidate velocity with the best evaluation function value as the current control variable, perform local trajectory tracking, and output the executable inspection track segment.

[0192] S606. Repeat steps S601 to S605 in each control cycle until the target node is reached.

[0193] S7. When at least one preset condition is detected, a closed-loop replanning is triggered, the 3D occupation grid is updated, and steps S4 to S6 are repeated until the drone reaches the destination.

[0194] It should be noted that the above preset conditions include: dynamic obstacles entering the preset warning distance, the deviation between the local trajectory and the global reference path exceeding a threshold, and the dynamic window method DWA failing to generate a feasible local trajectory that meets the safety distance constraint within the preset time window.

[0195] This application provides a hierarchical collaborative 3D path planning method for UAVs in converter valve hall inspection. By constructing a total cost function that includes path length, safety distance, altitude window, altitude difference, angle smoothing, and electromagnetic penalty, complex inspection constraints are uniformly incorporated into the optimization framework, effectively balancing flight efficiency, safety, and controllability. Secondly, the A* algorithm is used to quickly generate a globally discrete coarse path on a low-resolution grid and construct a guiding corridor, providing a clear boundary for subsequent continuous spatial search and significantly compressing invalid sampling areas. Based on this, the RRT* algorithm performs bias sampling and reconnection optimization within the guiding corridor, significantly improving convergence speed and path quality in narrow corridor scenarios and effectively suppressing detours caused by random expansion. By combining suboptimal paths with Dynamic Window Method (DWA) for dynamic feasibility screening and safety assessment in velocity space, this method achieves smooth tracking and online obstacle avoidance of the global reference path. This ensures both global optimality and real-time response requirements. In narrow, highly constrained valve hall environments, this method can stably maintain sufficient safety margins, with smooth path turning, strong controllability, high success rate, short planning time, and small fluctuations. The overall path quality and stability are significantly better than single algorithms or conventional metaheuristic methods. In addition, through an adaptive replanning mechanism, this method can trigger local repair or global replanning when dynamic obstacles appear, forming a closed-loop control framework, which further improves the robustness and reliability of inspection under complex working conditions.

[0196] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0197] To verify the effectiveness of the method proposed in this application, a simulation verification was conducted on an indoor inspection scenario in the valve hall of a converter station, and a three-dimensional path planning environment was built on the Matlab platform. The valve hall space was set as a 200×100×50 (m) cube space, in which the gray cylinders are converter valves (equivalent to cylinder / square cylinder obstacles), and the red cylinders are converter transformer valve side bushings (equivalent to cylindrical obstacles). Their spatial positions and geometric parameters are shown in Tables 1 and 2 above, respectively.

[0198] First, a three-dimensional occupancy grid is established according to the method of this application, and safety margin processing is performed on obstacles; in terms of path planning constraints and cost settings, the flight altitude range of the UAV is set to 10–50, the diameter of the UAV is 5, and the safety distance between the UAV and the obstacle is 10. The relevant parameter settings are shown in Table 3.

[0199] Table 3

[0200] To avoid the influence of chance on the results, ten different starting points and ending points were selected in the simulated valve hall map to form the test scenario. The starting point / ending point information is shown in Table 4.

[0201] Table 4

[0202] Next, the planning is executed according to the hierarchical collaborative process of "global-continuous-local": the A* algorithm is run on a low-resolution raster map to quickly generate a discrete coarse path that avoids static obstacles to ensure the reachability from the start point to the end point; then, guided by this coarse path, biased sampling and reconnection optimization of RRT* is performed in its neighborhood to generate a smoother reference trajectory that meets the constraints; finally, DWA is introduced to perform candidate velocity sampling and dynamic feasibility assessment in the velocity space to achieve tracking of the reference trajectory and online obstacle avoidance, thereby obtaining an executable 3D inspection track.

[0203] Meanwhile, during the RRT* and DWA fusion stage, points approximately 3–8 m away from the current UAV pose are selected from the discrete trajectory point sequence generated by RRT* as local guide points to improve the stability and real-time performance of local tracking.

[0204] To highlight the search efficiency advantage of this application in narrow corridors and strongly constrained scenarios, this embodiment first compares the hybrid algorithm proposed in this application with the traditional RRT in the same ten sets of scenarios, and counts the number of extended nodes, planning time and path length, etc. The results are shown in Table 5.

[0205] Table 5

[0206] As shown in Table 5, compared with the traditional RRT*, the hybrid algorithm significantly reduces the number of expanded nodes and planning time in various scenarios (by approximately 70%+ and 77% respectively), and the overall path length is shorter (by approximately 15%). This indicates that global corridor guidance effectively reduces invalid searches and improves planning efficiency and path quality.

[0207] For hybrid algorithms, comparisons with other single algorithms are insufficient. Therefore, this application focuses on comparing hybrid algorithms with some classic metaheuristic algorithms, including: Genetic Algorithm (GA), Whale Optimization Algorithm (WOA), and Improved Particle Swarm Optimization Algorithm (SPSO). Each scenario was repeated 10 times, and the mean and standard deviation (SD) of the total cost function are shown in Table 6, while the mean and standard deviation (SD) of the search time for each algorithm are shown in Table 7.

[0208] Table 6

[0209] Table 7

[0210] As shown in Tables 6 and 7, the proposed hybrid algorithm achieves both better cost levels and stronger statistical stability in 10 sets of start-end scenarios: its total cost has a mean of 5076.82~6421.43 and a standard deviation of 110.73~175.65; the planning time has a mean of 8.752~11.982 and a standard deviation of 1.281~1.772, demonstrating the consistent advantages of low mean and low dispersion. In contrast, metaheuristic methods such as GA, WOA, and SPSO are more prone to deviating from the main channel due to random initialization and search path drift, resulting in increased cost and time dispersion.

[0211] Furthermore, for the comprehensive performance of multiple indicators, we provide a comprehensive radar chart, a stacked bar chart of success rate, a box plot of minimum safety clearance, a box plot of maximum turning angle, and the CDF result of safety clearance (CDF). Figures 4-8 ).

[0212] Depend on Figure 4 It is evident that the hybrid algorithm demonstrates a more balanced performance across multiple dimensions, including success rate, path length, total cost, planning time, maximum turning angle, and minimum safety distance, resulting in superior overall performance.

[0213] Depend on Figure 5 It can be seen that the hybrid algorithm maintains full success under the current discrete grid occupancy and security expansion settings, while the success rate of the other algorithms drops in some use cases.

[0214] Depend on Figure 6It is evident that the overall distribution of the minimum safety distance of the hybrid algorithm has shifted significantly upward, with the median being over 3 m and exhibiting smaller dispersion, indicating a more ample safety margin.

[0215] Depend on Figure 7 It can be seen that the median maximum turning angle of the hybrid algorithm is about 30° to 40°, with smaller dispersion, smoother path turning, and better controllability;

[0216] Depend on Figure 8 As can be seen, the CDF curve of the safety margin of the hybrid algorithm is located to the far right, indicating that it can maintain a larger minimum safety margin on most samples.

[0217] In summary, this embodiment demonstrates that the A*–RRT*–DWA hierarchical collaborative three-dimensional path planning method proposed in this application can achieve stable and feasible path planning in the environment of converter valve halls with "high obstacles, narrow corridors, and limited feasible domains." It performs better in terms of total cost, success rate, safety distance, maximum turning angle, and planning stability, and can meet the safety and feasibility requirements of UAV inspection of converter valve halls.

[0218] Corresponding to the above embodiment, a hierarchical collaborative 3D path planning method for UAVs for converter valve hall inspection, Figure 9 The diagram shows a structural block diagram of a UAV hierarchical collaborative three-dimensional path planning device for converter valve hall inspection provided in an embodiment of this application. For ease of explanation, only the parts related to the embodiment of this application are shown.

[0219] Reference Figure 9 The device 900 includes:

[0220] The environment modeling module 901 is used to establish a three-dimensional environment model of the converter valve hall, which equates the converter valve tower of the converter valve hall to a square column obstacle and the valve side sleeve of the converter valve hall to a cylindrical obstacle, and constructs a three-dimensional occupation grid.

[0221] The feasible domain construction module 902 is used to expand the safety gap of the three-dimensional occupied grid to obtain a feasible flight domain that meets the minimum safety gap.

[0222] Total cost function construction module 903 is used to construct the total cost function within the feasible flight domain and transform path planning into a total cost minimization problem under safety clearance and altitude window constraints;

[0223] The first path generation module 904 is used to generate a first global reference path from the start point to the end point based on the total cost function and the A* algorithm within the feasible flight domain.

[0224] The second path generation module 905 is used to take the first global path as a guide, the total cost function as the optimization objective, and the RRT* algorithm to perform bias sampling, expansion and reconnection optimization to obtain a continuous and smooth second global reference path.

[0225] The dynamic obstacle avoidance module 906 is used to perform dynamic feasibility screening and safety assessment of candidate velocities in the velocity space using the dynamic window method DWA, perform local tracking of the global reference path and realize online obstacle avoidance, and output an executable inspection track.

[0226] The adaptive replanning module 907 is used to trigger closed-loop replanning when at least one preset condition is detected, update the three-dimensional occupancy grid, and repeatedly execute the steps corresponding to the first path generation module 904 to the dynamic obstacle avoidance module 906 until the UAV reaches the destination.

[0227] In practical use, the UAV hierarchical collaborative three-dimensional path planning device for converter valve hall inspection provided in this application embodiment can be configured in any terminal device to execute the aforementioned UAV hierarchical collaborative three-dimensional path planning method for converter valve hall inspection.

[0228] This application provides a hierarchical collaborative 3D path planning device for UAVs used in converter valve hall inspection. By constructing a total cost function that includes path length, safety distance, altitude window, altitude difference, angle smoothing, and electromagnetic penalty, complex inspection constraints are uniformly incorporated into the optimization framework, effectively balancing flight efficiency, safety, and controllability. Secondly, the A* algorithm is used to quickly generate a globally discrete coarse path on a low-resolution grid and construct a guiding corridor, providing a clear boundary for subsequent continuous spatial searches and significantly compressing invalid sampling areas. Based on this, the RRT* algorithm performs bias sampling and reconnection optimization within the guiding corridor, significantly improving convergence speed and path quality in narrow corridor scenarios and effectively suppressing detours caused by random expansion. By combining suboptimal paths with Dynamic Window Method (DWA) for dynamic feasibility screening and safety assessment in velocity space, this method achieves smooth tracking and online obstacle avoidance of the global reference path. This ensures both global optimality and real-time response requirements. In narrow, highly constrained valve hall environments, this method can stably maintain sufficient safety margins, with smooth path turning, strong controllability, high success rate, short planning time, and small fluctuations. The overall path quality and stability are significantly better than single algorithms or conventional metaheuristic methods. In addition, through an adaptive replanning mechanism, this method can trigger local repair or global replanning when dynamic obstacles appear, forming a closed-loop control framework, which further improves the robustness and reliability of inspection under complex working conditions.

[0229] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0230] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0231] To implement the above embodiments, this application also proposes a terminal device.

[0232] Figure 10 This is a schematic diagram of the structure of a terminal device according to an embodiment of this application.

[0233] like Figure 10 As shown, the terminal device 200 includes:

[0234] The system includes a memory 210 and at least one processor 220, and a bus 230 connecting different components (including the memory 210 and the processor 220). The memory 210 stores a computer program, which, when executed by the processor 220, implements a hierarchical collaborative three-dimensional path planning method for UAV inspection of converter valve hall as described in this application embodiment.

[0235] Bus 230 represents one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. For example, these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0236] Terminal device 200 typically includes various electronically readable media. These media can be any available media that can be accessed by terminal device 200, including volatile and non-volatile media, removable and non-removable media.

[0237] Memory 210 may also include computer system readable media in the form of volatile memory, such as random access memory (RAM) 240 and / or cache memory 250. Terminal device 200 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 260 may be used to read and write non-removable, non-volatile magnetic media (… Figure 10 Not shown; usually referred to as a "hard drive"). Although Figure 10 As not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 230 via one or more data media interfaces. Memory 210 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of this application.

[0238] A program / utility 280 having a set (at least one) of program modules 270 may be stored in, for example, memory 210. Such program modules 270 include—but are not limited to—an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 270 typically perform the functions and / or methods described in the embodiments of this application.

[0239] Terminal device 200 can also communicate with one or more external devices 290 (e.g., keyboard, pointing device, display 291, etc.), and with one or more devices that enable a user to interact with terminal device 200, and / or with any device that enables terminal device 200 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed via input / output (I / O) interface 292. Furthermore, terminal device 200 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 293. As shown, network adapter 293 communicates with other modules of terminal device 200 via bus 230. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with terminal device 200, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0240] The processor 220 performs various functional applications and data processing by running programs stored in the memory 210.

[0241] It should be noted that the implementation process and technical principles of the terminal device in this embodiment are explained in the foregoing description of a hierarchical collaborative three-dimensional path planning method for UAV inspection of converter valve hall in this application embodiment, and will not be repeated here.

[0242] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.

[0243] This application provides a computer program product that, when run on a terminal device, enables the terminal device to implement the steps described in the various method embodiments above.

[0244] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a photographing device / terminal device, a recording medium, a computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0245] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0246] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0247] In the embodiments provided in this application, it should be understood that the disclosed devices / terminal equipment and methods can be implemented in other ways. For example, the device / terminal equipment embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling or direct coupling or communication connection may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0248] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0249] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A hierarchical collaborative three-dimensional path planning method for UAVs in converter valve hall inspection, characterized in that, include: S1. Establish a three-dimensional environment model of the converter valve hall, and treat the converter valve tower in the converter valve hall as a square column obstacle and the valve side sleeve in the converter valve hall as a cylindrical obstacle, and construct a three-dimensional occupation grid. S2. Expand the safety gap of the three-dimensional occupied grid to obtain a feasible flight domain that meets the minimum safety gap. S3. Construct the total cost function within the feasible flight domain and transform the path planning into a total cost minimization problem under the constraints of safety distance and altitude window; S4. Within the feasible flight domain, the first global reference path from the origin to the destination is generated using the A* algorithm based on the total cost function. S5. Guided by the first global path and with the total cost function as the optimization objective, the RRT* algorithm is used to perform bias sampling, expansion and reconnection optimization to obtain a continuous and smooth second global reference path. S6. The Dynamic Window Method (DWA) is used to perform dynamic feasibility screening and safety assessment of candidate velocities in the velocity space, execute local tracking of the global reference path and realize online obstacle avoidance, and output an executable inspection track. S7. When at least one preset condition is detected, a closed-loop replanning is triggered, the 3D occupation grid is updated, and steps S4 to S6 are repeated until the drone reaches the destination.

2. The method according to claim 1, characterized in that, S3 involves constructing a total cost function within the feasible flight domain and transforming path planning into a total cost minimization problem under constraints of safety distance and altitude window. Specifically, this includes: S301. Calculate the path length cost by summing the Euclidean distances between adjacent discrete waypoints: ； In the formula, For path length cost, For the first on the path discrete location points and N is the number of waypoints; S302. Calculate the safety cost by calculating the minimum distance between any path point and the set of obstacles, and based on the minimum safety gap. With buffer bandwidth Construct a segmented penalty function to obtain the full path safety cost: set up Given a set of obstacles, for any path segment To the obstacle The minimum distance is denoted as: ； Define minimum safety clearance and buffer bandwidth The minimum safety gap Composed of the equivalent radius of the drone and the safety margin, the single-segment safety penalty is defined as a piecewise function: ； based on The safety cost obtained from a single-segment safety penalty is: ； S303. Calculate the cost of height constraints, including absolute height range constraints and height difference constraints between adjacent path points: Let the permissible flight altitude range be For each path point height Define and sum the height constraint costs to obtain the absolute height range constraints. : ； ； Based on each waypoint height Thus, the height difference constraint is obtained: ； In the formula, The height window penalty is applied to the i-th path point; S304, Calculate the cost of angle smoothing: Define the adjacent path segment vector as: ； The steering angle is defined as: ； In the formula, A constant greater than 0; The climb angle is defined as: ； The cost of angle smoothing is: ； In the formula, , Preset weights; S305, Calculate the electromagnetic penalty cost: The spatial electromagnetic risk field strength index is defined as follows: ； In the formula, , Positions The electric and magnetic field strengths at that location , This is a normalized reference value. , Preset weighting coefficients; Set two threshold levels: safety threshold No-fly threshold The electromagnetic environment penalty term is then defined as a piecewise function: ； In the formula, The penalty index; For path points The cumulative electromagnetic penalty results in the following electromagnetic penalty cost: ； S306. Construct the total cost function as follows: ； In the formula, The preset cost weighting coefficient.

3. The method according to claim 2, characterized in that, S4, within the feasible flight domain, generates the first global reference path from the starting point to the destination using the A* algorithm based on the total cost function, as follows: S401. Within the feasible flight domain, with the starting point as the initial node and the ending point as the target node, node expansion is performed using an evaluation function that includes a heuristic term: ； in, The cost is the path length from the starting point to the current node; S402, to To extend this further, we introduce a composite cost function and consider multidimensional constraints. Defined as: ； In the formula, The threat cost represents the nearest grid distance from a node to the threat region. This represents the altitude difference cost between nodes and their neighbors, used to limit drastic changes in flight altitude. Threat distance cost weighting, Threat distance cost weighting; S403 The heuristic distance estimate from the current node to the target node is calculated using three-dimensional Euclidean distance: ； In the formula, , , They are nodes of coordinate, coordinates and coordinate; , , These are the target nodes. coordinate, coordinates and coordinate; S404. The node expansion is guided by the evaluation function until the target node is found, and the first global reference path composed of discrete nodes is obtained by backtracking.

4. The method according to claim 3, characterized in that, S5, guided by the first global path and with the total cost function as the optimization objective, employs the RRT* algorithm for bias sampling, expansion, and reconnection optimization to obtain a continuous and smooth second global reference path, specifically including: S501. Construct a global guidance corridor with a preset width, centered on the first global path; S502. Under the constraint of the global guidance corridor, bias sampling is performed: the sampling points fall into the guidance corridor with a higher probability than uniform sampling, and the remaining sampling points are randomly generated in the feasible flight domain to guide the random tree to grow in the target direction. S503. Based on the sampling points generated by the bias sampling, perform nearest neighbor node search and new node expansion, generate new nodes and add them to the random tree; S504. Using the total cost function as the optimization objective, traverse the nearest neighbor nodes within a preset radius near the new node, and determine whether connecting through the new node can reduce the total path cost of the nearest neighbor nodes. If so, update the parent node connection relationship of the nearest neighbor nodes. S505. Repeat steps S502 to S504 until the distance between a node in the random tree and the target node is less than a first preset threshold. Backtrack to obtain a discrete point sequence composed of continuous path points, which serves as the second global reference path; wherein, the discrete point sequence is represented as: ； In the formula, and All of these are the current poses of the drone, i.e. , For each local target point in the sequence It represents a location in three-dimensional space.

5. The method according to claim 4, characterized in that, S6 employs the Dynamic Window Method (DWA) to dynamically screen and assess the feasibility of candidate velocities in the velocity space, performs local tracking of the global reference path and implements online obstacle avoidance, and outputs an executable inspection track, specifically including: S601, Select distance from current pose point As a local guiding point, that is: ； S602. Based on the dynamic constraints of the UAV, generate a candidate velocity set in the velocity space. And perform forward simulation on each candidate velocity to simulate the trajectory of the UAV within a preset time window; S603. Perform dynamic feasibility screening on the candidate trajectories obtained from the forward simulation, eliminating those that exceed the dynamic limits of the UAV, exceed the height constraints, or are combined with obstacles. Candidate trajectories whose safe distance is less than the second preset threshold; S604. Construct an evaluation function for the candidate trajectories that have passed the feasibility screening. : ； In the formula, For candidate velocity unit vectors, The unit direction pointing to the RRT* guide point. For each candidate velocity Uniform forward simulation, direction consistency weighting Safety distance weight and speed weight satisfy and ; S605. Select the candidate velocity with the best evaluation function value as the current control variable, perform local trajectory tracking, and output the executable inspection track segment. S606. Repeat steps S601 to S605 in each control cycle until the target node is reached.

6. The method according to claim 5, characterized in that, The preset conditions in S7 include: dynamic obstacles entering a preset warning distance, the deviation between the local trajectory and the global reference path exceeding a threshold, and the Dynamic Window Method (DWA) failing to generate a feasible local trajectory that meets the safety distance constraint within a preset time window.

7. A hierarchical collaborative three-dimensional path planning device for unmanned aerial vehicle (UAV) inspection of converter valve halls, characterized in that, include: The environment modeling module is used to create a three-dimensional environment model of the converter valve hall, which equates the converter valve tower of the converter valve hall to a square column obstacle and the valve side sleeve of the converter valve hall to a cylindrical obstacle, and constructs a three-dimensional occupation grid. The feasible domain construction module is used to expand the safety gap of the three-dimensional occupied grid to obtain a feasible flight domain that meets the minimum safety gap. The total cost function construction module is used to construct the total cost function within the feasible flight domain and transform path planning into a total cost minimization problem under the constraints of safety distance and altitude window; The first path generation module is used to generate the first global reference path from the start point to the end point based on the total cost function and the A* algorithm within the feasible flight domain. The second path generation module is used to obtain a continuous and smooth second global reference path by using the first global path as a guide, the total cost function as the optimization objective, and the RRT* algorithm for bias sampling, expansion and reconnection optimization. The dynamic obstacle avoidance module is used to perform dynamic feasibility screening and safety assessment of candidate velocities in the velocity space using the dynamic window method DWA, perform local tracking of the global reference path and realize online obstacle avoidance, and output an executable inspection track. The adaptive replanning module is used to trigger closed-loop replanning when at least one preset condition is detected, update the 3D occupancy grid, and repeatedly execute the steps corresponding to the first path generation module to the dynamic obstacle avoidance module until the UAV reaches the destination.

8. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 6.

Images