Power inspection path planning method and device and electronic equipment

Abstract

The invention provides an electric power inspection path planning method and device and electronic equipment, and relates to the technical field of unmanned aerial vehicle electric power inspection. The method comprises the following steps: acquiring obstacle information of an electric power facility environment, wherein the obstacle information comprises the type and position of an obstacle; based on the type of the obstacle and a preset safety distance coefficient, determining a differentiated safety distance, the type of the obstacle including a power transmission line, a transformer substation, a tower and other obstacles; based on the obstacle information and the differentiated safety distance, obtaining an initial global path through a path search algorithm; based on a preset multi-objective optimization function, the initial global path is optimized, a Pareto optimal path set is generated, and the multi-objective optimization function comprises a path length objective, a safety margin objective and an electromagnetic safety objective; and determining a target global path from the Pareto optimal path set based on a preset inspection task mode. According to the invention, the inspection efficiency and adaptability can be improved while the safety is guaranteed.

Description

Technical Field

[0001] This application relates to the field of unmanned aerial vehicle (UAV) power line inspection technology, and in particular to a power line inspection path planning method, device, and electronic equipment. Background Technology

[0002] With the rapid development of power systems and the continuous increase in the length of transmission lines, traditional power line inspection methods are no longer sufficient to meet the operation and maintenance needs of modern power systems. To improve inspection efficiency and safety, drone technology, due to its advantages of maneuverability, low cost, and high accessibility, has been widely applied in the field of power line inspection. Drone path planning, as one of the core technologies for achieving autonomous inspection, directly affects the efficiency, safety, and energy consumption of inspection tasks.

[0003] Currently, UAV path planning technologies are mainly divided into several categories: classic graph search algorithms (such as...) Path planning algorithms include Dijkstra's algorithm, biomimetic or swarm intelligence-based optimization algorithms (such as genetic algorithms, particle swarm optimization, ant colony optimization, etc.), sampling-based algorithms (such as PRM and RRT), local programming algorithms (such as artificial potential field method and dynamic window method), and deep learning and reinforcement learning algorithms. These algorithms each have their own focus in path planning. For example, graph search algorithms find the optimal path by expanding nodes in a discrete map and evaluating costs; biomimetic optimization algorithms simulate natural biological behavior for global search; sampling-based algorithms generate waypoints through random sampling; local programming algorithms emphasize real-time response and obstacle avoidance; and deep learning and reinforcement learning algorithms optimize path strategies through model learning and environmental interaction.

[0004] Despite the variety of existing technologies, they still suffer from insufficient adaptability and difficulty in balancing safety and inspection efficiency when dealing with complex power inspection scenarios. Summary of the Invention

[0005] The power inspection path planning method, device, and electronic equipment provided in this application are used to solve the problems that the existing technology still has insufficient adaptability and difficulty in balancing safety and inspection efficiency when dealing with complex power inspection scenarios.

[0006] In a first aspect, embodiments of this application provide a power line inspection route planning method, including:

[0007] Obtain obstacle information about the power facility environment, including the type and location of obstacles;

[0008] Differential safety distances are determined based on the type of obstacle and a preset safety distance coefficient. The types of obstacles include power transmission lines, substations, towers, and other obstacles.

[0009] An initial global path is obtained based on the obstacle information and the differentiated safety distance through a path search algorithm;

[0010] The initial global path is optimized based on a preset multi-objective optimization function to generate a Pareto optimal path set, and the multi-objective optimization function includes a path length objective, a safety margin objective and an electromagnetic safety objective;

[0011] A target global path is determined from the Pareto optimal path set based on a preset inspection task mode.

[0012] In a possible implementation, the differentiated safety distance is determined based on the type of the obstacle and a preset safety distance coefficient, including: obtaining a basic safety distance and a safety margin; determining a corresponding target safety distance coefficient from the preset safety distance coefficient based on the type of the obstacle; and determining the differentiated safety distance of the obstacle according to the basic safety distance, the safety margin and the target safety distance coefficient.

[0013] In a possible implementation, the method further includes: when the type of the obstacle is a high-voltage transmission line or a transformer substation, constructing an electromagnetic field strength attenuation model; determining a no-fly area based on the electromagnetic field strength attenuation model and a preset maximum electromagnetic field strength threshold for flight; regarding the no-fly area as an obstacle and adding the no-fly area to the obstacle information, and / or adjusting the electromagnetic safety objective in the multi-objective optimization function based on the no-fly area.

[0014] In a possible implementation, the initial global path is obtained based on the obstacle information and the differentiated safety distance through a path search algorithm, including: starting a bidirectional search process from a starting point and an ending point of path planning through the path search algorithm; in the bidirectional search process, dynamically adjusting the size of a search neighborhood based on the obstacle information and the differentiated safety distance around a current node; and performing node expansion based on the dynamically adjusted search neighborhood to obtain the initial global path.

[0015] In a possible implementation, the size of the search neighborhood is dynamically adjusted based on the obstacle information and the differentiated safety distance around the current node, including: determining a ratio of the number of obstacles within a search radius of the current node to the total number of grids based on the obstacle information around the current node; determining a local environment complexity index based on the differentiated safety distance and the ratio; and dynamically adjusting the size of the search neighborhood based on the local environment complexity index.

[0016] In a possible implementation, the size of the search neighborhood is dynamically adjusted based on the local environment complexity index, including: when the local environment complexity index is less than or equal to a lower threshold, the search neighborhood adopts a first search step; when the local environment complexity index is greater than or equal to an upper threshold, the search neighborhood adopts a second search step; when the local environment complexity index is greater than the lower threshold and less than the upper threshold, the search neighborhood adopts a third search step; wherein the first search step is greater than the third search step, and the third search step is greater than the second search step.

[0017] In a possible implementation, when the search neighborhood adopts the first search step, the constraint condition for generating the extended node includes: in the direction from the current node to the extended node, all intermediate nodes located between the current node and the extended node are not occupied by obstacles; and the shortest distance from the path segment connecting the current node and the extended node to all obstacles within a preset range is greater than the differentiated safety distance corresponding to the obstacle.

[0018] In a possible implementation, after determining the target global path from the set of Pareto optimal paths based on the preset inspection task mode, the method further includes: identifying all inflection points in the target global path; determining, as redundant inflection points, the inflection points that pass the geometric collision detection and the speed feasibility detection among all the inflection points, and removing the redundant inflection points; wherein the geometric collision detection is used to judge whether the line segment connecting the adjacent nodes after removal intersects with the obstacle; the speed feasibility detection is based on the cruise speed of the unmanned aerial vehicle, the turning dynamics model, and the turning angle of the line segment, to determine whether the unmanned aerial vehicle can complete the line segment flight without deceleration.

[0019] In a possible implementation, the method further includes: for each inflection point in the target global path after removing the redundant inflection points, dynamically determining an arc smoothing radius based on the current flight speed of the unmanned aerial vehicle and the turning angle of the inflection point; generating an arc trajectory for replacing the inflection point based on the arc smoothing radius; verifying whether the points on the arc trajectory satisfy the safety distance requirement with the adjacent obstacle.

[0020] In a possible implementation, the method further includes: obtaining a plurality of devices to be inspected and an access priority of each device; determining, based on the access priorities of all the devices, a device inspection sequence by a priority-punished sequence optimization model, the device inspection sequence including a plurality of inspection points, each inspection point being used to indicate an inspection order of a device corresponding to the inspection point; adjusting, based on the device inspection sequence, a passing point order of the target global path; determining, for each inspection point, a candidate position set that meets a preset photographing condition, the preset photographing condition including a spatial geometric constraint and a time behavior constraint, the spatial geometric constraint including a distance constraint and / or a shooting angle constraint between devices, and the time behavior constraint including a hovering stable time requirement; selecting, from the candidate position set, a position that meets a preset path cost requirement as an actual path point; and generating an updated target global path based on the actual path points corresponding to all the inspection points and the passing point order.

[0021] In a possible implementation, the method further includes: obtaining a current flight speed of the unmanned aerial vehicle when the unmanned aerial vehicle performs power inspection based on the target global path; detecting, in real time, a dynamic obstacle by an on-board sensor and predicting a motion trajectory of the dynamic obstacle; calculating a predicted collision time and a minimum expected distance based on a current path of the unmanned aerial vehicle and the predicted motion trajectory of the dynamic obstacle; triggering local path re-planning if the predicted collision time is less than a warning time threshold and / or the minimum expected distance is less than a dynamic obstacle safety distance; generating a detour path by using a simplified path planning algorithm in a local search range with the current position of the unmanned aerial vehicle as a starting point, the detour path being used to guide the unmanned aerial vehicle to re-enter the target global path after avoiding the dynamic obstacle; performing adaptive inflection point optimization on the target global path or the detour path to remove redundant inflection points; and dynamically calculating a circular arc smoothing radius based on the current flight speed of the unmanned aerial vehicle and a path turning angle, and performing speed-adaptive circular arc smoothing processing on the optimized path.

[0022] In a second aspect, an embodiment of the present application provides a power inspection path planning device, including:

[0023] An obtaining module is configured to obtain obstacle information of a power facility environment, the obstacle information including types and positions of obstacles;

[0024] A determining module is configured to determine a differentiated safety distance based on the types of the obstacles and a preset safety distance coefficient, wherein the types of the obstacles include power transmission lines, power substations, towers, and other obstacles.

[0025] An obtaining module is configured to obtain an initial global path by a path search algorithm based on the obstacle information and the differentiated safety distance.

[0026] The generation module is used to optimize the initial global path based on a preset multi-objective optimization function to generate a Pareto optimal path set. The multi-objective optimization function includes path length objective, safety margin objective, and electromagnetic safety objective.

[0027] The selection module is used to determine the target global path from the Pareto optimal path set based on the preset inspection task mode.

[0028] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0029] The memory stores the instructions that the computer executes;

[0030] The processor executes computer execution instructions stored in memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0031] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0032] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed, implements the first aspect and / or various possible implementations of the first aspect.

[0033] The power line inspection path planning method, apparatus, and electronic equipment provided in this application acquire obstacle type and location information, and determine differentiated safety distances based on the type and a preset safety distance coefficient. This allows for different safety strategies to be adopted for different types of obstacles during subsequent path planning, thereby reducing the probability of collisions between the UAV and high-risk obstacles during flight and improving flight safety. Furthermore, the differentiated safety distance setting avoids overly conservative path planning in low-risk areas, reducing unnecessary detours. The planned path enables the UAV to complete inspection tasks more efficiently, thus improving inspection efficiency. In addition, by introducing a multi-objective optimization function that includes path length, safety margin, and electromagnetic safety objectives to optimize the initial global path, the path planning process no longer solely pursues the shortest path length but comprehensively considers multiple key factors such as safety and electromagnetic safety, seeking an optimal balance among multiple objectives. This ensures that the target global path guarantees both efficiency and meets various safety requirements. Finally, the target global path is determined based on a preset inspection task mode, enabling path planning to adapt to different power line inspection task requirements. Attached Figure Description

[0034] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0035] Figure 1 A flowchart illustrating a power inspection route planning method provided in this application;

[0036] Figure 2 A diagram illustrating the safe distances corresponding to different obstacle types provided in this application;

[0037] Figure 3 A flowchart illustrating the initial global path determination method provided in this application;

[0038] Figure 4 This application provides a schematic diagram of adaptive neighborhood selection based on environmental complexity.

[0039] Figure 5 A schematic diagram of the inflection point path provided for this application;

[0040] Figure 6 This application provides a schematic diagram of speed-adaptive circular arc smoothing comparison.

[0041] Figure 7 This is a schematic diagram illustrating the geometric principle of the local circular arc smoothing method provided in this application;

[0042] Figure 8 This is a schematic diagram of the multi-inspection point sequence optimization provided in this application;

[0043] Figure 9 This is a schematic diagram of local replanning of dynamic obstacles provided in this application;

[0044] Figure 10 A comparative diagram of the power inspection route planning method provided in this application and the traditional planning method;

[0045] Figure 11 A schematic diagram of the power inspection route planning device provided in this application;

[0046] Figure 12 A schematic diagram of the structure of the electronic device provided in this application.

[0047] The accompanying drawings have illustrated specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to specific embodiments. Detailed Implementation

[0048] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0049] In power line inspection scenarios, high-voltage transmission lines, towers, substations, and other facilities are densely distributed in complex environments, exhibiting both high-risk areas (such as areas with strong electromagnetic interference in substations) and open areas. Drones, with their advantages of maneuverability, low cost, and wide coverage, have become a core tool for power line inspection. However, drone path planning, as a crucial aspect of autonomous inspection, directly impacts inspection efficiency, safety, and energy consumption.

[0050] In existing technologies, graph search algorithms (such as...) Dijkstra's algorithm expands nodes and evaluates costs using a discretized map, making it suitable for global path optimization in static environments. However, it suffers from high computational complexity, long search time, and the fact that paths only optimize a single objective (such as path length). Furthermore, its fixed-neighborhood search strategy is ill-suited for complex power environments. Biomimetic optimization algorithms (such as genetic algorithms, particle swarm optimization, and gray wolf optimizers) simulate biological evolution or cooperative behavior for global search, exhibiting strong adaptability. However, they generally suffer from slow convergence, parameter sensitivity, and susceptibility to local optima, and they do not optimize for the varying risk levels of power facilities. Sampling-based algorithms (such as PRM and RRT) generate path points through random sampling, making them suitable for high-dimensional complex environments. However, path quality depends on the randomness of the sampling, and they do not consider the specific needs of power line inspection. Local planning algorithms (such as artificial potential field methods and dynamic window methods) have fast response times and are suitable for dynamic obstacle avoidance, but they have limited visibility, are prone to local optima, and cannot guarantee the optimality of the global path. Deep learning and reinforcement learning methods optimize path strategies through model learning and are good at handling high-dimensional dynamic environments, but they face bottlenecks such as long training time, dependence on high-quality data, and insufficient adaptability to dynamic environments.

[0051] When faced with complex power line inspection scenarios, existing planning methods are overly simplistic in their environmental perception, failing to differentiate the risk levels and safety requirements of different obstacles. This results in insufficient safety margins in high-risk areas and overly conservative path planning in low-risk areas. Furthermore, most planning methods only pursue the shortest path length, neglecting multi-dimensional indicators such as the number of turns, trajectory smoothness, safety distance, and electromagnetic exposure duration. This makes the planned paths difficult to execute in actual flight, increasing the drone's energy consumption and operational complexity. In addition, the core parameters of existing planning methods are usually fixed, unable to be intelligently adjusted according to dynamic environmental conditions, limiting their performance in changing environments. Finally, existing planning methods almost completely ignore the special requirements of power line inspection tasks, such as the optimal access order for multiple inspection points, image posture constraints, return-to-home decisions, and the ability to respond instantly to dynamic environmental factors. Therefore, existing planning methods still suffer from difficulties in balancing safety and inspection efficiency, as well as insufficient adaptability, when facing complex power line inspection scenarios.

[0052] To address the aforementioned issues, this application provides a power line inspection path planning method, apparatus, and electronic device. By acquiring information on the type and location of obstacles in the power facility environment, it provides a data foundation for subsequent precise planning. Considering that different types of obstacles have varying degrees of impact on UAV flight safety—for example, power transmission lines and substations pose higher risks, while other common obstacles have relatively lower risks—differentiated safety distances are determined based on obstacle type and a preset safety distance coefficient. This avoids the safety issues caused by a single safety strategy, making path planning more aligned with actual conditions. Then, based on the above information, a path search algorithm is used to generate an initial global path, providing a preliminary framework for path planning. However, the initial path may not meet the various requirements of power line inspection. Therefore, a preset multi-objective optimization function is introduced, covering path length, safety margin, and electromagnetic safety objectives. Through multi-objective optimization, a Pareto optimal path set is generated, comprehensively evaluating and optimizing the path from multiple key dimensions. Finally, considering that power inspection tasks have different modes, such as different inspection point sequences and photo-taking posture requirements, the target global path is determined from the Pareto optimal path set based on the preset inspection task mode. This ensures that the path planning can flexibly cope with different inspection scenarios and requirements, and improves the adaptability and practicality of the method.

[0053] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0054] The execution entity of the power inspection path planning method provided in this application embodiment can be a computing device such as a server or server cluster. The server can be a mobile phone, computer, tablet, or other device. This application embodiment does not impose any particular restrictions on the implementation method of the execution entity, as long as the execution entity can obtain obstacle information of the power facility environment, including the type and location of obstacles; determine differentiated safety distances based on the type of obstacles and a preset safety distance coefficient, wherein the types of obstacles include transmission lines, substations, towers, and other obstacles; obtain an initial global path based on the obstacle information and differentiated safety distances through a path search algorithm; optimize the initial global path based on a preset multi-objective optimization function to generate a Pareto optimal path set, wherein the multi-objective optimization function includes path length objective, safety margin objective, and electromagnetic safety objective; and determine the target global path from the Pareto optimal path set based on a preset inspection task mode.

[0055] Figure 1 This is a flowchart illustrating a power line inspection route planning method provided in this application. The executing entity of this method can be a system server storing the power line inspection route planning method or other servers; this embodiment does not impose any particular limitations. Figure 1 As shown, the method may include:

[0056] S101. Obtain obstacle information of the power facility environment, including the type and location of obstacles.

[0057] The power facility environment refers to the spatial range that includes power-related facilities such as transmission lines, substations, and power poles, as well as other obstacles that may affect drone power line inspections. The type of obstacle is used to determine its threat level to drone inspections, while the location of the obstacle is used to determine its coordinates in space.

[0058] In this step, various sensors (such as lidar, cameras, etc.) can be used to scan and monitor the environment of power facilities, collect relevant data on obstacles, and determine the type and specific location of obstacles through data processing and analysis.

[0059] For example, using lidar scanning, such as by emitting a laser beam and receiving the reflected signal, a three-dimensional point cloud model of the power facility environment can be constructed based on information such as the time difference and intensity of the signal, thereby identifying different types of obstacles and their locations.

[0060] Another example is the use of camera vision recognition, such as using a high-definition camera to capture images of the environment of power facilities, using image recognition algorithms (such as convolutional neural networks in deep learning) to analyze the images, identify obstacles such as transmission lines and substations, and determine their positions by converting image coordinates to actual spatial coordinates.

[0061] This system can pre-define various obstacle types and determine obstacle information about the power facility environment using currently collected obstacle data. In some embodiments, an obstacle set is... Refine into multiple types of collections: ;

[0062] in, For power transmission lines; For poles and towers; For substations and their equipment; Other obstacles (such as trees and buildings in the surrounding environment).

[0063] S102. Based on the type of obstacle and the preset safety distance coefficient, determine the differentiated safety distance, wherein the type of obstacle includes power transmission lines, substations, poles and other obstacles.

[0064] In this step, the safety distance coefficient can be a pre-set coefficient based on the characteristics and safety requirements of different types of obstacles, used to determine the appropriate differentiated safety distance in combination with the type of obstacle.

[0065] Furthermore, based on pre-set rules, different obstacle types can be multiplied by their corresponding safety distance coefficients or subjected to other calculations to obtain the corresponding differentiated safety distances.

[0066] In one example, safety distance calculation formulas for different types of obstacles can be formulated based on experience accumulated from long-term power line inspection practices. For example, for transmission lines, the safety distance = transmission line voltage level × coefficient 1; for substations, the safety distance = substation scale parameter × coefficient 2. In some embodiments, when the obstacle type is a transmission line, it can be further refined into high-voltage transmission lines and low-voltage transmission lines. Based on factors such as wind sway and electromagnetic interference radius of high-voltage transmission lines, a safety distance coefficient of 2.5 is set; based on the wind sway of low-voltage transmission lines, a safety distance coefficient of 1.8 is set. Based on factors such as the structural dimensions of the tower and the lifting components, a safety distance coefficient of 1.5 is set; based on factors such as the strong electromagnetic field, equipment density, and no-fly zones of the substation, a safety distance coefficient of 3.0 is set; and for other obstacles, a safety distance coefficient of 1.0 can be set based on the basic safety distance.

[0067] In another example, an expert system incorporating the knowledge and experience of power experts can be established. By inputting obstacle types into the expert system, the system can output corresponding differentiated safety distances based on its internal knowledge rules.

[0068] S103. Based on obstacle information and differentiated safety distance, an initial global path is obtained through a path search algorithm.

[0069] The path search algorithm can be an algorithm used to find a feasible path from the starting point to the ending point in a given environment. It combines obstacle information and differentiated safety distance to ensure that the generated path can avoid obstacles and meet the safety distance requirements.

[0070] Furthermore, the power facility environment can be modeled as a graph or grid, taking into account obstacle information and differentiated safety distances. A path search algorithm can then be used to search for a path from the inspection start point to the end point in the model to obtain an initial global path.

[0071] Among them, it can be based on In a graph or mesh model, the algorithm selects the optimal path by evaluating a cost function for each node (typically including the actual cost from the starting point to the node and the estimated cost from the node to the ending point). When considering obstacle information and differentiated safety distances, areas within the safety distance are treated as impassable, thus guiding the algorithm to search for safe paths.

[0072] Alternatively, a genetic algorithm can be used to encode the path as chromosomes, continuously optimizing the path by simulating selection, crossover, and mutation processes in biological evolution. Factors such as path length, obstacle avoidance, and meeting safe distance requirements are considered in the fitness function, and an initial global path is obtained through multiple generations of evolution.

[0073] S104. Based on a preset multi-objective optimization function, optimize the initial global path to generate a Pareto optimal path set. The multi-objective optimization function includes path length objective, safety margin objective, and electromagnetic safety objective.

[0074] The multi-objective optimization function can include multiple objective functions such as path length objective, safety margin objective, and electromagnetic safety objective, and is used to comprehensively consider multiple key performance indicators of the path. The Pareto optimal path set can refer to the set of other paths in multi-objective optimization where no single path is better than any other path in all objectives.

[0075] In this step, the initial global path can be substituted into a multi-objective optimization function, and the parameters or shape of the path can be adjusted by the optimization algorithm so that the multiple objective functions reach a relatively optimal balance state, thereby generating a Pareto optimal path set.

[0076] Furthermore, a multi-objective particle swarm optimization algorithm can be used to represent the path parameters as particle positions. By searching the solution space with the particle swarm, the particle velocity and position are updated based on the fitness value of the multi-objective optimization function. During the search process, multiple objectives such as path length, safety margin, and electromagnetic safety are continuously optimized, ultimately yielding a Pareto-optimal path set.

[0077] Alternatively, the Non-Dominated Sorting Genetic Algorithm (NSGA-II) can be used to perform non-dominated sorting on the initial path population, and selection, crossover, and mutation operations can be performed based on the sorting results. During the evolutionary process, the diversity of the population is maintained, gradually approaching the optimal solution for multiple objectives and generating a Pareto optimal path set.

[0078] In some embodiments, the multi-objective optimization function further includes path length, number of turns, minimum safety margin, path smoothness, and electromagnetic safety, wherein:

[0079] (1) The path length satisfies: .

[0080] (2) The number of turns satisfies: ;in, For indicator functions; The angle at the inflection point; The threshold used to determine whether a point is an inflection point.

[0081] (3) The minimum safety margin is satisfied: ;in, For nodes Distance to the nearest obstacle; This is the safe distance from the obstacle; Safety margin (actual distance minus required distance); This indicates a change in the direction of optimization, because multi-objective optimization is usually about minimizing the objective function, but it is generally desirable to have as large a safety margin as possible.

[0082] (4) The path smoothness satisfies: .

[0083] (5) Electromagnetic safety requirements are met: ;in, Represents a node Distance to electromagnetic facilities.

[0084] However, these five objectives are often contradictory and cannot be optimal. Therefore, the NSGA-II algorithm is used to generate the optimal path set and select the most suitable path according to the task requirements to solve the fundamental problem that traditional single-objective path planning cannot take into account multiple requirements such as safety, efficiency and energy consumption.

[0085] Therefore, in some examples, the improved bidirectional [mechanism] is first utilized. The algorithm intelligently generates an initial population consisting of multiple high-quality, non-identical paths, providing an excellent starting point for subsequent optimization. Then, it enters the core iterative evolutionary process: in each generation, each path in the population is first comprehensively evaluated in terms of five dimensions: path length, smoothness, safety margin, etc. Next, all paths are stratified using non-dominated sorting, and crowding calculation is used to ensure solution diversity, thus avoiding path homogenization while selecting the best option. Based on this, better path solutions are explored by simulating crossover and mutation operations in biological evolution, such as randomly swapping the middle segments of two paths or fine-tuning nodes on the path, and using constraint repair mechanisms to ensure that all newly generated paths strictly avoid obstacles. Finally, an elite selection strategy is used to retain the best offspring for the next iteration, and this cycle continues until the maximum number of generations is reached. The final output is not a single optimal solution, but a Pareto optimal path set. It provides decision-makers with a "menu" of different path options (such as shortest, smoothest, safest, etc.) to meet the complex and ever-changing needs of actual inspection tasks.

[0086] S105. Based on the preset inspection task mode, determine the target global path from the Pareto optimal path set.

[0087] The inspection task mode can refer to the specific requirements and scenarios of power inspection, such as the order of inspection points, the posture for taking photos, and other different task requirements. Furthermore, based on the preset inspection task mode, corresponding selection rules can be formulated to select the path that best fits the task mode from the Pareto optimal path set as the target global path.

[0088] In this step, path selection rules corresponding to different inspection task modes can be predefined. For example, for task modes that require rapid completion of inspections, shorter paths can be selected; for task modes that require high-quality photography, paths that meet the photography posture requirements can be selected. Alternatively, multiple attributes in the inspection task mode (such as time, photography quality, etc.) can be considered, and each attribute can be assigned a certain weight. By calculating the comprehensive attribute value of each path in the Pareto optimal path set, the path with the highest comprehensive attribute value can be selected as the target global path.

[0089] In some examples, the system prioritizes the path with the shortest path length in fast inspection mode; prioritizes the path with the highest safety margin in safety inspection mode; and prioritizes the path with the best path smoothness in smooth inspection mode.

[0090] Furthermore, the inspection task modes can include rapid inspection mode, safe inspection mode, smooth inspection mode, and comprehensive inspection mode. Therefore, the selection method for the target global path can be:

[0091] (1) Rapid Inspection Mode: ;in, This indicates finding the path with the smallest calculated result within the parentheses; weight: (i.e., the preferred path length).

[0092] (2) Safety inspection mode: Among them, the weights are: (i.e., priority safety margin and electromagnetic safety);

[0093] (3) Smooth inspection mode: Among them, the weights are: (i.e., prioritizing full smoothness);

[0094] (4) Comprehensive inspection mode: Among them, the weights are: .

[0095] The power line inspection path planning method provided in this application acquires obstacle type and location information and determines differentiated safety distances. It fully considers the risk levels of different obstacles; for high-risk obstacles such as power transmission lines and substations, larger safety distances can be set to ensure the flight safety of drones. For other obstacles with relatively low risk, the safety distance can be appropriately reduced to avoid overly conservative path planning. This improves inspection efficiency while ensuring safety, effectively solving the problem of balancing safety and efficiency. Simultaneously, a multi-objective optimization function is established, including path length, safety margin, and electromagnetic safety objectives, to optimize the initial global path and generate a Pareto optimal path set. This means that the path planning process no longer solely pursues the shortest path length but comprehensively considers multiple key factors such as safety and electromagnetic safety, seeking an optimal balance among multiple objectives. This ensures that the final determined global path guarantees both efficiency and meets various safety requirements. Furthermore, the multi-objective optimization function reflects consideration of multi-dimensional indicators, changing the previous situation of a single optimization objective and helping to plan paths that better meet actual needs. Determining the target global path based on a preset inspection task mode can improve the versatility and practicality of this method.

[0096] Based on the above embodiments, a differentiated safety distance is determined based on the type of obstacle and a preset safety distance coefficient, including: obtaining a basic safety distance and a safety margin; determining a corresponding target safety distance coefficient from the preset safety distance coefficients based on the type of obstacle; and determining the differentiated safety distance of the obstacle based on the basic safety distance, the safety margin, and the target safety distance coefficient.

[0097] In this embodiment, the basic safety distance refers to a fundamental safety distance value set in a power inspection environment to ensure the basic safe operation of inspection equipment such as drones. It is a general initial safety distance reference value that does not consider the differences in specific obstacle types, and is usually determined based on factors such as the overall safety requirements of power facilities and the basic performance of inspection equipment.

[0098] Safety margin is an additional safety distance added to further enhance safety. Considering the uncertainties that may exist in the actual environment, such as measurement errors and environmental changes, safety margin can provide a certain buffer space on top of the basic safety distance to ensure that the inspection equipment can maintain a sufficient safe distance from obstacles under various conditions.

[0099] The target safety distance coefficient is a corresponding coefficient selected from a preset set of safety distance coefficients based on the specific type of obstacle. Different types of obstacles pose different levels of threat to inspection equipment. The target safety distance coefficient is used to adjust the basic safety distance and safety margin to determine the differentiated safety distance for that type of obstacle.

[0100] In some examples, the specific values ​​of the basic safety distance and safety margin can be determined by consulting relevant power inspection safety specifications, equipment performance parameters, and past experience data; a correspondence table between obstacle types and safety distance coefficients can be established in advance, and after identifying the type of obstacle, the corresponding target safety distance coefficient can be found and determined according to the table; using a specific calculation formula, the basic safety distance, safety margin, and target safety distance coefficient are substituted to calculate the differentiated safety distance for that obstacle.

[0101] In some embodiments, the differentiated safety factor is a quantitative risk model set based on the safety regulations of the power industry, the physical characteristics of various facilities, and their potential impact on UAV flight. It defines the collision risk of general static obstacles as a benchmark, and on this basis, comprehensively assesses the risks of complex structures of towers, electromagnetic interference and wind sway risks of high-voltage transmission lines, and the multiple superimposed risks of extremely strong electromagnetic interference and dense equipment in substations.

[0102] Furthermore, data on safety incidents related to obstacles during past power line inspections can be collected to analyze the probability and severity of safety accidents caused by different types of obstacles, and the corresponding target safety distance coefficient can be determined through data modeling. For example, if data analysis reveals that transmission lines have a high probability of causing safety accidents, their target safety distance coefficient can be determined to be 1.8.

[0103] Differential safety distances can be calculated in the following ways:

[0104] ;

[0105] in, This refers to the safety distance factor. The basic safety distance (can be taken as 2m) For safety margin (0.5m can be taken).

[0106] Figure 2 This application provides a diagram illustrating the safe distances corresponding to different obstacle types, such as... Figure 2 As shown, after calculation using the above method, when the target safety distance coefficient for high-voltage transmission lines is 2.5, the corresponding differentiated safety distance is 5.5m; when the target safety distance coefficient for poles and towers is 1.5, the corresponding differentiated safety distance is 3.5m; when the target safety distance coefficient for substation equipment is 3.0, the corresponding differentiated safety distance is 6.5m; and when the target safety distance coefficient for general obstacles (such as trees) is 1.0, the corresponding differentiated safety distance is 2.5m.

[0107] By assigning coefficients proportional to their risk levels to different facilities, abstract safety concepts can be transformed into concrete, calculable path constraints. This enables path planning to adjust safety distances reasonably based on different obstacle types and actual conditions when facing complex power environments, better ensuring the safe operation of power inspection equipment. Through intelligent and differentiated risk avoidance, it can optimize safety and efficiency, further improving the quality and reliability of power inspection path planning.

[0108] Based on the above embodiments, the method further includes: when the type of obstacle is a high-voltage transmission line or a substation, constructing an electromagnetic field strength attenuation model; determining a no-fly zone based on the electromagnetic field strength attenuation model and a preset maximum electromagnetic field strength threshold for flight; treating the no-fly zone as an obstacle and adding obstacle information; and / or adjusting the electromagnetic safety objective in the multi-objective optimization function based on the no-fly zone.

[0109] In this embodiment, the electromagnetic field strength attenuation model can refer to a mathematical model used to describe the attenuation of electromagnetic field strength as it changes with factors such as distance and spatial location. In power facility environments, high-voltage transmission lines and substations generate electromagnetic fields, and this model can help predict the magnitude of electromagnetic field strength at different locations. The maximum electromagnetic field strength threshold for flight can refer to the highest electromagnetic field strength value that can be withstood to ensure the normal and safe flight of inspection equipment such as drones. When the electromagnetic field strength in the environment exceeds this threshold, it may cause interference or even damage to the electronic systems and communication systems of the inspection equipment. In no-fly zones, if the electromagnetic field strength exceeds the maximum electromagnetic field strength threshold for flight, inspection equipment such as drones may face safety risks when entering the zone, and therefore, flight is prohibited.

[0110] In some examples, high-voltage transmission lines and substations generate strong electromagnetic fields during operation, and the distribution of these fields follows certain patterns. Therefore, it is necessary to construct appropriate models based on their characteristics and relevant physical principles to describe the attenuation of the electromagnetic field intensity. For instance, based on electromagnetic theory, the Biot-Savart law and other formulas relating electromagnetic field intensity to factors such as distance can be derived for high-voltage transmission lines, thus constructing an electromagnetic field intensity attenuation model. For example, for infinitely long direct-drive transmission lines, the electromagnetic field intensity is inversely proportional to the distance; this theoretical formula can be used to construct a model describing its attenuation law. In actual power facility environments, the electromagnetic field intensity around high-voltage transmission lines or substations can be measured at multiple points to obtain a large amount of measured data. Then, mathematical fitting methods, such as polynomial fitting and neural network fitting, can be used to construct an electromagnetic field intensity attenuation model based on the measured data. For example, by measuring the electromagnetic field intensity around a substation at different distances and angles, a neural network can be used to train this data to obtain a model that accurately describes the attenuation of the electromagnetic field intensity.

[0111] Using a pre-constructed electromagnetic field strength attenuation model, the electromagnetic field strength at different locations is calculated. The calculation results are compared with a preset maximum electromagnetic field strength threshold for flight, and areas where the electromagnetic field strength exceeds the threshold are identified as no-fly zones. For example, the environment of power facilities can be divided into small grids, and the electromagnetic field strength at the center point of each grid can be calculated using the electromagnetic field strength attenuation model. The calculation results are compared with the maximum electromagnetic field strength threshold for flight, and the grids with electromagnetic field strength exceeding the threshold are marked. The area formed by these marked grids is the no-fly zone. Alternatively, starting from the vicinity of high-voltage transmission lines or substations, the search gradually moves outwards, using the electromagnetic field strength attenuation model to calculate the electromagnetic field strength at different locations. When the electromagnetic field strength approaches the maximum electromagnetic field strength threshold for flight, the boundary points are precisely determined, and connecting these boundary points forms the boundary of the no-fly zone, thus defining the no-fly zone.

[0112] In one example, an electromagnetic field strength attenuation model is established for high-voltage transmission lines and substations:

[0113] ;

[0114] in, It is the electric field strength on the surface of the facility; It refers to the distance from the facility; It is the characteristic distance of electric field attenuation (related to voltage level).

[0115] when At that time, the area was marked as a no-fly zone. This is the threshold for the highest electromagnetic field strength during flight.

[0116] Then, the identified no-fly zones are treated as special obstacles, and their relevant information is added to the existing obstacle information for avoidance during path planning. Alternatively / simultaneously, the electromagnetic safety objective in the multi-objective optimization function is optimized and adjusted based on the no-fly zone situation, making the path planning more compliant with electromagnetic safety requirements. For example, the electromagnetic safety objective originally has a certain weight in the multi-objective optimization function. This weight is adjusted based on the size and location of the no-fly zone. If the no-fly zone is large and close to the inspection path, the weight of the electromagnetic safety objective is increased, making the path planning more inclined to avoid the area; conversely, the weight is appropriately reduced. Another example is adding no-fly zones as new constraints to the multi-objective optimization function to strengthen electromagnetic safety requirements. For instance, during optimization, not only is it required that the electromagnetic safety index of the path meet a certain range, but the path must also maintain a certain safe distance from the no-fly zone, thus adjusting the electromagnetic safety objective.

[0117] By constructing an electromagnetic field intensity attenuation model to determine no-fly zones and incorporating them into obstacle information or adjusting electromagnetic safety targets, the impact of electromagnetic safety factors on power line inspection path planning can be considered more accurately. This effectively prevents inspection equipment from entering areas with strong electromagnetic fields, ensuring the safe operation of inspection equipment. At the same time, it makes path planning more in line with actual electromagnetic environment requirements, improving the scientificity and reliability of power line inspection path planning.

[0118] Based on the above embodiments, an initial global path is obtained through a path search algorithm based on obstacle information and differentiated safety distances. This includes: concurrently initiating a bidirectional search process from the starting point and the ending point of the path planning using the path search algorithm; dynamically adjusting the size of the search neighborhood based on obstacle information and differentiated safety distances around the current node during the bidirectional search process; and expanding the node based on the dynamically adjusted search neighborhood to obtain the initial global path.

[0119] In this embodiment, the bidirectional search process can refer to a path search strategy that simultaneously starts from both the starting and ending points of the path planning and searches in the opposite direction. Compared to traditional unidirectional search, bidirectional search can find feasible paths faster and improve path search efficiency.

[0120] In pathfinding algorithms, the region surrounding the current node is called the search neighborhood. Within this region, the algorithm searches for the next expandable node. The size of the search neighborhood affects the search range and efficiency; a larger search neighborhood allows for exploring a wider area but increases computational cost, while a smaller search neighborhood has the opposite effect. During pathfinding, starting from the current node, a suitable node is selected as the next node within its search neighborhood according to certain rules; this process is called node expansion. By continuously expanding nodes, a path from the starting point to the destination is gradually constructed.

[0121] In some embodiments, (1) a suitable path search algorithm is selected, such as Algorithms such as Dijkstra's algorithm perform path searching simultaneously from both the starting and ending points. Bidirectional search fully utilizes information from both the starting and ending points, narrowing the search range more quickly and improving search efficiency. For example, using... The algorithm starts the search from the starting point and the ending point respectively. The algorithm guides the search direction by estimating a function, thus improving search efficiency. For example, in power line inspection route planning, it starts simultaneously from both the starting point and the end point. The algorithm selects the next node to expand based on the cost estimate of each node (including actual distance and heuristic distance). Another example is the use of Dijkstra's algorithm for bidirectional search, which expands nodes incrementally based on the shortest path principle in a graph. For instance, in a relatively regular power facility layout environment, Dijkstra's algorithm can be used to search simultaneously from the start and end points, calculating the shortest distance from each node to both.

[0122] (2) During the search process, for each current node, the search neighborhood is adjusted in real time based on the distribution of obstacles around it and the pre-determined differentiated safety distance. If there are many obstacles around the current node and the safety distance requirement is high, the search neighborhood is appropriately reduced to avoid searching infeasible areas; conversely, if there are few obstacles around it and the safety distance is relatively loose, the search neighborhood can be expanded to speed up the search. In some examples, the number of obstacles within a certain range around the current node can be counted, and the size of the search neighborhood can be adjusted according to the obstacle density. If the obstacle density is high, it indicates that the area is more complex and the search neighborhood needs to be reduced; conversely, if the obstacle density is low, the search neighborhood can be expanded. For example, an obstacle density threshold can be set, and when the number of surrounding obstacles exceeds the threshold, the search neighborhood is reduced to half of its original size. In another example, an influence weight can be set for each obstacle based on the differentiated safety distance of different obstacles. The size of the search neighborhood is dynamically adjusted by comprehensively considering the influence weight of surrounding obstacles. For example, for obstacles with high safety distance requirements, a larger influence weight is given, and the search neighborhood is appropriately reduced when such obstacles exist around it.

[0123] (3) Based on the adjusted search neighborhood, select a node that meets the conditions within the neighborhood of the current node for expansion. By continuously expanding nodes, the bidirectional search processes gradually move closer to each other. When the search processes in both directions meet, connect the meeting nodes to obtain the initial global path from the starting point to the ending point. For example, a cost function for node expansion can be defined, taking into account factors such as the distance from the node to the ending point and the safe distance from obstacles. Select the node with the minimum cost within the search neighborhood for expansion. For example, the cost function can be expressed as the sum of the actual distance from the node to the ending point and a safe distance penalty term. By calculating the cost of each node, select the node with the minimum cost as the next node to be expanded.

[0124] Alternatively, within the search neighborhood, some nodes can be expanded according to certain rules (such as the shortest distance principle), while others can be expanded randomly. This approach can increase search diversity while maintaining search efficiency and avoiding getting trapped in local optima. For example, within the search neighborhood, 80% of the nodes can be selected according to the shortest distance principle, and 20% can be selected randomly.

[0125] By employing a bidirectional search process, the efficiency of path search is improved, enabling the discovery of feasible paths more quickly. Simultaneously, the size of the search neighborhood is dynamically adjusted based on obstacle information and differentiated safety distances, and nodes are expanded according to the adjusted neighborhood. This makes path search more flexible and intelligent, better adapting to complex power facility environments and avoiding the problems of low search efficiency or failure to find feasible paths due to fixed search neighborhoods. This results in a more reasonable initial global path, providing a solid foundation for subsequent path optimization.

[0126] Based on the above embodiments, the size of the search neighborhood is dynamically adjusted based on obstacle information around the current node and differentiated safety distance, including: determining the ratio of the number of obstacles within the search radius of the current node to the total number of grid cells based on obstacle information around the current node; determining the local environment complexity index based on differentiated safety distance and the ratio; and dynamically adjusting the size of the search neighborhood based on the local environment complexity index.

[0127] In path planning, the search radius refers to a pre-defined fixed distance range centered on the current node. This range forms a circular (in two-dimensional space) or spherical (in three-dimensional space) area. Within this range, obstacle detection and analysis are performed to assist in path search decisions.

[0128] When the path planning space is divided into discrete grids, the total number of grids contained within the search radius of the current node is the total number of grids. Spatial rasterization is a common method for discretizing continuous space, which facilitates the quantification and computation of spatial information.

[0129] The local environment complexity index is a quantitative indicator used to measure the complexity of the local environment surrounding the current node. It comprehensively considers factors such as the number of obstacles within the current node's search radius and differentiated safety distances, and is calculated using a specific method to reflect the difficulty of path planning at the current location.

[0130] In this embodiment, firstly, the search range around the current node is determined based on a preset search radius. Then, the number of grid cells occupied by obstacles within this search range is counted, and the total number of grid cells within the search radius is determined. Finally, the ratio of the number of obstacles to the total number of grid cells is calculated; this ratio can preliminarily reflect the density of obstacles around the current node. For example, image processing or spatial analysis techniques are used to accurately scan and count the grid cells within the search radius of the current node. For instance, in a two-dimensional grid map, by traversing each grid cell and determining whether it is an obstacle cell, the number of obstacles and the total number of grid cells are accurately counted, and the ratio is calculated. As another example, when the search radius is large, a sampling estimation method can be used to save computation time. A certain number of sample grid cells are randomly selected within the search radius, the number of obstacles in the samples is counted, and the ratio of the number of obstacles to the total number of grid cells within the entire search radius is estimated based on the sample proportion. For example, 10% of the total number of grid cells can be selected as a sample for statistics.

[0131] Differentiated safety distances are different safety distance requirements set for different types of obstacles. Combining the previously calculated ratio of the number of obstacles to the total number of grid cells, a local environment complexity index is derived by fusing the differentiated safety distance and this ratio using a specific algorithm or formula. This index more comprehensively considers the distribution of obstacles and the impact of safety distance requirements on the local environment. Furthermore, a weighted summation method can be used to determine the index: weights are assigned to the differentiated safety distance and the ratio of the number of obstacles to the total number of grid cells, and the local environment complexity index is calculated by weighted summation. For example, the weight of the differentiated safety distance can be set to 0.6, and the weight of the ratio to 0.4; the weighted values ​​of the two can be added together to obtain the index value. Alternatively, the principles of fuzzy logic can be used to map the differentiated safety distance and the ratio to fuzzy sets, and inference and calculation can be performed using fuzzy rules to derive the local environment complexity index. For example, different fuzzy rules can be defined based on different ranges of safety distances and ratios, such as "if the safety distance is large and the ratio is small, then the local environment complexity is low."

[0132] Based on the calculated local environment complexity index, different threshold ranges are set. When the local environment complexity index falls within different threshold ranges, the size of the search neighborhood is adjusted accordingly. For example, when the local environment complexity is high, the search neighborhood is reduced to avoid searching too many infeasible regions; when the local environment complexity is low, the search neighborhood is expanded to improve search efficiency. Furthermore, multiple threshold ranges for the local environment complexity index can be pre-defined, each threshold range corresponding to a fixed search neighborhood size. When the calculated index value falls within a certain threshold range, the search neighborhood is directly adjusted to the corresponding fixed size. For example, three threshold ranges can be set, corresponding to small, medium, and large search neighborhood sizes respectively. Optionally, a continuous functional relationship between the local environment complexity index and the search neighborhood size can be established. Based on the calculated index value, the corresponding search neighborhood size is calculated in real time using this function. For example, the functional relationship could be that the search neighborhood size is inversely proportional to the local environment complexity index; the larger the index value, the smaller the search neighborhood.

[0133] By dynamically adjusting the search neighborhood size using a local environment complexity index, path search can more accurately adapt to the characteristics of different local environments. In areas with dense obstacles and complex environments, the search neighborhood is appropriately reduced to avoid unnecessary searches and calculations; in areas with simple environments, the search neighborhood is expanded to improve search efficiency. This overall optimizes the path search process and improves the quality and efficiency of path planning.

[0134] Based on the above embodiments, the size of the search neighborhood is dynamically adjusted based on the local environment complexity index, including: when the local environment complexity index is less than or equal to the lower threshold, the search neighborhood adopts a first search step size; when the local environment complexity index is greater than or equal to the upper threshold, the search neighborhood adopts a second search step size; when the local environment complexity index is greater than the lower threshold and less than the upper threshold, the search neighborhood adopts a third search step size; wherein, the first search step size is greater than the third search step size, and the third search step size is greater than the second search step size.

[0135] In this embodiment, the search step size refers to the distance or span moved when expanding outwards from the current node during path search. It determines the magnitude of each expansion within the search neighborhood; different search step sizes affect the search speed and accuracy. The lower threshold is a pre-defined minimum critical value for the local environment complexity index. When the local environment complexity index is less than or equal to this value, it indicates that the local environment around the current node is relatively simple. The upper threshold is a pre-defined maximum critical value for the local environment complexity index. When the local environment complexity index is greater than or equal to this value, it means that the local environment around the current node is relatively complex.

[0136] Furthermore, the calculated local environment complexity index is compared with a pre-set lower threshold. If the index is less than or equal to the lower threshold, it indicates that the current local environment is simple. In this case, the search neighborhood uses a first search step size; a larger search step size can speed up the search and quickly explore a larger area in a simple environment. The local environment complexity index is then compared with an upper threshold. If the index is greater than or equal to the upper threshold, it indicates that the current local environment is complex. In this case, the search neighborhood uses a second search step size; a smaller search step size can make the search more refined, avoiding missing feasible paths or entering infeasible regions. When the local environment complexity index is greater than the lower threshold but less than the upper threshold, it indicates that the current local environment complexity is at a moderate level. In this case, the search neighborhood uses a third search step size, which is between the first and second search step sizes to balance search speed and accuracy.

[0137] Optionally, based on past path planning experience in similar power facility environments and considering the environmental characteristics of different areas, lower and upper thresholds can be set. For example, in previous planning, when the obstacle ratio is less than 20%, the environment is relatively simple, and when it is greater than 50%, the environment is complex. In this case, the lower threshold can be set to 0.2, and the upper threshold to 0.5. Alternatively, extensive simulation experiments can be conducted to test path search under different local environmental complexities, and the lower and upper thresholds can be adjusted and determined based on the search results (such as search time and path quality).

[0138] Optionally, the first, second, and third search step sizes can be determined proportionally based on the size of the search space and overall planning requirements. For example, the maximum size of the search space can be set as a reference, with the first search step size set to 1 / 5 of the maximum size, the second search step size to 1 / 10 of the maximum size, and the third search step size to 1 / 8 of the maximum size. Alternatively, the search step size can be dynamically adjusted during the path search process based on real-time search conditions and remaining planning tasks. For instance, when the remaining area to be searched is large, the search step size can be appropriately increased; when approaching the target point or encountering complex areas, the search step size can be decreased.

[0139] By flexibly employing different search step sizes to adjust the search neighborhood, pathfinding can better adapt to local environments of varying complexity. Larger step sizes are used in simple environments to improve search efficiency, while smaller step sizes are used in complex environments to ensure search accuracy, further optimizing the pathfinding process.

[0140] Based on the above embodiments, when the search neighborhood adopts the first search step size, the constraints for generating the extended node include: in the direction from the current node to the extended node, all intermediate nodes located between the current node and the extended node are not occupied by obstacles; and the shortest distance from the path segment connecting the current node and the extended node to all obstacles within the surrounding preset range is greater than the differentiated safety distance of the corresponding obstacle.

[0141] In this embodiment, an extension node refers to a node that, during the path search process, is explored outward from the current node according to certain rules and may become part of the subsequent path. It is a key element in path search used to extend the path; by continuously generating extension nodes, a complete path is gradually constructed.

[0142] Intermediate nodes are the nodes traversed on the path from the current node to the extended node. When determining whether a path is feasible, it is necessary to check whether these intermediate nodes are occupied by obstacles to ensure the connectivity of the path.

[0143] A path segment is a straight line segment connecting the current node and the extended node; it represents the direct path from the current position to the extended position. When considering path safety, it is necessary to analyze the relationship between this path segment and surrounding obstacles.

[0144] The preset range is a pre-defined area surrounding the path segment, used to define the area where the impact of obstacles needs to be considered. Obstacles are detected within this range to assess the safety of the path.

[0145] Furthermore, when determining the first search step size for generating the extended node, the first step is to check all intermediate nodes located between the current node and the extended node in the direction from the current node to the extended node. If none of these intermediate nodes are occupied by obstacles, it indicates that the path in that direction is connected at the node level, meeting the preliminary conditions for generating the extended node. For example, each intermediate node can be checked sequentially from the current node to the extended node. Using pre-built environmental map information, the status of each intermediate node can be queried to determine whether it is occupied by an obstacle. Alternatively, a region containing all intermediate nodes can be defined using the current node and the extended node as endpoints, and the entire region can be scanned to quickly determine whether there are intermediate nodes occupied by obstacles.

[0146] Furthermore, the shortest distance from the path segment connecting the current node and the extended node to all obstacles within a preset range is calculated. This shortest distance is then compared to the differentiated safety distance of the corresponding obstacle. Only when the shortest distance is greater than the differentiated safety distance of the corresponding obstacle is another constraint for generating the extended node satisfied, ensuring a sufficient safety distance between the path and the obstacle. The comparison can be performed as follows: for each obstacle, the shortest distance from the path segment to that obstacle is calculated using geometric knowledge. For example, when the obstacle is circular, the distance from the path segment to the center of the circle is calculated minus the radius of the circle; when the obstacle is polygonal, the shortest distance is determined by calculating the distances between the path segment and each side of the polygon. The calculated shortest distance is then compared to the corresponding differentiated safety distance.

[0147] By clearly defining the constraints for generating extended nodes when using the first search step size, it is helpful to improve search efficiency by utilizing larger step sizes while ensuring the feasibility and safety of the generated paths. This avoids ignoring potential obstacles or creating unsafe paths due to larger step sizes, thereby further improving the quality of path planning.

[0148] As can be seen from the above embodiments, the path search algorithm can be dynamically adjusted according to the type of obstacles around the current node and the safe distance. At the same time, combined with bidirectional search starting from the start point and the end point simultaneously, the single large-scale search problem is decomposed into two small-scale searches, which greatly reduces the search space and computation time. Figure 3 A flowchart illustrating the initial global path determination method provided in this application is shown below. Figure 3 As shown, this embodiment, based on the above embodiments, provides a detailed description of the method for determining the initial global path using a path search algorithm. The path search algorithm includes... Algorithm and bidirectional search mechanism. Further, the method includes:

[0149] S301, Power Inspection Guided Smart Neighborhood algorithm.

[0150] It should be noted that, The path length planned by the algorithm can be expressed as: ;

[0151] in, This is the final path length. Indicates from node To the node The Euclidean distance is given by n, where n is the number of nodes on the path. Under the same heuristic function, the optimal node selected will differ due to different search neighborhoods, leading to variations in the final path length.

[0152] Let the search range be R. According to path planning theory, the path length is inversely proportional to the search range: ;

[0153] This formula shows that a larger search range corresponds to a larger exploration space, thus offering the opportunity to escape local optima and increase the probability of finding a shorter path. To solve the traditional... To address the problem of algorithmic search strategy fixation under different obstacle densities, this step employs an intelligent neighborhood expansion method based on environmental complexity quantification, where:

[0154] (1) Adaptive neighborhood switching of environmental complexity: The local environmental complexity index is defined as follows:

[0155] ;

[0156] in, This represents the number of obstacles within the current node's search radius. This represents the total number of grid cells within the search radius. This represents the average safety factor of the surrounding obstacles.

[0157] (2) Adaptive neighborhood selection strategy: Its core lies in relying on two key preset thresholds: a lower threshold and a lower threshold. With an upper limit threshold The algorithm divides the environmental complexity into intervals and selects the optimal search step size accordingly. Specifically, the lower threshold is defined as follows: when the environmental complexity is below this value, the collision risk of using the maximum step size for a jump-based search is extremely low, and the efficiency gain far outweighs the safety risk, marking the critical point where the environment transitions from "semi-open" to "open." The upper threshold is defined as follows: when the environmental complexity is above this value, the distribution of obstacles is sufficient to form narrow passages or complex clusters, and any search strategy with a step size larger than the standard will lead to a sharp drop in the success rate of path planning, marking the critical point where the environment transitions from "semi-open" to "dense." Therefore, the setting of these two thresholds essentially quantifies the inverse relationship between efficiency and safety, thus endowing the algorithm with the intelligence to select the optimal strategy in different environments.

[0158] Furthermore, when When, the expansion strategy is: expand the neighborhood (maximum step size is 3); when When, the expansion strategy is: expand the neighborhood (maximum step size is 2); when At that time, the expansion strategy is: standard eight-neighborhood.

[0159] Figure 4 This application provides a schematic diagram of adaptive neighborhood selection based on environmental complexity. Figure 4As shown, in the open region a, the step size of the extended neighbor node corresponding to the current node S is 3; in the moderately complex region b, the step size of the extended neighbor node corresponding to the current node S is 2; and in the dense region c, the current node S uses an eight-neighbor node with a step size of 1.

[0160] Based on the aforementioned adaptive neighborhood strategy, a specific extended node generation rule was designed to address the characteristics of power line inspection scenarios. Let the current node coordinates be... The collection of obstacles to power facilities is Based on the traditional eight-neighbor domain orientation, the algorithm adds extended nodes in a specific direction when certain directional constraints are met. Taking a two-dimensional plane as an example (the extension is similar in three-dimensional space), the specific constraints for the extended nodes are as follows:

[0161] 1) Extended node constraints:

[0162] AND ;

[0163] Add a new node (relative coordinates): (1,2).

[0164] 2) Extended node constraints:

[0165] AND ;

[0166] Add a new node (relative coordinates): (2,1).

[0167] 3) Extended node constraints:

[0168] AND ;

[0169] Add a new node (relative coordinates): (2, -1).

[0170] in, Candidate paths to obstacles The shortest distance; Obstacles The required safe distance.

[0171] S302, bidirectional search mechanism.

[0172] In this embodiment, to fundamentally reduce the search complexity in large-scale three-dimensional space, a bidirectional concurrent search strategy is adopted, decomposing the originally unidirectional search task into two smaller, opposite-direction subtasks that can be executed in parallel. Furthermore, combined with... The algorithm, which utilizes this bidirectional search mechanism to determine the initial global path, includes the following methods:

[0173] (1) Basic architecture of bidirectional search: The algorithm starts two independent search engines simultaneously from the initial position of the UAV (start) and the target position of the mission (target). The search process involves both forward and backward searches. The algorithm maintains two independent open and closed list data structures to manage the nodes to be evaluated and those already evaluated in each search direction, respectively.

[0174] When the search process is forward, the starting point is "start" and the target direction is "towards target," and its content is maintaining an independent open list. and close list When the search process proceeds backward, the starting point is `target`, the target direction is towards `start`, and its content is maintaining an independent open list. and close list .

[0175] Furthermore, the two search processes alternate in each iteration of the algorithm, each selecting a cost function from its open list. Expand the node with the smallest value. In the forward search, the cost function follows the classical... Definition, and its heuristic function Used to evaluate from any node The estimated cost to reach the target. Correspondingly, in the backward search, the heuristic function... This is used to evaluate from any node Estimated cost to reach the starting point:

[0176] ;

[0177] ;

[0178] To ensure the synergy and optimality of the two search processes, the heuristic function is designed following the consistency principle, that is, for any node... ,satisfy ,in This is a constant representing the actual minimum cost from the starting point to the ending point. If this condition is not met, then the following formula applies:

[0179] ;

[0180] ;

[0181] The heuristic function is corrected by taking the average value, and a new heuristic function is generated. and This ensures that the searches in both directions are based on the same evaluation benchmark.

[0182] Furthermore, in some embodiments, the aforementioned intelligent neighborhood expansion and bidirectional search mechanisms can be integrated into a unified path planning algorithm through a collaborative decision-making process. The top-level architecture of this algorithm is bidirectional search, where intelligent neighborhood expansion is called as the core subroutine in its node expansion phase. The algorithm execution begins with the concurrent initiation of two independent paths from the start and end points. In the search process, in each iteration, the algorithm bases its search on its respective cost function. Find the minimum value, and select an optimal node n from the forward or backward open list for processing.

[0183] The core decision-making basis of the algorithm is Cost function Its definition is: This function is used to quantitatively evaluate the potential of any node n as part of the optimal path. Wherein, It is the actual path cost, precisely calculated as the cumulative three-dimensional Euclidean distance from the starting point (starting point or ending point) in this search direction to the current node n. This is a heuristic cost, using three-dimensional Euclidean straight-line distance to estimate the shortest possible distance from node n to the target point (end point or start point) in the search direction. This cost function is applied symmetrically to both the forward and backward search processes and corrected by averaging, as described above, to ensure that the two search processes proceed collaboratively under a unified and balanced evaluation benchmark.

[0184] After determining the node n to be expanded, the decision logic described in S301 is immediately invoked. This logic first checks for obstacles such as power facilities around node n. Perform local environmental perception and based on obstacle complexity The algorithm dynamically switches between an eight-neighborhood for fine-grained search and an expanded neighborhood for efficient exploration. If the expanded neighborhood is chosen, the algorithm strictly adheres to the expanded node constraints outlined above, ensuring the safety of large-step movements through path clearing checks. This mechanism is applied equally to both search directions.

[0185] (2) Search intersection and termination conditions: The core advantage of bidirectional search lies in the efficient intersection of the two search fronts. The termination conditions of the algorithm are designed as follows:

[0186] ① Intersection detection: In each iteration, the current search proceeds from its open list. Extract a node When expanding, immediately check if the node already exists in the backward search's closed list. If it exists, it means that the two search directions are already at the node. The nodes successfully intersected. Symmetrically, when the backward search extracts nodes... At the same time, the same check is performed.

[0187] ② Path reconstruction: Once the intersection node is detected Immediately stop both search processes and perform path reconstruction.

[0188] a. from By traversing the parent node pointer forward to start, the sub-path is obtained. ;

[0189] b. From By backtracking the parent node pointer along the backward search path to the target, the sub-path is obtained. ;

[0190] c. Concatenate the two path segments: ,in, This indicates that the backward sub-path will be reversed.

[0191] This method can generate a high-quality initial global path that conforms to the characteristics of the power grid inspection environment with extremely high computational efficiency while ensuring safety. After obtaining the initial global path, multiple objective functions are defined based on a multi-objective optimization framework to generate an optimal path set. Finally, the optimal path is selected according to the inspection task mode, which improves the efficiency of path planning and adaptability to multiple scenarios while ensuring safety.

[0192] Since the target global path generated by the path planning algorithm usually contains a large number of redundant turning points, these redundant turning points not only increase the path length but also cause the UAV to turn frequently, significantly increasing energy consumption and flight time. Therefore, based on the above embodiment, after determining the target global path from the Pareto optimal path set based on a preset inspection task mode, the method may further include: identifying all turning points in the target global path; identifying the turning points that pass geometric collision detection and speed feasibility detection as redundant turning points and removing them; wherein, geometric collision detection is used to determine whether the line segment connecting adjacent nodes after removal intersects with the obstacle; speed feasibility detection is based on the UAV's cruise speed, turning dynamics model, and the turning angle of the line segment to determine whether the UAV can complete the line segment flight without deceleration.

[0193] In this embodiment, an inflection point refers to a point in the path where the direction changes. In the UAV inspection path, an inflection point is the location where the path direction changes, determining the change in the UAV's flight direction. Redundant inflection points, on the other hand, are those inflection points in the target's global path that, after detection, can be removed without affecting the basic functionality and safety of the path. Specifically, these are inflection points that meet the requirements of geometric collision detection and speed feasibility detection, and whose adjacent node connection segments remain feasible after removal.

[0194] Geometric collision detection is a method used to determine the positional relationship between path segments and obstacles. It determines the safety and feasibility of a path by detecting whether line segments connecting adjacent nodes after removing inflection points intersect with obstacles in the surrounding environment. Speed ​​feasibility detection is based on the performance parameters of the UAV and path characteristics. It comprehensively considers the UAV's cruise speed, turning dynamics model, and the turning angle of the line segments to determine whether the UAV can smoothly complete turns and flight between line segments without deceleration while flying along the path. The turning dynamics model can be a mathematical model describing the physical characteristics of the UAV during turning. It considers factors such as the UAV's mass, inertia, and propulsion system to analyze the UAV's motion state at different turning angles and speeds.

[0195] In some examples, a defined global path to the target is analyzed to identify all points along the path where the direction changes—the inflection points. These inflection points divide the path into multiple segments. For instance, inflection points are identified by analyzing the coordinate changes of points along the path; when the direction of coordinate change between adjacent points changes significantly, that point can be identified as an inflection point. For example, in a two-dimensional path, if the slope of the line connecting two adjacent points changes significantly, the midpoint may be an inflection point. Alternatively, the direction vectors of each point along the path can be calculated; when the direction vector changes significantly, the corresponding point is an inflection point. For example, a threshold for the change in direction vector can be set; when the change exceeds this threshold, it is identified as an inflection point.

[0196] In geometric collision detection, for each potentially culled inflection point, it is assumed that after culling, connecting its two adjacent nodes forms a new line segment. Then, a geometric collision detection method is used to check if this new line segment intersects with obstacles in the surrounding environment. If they intersect, the inflection point cannot be culled; if they do not intersect, the inflection point satisfies the culling condition in terms of geometric collision. One example uses collision detection algorithms from computer graphics, such as bounding box detection and polygon collision detection. Path segments and obstacles are represented as corresponding graphic objects, and algorithms determine whether they intersect. For example, obstacles and path segments can both be represented as polygons, and polygon collision detection algorithms can be used for judgment. In another example, the shortest distance from a path segment to an obstacle is calculated; if the shortest distance is less than or equal to zero, it indicates that the line segment intersects with the obstacle. This can be achieved by calculating the shortest distance from a point to a line segment, or from a line segment to another line segment, using geometric formulas.

[0197] During speed feasibility testing, based on the UAV's cruise speed, turning dynamics model, and the turning angle of the line segment, the analysis examines whether the UAV can complete flight without deceleration when turning from one line segment to another. If the turning and flight can be completed smoothly, the inflection point meets the elimination condition in terms of speed feasibility. For example, UAV simulation software can be used to simulate the UAV's flight process at different turning angles and speeds. The simulation results determine whether the UAV can complete line segment flight without deceleration. Alternatively, mathematical formulas can be established based on the UAV's turning dynamics model to calculate the minimum turning radius and speed changes required by the UAV at different turning angles. By comparing these calculations with the UAV's actual performance parameters, it can be determined whether flight can be completed without deceleration.

[0198] When a certain inflection point simultaneously meets the requirements of geometric collision detection and velocity feasibility detection, it is identified as a redundant inflection point and removed from the target global path.

[0199] Furthermore, Figure 5 The inflection point path diagram provided in this application is as follows: Figure 5 As shown in the figure, from the starting point to inflection point ②, inflection point ① is completely redundant. The goal of this method is to find and eliminate these redundant inflection points. In some embodiments, based on this embodiment, an inflection point optimization method combining environmental adaptive threshold and velocity constraint is further proposed to optimize inflection points in the target global path. This method may include:

[0200] (1) Inflection point identification and path representation: the target global path generated by global planning In, among them, Starting from, As the endpoint, These are nodes on the path.

[0201] 1) Calculate from point arrive Direction vector :

[0202] ;

[0203] 2) Calculation and the previous direction vector The angle between :

[0204] ;

[0205] 3) Unlike traditional methods that use a fixed angle threshold, this embodiment uses a node-based approach. The judgment threshold is dynamically adjusted based on the environmental complexity.

[0206] ;

[0207] in, It is a very small basic angle threshold. It is the environmental impact factor. It is a node The complexity of the environment. If Then determine Add the inflection point to the inflection point set. .

[0208] Using the above method, in areas with dense obstacles, the threshold is increased to retain more inflection points, ensuring the path can flexibly bypass obstacles; in open areas, the threshold is decreased to more aggressively identify and eliminate redundant inflection points. After identifying all candidate inflection points, it is necessary to further determine which inflection points are truly necessary and which can be safely eliminated. Traditional methods only check geometric collisions; this embodiment adds a velocity feasibility check based on UAV dynamic constraints.

[0209] That is, the starting point and ending point are also added to the inflection point set to form a new node sequence:

[0210] ;

[0211] in, , , Let the current inspection starting point be... ,index Perform an iteration.

[0212] (2) Feasibility check for eliminating the inflection point of speed perception:

[0213] 1) Geometric collision check: Attempt direct connection and Check line segments Does it intersect with obstacles?

[0214] ;

[0215] If they intersect, then It is a necessary inflection point, update Then start the check from the next point again.

[0216] 2) Speed ​​feasibility check: even if the line segment Even without navigating obstacles, it is necessary to verify whether the drone can complete this segment of flight at cruising speed. Calculate the average rate of change of angle for this line segment. :

[0217] ;

[0218] in This represents the change in angle from the direction of the previous path segment to the direction of the current line segment. This refers to the drone's cruising speed.

[0219] According to the turning dynamics of the drone, the minimum speed required is: ;in, This is the minimum turning radius of the drone (related to the drone's mass and maximum lateral thrust).

[0220] like The drone can complete the turn at cruising speed and continue to check the next point. ;like The turn was too aggressive, and the drone needed to slow down to complete it. It is a necessary inflection point and cannot be eliminated.

[0221] 3) Optimal Path Reconstruction: After the above inflection point removal, the remaining inflection points constitute the optimized path:

[0222] ;

[0223] in, To optimize the number of line segments retained, This represents a straight line segment connecting two necessary inflection points.

[0224] This embodiment introduces an environmentally adaptive threshold, enabling the path optimization process to possess intelligent perception capabilities. In complex environments with dense obstacles, it automatically tightens the judgment criteria, retaining crucial turning points for safe avoidance; while in open environments with sparse obstacles, it relaxes the criteria, maximizing path straightening and eliminating unnecessary redundant maneuvers, thus achieving an optimal balance between safety and efficiency. Furthermore, by introducing a speed feasibility check, the optimization dimension is elevated from purely geometric space to the physical dynamics level. This filters out treacherous paths that, while geometrically collision-free, cannot be smoothly completed at cruising speed physically, significantly reducing unnecessary deceleration and acceleration processes caused by unreasonable paths during actual flight.

[0225] Therefore, by identifying inflection points in the target's global path and using geometric collision detection and velocity feasibility detection to determine and eliminate redundant inflection points, the path structure is simplified, reducing unnecessary turning and flight operations. This not only helps improve inspection efficiency but also reduces the energy consumption of inspection equipment, while ensuring the safety and feasibility of the path, making the UAV inspection path more optimized.

[0226] However, while the path optimized by the inflection points is significantly simplified, it still consists of straight line segments, with sharp angle abrupt changes at the inflection points. As a physical inspection device, the drone cannot achieve instantaneous directional changes; flying directly along such a path would result in excessive acceleration, affecting flight stability and image acquisition quality. Therefore, based on the above embodiments, the method may further include: for each inflection point in the target global path after redundant inflection points are removed, dynamically determining the arc smoothing radius based on the drone's current flight speed and the inflection point's turning angle; generating an arc trajectory to replace the inflection point based on the arc smoothing radius; and verifying whether the points on the arc trajectory meet the safe distance requirements from nearby obstacles.

[0227] In this embodiment, the arc smoothing radius can be the size of the arc radius determined when smoothing the path inflection points after removing redundant inflection points. It determines the curvature of the generated arc trajectory; the larger the radius, the smoother the arc; the smaller the radius, the more curved the arc.

[0228] To make the drone's flight smoother when passing through inflection points, an arc with a certain radius is used to replace the original inflection point. The flight path formed by this arc is called the arc trajectory.

[0229] The safe distance requirement is the minimum distance that a point on the drone's flight path must maintain with respect to nearby obstacles to ensure safety during flight. This distance is preset based on factors such as the drone's performance and environmental characteristics.

[0230] In some examples, the smoothing radius of the arc can be determined using a specific algorithm or rule based on the drone's current flight speed and the inflection angle of the inflection point. Generally, the faster the flight speed and the larger the inflection angle, the larger the required smoothing radius of the arc may be to ensure that the drone can smoothly pass through the inflection point. Furthermore, a mathematical formula can be established relating the smoothing radius of the arc to the drone's current flight speed and the inflection angle of the inflection point. Alternatively, the corresponding smoothing radius of the arc can be pre-determined through experiments or simulations based on different combinations of flight speeds and inflection angles, and then compiled into a table. In practical applications, the corresponding smoothing radius of the arc can be directly looked up from the table based on the drone's current flight speed and the inflection angle of the inflection point.

[0231] Then, with a determined smoothing radius, an arc trajectory is generated at the inflection point to replace it. This arc trajectory should smoothly connect with the original path at the inflection point, ensuring the drone can naturally transition to the new flight direction. Furthermore, geometric drawing methods can be used to draw the arc trajectory in two-dimensional or three-dimensional space based on the smoothing radius and the inflection point's location. For example, in a two-dimensional plane, an arc is drawn with the inflection point as the center (in some cases, this may not be the center, depending on the specific geometric relationships) and according to the smoothing radius. Optionally, computer algorithms, such as Bézier curve algorithms and circular interpolation algorithms, can also be used to generate the arc trajectory based on given parameters such as the starting point, ending point, and smoothing radius. These algorithms can accurately calculate the coordinates of each point on the arc.

[0232] Finally, each point on the generated circular trajectory is checked, and the distances between these points and nearby obstacles are calculated to determine if they all meet the pre-set safety distance requirements. If all points meet the requirements, the circular trajectory is feasible; if any points do not meet the requirements, the smoothing radius of the arc needs to be readjusted or other measures need to be taken. The detection methods can be: point-by-point detection, which involves selecting points along the circular trajectory at certain intervals and calculating the distance between each point and nearby obstacles; or area scanning, which involves setting a scanning area centered on the circular trajectory and detecting obstacles within that area.

[0233] It should be noted that, in some embodiments, a bivariate adaptive circular arc smoothing method considering both flight speed and turning angle is further proposed based on this embodiment. This method performs circular arc smoothing only locally at inflection points, keeping the straight line segments between inflection points unchanged, thereby preserving the advantages of the original path to the greatest extent. This method may include:

[0234] (1) Velocity-Angle Bivariate Adaptive Smoothing Coefficient: In the traditional circular arc smoothing method, the smoothing coefficient... Typically, these values ​​are fixed or only related to the inflection point angle, making them unsuitable for varying flight speeds. According to UAV turning dynamics:

[0235] ;

[0236] in, For the quality of drones, For flight speed, The turning radius, For the required centripetal force.

[0237] Due to the maximum lateral force of the drone Limited, therefore: ;

[0238] Turning radius It must be proportional to the square of the velocity. Furthermore, the turning angle... The larger the radius, the greater the smoothing radius is needed for buffering.

[0239] Based on the above physical constraints, this method proposes a bivariate adaptive smoothing coefficient:

[0240] ;

[0241] in, This is the basic smoothing coefficient, with a value range of 0.8 to 1.2, and a default value of 1.0. This is the speed influence factor, with a value range of 0.3 to 0.7, used to adjust the degree of influence of speed on the smoothing radius; This is the angle influence factor, with a value range of 0.2 to 0.5, used to adjust the degree of influence of the turning angle on the smoothing radius; Current flight speed; This is the maximum flight speed of the drone.

[0242] The design of this formula ensures the dynamic feasibility of the trajectory, and it makes the smoothness coefficient... Able to respond simultaneously to changes in flight speed and turning angle (e.g.) Figure 6 As shown): During high-speed flight, it automatically increases the turning radius to meet the laws of physics; when facing sharp turns, it can provide more buffer space, thereby generating a smooth trajectory that is both safe and efficient.

[0243] Among them, parameters and The specific value should be determined based on the dynamic characteristics of the UAV and actual flight tests, and must meet the following constraints:

[0244] Speed ​​Constraint: At maximum flight speed, the turning radius should not exceed the drone's maximum maneuverability. By conducting turning flight tests at different speeds, the minimum radius at which the drone can stably complete a turn is recorded. A speed-radius relationship curve is fitted to determine... The appropriate value.

[0245] Angle Constraint: For different turning angles, the turning radius should increase with the angle. This is ensured through flight tests at different turning angles. The value of allows the drone sufficient room for smooth turning at large angles.

[0246] (2) Improved formula for calculating the radius of the arc: It should be noted that the geometric principle of the local arc smoothing method is as follows: Figure 7As shown, this method aims to replace a sharp corner with a smooth arc, and this arc must be precisely tangent to the two straight paths before and after the inflection point, thus ensuring a smooth transition for the drone. The method includes the following steps:

[0247] 1) Calculate the geometric parameters of the inflection point:

[0248] Turning angle Calculate the path segment before the inflection point. and the next path The angle between This angle describes the severity of the turn;

[0249] ;

[0250] Angle bisector : Calculate the unit vector along the direction of the angle bisector of this angle. The center of the tangent circle must lie on the angle bisector.

[0251] ;

[0252] 2) Adaptively determine the radius of the circle :

[0253] Based on current flight speed and turning angle Calculate the adaptive smoothing coefficient Then calculate the radius of the arc:

[0254] ;

[0255] 3) Determine the location of the center:

[0256] Calculate the distance from the center of the circle to the inflection point distance : ;

[0257] The coordinates of the center of the circle are from the inflection point Start along the angle bisector Directional movement Distance:

[0258] ;

[0259] 4) Generate circular arc trajectory:

[0260] When used to smooth inflection points Coordinates of the center of the tangential circle and radius Once determined, a circular arc trajectory is generated to replace the inflection point. :

[0261] Establish a reference coordinate system for generating the circular arc trajectory: Define the coordinates of the center of the circle. Pointing to the inflection point The polar angle of the vector of coordinates is the reference polar angle. The angle can be calculated using an arctangent function that can identify all quadrants, ensuring the global uniqueness and accuracy of the angle calculation.

[0262] ;

[0263] Calculate the angular range of the circular arc trajectory: The angular range is defined by a starting angle and an ending angle, and this range is relative to the aforementioned reference polar angle. Symmetrical distribution. The total angular span of the angular range is determined as follows: ,in For the path at the inflection point The turning angle at the point. Therefore, the angular parameters of the circular arc trajectory. The range of values ​​for is limited to:

[0264] ;

[0265] The parametric equation of a circle is used to generate a series of path points that constitute the trajectory of the circular arc: specifically, by adjusting the angle parameters... Within the previously calculated angle range, the algorithm iterates from the starting angle to the ending angle with a preset increment step. In each iteration, the formula is used:

[0266] ;

[0267] Calculate a set of corresponding path point coordinates The coordinates of all path points generated through iteration are combined sequentially to form a smooth circular arc trajectory. This generated circular arc trajectory is then used to replace the inflection points in the original path data. This allows for the smoothing of the inflection point, thereby forming an optimized path with continuous tangential direction at the original inflection point location.

[0268] 5) Safety Verification and Feedback Adjustment Mechanism: The generated arc may approach or even intrude into the safe distance of obstacles due to its excessive radius. Therefore, a sampling verification and feedback adjustment mechanism was designed, including:

[0269] Sampling point generation: Uniform sampling along the arc Let there be 1 point, denoted as 1. ;

[0270] Safety distance check: for each sampling point Calculate its distance to the nearest obstacle distance :

[0271] ;

[0272] Verification conditions: Check Does the safe distance requirement for the obstacle meet? ;

[0273] Feedback Adjustment: If all sampling points meet the safety distance requirements, the arc smoothness passes verification, and the original inflection point is replaced with the arc trajectory; if not, the following adjustments are performed. This reduces the radius of the arc, regenerates the arc, and verifies it again.

[0274] Iteration Termination: Repeat the above adjustment process until the arc passes the safety verification, or If the value shrinks below a certain minimum, then abandon the smoothing process at that inflection point and maintain the original path.

[0275] 6) Final output of the smoothed trajectory: After performing the above smoothing process on all inflection points on the path, the final smoothed trajectory is obtained:

[0276] ;

[0277] Each of them It could be a straight line segment (between two inflection points) or a circular arc segment (at the inflection point).

[0278] By smoothing the target global path inflection points after removing redundant inflection points, dynamically determining the smoothing radius and generating the arc trajectory, and then verifying the safe distance requirements of the points on the trajectory, the flight of the inspection equipment when passing through inflection points is made smoother and more stable, reducing energy loss and flight risks caused by sudden turns, while ensuring the safety of the flight process.

[0279] Because power line inspection involves complex path planning problems requiring sequential access to multiple devices (such as poles and insulator strings) according to specific priorities, this method, based on the above embodiments, may further include: acquiring multiple devices to be inspected and the access priority of each device; determining a device inspection sequence based on the access priorities of all devices using a priority-penalty-based sequence optimization model, wherein the device inspection sequence includes multiple inspection points, each inspection point indicating the inspection order of the corresponding device; adjusting the order of waypoints in the target global path based on the device inspection sequence; for each inspection point, determining a set of candidate locations that meet preset photographing conditions, wherein the preset photographing conditions include spatial geometric constraints and temporal behavior constraints, wherein the spatial geometric constraints include distance constraints and / or shooting angle constraints between the device and the device; and the temporal behavior constraints include hovering stabilization time requirements; selecting locations that meet preset path cost requirements from the candidate location set as actual path points; and generating an updated target global path based on the actual path points corresponding to all inspection points and the order of waypoints.

[0280] In this embodiment, access priority can be used to measure the urgency or importance of the equipment to be inspected. Higher-priority equipment typically needs to be inspected earlier to ensure that the status of critical equipment can be monitored in a timely manner. Access priorities can be manually set for each piece of equipment by professional inspection personnel based on factors such as equipment type, importance, and operating status. Alternatively, data analysis algorithms can be used to automatically calculate and generate the access priority for each piece of equipment based on data such as historical fault records, operating time, and impact on production.

[0281] Priority-penalty-based sequence optimization models consider device access priorities and determine the optimal device inspection sequence by evaluating and adjusting different inspection sequences. When high-priority devices in the inspection sequence are not prioritized, the model imposes a penalty, thereby prompting the generation of sequences that better meet priority requirements. For example, simulated annealing algorithms, by simulating the metal annealing process, allow a certain degree of "inferior solutions" during the search process to avoid getting trapped in local optima, thus finding a globally optimal inspection sequence that better meets priority requirements. Another example is using genetic algorithms, treating different inspection sequences as individuals and continuously optimizing them through selection, crossover, and mutation operations to better meet the evaluation criteria based on priority penalties.

[0282] An inspection point refers to an identifier point in the equipment inspection sequence, corresponding to each piece of equipment to be inspected, used to indicate the inspection order of that equipment. It is a key node when planning the inspection path.

[0283] Spatial geometric constraints refer to the restrictions imposed on the spatial relationship between the inspection point and the equipment when determining candidate locations. These include distance constraints (the appropriate distance between the inspection point and the equipment) and shooting angle constraints (the angle requirements that must be met when photographing the equipment from the inspection point). Temporal behavior constraints refer to the restrictions imposed on time-related factors during the inspection process. Here, this mainly refers to the hovering stability time requirement, which is the shortest time the drone needs to maintain a stable state when hovering at the inspection point for operations such as photography.

[0284] The candidate location set refers to the set of all possible locations that satisfy the preset imaging conditions (spatial geometric constraints and temporal behavior constraints) for each inspection point. The preset path cost requirement can be a constraint on path cost (such as path length, flight time, energy consumption, etc.) when selecting actual path points from the candidate locations. The selected location should keep the path cost within an acceptable range. The actual path points are the locations that satisfy the preset path cost requirement from the candidate location set corresponding to each inspection point; these points will serve as key nodes in generating the updated global path to the target.

[0285] Furthermore, spatial search methods can be used to search within a certain spatial range around the inspection point, following specific step sizes and directions, and to calculate whether each search point satisfies spatial geometric constraints and temporal behavioral constraints. A model of the equipment and its surrounding environment can also be built, and the location that meets the imaging conditions can be predicted based on the model. Then, by calculating the path cost corresponding to each location in the candidate location set, such as path length and flight time, the location with the lowest cost is selected as the actual path point. When the path cost includes multiple aspects (such as simultaneously considering path length and energy consumption), a multi-objective optimization algorithm can also be used to weigh the various objectives and select the location with the best overall performance as the actual path point.

[0286] In some examples, a list of devices requiring inspection can be clearly defined, and an access priority can be assigned to each device. Priorities can be determined manually or automatically based on device importance. The device access priorities are input into a sequence optimization model, which comprehensively considers various possible inspection sequences and selects the optimal sequence based on a priority penalty mechanism. This sequence determines the inspection order for each device. Based on the determined device inspection sequence, the order of waypoints in the original target global path is adjusted to match the device inspection order. For each inspection point, spatial geometric constraints (distance from the device, shooting angle) and temporal behavior constraints (hovering stabilization time) are considered, and all locations satisfying these conditions are identified, forming a candidate location set. From the candidate location set, suitable locations are selected as actual path points based on preset path cost requirements (such as shortest path, lowest energy consumption, etc.). All actual path points corresponding to all inspection points are connected according to the adjusted waypoint order to form the updated target global path.

[0287] In some embodiments, based on this embodiment, a power inspection task characteristic integration method is further proposed to improve the problem that existing methods do not consider the special needs of power inspection. This method may include:

[0288] (1) Multi-inspection point task sequence optimization: A priority-penalized sequence optimization model is constructed. The goal of this model is to find an optimal inspection point access order to minimize the overall cost. This overall cost function is defined as:

[0289] ;

[0290] in, This represents the total flight path length, aiming for optimal efficiency; It is a priority penalty item; It is a penalty imposed on delayed access to high-priority inspection points; This is the priority weight of that point. This design ensures that high-priority tasks (i.e., If a critical node (with a high value) is placed at the back of the access order, it will cause a sharp increase in the total cost, thus driving the algorithm to prioritize the planning of these critical nodes.

[0291] To address this issue, this embodiment employs an efficient genetic algorithm. The algorithm encodes each possible inspection point arrangement as a "chromosome" and uses the reciprocal of the total cost as the "fitness function" to evaluate its quality. Through genetic operations such as partial mapping crossover (PMX) and exchange mutation, the algorithm iteratively generates a population with a better access order until a preset convergence condition is met (e.g., reaching the maximum number of iterations or no improvement for multiple consecutive generations). Finally, it outputs an optimal inspection sequence that achieves the best balance between flight distance and task priority. The multi-inspection point sequence optimization result is shown below. Figure 8 As shown.

[0292] (2) Photo Taking Posture Constraints: To closely integrate path planning with the actual requirements of power line inspection operations, refined photo taking posture constraints are introduced to ensure that the UAV can not only "reach" the inspection point, but also "complete" the inspection task. Furthermore, effective photo taking behavior must simultaneously meet three core conditions:

[0293] 1) Distance between the drone and the inspection point It needs to be kept within the preset range ;

[0294] 2) The angle between the camera's optical axis and the normal to the device surface must be less than 30° to ensure image quality;

[0295] 3) There must be a hovering and stabilizing time of at least 2 seconds before taking the photo.

[0296] Therefore, during the path planning phase, abstract inspection points are no longer considered. Instead of acting as a necessary path node, it intelligently generates a set of candidate locations around it where all points meet the photo-taking criteria. .

[0297] Finally, from this set of feasible locations, a point that optimizes the global path (e.g., the shortest total length or the lowest energy consumption) is selected as the actual path transit point and seamlessly connected to the final flight trajectory, thus achieving a deep integration of mission objectives and path planning.

[0298] By acquiring device access priorities and using a sequence optimization model to determine the inspection sequence, adjusting the order of path points, and then combining the photo conditions to determine candidate locations and select actual path points, an updated target global path is finally generated, making the inspection path planning more reasonable. This not only considers the importance of the equipment and the inspection sequence, but also takes into account the photo requirements and path costs, thus improving inspection efficiency and quality.

[0299] To achieve accurate perception and proactive avoidance of dynamic environments, based on the above embodiments, the method may further include: acquiring the current flight speed of the UAV while it is conducting power line inspection based on the target global path; detecting dynamic obstacles in real time using onboard sensors and predicting their trajectories; calculating the estimated collision time and minimum expected distance based on the current path of the UAV and the predicted trajectory of the dynamic obstacle; triggering local path replanning if the estimated collision time is less than the warning time threshold and / or the minimum expected distance is less than the safe distance of the dynamic obstacle; generating a detour path using a simplified path planning algorithm within the local search range, starting from the current position of the UAV, the detour path is used to guide the UAV to avoid dynamic obstacles and reconnect to the target global path; performing adaptive inflection point optimization on the target global path or detour path and eliminating redundant inflection points; dynamically calculating the arc smoothing radius based on the current flight speed of the UAV and the path turning angle, and performing speed adaptive arc smoothing processing on the optimized path.

[0300] In this embodiment, dynamic obstacles can refer to objects that move and whose positions change over time in the drone inspection environment, such as birds, other aircraft, moving vehicles or people. The motion state of these obstacles is uncertain and may pose a threat to the flight safety of the drone.

[0301] The estimated collision time is calculated based on the drone's current path and the predicted trajectory of the dynamic obstacle, estimating the remaining time before a collision with the obstacle is likely. It is a crucial time indicator for determining whether obstacle avoidance measures are necessary. The warning time threshold can be a pre-set time value. When the estimated collision time is less than this threshold, it indicates a high probability of a collision between the drone and the dynamic obstacle, and time is of the essence, requiring the triggering of obstacle avoidance measures.

[0302] The minimum expected distance is the minimum possible distance between the drone and a dynamic obstacle during drone flight, based on the current path and the predicted trajectory of the dynamic obstacle. This distance measures the proximity of the drone to the dynamic obstacle. The dynamic obstacle safety distance is the minimum distance that must be maintained between the drone and the dynamic obstacle to ensure the drone's safety. This distance is set based on factors such as the drone's performance and the characteristics of the dynamic obstacle.

[0303] Local path replanning refers to replanning a path that avoids the dynamic obstacle within a certain range near the drone's current position when a potential collision with a dynamic obstacle is detected.

[0304] The local search range refers to a spatial area centered on the UAV's current position during local path replanning. Within this range, a suitable path is searched to avoid dynamic obstacles. Simplified path planning algorithms, compared to global path planning algorithms, are computationally less complex and faster in local path replanning, capable of generating feasible detour paths in a short time. The detour path, generated through local path replanning, guides the UAV to avoid dynamic obstacles and rejoin the target's global path after obstacle avoidance.

[0305] Adaptive inflection point optimization automatically identifies and removes unnecessary inflection points in the path based on the actual conditions, making the path simpler and smoother. Speed-adaptive circular smoothing dynamically determines the circular smoothing radius based on the drone's current flight speed and path turning angle, smoothing the path to make the drone more stable when turning.

[0306] In some examples, the drone's speed information during flight is acquired in real time via speed sensors or other relevant devices. Sensors onboard the drone, such as radar and cameras, monitor the surrounding environment, promptly detect dynamic obstacles, and predict their future trajectories based on their motion. Based on the drone's flight path and the predicted trajectory of the dynamic obstacle, relevant mathematical models and algorithms are used to calculate the estimated collision time and minimum expected distance. The estimated collision time is compared with a warning time threshold, and the minimum expected distance is compared with a safe distance from the dynamic obstacle. If the estimated collision time is less than the warning time threshold and / or the minimum expected distance is less than the safe distance from the dynamic obstacle, local path replanning is triggered. After determining the local search range, a simplified path planning algorithm is used to search within this range and generate a detour path that avoids the dynamic obstacle and ultimately rejoins the target global path. The original target global path or the newly generated detour path is analyzed to identify and remove unnecessary inflection points, further optimizing the path. Based on the drone's current flight speed and the turning angle of the path, a suitable arc smoothing radius is calculated, and this radius is then used to smooth the optimized path, making the drone more stable during flight.

[0307] Furthermore, simplification can be adopted. The algorithm generates a three-segment detour path; the three-segment detour path is then subjected to speed-adaptive circular arc smoothing, wherein the speed-adaptive circular arc smoothing dynamically adjusts the smoothing radius according to the flight speed and turning angle.

[0308] In some embodiments, based on this embodiment, a complete real-time dynamic obstacle detection and evaluation process can be further set up to dynamically optimize the target's global path. This process first receives and processes continuous data streams from airborne LiDAR or visual sensors, and initiates a three-stage processing pipeline: the first stage is target detection, utilizing an advanced YOLOv8 deep learning model to quickly identify potential dynamic threats such as birds and other aircraft from complex backgrounds; once a target is locked, the trajectory tracking stage immediately begins, employing a Kalman filter to optimally estimate the target's motion state, which not only smooths noise and stabilizes tracking but also predicts its trajectory in the near future based on its historical motion data; finally, in the collision risk assessment stage, the minimum expected distance between the predicted trajectory of the obstacle and the UAV's current target global path is accurately calculated. The entire process ultimately outputs a structured list of dynamic obstacles in real time. ={(location, speed, type)} provides an accurate and reliable basis for subsequent path replanning decisions.

[0309] The local path replanning triggering mechanism can be:

[0310] ;

[0311] in, It is the predicted closest distance. It is the safe distance for dynamic obstacles. This is the estimated collision time. It is the warning time threshold.

[0312] To achieve agile response to unexpected situations such as dynamic obstacles, this embodiment proposes an efficient and computationally low-cost local path replanning strategy. Once a collision risk assessment triggers the replanning mechanism, the existing globally optimal path is not completely discarded; instead, it is used as the final regression target. The core of the replanning lies in initiating a "micro" planning task within a limited local search radius, starting from the UAV's current position. To ensure absolute real-time decision-making, this local planning employs a simplified... The algorithm's optimization objective is simplified to the core factors of path length and safety margin, aiming to find a safe and efficient temporary detour as quickly as possible. For example... Figure 9 As shown, when a dynamic obstacle is detected and the collision point is determined, an optimal safe detour point is found. Then, a temporary trajectory consisting of three segments—"current position → detour point → smooth reconnection to the global path"—is generated. The original path segment is seamlessly replaced by this locally adjusted new path, thus achieving rapid and accurate avoidance of sudden threats without affecting the overall macro-optimal nature of the overall plan.

[0313] By acquiring the drone's flight speed in real time, detecting dynamic obstacles and predicting its trajectory, calculating collision-related indicators to trigger local path replanning, generating detour paths, and then performing inflection point optimization and speed-adaptive circular arc smoothing, the drone can effectively and timely avoid dynamic obstacles during power line inspections, while ensuring path smoothness and flight stability. This improves the safety and reliability of power line inspections and reduces inspection interruptions and accidents caused by collision risks.

[0314] Figure 10 A comparative diagram of the power line inspection route planning method provided in this application and traditional planning methods is shown below. Figure 10 As shown, when using poles and other obstacles as obstacle information for path planning, the power inspection path planning method provided in this application provides a smoother path. By reducing broken lines and height abrupt changes, the difficulty of motion control is reduced. It also actively detours around areas with dense obstacles, avoiding close contact with obstacle edges, thus improving safety. At the same time, it has a shorter path length and uniform height changes, reducing the energy consumption and movement time of the inspection equipment and improving inspection efficiency.

[0315] Based on the above embodiments, the power inspection path planning method provided in this application first classifies and marks power facilities, sets differentiated safety distances according to different types such as transmission lines, towers, and substations, and establishes an electromagnetic field influence model; then, it adopts an improved intelligent neighborhood... The algorithm and bidirectional search mechanism dynamically adjust the search strategy based on environmental complexity. Then, the NSGA-II algorithm is used to establish a multi-objective optimization framework, comprehensively considering five objectives: path length, number of turns, safety margin, smoothness, and electromagnetic safety. This generates a Pareto optimal solution set, and the optimal path is selected according to the task mode. Next, local path optimization is performed through environment-adaptive inflection point optimization and speed-adaptive circular arc smoothing. Simultaneously, it integrates power line inspection-specific functions such as multi-inspection point sequence optimization and image posture constraints. Finally, Kalman filtering is used to predict dynamic obstacle trajectories, achieving rapid local replanning. This improves the safety, efficiency, and completeness of power line inspection.

[0316] Figure 11 This is a schematic diagram of the power inspection route planning device provided in this application, such as... Figure 11 As shown, the power inspection route planning device 110 provided in this embodiment includes:

[0317] The acquisition module 1101 is used to acquire obstacle information of the power facility environment, including the type and location of the obstacles;

[0318] The determination module 1102 is used to determine a differentiated safety distance based on the type of obstacle and a preset safety distance coefficient, wherein the type of obstacle includes power transmission lines, substations, towers and other obstacles;

[0319] Module 1103 is obtained, which is used to obtain the initial global path based on obstacle information and differentiated safety distance through a path search algorithm;

[0320] The generation module 1104 is used to optimize the initial global path based on a preset multi-objective optimization function to generate a Pareto optimal path set. The multi-objective optimization function includes path length objective, safety margin objective and electromagnetic safety objective.

[0321] Select module 1105 is used to determine the target global path from the Pareto optimal path set based on the preset inspection task mode.

[0322] In one possible implementation, the determining module 1102 can also be used to: obtain a basic safety distance and a safety margin; determine a corresponding target safety distance coefficient from a preset safety distance coefficient based on the type of obstacle; and determine a differentiated safety distance of the obstacle based on the basic safety distance, the safety margin, and the target safety distance coefficient.

[0323] In one possible implementation, the determining module 1102 can also be used to: construct an electromagnetic field strength attenuation model when the type of obstacle is a high-voltage transmission line or a substation; determine a no-fly zone based on the electromagnetic field strength attenuation model and a preset maximum electromagnetic field strength threshold for flight; treat the no-fly zone as an obstacle and add obstacle information; and / or adjust the electromagnetic safety objective in the multi-objective optimization function based on the no-fly zone.

[0324] In one possible implementation, module 1103 can also be used to: concurrently initiate a bidirectional search process from the starting point and the ending point of the path planning using a path search algorithm; dynamically adjust the size of the search neighborhood based on obstacle information and differentiated safety distances around the current node during the bidirectional search process; and expand the node based on the dynamically adjusted search neighborhood to obtain an initial global path.

[0325] In one possible implementation, module 1103 can also be used to: determine the ratio of the number of obstacles within the search radius of the current node to the total number of grid cells based on obstacle information around the current node; determine the local environment complexity index based on the differentiated safety distance and the ratio; and dynamically adjust the size of the search neighborhood based on the local environment complexity index.

[0326] In one possible implementation, module 1103 can also be used to: when the local environment complexity index is less than or equal to the lower threshold, the search neighborhood adopts a first search step size; when the local environment complexity index is greater than or equal to the upper threshold, the search neighborhood adopts a second search step size; when the local environment complexity index is greater than the lower threshold and less than the upper threshold, the search neighborhood adopts a third search step size; wherein, the first search step size is greater than the third search step size, and the third search step size is greater than the second search step size.

[0327] In one possible implementation, the selection module 1105 can also be used to: identify all inflection points in the target global path; identify the inflection points that pass geometric collision detection and speed feasibility detection as redundant inflection points and remove the redundant inflection points; wherein, geometric collision detection is used to determine whether the line segment connecting adjacent nodes after removal intersects with the obstacle; speed feasibility detection is based on the UAV's cruise speed, turning dynamics model and the turning angle of the line segment to determine whether the UAV can complete the line segment flight without deceleration.

[0328] In one possible implementation, the selection module 1105 can also be used to: dynamically determine the arc smoothing radius for each inflection point in the target global path after removing redundant inflection points, based on the current flight speed of the UAV and the turning angle of the inflection point; generate an arc trajectory to replace the inflection point based on the arc smoothing radius; and verify whether the points on the arc trajectory meet the safe distance requirements with nearby obstacles.

[0329] In one possible implementation, the selection module 1105 can also be used to: acquire multiple devices to be inspected and the access priority of each device; determine a device inspection sequence based on the access priorities of all devices using a priority-penalty-based sequence optimization model, wherein the device inspection sequence includes multiple inspection points, each inspection point indicating the inspection order of the device corresponding to that inspection point; adjust the order of waypoints of the target global path based on the device inspection sequence; for each inspection point, determine a set of candidate locations that meet preset photography conditions, wherein the preset photography conditions include spatial geometric constraints and temporal behavior constraints, wherein the spatial geometric constraints include distance constraints and / or shooting angle constraints between the device and the device; and the temporal behavior constraints include hovering stabilization time requirements; select locations that meet preset path cost requirements from the candidate location set as actual path points; and generate an updated target global path based on the actual path points corresponding to all inspection points and the order of waypoints.

[0330] In one possible implementation, the selection module 1105 can also be used to: obtain the current flight speed of the UAV when it is performing power inspection based on the target global path; detect dynamic obstacles in real time through onboard sensors and predict their trajectories; calculate the expected collision time and minimum expected distance based on the current path of the UAV and the predicted trajectory of the dynamic obstacle; if the expected collision time is less than the warning time threshold and / or the minimum expected distance is less than the safe distance of the dynamic obstacle, trigger local path replanning; generate a detour path using a simplified path planning algorithm within the local search range, starting from the current position of the UAV, and use the detour path to guide the UAV to avoid dynamic obstacles and reconnect to the target global path; perform adaptive inflection point optimization on the target global path or detour path and eliminate redundant inflection points; dynamically calculate the arc smoothing radius based on the current flight speed of the UAV and the path turning angle, and perform speed adaptive arc smoothing processing on the optimized path.

[0331] The power inspection route planning device provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0332] Figure 12 A schematic diagram of the structure of the electronic device provided in this application. Figure 12 As shown, the electronic device 120 provided in this embodiment includes at least one processor 1201 and a memory 1202. Optionally, the device 120 further includes a communication component 1203. The processor 1201, the memory 1202, and the communication component 1203 are connected via a bus 1204.

[0333] In a specific implementation, at least one processor 1201 executes computer execution instructions stored in memory 1202, causing at least one processor 1201 to perform the above-described method.

[0334] The specific implementation process of processor 1201 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0335] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0336] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0337] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0338] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0339] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0340] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0341] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0342] The division of units is merely a logical functional division; 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 coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0343] 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.

[0344] In addition, the functional units in the various embodiments of the present invention 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.

[0345] If a function 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, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0346] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0347] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A power line inspection route planning method, characterized in that, include: Obtain obstacle information of the power facility environment, the obstacle information including the type and location of the obstacles; Based on the type of obstacle and a preset safety distance coefficient, a differentiated safety distance is determined, wherein the type of obstacle includes power transmission lines, substations, towers and other obstacles; Based on the obstacle information and the differentiated safety distance, an initial global path is obtained through a path search algorithm; Based on a preset multi-objective optimization function, the initial global path is optimized to generate a Pareto optimal path set. The multi-objective optimization function includes path length objective, safety margin objective, and electromagnetic safety objective. Based on the preset inspection task mode, the target global path is determined from the Pareto optimal path set.

2. The method according to claim 1, characterized in that, The determination of differentiated safety distances based on the type of obstacle and a preset safety distance coefficient includes: Obtain basic safety distances and safety margins; Based on the type of obstacle, a corresponding target safety distance coefficient is determined from a preset safety distance coefficient; The differentiated safety distance of the obstacle is determined based on the basic safety distance, the safety margin, and the target safety distance coefficient.

3. The method according to claim 1, characterized in that, The method further includes: When the type of obstacle is a high-voltage transmission line or a substation, an electromagnetic field strength attenuation model is constructed. Based on the electromagnetic field intensity attenuation model and the preset maximum electromagnetic field intensity threshold for flight, no-fly zones are determined; The no-fly zone is used as an obstacle, and the obstacle information is added. And / or, the electromagnetic safety objective in the multi-objective optimization function is adjusted based on the no-fly zone.

4. The method according to claim 1, characterized in that, The initial global path is obtained based on the obstacle information and the differentiated safety distance using a path search algorithm, including: The path search algorithm initiates a bidirectional search process concurrently from the starting point and the end point of the path planning. In the bidirectional search process, the size of the search neighborhood is dynamically adjusted based on obstacle information around the current node and the differential safety distance; The initial global path is obtained by expanding the nodes based on the dynamically adjusted search neighborhood.

5. The method according to claim 4, characterized in that, The dynamic adjustment of the search neighborhood size based on obstacle information and differential safety distance around the current node includes: Based on the obstacle information around the current node, determine the ratio of the number of obstacles within the search radius of the current node to the total number of grid cells; Based on the differentiated safety distance and the ratio, the local environment complexity index is determined; The size of the search neighborhood is dynamically adjusted based on the local environment complexity index.

6. The method according to claim 5, characterized in that, The dynamic adjustment of the search neighborhood size based on the local environment complexity index includes: When the local environment complexity index is less than or equal to the lower limit threshold, the search neighborhood adopts the first search step size; When the local environment complexity index is greater than or equal to the upper limit threshold, the search neighborhood adopts the second search step size; When the local environment complexity index is greater than the lower threshold and less than the upper threshold, the search neighborhood adopts a third search step size. Wherein, the first search step size is greater than the third search step size, and the third search step size is greater than the second search step size.

7. The method according to claim 6, characterized in that, When the search neighborhood adopts the first search step size, the constraints for generating the extended node include: in the direction from the current node to the extended node, all intermediate nodes located between the current node and the extended node are not occupied by obstacles; and the shortest distance from the path segment connecting the current node and the extended node to all obstacles within the surrounding preset range is greater than the differentiated safety distance of the corresponding obstacle.

8. The method according to any one of claims 1-7, characterized in that, After determining the target global path from the Pareto optimal path set based on the preset inspection task mode, the method further includes: Identify all inflection points in the target global path; Among all inflection points, those that pass both geometric collision detection and velocity feasibility detection are identified as redundant inflection points, and these redundant inflection points are then removed. The geometric collision detection is used to determine whether the line segment connecting adjacent nodes after removal intersects with the obstacle; the speed feasibility detection is based on the drone's cruising speed, turning dynamics model, and the turning angle of the line segment to determine whether the drone can complete the line segment flight without decelerating.

9. The method according to claim 8, characterized in that, The method further includes: For each inflection point in the target's global path after removing redundant inflection points, the radius of the circular arc is dynamically determined based on the UAV's current flight speed and the inflection point's turning angle. Based on the arc smoothing radius, generate an arc trajectory to replace the inflection point; Verify whether the points on the circular trajectory meet the safe distance requirements from nearby obstacles.

10. The method according to any one of claims 1-7, characterized in that, The method further includes: Obtain multiple devices to be inspected, as well as the access priority of each device; Based on the access priority of all devices, a device inspection sequence is determined through a priority-penalty-based sequence optimization model. The device inspection sequence includes multiple inspection points, and each inspection point is used to indicate the inspection order of the devices corresponding to that inspection point. Based on the equipment inspection sequence, adjust the order of the waypoints of the target global path; For each inspection point, a set of candidate locations that meet preset photography conditions is determined. The preset photography conditions include spatial geometric constraints and temporal behavior constraints. The spatial geometric constraints include distance constraints between the location and the device and / or shooting angle constraints. The temporal behavior constraints include hovering stabilization time requirements. Select a location from the candidate location set that meets the preset path cost requirement as the actual path point; Based on the actual path points corresponding to all inspection points and the order of the path points, an updated global target path is generated.

11. The method according to any one of claims 1-7, characterized in that, The method further includes: When the UAV is performing power line inspection based on the target global path, the current flight speed of the UAV is obtained; Dynamic obstacles are detected in real time using airborne sensors, and their trajectories are predicted. Based on the current path of the drone and the predicted trajectory of dynamic obstacles, calculate the estimated collision time and minimum expected distance; If the estimated collision time is less than the warning time threshold and / or the minimum expected distance is less than the dynamic obstacle safety distance, then local path replanning is triggered. Starting from the current position of the UAV, a simplified path planning algorithm is used to generate a detour path within the local search range. The detour path is used to guide the UAV to avoid dynamic obstacles and then reconnect to the target global path. Adaptive inflection point optimization is performed on the target global path or the detour path to eliminate redundant inflection points; Based on the current flight speed and path turning angle of the UAV, the radius of the circular arc is dynamically calculated, and the optimized path is subjected to speed-adaptive circular arc smoothing processing.

12. A power line inspection route planning device, characterized in that, include: The acquisition module is used to acquire obstacle information of the power facility environment, the obstacle information including the type and location of the obstacle; The determination module is used to determine a differentiated safety distance based on the type of obstacle and a preset safety distance coefficient, wherein the type of obstacle includes power transmission lines, substations, towers and other obstacles; The module is used to obtain an initial global path based on the obstacle information and the differentiated safety distance using a path search algorithm; The generation module is used to optimize the initial global path based on a preset multi-objective optimization function to generate a Pareto optimal path set. The multi-objective optimization function includes a path length objective, a safety margin objective, and an electromagnetic safety objective. The selection module is used to determine the target global path from the Pareto optimal path set based on a preset inspection task mode.

13. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-11.

Images