Distributed cooperative power inspection method and system for heterogeneous unmanned aerial vehicle cluster in complex dynamic environment

Abstract

The application provides a heterogeneous unmanned aerial vehicle cluster distributed cooperative power inspection method and system facing complex dynamic environments, relates to the technical field of heterogeneous unmanned aerial vehicle cluster distributed cooperative power inspection methods, and comprises the following steps: determining a range of allowed hop values at a current time, and determining a candidate hop value in combination with a priority inspection grid in the range of hop values; calculating an expected displacement vector of the unmanned aerial vehicle, so as to obtain a predicted motion trajectory of the unmanned aerial vehicle; performing cumulative calculation on the grid benefits covered by the predicted motion trajectory, and checking the candidate hop value; selecting an optimal hop value capable of maximizing the cumulative benefits from the candidate hop values that pass the check, and using the optimal hop value to update the motion state of the unmanned aerial vehicle at the next time; and performing multidimensional checking by introducing a hop value mechanism in combination with maneuverability constraints, environmental constraints and task conflict constraints, so that the motion planning of the unmanned aerial vehicle can dynamically adapt to the performance differences of heterogeneous platforms.

Description

Technical Field

[0001] This invention relates to the technical field of heterogeneous unmanned aerial vehicle (UAV) swarm distributed collaborative power line inspection methods, specifically to a heterogeneous UAV swarm distributed collaborative power line inspection method and system for complex dynamic environments. Background Technology

[0002] In recent years, UAV inspection technology has been widely used. However, existing methods mostly use single UAVs or homogeneous UAV swarms, and usually rely on preset routes or centralized command. They are difficult to adapt to real-time changes in complex dynamic environments, lack the ability to dynamically evaluate and prioritize mission benefits, and are prone to inefficient and repetitive inspections or omission of key targets. UAV motion planning does not fully consider the differences in maneuverability and environmental constraints of heterogeneous platforms, making it difficult to flexibly avoid obstacles and coordinate in scenarios with dense obstacles, overlapping no-fly zones, and dynamic weather changes. There is a lack of distributed coordination mechanism for mission conflicts and spatiotemporal conflicts between multiple UAVs, which can easily lead to target competition, route intersections, and even collision risks. Most existing methods use fixed step size or preset grid path planning, which cannot dynamically adjust the jump step size according to the real-time status, resulting in coarse trajectories and poor adaptability. Therefore, there is an urgent need for a distributed collaborative power inspection method and system for heterogeneous UAV swarms in complex dynamic environments.

[0003] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0004] The purpose of this invention is to provide a distributed collaborative power inspection method and system for heterogeneous UAV swarms in complex dynamic environments, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A distributed collaborative power line inspection method for heterogeneous UAV swarms in complex dynamic environments includes the following steps: S1: Acquire the boundary data of the inspection area, the data of the inspection target, and the real-time status data of each UAV. Divide the task grid based on the boundary data of the inspection area and the current position of the UAV, and generate a corresponding inspection benefit value for each task grid to build a task grid benefit table. Based on the task grid benefit table, filter the target grid from the grids that the UAV can reach, and determine the target grid with the highest benefit value as the current priority inspection grid. S2: Based on the relative orientation between the priority inspection grid and the current position of the UAV, the UAV's maneuverability parameters, and real-time status data, determine the allowable jump grid value range at the current moment, and within the jump grid value range, combine with the priority inspection grid to determine candidate jump grid values; S3: Based on the current heading angle, control steering command, and candidate jump values, calculate the heading angle of the UAV at the next moment, and normalize the heading angle at the next moment. S4: Map the normalized heading angle of the next moment to the unit circle, calculate the expected displacement vector of the UAV, and determine the coordinate increment of the next moment based on the expected displacement vector, thereby obtaining the predicted motion trajectory of the UAV. S5: Use the predicted motion trajectory to look up the task grid revenue table, accumulate the revenue of the grid covered by the predicted motion trajectory, and verify the candidate jump grid values ​​in combination with the maneuverability constraint, collision constraint and task conflict constraint; select the optimal jump grid value that maximizes the accumulated revenue from the candidate jump grid values ​​that pass the verification, and use the optimal jump grid value to update the UAV's motion state at the next moment.

[0006] Furthermore, a corresponding inspection reward value is generated for each task grid, including: Acquire the target type, defect level, historical anomaly records, and most recent inspection time of the corresponding inspection target within the task grid; acquire obstacle distribution information, meteorological risk information, and accessibility information for the area corresponding to the task grid. Based on target type, defect level, historical anomaly records, most recent inspection time, obstacle distribution information, meteorological risk information, and accessibility information, the benefits of each task grid are scored to form a task grid benefit table.

[0007] The benefit score includes an objective importance score, a regional accessibility score, and an environmental risk correction score, among which: The target importance score is used to characterize the inspection priority of the inspection target within the task grid, the regional accessibility score is used to characterize the ease with which the UAV can reach the corresponding task grid, and the environmental risk correction score is used to characterize the degree of influence of the external environment on the execution of the inspection task. The inspection benefit value is determined based on the weighted result of the target importance score, regional accessibility score and environmental risk correction score.

[0008] Furthermore, the target grid is filtered from the current drone-reachable grid, including: When determining the reachable grid set at the current moment, the criteria are based on the drone's current remaining battery power, current speed, maximum flight speed, maximum turning angular velocity, and the distance between the current position and each task grid: First, the maximum allowable range is calculated based on the remaining battery power. Then, the maximum displacement constraint is obtained by combining the product of the maximum flight speed and the decision cycle. Only when the distance from the task grid to the UAV's current position is less than or equal to the maximum flight speed multiplied by the decision cycle and the remaining battery power correction, is the grid included in the reachable grid set. Grids with inspection reward values ​​higher than a preset reward threshold are extracted from the reachable grid set as candidate target grids. When the number of candidate target grids is greater than one, the priority inspection grid is determined based on a comprehensive ranking result of inspection reward value, directional deviation from the current heading, and arrival cost. Specifically: For each task grid in the reachable grid set, the comprehensive score of each task grid is obtained by subtracting the direction deviation penalty factor from the inspection benefit value of the task grid, multiplying it by the absolute value of the difference between the grid direction angle of the task grid and the current heading angle of the UAV, and then dividing by pi. The task grid with the highest comprehensive score is selected as the priority inspection grid.

[0009] Based on the drone's current remaining battery power, current speed, maximum flight speed, maximum turning angular velocity, and the distance between the current position and each task grid, determine the set of reachable grids at the current moment; extract candidate target grids whose inspection reward value is higher than the preset reward threshold from the set of reachable grids.

[0010] Furthermore, determine the range of allowed jump values ​​at the current moment, including: The maximum displacement constraint and maximum steering constraint within a unit decision cycle are determined based on the maximum flight speed and maximum turning angular velocity of the UAV, and the minimum jump value and maximum jump value at the current moment are determined based on the maximum displacement constraint and maximum steering constraint. The minimum jump value is obtained by multiplying the UAV's minimum flight speed by the decision cycle and then dividing by the side length of the task grid; the maximum jump value is the smaller of the maximum flight speed multiplied by the decision cycle divided by the side length of the task grid and the maximum allowable range divided by the side length of the task grid; a jump value set is generated between the minimum jump value and the maximum jump value, where each jump value is an integer and must also satisfy the angle deviation constraint: that is, the absolute value of the angle deviation between the desired direction of the priority inspection grid relative to the current heading and the reachable direction corresponding to the jump value must be less than the preset angle deviation threshold.

[0011] Furthermore, a set of candidate jump values ​​is generated within the jump value range, including: The azimuth angle of the priority inspection grid relative to the current position of the UAV is used as the target guidance direction; the directional deviation between the theoretical arrival direction and the target guidance direction corresponding to each jump grid value is calculated; jump grid values ​​with directional deviation less than a preset angle threshold are retained as candidate jump grid values ​​and sorted in ascending order of directional deviation to obtain a set of candidate jump grid values.

[0012] Furthermore, a predicted motion trajectory is generated for each candidate hop value in the candidate hop value set, including: The heading angle for the next moment is calculated based on the current heading angle, control steering command, and corresponding candidate jump values. The specific calculation method is as follows: using the current heading angle of the UAV as a baseline, add the current control steering command multiplied by the UAV's maximum steering angle, and then multiply by the ratio of the candidate jump value to the maximum jump value. The control steering command ranges from -1 to +1, representing the intensity of a left or right turn. Subsequently, the heading angle is normalized by calculating the sine and cosine values ​​of the normalized heading angle, and then performing an arctangent operation on these two values ​​to limit the heading angle for the next moment within a preset angle range, thus obtaining the normalized heading angle corresponding to the candidate jump value for the UAV at the next moment. The normalized heading angle for the next time step is subjected to sine and cosine mapping to obtain the displacement direction component under the corresponding candidate jump grid value. Specifically, the cosine and sine values ​​of the heading angle corresponding to the candidate jump grid value of the UAV at the next time step are taken to form a two-dimensional column vector with a magnitude of 1, representing the displacement direction component of the UAV on the unit circle. At the same time, the displacement length component is obtained by combining the candidate jump grid value, that is, the candidate jump grid value multiplied by the side length of the task grid, representing the physical distance moved by the UAV within the decision cycle. Then, the coordinate increment of the next time step is determined based on the displacement direction component and the displacement length component, that is, the scalar of the displacement length component is multiplied by the two-dimensional vector of the displacement direction component to obtain the coordinate increment of the UAV's candidate jump grid value at the next time step. Further, the predicted motion trajectory corresponding to the candidate jump value is generated. Specifically, the coordinates of the current candidate jump value of the UAV are used as the reference, and the coordinate increment of the candidate jump value at the next moment is added to obtain the coordinates of the candidate jump value of the UAV at the next moment, thus completing the position state update. At the same time, the heading angle corresponding to the normalized candidate jump value of the UAV at the next moment is directly assigned as the heading angle of the candidate jump value of the UAV at the next moment, thus completing the heading state update.

[0013] Furthermore, by using the predicted motion trajectory corresponding to each candidate jump value to look up the task grid reward table, the cumulative reward value corresponding to each predicted motion trajectory is obtained, including: Identify the task grids that each predicted motion trajectory passes through in sequence, extract the inspection reward value of each passed task grid in the task grid reward table, and accumulate the corresponding inspection reward values ​​according to the order in which the predicted motion trajectory passes through each task grid. Then, correct the accumulation result according to the predicted dwell time of the UAV in each task grid, the proximity to the priority inspection grid, and the repeated inspection penalty factor to obtain the cumulative reward value of the corresponding candidate jump grid value. The specific correction method is as follows: For each task grid that the predicted trajectory passes through, its original inspection reward value is multiplied by an exponential function with the natural constant as the base and a negative dwell time decay coefficient multiplied by the predicted dwell time in that grid as the exponent, and then multiplied by an exponential function with the natural constant as the base and a negative repetition penalty coefficient multiplied by the number of times that grid has been inspected before the current decision cycle, to obtain the corrected reward contribution of that grid; the corrected reward contributions of all grids that have passed through are accumulated in the order of passing through, and an additional priority inspection grid proximity reward item is added. This reward item takes a value of one when the task grid passed through is exactly equal to the priority inspection grid, and a value of zero otherwise.

[0014] Furthermore, the validity of each candidate jump value is verified and the optimal jump value is selected, including: When the predicted motion trajectory corresponding to the candidate jump grid value does not exceed the maneuverability boundary of the UAV, the predicted motion trajectory does not enter any no-fly zone or obstacle-occupied area, and the minimum distance between the predicted motion trajectory of other UAVs is greater than the preset safety threshold, and the target task grid pointed to by the candidate jump grid value has not been preferentially occupied by other UAVs in the cluster during this decision cycle, the candidate jump grid value is considered to have passed the verification. At the same time, candidate jump grid values ​​that fail the verification are removed, and the candidate jump grid value with the largest cumulative benefit value is selected from the remaining candidate jump grid values ​​as the optimal jump grid value. When multiple candidate jump values ​​with the same cumulative benefit exist, the candidate jump value with the smallest directional deviation and the lowest energy consumption is selected as the optimal jump value. The optimal jump value is used to update the drone's motion state at the next moment. The specific update method is as follows: based on the coordinates of the drone's current candidate jump value, the coordinate increment of the drone's optimal jump value at the next moment is added to obtain the coordinates of the drone's optimal jump value at the next moment, thus completing the position state update; at the same time, the heading angle of the drone's optimal jump value at the next moment is directly assigned as the heading angle of the drone at the next moment after normalization, thus completing the heading state update.

[0015] The present invention also provides a heterogeneous UAV swarm distributed collaborative power inspection system for complex dynamic environments, the inspection system being used to execute the above-described inspection method, including: Priority Inspection Module: Used to acquire inspection area boundary data, inspection target data, and real-time status data of each UAV. Based on the inspection area boundary data and the current position of the UAV, it divides the task grid and generates a corresponding inspection benefit value for each task grid, and constructs a task grid benefit table. Based on the task grid benefit table, it selects target grids from the grids that the UAV can reach and determines the target grid with the highest benefit value as the current priority inspection grid. Candidate jump grid value calculation module: It is used to determine the allowable jump grid value range at the current moment based on the relative orientation relationship between the priority inspection grid and the current position of the UAV, the UAV's maneuverability parameters, and real-time status data, and to determine the candidate jump grid value within the jump grid value range in combination with the priority inspection grid. Heading analysis module: used to calculate the UAV's heading angle at the next moment based on the current heading angle, control turning command and candidate jump value, and to normalize the heading angle at the next moment; The motion state generation module is used to map the normalized heading angle of the next moment to the unit circle, calculate the expected displacement vector of the UAV, and determine the coordinate increment of the next moment based on the expected displacement vector, thereby obtaining the predicted motion trajectory of the UAV. Motion Update Module: This module uses the predicted motion trajectory to look up the task grid reward table, accumulates the grid rewards covered by the predicted motion trajectory, and verifies the candidate jump grid values ​​in conjunction with maneuverability constraints, collision constraints, and task conflict constraints. It selects the optimal jump grid value that maximizes the accumulated reward from the candidate jump grid values ​​that pass the verification, and uses the optimal jump grid value to update the UAV's motion state at the next moment.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This application addresses the problems of ambiguous task priorities and the easy omission of key targets in existing methods by constructing a task grid benefit table and dynamically selecting priority inspection grids based on multiple factors such as benefit value, reachability, and directional deviation. This enables proactive guidance and intelligent prioritization of high-value inspection targets. Secondly, by introducing a jump-grid value mechanism and combining it with multi-dimensional verification based on maneuverability constraints, environmental constraints, and task conflict constraints, the UAV's motion planning can dynamically adapt to the performance differences of heterogeneous platforms, while effectively avoiding no-fly zones, obstacles, and spatiotemporal conflicts with other UAVs, significantly improving flight safety and path flexibility in complex environments. Thirdly, by calculating the cumulative benefit of predicted motion trajectories and introducing a dwell time decay factor, a repeated inspection penalty factor, and a priority grid proximity reward, fine-grained optimization of inspection efficiency and resource utilization is achieved, effectively suppressing ineffective repeated inspection behavior. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 This is a graph showing the relationship between candidate jump values ​​and the corresponding UAV's heading angle at the next moment. Figure 3 This is a schematic diagram of the overall system of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0019] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0020] Example: Please see Figures 1-2 The present invention provides a technical solution: A distributed collaborative power line inspection method for heterogeneous UAV swarms in complex dynamic environments includes the following steps: S1: Acquire the boundary data of the inspection area, the data of the inspection target, and the real-time status data of each UAV. Divide the task grid based on the boundary data of the inspection area and the current position of the UAV, and generate a corresponding inspection benefit value for each task grid to build a task grid benefit table. Based on the task grid benefit table, filter the target grid from the grids that the UAV can reach, and determine the target grid with the highest benefit value as the current priority inspection grid. In the above process, the outer rectangular range of the entire area to be inspected is first determined based on the boundary data of the inspection area. This rectangular range is then defined according to the preset grid side length. The area is divided evenly to form a fixed grid coordinate system covering the entire inspection area. Each grid is uniquely identified by its row index and column index.

[0021] Generate corresponding inspection reward values ​​for each task grid, including: Acquire the target type, defect level, historical anomaly records, and most recent inspection time of the corresponding inspection target within the task grid; acquire obstacle distribution information, meteorological risk information, and accessibility information for the area corresponding to the task grid. Based on target type, defect level, historical anomaly records, most recent inspection time, obstacle distribution information, meteorological risk information, and accessibility information, the benefits of each task grid are scored to form a task grid benefit table.

[0022] The benefit score includes an objective importance score, a regional accessibility score, and an environmental risk correction score, among which: The target importance score is used to characterize the inspection priority of the inspection target within the task grid, the regional accessibility score is used to characterize the ease with which the UAV can reach the corresponding task grid, and the environmental risk correction score is used to characterize the degree of influence of the external environment on the execution of the inspection task. The inspection benefit value is determined based on the weighted result of the target importance score, regional accessibility score and environmental risk correction score.

[0023] In the above process, a corresponding inspection benefit value is generated for each task grid. First, the target type, defect level, historical anomaly records, and most recent inspection time of the corresponding inspection target within the task grid are obtained. Simultaneously, obstacle distribution information, meteorological risk information, and accessibility information for the corresponding area of ​​the task grid are acquired. Based on this, a benefit score is calculated for each task grid, with the score determined by the target importance. Regional accessibility score and environmental risk correction score It consists of three weighted parts, and the specific calculation formula is as follows: ,in , , The first, second, and third weighting coefficients are respectively, and they are full. ; Target Importance Score The inspection priority of targets within the task grid is determined by a combination of target type, defect level, historical anomaly records, and the time since the last inspection. Grids with higher defect levels, more frequent historical anomalies, and longer time since the last inspection receive a higher score. Area accessibility score. The score is used to characterize the ease with which a drone can reach a corresponding task grid. It is a comprehensive assessment based on obstacle distribution information and accessibility information. Grids with lower obstacle density and better accessibility receive higher scores; environmental risk correction score. It is used to characterize the degree of impact of the external environment on the execution of inspection tasks. It is determined by meteorological risk information. The higher the meteorological risk, the greater the correction score. It is also used as a deduction item in the benefit calculation to reduce the inspection priority of high-risk grids.

[0024] Filter the target grid from the currently reachable grid of the drone, including: When determining the reachable grid set at the current moment, the criteria are based on the drone's current remaining battery power, current speed, maximum flight speed, maximum turning angular velocity, and the distance between the current position and each task grid: First, the maximum allowable range is calculated based on the remaining battery power. Then, the maximum displacement constraint is obtained by combining the product of the maximum flight speed and the decision cycle. Only when the distance from the task grid to the UAV's current position is less than or equal to the maximum flight speed multiplied by the decision cycle and the remaining battery power is the grid included in the reachable grid set. Grids with inspection reward values ​​higher than a preset reward threshold are extracted from the reachable grid set as candidate target grids. When the number of candidate target grids is greater than one, the priority inspection grid is determined based on a comprehensive ranking result of inspection reward value, directional deviation from the current heading, and arrival cost. Specifically: For each task grid in the reachable grid set, the comprehensive score of each task grid is obtained by subtracting the direction deviation penalty factor from the inspection benefit value of the task grid, multiplying it by the absolute value of the difference between the grid direction angle of the task grid and the current heading angle of the UAV, and then dividing by pi. The task grid with the highest comprehensive score is selected as the priority inspection grid.

[0025] Based on the drone's current remaining battery power, current speed, maximum flight speed, maximum turning angular velocity, and the distance between the current position and each task grid, determine the set of reachable grids at the current moment; extract candidate target grids whose inspection reward value is higher than the preset reward threshold from the set of reachable grids.

[0026] The formula upon which the above process is based is: in, Indicates the current location of the drone. To the task grid The distance; Indicates the maximum flight speed; Indicates the decision-making cycle; Indicates the remaining battery power; Represents the task grid The inspection revenue value; Represents the task grid The grid direction angle; Indicates the current heading angle of the drone; This represents the penalty factor for directional deviation; Indicates the priority inspection grid; This represents the set of grids that can be reached by drones; This represents the task grid.

[0027] In the above process, the preset revenue threshold is determined as follows: first, the current reachable grid set is statistically analyzed. Inspection reward value for all task grids Calculate its average value with standard deviation Then set the preset revenue threshold to ,in This is an adjustable sensitivity coefficient, typically ranging from 0.5 to 1.0. This threshold design dynamically and adaptively adjusts based on the reward distribution of currently reachable grids, avoiding the problems of using a fixed threshold that misses high-value grids when overall rewards are high or fails to filter out any candidate target grids when overall rewards are low. When the inspection reward value of a certain task grid... Greater than or equal to the preset revenue threshold When the target grid is selected, it is extracted as a candidate target grid and enters the subsequent comprehensive sorting stage; otherwise, if the profit value of all grids is lower than the threshold, the system will automatically select the grid with the highest profit as the only candidate target grid, ensuring that the selection of priority inspection grids has effective output under any circumstances.

[0028] The maximum permissible range is estimated based on the current remaining battery power. ,in This is the current remaining battery level. The energy consumption per unit distance is used as a global energy constraint; secondly, the current speed is considered. Maximum flight speed and decision cycle Calculate the maximum displacement constraint within a unit decision period. And considering the maximum steering angular velocity The resulting feasibility of a shift; then, for each task grid Calculate its position relative to the drone's current location. Euclidean distance between If the distance simultaneously satisfies That is, not exceeding the maximum displacement in a single step and If the range does not exceed the maximum range allowed by the remaining battery power, then the task grid is included in the reachable grid set. Furthermore, for situations where the distance is close to the maximum turning boundary, it is also necessary to verify whether the UAV can complete the orientation adjustment to face the grid within a single step, based on the maximum turning angular velocity. If the turning requirement exceeds this limit... If so, it is also considered unreachable; Wherein, formula This calculation method is used to balance the inspection benefit value of the task grid with the consistency of the UAV's current flight direction during the priority inspection grid selection process, avoiding excessive turning maneuvers or even violations of maneuverability constraints due to blindly pursuing high-benefit grids; among other things, Directly reflects the task grid The inspection benefit value serves as a positive incentive to prioritize high-value grids; and The term represents the grid orientation angle of the task grid. relative to the current heading angle of the drone The normalized directional deviation between them is multiplied by the directional deviation penalty factor. This is then added as a subtraction term to the formula to penalize the final score of the grid. Through this structure of positive reward minus deviation, the final score tends to select grids with higher inspection reward values ​​and smaller directional deviations from the candidate grids, thus achieving a balance between high-value objectives and low turning costs at the decision-making level.

[0029] S2: Based on the relative orientation between the priority inspection grid and the current position of the UAV, the UAV's maneuverability parameters, and real-time status data, determine the allowable jump grid value range at the current moment, and within the jump grid value range, combine with the priority inspection grid to determine candidate jump grid values; Determine the range of allowed jump values ​​at the current moment, including: The maximum displacement constraint and maximum steering constraint within a unit decision cycle are determined based on the maximum flight speed and maximum turning angular velocity of the UAV, and the minimum jump value and maximum jump value at the current moment are determined based on the maximum displacement constraint and maximum steering constraint. The jump value is an integer value representing the number of task grids the UAV traverses when moving from the current task grid to the next task grid within a decision cycle. It represents the step size the UAV moves within a unit decision cycle, with the side length of the task grid as the basic unit of measurement. The minimum jump value is obtained by multiplying the UAV's minimum flight speed by the decision cycle and then dividing by the side length of the task grid. The maximum jump value is the smaller of the maximum flight speed multiplied by the decision cycle divided by the side length of the task grid and the maximum allowable range divided by the side length of the task grid. A jump value set is generated between the minimum and maximum jump values. Each jump value in this set is an integer and must also satisfy an angle deviation constraint: the absolute value of the angle deviation between the desired direction of the priority inspection grid relative to the current heading and the reachable direction corresponding to the jump value must be less than a preset angle deviation threshold.

[0030] The formula upon which the above process is based is: in, This indicates the desired direction of the priority inspection grid relative to the current course; Indicates the value of the jump. The corresponding reachable direction; Indicates the angular deviation threshold; Indicates the minimum number of ticks; Indicates the maximum number of ticks; Indicates the side length of the task grid; This indicates the maximum flight speed of the drone; Indicates the minimum flight speed of the drone; Indicates the decision-making cycle; Indicates the maximum permissible flight range of the drone; Indicates the number of tabs; Represents the set of tab values; Indicates a range of integers.

[0031] In the above process, the set of jump values This calculation method is used to discretize the continuous motion space of the UAV into a finite number of integer jump value candidates, and to establish a quantifiable mapping relationship between maneuverability constraints and directional guidance. Among these, the minimum jump value... The minimum flight speed of the drone and decision cycle The decision ensures that the drone moves at least a controllable minimum distance within a single decision cycle, avoiding trajectory oscillations or invalid decisions due to excessively small step sizes; maximum jump value. Take maximum displacement constraint With maximum permissible range constraints The smaller of the two values, the former being determined by the maximum flight speed. and decision cycle The decision is that the latter is based on the maximum permissible range calculated from the current remaining battery power. This decision allows for the simultaneous satisfaction of both single-step maneuverability and global energy constraints. Furthermore, an angle deviation constraint is introduced into the formula. ,in Prioritize inspecting the grid in the desired direction relative to the current heading. value of the jump The corresponding reachable direction is determined by a preset angle deviation threshold. Filter out skip grid values ​​with excessive directional deviation to ensure that candidate skip grid values ​​tend to point to the priority inspection grid in terms of direction.

[0032] Generate a set of candidate jump values ​​within the jump value range, including: The azimuth angle of the priority inspection grid relative to the current position of the UAV is used as the target guidance direction; the directional deviation between the theoretical arrival direction and the target guidance direction corresponding to each jump grid value is calculated; jump grid values ​​with directional deviation less than a preset angle threshold are retained as candidate jump grid values ​​and sorted in ascending order of directional deviation to obtain a set of candidate jump grid values.

[0033] In the above process, a preset angle threshold is used. The determination method is based on the maximum turning angular velocity of the UAV. and decision cycle Calculate the maximum steering capability within a unit decision cycle. Then set the preset angle threshold to ,in This is an adjustable leniency factor, typically ranging from 1.0 to 2.0. This threshold design directly links the angular deviation tolerance range to the actual steering maneuverability of the UAV: ​​when At this time, only jump values ​​with directional deviations not exceeding the maximum turning angle of the UAV in a single step are allowed to enter the candidate set, ensuring that the candidate jump values ​​can guide the direction by turning to approach the target within a single step; when At the same time, a certain degree of overshoot is allowed, meaning that the UAV may need to gradually correct its direction over multiple decision cycles to reach the target grid, thereby achieving a balance between directional accuracy and decision-making flexibility.

[0034] S3: Based on the current heading angle, control steering command, and candidate jump values, calculate the heading angle of the UAV at the next moment, and normalize the heading angle at the next moment. S4: Map the normalized heading angle of the next moment to the unit circle, calculate the expected displacement vector of the UAV, and determine the coordinate increment of the next moment based on the expected displacement vector, thereby obtaining the predicted motion trajectory of the UAV. Generate a predicted motion trajectory for each candidate jump value in the candidate jump value set, including: The heading angle for the next moment is calculated based on the current heading angle, control steering command, and corresponding candidate jump values. The specific calculation method is as follows: using the current heading angle of the UAV as a baseline, add the current control steering command multiplied by the UAV's maximum steering angle, and then multiply by the ratio of the candidate jump value to the maximum jump value. The control steering command ranges from -1 to +1, representing the intensity of a left or right turn. Subsequently, the heading angle is normalized by calculating the sine and cosine values ​​of the normalized heading angle, and then performing an arctangent operation on these two values ​​to limit the heading angle for the next moment within a preset angle range, thus obtaining the normalized heading angle corresponding to the candidate jump value for the UAV at the next moment. The normalized heading angle for the next time step is subjected to sine and cosine mapping to obtain the displacement direction component under the corresponding candidate jump grid value. Specifically, the cosine and sine values ​​of the heading angle corresponding to the candidate jump grid value of the UAV at the next time step are taken to form a two-dimensional column vector with a magnitude of 1, representing the displacement direction component of the UAV on the unit circle. At the same time, the displacement length component is obtained by combining the candidate jump grid value, that is, the candidate jump grid value multiplied by the side length of the task grid, representing the physical distance moved by the UAV within the decision cycle. Then, the coordinate increment of the next time step is determined based on the displacement direction component and the displacement length component, that is, the scalar of the displacement length component is multiplied by the two-dimensional vector of the displacement direction component to obtain the coordinate increment of the UAV's candidate jump grid value at the next time step. Further, the predicted motion trajectory corresponding to the candidate jump value is generated. Specifically, the coordinates of the current candidate jump value of the UAV are used as the reference, and the coordinate increment of the candidate jump value at the next moment is added to obtain the coordinates of the candidate jump value of the UAV at the next moment, thus completing the position state update. At the same time, the heading angle corresponding to the normalized candidate jump value of the UAV at the next moment is directly assigned as the heading angle of the candidate jump value of the UAV at the next moment, thus completing the heading state update.

[0035] The formula used in the above process is: in, Indicates the current heading angle of the drone; Indicates the current control steering command; Indicates the heading angle of the drone at the next moment; Indicates the maximum turning angle of the drone; Indicates the candidate tab value; Indicates the maximum number of ticks; In the above process, the dependent variable This represents the heading angle of the drone at the next moment. Its technical effect is to quantitatively connect the drone's current flight direction with the change in direction after the decision, providing a heading benchmark for subsequent trajectory prediction and benefit assessment. The independent variable in the formula includes the drone's current heading angle. Current control steering command The range of values ​​is The maximum turning angle of the drone and candidate jump values With the maximum jump value ratio ;in, It provides a starting point for direction as a basic heading. It reflects the driver's or higher-level controller's instructions regarding steering direction and steering intensity, while Then based on the candidate jump value Relative to the maximum jump value The proportional dynamic adjustment of the actual applied steering angle amplitude makes the steering angle larger when the step value is larger, thus achieving a reasonable coupling of "larger step size corresponds to stronger steering". when When, it means the drone will maintain its current heading angle and will not perform any turning; when When the value is 1, it indicates that the drone is making a right turn. The larger the value, the closer it is to 1, the stronger the turn, meaning the closer the desired turn angle is to the maximum turn angle. ;when When the value is -1, it indicates that the drone is making a left turn, and the larger the absolute value, the closer it is to the desired left turn angle. ; Regarding positive and negative correlations, and , , as well as Both show a positive correlation: current heading angle The larger the value, the greater the heading angle at the next moment; control steering commands. A positive value indicates an increase in the heading angle when turning right, while a negative value indicates a decrease in the heading angle when turning left; maximum turning angle. Larger or larger jump value ratio The larger the value, the greater the steering offset, thus ensuring that the candidate step size can be effectively mapped to different steering amplitudes, which facilitates the subsequent selection of the optimal step size.

[0036] In the above embodiments, 10 sets of data are given for candidate jump values ​​and the corresponding heading angle of the UAV at the next moment, so that the heading angle of the UAV at the next moment changes with the candidate jump values, as shown in Table 1: Table 1: Relationship between candidate jump values ​​and the corresponding UAV heading angle at the next moment. In Table 1 above, given , , , In the case of candidate jump values The larger the value, the greater the heading angle of the drone in the next moment. The larger the value, the stronger the positive correlation. When At that time, the drone maintains its current heading angle; when At that moment, the drone performs its maximum turning intensity, and the heading angle reaches 63° at the next moment.

[0037] The normalization process is completed by limiting the heading angle of the next moment to a preset angle range. in, This represents the normalized candidate hop value for the next time step of the drone. The corresponding heading angle; Represents the candidate hop value for the next moment of the drone. The value of the sine direction of the corresponding heading angle; Indicates the candidate jump value for the next moment of the drone. The value of the cosine direction of the corresponding heading angle; In the above process, because the heading angle is mathematically periodic, for example... and Indicating the same direction, in a series of consecutive decision-making processes, It may exceed Scope or generation and In cases of discontinuity, such as abrupt changes, directly using unnormalized angles for subsequent calculations can lead to incorrect mapping of displacement direction components. This can be addressed by calculating separately... The value in the sine direction The value of the cosine direction These two trigonometric function values ​​naturally retain the complete directional information of the original angle and are not constrained by the angle range. Then, using the arctangent function... Recover the unique and strictly fall into or Normalized candidate jump values ​​for the UAV within a preset angle range for the next moment Corresponding heading angle .

[0038] The normalized heading angle for the next time step is subjected to sine and cosine mapping to obtain the displacement direction component under the corresponding candidate jump value; the displacement length component is obtained by combining the corresponding candidate jump value, and the coordinate increment for the next time step is determined based on the displacement direction component and the displacement length component. in, Indicates candidate jump value The corresponding displacement length component; Indicates the side length of the task grid; This represents the displacement component of the normalized heading angle in the unit circle. Indicates the candidate tab value; The value of the cosine of the heading angle of the UAV at the next moment after normalization; This represents the sine value of the heading angle of the UAV at the next moment after normalization. Indicates the candidate hop value of the drone The coordinate increment at the next moment; Indicates the candidate hop value of the drone The heading angle at the next moment; In the above process, discrete candidate jump values ​​are... This is converted into two-dimensional coordinate increments of the UAV in the actual physical space, thereby quantifying the motion trajectory corresponding to different jump values. Among these, the displacement length component... Discrete candidate jump values Side length of the task grid Multiplication ensures that the jump value directly corresponds to the accumulated physical distance of the grids traversed by the drone, achieving a direct proportional mapping that "the larger the jump value, the longer the travel distance." Displacement direction component. Candidate hop values ​​for the next time step of the normalized UAV Corresponding heading angle By taking the values ​​in the cosine and sine directions respectively, a two-dimensional direction vector is constructed on the unit circle, ensuring that the displacement direction is strictly consistent with the normalized heading angle and has a magnitude of 1. The candidate jump values ​​are obtained by multiplying the displacement length component (scalar) with the displacement direction component (unit vector). coordinate increment at the next moment It also includes two independent dimensions: movement distance and movement direction, which allows different candidate jump values ​​to generate predicted movement trajectories that differ in both step length and orientation.

[0039] Further generate the predicted motion trajectory corresponding to the candidate jump values: in, The coordinates representing the current candidate jump value of the drone; The coordinates representing the candidate jump values ​​for the next moment of the drone; Indicates the candidate hop value of the drone The heading angle at the next moment; This represents the normalized candidate hop value for the next time step of the drone. The corresponding heading angle; In the above process, the candidate jump values ​​obtained in the previous step are... Corresponding coordinate increment Current position coordinates of the drone Vector superposition is performed to obtain the predicted position of the UAV after executing the given jump value. The coordinates of the current candidate jump value are then used. This indicates the spatial location of the drone at the start of the current decision-making cycle, serving as a reference point for trajectory estimation; candidate jump values. Corresponding coordinate increment This represents the expected displacement change of the drone from the current moment to the next moment, including the distance traveled determined by the jump value and the directional information determined by the normalized heading angle. By directly adding the two, the formula completes the recursive mapping from the current state to the future state in a concise vector addition form, so that each candidate jump value can generate the endpoint position of a complete predicted motion trajectory.

[0040] S5: Use the predicted motion trajectory to look up the task grid revenue table, accumulate the revenue of the grid covered by the predicted motion trajectory, and verify the candidate jump grid values ​​in combination with the maneuverability constraint, collision constraint and task conflict constraint; select the optimal jump grid value that maximizes the accumulated revenue from the candidate jump grid values ​​that pass the verification, and use the optimal jump grid value to update the UAV's motion state at the next moment.

[0041] Identify the task grids that each predicted motion trajectory passes through in sequence, extract the inspection reward value of each passed task grid in the task grid reward table, and accumulate the corresponding inspection reward values ​​according to the order in which the predicted motion trajectory passes through each task grid. Then, correct the accumulation result according to the predicted dwell time of the UAV in each task grid, the proximity to the priority inspection grid, and the repeated inspection penalty factor to obtain the cumulative reward value of the corresponding candidate jump grid value. The specific correction method is as follows: For each task grid that the predicted trajectory passes through, its original inspection reward value is multiplied by an exponential function with the natural constant as the base and a negative dwell time decay coefficient multiplied by the predicted dwell time in that grid as the exponent, and then multiplied by an exponential function with the natural constant as the base and a negative repetition penalty coefficient multiplied by the number of times that grid has been inspected before the current decision cycle, to obtain the corrected reward contribution of that grid; the corrected reward contributions of all grids that have passed through are accumulated in the order of passing through, and an additional priority inspection grid proximity reward item is added. This reward item takes a value of one when the task grid passed through is exactly equal to the priority inspection grid, and a value of zero otherwise.

[0042] The formula upon which the above process is based is: in, This represents the cumulative gain value of the predicted motion trajectory of the UAV corresponding to the corrected candidate jump grid value; Indicates the number of times the drone passed. The original inspection revenue value of each task grid; This indicates the task grid index through which the predicted trajectory of the drone will pass; This indicates the total number of task grids traversed by the predicted trajectory of the drone; Indicates that the drone is in Time spent in each task grid; Indicates the dwell time decay coefficient; Indicates the first The number of times each task grid has been inspected before the current decision cycle; Indicates the repeated penalty coefficient; Indicates when When the condition is met, the value is 1; otherwise, the value is 0. Indicates the priority inspection grid; This indicates the number of times the predicted trajectory of the drone passes through. Task grid.

[0043] The formula for cumulative profit value in the above process The reason for this calculation is to dynamically weight the revenue based on multiple correction factors after accumulating the original inspection revenue values. This ensures that the cumulative revenue value more accurately reflects the actual inspection benefits of the UAV executing the predicted motion trajectory. Specifically, the formula first identifies the task grids that each predicted motion trajectory passes through sequentially. Extract the inspection reward value of each task grid in the task grid reward table. The base payout is calculated by accumulating the payouts according to the order in which they were received. Based on this, a dwell time decay factor is introduced. ,in Indicates that the drone is in Predicted dwell time in each task grid This factor represents the dwell time decay coefficient, which means that the longer a UAV stays in a grid, the smaller the benefit contribution of that grid, thus suppressing ineffective hovering or low-speed loitering behavior; a repeated inspection penalty factor is introduced. ,in Indicates the first The number of times each task grid has been inspected before the current decision cycle. The factor used to represent the repetition penalty coefficient reduces the benefit contribution of grids that are inspected more frequently, thereby encouraging drones to explore uninspected or infrequently inspected areas and avoiding multiple drones repeatedly inspecting the same low-value grids. Finally, a priority grid proximity reward is introduced. When the predicted motion trajectory passes through the priority inspection grid Additional rewards are given to guide drones to approach or reach high-value targets as quickly as possible.

[0044] The validity of each candidate jump value is validated, and the optimal jump value is selected, including: When the predicted motion trajectory corresponding to the candidate jump grid value does not exceed the maneuverability boundary of the UAV, the predicted motion trajectory does not enter any no-fly zone or obstacle-occupied area, and the minimum distance between the predicted motion trajectory of other UAVs is greater than the preset safety threshold, and the target task grid pointed to by the candidate jump grid value has not been preferentially occupied by other UAVs in the cluster during this decision cycle, the candidate jump grid value is considered to have passed the verification. At the same time, candidate jump grid values ​​that fail the verification are removed, and the candidate jump grid value with the largest cumulative benefit value is selected from the remaining candidate jump grid values ​​as the optimal jump grid value. The verification and judgment logic described above is designed to simultaneously ensure the safety, feasibility, and task coordination of drones in distributed collaborative decision-making, avoiding the selection of candidate hop grid values ​​that appear to offer the highest benefit but are actually unexecutable or harmful to the cluster. Specifically, determining whether the predicted trajectory exceeds the drone's maneuverability boundaries ensures the selected trajectory remains within the drone's physical capabilities, such as maximum speed, minimum speed, and maximum turning angular velocity, preventing the planning of unexecutable trajectories. Determining whether the predicted trajectory enters any no-fly zones or obstacle-occupied areas avoids static hazards in the environment, preventing accidents caused by drones colliding with no-fly zones or obstacles. Determining whether the minimum distance between the predicted trajectories of other drones exceeds a preset safety threshold prevents multiple drones from intersecting trajectories or being too close in space and time, effectively preventing collision risks. Determining whether the target task grid pointed to by the candidate hop grid value has not yet been prioritized for use by other drones in the cluster during this decision-making cycle prevents multiple drones from simultaneously targeting the same task grid, leading to task conflicts and resource waste, achieving task decoupling and load balancing at the cluster level.

[0045] When multiple candidate jump values ​​with the same cumulative benefit exist, the candidate jump value with the smallest directional deviation and the lowest energy consumption is selected as the optimal jump value. The optimal jump value is used to update the drone's motion state at the next moment. The specific update method is as follows: based on the coordinates of the drone's current candidate jump value, the coordinate increment of the drone's optimal jump value at the next moment is added to obtain the coordinates of the drone's optimal jump value at the next moment, thus completing the position state update; at the same time, the heading angle of the drone's optimal jump value at the next moment is directly assigned as the heading angle of the drone at the next moment after normalization, thus completing the heading state update.

[0046] The formula upon which the above process is based is: in, Indicates the optimal jump value; The coordinates represent the optimal jump value for the drone at the next moment; This represents the optimal jump value for the drone. The coordinate increment at the next moment; This represents the optimal jump value for the drone. The heading angle at the next moment.

[0047] In the above process, when there are multiple candidate jump values ​​with the same cumulative profit value, the candidate jump value with the smallest directional deviation is selected as the optimal jump value. This is because a smaller directional deviation means the predicted trajectory is closer to the desired direction of the priority inspection grid, reducing unnecessary turning maneuvers and heading oscillations. If the directional deviation remains the same, the candidate jump grid value with the lowest energy consumption is further selected. The energy consumption cost is usually positively correlated with the displacement length component; that is, the smaller the jump grid value, the lower the energy consumption, thus saving the UAV's remaining battery power and extending mission endurance. Determining the optimal jump grid value... Afterwards, through Calculate the coordinates of the optimal jump value for the UAV at the next moment, where the coordinates of the current candidate jump value are... With the optimal jump value Corresponding coordinate increment By performing vector superposition, the updated spatial position is obtained; simultaneously, through... The optimal jump value of the normalized drone The heading angle for the next moment is directly assigned as the heading angle for the next moment. This method of synchronously updating position and angle ensures that the UAV's state is completely closed at the end of each decision cycle, providing continuous and consistent initial conditions for grid division, jump value generation and trajectory prediction in the next cycle, thereby realizing the recursive evolution of the UAV's motion state in space and time.

[0048] Please see Figure 3 The present invention also provides a heterogeneous UAV swarm distributed collaborative power inspection system for complex dynamic environments, wherein the inspection is used to perform the above-mentioned inspection method, including: Priority Inspection Module: Used to acquire inspection area boundary data, inspection target data, and real-time status data of each UAV. Based on the inspection area boundary data and the current position of the UAV, it divides the task grid and generates a corresponding inspection benefit value for each task grid, and constructs a task grid benefit table. Based on the task grid benefit table, it selects target grids from the grids that the UAV can reach and determines the target grid with the highest benefit value as the current priority inspection grid. Candidate jump grid value calculation module: It is used to determine the allowable jump grid value range at the current moment based on the relative orientation relationship between the priority inspection grid and the current position of the UAV, the UAV's maneuverability parameters, and real-time status data, and to determine the candidate jump grid value within the jump grid value range in combination with the priority inspection grid. Heading analysis module: used to calculate the UAV's heading angle at the next moment based on the current heading angle, control turning command and candidate jump value, and to normalize the heading angle at the next moment; The motion state generation module is used to map the normalized heading angle of the next moment to the unit circle, calculate the expected displacement vector of the UAV, and determine the coordinate increment of the next moment based on the expected displacement vector, thereby obtaining the predicted motion trajectory of the UAV. Motion Update Module: This module uses the predicted motion trajectory to look up the task grid reward table, accumulates the grid rewards covered by the predicted motion trajectory, and verifies the candidate jump grid values ​​in conjunction with maneuverability constraints, collision constraints, and task conflict constraints. It selects the optimal jump grid value that maximizes the accumulated reward from the candidate jump grid values ​​that pass the verification, and uses the optimal jump grid value to update the UAV's motion state at the next moment.

[0049] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0050] 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; 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.

[0051] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A distributed collaborative power line inspection method for heterogeneous UAV swarms in complex dynamic environments, characterized in that: include: S1: Obtain the inspection area boundary data, inspection target data, and real-time status data of each UAV. Divide the task grid based on the inspection area boundary data and the current position of the UAV, generate the corresponding inspection benefit value for each task grid, and construct the task grid benefit table. Based on the task grid benefit table, target grids are selected from the current drone reachable grids, and the target grid with the highest benefit value is determined as the current priority inspection grid; S2: Based on the relative orientation between the priority inspection grid and the current position of the UAV, the UAV's maneuverability parameters, and real-time status data, determine the allowable jump grid value range at the current moment, and within the jump grid value range, combine with the priority inspection grid to determine candidate jump grid values; S3: Based on the current heading angle, control steering command, and candidate jump values, calculate the heading angle of the UAV at the next moment, and normalize the heading angle at the next moment. S4: Map the normalized heading angle of the next moment to the unit circle, calculate the expected displacement vector of the UAV, and determine the coordinate increment of the next moment based on the expected displacement vector, thereby obtaining the predicted motion trajectory of the UAV. S5: Use the predicted motion trajectory to look up the task grid revenue table, accumulate the revenue of the grid covered by the predicted motion trajectory, and verify the candidate jump grid values ​​in combination with the maneuverability constraint, collision constraint and task conflict constraint; select the optimal jump grid value that maximizes the accumulated revenue from the candidate jump grid values ​​that pass the verification, and use the optimal jump grid value to update the UAV's motion state at the next moment.

2. The heterogeneous UAV swarm distributed collaborative power line inspection method for complex dynamic environments according to claim 1, characterized in that, Generate corresponding inspection reward values ​​for each task grid, including: Acquire the target type, defect level, historical anomaly records, and most recent inspection time of the corresponding inspection target within the task grid; acquire obstacle distribution information, meteorological risk information, and accessibility information for the area corresponding to the task grid. Based on target type, defect level, historical anomaly records, most recent inspection time, obstacle distribution information, meteorological risk information, and accessibility information, the benefit score of each task grid is calculated to form a task grid benefit table. The benefit score includes an objective importance score, a regional accessibility score, and an environmental risk correction score, among which: The target importance score is used to characterize the inspection priority of the inspection target within the task grid, the regional accessibility score is used to characterize the ease with which the UAV can reach the corresponding task grid, and the environmental risk correction score is used to characterize the degree of influence of the external environment on the execution of the inspection task. The inspection benefit value is determined based on the weighted result of the target importance score, regional accessibility score and environmental risk correction score.

3. The heterogeneous UAV swarm distributed collaborative power line inspection method for complex dynamic environments according to claim 2, characterized in that, Filter the target grid from the currently reachable grid of the drone, including: When determining the reachable grid set at the current moment, the criteria are based on the drone's current remaining battery power, current speed, maximum flight speed, maximum turning angular velocity, and the distance between the current position and each task grid: First, the maximum allowable range is calculated based on the remaining battery power. Then, the maximum displacement constraint is obtained by combining the product of the maximum flight speed and the decision cycle. Only when the distance from the task grid to the UAV's current position is less than or equal to the maximum flight speed multiplied by the decision cycle and the remaining battery power is the grid included in the reachable grid set. Grids with inspection reward values ​​higher than a preset reward threshold are extracted from the reachable grid set as candidate target grids. When the number of candidate target grids is greater than one, the priority inspection grid is determined according to the comprehensive ranking result of inspection reward value, directional deviation from the current heading, and arrival cost. Specifically: For each task grid in the reachable grid set, the comprehensive score of each task grid is obtained by subtracting the direction deviation penalty factor from the inspection benefit value of the task grid, multiplying it by the absolute value of the difference between the grid direction angle of the task grid and the current heading angle of the UAV, and then dividing it by pi. The task grid with the highest comprehensive score is selected as the priority inspection grid. Based on the drone's current remaining battery power, current speed, maximum flight speed, maximum turning angular velocity, and the distance between the current position and each task grid, determine the set of reachable grids at the current moment; extract candidate target grids whose inspection reward value is higher than the preset reward threshold from the set of reachable grids.

4. The heterogeneous UAV swarm distributed collaborative power line inspection method for complex dynamic environments according to claim 1, characterized in that, Determine the range of allowed jump values ​​at the current moment, including: The maximum displacement constraint and maximum steering constraint within a unit decision cycle are determined based on the maximum flight speed and maximum turning angular velocity of the UAV, and the minimum jump value and maximum jump value at the current moment are determined based on the maximum displacement constraint and maximum steering constraint. The minimum jump value is obtained by multiplying the UAV's minimum flight speed by the decision cycle and then dividing by the side length of the task grid; the maximum jump value is the smaller of the maximum flight speed multiplied by the decision cycle divided by the side length of the task grid and the maximum allowable range divided by the side length of the task grid; a jump value set is generated between the minimum jump value and the maximum jump value, where each jump value is an integer and must also satisfy the angle deviation constraint: that is, the absolute value of the angle deviation between the desired direction of the priority inspection grid relative to the current heading and the reachable direction corresponding to the jump value must be less than the preset angle deviation threshold.

5. The heterogeneous UAV swarm distributed collaborative power line inspection method for complex dynamic environments according to claim 1, characterized in that, Generate a set of candidate jump values ​​within the jump value range, including: The azimuth angle of the priority inspection grid relative to the current position of the UAV is used as the target guidance direction; the directional deviation between the theoretical arrival direction and the target guidance direction corresponding to each jump grid value is calculated; jump grid values ​​with directional deviation less than a preset angle threshold are retained as candidate jump grid values ​​and sorted in ascending order of directional deviation to obtain a set of candidate jump grid values.

6. The heterogeneous UAV swarm distributed collaborative power line inspection method for complex dynamic environments according to claim 4, characterized in that, Generate a predicted motion trajectory for each candidate jump value in the candidate jump value set, including: The heading angle for the next moment is calculated based on the current heading angle, control steering command, and corresponding candidate jump values. The specific calculation method is as follows: using the current heading angle of the UAV as a baseline, add the current control steering command multiplied by the UAV's maximum steering angle, and then multiply by the ratio of the candidate jump value to the maximum jump value. The control steering command ranges from -1 to +1, representing the intensity of a left or right turn. Subsequently, the heading angle is normalized by calculating the sine and cosine values ​​of the normalized heading angle, and then performing an arctangent operation on these two values ​​to limit the heading angle for the next moment within a preset angle range, thus obtaining the normalized heading angle corresponding to the candidate jump value for the UAV at the next moment. The normalized heading angle for the next time step is subjected to sine and cosine mapping to obtain the displacement direction component under the corresponding candidate jump grid value. Specifically, the cosine and sine values ​​of the heading angle corresponding to the candidate jump grid value of the UAV at the next time step are taken to form a two-dimensional column vector with a magnitude of 1, representing the displacement direction component of the UAV on the unit circle. At the same time, the displacement length component is obtained by combining the candidate jump grid value, that is, the candidate jump grid value multiplied by the side length of the task grid, which represents the physical distance moved by the UAV within the decision cycle. Then, the coordinate increment of the next time step is determined based on the displacement direction component and the displacement length component, that is, the scalar of the displacement length component is multiplied by the two-dimensional vector of the displacement direction component to obtain the coordinate increment of the UAV's candidate jump grid value at the next time step. Further, the predicted motion trajectory corresponding to the candidate jump value is generated. Specifically, the coordinates of the current candidate jump value of the UAV are used as the reference, and the coordinate increment of the candidate jump value at the next moment is added to obtain the coordinates of the candidate jump value of the UAV at the next moment, thus completing the position state update. At the same time, the heading angle corresponding to the normalized candidate jump value of the UAV at the next moment is directly assigned as the heading angle of the candidate jump value of the UAV at the next moment, thus completing the heading state update.

7. The method for distributed collaborative power line inspection of heterogeneous UAV swarms in complex dynamic environments according to claim 1, characterized in that, By using the predicted motion trajectory corresponding to each candidate jump value to look up the task grid reward table, the cumulative reward value corresponding to each predicted motion trajectory is obtained, including: Identify the task grids that each predicted motion trajectory passes through in sequence, extract the inspection reward value of each passed task grid in the task grid reward table, and accumulate the corresponding inspection reward values ​​according to the order in which the predicted motion trajectory passes through each task grid. Then, correct the accumulation result according to the predicted dwell time of the UAV in each task grid, the proximity to the priority inspection grid, and the repeated inspection penalty factor to obtain the cumulative reward value of the corresponding candidate jump grid value. The specific correction method is as follows: For each task grid that the predicted trajectory passes through, its original inspection reward value is multiplied by an exponential function with the natural constant as the base and a negative dwell time decay coefficient multiplied by the predicted dwell time in that grid as the exponent, and then multiplied by an exponential function with the natural constant as the base and a negative repetition penalty coefficient multiplied by the number of times that grid has been inspected before the current decision cycle, to obtain the corrected reward contribution of that grid; the corrected reward contributions of all grids that have passed through are accumulated in the order of passing through, and an additional priority inspection grid proximity reward item is added. This reward item takes a value of one when the task grid passed through is exactly equal to the priority inspection grid, and a value of zero otherwise.

8. The heterogeneous UAV swarm distributed collaborative power line inspection method for complex dynamic environments according to claim 7, characterized in that, The validity of each candidate jump value is validated, and the optimal jump value is selected, including: When the predicted motion trajectory corresponding to the candidate jump grid value does not exceed the maneuverability boundary of the UAV, the predicted motion trajectory does not enter any no-fly zone or obstacle-occupied area, and the minimum distance between the predicted motion trajectory of other UAVs is greater than the preset safety threshold, and the target task grid pointed to by the candidate jump grid value has not been preferentially occupied by other UAVs in the cluster during this decision cycle, the candidate jump grid value is considered to have passed the verification. At the same time, candidate jump grid values ​​that fail the verification are removed, and the candidate jump grid value with the largest cumulative benefit value is selected from the remaining candidate jump grid values ​​as the optimal jump grid value. When multiple candidate jump values ​​with the same cumulative benefit exist, the candidate jump value with the smallest directional deviation and the lowest energy consumption is selected as the optimal jump value. The optimal jump value is used to update the drone's motion state at the next moment. The specific update method is as follows: based on the coordinates of the drone's current candidate jump value, the coordinate increment of the drone's optimal jump value at the next moment is added to obtain the coordinates of the drone's optimal jump value at the next moment, thus completing the position state update; at the same time, the heading angle of the drone's optimal jump value at the next moment is directly assigned as the heading angle of the drone at the next moment after normalization, thus completing the heading state update.

9. A heterogeneous UAV swarm distributed collaborative power inspection system for complex dynamic environments, the inspection system being used to implement the inspection method according to any one of claims 1-8, characterized in that: Priority Inspection Module: Used to acquire inspection area boundary data, inspection target data and real-time status data of each UAV, divide task grids based on inspection area boundary data and current UAV position, generate corresponding inspection benefit value for each task grid, and construct task grid benefit table; Based on the task grid benefit table, target grids are selected from the current drone reachable grids, and the target grid with the highest benefit value is determined as the current priority inspection grid; Candidate jump grid value calculation module: It is used to determine the allowable jump grid value range at the current moment based on the relative orientation relationship between the priority inspection grid and the current position of the UAV, the UAV's maneuverability parameters, and real-time status data, and to determine the candidate jump grid value within the jump grid value range in combination with the priority inspection grid. Heading analysis module: used to calculate the UAV's heading angle at the next moment based on the current heading angle, control turning command and candidate jump value, and to normalize the heading angle at the next moment; The motion state generation module is used to map the normalized heading angle of the next moment to the unit circle, calculate the expected displacement vector of the UAV, and determine the coordinate increment of the next moment based on the expected displacement vector, thereby obtaining the predicted motion trajectory of the UAV. Motion Update Module: This module uses the predicted motion trajectory to look up the task grid reward table, accumulates the grid rewards covered by the predicted motion trajectory, and verifies the candidate jump grid values ​​in conjunction with maneuverability constraints, collision constraints, and task conflict constraints. It selects the optimal jump grid value that maximizes the accumulated reward from the candidate jump grid values ​​that pass the verification, and uses the optimal jump grid value to update the UAV's motion state at the next moment.

Images