Multi-unmanned aerial vehicle cooperative task planning method and system

Abstract

This invention discloses a multi-UAV collaborative task planning method and system, belonging to the field of UAV task planning technology. The method includes: collecting point cloud data of the guide line at the inspection site and constructing a three-dimensional sag model; calculating the sag change in the same coordinate system to form a continuous dancing potential field; based on the dancing potential field, adaptively slicing the guide line, attaching a dancing intensity label to each slice; determining whether each slice requires dual-UAV array observation based on the label; generating baseline length and baseline direction constraints for slices requiring array observation, and retaining single-UAV observation constraints for slices not requiring array observation, thus obtaining a set of attitude constraints; after completing the task slicing and constraint labeling, allocating task slices according to the current position and energy state of the UAVs. This invention effectively solves key bottlenecks in traditional collaborative inspections, such as unbalanced tasks, infeasible flight paths, and inability to guarantee array attitude, thus improving the overall accuracy and operational safety of multi-UAV recognition of dancing guide lines.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) mission planning technology, and in particular to a method and system for collaborative mission planning of multiple UAVs. Background Technology

[0002] With the continuous expansion of the power grid and increasingly close inter-regional interconnection, the galloping of transmission lines under severe weather conditions has become a significant hidden danger threatening power grid safety. When conductors are affected by strong winds, icing, or temperature changes, they undergo significant three-dimensional dynamic deformation, with their spatial positions constantly changing, posing a great challenge to real-time monitoring. In recent years, drone inspections have been increasingly used in the power industry, and technologies such as point cloud reconstruction, 3D fitting, and visual inspection have gradually matured. However, most of these technologies are currently only suitable for static or quasi-static line morphology detection, and are difficult to meet the requirements of spatial continuity, time sensitivity, and differentiated inspection density in conductor galloping scenarios. On the other hand, the inspection tasks of long-span lines are often long-distance, involve large amounts of data, and are subject to strong environmental interference. A single drone is prone to missed inspections, frequent returns, or unstable operation due to excessive load. Therefore, multi-drone collaborative operations are needed to improve coverage and inspection efficiency.

[0003] However, existing multi-drone collaborative inspection technologies still have two significant shortcomings when dealing with conductor galloping: First, current task division methods typically segment by fixed spans or intervals, lacking the ability to dynamically adjust task granularity based on galloping intensity. This results in insufficient sampling density in areas of intense galloping, while generating a large amount of redundant data in stable areas, making it impossible to achieve intelligent matching between galloping behavior and inspection tasks.

[0004] Second, current multi-UAV trajectory planning methods mostly rely on static maps and simple path search strategies. These methods cannot identify key passage points in real time and dynamically update the map based on the actual situation of conductor galloping, nor can they incorporate flight attitude constraints such as formation and shooting angle into the planning process. Therefore, in dynamic scenarios with violent conductor galloping, traditional methods struggle to generate reference routes that both ensure safety and meet observation requirements, making it difficult to effectively carry out high-precision multi-UAV collaborative inspection tasks. Summary of the Invention

[0005] In view of the aforementioned existing problems, the present invention is proposed.

[0006] Therefore, this invention provides a multi-UAV collaborative task planning method and system, which solves the problems in the prior art where the task division method of UAVs in complex line inspection scenarios is singular, it is difficult to flexibly adjust according to the on-site conditions, and the path planning lacks comprehensive consideration of dynamic environmental changes and attitude constraints.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a multi-UAV collaborative mission planning method, which includes, Collect point cloud data of the traverse line at the inspection site and construct a three-dimensional sag model. Calculate the sag change in the same coordinate system to form a continuous dynamic potential field. Based on the dancing potential field, the line is adaptively sliced, and each slice is attached with a dancing intensity label. Based on the label, it is determined whether each slice needs to be observed by a dual-machine array. Baseline length constraints and baseline direction constraints are generated for slices that need arrays, and single-machine observation constraints are retained for slices that do not need arrays, thus obtaining a set of attitude constraints. After completing the task slicing and constraint annotation, the task slices are allocated according to the current position and energy state of the UAV, and a task sequence is generated; Taking the task sequence as input, a reference trajectory is generated under the premise of satisfying attitude constraints. The reference trajectory is then mapped point by point to a 3D sag model. Based on the trajectory projection position, real-time statistics of the 3D sag model are retrieved to dynamically correct the reference trajectory for safe distance. The final execution trajectory is then generated and the UAV flies along this trajectory. After all task slices are completed, the data is stored in the database.

[0008] As a preferred embodiment of the multi-UAV cooperative mission planning method of the present invention, the step of generating a reference trajectory with the mission sequence as input and under the premise of satisfying attitude constraints includes: sequentially reading the currently to-be-executed mission slices from the mission sequence, using the real-time position of the UAV and the three-dimensional coordinates of the center of the slice as the initial input for trajectory planning, automatically identifying spatial nodes with significant changes in accessibility based on the dancing intensity of the area where the slice is located, forming a set of key nodes, and using this set as the basis for local environment planning. Incremental updates yield a mobility map reflecting the latest dancing state. Within this updated mobility map, a local feasible region is extracted between the UAV's current position and the slice center position. The node closest to the task slice center is selected from this feasible region as a short-term navigation point. Using this navigation point as a direction, a set of feasible UAV velocities is established, and local rolling planning is executed to progressively generate a discrete trajectory point sequence leading to the task slice. This discrete trajectory point sequence is input into a minimum snap spline model, while simultaneously retrieving the corresponding attitude constraint set C from the task slice metadata. j During the optimization process, a time-parameterized reference track that satisfies attitude constraints is applied simultaneously.

[0009] As a preferred embodiment of the multi-UAV collaborative task planning method of the present invention, the step of collecting the conductor point cloud data at the inspection site and constructing a three-dimensional sag model, and calculating the sag change to form a continuous dynamic potential field in the same coordinate system includes: at the start of the inspection task, taking the positioning benchmark of the tower as a reference, receiving the conductor point cloud data transmitted back by the UAV, establishing a unified spatial coordinate system with the line direction as the axis, and performing three-dimensional projection of all point cloud data according to the span direction to obtain the discrete spatial distribution of the conductor; The projected point cloud is used as input to perform quadratic curve fitting on each span of the conductor to obtain a continuous sag curve that represents the vertical shape of the conductor. The fitting results of consecutive frames are statistically analyzed within a set time window to calculate the average sag and sag fluctuation at each position. The sag fluctuation and the characteristic quantities reflecting the changes in conductor shape in the fitted curve are normalized and combined to form a continuously changing dynamic potential field along the line.

[0010] As a preferred embodiment of the multi-UAV collaborative task planning method of the present invention, the step of adaptively slicing the route based on the dancing potential field and attaching a dancing intensity label to each slice includes: establishing a task point density distribution according to the strength of the potential field within the entire arc length of the route based on the generated continuous dancing potential field, and performing cumulative distribution calculation from the starting point along the arc length direction. When the cumulative amount reaches an integer value for the first time, the corresponding arc length position is determined as a task center point, thereby forming a center point sequence that automatically changes with the dancing intensity. The interval between adjacent task center points is defined as a task slice. The dancing potential within each task slice is integrated and averaged to obtain a dancing intensity label, and finally, a task slice sequence arranged in the spatial order of the route is obtained.

[0011] As a preferred embodiment of the multi-UAV cooperative mission planning method of the present invention, the step of determining whether each slice needs to be observed using a dual-UAV array based on labels, generating baseline length constraints and baseline direction constraints for slices that require arrays, and retaining single-UAV observation constraints for slices that do not require arrays includes: Read the dance intensity label of the task slice and compare it with the preset dance intensity threshold. th Compare the labels; if the label is greater than or equal to... th Mark the slice as an array observation slice, otherwise mark it as a single-machine observation slice; If it is an array observation slice, the center arc length position of the slice is used as the reference, and adjacent sampling points are obtained on the traverse fitting curve. The tangential unit vector of the traverse is constructed and normalized through finite difference, and the normal unit vector of the traverse is derived by combining the gravity direction of the UAV's real-time IMU. The desired baseline vector is defined between the two UAVs participating in the slice operation, and an array attitude constraint set including baseline length constraint and baseline direction constraint is constructed based on the baseline. If it is a single-machine observation slice, the unit vector of the traverse normal is obtained based on the center position, and a single-machine observation constraint of the range of the angle between the observation optical axis and the traverse normal is applied only to the UAV that performs the slice. Generate a unified set of pose constraints C for each task slice. j .

[0012] As a preferred embodiment of the multi-UAV cooperative task planning method of the present invention, the following steps are performed: after completing the task slicing and constraint labeling, task slices are allocated according to the current position and energy state of the UAVs to generate a task sequence: obtaining the current position and remaining available energy of all online UAVs, extracting the center coordinates of each task slice on the three-dimensional sag model obtained by wire fitting, calculating the spatial distance from each UAV to the center of each slice and estimating the energy required to execute the slice, processing the task slices one by one according to the spatial order of the slices on the line, eliminating UAVs with insufficient energy to complete the flight and return, calculating the allocation cost of "UAV-slice" for the remaining candidate UAVs, selecting the UAV with the lowest cost as the executor, adding the slice and constraints to the end of the UAV's task sequence and updating the UAV's remaining energy synchronously, repeating the selection process until all task slices are allocated, forming the final task sequence of each UAV under energy constraints and spatial order.

[0013] As a preferred embodiment of the multi-UAV cooperative mission planning method of the present invention, the step of mapping the reference track point by point to the three-dimensional sag model means that after obtaining the reference track, each time point t on the reference track is processed sequentially, and the reference trajectory point of time point t is mapped to the three-dimensional sag model corresponding to the traverse.

[0014] As a preferred embodiment of the multi-UAV collaborative mission planning method of the present invention, the step of dynamically correcting the reference trajectory with a safe distance, generating the final execution trajectory, and having the UAV fly according to the trajectory includes: reading the projected arc length corresponding to each reference trajectory point in chronological order, directly querying the sway statistics of the corresponding position in the statistical library of the three-dimensional sag model, constructing a dynamic safe distance at each trajectory moment, reading the traverse normal unit vector at each arc length point, and offsetting the reference trajectory moment by moment in combination with the dynamic safe distance to form the final execution trajectory and use it as the control target sequence, which is then sent to the UAV flight control module moment by moment.

[0015] As a preferred embodiment of the multi-UAV collaborative task planning method of the present invention, the step of storing data in the database after all task slices are completed means that after the observation and execution trajectory tracking of all task slices are completed, the UAV summarizes all the data generated during the inspection process and writes it into the backend database in chronological order.

[0016] Secondly, this invention provides a multi-UAV collaborative mission planning system, comprising, The data acquisition and sag modeling module is used to acquire conductor point cloud data, construct a three-dimensional sag model, and generate a continuous sag potential field. The posture constraint annotation module is used to perform adaptive task slicing based on the dance potential field and generate corresponding dance intensity labels and posture constraint sets for each slice. The multi-UAV task allocation module is used to calculate the allocation cost for task slices based on the current position and energy status of the UAVs and generate the task execution sequence for each UAV. The trajectory planning module is used to generate a reference trajectory under attitude constraints, map it to a 3D sag model, and perform dynamic safety distance correction to form the final execution trajectory. The data storage module is used to archive and store the inspection data after all tasks are completed.

[0017] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the multi-UAV cooperative mission planning method as described in the first aspect of the present invention.

[0018] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the multi-UAV cooperative mission planning method as described in the first aspect of the present invention.

[0019] The beneficial effects of this invention are as follows: By constructing a continuous dancing potential field based on a three-dimensional sag model of a conductor, this invention enables the task slices to adaptively adjust their granularity according to the intensity of the dancing motion. This fundamentally overcomes the limitations of traditional methods, which suffer from static task division and inability to highlight high-risk areas. Furthermore, it introduces a key node identification mechanism based on dancing motion in the trajectory planning stage, and combines it with... By performing real-time incremental updates to the accessibility map, multiple UAVs can continuously obtain navigation information reflecting the latest risk distribution in constantly changing traverse environment, thereby avoiding the problems of long-term path lag and inability to cope with severe traverse disturbances in existing technologies. In the process of generating the reference trajectory, this invention simultaneously applies attitude constraints such as array baseline and optical axis angle, realizing unified planning of trajectory and observation geometry, significantly improving the geometric stability and data consistency of multi-UAV collaborative observation. Through the above mechanism, this invention effectively solves key bottlenecks in traditional collaborative inspection such as unbalanced tasks, infeasible tracks, and inability to guarantee array attitude, and improves the overall recognition accuracy and operational safety of multi-UAVs for dancing traverses. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of the multi-UAV collaborative task planning method in Example 1.

[0022] Figure 2 This is a structural diagram of the multi-UAV collaborative task planning system in Example 1. Detailed Implementation

[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0024] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0025] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0026] Example 1, referring to Figures 1-2 This is the first embodiment of the present invention, which provides a multi-UAV cooperative mission planning method, including the following steps: S1. Collect point cloud data of the guide wire at the inspection site and construct a three-dimensional sag model. Calculate the sag change in the same coordinate system to form a continuous dynamic potential field. Specifically, at the start of the inspection mission, using the high-precision positioning benchmarks of the towers as a reference, the system receives real-time point cloud data of the conductor transmitted by the inspection drone. A line alignment coordinate axis 's' is established based on the direction of the connection between towers, and all point cloud coordinates are transformed to the span direction through three-dimensional geometric projection to obtain projection sample pairs. ,in It is the arc length position of the i-th point in the span direction. This represents the vertical coordinate of the point in the direction of gravity; To transform discrete point cloud data into continuous sag curves that reflect the overall shape of the conductor, projection points are used as fitting data. A quadratic polynomial fitting method is employed to model the vertical shape of each span of the conductor. The continuous sag curves within the span are obtained by least squares solution.

[0027]

[0028] In the formula, y(s,t) is the vertical displacement of the traverse at time t at the span coordinate s. It is the coefficient of the quadratic term, which determines the sag curvature of the fitted curve and reflects the degree of conductor sag. It is a linear coefficient that reflects the conductor tilt trend (caused by span asymmetry or wind deflection). This is a constant term, where N is the number of point cloud samples. These are the true vertical coordinates of the i-th point cloud. It is the span coordinate of the i-th point cloud. Is the fitted model at the position Output value at; The fitted displacement sequence of consecutive frames is collected according to a time window, and the average sag and standard deviation of sag at each position within the time window are calculated point by point:

[0029]

[0030] In the formula, M is the total number of sampling frames within the time window. It is the fitted vertical displacement at position s in the m-th frame. It is the average sag at position s. It is the amplitude of the sag (the intensity of the sag). The standard deviation of the sag and the absolute value of the coefficient of the quadratic term of the fitted curve are normalized to construct the dynamic potential function:

[0031] In the formula, It is the dynamic potential value at position s. It is the weight of the sag ripple, used to adjust the effect of random vibration on the potential field. It is the curvature weight, used to adjust the influence of the conductor sag shape on the potential field. It is the maximum standard deviation of the sag along the entire line. It is the absolute value of the maximum curvature coefficient of the entire line. and The values ​​of w1 and w2 are all in the range of [0,1] and satisfy w1+w2=1. They are determined by multi-index normalization weighting method (such as analytic hierarchy process AHP or empirical weighting method) based on the relative importance of sag fluctuation and curvature in risk assessment.

[0032] The 3D sag modeling method based on projection fitting and time window statistics greatly improves data utilization efficiency, enabling UAVs to reconstruct the 3D morphology of conductors in real time under conditions that rely solely on forward-looking or side-looking perception. By constructing a continuous dancing potential field, it achieves a continuous spatial characterization of the dynamic behavior of conductors, allowing the risk distribution of the inspection area to be quantified, visualized, and used for real-time planning and decision-making. By introducing a multi-index normalization weighting strategy, the dancing intensity assessment is upgraded from the previous single-factor indicators (such as acceleration or single displacement) to a two-dimensional evaluation system that integrates geometric and statistical information, achieving higher robustness and more accurate dynamic response.

[0033] S2. Based on the dancing potential field, the line is adaptively sliced ​​and each slice is attached with a dancing intensity label. Based on the label, it is determined whether each slice needs to be observed by a dual-machine array. Baseline length constraints and baseline direction constraints are generated for slices that need arrays, and single-machine observation constraints are retained for slices that do not need arrays. Specifically, based on the dance dynamics field, the course is adaptively sliced ​​into task segments, and each segment is labeled with a dance intensity tag, including: Based on the generated continuous dynamic potential field, the task point density function is defined in the arc length interval [0,L] of the line:

[0034] In the formula, λ(s) is the task density per unit length corresponding to position s. It is the minimum task density allowed by the system. This is the maximum task density allowed by the system. It is the dance posture of position s. and These are the minimum and maximum values ​​of the dynamic potential field, respectively; Integrate along the arc length of the line starting from the starting point s=0 to calculate the cumulative task quantity:

[0035] In the formula, In position Task point density at the location It is an integral variable, representing the position of the arc length of the line. It is the cumulative task volume from the starting point of the route to position s; Starting from the cumulative task quantity equal to 1, whenever the cumulative task quantity reaches an integer j for the first time, the corresponding position is taken as the j-th task center point, and a sequence of task center points evenly distributed in the cumulative task quantity space is automatically generated. The density of the center points in space corresponds one-to-one with the strength of the dynamic potential: the larger the potential field, the larger λ(s), the faster Λ(s) grows, and the denser the task center points. And the interval between adjacent task center points is defined as a task slice:

[0036] In the formula, It is the spatial range of the j-th task slice. It is the jth The position of the center point of a task on the arc length of the line. It is the position of the j-th task center point; The length is entirely determined by the density integral result. Therefore, the slice interval in the high-galloping region naturally shortens, while the slice interval in the low-galloping region naturally lengthens, thus forming a task space partitioning method that automatically adapts to the dynamic characteristics of the conductor. To enable subsequent steps to set different attitude constraints, imaging arrays, or path planning priorities based on the degree of galloping of different slices, the galloping potential within each task slice is further integrated and averaged to obtain the galloping intensity label.

[0037] In the formula, It is the arc length (in meters) of the task slice. It represents the overall intensity of the dance of the task slice, and ds is the infinitesimal length increment along the arc length of the line. The greater the intensity of the galloping, the more significant the impact of wind-induced vibration on the line segment corresponding to the slice, and the more likely it is to trigger array observation or attitude rigid constraints in the future. The final result is a sequence of task slices arranged in line space order: .

[0038] By introducing an adaptive task slicing mechanism based on the dancing potential field, multi-UAV collaborative task planning achieves real-time response capability to the dynamic behavior of the traverse line for the first time. By mapping the potential field strength to task point density and uniformly sampling it in the cumulative task space, the system can automatically generate denser task slices in high-dancing regions and sparser slices in low-dancing regions, achieving a one-to-one correspondence between the task space and the actual risk distribution. Furthermore, by integrating the potential field within each slice to obtain a dancing intensity label, each slice possesses quantifiable dynamic risk characteristics, providing directly callable physical basis for subsequent array observation judgment, attitude constraint setting, and local trajectory planning. Compared to traditional inspection methods that divide by fixed span or equal intervals, this invention significantly improves the rationality of task allocation and the efficiency of observation resource utilization, achieves adaptive adjustment of task granularity, and enhances the safety, observation accuracy, and overall collaborative planning effect of multi-UAVs in complex wind-induced dancing environments.

[0039] Furthermore, based on the labels, it is determined whether each slice requires dual-machine array observation. Baseline length and baseline direction constraints are generated for slices requiring array observation, while single-machine observation constraints are retained for slices not requiring array observation. Preset dance intensity threshold th If the dance intensity label is greater than or equal to th If the condition is met, the slice is marked as an array observation slice; otherwise, it is marked as a single-machine observation slice. In this embodiment, the dancing intensity threshold th The typical range of values ​​is 0.6 to 0.8 of the normalized value of the dancing potential field. The setting is based on the statistical sensitivity analysis of the location of the sudden change in the dancing trend of the conductor, and can be automatically determined by existing technologies such as normalized gradient detection based on trend change (such as Otsu adaptive threshold method). If it is an array observation slice, calculate the center position of the arc length of the slice:

[0040] In the formula, It represents the starting arc length position of the j-th task slice, indicating the starting position of this slice on the route coordinate axis. It represents the terminating arc length of the j-th task slice, indicating the endpoint position of that slice on the route coordinate axis. It is the center arc length of the j-th task slice, obtained by averaging the start and end points, and is used to represent the spatial center position of the slice; Based on the center position, a fixed small step size is selected on the arc length axis, and two adjacent sampling points are taken on the conductor fitting curve. and ,in, It is the nearest neighbor sampling step size, r wire () is a three-dimensional curve function for the conductor. It returns the coordinates of the conductor at the arc length coordinate s. The original tangential vector at that position is constructed using the finite difference method and then normalized to obtain the conductor's tangential unit vector. :

[0041]

[0042] In the formula, It is the local direction vector of the conductor. These are the coordinates of the sampling point on the right side of the guide curve. These are the coordinates of the sampling point on the left side of the guide curve. It is the magnitude of the original tangential vector; Real-time reading of three-axis acceleration output from the IMU of any drone undergoing inspection. ,in, These are the acceleration components along the x, y, and z axes of the body coordinate system measured by the IMU, which are then normalized to obtain the unit vector of gravity. :

[0043] In the formula, It is the magnitude of the acceleration vector; Constructing a local normal plane for the traverse using the direction of gravity: Subtracting the tangential component of the gravity direction, we obtain a temporary normal vector within the normal plane, which is then normalized to obtain the unit normal vector of the traverse. :

[0044]

[0045] In the formula, It is the unit vector of the direction of gravity. It is the projection of the direction of gravity onto the tangential direction of the conductor. It is a temporary normal vector within the normal plane; For all slices marked "array observation", select two drones participating in the observation and record their current three-dimensional positions. and And construct the array desired baseline vector :

[0046] In the formula, This is the spatial location of the first drone. This is the spatial location of the second drone; Based on the array imaging resolution requirements, two types of attitude constraints are applied to this baseline: one is the baseline length constraint.

[0047] In the formula, It is the distance between the two array units. It is the preset array baseline length, set according to coherent imaging requirements; Second, baseline direction constraints:

[0048] In the formula, It is the unit direction vector of the array baseline. It is the cosine of the expected incident angle of the array. It is the incident angle required for array imaging; In a dual-machine array, the baseline length constraint between the two machines determines the relative spatial configuration of the array, while the baseline direction constraint limits the orientation of this configuration relative to the observed target. Therefore, the two together completely define the attitude state of the array from the two dimensions of "shape" and "orientation", which essentially constitute the array attitude constraint. For slices marked as "single-machine observation," the mandatory configuration of a two-machine baseline is no longer enforced. Instead, an observation line-of-sight constraint is imposed only on the single UAV performing the slice. This means that the angle between the observation optical axis of its onboard sensor and the normal to the traverse line must be within an allowable range, for example, satisfying the following:

[0049] In the formula, It is the unit vector along the optical axis direction observed by a single machine. and These are the minimum and maximum acceptable viewing angles, respectively. Generate a unified set of pose / array constraints C for each task slice. j If the slice requires array observation, then C j Includes two constraints: baseline length constraint and baseline direction constraint. If the slice only requires single-machine observation, then C j Including the constraint of the range of the angle between the observation line and the traverse normal, the final result is sequence.

[0050] By introducing a unified mechanism of "galloping intensity labeling—array / single-machine observation switching—attitude constraint generation" into the mission planning chain, high dynamic adaptability and high geometric consistency, which are unattainable by traditional multi-UAV inspections, are achieved. First, the traverse galloping intensity can be quantified and directly drive the selection of the observation mode, enabling the UAV to automatically adopt a dual-UAV array in high galloping areas and a single-UAV inspection in stable areas, thus ensuring both imaging accuracy and inspection efficiency. Second, through the construction of a local orthogonal reference system using tangential and normal vectors, as well as constraints such as baseline length and baseline direction, the UAV maintains a stable configuration when performing array tasks, avoiding coherent imaging failures caused by attitude inconsistencies in traditional methods. Furthermore, the optical axis angle constraint in the single-UAV observation area further enhances the safety and effectiveness of observations.

[0051] S3. After completing the task slicing and constraint annotation, allocate task slices according to the current position and energy status of the UAV and generate a task sequence. Specifically, the real-time location, remaining battery percentage, and rated battery capacity of each currently online drone are read, and the remaining battery percentage and rated battery capacity are multiplied to obtain the current energy state of the drone; Values ​​are taken from the 3D curve obtained by wire fitting to obtain the coordinates of the 3D center point of the slice. :

[0052] In the formula, These represent the lateral, vertical, and side coordinates of the conductor at the center arc length position, respectively. The three-dimensional curve refers to , where x(s), y(s), and z(s) are the coordinates of the conductor obtained at the arc length position s of the three-dimensional sag model in the horizontal, vertical (usually the direction of gravity), and lateral directions, respectively. Calculate the spatial distance from the current position of each drone to the center of the mission slice. :

[0053]

[0054] In the formula, This is the drone's current location. These are the drone's horizontal, vertical, and lateral coordinates in space; Simultaneously estimate the energy consumption required for slice execution:

[0055] In the formula, This is the total energy required to execute task slice j. It is the energy consumption per unit flight distance of the drone. This refers to the energy consumption per unit distance when a drone performs observations. and All of these results were obtained by calibrating and fitting the changes in battery power in the historical flight records of the UAV with the corresponding flight distance and observation maneuver distance. It is the spatial distance from the drone to the center point of the slice. It is the arc length of the task slice. It is the estimated return distance from the mission area to the base station. This refers to the energy required for the drone to fly from its current location to the mission slice center. It refers to the energy required to complete the observation maneuver within the mission slice. It is reserved for return flight; Each unassigned task slice is processed sequentially in spatial order. When a slice is processed, drones with insufficient energy are removed to ensure that tasks are not assigned to crews that cannot safely arrive and return. For the remaining drones that meet the energy requirements, the allocation cost is calculated for each pair of "drone-slice".

[0056] In the formula, The value of allocating the j-th task slice to the u-th UAV is... It is the spatial distance between the current position of the uth UAV and the center point of the mission slice. It represents the remaining available energy of the uth drone, which is the product of its remaining power percentage and the rated battery capacity. This is the weighting coefficient for the flight distance term, used to adjust the influence of "the distance between the UAV and the mission slice" on the allocation decision. These are the weighting coefficients for the energy consumption item, used to adjust the degree of influence of "how much energy the drone has remaining" on the task allocation result. The values ​​of α and β are usually set to positive weighting coefficients greater than zero, and generally range from one percent to ten percent. Their specific values ​​are determined by common parameter calibration methods such as multiple linear regressions or grid search optimizations on the drone's historical task data. Select the drone with the lowest cost from the set of drones that meet the energy conditions as the executor of the task slice, add the slice and constraints to the end of the drone's task sequence, remove the slice from the set of unassigned tasks, and directly deduct the energy required to complete the task slice from the drone's current remaining available energy value, so that the updated energy value can truly reflect the remaining energy that the drone can still use for subsequent tasks after completing the slice. The process of repeatedly filtering energy, calculating costs, selecting the drone with the lowest cost, and updating the energy state continues until all task slices have been allocated, ultimately forming a task slice execution sequence for each drone that is reasonably arranged according to spatial order and energy constraints.

[0057] By introducing a comprehensive allocation mechanism based on energy state, spatial distance, and task slice characteristics, efficient collaboration among multiple UAVs is achieved. The method uses a realistic 3D traverse model to accurately estimate energy consumption at different stages, enabling task allocation to dynamically adapt to the remaining energy and spatial position of the UAVs. This avoids the risk of "assigning tasks despite insufficient energy" in traditional methods. Simultaneously, a cost function quantifies distance and energy uniformly, allowing task slices to automatically match the most suitable UAV, significantly improving overall operational efficiency. The method continuously updates the energy state during allocation, ensuring that the generated task sequence satisfies both spatial continuity and return-to-home safety, making it particularly suitable for high-frequency UAV switching collaboration in high-gallery sections. Furthermore, task slices are synchronously bound to attitude constraints, providing complete preconditions for subsequent trajectory planning, array observation, and safe distance control. This enables multiple UAVs to execute stably in complex traverse environments, reducing conflicts and improving observation quality, significantly enhancing the overall safety, intelligence, and engineering feasibility of the inspection system.

[0058] S4. Taking the task sequence as input, a reference trajectory is generated under the premise of satisfying attitude constraints. The reference trajectory is then mapped point by point to the three-dimensional sag model. Based on the trajectory projection position, the real-time statistics of the three-dimensional sag model are retrieved to dynamically correct the reference trajectory for safe distance. The final execution trajectory is generated and the UAV flies along this trajectory. After all task slices are completed, the data is stored in the database. Preferably, generating a reference track, using the task sequence as input and satisfying attitude constraints, includes: Read the currently pending slice from the task sequence and combine the slice's center 3D coordinates with the UAV's real-time position as the initial input for trajectory planning; Because conductor galloping causes changes in local traversable space, based on the galloping intensity and local curvature changes in the space where the current task slice is located, key spatial nodes with "significant changes in traversability" are automatically identified around the navigation area, forming a set of key nodes, specifically including: Since the spatial impact of conductor galloping is usually concentrated within a certain range near the slice, based on the arc length s of the slice center in the previous stage of the task... cj Define a symmetric neighborhood interval:

[0059] In the formula, Δs is the center arc length of the j-th task slice, and Δs is the local search bandwidth that extends to both sides, used to limit the range of dancing geometry determination (such as tangential, normal, curvature calculation, etc.). Within this strip region, the sag average μ(s), dancing wave σ(s), and dancing potential Φ(s) obtained in the dancing potential field calculation step are sampled, and the gradient of change is calculated point by point:

[0060]

[0061]

[0062] In the formula, and These are the arc length coordinates of the k-th and (k-1)-th discrete points, respectively. , , These are the gradients of average sag, fluctuation amplitude, and rate of change of situation, respectively. Set corresponding thresholds for the three gradients. , and Filter key arc length points, including: If a point satisfies any condition:

[0063] In the formula, This is the threshold for abrupt changes in the average sag, used to determine whether there is a significant jump in the average sag value. A typical value is 5%–15% of the change in the average sag, set using statistical abrupt change detection based on a sliding window (CUSUM or Shewhart control chart). This is the standard deviation mutation threshold for the sag, used to identify whether the amplitude of the sag fluctuation has increased abnormally. A typical value is the 70% to 90th percentile of the global standard deviation distribution, set using the Levene test based on variance mutation or the moving variance detection method. It is the dynamic potential change threshold, used to determine the location of a sudden increase in the dynamic potential field. The typical value is in the range of 0.6 to 0.8 of the normalized value of the potential field, and is set based on the normalized gradient detection of trend changes or the Otsu adaptive threshold method. It is then assumed that this arc length point may cause changes to the actual safe passage of the UAV, and the three-dimensional coordinates of the arc length point are immediately calculated using a three-dimensional sag model. ,in, These are the positions of the conductor at the arc length. The horizontal, vertical, and lateral coordinates are determined, and spatial grid nodes near the three-dimensional coordinates are added to the key node set; Take the set of key nodes as input and execute Incremental updates yield the latest mobility graph:

[0064] In the formula, g(s) is the known minimum cost estimate from the current node to the target node, and rhs(s) is the minimum estimated cost derived from the neighboring nodes. It's a heuristic function, typically Euclidean distance, used to improve search efficiency. It is the incremental offset, which records the cumulative Key offset when the map changed last time. Key(s) is a tuple used for D* Lite sorting, which determines the node expansion order. Based on the updated local mobility map, and combined with the current UAV location and the center point of the mission slice, a 3D traversable region between two points is extracted from the updated mobility map, forming a local feasible region R. j Specifically, it includes: Calculate the straight line direction between the UAV's current position and the center of the mission slice, and normalize it to obtain the unit vector of the main flight direction:

[0065]

[0066] In the formula, This is the three-dimensional real-time position of the uth UAV in the current planning cycle. These are the coordinates of the 3D center point of the j-th task slice. It is the three-dimensional direction vector pointing from the current point of the UAV to the center point of the task slice. It is the unit direction vector pointing from the current position of the drone to the task slice. It is the Euclidean distance from the drone to the center of the mission slice. It is the unit direction vector of the previous planning cycle; Sampling is performed sequentially from the UAV position to the center point of the task slice along the main direction according to a preset spatial step size, forming center sampling points arranged in chronological order:

[0067]

[0068] In the formula, It is the sampling step size along the main flight direction, and N is the number of samples along the main direction line. These are the three-dimensional coordinates of the nth sampling center point; Connecting the central sampling points together forms a "central skeleton line" pointing from the current position of the UAV to the center of the mission slice, which is the spatial axis for constructing the local feasible domain; Using each central sampling point as the center and the buffer radius calibrated according to the UAV's body size, safe lateral clearance, and attitude maneuver requirements as the radius, find all grid cells in the three-dimensional grid space that are no more than the buffer radius from the central sampling point, and form a local neighborhood set:

[0069]

[0070] In the formula, It is the radius of the neighborhood sphere, that is, the radius of local spatial expansion. Based on the center point Center and radius of the sphere The Grid is a three-dimensional raster map (occupancy map + risk map) maintained internally by the system. It is the set of all grid nodes in the neighborhood of the sphere, x is any point in the three-dimensional continuous space, used to define the geometric range of the neighborhood sphere, and g is a discrete grid node in the three-dimensional grid map Grid, whose coordinates are the grid center point at a fixed resolution, used as a state node for path planning; Each grid cell in the local neighborhood set is matched and filtered with the local accessibility map one by one. Only grid cells that are marked as "accessible", are not located in obstacles or their expansion zones, are not in spatial areas where the galloping risk coefficient exceeds the threshold, and maintain a distance greater than the minimum safe distance from the three-dimensional sag curve of the conductor are retained and formed into a local accessible subset. Perform a union operation on all locally feasible subsets obtained along the entire central skeleton line to obtain the overall 3D locally feasible region corresponding to the current task slice:

[0071]

[0072] In the formula, It is the set of grid nodes that are allowed to pass after filtering. It is the complete three-dimensional local feasible region of the j-th task slice. g is safe and free means that the grid node g is neither occupied by obstacles (free) nor does it meet all safe distance and risk constraints (safe). exist Query all reachable nodes, calculate their distances to the center point of the task slice, and select the node with the smallest distance as the short-term navigation point. :

[0073] Candidate velocity pairs are generated discretely based on the UAV's dynamic upper limit (a velocity set is generated based on the UAV's known physical performance parameters, including maximum horizontal velocity, maximum climb / descent velocity, maximum angular velocity, and maximum acceleration; the UAV's linear velocity and angular velocity must be less than or equal to the maximum horizontal velocity and maximum angular velocity). The motion direction generated by each candidate velocity in the next moment is checked for consistency with the attitude constraints of the mission slice. If a velocity causes the motion direction to deviate from the observation line-of-sight limit of a single UAV or the baseline direction limit of the array, the velocity pair is discarded. The remaining velocity pairs form a feasible velocity set V. The feasible velocity set V of the UAV is established, and the position of each velocity in the feasible velocity set V in the next moment is calculated. :

[0074]

[0075] In the formula, θ(t) is the heading angle of the UAV at the current time t (the deflection angle relative to true north or the x-axis of the global coordinate system), θ(t+Δt) is the heading angle of the UAV at the next discrete time t+Δt, which is the updated orientation direction used in trajectory prediction, w is the angular velocity (rotational speed) in the current candidate velocity pair, representing the change in heading angle per unit time; a positive value usually indicates counterclockwise rotation, and a negative value indicates clockwise rotation, Δt is the discrete time step of the roll prediction, used to approximate the continuous motion of the UAV in that time period, p(t) is the three-dimensional position vector of the UAV at the current time t, p(t+Δt) is the predicted position of the UAV at the next time step, and v is the horizontal linear velocity in the candidate velocity pair. It is the velocity component along the horizontal x-axis, determined by the updated heading angle. It is the velocity component along the horizontal y-axis. It refers to the drone's vertical speed (climbing speed or descent speed). Using short-term navigation points as the direction of advancement and predicted locations as input, local rolling planning is performed:

[0076] In the formula, Φ() is the dynamic potential value of the dance. It is the distance from the point to the nearest obstacle (including wire entities). This is the dynamic potential weighting coefficient, which converts the dimensionless potential field into a "cost length". In this embodiment, its value is set between 0.1 and 10. Its value is obtained through grid search of multiple sets of line-following inspection flight data, based on the cost weight calibration mechanism of the existing DWA dynamic window algorithm. is the obstacle distance weighting coefficient, which is set between 0.1 and 10 in this embodiment. Its value is obtained by obstacle distance-cost regression calibration in real scene based on the existing obstacle distance weighting parameter tuning method of dynamic obstacle avoidance cost function. c(v,w) is the single-step motion cost, which is used to select the optimal speed combination for the next step of the UAV. Select the speed pair with the lowest cost to make the UAV's position continuously approach the short-term navigation point, thereby generating a discrete trajectory point series point by point; Input the discrete trajectory point sequence into the minimum snap optimization model:

[0077] In the formula, r(t) is the three-dimensional position function of the UAV's spatial trajectory. It is the fourth derivative of the trajectory (the derivative of jerk, snap). and It refers to the start and end times of the trajectory. It is a smoothness cost used to generate continuous, flyable, high-order differentiable trajectories; Simultaneously, the corresponding attitude constraint set C is read from the task slice metadata. j Simultaneously apply the following during the optimization process: If it is a single-machine slice → limit the angle between the observation line of sight and the normal of the traverse; If it is a dual-machine slice → limit the baseline length + the included angle of the baseline direction; Obtain a time-parameterized reference trajectory that satisfies the attitude constraints; For each time point of the reference trajectory, the nearest projection is performed to obtain the arc length sequence of the traverse corresponding to the reference trajectory. Specifically, for a given reference trajectory point, the model point with the smallest spatial distance to the reference trajectory point is searched on the 3D sag model of the traverse. By minimizing the Euclidean distance between the reference trajectory point and the sag model point, the corresponding position point of the reference trajectory point on the sag model is determined.

[0078] In the formula, It is the reference trajectory point of the uth UAV at time t. s(t) is the coordinate of the 3D model of the conductor at the arc length position s, and s(t) is the nearest projection arc length of the reference trajectory point on the conductor curve.

[0079] By sampling the average sag, gyration fluctuations, and situation function in a spatiotemporal manner, and combining this with a gradient mutation identification method, this invention can determine in real time the spatial regions that significantly affect the flight path. This ensures that UAV trajectory planning is always based on the latest line morphology and gyration state, thereby significantly improving the planning accuracy and adaptability in dynamic scenarios. Secondly, this invention combines a key node set with... The incremental update strategy only performs local map corrections on sensitive areas of movement, avoiding the latency caused by global recalculation. This enables multiple UAVs to maintain millisecond-level path update efficiency during long-distance inspection missions, achieving rapid autonomous avoidance and dynamic replanning under changes in accessibility.

[0080] At the task execution level, this invention employs a task granularity adaptive mechanism driven by sway intensity labels, enabling multiple UAVs to perform differentiated task allocation based on the sway complexity of different sections of the line, thus improving overall task execution efficiency. Simultaneously, by constraining baseline length and direction based on array requirements, it ensures that dual-UAV observation tasks can achieve high-precision imaging and synchronous inspection under controlled attitude consistency and geometric conditions. Furthermore, the use of locally optimal velocities generated by rolling prediction and dynamic window mechanisms, along with high-order smooth trajectories optimized by minimum snap splines, allows UAVs to maintain continuous, trackable, and safe flight paths even under complex attitude constraints and dynamic environments. Finally, combined with a dynamic safety distance compression mechanism driven by sag standard deviation, the UAVs can automatically approach or move away from the guideline based on real-time sway intensity, not only improving inspection imaging resolution but also significantly enhancing safety redundancy.

[0081] Furthermore, based on the projected position of the flight path, real-time statistics from the 3D sag model are retrieved to dynamically correct the reference flight path with a safe distance, generating the final execution flight path, which the UAV then follows. Read the projected arc length corresponding to each reference trajectory point in chronological order, and directly query the corresponding position's dancing statistics in the statistical library of the 3D sag model, including sag standard deviation, mean sag, and dancing potential. To improve resolution, drones automatically move away from power lines when the galloping intensifies and automatically move closer to the power lines when the galloping intensifies, a dynamic safe distance is established at each trajectory moment. :

[0082] In the formula, It is the minimum lateral distance at which a drone can safely fly under static conditions (no wind or low wind with the conductor swaying). This is the galloping sensitivity coefficient, which in this embodiment is set to a range of 1 to 5. Its value can be determined by calibrating the standard deviation amplification factor (based on the 3σ–6σ safety boundary empirical model) using historical conductor galloping measurement data. It is the standard deviation of the sag of the reference trajectory point at the position s(t) of the projected arc length, representing the amplitude of the vertical random fluctuation of the conductor within the most recent time window; At each arc length point, the unit vector of the traverse normal provided by the sag model is read, and the reference trajectory is offset time-by-time by combining the dynamic safety distance to form the final execution trajectory:

[0083] In the formula, It is the normalized unit normal vector of the conductor's three-dimensional geometry at the arc length position s(t). This is the final trajectory point of the uth drone after safety distance correction; The generated execution trajectory is used as a control target sequence and sent to the UAV flight control module at each moment. The flight control module performs closed-loop control according to the time parameterization of the trajectory points and the target points at each moment, so that the UAV always flies close to the guide wire in a way that meets safety constraints and has high imaging efficiency.

[0084] This invention can update the safety distance in real time and dynamically correct the trajectory based on the conductor galloping, without relying on a preset fixed safety interval. This significantly improves the UAV's adaptability to sudden wind disturbances and severe conductor swaying. Through point-by-point mapping with a three-dimensional sag model, this invention achieves integrated coupling of conductor risk statistics and trajectory planning, avoiding the problem of trajectory being disconnected from the actual geometry of the conductor in traditional methods. By using a normal unit vector for safety offset, this method ensures that the final trajectory maintains a consistent, safe, and predictable offset direction under any galloping conditions. This enables the UAV to robustly avoid high-risk areas and precisely approach low-risk areas, effectively improving inspection accuracy, safety, and mission completion reliability.

[0085] Furthermore, after all task slices are completed, the data is stored in the database. This means that after completing the observation and execution trajectory tracking of all task slices, the UAV will summarize all the data generated during this inspection in chronological order, including the discrete point sequence of the execution trajectory, the corresponding traverse projection arc length, dynamic safety distance, attitude parameters, original sensor images and video streams, and real-time sampled values ​​of the dance statistics. The ground control center will then write this data into the backend database after the mission end signal is triggered. During the data entry process, a multi-level index is established according to the mission number, slice number, and UAV number. At the same time, to ensure data security, a role-based access control model is used to configure different query permissions for maintenance personnel, analysts, and the model training module. Only authorized users are allowed to access time-series data, image data, or statistical results within their corresponding permission scope, and all access logs are recorded, thus forming a complete data closed loop for the inspection mission.

[0086] This embodiment also provides a multi-UAV collaborative mission planning system, including: The data acquisition and sag modeling module is used to acquire conductor point cloud data, construct a three-dimensional sag model, and generate a continuous sag potential field. The posture constraint annotation module is used to perform adaptive task slicing based on the dance potential field and generate corresponding dance intensity labels and posture constraint sets for each slice. The multi-UAV task allocation module is used to calculate the allocation cost for task slices based on the current position and energy status of the UAVs and generate the task execution sequence for each UAV. The trajectory planning module is used to generate a reference trajectory under attitude constraints, map it to a 3D sag model, and perform dynamic safety distance correction to form the final execution trajectory. The data storage module is used to archive and store the inspection data after all tasks are completed.

[0087] This embodiment also provides a computer device applicable to multi-UAV collaborative task planning methods, including: a memory and a processor; the memory stores computer-executable instructions, and the processor executes the computer-executable instructions to implement the multi-UAV collaborative task planning method proposed in the above embodiment. The computer device can be a terminal, and includes a processor, memory, communication interface, display screen, and input device connected via a system bus. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used for wired or wireless communication with external terminals. Wireless communication can be achieved through WIFI, carrier networks, NFC (Near Field Communication), or other technologies. The display screen of the computer device can be a liquid crystal display or an e-ink display. The input device of the computer device can be a touch layer covering the display screen, or buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse, etc.

[0088] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the multi-UAV cooperative mission planning method proposed in the above embodiments. The 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 Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0089] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A multi-UAV collaborative task planning method, characterized in that: include, Collect point cloud data of the traverse line at the inspection site and construct a three-dimensional sag model. Calculate the sag change in the same coordinate system to form a continuous dynamic potential field. Based on the dancing potential field, the line is adaptively sliced, and each slice is attached with a dancing intensity label. Based on the label, it is determined whether each slice needs to be observed by a dual-machine array. Baseline length constraints and baseline direction constraints are generated for slices that need arrays, and single-machine observation constraints are retained for slices that do not need arrays, thus obtaining a set of attitude constraints. After completing the task slicing and constraint annotation, the task slices are allocated according to the current position and energy state of the UAV, and a task sequence is generated; Taking the task sequence as input, a reference trajectory is generated under the premise of satisfying attitude constraints. The reference trajectory is then mapped point by point to a 3D sag model. Based on the trajectory projection position, real-time statistics of the 3D sag model are retrieved to dynamically correct the reference trajectory for safe distance. Finally, the execution trajectory is generated and the UAV flies along the trajectory.

2. The multi-UAV cooperative mission planning method as described in claim 1, characterized in that: The process of generating a reference trajectory, using a task sequence as input and satisfying attitude constraints, includes: sequentially reading the currently executed task slices from the task sequence; using the UAV's real-time position and the center 3D coordinates of the slice as the initial input for trajectory planning; automatically identifying spatial nodes with significant changes in accessibility based on the agitation intensity of the slice's region, forming a set of key nodes; and using this set as a basis for analyzing the local environment. Incremental updates yield a mobility map reflecting the latest dancing state. Within this updated mobility map, a local feasible region is extracted between the UAV's current position and the slice center position. The node closest to the task slice center is selected from this feasible region as a short-term navigation point. Using this navigation point as a direction, a set of feasible UAV velocities is established, and local rolling planning is executed to progressively generate a discrete trajectory point sequence leading to the task slice. This discrete trajectory point sequence is input into a minimum snap spline model, while simultaneously retrieving the corresponding attitude constraint set C from the task slice metadata. j During the optimization process, a time-parameterized reference track that satisfies attitude constraints is applied simultaneously.

3. The multi-UAV cooperative mission planning method as described in claim 2, characterized in that: The process of collecting point cloud data of the conductor at the inspection site and constructing a three-dimensional sag model, and calculating the sag change to form a continuous dynamic potential field under the same coordinate system includes: at the beginning of the inspection task, taking the positioning benchmark of the tower as a reference, receiving the conductor point cloud data transmitted back by the UAV, establishing a unified spatial coordinate system with the line direction as the axis, and performing three-dimensional projection of all point cloud data according to the span direction to obtain the discrete spatial distribution of the conductor. The projected point cloud is used as input to perform quadratic curve fitting on each span of the conductor to obtain a continuous sag curve that represents the vertical shape of the conductor. The fitting results of consecutive frames are statistically analyzed within a set time window to calculate the average sag and sag fluctuation at each position. The sag fluctuation and the characteristic quantities reflecting the changes in conductor shape in the fitted curve are normalized and combined to form a continuously changing dynamic potential field along the line.

4. The multi-UAV cooperative mission planning method as described in claim 3, characterized in that: The process of adaptively slicing the line based on the dynamic potential field and attaching a dynamic intensity label to each slice includes: establishing a task point density distribution according to the strength of the potential field within the entire arc length of the line based on the generated continuous dynamic potential field; performing cumulative distribution calculations from the starting point along the arc length direction; determining the corresponding arc length position as a task center point when the cumulative amount first reaches an integer value; forming a center point sequence that automatically changes with the dynamic intensity; defining the interval between adjacent task center points as a task slice; performing an integral average of the dynamic potential within each task slice to obtain a dynamic intensity label; and finally obtaining a task slice sequence arranged in the spatial order of the line.

5. The multi-UAV cooperative mission planning method as described in claim 4, characterized in that: The step of determining whether each slice requires dual-machine array observation based on labels, generating baseline length and baseline direction constraints for slices requiring array observation, and retaining single-machine observation constraints for slices not requiring array observation includes: Read the dance intensity label of the task slice and compare it with the preset dance intensity threshold. th Compare the labels; if the label is greater than or equal to... th Mark the slice as an array observation slice, otherwise mark it as a single-machine observation slice; If it is an array observation slice, the center arc length position of the slice is used as the reference, and adjacent sampling points are obtained on the traverse fitting curve. The tangential unit vector of the traverse is constructed and normalized through finite difference, and the normal unit vector of the traverse is derived by combining the gravity direction of the UAV's real-time IMU. The desired baseline vector is defined between the two UAVs participating in the slice operation, and an array attitude constraint set including baseline length constraint and baseline direction constraint is constructed based on the baseline. If it is a single-machine observation slice, the unit vector of the traverse normal is obtained based on the center position, and a single-machine observation constraint of the range of the angle between the observation optical axis and the traverse normal is applied only to the UAV that performs the slice. Generate a unified set of pose constraints C for each task slice. j .

6. The multi-UAV cooperative mission planning method as described in claim 5, characterized in that: After completing the task slicing and constraint annotation, the task slices are allocated according to the current position and energy status of the UAVs, and the task sequence is generated by: obtaining the current position and remaining available energy of all online UAVs, extracting the center coordinates of each task slice on the 3D sag model obtained by wire fitting, calculating the spatial distance from each UAV to the center of each slice and estimating the energy required to execute the slice, processing the task slices one by one according to the spatial order of the slices on the line, eliminating UAVs with insufficient energy to complete the flight and return, calculating the allocation cost of "UAV-slice" for the remaining candidate UAVs, and selecting the UAV with the lowest cost as the executor, adding the slice and constraints to the end of the UAV's task sequence and updating the UAV's remaining energy synchronously, repeating the selection process until all task slices are allocated, forming the final task sequence of each UAV under energy constraints and spatial order.

7. The multi-UAV cooperative mission planning method as described in claim 6, characterized in that: The step of mapping the reference track point by point to the three-dimensional sag model refers to processing each time point t on the reference track in sequence after obtaining the reference track, and mapping the reference trajectory point of time point t to the three-dimensional sag model corresponding to the traverse.

8. The multi-UAV cooperative mission planning method as described in claim 7, characterized in that: The process of dynamically correcting the reference trajectory with a safe distance, generating the final execution trajectory, and having the UAV fly according to the trajectory includes: reading the projected arc length corresponding to each reference trajectory point in chronological order, directly querying the galloping statistics of the corresponding position in the statistical library of the three-dimensional sag model, constructing a dynamic safe distance at each trajectory moment, reading the unit vector of the traverse normal at each arc length point, and offsetting the reference trajectory moment by moment in combination with the dynamic safe distance to form the final execution trajectory and use it as the control target sequence, which is then sent to the UAV flight control module moment by moment.

9. The multi-UAV cooperative mission planning method as described in claim 1, characterized in that, Also includes: After completing the observation and execution trajectory tracking of all task slices, the UAV will summarize all the data generated during the inspection process and write it into the backend database in chronological order.

10. A multi-UAV collaborative task planning system, based on the multi-UAV collaborative task planning method according to any one of claims 1 to 9, characterized in that: include, The data acquisition and sag modeling module is used to acquire conductor point cloud data, construct a three-dimensional sag model, and generate a continuous sag potential field. The posture constraint annotation module is used to perform adaptive task slicing based on the dance potential field and generate corresponding dance intensity labels and posture constraint sets for each slice. The multi-UAV task allocation module is used to calculate the allocation cost for task slices based on the current position and energy status of the UAVs and generate the task execution sequence for each UAV. The trajectory planning module is used to generate a reference trajectory under attitude constraints, map it to a 3D sag model, and perform dynamic safety distance correction to form the final execution trajectory. The data storage module is used to archive and store the inspection data after all tasks are completed.

11. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the multi-UAV cooperative mission planning method according to any one of claims 1 to 9.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the multi-UAV cooperative mission planning method according to any one of claims 1 to 9.

Images