A method and system for scheduling multi-drone collaborative unmanned aerial vehicles (UAVs)

Abstract

This invention discloses a multi-drone collaborative UAV scheduling method and system, relating to the field of UAV scheduling technology. The method includes the following steps: acquiring path response data during UAV task switching; identifying execution interference regions using a clustering algorithm; performing feature separation on the interference regions and evaluating scheduling based on the results; reconstructing buffer time segments based on the evaluation to form scheduling windows and extending them to obtain multiple first scheduling windows; within these windows, deducing avoidance areas for UAV path intersections to generate a candidate scheduling path set; inserting tasks using a multi-target matching algorithm to obtain a multi-drone task timing coordination matrix for UAV scheduling. This method effectively solves the execution interference and conflict problems caused by path intersections and timing overlaps due to task switching in multi-UAV collaborative tasks, improving the safety and efficiency of scheduling.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) scheduling technology, and more specifically, to a method and system for scheduling multi-UAV collaboratively. Background Technology

[0002] With the rapid development of drone technology and the continuous expansion of its application fields, the capabilities of a single drone in mission execution are becoming increasingly apparent. However, faced with large-scale, complex, and diverse mission requirements, single-drone operation is no longer sufficient to meet the dual requirements of efficiency and safety. Therefore, multi-drone collaborative scheduling has become an important direction for the development of drone systems.

[0003] Multi-drone collaborative drone scheduling involves key technologies such as path planning, task allocation, time coordination, and conflict avoidance for multiple drones. The goal is to enable multiple drones to conduct efficient, coordinated, and safe joint operations within a limited airspace. This scheduling process must not only consider the flight performance and mission requirements of individual drones but also dynamically respond to factors such as mission switching, changes in the flight environment, communication limitations, and sudden interference.

[0004] However, when multiple drones are performing tasks collaboratively, the intersection of paths and the overlap of timing during task switching can cause execution interference and conflicts between drones. It is difficult to effectively identify and quantify the interference area, which in turn affects the continuity and flexibility of task scheduling and makes it impossible to achieve safe scheduling and path optimization adjustment for inserting new tasks. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a multi-drone collaborative drone scheduling method and system. By identifying interference areas during task switching and dynamically reconstructing the scheduling window, the method addresses the problem of execution interference and conflicts between drones caused by path convergence and timing overlap during task switching when multiple drones are collaboratively executing tasks.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A multi-drone collaborative UAV scheduling method includes the following steps: acquiring path response data during UAV task switching and using a clustering algorithm to identify execution interference regions during the UAV task switching process; performing feature separation on the execution interference regions and evaluating UAV task scheduling based on the feature separation results; reconstructing the buffer time segment in UAV task scheduling based on the evaluation results to obtain a scheduling window, and extending the time range before and after the scheduling window to obtain several first scheduling windows; performing avoidance region deduction on multiple UAV path intersection segments within the several first scheduling windows to obtain a candidate scheduling path set; inserting task matching into the candidate scheduling path set based on a multi-target matching algorithm to obtain a multi-drone task temporal coordination matrix for UAV scheduling.

[0008] In a preferred embodiment, the step of acquiring path response data during the UAV mission switching process and using a clustering algorithm to identify the execution interference region during the UAV mission switching process specifically involves: real-time acquisition of path response data during the UAV mission switching process and construction of a spatiotemporal response matrix, wherein the path response data includes three-dimensional trajectory coordinates, velocity vectors, and heading angle deviations; extraction of acceleration mutation points and heading oscillation intervals from the spatiotemporal response matrix to identify potential interference feature points; using a density peak clustering algorithm to perform spatial density scanning on the potential interference feature points to obtain a cluster core area; and expanding outward from the cluster core area with a preset radius, combined with preset geofencing constraints, to obtain the execution interference region.

[0009] In a preferred embodiment, the feature separation of the execution interference region specifically includes:

[0010] A topology map of the execution interference region is constructed, and the Laplacian matrix eigenvector of the topology map is calculated using a spectral clustering algorithm. Based on the Laplacian matrix eigenvector, the execution interference region is divided to obtain several interference core regions. The convex hull boundaries of several interference core regions are extracted to obtain a dynamic separation interface.

[0011] In a preferred embodiment, the step of scheduling and evaluating UAV missions based on feature separation results specifically involves: constructing a three-dimensional grid cell within the dynamic separation interface and using the Poisson distribution algorithm to calculate the UAV spacing conflict probability between grid cells; using the Markov chain algorithm to predict the heading angle conflict state transition path; and constructing a scheduling evaluation matrix based on the UAV spacing conflict probability.

[0012] In a preferred embodiment, the step of reconstructing the buffer time segment in the UAV mission scheduling based on the evaluation results to obtain a scheduling window, and extending the time range before and after the scheduling window to obtain several first scheduling windows, specifically involves: analyzing the maximum tolerable delay in the scheduling evaluation matrix to obtain the initial buffer time segment; taking the core interference region as the center, tracing back the time sequence of the mission preparation stage and extending the time sequence of the mission exit stage to obtain the extended time axis.

[0013] On the extended time axis, the initial buffer time segment is expanded according to a preset ratio to obtain the scheduling window; the overlapping interval of the scheduling windows of adjacent UAVs is detected, and the overlapping area is nonlinearly stretched on the time axis to obtain several first scheduling windows.

[0014] In a preferred embodiment, the step of performing avoidance area deduction on the intersection segments of multiple UAV paths within several first scheduling windows to obtain a candidate scheduling path set specifically involves: for each first scheduling window, extracting the predicted path centerline of all UAVs within the window; calculating the intersection angle and spatiotemporal coordinates of the intersection point of the path centerline to construct a three-dimensional conflict heat map; identifying gradient abrupt change regions in the conflict heat map, and generating a dynamic avoidance sphere with the abrupt change point as the center; deforming the dynamic avoidance sphere along the path centerline direction according to a preset deformation degree to obtain an ellipsoid; and expanding the ellipsoid along the tangential direction to form an avoidance channel to obtain a candidate scheduling path set.

[0015] In a preferred embodiment, the step of inserting task matching into the candidate scheduling path set based on the multi-objective matching algorithm to obtain the multi-machine task timing coordination matrix specifically involves: constructing a multi-objective function and converting each path solution in the candidate scheduling path set into a chromosome gene sequence, wherein the chromosome gene sequence includes waypoint sequences and timestamps of multiple UAVs; randomly exchanging the UAV waypoint sequences in the chromosome gene sequence to obtain several populations; based on the several populations, using the improved NSGA-II algorithm to solve the multi-objective function to obtain the Pareto optimal solution set; and encoding the Pareto optimal solution set into the multi-machine task timing coordination matrix.

[0016] In a preferred embodiment, encoding the Pareto optimal solution set into a multi-machine task temporal coordination matrix specifically involves: sorting the Pareto optimal solutions in descending order and selecting several non-dominated solutions; performing temporal deconvolution on the path nodes of each non-dominated solution to obtain the action sequence of each UAV; discretizing the action sequence according to preset time slices and labeling the UAV state in each time slice to obtain several labeled non-dominated solutions; merging several labeled non-dominated solutions to obtain a three-dimensional tensor; and compressing the three-dimensional tensor to obtain the multi-machine task temporal coordination matrix.

[0017] The technical effects and advantages of the multi-drone collaborative UAV scheduling method and system of the present invention are as follows:

[0018] 1. This invention enhances the ability to perceive potential conflict points during mission switching by real-time acquisition of 3D trajectory, velocity, and heading deviation information during UAV mission switching. It employs a density peak clustering algorithm to accurately identify execution interference areas. Secondly, it combines spectral clustering to perform feature separation of interference areas, constructing a dynamic separation interface to achieve fine-grained division of complex intersection areas and precise location of core interference areas, thus improving the accuracy of conflict assessment. Based on this, it uses Poisson distribution and Markov chain models to predict the probability of UAV spacing conflict and heading angle state transitions, constructing a scheduling evaluation matrix to provide data support for the scientific reconstruction of buffer time segments and the dynamic extension of scheduling windows.

[0019] 2. This invention generates dynamically deformable avoidance channels by extrapolating the avoidance area of ​​path intersections within the scheduling window, ensuring safe flight distances between UAVs. Finally, combined with an improved multi-objective optimization algorithm, it achieves task insertion matching of candidate paths, generating a multi-UAV task timing coordination matrix. This effectively coordinates UAV timing, avoids task conflicts, and realizes real-time dynamic scheduling and conflict avoidance during multi-UAV task switching. This not only improves the safety and reliability of task execution but also optimizes path utilization and scheduling efficiency, significantly reducing flight delays and conflict risks. It meets the complex needs of large-scale, multi-task UAV collaborative operations and has broad application prospects and practical value. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a multi-drone collaborative UAV scheduling method according to the present invention.

[0021] Figure 2 This is a schematic diagram of the structure of a multi-machine collaborative unmanned aerial vehicle (UAV) scheduling system according to the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0023] Example 1, Figure 1 This invention presents a multi-drone collaborative UAV scheduling method, comprising the following steps:

[0024] S1. Obtain path response data during the UAV mission switching process and use clustering algorithm to identify the execution interference area during the UAV mission switching process.

[0025] In this example, path response data during the UAV mission switching process is acquired, and a clustering algorithm is used to identify the execution interference region during the UAV mission switching process. Specifically:

[0026] The path response data of the UAV mission switching process is collected in real time and a spatiotemporal response matrix is ​​constructed. The path response data includes three-dimensional trajectory coordinates, velocity vector and heading angle deviation.

[0027] Extract acceleration abrupt changes and heading oscillation intervals from the spatiotemporal response matrix to identify potential interference feature points;

[0028] Density peak clustering algorithm is used to perform spatial density scanning on potential interference feature points to obtain the cluster core area;

[0029] The execution interference area is obtained by expanding outward from the cluster core area with a preset radius and combining it with preset geofencing constraints.

[0030] It should be noted that by calculating the rate of change of velocity at consecutive time points, acceleration exceeding a set threshold (e.g., 2 m / s²) is identified. 2 The moment of change is taken as the abrupt change point, and the continuous change of the heading angle is analyzed. If there is a high-frequency back-and-forth oscillation (e.g., the heading angle fluctuation frequency is greater than 2Hz within 5 seconds), it is marked as a heading oscillation interval, and potential interference feature points are obtained. These interference feature points are then input into the density peak clustering algorithm for spatial clustering analysis. This method automatically identifies the core points of the clusters by calculating the local density of each point and the distance to points with higher density. The steps are as follows:

[0031] Calculate the local density ρ for each feature point;

[0032] For each point, calculate its minimum distance δ to all points with higher density;

[0033] Points with high ρ and high δ values ​​are selected as cluster centers, and multiple cluster core regions are automatically formed.

[0034] Centered on the core region of each cluster, a spherical region with radius R is extended outward to initially generate a candidate set of execution interference regions.

[0035] To ensure the spatial rationality and security of the identified areas, further screening is performed using pre-set geofencing constraints. For example, if a candidate area overlaps with a no-fly zone, buildings, or areas with abnormal ground elevation differences, the boundary of that area is removed or adjusted. Ultimately, a set of high-confidence execution interference areas is obtained, each described by attributes such as center coordinates, boundary radius, and time window, for subsequent feature separation and scheduling reconstruction.

[0036] S2, perform feature separation on the interference area and evaluate the scheduling of UAV missions based on the feature separation results;

[0037] In this example, feature separation is performed on the interference region, specifically as follows:

[0038] Construct a topology map of the execution interference region, and use a spectral clustering algorithm to calculate the eigenvectors of the Laplacian matrix of the topology map of the execution interference region;

[0039] The execution interference region is divided based on the eigenvectors of the Laplacian matrix, resulting in several core interference regions;

[0040] Extract the convex hull boundaries of several interference core regions to obtain the dynamic separation interface.

[0041] It should be noted that, firstly, all interference feature points are extracted from the interference area. These feature points typically include locations where the heading angle fluctuates frequently, locations where acceleration changes abruptly, etc., and their spatial positions are marked using three-dimensional coordinates.

[0042] Then, the spatial connections between these feature points are constructed into a topological graph. Based on the spatial proximity of the feature points, the system assigns a correlation strength to each pair of points to describe their mutual interference. Points that are closer together and have more similar interference changes have a higher correlation strength.

[0043] Based on the constructed topology map, spectral clustering algorithm is used to divide the interference region. Spectral clustering is a graph-based clustering method that analyzes the spatial distribution relationship between interference points to divide the interference region into several sub-regions, each with highly consistent interference characteristics. These sub-regions are the core interference regions, each with high feature point density, consistent fluctuation patterns, and strong spatial coupling. In actual deployment, the system can dynamically adjust the number of regions based on the distribution density of interference points, typically set to three to five core regions.

[0044] For each interference core region, the system further extracts its boundary shape. Specifically, it calculates the minimum bounding boundary of all feature points in the region to form a three-dimensional convex hull model. This convex hull boundary effectively describes the external shape and distribution contour of the core region. The boundaries of multiple interference core regions together constitute a dynamic separation interface, serving as the basis for dividing the internal structure of the interference region.

[0045] In this example, the scheduling and evaluation of UAV missions are performed based on the feature separation results, specifically as follows:

[0046] A three-dimensional mesh cell is constructed within the dynamic separation interface, and the Poisson distribution algorithm is used to calculate the probability of collision between UAVs in the mesh cell spacing.

[0047] The Markov chain algorithm is used to predict the heading angle conflict state transition path, and a scheduling evaluation matrix is ​​constructed based on the conflict probability between UAVs.

[0048] It should be noted that within each dynamic separation interface, a three-dimensional evaluation grid with spatial resolution is established. This grid is divided at set intervals to form multiple three-dimensional cells, covering the entire interference core area. Each cell represents a local evaluation area, used to statistically evaluate the probability of UAV interference within that area.

[0049] The system records information such as the number of drones passing through, their flight speed, heading changes, and flight time in each grid cell. Statistical analysis is performed using the relative distances between drones, the probability of simultaneous occurrence, and historical mission data to determine the spatial congestion level of each grid cell at different time periods. Based on the statistical results, the system probabilistically assesses the probability of distance conflict between any two drones within the same grid cell. A higher probability value indicates that the area is more likely to become a mission conflict hotspot. Based on known drone flight paths and historical heading data, the system models the heading change trends of drones within the core interference area. By analyzing the heading angle change sequence, the system predicts the heading status change paths of each drone within different time periods. Combining flight speed and trajectory curvature, the system can predict time segments where intersections, approaching, or collisions are possible, thereby inferring the likelihood of heading conflicts between drones within a specific grid cell.

[0050] After obtaining the spatial conflict probability and heading state transition prediction for each grid cell, the system aggregates all evaluation results to construct a task scheduling evaluation matrix. This matrix, with the UAV number as the dimension, records the conflict risk level of each UAV in each core interference area. The scheduling evaluation matrix will serve as a key reference for subsequent scheduling window construction, buffer time adjustment, and path extrapolation, quantifying the interference level and scheduling risk of each UAV at a specific time and spatial location.

[0051] S3. Based on the evaluation results, the buffer time segment in the UAV mission scheduling is reconstructed to obtain the scheduling window, and the time range before and after the scheduling window is extended to obtain several first scheduling windows.

[0052] In this example, based on the evaluation results, the buffer time segment in UAV mission scheduling is reconstructed to obtain a scheduling window. Furthermore, the time range before and after the scheduling window is extended to obtain several first scheduling windows, specifically:

[0053] The initial buffer time segment is obtained by analyzing the maximum tolerable delay in the scheduling evaluation matrix.

[0054] Centered on the core interference area, the time sequence of the mission preparation phase is traced backward, and the time sequence of the mission exit phase is extended backward to obtain the extended timeline.

[0055] On the extended timeline, the initial buffer time segment is expanded by a preset ratio to obtain the scheduling window;

[0056] The overlapping intervals of scheduling windows of adjacent UAVs are detected, and the overlapping intervals are nonlinearly stretched along the time axis to obtain several first scheduling windows.

[0057] It should be noted that, firstly, the maximum allowable time delay for each UAV within the core interference area is extracted from the scheduling evaluation matrix. This delay value is derived by combining historical mission performance, flight speed, path complexity, and mission type, representing the longest acceptable delay for the mission without affecting the overall process. Based on this maximum tolerable delay, an initial buffer time interval is set for each mission segment as a time protection zone during mission execution to absorb time changes caused by local path adjustments.

[0058] Furthermore, by tracing back the time period of the mission preparation phase along the timeline, centered on the core interference area, we can obtain the adjustable time before the mission enters the area. Simultaneously, we extend the time of the exit phase after the mission is completed, forming a complete mission interference impact zone. By analyzing the mission state transition moments of the UAV (such as takeoff, turning, cruising, and landing), these tracing back and extended time periods are integrated into the mission timeline, forming an extended time zone, providing greater adjustability for scheduling window generation.

[0059] In the extended timeline, the initial buffer time segment is magnified by a preset ratio. For example, a larger magnification factor can be set for high-risk areas to increase scheduling redundancy time; for low-risk areas, the original buffer ratio can be retained. The magnified time period is defined as the scheduling window. This window covers high-risk periods before and after critical nodes of the task, providing time tolerance for operations such as path avoidance and intersection prediction.

[0060] Finally, an overlap analysis was performed on the scheduling windows of all UAVs to identify overlapping areas where multiple scheduling windows occurred within the same time period. These overlapping areas are high-risk conflict zones and require focused management. To avoid task conflicts, the time axis of these overlapping areas was non-linearly stretched, i.e., the time interval of some time periods was adjusted differently to make the tasks more sparsely distributed within the same physical time period, thereby reducing the probability of convergence. After the stretching process, several first scheduling windows were generated. Each first scheduling window has a clear start and end time, information on the UAV task segments it covers, and a redundancy duration, providing a temporal structure basis for subsequent path deduction and conflict avoidance.

[0061] S4. In several first scheduling windows, avoidance areas are simulated for multiple UAV path intersection segments to obtain a set of candidate scheduling paths.

[0062] In this example, within several first scheduling windows, avoidance zones are simulated for multiple UAV path intersection segments to obtain a candidate scheduling path set, specifically:

[0063] For each first scheduling window, extract the predicted path centerline of all UAVs within the window;

[0064] Calculate the intersection angle of the path centerlines and the spatiotemporal coordinates of the intersection point to construct a three-dimensional conflict heat map;

[0065] Identify gradient abrupt change regions in the conflict heatmap and generate dynamic avoidance spheres with the abrupt change points as the sphere centers;

[0066] The dynamically avoiding sphere is deformed along the path centerline according to a preset deformation degree to obtain an ellipsoid.

[0067] The ellipsoid is extended along the tangential direction to form an avoidance channel, resulting in a set of candidate scheduling paths.

[0068] It should be noted that for each first scheduling window, the predicted flight paths of all UAVs participating in the scheduling within that time period are extracted. Each path consists of multiple consecutive flight points. Through fitting analysis, the spatial center trajectory is extracted as the path centerline. The path centerline represents the main flight direction and shape of the UAV within a specific time period and serves as the basis for subsequent intersection analysis and conflict assessment.

[0069] Furthermore, spatial analysis is performed on each pair of path centerlines to identify their intersection points and convergence angles. If there is spatial overlap between paths and the convergence angle is less than a set threshold, it is identified as a high-risk intersection segment. For each high-risk intersection point, its corresponding time coordinates are extracted to establish an intersection dataset that combines temporal and spatial factors, which is used for subsequent dynamic avoidance zone construction.

[0070] A 3D conflict heatmap is constructed within each scheduling window using the intersection dataset. This heatmap assesses the conflict risk level at various locations in space based on path density, intersection angle, and temporal overlap, and is labeled with color intensity or numerical value. The heatmap identifies locations and regions with significant risk changes in local space, providing intuitive avoidance reference information. Regions with prominent risk value changes, i.e., risk gradient abrupt change points, are detected in the heatmap. A 3D spherical avoidance zone is constructed centered on these abrupt change points. The size of this sphere is determined by the UAV speed, intersection density, and scheduling window width. This avoidance sphere dynamically moves over time, representing path conflict areas that require priority avoidance.

[0071] The dynamic avoidance sphere is deformed according to the path centerline, elongating it appropriately along the flight direction to form an ellipsoidal structure, thus enhancing the path maneuverability. Subsequently, the system extends multiple ellipsoids tangentially along the path direction, constructing a more connected and maneuverable avoidance channel. This channel possesses flexible bending capabilities to adapt to different mission speeds and headings.

[0072] Ultimately, based on the original path, the system constructs multiple bypass path schemes according to the spatial structure of the avoidance channel. Each path avoids major intersection risk points and is feasible within the scheduling window. After all path schemes are summarized, a candidate scheduling path set is formed, providing path support for subsequent task insertion and multi-machine coordination.

[0073] S5 uses a multi-objective matching algorithm to insert task matching into the candidate scheduling path set, obtains a multi-machine task timing coordination matrix, and uses it for UAV scheduling.

[0074] In this example, a multi-objective matching algorithm is used to perform insertion task matching on the candidate scheduling path set to obtain the multi-machine task timing coordination matrix, specifically:

[0075] A multi-objective function is constructed, and each path solution of the candidate scheduling path set is converted into a chromosome gene sequence, which includes waypoint sequences and timestamps of multiple UAVs;

[0076] Randomly swapping the UAV waypoint sequences in chromosome gene sequences yields several populations;

[0077] Based on several populations, the improved NSGA-II algorithm is used to solve the multi-objective function and obtain the Pareto optimal solution set.

[0078] The Pareto optimal solution set is encoded as a multi-machine task timing coordination matrix.

[0079] It should be noted that the multi-objective function includes: minimizing the total task duration, minimizing path intersection conflicts, minimizing waiting time between tasks, maximizing the utilization of the available scheduling window, and minimizing flight energy consumption. Furthermore, for each path in the candidate scheduling path set, the system serializes it, including the following: the path point sequence for each UAV, the time node of each path point, and the task segment information contained in each path.

[0080] Furthermore, by shuffling and rearranging the task order of multiple scheduling paths, multiple initial matching solutions are formed, called a population. Each population represents a possible UAV task ordering scheme with different path combinations and execution time structures.

[0081] In this example, based on several populations, an improved NSGA-II algorithm is used to solve the multi-objective function and obtain the Pareto optimal solution set, specifically:

[0082] Evaluate the fitness of each population solution under the multi-objective function;

[0083] Compare the nondominance of each solution and select the optimal solution;

[0084] Crossover and mutation processes are applied to optimal solutions to generate new solutions;

[0085] The evaluation and screening process is repeated until the optimal matching set is reached.

[0086] In this example, the Pareto optimal solution set is encoded as a multi-machine task timing coordination matrix, specifically:

[0087] Sort the Pareto optimal solutions in descending order and select several non-dominated solutions;

[0088] Perform temporal deconvolution on the path nodes of each non-dominated solution to obtain the action sequence of each UAV;

[0089] The action sequence is discretized according to a preset time slice, and the UAV state is labeled in each time slice to obtain several labeled non-dominated solutions;

[0090] By merging several labeled non-dominated solutions, a three-dimensional tensor is obtained;

[0091] The three-dimensional tensor is compressed to obtain the multi-machine task timing coordination matrix.

[0092] It should be noted that, firstly, the non-dominated solution set is comprehensively ranked according to scheduling objectives. Ranking indicators include total task time, path conflict intensity, scheduling window utilization, and energy consumption assessment. Based on the ranking results, several optimal non-dominated solutions are selected as representative scheduling schemes to ensure coverage of various scheduling preference types, such as speed priority, safety priority, or energy saving priority.

[0093] In addition, for each selected non-dominated solution, the system extracts the corresponding sequence of multiple UAV path nodes, including the time information, spatial location, and mission status of the flight points.

[0094] The system employs a time-series backward inference algorithm to reconstruct the complete mission action sequence of each UAV point-by-point from the final mission objective. This sequence includes: initiation phase; takeoff process; mid-course navigation; path avoidance; mission execution; return path; landing and exit. The action sequence is divided into equally spaced time slices with a uniform time resolution, for example, each slice consisting of five time units. Within each time slice, the UAV's state is labeled, such as "waiting," "moving," "mission execution," "avoidance," or "idle." All the labeled time slices of all UAVs form multiple separate two-dimensional matrices, each matrix corresponding to a set of non-dominated solutions.

[0095] Subsequently, multiple annotation matrices were encoded and integrated, forming a three-dimensional structure based on task number, UAV number, and time slice number, represented as follows: the first dimension is the UAV number, and the second dimension is the time slice number.

[0096] The third dimension is the action state number.

[0097] Finally, to reduce scheduling and control costs and improve real-time computing efficiency, the system performs data compression on the aforementioned three-dimensional structure. Compression methods include: state redundancy identification within adjacent time slices; action segment merging; invalid segment removal; and state segment encoding compression.

[0098] Example 2, Figure 2 This invention provides a multi-drone collaborative UAV scheduling system, comprising an interference identification module, a scheduling evaluation module, a scheduling reconstruction module, a path deduction module, and a task coordination module.

[0099] The interference identification module is used to acquire path response data during the UAV mission switching process and to use clustering algorithms to identify the execution interference area during the UAV mission switching process.

[0100] The scheduling evaluation module is used to perform feature separation of the execution interference area and to perform scheduling evaluation of the UAV mission based on the feature separation results.

[0101] The scheduling reconstruction module is used to reconstruct the buffer time segment in UAV mission scheduling based on the evaluation results, obtain the scheduling window, and extend the time range before and after the scheduling window to obtain several first scheduling windows.

[0102] The path deduction module is used to perform avoidance area deduction for multiple UAV path intersection segments in several first scheduling windows to obtain a set of candidate scheduling paths;

[0103] The task coordination module is used to perform insertion task matching on the candidate scheduling path set based on the multi-objective matching algorithm, obtain the multi-machine task timing coordination matrix, and use it for UAV scheduling.

[0104] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0105] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0106] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0107] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0108] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0109] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for scheduling multi-drone collaborative unmanned aerial vehicles (UAVs), characterized in that, Includes the following steps: The path response data during the UAV mission switching process is acquired, and a clustering algorithm is used to identify the execution interference area during the UAV mission switching process. Specifically: The path response data of the UAV mission switching process is collected in real time and a spatiotemporal response matrix is ​​constructed. The path response data includes three-dimensional trajectory coordinates, velocity vector and heading angle deviation. Acceleration mutation points and heading oscillation intervals are extracted from the spatiotemporal response matrix to identify potential interference feature points. The density peak clustering algorithm is used to perform spatial density scanning on the potential interference feature points to obtain the cluster core area. The cluster core area is expanded outward with a preset radius, and combined with preset geofence constraints, the execution interference area is obtained. The interference area is segmented by features, and the scheduling and evaluation of UAV missions are performed based on the feature segmentation results, specifically as follows: A topology map of the execution interference region is constructed, and the Laplacian matrix eigenvector of the topology map is calculated using a spectral clustering algorithm. Based on the Laplacian matrix eigenvector, the execution interference region is divided into several interference core regions. The convex hull boundaries of several interference core regions are extracted to obtain a dynamic separation interface. A three-dimensional mesh cell is constructed within the dynamic separation interface, and the Poisson distribution algorithm is used to calculate the UAV spacing conflict probability between mesh cells. The Markov chain algorithm is used to predict the heading angle conflict state transition path, and a scheduling evaluation matrix is ​​constructed based on the UAV spacing conflict probability. Based on the evaluation results, the buffer time segment in the UAV mission scheduling is reconstructed to obtain the scheduling window, and the time range before and after the scheduling window is extended to obtain several first scheduling windows. In several first scheduling windows, avoidance areas are simulated for multiple UAV path intersection segments to obtain a set of candidate scheduling paths; Based on the multi-objective matching algorithm, the candidate scheduling path set is inserted for task matching to obtain the multi-machine task timing coordination matrix, which is then used for UAV scheduling.

2. The multi-drone collaborative UAV scheduling method according to claim 1, characterized in that, Based on the evaluation results, the buffer time segment in the UAV mission scheduling is reconstructed to obtain a scheduling window, and the time range before and after the scheduling window is extended to obtain several first scheduling windows, specifically: The initial buffer time segment is obtained by analyzing the maximum tolerable delay in the scheduling evaluation matrix. Centered on the core interference area, the time sequence of the mission preparation phase is traced backward, and the time sequence of the mission exit phase is extended backward to obtain the extended timeline. On the extended timeline, the initial buffer time segment is expanded by a preset ratio to obtain the scheduling window; The overlapping intervals of scheduling windows of adjacent UAVs are detected, and the overlapping intervals are nonlinearly stretched along the time axis to obtain several first scheduling windows.

3. The multi-drone collaborative UAV scheduling method according to claim 2, characterized in that, In several first scheduling windows, avoidance areas are simulated for multiple UAV path intersection segments to obtain a candidate scheduling path set. Specifically: For each first scheduling window, extract the predicted path centerline of all UAVs within the window; Calculate the intersection angle of the path centerlines and the spatiotemporal coordinates of the intersection point to construct a three-dimensional conflict heat map; Identify gradient abrupt change regions in the conflict heatmap and generate dynamic avoidance spheres with the abrupt change points as the sphere centers; The dynamic avoidance sphere is deformed along the path centerline according to a preset deformation degree to obtain an ellipsoid. The ellipsoid is extended along the tangential direction to form an avoidance channel, resulting in a set of candidate scheduling paths.

4. The multi-drone collaborative UAV scheduling method according to claim 3, characterized in that, The process of inserting tasks into the candidate scheduling path set using a multi-objective matching algorithm to obtain a multi-machine task timing coordination matrix is ​​as follows: A multi-objective function is constructed, and each path solution of the candidate scheduling path set is converted into a chromosome gene sequence, which includes waypoint sequences and timestamps of multiple UAVs; Randomly swapping the UAV waypoint sequences in chromosome gene sequences yields several populations; Based on several populations, the improved NSGA-II algorithm is used to solve the multi-objective function and obtain the Pareto optimal solution set. The Pareto optimal solution set is encoded as a multi-machine task timing coordination matrix.

5. The multi-drone collaborative UAV scheduling method according to claim 4, characterized in that, The process of encoding the Pareto optimal solution set into a multi-machine task timing coordination matrix is ​​as follows: Sort the Pareto optimal solutions in descending order and select several non-dominated solutions; Perform temporal deconvolution on the path nodes of each non-dominated solution to obtain the action sequence of each UAV; The action sequence is discretized according to a preset time slice, and the UAV state is labeled in each time slice to obtain several labeled non-dominated solutions; By merging several labeled non-dominated solutions, a three-dimensional tensor is obtained; The three-dimensional tensor is compressed to obtain the multi-machine task timing coordination matrix.

6. A multi-drone collaborative UAV scheduling system, applied to the multi-drone collaborative UAV scheduling method according to any one of claims 1-5, characterized in that, It includes an interference identification module, a scheduling evaluation module, a scheduling reconstruction module, a path deduction module, and a task coordination module: The interference identification module is used to acquire path response data during the UAV mission switching process and to identify the execution interference area during the UAV mission switching process using a clustering algorithm. Specifically: The path response data of the UAV mission switching process is collected in real time and a spatiotemporal response matrix is ​​constructed. The path response data includes three-dimensional trajectory coordinates, velocity vector and heading angle deviation. Acceleration mutation points and heading oscillation intervals are extracted from the spatiotemporal response matrix to identify potential interference feature points. The density peak clustering algorithm is used to perform spatial density scanning on the potential interference feature points to obtain the cluster core area. The cluster core area is expanded outward with a preset radius, and combined with preset geofence constraints, the execution interference area is obtained. The scheduling evaluation module is used to perform feature separation of the execution interference area and to evaluate the scheduling of UAV missions based on the feature separation results. Specifically: A topology map of the execution interference region is constructed, and the Laplacian matrix eigenvector of the topology map is calculated using a spectral clustering algorithm. Based on the Laplacian matrix eigenvector, the execution interference region is divided into several interference core regions. The convex hull boundaries of several interference core regions are extracted to obtain a dynamic separation interface. A three-dimensional mesh cell is constructed within the dynamic separation interface, and the Poisson distribution algorithm is used to calculate the UAV spacing conflict probability between mesh cells. The Markov chain algorithm is used to predict the heading angle conflict state transition path, and a scheduling evaluation matrix is ​​constructed based on the UAV spacing conflict probability. The scheduling reconstruction module is used to reconstruct the buffer time segment in UAV mission scheduling based on the evaluation results, obtain the scheduling window, and extend the time range before and after the scheduling window to obtain several first scheduling windows. The path deduction module is used to perform avoidance area deduction for multiple UAV path intersection segments in several first scheduling windows to obtain a set of candidate scheduling paths; The task coordination module is used to perform insertion task matching on the candidate scheduling path set based on the multi-objective matching algorithm, obtain the multi-machine task timing coordination matrix, and use it for UAV scheduling.

Images