Multi-uav cooperative task allocation and conflict resolution system and method

Abstract

The application relates to the technical field of unmanned aerial vehicle control, in particular to a task allocation and conflict resolution system and method for multi-unmanned aerial vehicle cooperative operation. The application proposes the following scheme: a crop growth state atlas and a three-dimensional land model are constructed, an initial operation area is divided, and a response weight is generated; based on path planning and resource adaptation, a flight route is dynamically generated, and the crop state and the unmanned aerial vehicle state are monitored in real time; when the adjustment condition is met, a multi-dimensional dynamic task evaluation model is constructed, a particle swarm optimization and an autonomous negotiation mechanism are combined, and task migration and conflict decoupling are completed. The application is suitable for the precision agriculture scene, realizes dynamic scheduling of the unmanned aerial vehicle path in the operation area and precise coverage of the high-priority area, and improves the operation efficiency and the resource cooperation ability.

Description

Technical Field

[0001] This application relates to the field of unmanned aerial vehicle (UAV) control technology, and in particular to a system and method for task allocation and conflict resolution in multi-UAV collaborative operations. Background Technology

[0002] Existing multi-drone collaborative operation solutions in agriculture often face problems such as high overlap of operation paths, delayed task scheduling response, and insufficient coverage of high-priority areas. In actual execution, multiple drones are prone to issues like unreasonable task area division, uneven resource load, and frequent flight path conflicts, leading to a decline in overall operational efficiency. Furthermore, when faced with sudden changes in crop conditions or local anomalies, existing methods cannot achieve timely and effective task migration and path adjustment, resulting in delayed response in hotspot areas and even omissions of unmanned areas. In addition, due to insufficient task priority identification capabilities, some high-value crop areas are not adequately covered, affecting the effectiveness of precision operations and the quality of agronomic output.

[0003] To address the above issues, this application presents a system and method for task allocation and conflict resolution in multi-UAV collaborative operations. Summary of the Invention

[0004] The technical problem this application aims to solve is to address the shortcomings of existing technologies by providing a task allocation and conflict resolution system and method for multi-UAV collaborative operations. This system constructs a crop growth status map and a 3D plot model to divide the initial operation area and generate response weights. Based on path planning and resource adaptation, it dynamically generates flight routes and monitors crop and UAV status in real time. When adjustment conditions are met, a multi-dimensional dynamic task evaluation model is constructed, combining particle swarm optimization and autonomous negotiation mechanisms to complete task migration and conflict decoupling. This method is applicable to precision agriculture scenarios, achieving dynamic scheduling of UAV paths and precise coverage of high-priority areas within the operation area, thus improving operational efficiency and resource coordination capabilities.

[0005] To achieve the above objectives, this application provides the following technical solution:

[0006] A method for task allocation and conflict resolution in multi-UAV collaborative operations, the method comprising:

[0007] A three-dimensional land parcel model of the target area is obtained, and the flight route of the UAV is planned through an optimization algorithm based on the terrain features and distribution information of the three-dimensional land parcel model.

[0008] The flight routes of the drones are assigned to the corresponding drones. During the operation of the drones, the growth status information of the crop area and the status information of the drones are monitored to determine whether the conditions for entering the adjustment phase are met.

[0009] If the conditions are met, the area to be adjusted is obtained from the drone flight path, and the drones that will perform the corresponding tasks for the area to be adjusted are reassigned according to the crop growth status map of the area to be adjusted.

[0010] The process of planning the drone flight path using an optimization algorithm includes:

[0011] Based on the three-dimensional land parcel model and the number of drones to be operated, multiple initial operation areas are divided, wherein the multiple initial operation areas do not overlap with each other;

[0012] Based on the priority, workload, and current status of the drones to be operated in each initial work area, the initial work areas are assigned to the corresponding drones.

[0013] Based on the initial operating area corresponding to each UAV, and combined with the preset crop growth status map, a complete flight plan is generated. The crop growth status map represents the crop growth activity index and is constructed based on historically collected crop data.

[0014] The process of generating a complete flight plan by combining a preset crop growth status map includes:

[0015] Extract the boundary information and terrain features of the crop area from the initial working area to determine the operating range and flight restrictions of the UAV;

[0016] The crop area is divided into regular grids, where the size of each grid is sufficient to cover the effective area captured by a single drone shot. The crop response weight associated with each grid is marked by the crop growth status map, and the crop response weight is determined based on the growth status of the crop area in the crop growth status map.

[0017] Extract the vertex set of the boundary of the regular grid, calculate the minimum bounding rectangle based on the vertex set, and take the long side direction of the rectangle as the working direction of the UAV;

[0018] Generate parallel paths along the work direction and mark the start and end coordinates of each path;

[0019] The generated path is subjected to obstacle detection. If there is an obstacle grid in the path, the path is replanned in the area corresponding to the obstacle grid using the A* algorithm. The starting point and ending point of the path after bypassing the obstacle are reconnected. If there is no obstacle grid in the path, the path is added to the path set. The path is traversed repeatedly until all paths are added to the path set.

[0020] A multi-objective evaluation function is established, wherein the objectives include path length, number of turns, coverage repetition rate, energy consumption, and crop response weight. The evaluation score of each path in the path set is calculated according to the multi-objective evaluation function. The path corresponding to the highest evaluation score is taken as the optimal path scheme. The UAV flight route is generated based on the path nodes, turning points, and obstacle avoidance parameters of the optimal path scheme.

[0021] Generate parallel paths along the stated work direction, and mark the start and end coordinates of each path, including:

[0022] The starting point of the path is determined based on the crop response weights of the rule grid;

[0023] The endpoint of the path is determined based on the drone's remaining battery power, wind direction, and return route.

[0024] The starting points of each parallel path are different.

[0025] If there are multiple drones to be operated in the initial operation area, the method further includes:

[0026] Obtain the path set and the evaluation scores corresponding to the paths in the path set, and prioritize the paths.

[0027] Based on the current positions of multiple drones awaiting operation, and in descending order of the path priority, the drone that meets the flight requirements of the current path and is closest to the starting point is selected as the execution drone for the current path, until there are no more drones awaiting operation.

[0028] The conditions for entering the adjustment phase include at least one of the following:

[0029] The change in crop growth status indicators in any sub-region of the crop region is greater than or equal to a preset generation threshold.

[0030] If any sub-region of the crop region is found to have spatially clustered pests and diseases, high temperature or low humidity anomalies, the high temperature and low humidity can be determined by preset high temperature thresholds and low humidity thresholds. If the temperature of the sub-region is greater than or equal to the high temperature threshold, it is a high temperature anomaly; if the humidity of the sub-region is less than or equal to the low humidity threshold, it is a low humidity anomaly.

[0031] The drone's battery level is less than or equal to a preset flight safety threshold during operation;

[0032] The drone may experience deviation, mission interruption, or path execution timeout during the operation.

[0033] The flight paths of multiple drones intersect in space, and the safety interval is less than the preset tolerance threshold.

[0034] The drones assigned to perform the corresponding tasks in the area to be adjusted are reassigned based on the crop growth status map of the area to be adjusted, including:

[0035] A multidimensional dynamic task evaluation model is constructed for the region to be adjusted. The multidimensional dynamic task evaluation model includes the crop growth response dimension, the task urgency dimension, the historical task effect dimension, and the task migration cost dimension.

[0036] The task migration evaluation score of the region to be adjusted is calculated based on the multidimensional dynamic task evaluation model. The task migration evaluation score is used to determine the task migration priority of the region to be adjusted.

[0037] Based on the task migration evaluation score and the spatial distribution characteristics of the area to be adjusted, a set of candidate UAV tasks is generated;

[0038] Using the particle swarm optimization algorithm, based on the remaining battery power, flight distance, current load, historical operation efficiency, and location status of the remaining drones, drones that can execute the candidate task set are obtained, and the corresponding tasks in the candidate task set are assigned to the corresponding drones.

[0039] The drones that obtain the drone candidate task set include:

[0040] Based on the preset resource fitness evaluation function in the particle swarm optimization algorithm, calculate the resource fitness of each UAV for executing the UAV candidate task set;

[0041] Based on the resource adaptability, determine the initial task allocation scheme for the UAV;

[0042] Based on the conflict situation in the initial task allocation scheme of the UAV, the initial task allocation scheme of the UAV is re-allocated through the autonomous negotiation logic between UAVs. The conflict situation includes UAV path intersection, spatiotemporal overlap and unbalanced task load.

[0043] The autonomous negotiation logic includes conflict resolution between drones based on game theory strategies, wherein:

[0044] Each drone broadcasts negotiation information to other drones based on its own status. The order in which drones perform tasks is updated based on the results of the negotiation information exchange, until a balance is reached where path conflicts between drones are minimized and the overall benefits of the task are maximized.

[0045] A task allocation and conflict resolution system for multi-UAV collaborative operations, the system comprising:

[0046] The modeling module is used to acquire a 3D land parcel model of the target area and construct the work area based on terrain features, obstacle information, and crop growth status maps.

[0047] The path planning module is used to generate initial flight routes for multiple drones based on the operation area model.

[0048] The task allocation module is used to dynamically allocate tasks to flight paths based on crop response weights, task priorities, and UAV status parameters.

[0049] The conflict resolution module is used to perform path adjustment, time window misalignment, or flight altitude reconstruction when path conflicts exist, and output updated flight control commands.

[0050] Compared with the prior art, the beneficial effects of this application are:

[0051] This application achieves accurate identification and priority coverage of high-response areas within the operational region through task value modeling based on crop growth status maps and multi-objective path optimization logic, thereby improving the operational targeting and resource allocation efficiency of drone swarms. It can respond in real-time to changes in crop status and fluctuations in drone resources during task execution, enabling flexible task reconfiguration and path conflict resolution, improving the stability and adaptability of multi-drone collaborative operations, and reducing operational redundancy and omissions. Attached Figure Description

[0052] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0053] Figure 1 This is a schematic diagram illustrating an exemplary application scenario of an embodiment of this application;

[0054] Figure 2 This is a flowchart illustrating the task allocation and conflict resolution method for multi-UAV collaborative operations according to an embodiment of this application.

[0055] Figure 3 This is a diagram of the crop identification network structure in an embodiment of this application;

[0056] Figure 4 This is a schematic diagram illustrating the process of generating a complete flight plan for an embodiment of this application. Detailed Implementation

[0057] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0058] The term "embodiment" as used herein means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0059] The task allocation and conflict resolution method for multi-UAV collaborative operations proposed in this application is applicable to UAV operation systems that require collaborative execution of multiple autonomous units, real-time changes in environmental conditions, and dynamic reconstruction of operational scenarios. The method is particularly suited to application scenarios such as those where crop growth exhibits spatial non-uniformity, task area division has ambiguous boundaries, or complex agricultural operation environments where UAVs may experience path congestion, task anomalies, or state imbalances during operation.

[0060] This application does not rely on rigid plot division of the mission area, nor does it require prior division of labor or fixed flight path templates among UAVs. Instead, it completes mission reconstruction and scheduling optimization for dynamically changing target operation areas by constructing crop growth status maps, UAV resource parameter models, and path conflict detection logic.

[0061] Application scenarios include, but are not limited to, the following:

[0062] Precision agriculture involves multiple drones spraying, fertilizing or sowing large areas of fields, where there are significant differences in the growth status of crops in different regions.

[0063] When multiple drones simultaneously enter high-response crop areas to perform tasks, there are operational risks such as battery depletion, operation failure, or path overlap.

[0064] During mission execution, unexpected changes may occur in the site environment, climate conditions, or crop health status, necessitating local reconstruction and redistribution of the flight path;

[0065] There are some high-value operation points or hot spots in the crop area, and their priority needs to be dynamically adjusted during the task planning process.

[0066] It should be noted that the scheduling and conflict resolution logic in this application is not limited to agricultural operations and can also be extended to:

[0067] Flight inspection missions targeting urban building complexes often encounter environments with obstructions and limited airspace resources.

[0068] During forest fire patrol operations, the distribution of heat sources for monitoring targets is uneven, and the path is restricted by wind direction and obstacles;

[0069] Collaborative inspection tasks in large industrial areas or transportation networks require path allocation to be dynamically optimized based on target anomaly priority and equipment status.

[0070] This application is particularly well-suited for systems that have one or more of the following characteristics:

[0071] The work area contains heterogeneous response regions, and the spatial distribution of task density is uneven.

[0072] The operating status of multiple drones varies in real time, such as differences in remaining battery power, operating speed, or historical load.

[0073] The system lacks high-frequency global command and dispatch capabilities and needs to perform local autonomous reallocation based on task weights and resource matching relationships.

[0074] The paths of multiple drones may dynamically intersect, requiring adaptive multi-objective path planning to be achieved without collisions.

[0075] Please see Figure 1 This figure is a schematic diagram of an exemplary application scenario provided by an embodiment of this application.

[0076] like Figure 1 As shown, multiple drones, including Drone 1, Drone 2, and Drone 3, are deployed above the farmland area to perform collaborative tasks. The farmland area is divided using a regular grid to clearly define the spatial boundaries of different sub-areas. The crop symbols drawn in each grid represent the actual crop planting distribution. The lower right area is the target area with a strong crop response, i.e., the area to be adjusted. It is marked in the figure with a dotted shaded area, indicating that this area needs to be prioritized for operation or enter the task reconfiguration and scheduling phase.

[0077] Figure 1 The three drones shown are in different operational states and flight phases. The dashed lines indicate operational states, where a drone detects a response hotspot in the work area and needs to cooperate with other drones. The drones are connected by solid arrows, indicating the dynamic migration path of the task. The curves and dashed arrows represent the adjustment process of the current flight path.

[0078] Figure 1 The illustrated scenario applies to the following situations: the crop growth status is uneven, there are overlapping or gaps in the drone path, the system dynamically adjusts the task allocation results based on the crop status map, and replans the drone flight path to maximize resource utilization efficiency and minimize path conflict risk.

[0079] It is understandable that the method of this application is also applicable when a certain drone is unable to complete the task.

[0080] Next, with reference to the accompanying drawings, the task allocation and conflict resolution method for multi-UAV collaborative operations provided in the embodiments of this application will be introduced. Figure 2 The method shown includes the following steps S1-S3:

[0081] S1: Obtain a three-dimensional land parcel model of the target area, and plan the UAV flight route using an optimization algorithm based on the terrain features and distribution information of the three-dimensional land parcel model;

[0082] In this embodiment, the target area can be a crop planting area, an economic forest area, or cultivated land extending from a greenhouse. The plot model is obtained through multi-view image modeling or laser point cloud reconstruction, without limiting the specific modeling method. The 3D model includes data such as slope, elevation difference, and boundaries to determine the workable area and obstacle boundaries. To improve operational efficiency and reduce flight redundancy, a minimum turning route set generated based on the plot's geometric contour and crop row and column distribution direction is further combined with operational priority, plant height map, and area avoidance attributes for trajectory constraint optimization, ultimately outputting a set of flight routes that meet the constraints. The optimization algorithm in this step does not rely on static plot division and has flexible adaptability to complex planting areas, which helps improve flight path quality and reduce path overlap problems caused by irregular areas.

[0083] S2: Assign the drone flight route to the corresponding drone, monitor the growth status information of the crop area and the status information of the drone during the drone operation, and determine whether the conditions for entering the adjustment phase are met.

[0084] In this embodiment, the initial allocation of flight routes considers the UAV's current starting position, remaining battery power, and payload, and is achieved through a track priority and resource matching degree calculation strategy. Growth status information can be generated into a response layer based on historical multispectral maps, red-edge index maps, or remote sensing time-series data, reflecting the current target area's health distribution, abnormal hotspots, and potential pest and disease risk points. UAV status information includes, but is not limited to, battery level, heading stability, mission execution delay, and number of mission failures. The methods for collecting UAV status information are clear to those skilled in the art; only the minimum criteria for determining whether to enter the adjustment phase are required.

[0085] S3: If satisfied, obtain the area to be adjusted from the UAV flight path, and reallocate the UAVs that will perform the corresponding tasks for the area to be adjusted according to the crop growth status map of the area to be adjusted.

[0086] In this embodiment, the reallocation process considers the task migration cost function and UAV resource availability, combining remaining range, current task density, and path offset cost, and uses a particle swarm optimization algorithm to optimize the matching of UAV and task path pairs. Simultaneously, to avoid multiple UAVs concentrating in hotspot areas, causing path conflicts or resource overload, path conflict detection and dynamic conflict resolution logic is implemented, specifically including time-window staggered scheduling, altitude-layer redistribution, and path micro-offset schemes. This ensures that even when tasks are concentrated at high density in local areas, the system can still maintain path decoupling, reasonable resource allocation, and stable overall operational efficiency.

[0087] Before elaborating on the specific implementation logic for each step, this application's embodiments need to reiterate:

[0088] In multi-drone collaborative operation scenarios, not all operational areas possess the characteristics of clear structure, uniform load, and stable response. Taking agricultural applications as an example, the following uncertainties often exist in medium to large-sized fields or mixed crop areas:

[0089] First, crop growth conditions exhibit significant spatiotemporal heterogeneity. For example, humidity gradients, pest and disease spread areas, or areas of uneven fertility within the same plot may change rapidly over time.

[0090] Secondly, the operating status of drone swarms is not absolutely synchronized. For example, power depletion, path deviation, and communication link delays can all lead to a decrease in operational consistency.

[0091] Third, due to factors such as flight airspace, obstacle distribution, and vegetation height, some areas already have the risk of resource coverage edges or path overlap during the initial track generation stage.

[0092] Understandably, traditional path allocation methods are mostly based on fixed-region partitioning or cost matrix matching. When faced with the aforementioned disturbances, they often can only compensate by increasing redundant flights or setting buffers, which can easily lead to low resource utilization or operational delays in high-response areas. In addition, most existing methods handle path conflict identification and task adjustment separately, i.e., allocation first, conflict identification later, and then backtracking correction. This fails to consider the dynamic trade-off between the probability of path intersection and priority value in the early stages of allocation, resulting in a decrease in global path efficiency.

[0093] The core of the task allocation and conflict resolution method proposed in this application is not to simply optimize the flight path itself, but to model the path allocation problem as a multi-objective cooperative resource scheduling game scenario:

[0094] The operational path not only carries spatial information about the crop's status but also implicitly includes multi-dimensional factors such as task priority, resource availability, and potential for track interference. In actual execution, task coordination between UAVs is no longer viewed as a static resource allocation problem but rather as a dynamic process of adjusting and balancing resources and targets.

[0095] Specifically, in task allocation, a task scoring model is constructed by integrating response weights from crop growth status maps, time window values, and the UAV's own status. This is then combined with a multi-objective optimization algorithm to dynamically search for and match the optimal path allocation strategy. Compared to the traditional approach of planning paths and then making local adjustments, this application introduces conflict potential as a constraint variable at the initial stage of task allocation, ensuring safe isolation even when UAVs concentrate in high-response areas through time-series scheduling, flight altitude layering, or path misalignment.

[0096] Furthermore, during conflict resolution, a fully centralized command-issuing model was not adopted. Instead, each drone was allowed to broadcast its own resource status and execution intentions to achieve local autonomous negotiation-based path reallocation. Based on a collaborative peer-to-peer bidding model in game theory, intersecting path segments were weighted and voted on, and the operation sequence or control parameters were adjusted accordingly. Without relying on a central unified scheduling system, the overall scalability and robustness of the system were enhanced, making it particularly suitable for typical scenarios such as farmland, mountainous areas, and forest areas with unstable communication links or strong environmental constraints.

[0097] Next, we will further elaborate on the part of the method in this application regarding the three-dimensional land parcel model.

[0098] Before executing a mission on the target area, it is necessary to first collect multi-angle images of the target area using drones, and then use feature matching and structured light technology to generate 3D point cloud data to construct a complete land parcel model. The 3D model provides accurate terrain and boundary information, which is helpful for subsequent flight route planning and regional analysis.

[0099] In one example, the construction of a 3D land parcel model includes the following four aspects:

[0100] Firstly, multi-angle images of the target area are collected by the image sensor carried by the drone, and the multi-angle images are stitched together by the feature matching algorithm to generate continuous images;

[0101] Specifically, aerial photography of the target area is conducted from different angles and heights to collect multi-angle images. These images are then input into feature matching algorithms, such as the ORB algorithm, to extract key feature points. Image stitching technology is then used to generate continuous images. The acquisition of multi-angle images ensures the integrity of the covered area, while feature matching and stitching techniques can integrate discrete images into continuous images, providing a comprehensive description of the target area.

[0102] Secondly, based on the continuous images, three-dimensional point cloud data of the target area is generated using structured light technology, and the three-dimensional point cloud data is processed to extract terrain features.

[0103] Specifically, the spatial position of each pixel is calculated using the principle of ray intersection in multi-view images to generate a dense point cloud. The generation of the dense point cloud can refer to the specific techniques of image point cloudification, which will not be elaborated here. The dense point cloud is filtered to remove noise, and parameters such as terrain elevation, slope, and aspect are extracted from the point cloud data. The average elevation difference and aspect change rate of each cell are calculated through grid division, i.e., terrain features.

[0104] Thirdly, a crop identification network is constructed, and the continuous images are used as input parameters of the crop identification network. The crop identification network processes the input parameters and outputs crop regional distribution information.

[0105] In this embodiment, the crop recognition network is a deep learning-based semantic segmentation model. The input is continuous imagery, and the output is a crop region distribution map. The network employs an improved DeepLabv3+ architecture, with a lightweight ResNet backbone, combined with dilated convolutional pyramid pooling modules to enhance multi-scale feature extraction capabilities. During model training, a labeled dataset covering crop images at different growth stages, terrain conditions, and lighting conditions is used. The network optimizes its ability to distinguish crop regions from the background (such as soil and weeds) through contextual feature fusion and spatial attention. Specifically, the model automatically outputs accurate distribution information of crop regions, reducing errors from manual annotation and judgment. Deep learning enhances the segmentation capability for complex boundaries and small areas of crop regions.

[0106] Please see Figure 3 The crop identification network structure diagram of this application embodiment.

[0107] In an optional embodiment, the crop identification network includes:

[0108] The residual layer is used to expand the planar dimension of the input parameters through bilinear interpolation. The residual network is used as the backbone network to extract deep features and shallow features. The deep features are used to extract semantic information, and the shallow features are used to recover boundary detail information.

[0109] Specifically, bilinear interpolation is used to preprocess the input image, expanding its planar dimension and enabling the feature extraction process to better preserve detailed information. Simultaneously, a residual network is employed as the backbone to extract multi-level features. Deep features, through multi-layer convolutions and non-linear activation functions, extract global semantic information from the image. These features are primarily used to identify the overall distribution patterns and shape characteristics of crop regions. Shallow features, through fewer convolutional operations, preserve the image's boundaries and texture information, and are used to restore details at the boundaries of crop regions.

[0110] The region recognition layer is used to fuse the deep features and shallow features to generate a feature map. The feature map is processed by convolution kernels and attention mechanisms to extract region semantic information. The label weights are calculated based on the region semantic information, and the label weights are classified by a Softmax classifier to determine the category information of the region.

[0111] Specifically, deep and shallow features are combined using a pixel-by-pixel weighted summation to generate a feature map that contains both global semantic information and preserves boundary details. Multiple convolutional kernels are applied to the fused feature map to further extract spatial correlation information. Kernel sizes of 1×1, 3×3, and 5×5 are used to capture spatial features at different scales. Channel attention and spatial attention mechanisms are employed to dynamically adjust the response values ​​of important regions in the feature map, focusing on key features in the crop region and reducing interference from non-target regions.

[0112] The output layer is used to segment crop regions based on the classification results of the region recognition layer, perform pixel-level statistics on crop regions, calculate crop coverage area and distribution characteristics, and generate crop region distribution information.

[0113] Specifically, based on the classification results output by the region recognition layer, crop region pixels are separated individually, and a binary crop region segmentation map is generated. Based on the segmentation results, the shape features (such as aspect ratio and compactness) and spatial distribution features (such as the area distribution of the Voronoi diagram) of the crop region are calculated. The generation and statistical calculation of the segmentation map visualize the spatial distribution information of the crop region, and the statistical results such as coverage area provide an intuitive reference for agricultural management. Through the extraction of distribution features, the planting patterns and density distribution of crops can also be further analyzed.

[0114] Fourthly, a three-dimensional land parcel model is constructed based on the crop area distribution information and the terrain features;

[0115] Specifically, crop distribution information is fused with terrain features, and a complete 3D plot model is generated using 3D modeling tools. During the fusion process, crop areas are cropped based on point cloud data, eliminating point clouds from non-target areas; weighted interpolation methods are used to smooth boundary areas, ensuring the continuity and accuracy of the model. The 3D plot model includes the boundary shapes of crop areas, terrain undulations, and elevation differences, forming the basic model for flight path planning and data analysis.

[0116] Next, we will further elaborate on the optimization algorithm in this application.

[0117] In one example, the specific steps of S1 are as follows:

[0118] S1.1: Based on the three-dimensional land parcel model and the number of drones to be operated, divide the area into multiple initial operation zones, wherein the multiple initial operation zones do not overlap with each other;

[0119] Specifically, in the early stages of mission scheduling, in order to ensure that multiple drones can cover the target area within the same operation cycle and that their operation paths are independent of each other, it is necessary to spatially divide the overall area based on the constructed three-dimensional land parcel model.

[0120] In this embodiment, a regularized partitioning strategy is used to complete the initial operation area division. Area division can be based on grid equalization, region segmentation along the main axis, or clustering segmentation based on operation density. In some optional implementations, the main axis scanning slicing method can be used, combining the main direction of the plot with the crop row structure, to divide the target area into strip-shaped regions according to the direction of convenient operation. This is further reinforced by a terrain obstacle buffer zone to correct the boundary lines, ensuring that each initial operation area is both executable and safe.

[0121] Furthermore, during the partitioning process, the operating radius, coverage bandwidth, and flight safety buffer zone settings of the drones are considered to ensure that each sub-region can be independently completed by a single drone in space, and that the edge regions have minimal overlapping boundaries. The resulting initial operating areas meet the spatial geometric requirements of no overlap between them and a completely closed coverage area.

[0122] Furthermore, the number of designated areas corresponds one-to-one with the number of available drones. The partitioning strategy is not a simple equal-area division, but rather an optimization based on slope aspect information from the 3D model. This avoids forcibly including areas with excessively steep slopes or obvious obstacles within a single area, ensuring the consistency and feasibility of subsequent flight path planning. This achieves a reasonable initial task space partitioning, providing a high-resolution and clearly defined input basis for subsequent task allocation, and preventing overlap and intersection between drone operating areas and the generation of blind spots.

[0123] In some optional embodiments, the number of divided work areas may be less than the number of available drones. In this case, multiple drones are allowed to collaboratively execute the same work area to improve the processing capacity for large, high-priority areas. During collaborative execution, the work area is further divided into multiple parallel sub-paths during the path planning stage and distributed accordingly. Conversely, to reduce path switching complexity and scheduling latency, this application does not recommend assigning a single drone to multiple initial work areas simultaneously at the initial stage of task allocation. That is, the number of initial work areas should not exceed the number of available drones to avoid wasting air resources and scheduling chaos caused by task switching, and to maintain the continuity and logical consistency of drone flight paths.

[0124] S1.2: Based on the priority of each initial work area, the workload, and the current status of the drones to be worked, the initial work areas are assigned to the corresponding drones;

[0125] Specifically, after the initial area division is completed, path scheduling and resource matching need to be performed based on the complexity of the task and the carrying capacity of the drones. The priority of the task area can be determined by the historical drone operation results. Drone status parameters include real-time remaining battery power, distance from starting position, maximum range, historical operation coverage, and remaining payload capacity.

[0126] In this embodiment, a weighted greedy allocation strategy is used for initial matching. The non-operational flight distance from the UAV's starting point to the target area is also considered to ensure that resource scheduling achieves an optimal trade-off between coverage efficiency and flight path cost, thereby allocating the initial operational area to the corresponding UAV based on the matching results.

[0127] S1.3: Based on the initial operating area corresponding to each UAV and combined with the preset crop growth status map, generate a complete flight plan, wherein the crop growth status map represents the crop growth activity index and is constructed based on historically collected crop data;

[0128] Specifically, after task allocation, to ensure that each UAV can execute precise operational strategies based on the differences in crop status within the target area during actual operations, path refinement is required based on the crop growth status map. The crop growth status map can be constructed from multi-temporal UAV remote sensing data or ground sensor feedback data. Response indicators in the crop growth status map may include the normalized vegetation index, spectral shift threshold, and crop growth rate gradient. The map is represented in the form of a three-dimensional heatmap. For each UAV's matched initial operational area, the corresponding response data matrix is ​​extracted from the crop growth status map as a reference for the flight mission's response characteristics.

[0129] See Figure 4 , Figure 4This is a schematic diagram illustrating the process of generating a complete flight plan for an embodiment of this application.

[0130] In one example, generating a complete flight plan by combining a preset crop growth status map includes:

[0131] S1.3.1: Extract the boundary information and terrain features of the crop area from the initial working area to determine the operating range and flight restrictions of the UAV;

[0132] Specifically, in order to achieve efficient crop coverage and path feasibility assessment in the target area, key spatial information related to the mission space needs to be extracted before the flight mission is generated.

[0133] In this embodiment, the boundary information includes the outer boundary contour of the initial working area and the distribution range of the internal crop planting area, which are parsed from the 3D model of the land parcel. The boundary information may include a set of vertex coordinates, a combination of boundary line segments, and closure verification logic. The extraction process of terrain features includes identifying slope information, elevation differences, local abrupt change areas, and obstacle-occupying areas from the point cloud model or structured light reconstruction model. By fusing the above boundary information and terrain features, a raster layer is formed for flight path constraints and flight passability judgment, and based on this, the UAV's flyable range, minimum flight altitude, flight buffer zone, and obstacle avoidance strategy applicable area are determined.

[0134] Understandably, the operational range is not limited to the flyable area; it also needs to be constrained by the crop's effective response range to avoid redundant paths in low-response, shaded, or unvegetated areas at the edge of farmland. Flight restrictions are overlaid on the operational map using a spatial mask to achieve real-time discrimination during path generation, ensuring that the final generated path does not encounter high obstacles, stays away from abrupt slope changes, and avoids edge flight error zones.

[0135] S1.3.2: Divide the crop area into regular grids, wherein the size of each grid is sufficient to meet the effective coverage area of ​​a single drone shot, and label the crop response weight associated with each grid through the crop growth status map, wherein the crop response weight is determined based on the growth status of the crop area in the crop growth status map;

[0136] Specifically, in order to achieve precise response and dynamic task weight management of crop growth during drone flight, it is necessary to divide the crop area into regular grids and assign quantifiable growth response indicators to each grid.

[0137] In this embodiment, after dividing the crop region into regular grids, to achieve a high level of precision in flight mission value assessment, it is necessary to establish an independent crop response weight for each grid. The crop response weight is not a direct pixel value or physical measurement value in the map, but a mission priority index that combines spatial status, growth trend, anomaly factors, and regional comparison.

[0138] Furthermore, the response data matrix corresponding to each raster cell in the crop growth status map is first extracted. The response data matrix contains multiple spectral channel information at the current time phase, as well as some optional environmental factor layers, including but not limited to temperature and humidity, vegetation height models, and historical drought records. The response data matrix is ​​mapped into the raster area in the form of a two-dimensional array.

[0139] Furthermore, the calculation of crop response weights is not simply done by using average or maximum values, but rather by constructing a regional sensing function. This regional sensing function considers the following four types of indicators as input variables:

[0140] The first category is the local response intensity index, which is the mean, peak value and standard deviation of the spectral image within the grid, reflecting the fluctuation range and density of vegetation status.

[0141] The second category is the trend factor, which calculates the response slope, inflection point and fluctuation frequency of the region over the past several periods based on historical multi-temporal maps, in order to identify the instability or potential risks of the growth state.

[0142] The third category is boundary neighborhood features, which introduces the response value gradient, boundary contrast and directional entropy between the grid and its neighboring grids to identify possible abnormal transition areas or disease spread edges.

[0143] The fourth category is regional importance factors, such as whether the grid is close to known traffic arteries, waterways, monitoring nodes, or high-contrast edge areas, serving as an external marker of management attention in actual deployment;

[0144] Furthermore, the crop response weight is calculated by weighting the aforementioned four types of indicators. Different weighting schemes can be selected for each type of indicator according to the type of operation. The specific weighting scheme and weight can be determined through a large number of experiments, which will not be elaborated here.

[0145] In some alternative implementations, the aforementioned four types of indicators can be normalized using a gradient-enhanced regional scoring logic. This enhances the ability to identify local fluctuations or anomalous hotspots through nonlinear function mapping, thereby forming a comprehensive assessment of the priority of operations for that grid. The final generated crop response weight is a quantitative indicator within a preset range.

[0146] S1.3.3: Extract the vertex set of the boundary of the regular grid, calculate the minimum bounding rectangle based on the vertex set, and take the long side direction of the rectangle as the working direction of the UAV;

[0147] Specifically, to construct an operational flight direction that closely matches the shape of the land parcel, it is necessary to extract effective geometric features from the boundary of the regular grid to determine the flight path direction that conforms to the principle of optimal shape adaptation. Traditional path planning often directly uses the orientation of the land parcel coordinate system as the default operational direction. However, this method often leads to problems such as an increased number of paths and insufficient edge cutting when the land parcel shape is oblique, strip-shaped, elongated, or has an irregular contour, affecting operational efficiency. This application constructs a minimum bounding rectangle to obtain the main directional features of the regular grid boundary in two-dimensional space. The construction of the minimum bounding rectangle is based on the boundary point set. The rotation caliper method or the orientation projection extreme value scanning method is used to traverse the boundary projection point set at multiple rotation angles, finally obtaining the envelope rectangle with the smallest area and its orientation angle.

[0148] In this embodiment, the long side of the smallest bounding rectangle is used as the working direction. This allows the flight path to extend to the maximum extent along the crop rows without changing the actual structure of the plot, thereby reducing the number of turns and shortening the flight path length in non-working sections. In addition, the working direction has strong spatial stability, avoiding directional shifts caused by map coordinate rotation or multi-temporal map resampling, and ensuring the consistency and generalization ability of path planning at different plot scales.

[0149] Furthermore, after the work direction is determined, a set of direction reference baselines can be generated as the input basis for subsequent algorithm modules such as path parallel generation, edge alignment control, and work overlap rate adjustment.

[0150] S1.3.4: Generate parallel paths along the stated work direction, and mark the start and end coordinates of each path, wherein:

[0151] The starting point of the path is determined based on the crop response weights of the rule grid;

[0152] The endpoint of the path is determined based on the drone's remaining battery power, wind direction, and return route.

[0153] Each parallel path has a different starting point;

[0154] Specifically, after obtaining the operational direction, the path generation process needs to unfold linearly along the predetermined direction to obtain a set of paths. To improve the uniformity of path distribution and operational efficiency, this application adopts a strategy of generating parallel paths at equal intervals along the operational direction. The spacing between each path is jointly determined by the UAV's operational width, image overlap requirements, and path coverage overlap control strategies. The path itself is a continuous line segment, and its starting and ending points need to be dynamically set based on multiple factors, avoiding a uniform direction or fixed-point takeoff method.

[0155] Regarding path start point selection, this application uses a gradient ranking logic based on crop response weights to perform crop response aggregation analysis on the grids covered by parallel paths. If the starting point of a path has a higher crop response, then that end is set as the starting point of the path, allowing the UAV to prioritize tasks in areas with higher response intensity, forming a task flow sequence from high-priority areas to low-priority areas. This effectively controls the priority response issue of task hotspots and reduces the omission of response areas or processing delays caused by sudden crop stress responses.

[0156] The determination of the destination path comprehensively considers the UAV's current battery level, total path length, the impact of wind speed and direction on operational speed, energy consumption differences caused by changes in altitude during flight, and the estimated energy consumption of the return flight path. Specifically, this includes flight energy consumption budgeting, performing energy consumption simulations and predictions for each path. If the prediction results indicate that the path cannot guarantee the safe return of the aircraft to the starting point or mission assembly point, the destination is pre-set, and path interruption points are tagged for subsequent mission inheritance scheduling. Wind direction has a significant impact at the end of the path; if the final segment of the flight path is opposite to the wind direction, the destination distance needs to be adjusted accordingly to adapt to flight control safety strategies.

[0157] Specifically, the flight control safety strategy can be understood as a set of decision-making and control criteria used to ensure flight safety and mission recoverability under the constraints of UAV trajectory planning and energy consumption. Its core objective is to ensure that the UAV has the ability to be controllable, returnable, interruptible, and recoverable when performing path tasks in complex environments (such as wind field interference, energy attenuation, and weather changes).

[0158] Furthermore, to prevent air resource congestion caused by overlapping paths or starting points of multiple UAVs during the initial takeoff phase, this application employs a staggered path start sequence or a staggered activation method to separate the execution timelines of adjacent paths. This facilitates the construction of a staggered path rhythm structure in high-density mission areas, improving the quality of track spatial distribution and execution fault tolerance from a scheduling perspective.

[0159] S1.3.5: Perform obstacle detection on the generated path. If there is an obstacle grid in the path, replan the path in the area corresponding to the obstacle grid using the A* algorithm, and reconnect the starting point and ending point of the path after bypassing the obstacle. If there is no obstacle grid in the path, add the path to the path set. Repeat the traversal of the path until all paths are added to the path set.

[0160] Specifically, the generated flight path, as the theoretical execution route, must undergo environmental constraint verification before being officially deployed to the UAV for execution. This verification mainly includes static obstacle detection and flight safety boundary review. This application uses elevation point cloud and structural texture information from the 3D model of the land parcel to extract obstacle regions, forming an obstacle grid mask. Spatial overlap calculations are then performed on the path grid set to identify whether any obstacle-crossing sections exist within the path segment. If the identification result is negative, meaning the entire path segment is within a passable area, the path is directly added to the path set; if the identification result is positive, path repair of the obstacle grid is required.

[0161] In this embodiment, the classic heuristic path planning algorithm A* is used as the core algorithm skeleton. The starting point and the ending point are defined as the source and target nodes of the heuristic search. The path search space is constructed with the path direction as a constraint, allowing the obstacle grid to be bypassed during the search process. The smoothness of the trajectory, the turning cost and the flight stability parameters are considered during the search process.

[0162] S1.3.6: Establish a multi-objective evaluation function, wherein the objectives include path length, number of turns, coverage repetition rate, energy consumption and crop response weight. Calculate the evaluation score of each path in the path set according to the multi-objective evaluation function, take the path corresponding to the highest evaluation score as the optimal path scheme, and generate the UAV flight route according to the path nodes, turning points and obstacle avoidance parameters of the optimal path scheme.

[0163] Specifically, to select the optimal path execution plan under the current operational conditions, it is necessary to quantitatively evaluate each path in the path set. This application uses a multi-objective path evaluation function, considering five key influencing factors:

[0164] First, route length, as an important indicator of flight time and operating costs, is better the shorter the route.

[0165] Second, the number of turns. More turns mean increased flight control complexity and energy consumption. A penalty factor is applied to continuous zigzag paths during the evaluation.

[0166] Third, cover duplication rate: by analyzing the overlap ratio between the route coverage area and historical operation records, resources are avoided from being wasted in duplicate areas;

[0167] Fourth, energy consumption budget, which combines path length, altitude changes and UAV parameters to simulate the total energy consumption of the path execution;

[0168] Fifth is the crop response weight, which is the average crop value density of the route coverage area, reflecting the operational priority of the route.

[0169] In this embodiment, the evaluation function does not use an equal weighting method, but instead constructs dynamic weights based on the task type. For example, if the task objective is rapid intervention in pests and diseases, the weight of the crop response weight factor is increased; if the task objective is full coverage detection, the weights of the coverage integrity and path smoothness factors are increased. After each path is evaluated by the evaluation function, a total score is output. The path with the highest score is selected as the priority execution path, and its corresponding path nodes, turning point information, and obstacle avoidance strategy parameters are extracted to form the optimal flight control command set for the UAV in the current operation cycle.

[0170] It is understood that if there are multiple drones to be operated in the initial operation area, the method further includes:

[0171] Obtain the path set and the evaluation scores corresponding to the paths in the path set, and prioritize the paths.

[0172] Based on the current positions of multiple drones awaiting operation, and in descending order of the path priority, the drone that meets the flight requirements of the current path and is closest to the starting point is selected as the execution drone for the current path, until there are no more drones awaiting operation.

[0173] Next, we will further elaborate on the adjustment phase of the method in this application.

[0174] In one example, the growth status information and UAV status information mentioned in S2 are derived from the real-time broadcast of the UAV.

[0175] It is understood that growth status information is collected through remote sensing devices carried by UAVs. These devices may include multispectral cameras, thermal infrared imagers, high-resolution visible light cameras, laser rangefinders, or narrowband filter components for vegetation reflectance monitoring. Different devices are selected based on the operational objectives. In this embodiment, normalized difference vegetation index (NDE), red edge index, and blue-green band reflectance are preferentially selected as the basic monitoring indicators for growth status information, used to determine the growth intensity, density changes, and stress response of the crop area.

[0176] In some optional implementations, vegetation elevation data and temperature layers can be combined for joint analysis to determine whether there is wilting, dehydration, or disease spread in a local area. After the aforementioned information is collected, it will be structured into growth status data blocks, each corresponding to a certain area below or around the current drone, and will be dynamically updated continuously with the flight trajectory to generate an approximately continuous crop growth status.

[0177] In one example, the conditions for entering the adjustment phase include at least one of the following:

[0178] The change in crop growth status indicators in any sub-region of the crop region is greater than or equal to a preset generation threshold.

[0179] Any sub-region within the crop region is found to have spatially clustered pests and diseases, or abnormal high or low humidity conditions. High temperature and low humidity can be determined by preset high temperature and low humidity thresholds. If the temperature of the sub-region is greater than or equal to the high temperature threshold, it is a high temperature abnormality area; if the humidity of the sub-region is less than or equal to the low humidity threshold, it is a low humidity abnormality area.

[0180] The generation threshold, high temperature threshold, and low humidity threshold can all be determined by those skilled in the art through a large number of experiments.

[0181] The drone's battery level is less than or equal to a preset flight safety threshold during operation;

[0182] The drone may experience deviation, mission interruption, or path execution timeout during the operation.

[0183] The flight paths of multiple drones intersect in space, and the safety interval is less than the preset tolerance threshold.

[0184] It is understood that the aforementioned thresholds are derived from experimental calibration, and this application will not elaborate on them here.

[0185] In one example, the specific steps of S3 are as follows:

[0186] S3.1: Construct a multi-dimensional dynamic task evaluation model for the area to be adjusted. The multi-dimensional dynamic task evaluation model includes the crop growth response dimension, the task urgency dimension, the historical task effect dimension, and the task migration cost dimension.

[0187] Specifically, during drone operations, certain areas may become areas requiring adjustment due to sudden changes in crop conditions, environmental disturbances, or changes in drone resources. To determine whether these areas need to be rescheduled and which drone should take over, a multi-dimensional dynamic task evaluation model needs to be established first to quantify the scheduling value of each area requiring adjustment.

[0188] In this embodiment, the multidimensional dynamic task evaluation model consists of four main dimensions. The crop growth response dimension measures the current agronomical importance of the region based on indicators such as NDVI, red-edge index, spectral anomaly gradient, or drought stress, and is a direct basis for determining whether an operation must be completed. The operation urgency dimension assesses the region's time sensitivity by combining crop cycle, number of delayed operations, and slope of response indicator changes; it is a scheduling priority dimension. The historical operation effectiveness dimension calculates the coverage of recent rounds of operations in the region, UAV trajectory density, and image acquisition completeness. If historical operations have records of insufficient coverage, trajectory deviation, or image acquisition failures, the region needs further subdivision of sub-tasks during reallocation. The task migration cost dimension measures the rationality of transferring the current task to other UAVs, mainly considering the UAV switching scheduling distance, path reconstruction complexity, and transfer cost from the operation interruption point to the adjacent UAV trajectory.

[0189] S3.2: Calculate the task migration evaluation score of the region to be adjusted based on the multidimensional dynamic task evaluation model. The task migration evaluation score is used to determine the task migration priority of the region to be adjusted.

[0190] The task migration evaluation score is calculated by weighting the aforementioned four dimensions. The essence of the task migration evaluation score is to calculate the comprehensive importance of each region in the question of whether it should be migrated, and then turn it into an ordered numerical result to support the downstream links of task sorting, resource allocation and path control.

[0191] S3.3: Generate a set of candidate UAV tasks based on the task migration evaluation score and the spatial distribution characteristics of the area to be adjusted;

[0192] Specifically, after calculating the task migration evaluation score, it is necessary to perform cluster analysis and task structure generation on high-scoring areas by considering factors such as the relative geographical location, shape boundaries, and connectivity of the regions. Directly assigning high-scoring areas to UAVs one by one may result in path fragmentation, flight route skipping, and a sharp increase in energy consumption. Therefore, it is necessary to construct a reasonable candidate task set, merging logically adjacent areas with similar operational characteristics and similar scheduling costs into a group of scheduling units for matching UAV resources.

[0193] In this embodiment, the candidate task set is generated based on the task migration evaluation score as the main ranking index, supplemented by spatial clustering for regional grouping. Common clustering algorithms can be selected for spatial clustering, which will not be elaborated here.

[0194] S3.4: Using the particle swarm optimization algorithm, based on the remaining battery power, flight distance, current load, historical operation efficiency and position status of the remaining drones, obtain the drones that will execute the drone candidate task set, and assign the corresponding tasks in the drone candidate task set to the corresponding drones;

[0195] Specifically, after obtaining the candidate drone task set, it is necessary to perform an optimal matching between drones and the candidate task set, taking into account the current schedulable drone resources. Due to the numerous dimensions of the matching objectives and the complex resource constraints, conventional greedy matching and local heuristic allocation methods struggle to simultaneously satisfy the triple requirements of matching efficiency, resource balance, and path executability. This embodiment employs a particle swarm optimization (PSO) algorithm, using drones and the candidate task set as particle matching units, and optimizing the scheduling score objective function through a population-parallel search strategy.

[0196] Understandably, each particle represents an allocation scheme, and the particle's dimension is determined by the number of candidate task sets, with each dimension representing the drone number assigned to that task set. The particle fitness function includes several scheduling metrics:

[0197] One aspect is the degree of matching between the drone's remaining battery power and the expected energy consumption for operation;

[0198] Secondly, there is the cost of the flight distance between the current position of the UAV and the centroid of the mission set;

[0199] Thirdly, the load ratio of the current drone load to the task workload;

[0200] Fourthly, the execution efficiency of historical tasks (such as coverage per unit time and path offset rate).

[0201] Fifthly, it involves estimating the continuity of path connectivity and the return cost.

[0202] By setting these fitness function components, the algorithm can continuously optimize the combination of UAV task allocation during the iteration process, and find the optimal matching structure that balances efficiency and stability.

[0203] Furthermore, after the allocation is completed, the feasibility of the optimal matching structure output by the particle swarm algorithm is reviewed. If a certain allocation scheme has problems such as the drone's carrying capacity exceeding the limit, overlapping paths, or unconnected flight paths, the scheme is rejected and the search direction is redirected. If there are multiple near-optimal solutions, a multi-branch decision model can be used to generate backup scheduling schemes, which facilitates rapid switching in case of task interruption or resource failure.

[0204] In an optional implementation, obtaining the drones that will perform the drone candidate task set includes:

[0205] Based on the preset resource fitness evaluation function in the particle swarm optimization algorithm, calculate the resource fitness of each UAV for executing the UAV candidate task set;

[0206] The resource fitness evaluation function can be determined by weighting and comprehensively considering the energy consumption, mission benefits, time constraints, payload status, and environmental factors when the UAV performs a mission.

[0207] Specifically, the resource fitness evaluation function can normalize indicators such as task completion rate, path energy consumption, flight time, payload utilization rate, and wind speed and direction, and then perform linear or nonlinear weighted calculations based on preset weights to obtain a resource fitness value that reflects the overall matching degree of the UAV's task execution.

[0208] Based on the resource adaptability, determine the initial task allocation scheme for the UAV;

[0209] Based on the conflict situation in the initial task allocation scheme of the UAV, the initial task allocation scheme of the UAV is re-allocated through the autonomous negotiation logic between UAVs. The conflict situation includes UAV path intersection, spatiotemporal overlap and unbalanced task load.

[0210] In one example, the autonomous negotiation logic includes conflict resolution between drones based on game-theoretic strategies, wherein:

[0211] Each drone broadcasts negotiation information to other drones based on its own status. The order in which drones perform tasks is updated based on the results of the negotiation information exchange, until a balance is reached where path conflicts between drones are minimized and the overall benefits of the task are maximized.

[0212] In this embodiment, the core idea of ​​the autonomous negotiation logic lies in introducing the Nash equilibrium game model from game theory as a framework for resolving task conflicts. Specifically, it is assumed that each drone is an independent participant, and its policy space is a set of optional task paths. The set of optional task paths consists of tasks other than the tasks that each drone must complete. The utility function is the task execution benefit minus the path conflict penalty factor and energy consumption prediction cost. Drones push their current path plans, state parameters, and priority execution information to the cluster network through a broadcast mechanism and receive similar information from other drones. After each round of information exchange, each drone calculates the conflict cost of its currently selected task path in the overall cluster based on the received path intersections, time overlap windows, and spatial adjacency relationships. It then readjusts its task selection based on the local optimal response function to minimize the total number of conflicts and improve its own utility.

[0213] Understandably, the negotiation process can be viewed as a discrete-time game, where each drone, based on its own strategy and collective feedback, continuously approaches the equilibrium state under a finite set of strategies. To prevent path oscillations or negotiation failures, a tolerance range is set for each iteration. When the change in task selection is lower than a preset disturbance threshold for two consecutive iterations, the negotiation is considered complete. If convergence is not achieved within a specified number of steps, a forced conflict resolution mechanism (such as time window staggered execution, Z-axis height staggered execution, etc.) is triggered for scheduling as a fallback.

[0214] Furthermore, after negotiation, the output is a set of UAV-task path mappings, where each path is the minimum conflict path collectively accepted by the cluster after negotiation, taking into account spatiotemporal staggering, resource balancing, and priority differences. Compared to a centralized matching strategy based solely on particle swarm optimization, this reduces the risks of path overlap, task crosstalk, and operation failure caused by multiple UAVs simultaneously entering adjacent high-response zones, and possesses collaborative adaptive capabilities under complex conditions such as stable local communication links and frequent updates to the task map.

[0215] In one example, this application provides a task allocation and conflict resolution system for multi-UAV collaborative operations, the system comprising:

[0216] The modeling module is used to acquire a 3D land parcel model of the target area and construct the work area based on terrain features, obstacle information, and crop growth status maps.

[0217] The path planning module is used to generate initial flight routes for multiple drones based on the operation area model.

[0218] The task allocation module is used to dynamically allocate tasks to flight paths based on crop response weights, task priorities, and UAV status parameters.

[0219] The conflict resolution module is used to perform path adjustment, time window misalignment, or flight altitude reconstruction when path conflicts exist, and output updated flight control commands.

[0220] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for task allocation and conflict resolution in multi-UAV collaborative operations, characterized in that, The method includes: A three-dimensional land parcel model of the target area is obtained, and the flight route of the UAV is planned through an optimization algorithm based on the terrain features and distribution information of the three-dimensional land parcel model. The flight routes of the drones are assigned to the corresponding drones. During the operation of the drones, the growth status information of the crop area and the status information of the drones are monitored to determine whether the conditions for entering the adjustment phase are met. If the conditions are met, the area to be adjusted is obtained from the drone flight path, and the drones that perform the corresponding tasks for the area to be adjusted are reassigned according to the crop growth status map of the area to be adjusted. The process of planning the drone flight path using an optimization algorithm includes: Based on the three-dimensional land parcel model and the number of drones to be operated, multiple initial operation areas are divided, wherein the multiple initial operation areas do not overlap with each other; Based on the priority, workload, and current status of the drones to be operated in each initial work area, the initial work areas are assigned to the corresponding drones. Based on the initial operating area corresponding to each UAV, and combined with the preset crop growth status map, a complete flight plan is generated. The crop growth status map represents the crop growth activity index and is constructed based on historically collected crop data. The process of generating a complete flight plan by combining a preset crop growth status map includes: Extract the boundary information and terrain features of the crop area from the initial working area to determine the operating range and flight restrictions of the UAV; The crop area is divided into regular grids, where the size of each grid is sufficient to cover the effective area captured by a single drone shot. The crop response weight associated with each grid is marked by the crop growth status map, and the crop response weight is determined based on the growth status of the crop area in the crop growth status map. Extract the vertex set of the boundary of the regular grid, calculate the minimum bounding rectangle based on the vertex set, and take the long side direction of the rectangle as the working direction of the UAV; Generate parallel paths along the work direction and mark the start and end coordinates of each path; The generated path is subjected to obstacle detection. If there is an obstacle grid in the path, the path is replanned in the area corresponding to the obstacle grid using the A* algorithm. The starting point and ending point of the path after bypassing the obstacle are reconnected. If there is no obstacle grid in the path, the path is added to the path set. The path is traversed repeatedly until all paths are added to the path set. A multi-objective evaluation function is established, wherein the objectives include path length, number of turns, coverage repetition rate, energy consumption, and crop response weight. The evaluation score of each path in the path set is calculated according to the multi-objective evaluation function. The path corresponding to the highest evaluation score is selected as the priority execution path. The UAV flight route is generated based on the path nodes, turning points, and obstacle avoidance parameters of the priority execution path. The drones assigned to perform the corresponding tasks in the area to be adjusted are reassigned based on the crop growth status map of the area to be adjusted, including: A multidimensional dynamic task evaluation model is constructed for the region to be adjusted. The multidimensional dynamic task evaluation model includes the crop growth response dimension, the task urgency dimension, the historical task effect dimension, and the task migration cost dimension. The task migration evaluation score of the region to be adjusted is calculated based on the multidimensional dynamic task evaluation model. The task migration evaluation score is used to determine the task migration priority of the region to be adjusted. Based on the task migration evaluation score and the spatial distribution characteristics of the area to be adjusted, a set of candidate UAV tasks is generated; Using the particle swarm optimization algorithm, based on the remaining battery power, flight distance, current load, historical operation efficiency, and location status of the remaining drones, drones that can execute the candidate task set are obtained, and the corresponding tasks in the candidate task set are assigned to the corresponding drones.

2. The task allocation and conflict resolution method for multi-UAV collaborative operations according to claim 1, characterized in that, Generate parallel paths along the stated work direction, and mark the start and end coordinates of each path, including: The starting point of the path is determined based on the crop response weights of the rule grid; The endpoint of the path is determined based on the drone's remaining battery power, wind direction, and return route. The starting points of each parallel path are different.

3. The task allocation and conflict resolution method for multi-UAV collaborative operations according to claim 1, characterized in that, If there are multiple drones to be operated in the initial operation area, the method further includes: Obtain the path set and the evaluation scores corresponding to the paths in the path set, and prioritize the paths. Based on the current positions of multiple drones awaiting operation, and in descending order of the path priority, the drone that meets the flight requirements of the current path and is closest to the starting point is selected as the execution drone for the current path, until there are no more drones awaiting operation.

4. The task allocation and conflict resolution method for multi-UAV collaborative operations according to claim 1, characterized in that, The conditions for entering the adjustment phase include at least one of the following: The change in crop growth status indicators in any sub-region of the crop region is greater than or equal to a preset generation threshold. If any sub-region of the crop region is found to have spatially clustered pests and diseases, high temperature or low humidity anomalies, the high temperature and low humidity can be determined by preset high temperature thresholds and low humidity thresholds. If the temperature of the sub-region is greater than or equal to the high temperature threshold, it is a high temperature anomaly; if the humidity of the sub-region is less than or equal to the low humidity threshold, it is a low humidity anomaly. The drone's battery level is less than or equal to a preset flight safety threshold during operation; The drone may experience deviation, mission interruption, or path execution timeout during the operation. The flight paths of multiple drones intersect in space, and the safety interval is less than the preset tolerance threshold.

5. The task allocation and conflict resolution method for multi-UAV collaborative operations according to claim 1, characterized in that, The drones that obtain the drone candidate task set include: Based on the preset resource fitness evaluation function in the particle swarm optimization algorithm, calculate the resource fitness of each UAV for executing the UAV candidate task set; Based on the resource adaptability, determine the initial task allocation scheme for the UAV; Based on the conflict situation in the initial task allocation scheme of the UAV, the initial task allocation scheme of the UAV is re-allocated through the autonomous negotiation logic between UAVs. The conflict situation includes UAV path intersection, spatiotemporal overlap and unbalanced task load.

6. The task allocation and conflict resolution method for multi-UAV collaborative operations according to claim 5, characterized in that, The autonomous negotiation logic includes conflict resolution between drones based on game theory strategies, wherein: Each drone broadcasts negotiation information to other drones based on its own status. The order in which drones perform tasks is updated based on the results of the negotiation information exchange, until a balance is reached where path conflicts between drones are minimized and the overall benefits of the task are maximized.

7. A task allocation and conflict resolution system for multi-UAV collaborative operations, used to implement the task allocation and conflict resolution method for multi-UAV collaborative operations as described in any one of claims 1-6, characterized in that, The system includes: The modeling module is used to acquire a 3D land parcel model of the target area and construct the work area based on terrain features, obstacle information, and crop growth status maps. The path planning module is used to generate initial flight routes for multiple drones based on the operation area model. The task allocation module is used to dynamically allocate tasks to flight paths based on crop response weights, task priorities, and UAV status parameters. The conflict resolution module is used to perform path adjustment, time window misalignment, or flight altitude reconstruction when path conflicts exist, and output updated flight control commands.

Images