An unmanned aerial vehicle swarm cooperative path planning method, system, product and medium

Abstract

The application discloses a method, system, product and medium for path planning of a UAV group, and relates to the field of non-electric variable control or regulation systems.The method comprises the following steps: collecting environmental data such as terrain height, obstacle distribution and dynamic targets within the detection range of a UAV, converting the environmental data into a grid cell space occupation state, dividing a search space into three-dimensional grids, calculating the unexplored probability value of each cell, extracting the time dimension change data and the gradient information between cells, identifying the dynamic change area of the environment, determining the boundary features and generating path node data.The node data is compressed into lightweight path information containing coordinates, priority and time sequence constraints, and a flight path sequence is generated in combination with the position of the UAV and is distributed.The method can improve the detection efficiency of the UAV.

Description

Technical Field

[0001] This application relates to the field of control or regulation systems for non-electrical variables, and more particularly to a method, system, product, and medium for collaborative path planning of unmanned aerial vehicle (UAV) swarms. Background Technology

[0002] With the increasing demand for large-scale environmental sensing in fields such as natural disaster relief and urban security monitoring, utilizing multiple drones in collaboration to complete complex environmental detection tasks has become an important development trend. Drone swarm collaborative detection can significantly improve the efficiency and coverage of environmental information acquisition, which is of great significance for enhancing the ability to respond to emergencies.

[0003] Existing technologies primarily employ path planning based on preset waypoints for collaborative UAV swarm reconnaissance. This approach first divides the mission area into multiple sub-regions and assigns a fixed sequence of waypoints to each UAV. Each UAV completes its regional reconnaissance task according to the preset waypoints and transmits the collected environmental data to a ground control station for processing and analysis via a communication network. The ground control station then determines whether manual intervention to adjust the waypoint settings is necessary based on the received data.

[0004] However, in complex and dynamic environments, environmental features such as terrain and obstacles often exhibit continuous changes. Pre-set waypoints require frequent manual path adjustments, and the static waypoint allocation method makes it difficult for UAVs to flexibly adjust their detection strategies according to real-time environmental changes. At the same time, the transmission and processing of environmental data can also cause system response delays, affecting detection efficiency. Summary of the Invention

[0005] This application provides a method, system, product, and medium for collaborative path planning of UAV swarms, which can improve the detection efficiency of UAVs.

[0006] Firstly, this application provides a method for collaborative path planning of UAV swarms, applied to a collaborative path planning system for UAV swarms. The method includes: collecting environmental data within the detection range of each UAV; determining the spatial occupancy status of multiple grid cells in a probability map based on the environmental data, including terrain height data, obstacle distribution data, and dynamic target data; dividing the search space into three-dimensional grid cells; calculating the unexplored probability value of each grid cell based on the spatial occupancy status, where the unexplored probability value characterizes the possibility of unsearched areas existing within each grid cell; extracting the change data of the unexplored probability value of the grid cells in the time dimension; calculating the probability difference between the unexplored probability values ​​of grid cells at adjacent time points; identifying dynamically changing areas of the environment based on the change data; extracting gradient information of the unexplored probability values ​​between grid cells; determining the boundary features of the environmental change areas based on the probability difference and gradient information; and generating path node data based on the boundary features; compressing the path node data into lightweight path information, which includes node coordinates, node priority calculated based on probability distribution, and temporal constraint relationships determined based on environmental change trends; generating a trajectory sequence that satisfies the temporal constraints based on the lightweight path information and the current position information of each UAV; and distributing the trajectory sequence to the corresponding UAVs.

[0007] In the above embodiments, unexplored probability value changes and gradient information of grid cells are extracted based on environmental data to identify dynamically changing environmental regions and determine boundary features. The generated lightweight path information includes node priorities and temporal constraints, guiding the UAV swarm to adaptively adjust its detection strategy according to environmental change trends. The UAVs can dynamically plan routes based on real-time environmental characteristics, eliminating the delay of manual path adjustments and improving the system's response speed and detection efficiency to environmental changes.

[0008] In conjunction with some embodiments of the first aspect, in some embodiments, the step of collecting environmental data within the detection range of each UAV, converting the environmental data into the spatial occupancy status of grid cells, and establishing a probabilistic map based on the spatial occupancy status specifically includes: receiving ranging data and image data collected by the lidar and camera carried by each UAV; performing noise filtering on the ranging data to generate three-dimensional point cloud data; performing coordinate alignment and stitching on the three-dimensional point cloud data, extracting ground points and obstacle points from the point cloud data to obtain terrain height data and obstacle distribution data; comparing the differences in target positions in images at different times and marking moving targets in the images to obtain dynamic target data; mapping the terrain height data, obstacle distribution data, and dynamic target data onto a three-dimensional grid map to obtain the spatial occupancy status of each grid cell; and generating a probabilistic map based on the spatial occupancy status of each grid cell.

[0009] In the above embodiments, noise filtering was applied to the lidar ranging data to generate a 3D point cloud. Moving targets were then identified by combining this data with image data, establishing a complete environmental representation model. Terrain, obstacle, and dynamic target data were uniformly mapped onto a grid map, achieving the fusion of multi-source environmental information. The establishment of the probabilistic map provides an accurate description of the environmental structure for subsequent path planning, enhancing the UAV swarm's perception and understanding of complex environments.

[0010] In conjunction with some embodiments of the first aspect, in some embodiments, the steps of extracting the change data of the unexplored probability value of grid cells in the time dimension, calculating the probability difference of the unexplored probability value of grid cells at adjacent time points, and identifying the dynamic change area of ​​the environment based on the change data specifically include: collecting and recording the unexplored probability value sequence of each grid cell within a set time window, calculating the difference of the unexplored probability value of grid cells at adjacent sampling time points within the time window; filtering out grid cells whose probability value changes exceed a preset change threshold, and counting the change frequency of each significantly changing grid cell within the time window; clustering grid cells based on spatial proximity and change pattern similarity, extracting the spatial range and change characteristics of each clustered region, and outputting the dynamic change area of ​​the environment marked with the change frequency.

[0011] In the above embodiments, probability value sequences are recorded and differences are calculated within a time window to filter out grid cells with significant changes. Clustering is performed based on spatial proximity and similarity of change patterns to extract the spatiotemporal characteristics of environmental change areas. The output change frequency information quantitatively describes the dynamic characteristics of the environment, providing a quantitative basis for path planning and optimizing the detection strategy of UAV swarms in dynamic areas.

[0012] In conjunction with some embodiments of the first aspect, in some embodiments, the step of extracting the gradient information of unexplored probability values ​​between grid cells, determining the boundary features of the environmental change area based on the probability value difference and gradient information, and generating path node data based on the boundary features specifically includes: calculating the gradient of unexplored probability values ​​between adjacent grid cells, extracting grid cells with abrupt probability value changes based on the gradient magnitude; connecting spatially adjacent high-gradient grid cells to construct a boundary contour, performing feature analysis on the boundary contour and extracting key nodes, calculating the environmental change intensity at each key node; generating priorities based on node positions and change intensity, and outputting path node data containing node positions and priorities.

[0013] In the above embodiments, environmental boundary features are extracted using probability value gradients to construct continuous contours of high-gradient units, calculate the intensity of environmental changes at key nodes, and generate path priorities. Based on the quantitative analysis of gradient information and change intensity, an automatic mapping mechanism from environmental features to path nodes is formed, reducing the workload of manual path point setting and improving the sensitivity and adaptability of path planning to environmental changes.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, the step of generating a trajectory sequence that satisfies temporal constraints based on lightweight path information and the current position information of each UAV, and distributing the trajectory sequence to the corresponding UAV, specifically includes: obtaining the real-time position and motion state parameters of each UAV, calculating the shortest travel time from each UAV to each path node; assigning exploration tasks to the UAVs based on travel time and node priority, generating an initial trajectory that satisfies the task assignment, verifying the feasibility of the trajectory based on the UAV performance parameters; optimizing the trajectory segments in the initial trajectory that do not meet the constraints, establishing a multi-UAV trajectory temporal relationship and generating a trajectory sequence, and distributing the trajectory sequence to the corresponding UAVs.

[0015] In the above embodiments, the flight time between nodes is calculated based on the real-time status of the UAV, and task allocation is performed in combination with node priority to verify the feasibility of the flight path. A multi-UAV time-series constraint optimization flight path sequence is established. Through the unified processing of flight path planning and task allocation, a collaborative planning framework that takes into account both task efficiency and aircraft performance is established, effectively improving the overall detection efficiency and task completion quality of the UAV swarm.

[0016] In conjunction with some embodiments of the first aspect, in some embodiments, after generating a trajectory sequence that satisfies timing constraints based on lightweight path information and the current position information of each UAV, and distributing the trajectory sequence to the corresponding UAV, the method further includes: acquiring dynamic obstacle feature information in environmental data, updating the obstacle avoidance strategy of the trajectory sequence according to the feature information, and obtaining an adjusted trajectory sequence; generating an optimized trajectory sequence based on the obstacle avoidance strategy, determining the execution state of the optimized trajectory sequence, and responding to the execution state to implement task control, and distributing the adjusted trajectory sequence to the corresponding UAV for execution.

[0017] In the above embodiments, dynamic obstacle features and obstacle avoidance strategies are updated in real time, flight path sequences are optimized, and execution status is monitored to achieve dynamic adjustments to the mission process. A closed-loop control mechanism from environmental perception to flight path optimization is established to ensure the continuous detection capability and mission execution stability of the UAV swarm in dynamic environments.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, after generating a flight path sequence that satisfies the time constraints based on lightweight path information and the current position information of each UAV, and distributing the flight path sequence to the corresponding UAV, the method further includes: obtaining the remaining power status parameters of each UAV, calculating the available flight time corresponding to the remaining power status parameters; determining the task allocation weight based on the available flight time, adjusting the UAV flight parameters according to the task allocation weight, and generating a return path plan that satisfies the power constraints.

[0019] In the above embodiments, the available flight time is calculated based on the remaining battery status parameters. The task allocation weights and flight parameters are adjusted according to the flight time to generate a return path that meets the battery constraints. This establishes a proactive management mechanism from battery monitoring to task adjustment, integrating battery constraints into the task planning process. This ensures the continuous operation capability and safety of the UAV swarm during long-duration reconnaissance missions, avoiding mission interruptions or equipment losses due to battery depletion.

[0020] In a second aspect, embodiments of this application provide a collaborative path planning system for a swarm of unmanned aerial vehicles (UAVs). The UAV swarm collaborative path planning system includes: one or more processors and a memory; the memory is coupled to the one or more processors and is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the UAV swarm collaborative path planning system to perform the method described in the first aspect and any possible implementation thereof.

[0021] Thirdly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on a drone swarm cooperative path planning system, cause the drone swarm cooperative path planning system to perform the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a UAV swarm cooperative path planning system, cause the UAV swarm cooperative path planning system to perform the method described in the first aspect and any possible implementation thereof.

[0023] Understandably, the UAV swarm cooperative path planning system provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the methods provided in the embodiments of this application. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0024] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0025] 1. This application constructs a probabilistic map based on environmental data, extracts the unexplored probability value changes and gradient information of grid cells, identifies dynamically changing environmental regions, and determines boundary features. The generated lightweight path information includes node priorities and temporal constraints, guiding the UAV swarm to adaptively adjust its detection strategy according to environmental change trends. The UAVs can dynamically plan their routes based on real-time environmental characteristics, eliminating the delay of manual path adjustments and improving the system's response speed and detection efficiency to environmental changes.

[0026] 2. This application establishes a complete environmental representation model by noise filtering of lidar ranging data and generating a 3D point cloud, combined with image data to identify moving targets. Terrain, obstacle, and dynamic target data are uniformly mapped onto a grid map, achieving the fusion of multi-source environmental information. The establishment of the probabilistic map provides an accurate environmental structure description for subsequent path planning, enhancing the UAV swarm's perception and understanding of complex environments.

[0027] 3. This application identifies grid cells with significant changes by recording probability value sequences and calculating differences within a time window. Clustering is performed based on spatial proximity and similarity of change patterns to extract the spatiotemporal characteristics of environmental change regions. The output change frequency information quantitatively describes the dynamic characteristics of the environment, providing a quantitative basis for path planning and optimizing the detection strategy of UAV swarms in dynamic regions. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating the UAV swarm cooperative path planning method in the embodiments of this application;

[0029] Figure 2 This is another flowchart illustrating the UAV swarm cooperative path planning method in the embodiments of this application;

[0030] Figure 3 This is a schematic diagram of the physical device structure of a drone swarm collaborative path planning system in the embodiments of this application. Detailed Implementation

[0031] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0032] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0033] To facilitate understanding, the application scenarios of the embodiments of this application are described below.

[0034] In large-scale natural disaster relief operations, search and rescue teams need to quickly obtain dynamic information about the disaster area. Traditional single-drone reconnaissance methods have limited coverage and cannot respond promptly to complex and ever-changing terrain environments. For example, at a landslide site, areas with a continuous risk of collapse need to be monitored closely while simultaneously searching for potentially trapped individuals. Terrain elevation, vegetation cover, and weather conditions are constantly changing, posing significant challenges to search and rescue planning. Therefore, there is an urgent need for a technological solution that can coordinate multiple drones, adapt to environmental changes, and optimize search paths in real time. The key issue is how to enable drone swarms to dynamically adjust their search strategies based on real-time environmental perception data and maximize search efficiency within limited flight time.

[0035] Existing drone swarm collaboration technologies primarily rely on pre-set flight paths. For example, in a forest fire monitoring mission, multiple drones are assigned to different areas to perform fixed-path patrols. Each drone flies independently, acquiring only local environmental information. When a new fire is detected, manual intervention is required to replan the flight path. This approach has significant drawbacks: first, pre-set flight paths cannot adapt to dynamically changing environments, such as changes in no-fly zones due to fire spread; second, the lack of information sharing and mission coordination mechanisms among drones leads to overlapping coverage or missed detections; and third, it fails to consider resource constraints such as battery life, easily causing mission interruptions. These problems severely impact the search efficiency and mission completion quality of the drone swarm.

[0036] By employing the path planning method of this invention, multiple drones can work efficiently and collaboratively in a search and rescue mission in an earthquake-stricken area. The system first acquires environmental data using lidar and cameras to establish a precise 3D grid probabilistic map. When a drone detects new signs of building collapse, this information is updated in real time to the probabilistic map. The system automatically calculates the boundary features of the environmentally changed area and generates new path nodes. Other drones automatically adjust their flight paths based on node priority and their own location, focusing on high-risk areas. Simultaneously, the system considers the remaining battery power of each drone to ensure its safe return. For example, when new displacement occurs in area A, the system prioritizes dispatching the nearest drone with sufficient battery power to investigate, while other drones continue monitoring the original designated area. This intelligent collaborative approach significantly improves search and rescue efficiency and avoids delays caused by manual intervention.

[0037] By employing the path planning method of this invention, multiple drones can work efficiently and collaboratively in a search and rescue mission in an earthquake-stricken area. The system first acquires environmental data using lidar and cameras to establish a precise 3D grid probabilistic map. When a drone detects new signs of building collapse, this information is updated in real time to the probabilistic map. The system automatically calculates the boundary features of the environmentally changed area and generates new path nodes. Other drones automatically adjust their flight paths based on node priority and their own location, focusing on high-risk areas. Simultaneously, the system considers the remaining battery power of each drone to ensure its safe return. For example, when new displacement occurs in area A, the system prioritizes dispatching the nearest drone with sufficient battery power to investigate, while other drones continue monitoring the original designated area. This intelligent collaborative approach significantly improves search and rescue efficiency and avoids delays caused by manual intervention.

[0038] To facilitate understanding, the method provided in this implementation will be described in detail below, using the above scenario as an example. Please refer to [link / reference]. Figure 1 This is a flowchart illustrating the UAV swarm cooperative path planning method in an embodiment of this application.

[0039] S101. Collect environmental data within the detection range of each UAV, and determine the spatial occupancy status of multiple grid cells in the probability map based on the environmental data. The environmental data includes terrain height data, obstacle distribution data, and dynamic target data.

[0040] Among them, environmental data represents the surrounding environment information acquired by the UAV during the detection process, including digital representations of elements such as terrain, obstacles, and dynamic targets; spatial occupancy status refers to the state description of whether a specific grid cell is occupied by environmental elements, usually represented by a value between 0 and 1; probabilistic map represents a spatial distribution map of environmental data in probabilistic form, used to quantify the uncertainty of environmental characteristics. Terrain height data refers to the elevation information of ground undulations; obstacle distribution data represents the spatial location information of static obstacles; dynamic target data refers to time-varying characteristic data such as the position and velocity of moving objects.

[0041] This step is performed when the UAV begins its reconnaissance mission and is used to establish an initial probabilistic model of the environment. Specifically, the system first acquires raw environmental data within the detection range using sensors on each UAV, including ranging data from lidar and image data from cameras. This data is then transformed into a unified three-dimensional spatial representation through coordinate transformation and data fusion. For each grid cell, the corresponding spatial occupancy probability value is calculated based on the characteristics of the environmental elements it contains. Finally, based on the occupancy status of all grid cells, a complete environmental probabilistic map is constructed.

[0042] In some embodiments, environmental data acquisition and probabilistic map construction can be achieved in multiple ways: Optionally, LiDAR can be used for environmental scanning to acquire point cloud data, which can then be denoised and registered to extract ground points and obstacle points. The spatial distribution characteristics of each point can be calculated, and finally, the grid occupancy probability can be estimated based on the point density distribution. Optionally, a binocular camera can be used to acquire environmental images, and a 3D scene can be reconstructed using stereo vision algorithms. Target objects in the images can be identified, and the motion trajectory of dynamic targets can be tracked. The grid occupancy probability can be calculated by combining the spatial location and motion characteristics of the targets. It is understood that other sensors and processing methods can also be used to achieve environmental perception and probabilistic modeling, which are not limited here.

[0043] S102. Divide the search space into three-dimensional grid cells, and calculate the unexplored probability value of each grid cell based on the space occupancy status. This unexplored probability value represents the possibility that there is an unsearched area within each grid cell.

[0044] Among them, the search space refers to the three-dimensional spatial region that needs to be explored; the three-dimensional grid cell refers to the cubic unit formed by uniformly dividing the search space; the unexplored probability value represents the probability that a specific grid cell has not yet been explored, and the value ranges from 0 to 1; the unsearched area refers to the spatial region that has not yet been effectively explored by the UAV.

[0045] This step is performed after the environmental probabilistic map is constructed, and it is used to quantitatively describe the distribution of the area to be explored. Specifically, the entire task space is first divided into three-dimensional grid cells of equal size. For each grid cell, based on its current spatial occupancy, combined with the effective detection range of the detection sensors and historical detection records, the probability value of that cell not being explored is calculated. The higher this probability value, the more likely that there is important undetected information in that area.

[0046] In some embodiments, the unexplored probability value can be calculated in several ways: Optionally, a probability calculation method based on a sensor detection model can be established, considering the sensor's detection range, resolution, and noise characteristics, and combining factors such as detection distance and incident angle, to calculate the effective detection probability of each grid cell, and subtract the effective detection probability from 1 to obtain the unexplored probability value; Optionally, a probability calculation method based on information entropy can be used to analyze the information uncertainty of the grid cell, consider the frequency of environmental changes and the detection time interval, evaluate the timeliness of the information, and comprehensively determine the unexplored probability value. It is understood that other probability calculation methods can also be used to quantify the distribution characteristics of unexplored areas, which are not limited here.

[0047] S103. Extract the change data of the unexplored probability value of the grid cell in the time dimension, calculate the probability difference of the unexplored probability value of the grid cell at adjacent time points, and identify the dynamic change area of ​​the environment based on the change data.

[0048] Among them, the time dimension variation data represents the historical sequence of the unexplored probability value of a grid cell changing over time; the probability value difference refers to the numerical difference in the unexplored probability value of the same grid cell at two adjacent time points; the dynamic change region represents the spatial range in which environmental characteristics have changed significantly. Adjacent time points refer to two consecutive data sampling time points; the variation data is used to represent a quantitative description of the dynamic characteristics of the environment.

[0049] This step is performed after acquiring environmental data from multiple consecutive time points to identify dynamic changes in the environment. Specifically, the system first establishes a time-series database, recording the unexplored probability value of each grid cell at different times. For each grid cell, the difference in probability values ​​between adjacent sampling times is calculated, forming a probability value change sequence. By analyzing the magnitude and frequency of these probability value changes, regions in the environment that have undergone significant changes are identified. These regions may contain important dynamic targets or information about environmental changes and require focused attention.

[0050] In some embodiments, the identification of dynamically changing environmental regions can be achieved in several ways: Optionally, a sliding time window method can be used, setting a fixed-size observation window, calculating the statistical characteristics of probability values ​​within the window, including mean, variance, and rate of change, and determining regions with significant changes based on thresholds of these characteristics. Finally, the judgment results of multiple consecutive time windows are fused to obtain a stable dynamic region identification result. Optionally, a time series pattern analysis method can be used, performing a Fourier transform on the probability value change sequence to extract the frequency characteristics of the changes, calculating the energy distribution of different frequency components, identifying regions with periodic and sudden changes based on the energy distribution characteristics, and finally combining spatial correlation analysis to determine the complete dynamically changing region. It is understood that other time series analysis methods can also be used to identify dynamic environmental characteristics, which are not limited here.

[0051] S104. Extract the gradient information of unexplored probability values ​​between grid cells, determine the boundary features of the environmental change area based on the probability value difference and gradient information, and generate path node data based on the boundary features.

[0052] Among them, gradient information represents the rate of change of the unexplored probability value between adjacent grid cells; boundary features refer to the spatial contour and morphological characteristics of dynamically changing regions; path node data is used to represent critical path point information when planning routes. Probability value difference refers to the amount of probability change at different times; boundary features include geometric features such as the position, direction, and curvature of the boundary.

[0053] This step is executed after identifying dynamically changing regions to determine their specific extent and generate probe waypoints. Specifically, it first calculates the gradient of unexplored probability values ​​between adjacent grid cells in space to obtain the spatial variation characteristics of these probability values. Combined with the probability value differences over time, it analyzes the spatial distribution patterns of environmental changes. By connecting grid cells with larger probability gradients, it constructs the boundary contour of the changing region. Finally, based on the geometric features of the boundary contour, key locations are selected as path nodes for subsequent route planning.

[0054] In some embodiments, boundary feature extraction and path node generation can be achieved in multiple ways: Optionally, morphological processing methods can be used to perform edge detection and contour extraction on the probability gradient image, optimize the continuity of the boundary contour through morphological operations, calculate the curvature and inflection point features of the contour, select path nodes at the locations of significant features, and finally optimize the spatial uniformity of the nodes to ensure a reasonable distribution of path points; Optionally, a region growing algorithm can be used, starting from seed points with significant changes in probability values, continuously expanding the region range through similarity criteria, extracting the complete boundary of the changing region, calculating the main direction and branch points of the boundary, determining the path node positions based on the spatial distribution of these feature points, and finally deleting redundant nodes and fine-tuning their positions. It is understood that other boundary detection and feature extraction methods can also be used to generate path nodes, which are not limited here.

[0055] S105. Compress the path node data into lightweight path information, which includes node coordinates, node priorities calculated based on probability distribution, and temporal constraint relationships determined based on environmental change trends.

[0056] Among them, lightweight path information represents a simplified path description after data compression and optimization; node coordinates refer to the three-dimensional spatial location information of key points on the path; node priority represents the importance of path node detection; temporal constraint relationship refers to the access order requirements between path nodes; probability distribution refers to the spatial distribution characteristics of the probability values ​​of unexplored grid cells; environmental change trend represents the evolution law of dynamic region characteristics over time; data compression refers to the process of reducing data redundancy while maintaining key information.

[0057] This step, performed after generating the initial path nodes, optimizes the data structure for the path description. Specifically, it first performs spatial clustering and filtering on the original path nodes, removing redundant nodes and retaining representative key nodes. Then, based on the unexplored probability distribution characteristics of the grid cell where the node resides, the detection priority of each node is calculated. Simultaneously, the temporal characteristics of environmental changes are analyzed to determine the access timing constraints between nodes. Finally, the node coordinates, priorities, and timing constraints are integrated into a unified, lightweight data structure.

[0058] In some embodiments, lightweight processing of path information can be achieved in several ways: Optionally, a spatial feature-based compression method can be used. First, the distance and direction features between adjacent path nodes are calculated. Nodes that are close together and have similar directions are merged. Then, a multi-level grid structure is used to spatially index the nodes. Finally, the compression ratio is determined based on the environmental complexity of the region where the node is located, generating a simplified path description. Optionally, an importance-based compression method can be used. First, a node importance evaluation model is established, comprehensively considering factors such as the probability of unexplored areas, the frequency of environmental changes, and location coverage. The importance score of each node is calculated. Then, the nodes are sorted and filtered according to the scores. Finally, the spatial distribution of the retained nodes is optimized to ensure the detection effect in key areas. It is understood that other data compression and optimization methods can also be used to achieve lightweight representation of path information, which are not limited here.

[0059] S106. Generate a trajectory sequence that meets the time constraints based on the lightweight path information and the current position information of each UAV, and distribute the trajectory sequence to the corresponding UAV.

[0060] The flight path sequence represents the complete flight path of the UAV performing the reconnaissance mission; the timing constraint refers to the access time requirements of the path nodes; and the current location information represents the real-time spatial coordinates and attitude data of each UAV. Flight path distribution refers to sending the planned flight path information to the corresponding UAV; satisfying the timing constraint means that flight path planning needs to consider the order in which nodes are accessed.

[0061] This step is executed after the path information is lightweighted, and is used to generate specific UAV flight tracks. Specifically, first, the real-time position and status information of each UAV is acquired. Combined with the node distribution in the lightweight path information, the flight cost from each UAV to each path node is calculated. Then, based on node priority and timing constraints, appropriate detection tasks are assigned to each UAV to generate an initial track. The generated track is then verified for feasibility and optimized to ensure that it meets UAV performance constraints. Finally, the optimized track sequence is distributed to each UAV for execution.

[0062] In some embodiments, trajectory planning and distribution can be implemented in several ways: Optionally, a hierarchical planning method can be used. First, a coarse trajectory plan is performed at the global level based on task priority and timing constraints to determine the approximate task area for each UAV. Then, at the local level, the motion characteristics and energy constraints of the UAVs are considered to refine the specific path points of the trajectory. Finally, a collaborative mechanism among multiple UAVs is established to ensure that conflicts are avoided during trajectory execution. Optionally, a dynamic programming method can be used. First, a path cost model considering multiple constraints is established, including factors such as flight distance, time window, and energy consumption. Then, a heuristic search algorithm is used to find the optimal trajectory that satisfies the constraints. Finally, the trajectory parameters are dynamically adjusted according to environmental changes to achieve real-time path optimization. It is understood that other path planning and task allocation methods can also be used to generate and distribute trajectories, which are not limited here.

[0063] The following provides a more detailed description of the process of the method provided in this implementation. Please refer to [link / reference]. Figure 2 This is another flowchart illustrating the UAV swarm cooperative path planning method in this application embodiment.

[0064] S201: Receive ranging data and image data collected by the lidar and camera carried by each UAV; perform noise filtering on the ranging data to generate three-dimensional point cloud data.

[0065] Among them, lidar ranging data represents the distance information of spatial points obtained using the principle of laser ranging; image data refers to the two-dimensional images of the environment captured by the camera; noise filtering refers to removing outliers and interference information from the ranging data; and three-dimensional point cloud data represents a three-dimensional scene composed of a large number of spatial scattered points.

[0066] During the environmental data acquisition phase, each UAV uses its onboard LiDAR to emit laser beams to scan the environment, recording the time difference between emission and reception to calculate target distances and simultaneously acquiring the angle information of reflection points, thus obtaining the three-dimensional coordinates of spatial points. Due to the influence of atmospheric conditions and target surface characteristics on the ranging process, noisy points exist in the raw data. Statistical analysis is performed on this data; outliers are removed by setting distance and density thresholds, and then Gaussian filtering or median filtering is used to smooth the data, ultimately generating a three-dimensional point cloud representing environmental characteristics. Each point in the point cloud contains spatial location and reflection intensity information for subsequent scene reconstruction.

[0067] S202. Align and stitch the 3D point cloud data according to coordinates, extract the ground points and obstacle points from the point cloud data, and obtain the terrain height data and obstacle distribution data.

[0068] Among them, coordinate alignment refers to converting point cloud data acquired from different locations to a unified coordinate system; point cloud stitching means merging multiple point cloud datasets into a complete scene description; ground points refer to the set of spatial points belonging to the ground; obstacle points refer to the set of spatial points belonging to obstacles.

[0069] When processing point cloud data collected from multiple locations, the first step is to extract feature points from the point clouds, calculate the matching relationships between these feature points, and establish a rigid body transformation matrix to achieve precise alignment between different point clouds. Then, the aligned point cloud data are stitched together into a unified dataset. The complete point cloud is then segmented into a ground plane, and the RANSAC algorithm is used to fit the ground plane, obtaining a set of ground points and extracting terrain height information. The remaining non-ground points are used to identify independent obstacles through cluster analysis, recording the obstacle's location, size, and other feature parameters to construct an obstacle distribution map.

[0070] S203. Compare the differences in target positions in images at different times and mark moving targets in the images to obtain dynamic target data; map terrain height data, obstacle distribution data and dynamic target data onto a three-dimensional grid map to obtain the spatial occupancy status of each grid cell.

[0071] Among them, target position difference refers to the change in position of the same target in images at different times; dynamic target data represents the position and velocity information of moving objects; space occupancy status refers to the state quantity of whether environmental elements exist in the grid cell.

[0072] The system processes continuously acquired image sequences, tracking the motion of targets within the images through methods such as feature matching and optical flow calculation. The displacement of the target on the image plane is calculated, and combined with camera parameters, the planar displacement is converted into actual spatial displacement to obtain the target's trajectory. The acquired terrain, obstacle, and dynamic target data are uniformly mapped onto a 3D mesh map. For each mesh cell, an occupancy status value is calculated based on the type and quantity of environmental features it contains. The occupancy status can be binary (occupied / unoccupied) or probabilistic (occupancy probability between 0 and 1), used to characterize the spatial structure features of the environment.

[0073] S204. Generate a probability map based on the spatial occupancy status of each grid cell. This environmental data includes terrain height data, obstacle distribution data, and dynamic target data.

[0074] Among them, the probability map represents the probability distribution model of environmental features; the space occupancy status refers to the existence status of environmental elements in the grid cell; the terrain height data represents the numerical description of ground undulation; the obstacle distribution data refers to the spatial location information of static obstacles; and the dynamic target data represents the spatiotemporal characteristics of moving objects.

[0075] When converting the spatial occupancy status of each grid cell into a probabilistic representation, it is necessary to comprehensively consider the characteristics of different types of environmental data. For terrain height data, the terrain probability is calculated based on the continuity of height values; for obstacle data, the occupancy probability is determined based on the density and distribution range of obstacles; for dynamic targets, the probability distribution of their influence range is calculated by combining the target's motion characteristics. A Bayesian probabilistic model is used to fuse the probabilities of multiple environmental features to generate a unified probabilistic map. For each grid cell, its probability value reflects the confidence level of the presence of environmental elements at that location; the higher the probability value, the more attention that area needs to be paid to it.

[0076] S205. Divide the search space into three-dimensional grid cells, and calculate the unexplored probability value of each grid cell based on the space occupancy status. This unexplored probability value represents the possibility that there is an unsearched area within each grid cell.

[0077] Among them, the search space refers to the three-dimensional area that needs to be explored; the grid cell represents the basic cubic unit after the space is divided; the unexplored probability value refers to the degree to which a specific location has not been fully explored; and the unsearched area represents the spatial area that has not yet been explored.

[0078] The task area is divided into a regular three-dimensional grid structure, with the grid size determined based on environmental characteristics and detection accuracy requirements. For each grid cell, an unexplored probability value is calculated based on its current spatial occupancy. The calculation process considers the following factors: the sensor's detection range and accuracy characteristics, assigning a higher unexplored probability to areas outside the detection range; historical detection records, adjusting the probability value based on the number of detections and time intervals; and environmental complexity, increasing the unexplored probability for areas with dense obstacles or frequent dynamic changes. This method constructs a complete unexplored probability distribution map to guide subsequent path planning.

[0079] S206. Collect and record the sequence of unexplored probability values ​​for each grid cell within the set time window, and calculate the difference in unexplored probability values ​​of grid cells at adjacent sampling times within the time window.

[0080] Among them, the time window represents the period of continuous observation; the unexplored probability value sequence refers to the time series of probability values ​​of a specific grid cell; the probability value difference represents the change in probability values ​​between adjacent time points; and the sampling time refers to the discrete time point of data acquisition.

[0081] Within a fixed-size time window, the changes in the unexplored probability value of each grid cell are recorded at a set sampling frequency. The sampling frequency needs to meet the sampling theorem requirements for environmental change characteristics to ensure that important dynamic information is not lost. For each grid cell, its probability value within the time window is constructed into a time series. The difference in probability values ​​between adjacent sampling times is calculated by subtracting the probability value of the previous time from the current time. These differences reflect the rate of change of environmental characteristics; larger differences indicate significant changes in the area, requiring focused exploration. Statistical analysis is performed on the probability value differences of all grid cells to establish a temporal characteristic model of environmental change.

[0082] S207. Filter out grid cells whose probability values ​​change more than a preset change threshold, and count the frequency of change of each significantly changing grid cell within the time window.

[0083] Among them, the preset change threshold represents the standard value for judging the significance of the change in probability value; the significantly changed grid cell refers to the cell whose probability value changes by more than the threshold; the change frequency represents the number of times the probability value changes significantly; and the time window refers to the time range of the observed data.

[0084] The probability value difference sequence calculated for each grid cell is analyzed, and a uniform change threshold is set as the judgment criterion. When the absolute value of the probability value difference at a certain moment exceeds the threshold, it is recorded as a significant change. The change threshold is set based on statistical analysis of environmental characteristics, typically taking 2-3 times the standard deviation of the probability values. For the identified significantly changing grid cells, the total number of times significant changes occur within the entire time window is counted. The statistical process uses a counter to record the number of times the threshold is exceeded, and simultaneously records the timestamp of the change. These statistical data reflect the temporal distribution characteristics of environmental changes; the higher the frequency of changes, the more significant the dynamic characteristics of the area.

[0085] S208. Cluster the grid cells based on spatial proximity and similarity of change patterns, extract the spatial range and change characteristics of each cluster region, and output the dynamic environmental change region marked with the change frequency.

[0086] Among them, spatial proximity refers to the distance relationship between grid cells in spatial location; change pattern similarity represents the degree of similarity of probability value change characteristics; clustering region refers to a set of grid cells with similar change characteristics; change frequency represents the time frequency characteristics of environmental feature changes.

[0087] Significantly changing grid cells are grouped according to their spatial location and variation characteristics. Spatial proximity is determined by calculating the Euclidean distance between grid cells and setting a distance threshold. Variation pattern similarity is determined by calculating the correlation coefficient or distance metric of probability value variation sequences. A density clustering algorithm such as DBSCAN is used to aggregate spatially adjacent grid cells with similar variation patterns into independent regions. For each clustered region, the boundary coordinates of its spatial extent are calculated, and characteristic parameters such as the average variation frequency and amplitude of grid cells within the region are extracted. The final output dynamic variation region contains a complete description of its location range and variation characteristics.

[0088] S209. Calculate the gradient of unexplored probability values ​​between adjacent grid cells, and extract grid cells with abrupt probability changes based on the gradient magnitude.

[0089] Among them, the unexplored probability value gradient refers to the rate of change of probability values ​​between adjacent grid cells; the gradient magnitude represents the drastic degree of change in the probability value space; and the grid cell with abrupt probability value change refers to the grid cell with a large gradient value.

[0090] Calculate the probability gradient of adjacent mesh cells in 3D space. For each mesh cell, calculate the first-order difference of the probability values ​​in six adjacent directions (up, down, left, right, front, back), obtaining the gradient components for each direction. Calculate the gradient using the central difference scheme, i.e., divide the difference in probability values ​​between the current mesh cell and its neighboring cells by the mesh spacing. Combine the gradient components in each direction to obtain the gradient vector, and calculate the magnitude of the gradient vector as the gradient value. Set a gradient threshold, and extract mesh cells whose gradient values ​​exceed the threshold as locations of abrupt probability changes. These locations typically correspond to boundaries of environmental features or areas of drastic change, and need to be given priority consideration in path planning.

[0091] S210. Connect adjacent high-gradient grid cells to construct the boundary profile, perform feature analysis on the boundary profile and extract key nodes, and calculate the intensity of environmental change at each key node.

[0092] Among them, high-gradient grid cells refer to grid cells whose probability gradient exceeds a threshold; boundary contours represent the spatial boundary lines of areas of environmental change; key nodes refer to the salient locations of features on the boundary contours; and environmental change intensity represents the degree of drastic change in environmental features at a specific location.

[0093] High-gradient grid cells with spatially adjacent locations are connected to form a complete boundary profile using a boundary tracing algorithm. The tracing process starts from any high-gradient cell and sequentially searches for and connects neighboring high-gradient cells using an eight-neighborhood search. Geometric features, including curvature and rotation angles, are calculated for the formed boundary profile. Key nodes are extracted at locations with significant curvature or rotation angles. For each key node, the gradient and probability values ​​within its neighborhood are statistically analyzed to calculate the intensity of environmental change. The calculation of change intensity comprehensively considers three factors: gradient magnitude, probability value difference, and change duration, and a weighted summation method is used to obtain a normalized intensity value.

[0094] S211. Generate priorities based on node positions and change intensity, and output path node data containing node positions and priorities.

[0095] Among them, node location refers to the three-dimensional spatial coordinates of key nodes; change intensity indicates the degree of change in environmental features at the node; priority refers to the importance of path node detection; path node data contains complete attribute information of the node.

[0096] Detection priorities are determined based on the location and intensity of change of key nodes. Priority calculation employs a multi-factor scoring method, including the following steps: First, a coverage score is calculated based on the spatial distribution of nodes, with dispersed nodes receiving higher scores; then, an importance score is calculated based on the intensity of change, with higher intensity yielding higher scores; finally, the two scores are weighted and summed according to set weights to obtain the node's comprehensive priority score. The node's location coordinates, change intensity, and priority score are organized into a unified data structure as the foundation for path planning. Priority scores are represented using normalized values ​​of 0-1 for ease of subsequent processing.

[0097] S212. Compress the path node data into lightweight path information, which includes node coordinates, node priorities calculated based on probability distribution, and temporal constraint relationships determined based on environmental change trends.

[0098] Among them, lightweight path information refers to path description data that has been optimized and compressed; node coordinates represent the spatial location of path points; node priority refers to the importance of the detection order; and temporal constraints represent the time sequence requirements for node access.

[0099] The original path node data is compressed and structurally optimized to generate a more efficient data representation. First, nodes are spatially clustered, merging nodes with close proximity and retaining higher-priority nodes as representative points. Then, node priorities are adjusted based on probability distribution characteristics. The calculation method involves weighting the probability value of a node's location with its original priority to obtain an updated priority value. Finally, the temporal characteristics of environmental changes are analyzed to establish access order constraints between nodes. These temporal constraints are represented by a directed graph structure, where edge weights reflect the order of access between nodes. The optimized node coordinates, priorities, and temporal constraints are packaged into a compact data structure for subsequent path planning.

[0100] S213. Obtain the real-time position and motion status parameters of each UAV, and calculate the shortest travel time from each UAV to each path node.

[0101] Among them, real-time position refers to the current three-dimensional spatial coordinates of the UAV; motion state parameters include dynamic characteristics such as speed, acceleration and attitude angle; shortest flight time represents the minimum time required for the UAV to reach the target node from its current position; path node refers to the critical path point when planning the route.

[0102] The drone's real-time position information, including latitude, longitude, and altitude, is acquired through its onboard positioning system. Simultaneously, motion parameters such as current velocity vector, acceleration, and Euler angles are recorded. For each drone and each path node, the shortest flight time is calculated. The calculation process is based on the A* algorithm: First, a 3D path planning space considering obstacles is established, and kinematic constraints for the drone are set. Then, a heuristic distance estimate from the starting point to the destination is calculated, and the distance is divided by the drone's maximum cruising speed to obtain a time estimate. Finally, the shortest path satisfying the constraints is found through iterative search, and the actual flight time for this path is calculated. The flight time calculation considers acceleration and deceleration processes, and accurate time values ​​are obtained by solving the kinematic equations.

[0103] S214. Assign exploration tasks to the UAV based on flight time and node priority, generate an initial trajectory that satisfies the task assignment, and verify the feasibility of the trajectory based on the UAV performance parameters.

[0104] Among them, flight time refers to the time required for the UAV to complete the path flight; node priority indicates the importance of the path point detection; initial trajectory refers to the initially planned flight path; performance parameters include flight performance indicators such as maximum speed, turning radius, and climb rate.

[0105] Based on the calculated flight time matrix and node priorities, a task allocation algorithm is used to assign reconnaissance tasks to UAVs. The allocation process is implemented using the Hungarian algorithm: first, a cost matrix is ​​constructed, with each element being a weighted combination of flight time and node priority; then, the optimal matching scheme is found through matrix transformation, determining the set of path nodes responsible for each UAV. Initial tracks are generated for each UAV based on the allocation results. Cubic spline curves are used for track smoothing to ensure path continuity. The generated tracks are then validated for feasibility, checking whether they meet performance constraints such as the UAV's minimum turning radius and maximum climb angle. The validation process involves simulating and calculating characteristic parameters such as track curvature and slope, comparing them with performance limitations to ensure the tracks meet flight requirements.

[0106] S215. Optimize the track segments in the initial track that do not meet the constraints, establish the time sequence relationship of multiple UAV tracks and generate track sequences, and distribute the track sequences to the corresponding UAVs.

[0107] Among them, a track segment refers to a local path segment in a track; a constraint refers to the kinematic and dynamic restrictions imposed on the UAV; a track timing relationship indicates the sequential requirements for the execution of multiple UAV tracks; and a track sequence refers to a complete list of flight path points.

[0108] Flight path segments that do not meet the constraints are optimized and adjusted. The optimization process uses an iterative method: first, flight path segments that violate the constraints are identified, including path segments with excessively small turning radii, excessively large climb angles, or excessive speeds; then, these flight path segments are refitted using Bézier curves, and the control point positions are adjusted to make the curves meet the constraint requirements; finally, the performance indicators of the optimized flight path are verified. A temporal relationship is established for the flight paths of multiple UAVs, and the execution order is calculated based on task priority and spatial location. The establishment of the temporal relationship uses a graph theory method: the flight path sequence is represented as a directed graph, where nodes represent path points, edges represent flight paths, and edge weights contain flight time and priority information. A topological sort is used to determine the execution order that meets the temporal constraints. The generated flight path sequence contains precise position coordinates, timestamps, and speed commands, and is distributed to the flight control systems of each UAV in a standard data format. The distribution process is carried out through a secure communication channel to ensure the reliability of data transmission. After receiving the flight path sequence, each UAV executes the planned path through the flight path tracking controller.

[0109] Following step S215, the following steps are also included:

[0110] The system acquires dynamic obstacle feature information from environmental data, updates the obstacle avoidance strategy of the trajectory sequence based on the feature information, and obtains the adjusted trajectory sequence.

[0111] In this step, dynamic obstacle feature information represents the position, speed, and motion trend of moving obstacles; obstacle avoidance strategy refers to the specific methods and path adjustment rules for the UAV to avoid obstacles; and the adjusted flight path sequence refers to the flight path optimized for obstacle avoidance.

[0112] Feature extraction and motion prediction are performed on the acquired dynamic obstacle data. Feature extraction includes calculating the current position and velocity vector of the obstacle and analyzing historical trajectory data to predict future motion trends. The obstacle avoidance strategy is updated using a dynamic window method: first, the danger zone of the obstacle is calculated, taking into account the obstacle's size and speed; then, it is detected whether the existing trajectory intersects with the danger zone; for trajectory segments with collision risk, obstacle avoidance paths are generated based on the artificial potential field method. Potential field construction treats obstacles as repulsive sources and target points as attractive sources, and a safe trajectory is obtained through force field superposition calculations. Path adjustments must ensure that the UAV's motion constraints are met while maintaining the temporal relationship of the original trajectory.

[0113] The system generates optimized flight path sequences based on obstacle avoidance strategies, determines the execution status of the optimized flight path sequences, responds to the execution status to implement task control, and distributes the adjusted flight path sequences to the corresponding UAVs for execution.

[0114] In this step, the optimized flight path sequence refers to the complete flight path after obstacle avoidance; the execution status indicates the progress and abnormal situations of the flight path execution; and the mission control refers to the real-time management and adjustment of the flight mission.

[0115] Based on the updated obstacle avoidance strategy, a new trajectory sequence is generated using a trajectory optimization algorithm. Multiple evaluation metrics are set during the optimization process: path safety, flight time, and energy consumption. The trajectory parameters are iteratively optimized using gradient descent until the evaluation requirements are met. The optimized trajectory is then monitored for execution status: the actual position of the UAV is compared with the planned path in real time, and the tracking error is calculated; any anomalies such as timeouts or yaws are detected. Task control is based on the status monitoring results: path replanning is triggered when the tracking error exceeds a threshold; and emergency handling procedures are initiated when an abnormal state occurs. Task control commands are transmitted to each UAV in real time via the communication network to ensure the reliable execution of collaborative tasks.

[0116] Obtain the remaining battery status parameters of each drone and calculate the available flight time corresponding to the remaining battery status parameters.

[0117] In this step, the remaining power status parameter refers to the numerical characteristics of the battery's remaining capacity; the available flight time represents the flight duration that can be maintained under the current load; and the status parameters include real-time data such as voltage, current, and temperature.

[0118] The battery management system collects state-of-the-art (SOP) data from the drone, including real-time voltage, discharge current, and battery temperature. A battery prediction model is built based on these parameters: first, the remaining percentage of charge is estimated based on the current voltage and discharge curve; then, the actual usable capacity is calculated by combining load power and ambient temperature; finally, the available flight time is derived using an energy consumption model. The calculation process considers multiple influencing factors: the impact of flight speed and maneuverability on power consumption, the impact of temperature on battery performance, and the provision for a safety margin. These calculations yield an accurate estimate of the remaining flight time.

[0119] The task allocation weights are determined based on the available flight time, and the UAV flight parameters are adjusted accordingly to generate a return path plan that meets the power constraints.

[0120] In this step, the task allocation weight refers to the task priority coefficient determined based on the remaining battery power; flight parameters include control variables such as speed and altitude; and the return path refers to the flight path through which the drone safely returns to its starting point.

[0121] Based on the calculated available flight time, a priority evaluation system for task allocation is established. Weight calculation considers the following factors: the safety margin of remaining battery power, the progress of the current task, and the energy requirement for the return distance. Flight parameters are dynamically adjusted: cruise speed and climb rate are adjusted according to the remaining battery power to optimize energy utilization efficiency. The A* algorithm is used for return path planning, with minimum energy consumption as the optimization objective. The influence of wind field is considered during path generation, and a flight altitude conducive to energy saving is selected. These measures ensure that the UAV can safely return when battery power is insufficient, avoiding unexpected forced landings.

[0122] The following describes the UAV swarm cooperative path planning system in the embodiments of this invention from the perspective of hardware processing. Please refer to [link / reference needed]. Figure 3 This is a schematic diagram of a physical device structure of a drone swarm cooperative path planning system in this application embodiment.

[0123] It should be noted that, Figure 3 The structure of the drone swarm collaborative path planning system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0124] like Figure 3As shown, the UAV swarm cooperative path planning system includes a Central Processing Unit (CPU) 301, which can perform various appropriate actions and processes based on programs stored in Read-Only Memory (ROM) 302 or programs loaded from storage section 308 into Random Access Memory (RAM) 303, such as executing the methods described in the above embodiments. The RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An Input / Output (I / O) interface 305 is also connected to the bus 304.

[0125] The following components are connected to I / O interface 305: input section 306 including audio input devices, push-button switches, etc.; output section 307 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 308 including a hard disk, etc.; and communication section 309 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 309 performs communication processing via a network such as the Internet. Drive 310 is also connected to I / O interface 305 as needed. Removable media 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0126] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by central processing unit (CPU) 301, it performs the various functions defined in the present invention.

[0127] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0128] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0129] Specifically, the UAV swarm cooperative path planning system of this embodiment includes a processor and a memory. The memory stores a computer program, and when the computer program is executed by the processor, it implements the UAV swarm cooperative path planning method provided in the above embodiment.

[0130] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the UAV swarm cooperative path planning system described in the above embodiments; or it may exist independently and not assembled into the UAV swarm cooperative path planning system. The storage medium carries one or more computer programs, which, when executed by a processor of the UAV swarm cooperative path planning system, cause the UAV swarm cooperative path planning system to implement the UAV swarm cooperative path planning method provided in the above embodiments.

[0131] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0132] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0133] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for cooperative path planning of a UAV swarm, characterized in that, The method, applied to a collaborative path planning system for unmanned aerial vehicle (UAV) swarms, includes: Collect environmental data within the detection range of each UAV, and determine the spatial occupancy status of multiple grid cells in the probability map based on the environmental data. The environmental data includes terrain height data, obstacle distribution data, and dynamic target data. The search space is divided into three-dimensional grid cells, and the unexplored probability value of each grid cell is calculated based on the space occupancy status. The unexplored probability value represents the possibility that there is an unsearched area within each grid cell. Extract the change data of the unexplored probability value of the grid cell in the time dimension, calculate the probability difference of the unexplored probability value of the grid cell at adjacent time points, and identify the dynamic change area of ​​the environment based on the change data; Extract the gradient information of unexplored probability values ​​between the grid cells, determine the boundary features of the environmental change area based on the probability value difference and the gradient information, and generate path node data based on the boundary features; The path node data is compressed into lightweight path information, which includes node coordinates, node priorities calculated based on probability distribution, and temporal constraint relationships determined based on the environmental change trend. Based on the lightweight path information and the current location information of each UAV, a flight path sequence that satisfies the time constraints is generated, and the flight path sequence is distributed to the corresponding UAV.

2. The method of claim 1, wherein, The step of collecting environmental data within the detection range of each UAV and determining the spatial occupancy status of multiple grid cells in the probability map based on the environmental data specifically includes: Receive ranging data and image data collected by the lidar and camera carried by each UAV; perform noise filtering on the ranging data to generate three-dimensional point cloud data; The three-dimensional point cloud data is aligned and stitched together to extract ground points and obstacle points from the three-dimensional point cloud data, thereby obtaining terrain height data and obstacle distribution data. By comparing the differences in target positions in images at different times and marking moving targets in the images, dynamic target data is obtained; the terrain height data, obstacle distribution data, and dynamic target data are mapped onto a three-dimensional grid map to obtain the spatial occupancy status of each grid cell; A probability map is generated based on the spatial occupancy status of each grid cell.

3. The method according to claim 1, characterized in that, The steps of extracting the change data of the unexplored probability value of the grid cell in the time dimension, calculating the probability difference of the unexplored probability value of the grid cell at adjacent time points, and identifying the dynamically changing areas of the environment based on the change data specifically include: Collect and record the sequence of unexplored probability values ​​for each grid cell within a set time window, and calculate the difference in unexplored probability values ​​of grid cells at adjacent sampling times within the time window; Filter out grid cells whose probability values ​​change more than a preset change threshold, and count the frequency of change of each significantly changing grid cell within a time window; The grid cells are clustered based on spatial proximity and similarity of change patterns. The spatial extent and change characteristics of each cluster region are extracted, and the dynamic environmental change regions marked with change frequency are output.

4. The method according to claim 1, characterized in that, The steps of extracting the gradient information of unexplored probability values ​​between the grid cells, determining the boundary features of the environmental change region based on the probability value difference and the gradient information, and generating path node data based on the boundary features specifically include: Calculate the gradient of unexplored probability values ​​between adjacent grid cells, and extract grid cells with abrupt probability changes based on the gradient magnitude; Connect adjacent high-gradient grid cells to construct the boundary profile, perform feature analysis on the boundary profile and extract key nodes, and calculate the intensity of environmental change at each key node. Priorities are generated based on node location and change intensity, and path node data containing node location and priority is output.

5. The method according to claim 1, characterized in that, The step of generating a trajectory sequence that satisfies the temporal constraints based on the lightweight path information and the current position information of each UAV, and distributing the trajectory sequence to the corresponding UAV, specifically includes: Obtain the real-time position and motion status parameters of each UAV, and calculate the shortest travel time from each UAV to each path node; Based on flight time and node priority, exploration tasks are assigned to UAVs, an initial flight path that meets the task assignment is generated, and the feasibility of the flight path is verified based on the UAV performance parameters. The track segments that do not meet the constraints in the initial track are optimized, the time sequence relationship of multiple UAV tracks is established and the track sequence is generated, and the track sequence is distributed to the corresponding UAVs.

6. The method according to claim 1, characterized in that, After the steps of generating a trajectory sequence that satisfies the temporal constraints based on the lightweight path information and the current location information of each UAV, and distributing the trajectory sequence to the corresponding UAVs, the method further includes: The dynamic obstacle feature information in the environmental data is obtained, and the obstacle avoidance strategy of the trajectory sequence is updated according to the feature information to obtain the adjusted trajectory sequence. Based on the obstacle avoidance strategy, an optimized flight path sequence is generated, the execution status of the optimized flight path sequence is determined, and task control is implemented in response to the execution status, distributing the adjusted flight path sequence to the corresponding UAV for execution.

7. The method according to claim 6, characterized in that, After the steps of generating a trajectory sequence that satisfies the temporal constraints based on the lightweight path information and the current location information of each UAV, and distributing the trajectory sequence to the corresponding UAVs, the method further includes: Obtain the remaining battery status parameters of each drone and calculate the available flight time corresponding to the remaining battery status parameters; The task allocation weight is determined based on the available flight time, and the UAV flight parameters are adjusted according to the task allocation weight to generate a return path plan that meets the power constraints.

8. A collaborative path planning system for a swarm of unmanned aerial vehicles (UAVs), characterized in that, The UAV swarm cooperative path planning system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to cause the UAV swarm cooperative path planning system to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instructions are executed on the UAV swarm cooperative path planning system, the UAV swarm cooperative path planning system performs the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on the UAV swarm collaborative path planning system, the UAV swarm collaborative path planning system performs the method as described in any one of claims 1-7.

Images