An emergency rescue unmanned aerial vehicle group cooperative partition coverage search and rescue method

Abstract

The application discloses an emergency rescue unmanned aerial vehicle group cooperative partition coverage search and rescue method, and particularly relates to the field of aircraft cooperative route control and visual environment sensing technology, and is used for solving the problem that the time inconsistency of local environment maps asynchronously constructed by each unmanned aerial vehicle in a dynamic disaster site directly leads to cooperative route planning conflicts and reduced safety. The method is achieved by the following steps: in the cooperative search and rescue process, acquiring the local environment maps constructed by each unmanned aerial vehicle based on visual SLAM, identifying the spatial overlapping areas between the maps and judging the obstacle state consistency thereof, if the obstacle state consistency is inconsistent, deducing the physical reversibility of state transition, and then based on the reversibility, deducing the potential influence of obstacle change on the topological connectivity network used for route planning to identify conflict-derived paths, subsequently evaluating the navigation conflict risk level according to the size and criticality of the conflict paths, and finally dynamically adjusting the cooperative route planning of the relevant unmanned aerial vehicles according to the risk level.

Description

Technical Field

[0001] This invention relates to the field of aircraft cooperative flight path control and visual environment perception technology, and more specifically, to a method for cooperative zoned coverage search and rescue of emergency rescue drone swarms. Background Technology

[0002] In the field of emergency rescue, drone swarms can be used to conduct coordinated, zoned search and rescue operations across a wide disaster area to improve search and rescue efficiency. Current technologies typically equip drones with visual intelligence systems, enabling them to autonomously locate, navigate, and build local environmental maps in real time in unknown or partially known environments based on visual SLAM (Simultaneous Localization and Mapping) technology. Through mission planning, the target area is divided into multiple sub-areas and assigned to various drones. The swarm, through information exchange and coordinated control, aims to achieve rapid and comprehensive search and rescue of the entire area, and then transmit the perceived information back to the command center.

[0003] However, in real, dynamically changing disaster sites, such as environments with aftershocks, falling rocks, or persistent structural instability, the aforementioned visual SLAM-based collaborative search and rescue methods face contradictions: because the disaster environment itself is dynamically changing, and the perception behaviors of each UAV in the swarm of the same spatial area occur at different times, this leads to inherent temporal inconsistencies in the local environmental maps constructed in real time by each UAV based on its own visual perception. When the UAV swarm performs collaborative route planning based on these locally discrepancies in map information, it directly causes conflicts in the planned paths, reduces feasibility, and even increases the risk of collisions between UAVs, severely restricting the overall reliability and safety of the UAV swarm in performing collaborative zonal coverage search and rescue missions in dynamic emergency scenarios. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method for collaborative zoned coverage search and rescue of emergency rescue drone swarms to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A method for collaborative zoned coverage search and rescue using emergency rescue drone swarms includes:

[0007] S1. During the collaborative search and rescue operation of a drone swarm, acquire the local environment map constructed in real time by each drone in the swarm based on visual SLAM.

[0008] S2. Identify the spatially overlapping areas in each local environment map and determine whether the obstacle status of the spatially overlapping areas is consistent in each local environment map.

[0009] S3. If the obstacle states in the spatially overlapping areas are inconsistent, then deduce and evaluate the physical reversibility of the transition paths between the inconsistent obstacle states in the spatially overlapping areas.

[0010] S4. Based on the physical reversibility, extrapolate the potential impact of obstacle state changes on the topological connectivity network on which UAV swarm collaborative route planning depends, and identify the conflict-derived paths that may be caused by obstacle state changes.

[0011] S5. Based on the scale and criticality of the conflict-derived paths, comprehensively assess the navigation conflict risk level of the spatially overlapping areas on the collaborative route planning of UAV swarms.

[0012] S6. Based on the navigation conflict risk level, dynamically adjust the collaborative route planning of relevant UAVs in the UAV swarm for areas with spatial overlap.

[0013] Furthermore, during the collaborative search and rescue operation involving a swarm of drones, local environmental maps constructed in real-time by each drone within the swarm based on visual SLAM are acquired, including:

[0014] Real-time construction of local environment maps for each UAV based on visual SLAM;

[0015] Create timestamps for each local environment map associated with the map;

[0016] Local environment maps with associated map building timestamps are shared and distributed within the drone swarm.

[0017] Furthermore, the system identifies spatially overlapping regions in each local environment map and determines whether the obstacle states in these overlapping regions are consistent across the various local environment maps, including:

[0018] Receive local environment maps that are shared and distributed within the drone swarm and are associated with map building timestamps;

[0019] Spatiotemporal coordinate alignment of local environment maps is performed based on map construction timestamps;

[0020] In the local environment maps after coordinate alignment, spatially overlapping areas with common three-dimensional spatial extent identifiers are identified;

[0021] Within the identified spatially overlapping areas, the obstacle outlines and spatial occupancy information at corresponding locations in each local environment map are extracted.

[0022] The obstacle outlines and space occupancy information from different local environment maps are overlaid and compared.

[0023] If the overlay comparison results show that the obstacles have differences in spatial location, geometry, or occupied area that exceed a preset threshold, then the obstacle status in the spatially overlapping area is determined to be inconsistent.

[0024] Furthermore, if the obstacle states in the spatially overlapping regions are inconsistent, the physical reversibility of the transition paths between inconsistent obstacle states is deduced and evaluated for the spatially overlapping regions, including:

[0025] Based on the judgment results of the inconsistency of obstacle states within the spatially overlapping area, determine the specific state representation of the obstacle state before and after the inconsistency is caused.

[0026] Based on the specific state representation of the obstacle before and after the state change, infer the object attributes or change type involved in the state change.

[0027] Determine the physical reversibility of the obstacle state change based on the inferred object properties or change type.

[0028] Furthermore, inferring the object attributes or change types involved in the state change includes: matching the specific state representation of the obstacle before and after the state change with a predefined typical change pattern library, which includes at least one of landslide, rockfall, moving object, dust, and smoke, and determining its physical reversibility based on the matched typical change patterns.

[0029] Furthermore, based on the principle of physical reversibility, the potential impact of obstacle state changes on the topological connectivity network upon which UAV swarm collaborative route planning relies is extrapolated, identifying conflict-derived paths that may be triggered by obstacle state changes, including:

[0030] Construct a topological connectivity network representing the traversable areas of UAVs and their connections based on all local environment maps;

[0031] Based on the determination of physical reversibility, if the change is determined to be irreversible, the path connection relationship that has failed due to the change of obstacle state will be removed or modified in the topological connectivity network.

[0032] In the updated topological connectivity network, we analyze additional paths that affect the connectivity of other unchanged regional airways due to changes in the connection relationships of some travel paths, and identify these additional paths as conflict-derived paths.

[0033] Furthermore, constructing a topological connectivity network representing the traversable areas of the UAV and their connections based on all local environment maps includes: merging the spatial voxels not occupied by obstacles in each local environment map to generate a global traversable space 3D grid; and extracting the connection relationships between adjacent and connected grid cells in the global traversable space 3D grid to form a topological connectivity network.

[0034] Furthermore, based on the scale and criticality of conflict-derived paths, a comprehensive assessment is conducted to evaluate the navigation conflict risk level of overlapping spatial areas on UAV swarm collaborative route planning, including:

[0035] The number of conflict-derived paths identified is used to obtain a size measure;

[0036] For each conflict-derived path, the corresponding conflict-derived path is simulated to be removed in the updated topological connectivity network, and the resulting decrease in network connectivity is calculated to obtain a critical metric.

[0037] The navigation conflict risk level is determined by querying a pre-defined risk level table based on size and criticality metrics.

[0038] Furthermore, based on the navigation conflict risk level, the collaborative route planning of relevant UAVs in the UAV swarm for areas of spatial overlap is dynamically adjusted, including:

[0039] Select the corresponding route planning adjustment strategy from the preset adjustment strategy library based on the navigation conflict risk level;

[0040] Based on the route planning adjustment strategy, flight paths for affected UAVs are replanned in the updated topological connectivity network to bypass spatially overlapping areas or conflict-derived paths;

[0041] The replanned flight paths will be issued to the relevant drones to replace their original collaborative flight path plans.

[0042] Furthermore, the preset adjustment strategy library stores route planning adjustment strategies corresponding to different navigation conflict risk levels. The route planning adjustment strategies include: for low risk levels, instructing relevant UAVs to maintain their original routes but reduce their flight speed; for medium risk levels, instructing relevant UAVs to replan local paths to avoid spatial overlap areas; and for high risk levels, instructing relevant UAVs to pause their advance and wait for a reassigned global path.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. Through in-depth analysis and evaluation of temporal inconsistencies between local environmental maps, a refined dynamic decision-making basis is provided for collaborative route planning of UAV swarms. The differences in map state at the perception level are transformed into judgments on the physical properties of obstacle state changes, thereby predicting the potential chain reaction impact on the connectivity of the entire route network and quantifying the risk level. This allows the UAV swarm's response to no longer be based solely on immediate, local, and potentially contradictory perception snapshots, but rather on an understanding of the nature and scope of environmental changes. It can distinguish between temporary interference and permanent changes, and adopt graded adjustment strategies for different levels of risk. Thus, while ensuring the continuity of search and rescue missions, it effectively avoids path planning conflicts and collision risks directly caused by map information conflicts, significantly improving the overall reliability and safety of multi-aircraft collaborative operations in complex dynamic environments.

[0045] 2. This achievement represents a leap from relying on visual intelligence for environmental reconstruction to utilizing visual intelligence to understand environmental dynamics and guide collaborative control. By constructing a topological connectivity network model directly related to flight path planning and mapping the physical reversibility assessment results to this network for impact deduction, advanced environmental perception information is deeply embedded into the flight path control decision-making closed loop of the aircraft. This enables the UAV swarm to shift from passively avoiding obstacles based on the latest map to proactively predicting the systemic impact of environmental changes on the collaborative flight network and performing local or global route optimization in advance. This not only solves the problem of information inconsistency caused by asynchronous perception but also achieves dynamic, flexible, and safe control of the swarm's flight path while ensuring coverage of search and rescue mission objectives. This demonstrates the intelligent decision-making advantages of the new generation of autonomous collaborative control systems for aircraft in dealing with unstructured and dynamically changing scenarios. Attached Figure Description

[0046] Figure 1 This is a flowchart of a collaborative zoned coverage search and rescue method for emergency rescue drone swarms according to the present invention. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] Example: Figure 1 This invention provides a method for collaborative zoned coverage search and rescue using a swarm of emergency rescue drones, comprising:

[0049] S1. During the collaborative search and rescue operation of a drone swarm, acquire the local environment map constructed in real time by each drone in the swarm based on visual SLAM.

[0050] S2. Identify the spatially overlapping areas in each local environment map and determine whether the obstacle status of the spatially overlapping areas is consistent in each local environment map.

[0051] S3. If the obstacle states in the spatially overlapping areas are inconsistent, then deduce and evaluate the physical reversibility of the transition paths between the inconsistent obstacle states in the spatially overlapping areas.

[0052] S4. Based on the physical reversibility, extrapolate the potential impact of obstacle state changes on the topological connectivity network on which UAV swarm collaborative route planning depends, and identify the conflict-derived paths that may be caused by obstacle state changes.

[0053] S5. Based on the scale and criticality of the conflict-derived paths, comprehensively assess the navigation conflict risk level of the spatially overlapping areas on the collaborative route planning of UAV swarms.

[0054] S6. Based on the navigation conflict risk level, dynamically adjust the collaborative route planning of relevant UAVs in the UAV swarm for areas with spatial overlap.

[0055] S1. During the collaborative search and rescue operation involving a swarm of drones, acquire the local environment map constructed in real time by each drone in the swarm based on visual SLAM. The specific implementation is as follows:

[0056] Each drone is equipped with a visual sensor and an onboard computing unit. The visual sensor uses an RGB-D camera with depth perception or a calibrated stereo camera, with a frame rate set to at least 30 frames per second. The onboard computing unit runs a real-time visual localization and mapping (LOM) algorithm. The LOM algorithm processes the image stream from the visual sensor. It extracts features from each input frame, including ORB feature points and SIFT feature descriptors. The algorithm associates the feature points extracted in the current frame with feature points in previous frames or existing maps through feature matching. Feature matching is achieved by calculating the Hamming or Euclidean distance between feature descriptors and finding the nearest neighbor match. The algorithm uses a random sampling consensus algorithm to eliminate erroneous feature matching pairs. Based on correct feature matching pairs, the algorithm calculates the drone's relative pose transformation from the previous moment to the current moment by solving the camera motion equations. The real-time visual localization and mapping (VMR) algorithm constructs a map using continuously estimated poses. It employs triangulation to backproject matched feature points into 3D space, calculating the 3D coordinates of these feature points in the world coordinate system to form a 3D point cloud map. The VMR algorithm undergoes bundled adjustment and optimization, minimizing reprojection errors to optimize the camera pose and the position of points on the 3D map. The onboard computing unit generates and updates a local environment map in a local coordinate system with the UAV's initial position as the origin. This local environment map is stored in the onboard memory as a 3D point cloud.

[0057] A map building timestamp is associated with each local environment map. This timestamp is obtained from the system clock of the onboard computing unit. The system clock is synchronized with the time signal output by the onboard high-precision GPS receiver during UAV power-on initialization. After the visual real-time positioning and map building algorithm integrates and updates a frame of map data, it reads the system clock value of the onboard computing unit at that moment and binds this system clock value as the map building timestamp with the local environment map for storage. The accuracy of the map building timestamp reaches the millisecond level.

[0058] Local environment maps, associated with map build timestamps, are shared and distributed within a drone swarm. This sharing and distribution occurs through a self-organizing wireless network established within the swarm. After updating the local environment map and associating it with a map build timestamp, each drone encapsulates the local environment map data and its associated map build timestamp into a data packet via its onboard wireless communication module. The data packet structure includes a header, payload, and checksum. The header indicates the data type as a local environment map and the drone identifier of the data source. The payload contains the compressed local environment map data body and the map build timestamp. The drone transmits the data packet according to a time-division multiple access-based communication schedule. Within communication range, the drone broadcasts the data packet to neighboring drones. Upon receiving the data packet, the receiving drone verifies the checksum, decompresses the payload, reconstructs the local environment map data and its map build timestamp, and stores the local environment map data and its map build timestamp in its local cache. The sharing and distribution process continues.

[0059] S2. Identify spatially overlapping areas in each local environment map and determine whether the obstacle states in each local environment map are consistent. Specifically, this is implemented as follows:

[0060] The system receives local environment maps, each associated with a map build timestamp, shared and distributed within a drone swarm via a wireless ad hoc network. The receiving process includes reading data packets parsed and stored by the communication module from a local cache, and decompressing the payload of the data packets to restore structured local environment map data and its corresponding map build timestamp.

[0061] Spatiotemporal coordinate alignment of local environment maps is performed based on map construction timestamps. This alignment unifies local environment maps built by different UAVs at different times into a single global reference coordinate system. The alignment process sorts all local environment maps chronologically based on their map construction timestamps. For each local environment map, the continuous pose estimation sequence generated by real-time visual localization and map construction is the input. By interpolating the pose estimation sequence using timestamps, the estimated position and attitude of the UAV in its own local coordinate system at a selected reference time point are calculated. Common environmental features observed in different local environment maps are used as spatial constraints. These common environmental features are stable 3D point cloud clusters with significant geometric features. A feature matching algorithm is used to find the correspondence between 3D feature points in the common environmental features between different local environment maps. The feature matching algorithm employs an iterative nearest-point algorithm. This algorithm iteratively calculates rotation matrices and translation vectors to minimize the overall distance between two sets of matching point clouds from two local environment maps, thereby solving for the transformation relationship between the two local coordinate systems. A pose map optimization model is constructed by combining temporal sorting and spatial constraints. The pose map optimization model uses the UAV pose at a reference time corresponding to each local environment map as the optimization variable. It uses the relative pose measurements provided by real-time visual localization and map construction, as well as the spatial relative pose constraints between different local environment maps calculated through feature matching, as edges. All optimization variables are jointly optimized by minimizing the error terms of all constraints. After optimization, the optimal estimate of the UAV pose corresponding to each local environment map in the same global coordinate system is obtained. Using this optimal estimate, the 3D point cloud data in each local environment map is transformed to this global coordinate system, completing spatiotemporal coordinate alignment and generating an aligned local environment map with the same coordinate system.

[0062] In the aligned local environment maps, spatially overlapping regions with a common 3D spatial extent identifier are identified. A spatially overlapping region is a physical spatial extent jointly covered by the 3D point cloud data of two or more aligned local environment maps. The identification method employs a 3D spatial grid partitioning approach. The spatial extent to be analyzed in the global coordinate system is divided into a series of regular 3D voxel grids, with a side length of, for example, 0.2 meters. The 3D point clouds in all aligned local environment maps are traversed, and each 3D point is classified into its corresponding 3D voxel grid based on its global coordinates. For a 3D voxel grid, if it contains 3D points from at least two different aligned local environment maps, it is considered part of the spatially overlapping region. All sets of 3D voxel grids satisfying this condition are spatially clustered, aggregating spatially adjacent 3D voxel grids into a continuous spatial volume block. Each continuous spatial volume block is identified as an independent spatially overlapping region and assigned a unique identifier.

[0063] Within the identified spatially overlapping regions, obstacle contours and spatial occupancy information at corresponding locations in each local environment map are extracted. For a spatially overlapping region, the set of all 3D voxel meshes constituting that region is first determined. For each aligned local environment map, all 3D points in its 3D point cloud that fall within the corresponding 3D voxel mesh of the spatially overlapping region are checked. Based on the spatial distribution of the 3D points, the surface morphology of the obstacles within this spatially overlapping region from the perspective of the local environment map is reconstructed. Obstacle contour information is obtained by extracting the boundary point set of the surface formed by the 3D points, for example, by calculating the 3D convex hull of the point cloud. Spatial occupancy information is defined by determining whether each 3D voxel mesh is occupied by the point cloud data. If the number of 3D points contained in a 3D voxel mesh exceeds a minimum point cloud density threshold, such as 5 points, then the 3D voxel mesh is marked as occupied in the local environment map. For this spatially overlapping region, each aligned local environment map outputs a set of 3D boundary point sequences describing the external contours of the obstacles and a 3D voxel occupancy state map representing the internal spatial occupancy state, constituting the obstacle contours and spatial occupancy information of the map in this region.

[0064] Obstacle contours and spatial occupancy information corresponding to the same spatially overlapping region from different aligned local environment maps are overlaid and compared. Overlay comparison includes geometric contour comparison and occupancy state comparison. Geometric contour comparison calculates the spatial differences between sequences of 3D boundary points describing the same spatially overlapping region from different aligned local environment maps. A set of keypoints is sampled from each 3D boundary point sequence, and the Euclidean distance between corresponding keypoint pairs in different aligned local environment maps is calculated. The average of these distances is used as a measure of geometric shape difference. Occupancy state comparison performs voxel-by-voxel logical operations on the 3D voxel occupancy state maps of different aligned local environment maps. For each 3D voxel grid within the spatially overlapping region, its occupancy state is compared to see if it is consistent in the two aligned local environment maps. The number of 3D voxel grids with inconsistent occupancy states is counted, and this number is divided by the total number of 3D voxel grids in the spatially overlapping region to obtain the occupancy state inconsistency rate.

[0065] If the overlay comparison results show that obstacles differ from each other in spatial location, geometry, or occupancy range beyond a preset threshold, the obstacle states in the spatially overlapping area are determined to be inconsistent. The preset thresholds include a geometric difference threshold and an occupancy inconsistency rate threshold. The geometric difference threshold is set based on the depth measurement error of the visual sensor and the typical reconstruction accuracy of the visual real-time localization and mapping (VMR) algorithm. The depth measurement error of the visual sensor at a specific distance is obtained through sensor calibration. The uncertainty in the map point position estimation of the VMR algorithm is evaluated through the algorithm's internal covariance. The uncertainty range of the 3D point position is estimated using an error propagation model, and the geometric difference threshold is set to a multiple of this uncertainty range, for example, 2 times. The occupancy inconsistency rate threshold is set based on the perceptual consistency requirements of the actual application, for example, 0.3. During the determination, if the geometric shape difference metric obtained from the geometric contour comparison is greater than the set geometric difference threshold, or the occupancy inconsistency rate obtained from the occupancy state comparison is greater than the set occupancy inconsistency rate threshold, the overlay comparison result is determined to show a difference exceeding the preset threshold. If, for a given spatially overlapping region, an overlay comparison between any two aligned local environment maps reveals a difference exceeding a preset threshold, then the obstacle states in that spatially overlapping region are ultimately determined to be inconsistent. This determination result is then output.

[0066] S3. If the obstacle states in the spatially overlapping regions are inconsistent, the physical reversibility of the transition paths between the inconsistent obstacle states is deduced and evaluated for the spatially overlapping regions. Specifically, the implementation is as follows:

[0067] Based on the judgment results of inconsistencies in obstacle states within spatially overlapping areas, the specific state representations of obstacles before and after the state changes that caused the inconsistencies are determined. The process of determining specific state representations involves extracting the specific 3D voxel mesh sets with differences identified during overlay comparison and their states in different aligned local environment maps. For each 3D voxel mesh with inconsistent states, its occupancy state in the aligned local environment map at an earlier timestamp is recorded as state A, and its occupancy state in the aligned local environment map at a later timestamp is recorded as state B. The values ​​of state A and state B are derived from the 3D voxel occupancy state map, and are either occupied or idle. Simultaneously, the difference portion of the 3D boundary point sequence related to the 3D voxel meshes with inconsistent states is extracted to obtain the specific shape descriptions of the obstacle's geometric contours before and after the change. These specific shape descriptions include point cloud density distribution, surface normal vector direction distribution, and the 3D envelope volume of the occupied space. The specific state representation before the change might be described as a continuous region marked as occupied, with its surface point cloud exhibiting approximately vertical wall features. The specific state representation after the change may be described as the same area becoming mostly idle and isolated point clusters scattered on the ground point cloud.

[0068] Based on the specific state representations of obstacles before and after the state change, the object attributes or change types involved in the state change are inferred. The inference process analyzes the spatial change pattern features, temporal change rate features, and spectral or reflectance features contained in the specific state representations. Spatial change pattern features describe the spatial distribution of the three-dimensional voxel grid involved in the state change, including local small-scale isolated changes, large-area continuous changes, and strip-like changes extending along a specific direction. Temporal change rate features use map construction timestamps to calculate the time interval from state A to state B, and combine this with the spatial volume involved in the change to estimate the average rate of physical change. Spectral or reflectance features extract the average pixel brightness values ​​or color distribution corresponding to the changed area before and after the change from the original visual sensor image that generated the local environment map. The extracted spatial change pattern features, temporal change rate features, and spectral or reflectance features are combined into a feature vector.

[0069] Inferring the object attributes or change type involved in state changes is achieved by matching feature vectors with a predefined library of typical change patterns. This predefined library of typical change patterns is a pre-built knowledge base storing obstacle change pattern templates for various typical disaster-related scenarios. The library includes landslide patterns, rockfall patterns, moving object patterns, dust patterns, and smoke patterns. Each typical change pattern is defined by a standard feature vector template and a physical reversibility label. The standard feature vector template defines the typical numerical range or feature description of the pattern in terms of spatial change pattern characteristics, temporal change rate characteristics, and spectral or reflectivity characteristics. For example, the standard feature vector template for a landslide pattern describes a spatial change pattern characterized by large-area continuous change accompanied by a significant decrease in terrain height; a temporal change rate characteristic of rapid or instantaneous change; and a spectral or reflectivity characteristic of abrupt changes in material reflectivity that are similar to the surrounding ruins. Similarly, the standard feature vector template for a moving object pattern describes a spatial change pattern characterized by isolated, compact changes with the total volume occupied by the area before and after the change remaining essentially unchanged; a temporal change rate characteristic of slow or medium speed; and a spectral or reflectivity characteristic that may not show significant abrupt changes. The physical reversibility label directly indicates the physical reversibility that the mode usually corresponds to. The physical reversibility label for the collapse mode is irreversible, while the physical reversibility label for the moving object mode is reversible.

[0070] The matching process calculates the similarity between the feature vector extracted from the specific state representation and each standard feature vector template in the typical change pattern library. The similarity calculation uses Euclidean distance. Euclidean distance is calculated by taking the square root of the sum of the squares of the differences between the feature vector and each standard feature vector template in corresponding dimensions. The typical change pattern corresponding to the standard feature vector template with the smallest Euclidean distance is inferred as the object attribute or change type involved in the current obstacle state change.

[0071] Based on the inferred object attributes or change type, the physical reversibility of the obstacle state change is determined. The determination process directly queries the physical reversibility label associated with the matched typical change patterns. The physical reversibility label is a classification conclusion, which includes reversible or irreversible. If the matched typical change pattern is a collapse pattern, its physical reversibility label is irreversible, thus determining that the current obstacle state change is physically irreversible. If the matched typical change pattern is a moving object pattern, its physical reversibility label is reversible, thus determining that the current obstacle state change is physically reversible. The matching process generates a similarity score, which is the reciprocal of the Euclidean distance or obtained through other normalization methods. When the highest similarity score is lower than a preset matching confidence threshold (e.g., 0.7), the matching result is considered unreliable. In the case of an unreliable matching result, a default determination rule is adopted, which determines that the physical reversibility of the obstacle state change is irreversible. The final output of the physical reversibility determination result is a clear classification conclusion.

[0072] S4. Based on the principle of physical reversibility, extrapolate the potential impact of obstacle state changes on the topological connectivity network upon which UAV swarm collaborative route planning relies, and identify potential conflict-derived paths caused by obstacle state changes. Specifically, this is implemented as follows:

[0073] A topological connectivity network representing the navigable areas of the UAV and their connections is constructed based on all local environment maps. The input for constructing the topological connectivity network is each local environment map after spatiotemporal coordinate alignment. The construction of the topological connectivity network merges the spatial voxels not occupied by obstacles in each local environment map to generate a global navigable space 3D mesh. All aligned local environment maps are traversed, and for each 3D voxel mesh in the global coordinate system, its state in the 3D voxel occupancy state map of all aligned local environment maps is checked. If a 3D voxel mesh is marked as idle in a certain aligned local environment map, then the 3D voxel mesh is added to the set of global navigable space 3D meshes. Each mesh cell in the global navigable space 3D mesh represents a discrete spatial location that the UAV can theoretically occupy.

[0074] The connectivity relationships between adjacent and connected grid cells in the globally traversable 3D grid are extracted to form a topological connectivity network. Adjacency is defined based on the spatial positional relationship of grid cells, considering six-connected neighborhoods. A six-connected neighborhood refers to a grid cell that is immediately adjacent to other grid cells in the six directions: up, down, left, right, front, and back. Connectivity is determined by two conditions: First, both grid cells belong to the globally traversable 3D grid. Second, when a UAV moves from the center point of one grid cell to the center point of another, its physical dimensions do not interfere with any grid cells marked as occupied. The UAV's physical dimensions are represented by a minimum safe bounding box. Connectivity is determined by checking all grid cells traversed by the line segment connecting the center points of two grid cells, ensuring that these traversed grid cells belong to the globally traversable 3D grid. The topological connectivity network is represented by a graph structure. Nodes in the graph correspond to each traversable grid cell in the globally traversable 3D grid. Edges in the graph connect pairs of nodes that satisfy the adjacency and connectivity conditions. The weight of each edge is set to the Euclidean distance between the two nodes. This diagram represents a topological connectivity network that symbolizes the traversable structure of the environment.

[0075] Based on the physical reversibility determination, if the change is deemed irreversible, the connectivity relationships of the traversable paths that have become invalid due to the obstacle state change are removed or modified in the topological connectivity network. The physical reversibility determination identifies the specific spatial overlapping region where the irreversible obstacle state change occurs. For this spatial overlapping region, the specific set of grid cells corresponding to it in the global traversable space 3D grid is located. An update operation is performed in the topological connectivity network. The update operation is divided into removing connectivity relationships and modifying connectivity relationships. Removing connectivity relationships means deleting all topological connectivity network nodes located within the changed region from the graph, along with all edges connected to them. Modifying connectivity relationships means setting the weight of edges that partially cross the boundary of the changed region to a maximum value, such as 10,000 meters. The choice between removal or modification depends on the analysis of the specific state representation before and after the obstacle state change.

[0076] In the updated topological connectivity network, we analyze the additional paths that affect the connectivity of other unchanged areas due to changes in the connection relationships of some travel paths, and identify these additional paths as conflict-derived paths. The analysis process employs the shortest path recalculation and comparison method from graph theory. In the original topological connectivity network before the update, a set of key source and target node pairs are selected. The source and target node pairs represent typical start and end points or key points of the planned flight path of the UAV swarm. For each pair of source and target nodes, Dijkstra's algorithm is used to calculate an optimal path in the original network, which is recorded as the original path. Dijkstra's algorithm finds the sequence of nodes from the source node to the target node that minimizes the sum of the weights of all edges on the path. In the updated topological connectivity network, for the same pair of source and target nodes, Dijkstra's algorithm is used again to calculate a new optimal path, which is recorded as the updated path. The original path and the updated path are then compared. If the updated path differs from the original path, and the updated path needs to detour through some regions or nodes that were not originally optimal choices, then this new detour path is considered an additional path derived from this irreversible change. If multiple different pairs of source and target nodes converge on the calculated updated path and depend on a common alternative channel or region, then that channel or region is identified as a high-priority conflict-derived path. The set of all identified additional paths of this kind is merged and organized, removing duplicate path segments, and outputting a clear list of conflict-derived paths. Each conflict-derived path is defined by a sequence of nodes in the topological connectivity network.

[0077] S5. Based on the scale and criticality of the conflict-derived paths, comprehensively assess the navigation conflict risk level of the spatially overlapping areas on the collaborative route planning of UAV swarms, specifically implemented as follows:

[0078] The number of conflict-derived paths identified is used to obtain a scale metric. The scale metric is based on the conflict-derived path list output from the preceding steps. This list contains a series of path entries, each defined by a sequence of nodes in the topological connectivity network. The scale metric is obtained directly by counting the total number of path entries in the conflict-derived path list. For example, if there are 5 entries in the list, the scale metric is 5. The scale metric is a non-negative integer reflecting the total number of new alternative paths that may cause airway congestion due to irreversible environmental changes.

[0079] For each conflict-derived path, the corresponding conflict-derived path is simulated for removal in the updated topological connectivity network, and the resulting decrease in network connectivity is calculated to obtain a criticality metric. The calculation of the criticality metric requires defining a network connectivity evaluation index. This index quantifies the overall traffic efficiency of the topological connectivity network. One such index is the global average shortest path length. Calculating the global average shortest path length requires predefining a set of critical node pairs. The selection of the critical node pair set is based on the predetermined flight mission or route network structure of the UAV swarm, including the topological connectivity network nodes corresponding to the entry point, exit point, core waypoints of each sub-region, and locations of UAV charging stations or supply drop points within the mission area. The critical node pair set contains multiple node pairs. For each pair of nodes in the critical node pair set, the shortest path length between them is calculated using Dijkstra's algorithm in the updated topological connectivity network. Dijkstra's algorithm finds the sequence of nodes from the starting node to the ending node that minimizes the sum of the weights of all edges on the path; this sum of weights is the shortest path length. If there is no connected path between two nodes, the shortest path length is assigned a maximum value to indicate unreachability. This maximum value is much larger than the length of any actually reachable path in the network, for example, 1,000,000 meters. The shortest path lengths of all node pairs in the critical node pair set are calculated, and the arithmetic mean of these lengths is taken to obtain the global average shortest path length of the current network, denoted as the original average shortest path length.

[0080] Simulated removal of conflict-derived paths refers to temporarily modifying the structure of the updated topological connectivity network when evaluating the criticality of a specific conflict-derived path. Specifically, for the node sequence defined by the conflict-derived path, each node in the sequence is processed sequentially. For each node in the sequence, the node is temporarily deleted from the topological connectivity network graph, along with all edges connected to it. This means that in the simulation, the physical space represented by this node is considered completely impassable. After deleting all nodes of the conflict-derived path, a temporary, weakened network topology is obtained, called the simulated post-removal network. In the simulated post-removal network, for the same set of critical node pairs, Dijkstra's algorithm is recalculated to determine the shortest path length between each pair of nodes. Similarly, for unreachable node pairs, the shortest path length is assigned the same maximum value. The arithmetic mean of the shortest path lengths of all node pairs in the critical node pair set in the simulated post-removal network is calculated and denoted as the post-removal average shortest path length.

[0081] The degree of network connectivity degradation caused by the removal is calculated. This degradation is quantified by comparing the original average shortest path length with the average shortest path length after removal. The degradation is equal to the difference between the original and the removed average shortest path length, divided by the original average shortest path length. The result is a dimensionless proportional value. This proportional value represents the relative increase in the average travel cost of the entire network due to the simulated removal of the conflict-derived path. If removal renders some node pairs unreachable, the post-removal average shortest path length will contain a maximum value, significantly raising the average and thus making the degradation of network connectivity substantial. For each conflict-derived path, the above simulation removal and calculation process is repeated to obtain the corresponding degradation of network connectivity. This value serves as the criticality metric for that conflict-derived path. The criticality metric is a non-negative real number; a larger value indicates a higher structural importance of the path in the entire route network, and a more severe impact of its failure on overall travel efficiency.

[0082] The navigation conflict risk level is determined by querying a pre-defined risk level table based on scale and criticality metrics. This table is a two-dimensional lookup table where the horizontal axis corresponds to the grading intervals of scale metrics, the vertical axis to the grading intervals of criticality metrics, and the intersection of the two axes indicates the final risk level. Risk levels are categorized as low, medium, and high. First, scale and criticality metrics are graded. Scale metric grading is based on quantity: 0-2 items are classified as low, 3-5 as medium, and 6 or more as high. Criticality metric grading is based on the degree of network connectivity degradation: 0-0.1 is classified as low impact, 0.1-0.3 as medium impact, and greater than 0.3 as high impact. The threshold values ​​for these intervals (e.g., 0.1 and 0.3) can be set based on historical task data analysis or simulation results. By analyzing a large amount of UAV route planning simulation data under similar disaster scenarios, we can observe the actual impact of different degrees of network connectivity degradation on task completion time or path conflict probability, thereby determining threshold points with significant differentiation.

[0083] The pre-defined risk level table is populated based on a combination of scale metric and criticality metric classifications. One logic is that when the scale metric classification is low and the criticality metric classification is low impact, the navigation conflict risk level is low. When the scale metric classification is high or the criticality metric classification is high impact, the navigation conflict risk level is high. Other combinations result in a medium risk level. For example, a medium scale metric classification and a low criticality metric classification correspond to a medium risk level; a low scale metric classification but a high criticality metric classification corresponds to a high risk level. To determine the navigation conflict risk level, the actual calculated scale metric value is used to determine its scale metric classification interval, and the actual calculated criticality metric value is used to determine its criticality metric classification interval. These two classification intervals are then used as indices to find the corresponding intersection point in the pre-defined risk level table. The risk level identified by this intersection point is the final determined navigation conflict risk level. The final output navigation conflict risk level is a discrete level label, which directly guides the selection of subsequent route planning adjustment strategies. The entire assessment process combines the quantitative characteristics of the path with the structural importance characteristics, thereby enabling the quantification and classification of navigation conflict risks caused by dynamic environmental changes.

[0084] S6. Based on the navigation conflict risk level, dynamically adjust the cooperative route planning of relevant UAVs in the UAV swarm for areas with spatial overlap, specifically as follows:

[0085] The system selects a corresponding route planning adjustment strategy from a preset adjustment strategy library based on the navigation conflict risk level. The preset adjustment strategy library is a data structure stored in the mission planning computer. It maps different navigation conflict risk levels to a set of specific executable route planning adjustment instructions. Navigation conflict risk levels are categorized as low, medium, and high. For low-risk levels, the route planning adjustment strategies stored in the preset adjustment strategy library include generating airspeed adjustment instructions, which contain a reduced flight speed value. The flight speed value is set based on environmental uncertainty and the UAV's minimum safe flight speed parameter. Environmental uncertainty is quantified by the degree of inconsistency in obstacle states within spatially overlapping areas. The UAV's minimum safe flight speed parameter is read from the UAV performance parameter database. For example, the flight speed value can be set to 50% of the maximum level flight speed of the UAV model read from the UAV performance parameter database, but must not be lower than a preset minimum safe flight speed threshold, such as 1 meter per second. For medium-risk levels, the route planning adjustment strategies stored in the preset adjustment strategy library include generating local path replanning instructions. Local path replanning instructions require planning an alternative flight segment for the UAV that avoids spatially overlapping areas. For high-risk levels, the preset adjustment strategy library stores route planning adjustment strategies, including generating hovering wait commands and global path replanning requests. The hovering wait command instructs the UAV to enter a hovering state. The global path replanning request triggers the mission planning computer to recalculate the coverage partition tasks for the entire UAV swarm.

[0086] Based on the route planning adjustment strategy, flight paths are replanned in the updated topological connectivity network for affected UAVs to bypass spatial overlap areas or conflict-derived paths. The determination of affected UAVs is based on the pre-assigned route mission data for each UAV. The system checks the node sequence corresponding to the pre-planned waypoint sequence of each UAV in the updated topological connectivity network. It determines whether the node sequence intersects with the set of grid cells identified as spatial overlap areas, or whether it overlaps with the node sequence of any path in the conflict-derived path list. If there is an intersection or overlap, the UAV is marked as an affected UAV. For UAVs marked as affected, path replanning is performed according to the selected route planning adjustment strategy. If the strategy is a local path replanning instruction, the path replanning process is as follows: In the updated topological connectivity network, the nearest node corresponding to the UAV's current position is located as the path start point. The node corresponding to the first subsequent waypoint in the UAV's original planned route that is outside the spatial overlap area is located as the path end point. Before path search, all nodes in the updated topological connectivity network that belong to the spatial overlap area are marked as prohibited nodes. The A* graph search algorithm is used to find the shortest path from the starting point to the ending point in the updated topological connectivity network. The A* algorithm uses the Euclidean distance from the current node to the ending point as a heuristic function. The A* algorithm ultimately outputs a new local path consisting of a sequence of nodes. If the strategy is to hover and wait for the command, an emergency return path from the current position to a preset safe assembly point is calculated for the UAV. The emergency return path is also calculated using the A* algorithm in the updated topological connectivity network, with the target node set as the preset safe assembly point. If the A* algorithm cannot find a feasible path from the starting point to the ending point within the preset maximum number of search iterations, the path planning fails. In the case of path planning failure, a straight path to the nearest known safe position is generated as an emergency path. The calculated new flight path consists of a series of three-dimensional coordinate points, which correspond to the center positions of nodes in the updated topological connectivity network.

[0087] The replanned flight path is issued to the relevant UAVs to replace their original cooperative route planning. The issuance process is conducted via a wireless data link within the UAV swarm. For each affected UAV, the mission planning computer generates a specific path update command message. The path update command message includes a header, UAV identifier, path sequence data, and a checksum. The header contains a field indicating that the command type is a route update. The UAV identifier field specifies the unique number of the target UAV receiving the command. The path sequence data field stores all three-dimensional coordinates of the replanned flight path in array form. Each coordinate point includes X, Y, and Z coordinate values. The coordinate values ​​are in meters and reference the global coordinate system. The checksum field ensures the integrity of data transmission. The command message is sent to the target UAV via a wireless ad hoc network. A reliable transmission protocol with acknowledgment and retransmission mechanisms is used. Upon receiving the path update command message, the target UAV's flight control system verifies the checksum field. It then parses the path sequence data field and loads it into the onboard waypoint execution queue. The method for replacing the original cooperative route planning is to clear the existing waypoints in the airborne waypoint execution queue and use the newly resolved path sequence as the new waypoint execution queue. The flight control system starts flight from the first waypoint in the waypoint execution queue. For UAVs that receive a hovering command, the flight control system controls the UAV to enter a hovering state and maintain this state until a new global path update command is received. Once the global mission replanning is complete, the new global path is sent to all UAVs in the same path update command message format, thus completing the dynamic adjustment of the entire fleet's route planning.

[0088] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0089] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0090] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions according to the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. Computer-readable storage media can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0091] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0092] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0093] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0094] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0095] If a function is implemented as a software module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0096] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

[0097] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for collaborative zoned coverage search and rescue using a swarm of emergency rescue drones, characterized in that, include: S1. During the collaborative search and rescue operation of a drone swarm, acquire the local environment map constructed in real time by each drone in the swarm based on visual SLAM. S2. Identify the spatially overlapping areas in each local environment map and determine whether the obstacle status of the spatially overlapping areas is consistent in each local environment map. S3. If the obstacle states in the spatially overlapping areas are inconsistent, then deduce and evaluate the physical reversibility of the transition paths between the inconsistent obstacle states in the spatially overlapping areas. S4. Based on the physical reversibility, extrapolate the potential impact of obstacle state changes on the topological connectivity network on which UAV swarm collaborative route planning depends, and identify the conflict-derived paths that may be caused by obstacle state changes. S5. Based on the scale and criticality of the conflict-derived paths, comprehensively assess the navigation conflict risk level of the spatially overlapping areas on the collaborative route planning of UAV swarms. S6. Based on the navigation conflict risk level, dynamically adjust the collaborative route planning of relevant UAVs in the UAV swarm for areas with spatial overlap.

2. The emergency rescue drone swarm collaborative zoned coverage search and rescue method according to claim 1, characterized in that, During collaborative search and rescue operations using a swarm of drones, local environmental maps constructed in real-time by each drone within the swarm based on visual SLAM are acquired, including: Real-time construction of local environment maps for each UAV based on visual SLAM; Create timestamps for each local environment map associated with the map; Local environment maps with associated map building timestamps are shared and distributed within the drone swarm.

3. The emergency rescue drone swarm collaborative zoned coverage search and rescue method according to claim 1, characterized in that, Identify spatially overlapping regions in each local environment map and determine whether the obstacle states in these overlapping regions are consistent across different local environment maps, including: Receive local environment maps that are shared and distributed within the drone swarm and are associated with map building timestamps; Spatiotemporal coordinate alignment of local environment maps is performed based on map construction timestamps; In the local environment maps after coordinate alignment, spatially overlapping areas with common three-dimensional spatial extent identifiers are identified; Within the identified spatially overlapping areas, the obstacle outlines and spatial occupancy information at corresponding locations in each local environment map are extracted. The obstacle outlines and space occupancy information from different local environment maps are overlaid and compared. If the overlay comparison results show that the obstacles have differences in spatial location, geometry, or occupied area that exceed a preset threshold, then the obstacle status in the spatially overlapping area is determined to be inconsistent.

4. The emergency rescue drone swarm collaborative zoned coverage search and rescue method according to claim 1, characterized in that, If the obstacle states in the spatially overlapping regions are inconsistent, the physical reversibility of the transition paths between the inconsistent obstacle states is deduced and evaluated for the spatially overlapping regions, including: Based on the judgment results of the inconsistency of obstacle states within the spatially overlapping area, determine the specific state representation of the obstacle state before and after the inconsistency is caused. Based on the specific state representation of the obstacle before and after the state change, infer the object attributes or change type involved in the state change. Determine the physical reversibility of the obstacle state change based on the inferred object properties or change type.

5. The emergency rescue drone swarm collaborative zoned coverage search and rescue method according to claim 4, characterized in that, Inferring the object attributes or change types involved in the state change includes: matching the specific state representation of the obstacle before and after the state change with a predefined typical change pattern library, which includes at least one of landslide, rockfall, moving object, dust, and smoke, and determining its physical reversibility based on the matched typical change patterns.

6. The emergency rescue drone swarm collaborative zoned coverage search and rescue method according to claim 1, characterized in that, Based on the physical reversibility of obstacle state changes, the potential impact on the topological connectivity network upon which UAV swarm collaborative route planning relies is extrapolated. This identifies potential conflict-derived paths that may arise due to obstacle state changes, including: Construct a topological connectivity network representing the traversable areas of UAVs and their connections based on all local environment maps; Based on the determination of physical reversibility, if the change is determined to be irreversible, the path connection relationship that has failed due to the change of obstacle state will be removed or modified in the topological connectivity network. In the updated topological connectivity network, we analyze additional paths that affect the connectivity of other unchanged regional airways due to changes in the connection relationships of some travel paths, and identify these additional paths as conflict-derived paths.

7. The emergency rescue drone swarm collaborative zoned coverage search and rescue method according to claim 6, characterized in that, Constructing a topological connectivity network based on all local environment maps to represent the traversable areas of UAVs and their connections includes: merging the spatial voxels not occupied by obstacles in each local environment map to generate a global traversable space 3D mesh; and extracting the connection relationships between adjacent and connected mesh cells in the global traversable space 3D mesh to form a topological connectivity network.

8. The emergency rescue drone swarm collaborative zoned coverage search and rescue method according to claim 1, characterized in that, Based on the scale and criticality of conflict-derived paths, a comprehensive assessment is conducted to determine the navigation conflict risk level of overlapping spatial areas for UAV swarm collaborative route planning, including: The number of conflict-derived paths identified is used to obtain a size measure; For each conflict-derived path, the corresponding conflict-derived path is simulated to be removed in the updated topological connectivity network, and the resulting decrease in network connectivity is calculated to obtain a critical metric. The navigation conflict risk level is determined by querying a pre-defined risk level table based on size and criticality metrics.

9. A method for collaborative zoned coverage search and rescue using an emergency rescue drone swarm according to claim 1, characterized in that, Based on the navigation conflict risk level, dynamically adjust the cooperative route planning of relevant drones in the drone swarm for areas of spatial overlap, including: Select the corresponding route planning adjustment strategy from the preset adjustment strategy library based on the navigation conflict risk level; Based on the route planning adjustment strategy, flight paths for affected UAVs are replanned in the updated topological connectivity network to bypass spatially overlapping areas or conflict-derived paths; The replanned flight paths will be issued to the relevant drones to replace their original collaborative flight path plans.

10. A method for collaborative zoned coverage search and rescue using an emergency rescue drone swarm according to claim 9, characterized in that, The preset adjustment strategy library stores route planning adjustment strategies corresponding to different navigation conflict risk levels. The route planning adjustment strategies include: for low risk levels, instructing relevant UAVs to maintain their original routes but reduce their flight speed; for medium risk levels, instructing relevant UAVs to replan local paths to avoid spatial overlap areas; and for high risk levels, instructing relevant UAVs to pause their progress and wait for a reassigned global path.

Images