An unmanned aerial vehicle landing site deployment point generation method, system, medium and product

Abstract

The application discloses a UAV landing site deployment point generation method and system, a medium and a product, relates to the technical field of UAVs, and comprises the following steps: taking the maximum service radius as the inter-cluster distance termination threshold of a full connection hierarchical clustering, and using a candidate center point subset weighting solving mechanism, so that the number of landing sites can be adaptively determined according to the service radius hard constraint, and the demand weight grid is coupled with the historical order space-time density to drive center point optimization, so that in the case of a large number of candidate points, the landing site point generation is realized without exceeding the radius, the demand coverage cost is minimized, and the engineering deployability and airspace availability of the site selection result are improved.

Description

Technical Field

[0001] This application relates to the field of unmanned aerial vehicle (UAV) technology, and in particular to a method, system, medium, and product for generating UAV take-off and landing site deployment locations. Background Technology

[0002] With the accelerated development of the low-altitude economy and urban air mobility (UAM), logistics drones are playing an increasingly important role in scenarios such as last-mile delivery, medical emergency transport, and short-haul freight. As the infrastructure of the drone logistics network, the deployment locations of take-off and landing sites directly determine transportation efficiency, airspace utilization, and service coverage quality. Therefore, how to scientifically deploy drone take-off and landing sites under complex urban geographical environments and airspace constraints has become a key issue that urgently needs to be studied and solved in the field of UAM infrastructure planning.

[0003] In related technologies, the commonly used methods for selecting UAV take-off and landing sites mainly rely on traditional K-means clustering or its variants. Taking the coordinates of the points of interest as input, the candidate points are spatially grouped by a preset number of clusters, and the center point is solved by minimizing the distance within each group. Some solutions introduce simple buffer analysis on this basis to perform single-level filtering on the boundaries of military restricted areas or nature reserves.

[0004] However, existing technologies require pre-specifying the number of clusters K, making it difficult to adapt to hard constraints on the service radius. This can easily lead to situations where the farthest point within a cluster exceeds the drone's range. Furthermore, relying solely on a single dimension of spatial distance results in a large number of output site selection points falling into practically undeployable areas. Further improvements are needed to address these issues. Summary of the Invention

[0005] This application provides a method, system, medium, and product for generating UAV take-off and landing site deployment locations. It uses a fully connected hierarchical clustering mechanism with the maximum service radius as the inter-cluster distance termination threshold and a weighted solution mechanism for candidate center point subsets to adaptively determine the number of take-off and landing sites based on the hard constraint of the service radius. It also uses a demand weight grid coupled with the spatiotemporal density of historical orders to drive the optimization of center points. This enables the generation of take-off and landing site locations with the radius not exceeding the limit and the demand coverage cost to be minimized in large-scale candidate point scenarios, thereby improving the engineering deployability and airspace availability of the site selection results.

[0006] In a first aspect, this application provides a method for generating UAV take-off and landing site deployment locations, the method comprising:

[0007] Multi-source geographic information data of the target area is acquired to obtain point of interest vector layers, airspace control data, land use classification raster, and demand weight raster; the airspace control data includes polygonal boundaries of restricted areas, restricted areas, and danger zones; the demand weight raster is generated from the spatiotemporal distribution density of historical orders and is used to characterize the demand intensity of each geographic unit.

[0008] Based on the airspace control data and the land use classification raster, spatial topology clipping is performed on the point of interest vector layer to obtain a candidate point set.

[0009] Calculate the spherical distance between any two points in the candidate point set to obtain a distance matrix. Use the preset maximum service radius as the inter-cluster distance termination threshold, and perform fully connected hierarchical clustering on the distance matrix to obtain several initial clusters.

[0010] Within each of the initial clusters, points whose distance to all other points within the cluster does not exceed the maximum service radius are selected to obtain a subset of candidate center points;

[0011] Based on the demand weight grid, the weighted total distance from each point in the candidate center point subset to all points in its cluster is calculated to obtain the weighted cost value.

[0012] The candidate center point with the smallest weighted cost value in the candidate center point subset is taken as the take-off and landing anchor point of the corresponding cluster, thus obtaining the anchor point set; the anchor point set is sent to the UAV traffic management system as the take-off and landing anchor point parameters for route planning in the UAV traffic management system.

[0013] In the above embodiments, this application first obtains multi-source geographic information such as point-of-interest vector layers, airspace control data, land use classification rasters, and demand weight rasters of the target area through a data interface, using this as the data basis for site selection. Then, based on airspace control and land use classification, spatial topology pruning is performed on the points of interest to obtain a set of candidate points. Subsequently, a distance matrix is ​​constructed based on Haversine spherical distance, and fully connected hierarchical clustering is performed with the maximum service radius as the inter-cluster distance termination threshold. This adaptively generates several initial clusters without pre-setting a K value. Then, within each cluster, points whose distance to other points does not exceed the service radius are selected to form a subset of candidate center points. The weighted cost value is obtained by calculating the weighted distance sum in combination with the demand weight rasters. Finally, the anchor point result corresponding to the weighted cost value is sent to the UAV traffic management system as the take-off and landing anchor point parameters for flight path planning. This not only solves the technical bottleneck of traditional methods requiring preset K values ​​and difficulty in guaranteeing hard radius constraints, but also improves the engineering deployability and airspace availability of the site selection results through multi-source geographic constraint fusion and interface coupling with the UAV traffic management system.

[0014] In some embodiments, the method further includes preloading an ecological protection red line layer and an electromagnetic field strength raster; and performing spatial topology clipping on the point of interest vector layer based on the airspace control data and the land use classification raster to obtain a candidate point set, including the following steps:

[0015] Perform a spatial difference operation between the ecological protection red line layer and the point of interest vector layer, and remove the points in the point of interest vector layer that fall into the ecological protection red line layer to obtain the first set of filtered points.

[0016] Based on the polygonal boundaries of restricted areas, restricted zones, and dangerous zones in the airspace control data, a point-to-surface inclusion judgment is performed on the first set of filtered points, and points in the first set of filtered points that fall within the polygonal boundaries are removed to obtain the second set of filtered points.

[0017] Based on a preset electromagnetic threshold, points in the second set of filter points that fall within the electromagnetic field strength grid and whose pixel intensity exceeds the electromagnetic threshold are removed, resulting in a third set of filter points.

[0018] Based on the land use classification grid, preset land use types are retained in the third filter point set to obtain the candidate point set. The preset land use types include commercial land, industrial land, transportation land, residential land, and public management and service land.

[0019] In the above embodiments, this application preloads an ecological protection red line layer and an electromagnetic field strength grid. It then sequentially performs a spatial difference operation on the interest point vector layer to obtain a first set of filtered points. Next, it performs a point-to-surface inclusion judgment on the first set of filtered points to obtain a second set of filtered points. Finally, it removes points from the second set of filtered points that fall into high-field-strength pixels based on electromagnetic thresholds to obtain a third set of filtered points. Finally, it retains five preset land use types—commercial, industrial, transportation, residential, and public management and services—based on a land use classification grid, thus forming a candidate point set that has undergone multiple compliance screenings. This not only solves the problem of single-dimensional filtering easily generating undeployable points but also significantly reduces subsequent approval and operational risks through a chain-like filtering mechanism, improving the compliance and feasibility of the site selection results.

[0020] In some embodiments, the method further includes the steps of preloading a digital elevation raster layer and building height data; and after obtaining the candidate point set by retaining a preset land use type in the third filter point set based on the land use classification raster, the method further includes:

[0021] Based on the digital elevation raster layer, the terrain slope of each candidate point in the candidate point set is calculated to obtain the slope value corresponding to each candidate point.

[0022] Based on a preset slope threshold, candidate points in the candidate point set whose slope values ​​are greater than the preset slope threshold are removed to obtain a first set of filtered points;

[0023] Based on the building height data, detect whether there are buildings with heights exceeding a preset occlusion threshold within a preset clearance range above each candidate point in the first set of filter points, and remove candidate points containing such buildings to obtain the second set of filter points;

[0024] The second set of filtered points is used as the updated set of candidate points.

[0025] In the above embodiments, after the candidate point set is generated, this application preloads a digital elevation raster layer and building height data. First, it calculates the terrain slope of each candidate point based on the digital elevation raster to obtain the slope value. Then, it removes points with excessively large slopes based on a preset slope threshold to obtain a first set of filtered points. Next, it detects whether there are buildings with heights exceeding the obstruction threshold within a preset clearance range above each candidate point based on the building height data. Points with obstruction are removed to obtain a second set of filtered points, and the candidate point set is updated with the second set of filtered points. This not only solves the problem of missing deployment verification of candidate points in the vertical dimension, but also provides a terrain-suitable and airspace-safe input set for subsequent clustering solutions, improving the engineering construction feasibility and flight safety margin of the take-off and landing site selection results.

[0026] In some embodiments, prior to the step of acquiring multi-source geographic information data of the target area, the method further includes:

[0027] Based on the business type, the deployment requirements are divided into urban last-mile delivery scenarios, medical supply delivery scenarios, and short-distance logistics transportation scenarios, resulting in multiple deployment scenarios;

[0028] For each of the multiple deployment scenarios, the corresponding maximum service radius, candidate point source category, and demand weight grid calculation method are set to obtain the scenario parameter set corresponding to each deployment scenario;

[0029] Based on the scenario parameter set, for each deployment scenario, the scenario parameter set corresponding to the deployment scenario is used as the input parameter to obtain the take-off and landing field configuration data corresponding to each deployment scenario.

[0030] In the above embodiments, since different logistics businesses have significant differences in the service radius, candidate point source, and demand intensity representation methods of take-off and landing yards, if a unified parameter is used for full-domain site selection calculation, it will lead to overly dense or sparse take-off and landing yards in some scenarios. For example, medical emergency scenarios require extremely short service radii to ensure timeliness, while short-distance logistics scenarios need to cover a large geographical span. A unified parameter will not be able to take into account the needs of both ends. Before obtaining multi-source geographic information, this application first divides the deployment requirements into three deployment scenarios based on business type: urban last-mile delivery, medical supply delivery, and short-distance logistics transportation. For each scenario, a corresponding maximum service radius, candidate point source category, and demand weight grid calculation method are set to form a scenario parameter set. The take-off and landing yard generation process is executed separately using each scenario parameter set as input parameters to obtain the take-off and landing yard configuration data corresponding to each scenario. By decoupling the scenario parameters, the same algorithm can be flexibly adapted to multiple business parallelism, improving the versatility of the method and the service matching accuracy.

[0031] In some embodiments, after calculating the weighted sum of distances from each point in the candidate centroid subset to all points within its cluster based on the demand weight grid, and obtaining the weighted cost value, the method further includes:

[0032] Within the subset of candidate centroids of each initial cluster, the candidate centroids are sorted in ascending order of the weighted cost value to obtain the sequence of candidate centroids for each cluster.

[0033] The first candidate center point of each candidate center point sequence is selected as the current anchor point of the corresponding cluster. Based on the preset take-off and landing corridor geometric parameters and the prevailing wind direction data at the location of the current anchor point, a three-dimensional take-off and landing corridor envelope corresponding to the current anchor point is constructed.

[0034] Perform pairwise spatial intersection detection on the three-dimensional take-off and landing corridor envelopes corresponding to different clusters to obtain conflicting cluster pairs;

[0035] For each conflict cluster pair, within the cluster where the weighted cost value is greater than a preset threshold, the next candidate center point is selected sequentially as the updated current anchor point according to the order of the candidate center point sequence. The construction of the three-dimensional take-off and landing corridor envelope and the spatial intersection detection are re-executed until the three-dimensional take-off and landing corridor envelopes do not intersect with each other, thus obtaining the optimized anchor point set.

[0036] The optimized set of anchor points is sent to the UAV traffic management system as the updated set of anchor points, serving as the take-off and landing anchor point parameters.

[0037] In the above embodiments, conventional techniques do not consider the three-dimensional spatial corridors formed during the takeoff and landing of UAVs. The corridor envelopes of adjacent anchor points may still overlap in the height direction. For example, the anchor points of two adjacent office building rooftops may meet the safe distance on the plane, but their takeoff and landing corridors may overlap in the air due to the shift in prevailing wind direction, causing potential flight path conflicts. After obtaining the weighted cost value, this application first sorts the candidate center points in the subset of candidate center points in each cluster according to the weighted cost value from smallest to largest to form a candidate center point sequence. The first candidate point is taken as the current anchor point and its three-dimensional structure is constructed by combining the geometric parameters of the takeoff and landing corridor and the prevailing wind direction data. The takeoff and landing corridor envelope is used to perform spatial intersection detection and identify conflicting cluster pairs for each pair of envelopes from different clusters. For clusters with a weighted cost value greater than a preset threshold, the current anchor point is replaced in sequence and the envelope is reconstructed and detected until all corridors are mutually non-intersecting, resulting in an optimized set of anchor points, which is then sent to the UAV traffic management system. This solves the technical blind spot of two-dimensional site selection being unable to perceive vertical space conflicts. At the same time, through adaptive iterative replacement of candidate center point sequences, three-dimensional space conflicts are eliminated while ensuring minimum cost of demand coverage, thus improving the route safety margin during parallel operation of multiple anchor points.

[0038] In some embodiments, after replacing the anchor point result corresponding to the weighted cost value with the optimized anchor point set as the take-off and landing anchor point parameter sent to the UAV traffic management system, the method further includes:

[0039] Receive temporary airspace control updates pushed by the UAV traffic management system, and obtain the dynamic control polygon and the effective time window corresponding to the dynamic control polygon;

[0040] Based on the preset spatial buffer distance, the dynamic control polygon is buffered and expanded to obtain the influence range polygon.

[0041] In the optimized set of anchor points, anchor points that fall within the polygon of the influence range are selected to obtain the set of affected anchor points.

[0042] Using the initial cluster corresponding to the affected anchor point set as the local reconstruction range, the dynamic control polygon is incorporated into the spatial control data, and the spatial topology clipping, the fully connected hierarchical clustering, and the weighted cost value are recalculated for the points within the local reconstruction range to obtain the local replacement anchor point set within the effective time window;

[0043] Within the effective time window, the local replacement anchor point set replaces the affected anchor point set, which is then used as the updated take-off and landing anchor point parameters sent to the UAV traffic management system.

[0044] In the above embodiments, since the issued take-off and landing site anchor point parameters cannot respond immediately when encountering temporary airspace control, resolving the entire target area globally would waste computing resources and cause unnecessary deployment changes in unaffected areas. For example, if a city is closed to flights due to a temporary airshow, recalculating the entire area would result in the redistribution of already operating anchor points tens of kilometers away from the event area. In this application, after the optimized anchor point set is issued, the system receives the temporary airspace control update pushed by the UAV traffic management system to obtain the dynamic control polygon and the effective time window. The polygon is buffered and expanded according to the preset airspace buffer distance to obtain the affected range polygon. Anchor points falling within the affected range are selected from the optimized anchor point set to form the affected anchor point set. The initial cluster corresponding to the affected anchor point is used as the local reconstruction range. After incorporating the dynamic control polygon into the airspace control data, spatial topology pruning, hierarchical clustering, and weighted cost value calculation are re-performed only on the local range to obtain the local replacement anchor point set within the effective time window to replace the affected anchor points. This solves the problem of computing power waste and service disturbance caused by global recalculation under dynamic control and improves the system's real-time adaptability in dynamic airspace environments.

[0045] In some embodiments, the method further includes:

[0046] The system receives the actual number of takeoffs and landings and the average queuing time of each anchor point within a preset statistical period from the UAV traffic management system, and obtains the corresponding operational load index for each anchor point.

[0047] Based on a preset throughput threshold, anchor points where the operating load index exceeds the throughput threshold are selected to obtain an overload anchor point set.

[0048] For each overload anchor point in the set of overload anchor points, the cluster in which the overload anchor point is located is divided into two sub-clusters based on the weight distribution of the demand weight grid within the cluster in which the overload anchor point is located, thus obtaining a set of secondary split clusters.

[0049] For each sub-cluster in the secondary split cluster set, the screening of the candidate centroid subset, the calculation of the weighted cost value, and the construction of the three-dimensional take-off and landing corridor envelope are re-executed to obtain the split anchor point corresponding to the sub-cluster.

[0050] The overload anchor point is replaced with the split anchor point, the optimized anchor point set is updated, and the updated optimized anchor point set is sent to the UAV traffic management system.

[0051] In the above embodiments, this application receives the actual takeoff and landing frequency and average queuing time of each anchor point within a preset statistical period from the UAV traffic management system to form an operational load index. Based on a preset throughput threshold, an overloaded anchor point set is selected. For each overloaded anchor point, the cluster is divided into two sub-clusters based on the weight distribution of the demand weight grid within its cluster. For each sub-cluster, the candidate center point subset selection, weighted cost value calculation, and three-dimensional takeoff and landing corridor envelope construction are re-executed to obtain split anchor points. The split anchor points replace the overloaded anchor points, and the updated and optimized anchor point set is sent back. This not only solves the problem that static site selection cannot respond to traffic changes during operation, but also realizes on-demand expansion of hotspot areas through a weight distribution-driven cluster splitting mechanism, improving the throughput capacity and service stability of the takeoff and landing field network under peak load.

[0052] In a second aspect, embodiments of this application provide a drone take-off and landing site deployment point generation system, comprising: 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 including computer instructions, and the one or more processors call the computer instructions to cause the system to perform the method described in the first aspect and any possible implementation thereof.

[0053] Thirdly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a system, cause the system to perform the method described in the first aspect and any possible implementation thereof.

[0054] Fourthly, embodiments of this application provide a computer program product that, when run on a system, causes the system to execute the method described in any possible implementation of the first aspect.

[0055] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0056] 1. By using a fully connected hierarchical clustering and candidate center point subset weighted solution mechanism with the maximum service radius as the inter-cluster distance termination threshold, the limitation of traditional K-means algorithms that must preset the K value is eliminated. This allows the number of take-off and landing fields to be adaptively determined based on the hard constraint of the service radius. Furthermore, the optimization of center points is driven by demand weight grid coupling with historical order spatiotemporal density. This enables the generation of take-off and landing field locations with the radius not exceeding the limit and the demand coverage cost to be minimized in large-scale candidate point scenarios, thereby improving the engineering deployability and airspace availability of the site selection results.

[0057] 2. By introducing the construction of a three-dimensional take-off and landing corridor envelope and pairwise spatial intersection detection on the basis of the optimal plane anchor point, and iteratively replacing the candidate center point sequence according to the weighted cost value, the dimensional leap of take-off and landing site selection from two-dimensional plane to three-dimensional airspace is realized. Under the premise of ensuring the suboptimal cost of demand coverage, the airspace conflict between multiple anchor points is adaptively resolved, filling the technical blind spot of traditional facility site selection methods that cannot perceive the vertical dimension corridor overlap, and greatly improving the route safety margin when multiple anchor points are running in parallel.

[0058] 3. By using the temporary airspace control push and operational load feedback from the UAV traffic management system as feedback inputs, a dynamic closed loop is constructed. This allows the take-off and landing field layout to be precisely reconstructed in local areas within the temporary control effective time window, and to perform secondary clustering and splitting based on weight distribution at overload anchor points. This avoids the waste of computing power in full-domain recalculation and enables on-demand expansion of hotspot areas. It solves the pain point that static site selection methods cannot respond to dynamic changes in airspace and traffic fluctuations, and improves the real-time adaptability and throughput stability of the take-off and landing field network in UAM scenarios. Attached Figure Description

[0059] Figure 1 This is a flowchart illustrating a method for generating UAV take-off and landing site deployment locations in an embodiment of this application;

[0060] Figure 2 This is another flowchart illustrating a method for generating UAV take-off and landing site deployment locations in an embodiment of this application;

[0061] Figure 3 This is a schematic diagram of the physical device structure of a drone take-off and landing site deployment point generation system according to an embodiment of this application. Detailed Implementation

[0062] 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.

[0063] 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.

[0064] To facilitate understanding, the application scenarios of the embodiments of this application are described below.

[0065] In the field of urban air traffic infrastructure planning, with the development of the low-altitude economy and the large-scale deployment of logistics drones in last-mile delivery, medical emergency and short-haul freight, the deployment location of take-off and landing sites has become a key factor in determining the efficiency of drone logistics networks and airspace utilization.

[0066] In related technologies, traditional K-means clustering or its variants are typically used to group candidate points and then solve for the center coordinates. This type of scheme requires pre-specifying the number of clusters K, which is difficult to adapt to the hard constraint of service radius. The farthest point within a cluster often exceeds the drone's endurance. Moreover, it only relies on a single-dimensional planar distance and does not integrate multi-source constraints such as airspace, ecology, electromagnetic, land use, terrain, and airspace clearance. As a result, a large number of output points fall into areas that are actually undeployable. Furthermore, the output results remain at the static coordinate level and lack effective connection with downstream airspace management and operational feedback.

[0067] This application is primarily applied to the automated deployment planning of urban-level drone take-off and landing sites under the UAM (Unmanned Aerial Vehicle) framework, covering multiple business models such as urban last-mile food delivery, medical emergency supply transportation, and short-distance inter-regional express delivery. In these application scenarios, the site selection results not only need to be compliant and deployable, and match the drone's endurance parameters, but also need to be compatible with temporary airspace control and throughput fluctuations during operation. To address the aforementioned technical issues, this application provides a method for generating drone take-off and landing site deployment locations. An embodiment is described below, combined with… Figure 1 This application describes a method for generating UAV take-off and landing site deployment locations in its embodiments.

[0068] The following describes the process of the method provided in this implementation. Please refer to [link / reference]. Figure 1 This is a flowchart illustrating a method for generating UAV take-off and landing site deployment locations in an embodiment of this application.

[0069] S101. Obtain multi-source geographic information data of the target area to obtain point of interest vector layer, airspace control data, land use classification raster and demand weight raster.

[0070] Among them, point-of-interest vector layer refers to a set of urban geographic location data recorded in the form of point features; airspace control data refers to an airspace control dataset released by the civil aviation administration department that includes the polygonal boundaries of restricted areas, prohibited areas, and dangerous areas and their attributes; land use classification raster refers to raster data that labels the surface raster units with attributes according to categories such as commercial, industrial, transportation, residential, and public management and services; and demand weight raster refers to raster data that uses raster units as carriers and is generated from the spatiotemporal distribution density of historical orders, used to characterize the intensity of logistics demand in each geographic unit.

[0071] Specifically, the system first retrieves the administrative boundaries of the target area and defines the data retrieval scope through the data interface; then, it downloads the above four types of data from public geographic information platforms or authorized commercial data sources according to the preset data source priority; next, it performs coordinate system one and spatial alignment processing on each layer; finally, it loads the aligned layers into memory in standard raster and vector formats for subsequent calls.

[0072] In some embodiments, the loading of the aforementioned multi-source geographic information data can be accomplished using desktop geographic information system software (such as ArcGIS Pro). This type of software has integrated 2D and 3D design, spatial analysis, and visualization capabilities, facilitating preliminary screening of points and display of results. The required boundary, SHP dataset, POI dataset, DEM terrain elevation data, and statistical data can be purchased or downloaded from publicly available geographic information platforms. In addition, the demand weight raster, besides being generated from the spatiotemporal distribution density of historical orders, can be further integrated with population density raster and GDP distribution raster for weighted overlay. That is, the target area is divided into geographic units according to a preset grid size, and the population, GDP output, and historical order quantity of each geographic unit are read, normalized, weighted, and synthesized, and written into the demand weight raster to simultaneously achieve the optimization goals of maximum population coverage, maximum GDP coverage, and maximum area coverage in subsequent steps.

[0073] S102. Based on airspace control data and land use classification raster, perform spatial topology clipping on the point of interest vector layer to obtain a set of candidate points.

[0074] Spatial topology clipping refers to the operation of filtering and eliminating point feature sets based on topological relationships such as inclusion, intersection, and difference between spatial features.

[0075] Specifically, the system first uses the restricted and dangerous zone polygons in the airspace control data as boundaries to perform point-to-surface inclusion judgment on points of interest and removes points that fall within the polygons; then, it uses the land use classification raster as constraints to retain points belonging to the preset land use type; next, it performs spatial index reconstruction on the removed point set; finally, it outputs the filtered point set as a candidate point set.

[0076] Furthermore, the system will introduce joint filtering of ecological compliance and electromagnetic compliance when performing spatial topology pruning. That is, the system will preload the ecological protection red line layer and the electromagnetic field strength grid, and then sequentially perform the ecological red line spatial difference operation, the inclusion judgment of prohibited and restricted dangerous areas, the electromagnetic threshold filtering, and the land use type retention to obtain a set of candidate points.

[0077] Among them, the ecological protection red line layer refers to the ecological protection space control vector map layer delineated and publicly released by the national or local ecological and environmental authorities, and the electromagnetic field strength grid refers to the grid data that records the intensity of spatial electromagnetic radiation using pixels as a carrier.

[0078] Specifically, the system first performs a spatial difference operation on the ecological protection red line layer and the point of interest vector layer, and removes points falling within the ecological red line to obtain the first filter point set; then, it performs a point-to-surface inclusion judgment on the first filter point set using the restricted and dangerous area polygons in the airspace control data, and removes points falling within the polygons to obtain the second filter point set; next, it reads the field strength value of the pixel where each point is located in the electromagnetic field strength raster and removes points that are higher than the preset electromagnetic threshold to obtain the third filter point set; finally, based on the land use classification raster, it retains only points in the third filter point set that belong to the five land use types of commercial, industrial, transportation, residential, and public management and service, to obtain the candidate point set.

[0079] In other embodiments, to further improve the actual deployability of the candidate point set, the system can introduce two additional filtering steps—air route reference filtering and ground restricted area filtering—on top of the aforementioned four-fold filtering based on ecological red lines, restricted and dangerous areas, electromagnetic thresholds, and land use types. Specifically, the system preloads air route data and ground restricted area vector layers for the target city. The air route data records the geometric elements of the pilot areas and flight routes already in operation, undergoing trial operation, and planned operation in the target city. The ground restricted area vector layer records the vector boundaries of military restricted areas, prisons, large energy facilities, confidential locations, and their surrounding sensitive buffer zones. The system first performs spatial proximity calculations on the air route data and the candidate point set, assigning priority bonuses to candidate points falling within the vicinity of the corridors of already in operation, undergoing trial operation, and planned operation routes, and assigning priority deductions to candidate points that significantly deviate from the existing route system. Then, it performs point-to-surface inclusion judgments on the ground restricted area vector layer and the candidate point set, eliminating candidate points falling within military restricted areas, prisons, large energy facilities, confidential locations, and their sensitive buffer zones. The point set after the above additional filtering is output as the updated candidate point set. Therefore, combining the original four constraints of land use, ecological protection, surrounding airspace, and electromagnetic environment of S102, as well as the two constraints of air routes and ground restricted areas in this supplementary paragraph, together constitute six strong constraints that run through the candidate site selection stage.

[0080] In some embodiments, the system will further verify the vertical deployability after the candidate point set is generated. That is, the system will preload the digital elevation raster layer and building height data, calculate the slope value of each candidate point and remove points that exceed the threshold to obtain the first set of filtered points, then detect whether there are any super-tall obstructing buildings within the preset clearance range above each candidate point, remove the obstructing points to obtain the second set of filtered points, and update the candidate point set with the second set of filtered points.

[0081] Among them, the digital elevation raster layer refers to raster data that represents the ground elevation with pixel values, and the building height data refers to the dataset that records the vector outline of a building and its roof height attribute.

[0082] Specifically, the system first calculates the elevation gradient of each candidate point and its surrounding neighborhood based on the digital elevation grid to form a slope value; then, points with slope values ​​greater than a preset slope threshold are removed to form a first set of filtered points; next, the system searches for buildings within a preset clearance cylinder range above each candidate point in the building height data and determines whether their height exceeds a preset shading threshold; finally, points with excessively tall shading buildings are removed to form a second set of filtered points, and the candidate point set is updated accordingly.

[0083] S103. Calculate the spherical distance between any two points in the candidate point set to obtain the distance matrix. Use the preset maximum service radius as the inter-cluster distance termination threshold, and perform fully connected hierarchical clustering on the distance matrix to obtain several initial clusters.

[0084] Among them, spherical distance refers to the arc length of the great circle between two points on the Earth's ellipsoid calculated according to the Havesing formula; distance matrix refers to a two-dimensional symmetric matrix with candidate point numbers as row and column indices and spherical distances between two points as elements; maximum service radius refers to the maximum planar distance that a single anchor point can cover, determined based on the selected UAV model's range, payload capacity, and cruise speed; and fully connected hierarchical clustering refers to an agglomerative clustering method that uses the distance between the farthest points in the cluster as the inter-cluster distance metric and gradually merges clusters from the bottom up.

[0085] Specifically, the system first uses the latitude and longitude pairs of candidate points as input to calculate the pairwise spherical distances in batches according to the Havesing formula and fills them into the distance matrix; then, it initializes each candidate point as an independent cluster and starts fully connected hierarchical clustering with the maximum service radius as the termination threshold for inter-cluster distance; next, in each merging iteration, it selects the two clusters with the smallest distance between the farthest point pairs and that does not exceed the termination threshold to perform merging; finally, it stops merging when the distance between the farthest point pairs of any two clusters exceeds the termination threshold, and outputs all merging results as several initial clusters.

[0086] It should be noted that traditional K-means algorithms rely on manually specifying the K value, which makes it difficult to balance coverage and radius constraints in large-scale urban candidate point scenarios. This embodiment uses the maximum service radius as the termination threshold for inter-cluster distance in hierarchical clustering, so that the number of clusters is adaptively determined by the threshold and the spatial distribution of candidate points. The spherical distance between any two points in any initial cluster does not exceed the maximum service radius. The value of the maximum service radius needs to be determined comprehensively based on the specific drone model's range, payload capacity, cruising speed, and reserved safety margin. For example, for a logistics rotorcraft drone with a range of about 15 kilometers and a cruising speed of about 60 kilometers per hour, considering round-trip requirements and safety margin, the maximum service radius can be taken as 500 to 1000 meters. For medical emergency scenarios with extremely high time requirements, it can be further compressed to within 200 meters. For short-distance logistics transportation scenarios, it can be extended to more than 1000 meters. The distance between clusters is measured using a fully connected (CompleteLinkage) approach. Compared to single-connected or average-connected approaches, this fundamentally ensures that the distance between the farthest point pairs within a cluster does not exceed a termination threshold, thus implementing a hard constraint on the service radius. The input to this step is a list of candidate point latitude and longitude coordinates and a maximum service radius parameter. The output is the cluster label to which each candidate point belongs and the number of clusters.

[0087] In some embodiments, the system as a whole constitutes a hierarchical radius-weighted K-medoids algorithm, the core of which is a two-stage solution framework of "hierarchical clustering to determine the cluster structure + weighted selection of intra-cluster centroids". In this algorithm, the S103 Havesing formula can be specifically expressed as: ;in, =6371km is the Earth's radius , , , Points and latitude and longitude and These represent the latitude and longitude differences between the two points, respectively. Compared to the traditional K-means algorithm's strong dependence on the K value, this algorithm has the following characteristics: First, it provides dual constraint guarantees, determining the minimum number of clusters through hierarchical clustering while ensuring service quality through strict distance constraints; second, it employs population-weighted optimization, considering the population size of each point to make the selection of the center point more in line with actual service needs, and using equal weights to substitute when population data is lacking still yields the optimal distance location; third, it provides multiple alternative solutions, offering several alternative center points for each cluster to enhance the feasibility of the solution; fourth, it includes an anomaly handling mechanism, providing an approximate optimal solution for situations where strict constraints cannot be met; and fifth, it offers comprehensive visualization, generating a clustering result map that intuitively displays the spatial distribution and service coverage.

[0088] S104. Within each cluster of the initial clustering, select points whose distance to other points within the cluster does not exceed the maximum service radius to obtain a subset of candidate center points.

[0089] The candidate center point subset refers to the set of all candidate points within a single cluster that can serve as center anchor points and guarantee that all other points within the cluster are within its service range.

[0090] Specifically, the system first iterates through each initial cluster and extracts all candidate points within the cluster; then it checks the spherical distance of each candidate point within the cluster to all other points in the same cluster; next, it filters out all points whose distances do not exceed the maximum service radius; finally, it summarizes the filtered points to form a subset of candidate center points for the cluster.

[0091] It should be noted that the termination condition of hierarchical clustering guarantees that the distance between any two points within a cluster does not exceed the maximum service radius, but does not guarantee that there exists a point within a cluster that can independently cover all other points. This step introduces a selection process for candidate centroid subsets, further strengthening the clustering property of "pairwise distance within a cluster ≤ R" into an anchor point reachability constraint of "there exists a point whose distance to all other points within the cluster is ≤ R". If the candidate centroid subset of a cluster is empty, it indicates that the cluster cannot be completely covered by a single anchor point. In this case, the system can output an exception handling result through a preset relaxation mechanism (such as resolving the maximum service radius after relaxing it by a relaxation coefficient, or performing a binary split within the cluster). The input of this step is the cluster label and the distance matrix, and the output is a list of candidate centroid numbers for each cluster.

[0092] S105. Based on the demand weight grid, calculate the sum of weighted distances from each point in the candidate center point subset to all points in its cluster to obtain the weighted cost value.

[0093] The weighted distance sum refers to the scalar value obtained by using the weight value of the demand weight grid in the grid cell where each candidate point is located as the distance coefficient and weighting the spherical distance from the candidate center point to all points in the cluster. The weighted cost value refers to the cost evaluation result formed by the above weighted distance sum in the candidate center point subset.

[0094] Specifically, the system first traverses each candidate point in the candidate center point subset according to clusters; then, for each candidate point, it reads the weight value of the grid cell in which each point in the cluster is located from the demand weight grid; next, it multiplies each weight value by the spherical distance from that point to the candidate center point and sums them within the cluster to obtain the weighted total distance of the candidate center point; finally, it summarizes the weighted total distance of all candidate center points in each cluster to form the weighted cost value of that cluster.

[0095] It should be noted that the introduction of weighted cost values ​​expands the selection of takeoff and landing sites from simple geometric minimization to demand coverage minimization coupled with the intensity of actual logistics demand. The weight values ​​of the demand weight grid are derived from the spatiotemporal distribution density of historical orders. This can be obtained by dividing the target area into geographical units according to a preset grid size, statistically analyzing the number of orders and the variance of order time distribution in each unit within a preset historical time window, and then normalizing and merging these values ​​before writing them into the grid. Different weight generation methods can be adopted for different business scenarios. For example, in the medical scenario, the density of emergency calls is the main factor; in the last-mile delivery scenario, the density of takeout orders is the main factor; and in the short-distance logistics scenario, the volume of parcels received and dispatched is the main factor. The input of this step is a subset of candidate center points, a distance matrix, and a demand weight grid. The output is a list of weighted cost values ​​corresponding to each candidate center point in each cluster. This weighted cost value not only serves as the basis for selecting the main center point (taking the one with the smallest weighted cost value), but also provides the sorting basis for the candidate center point sequence for the three-dimensional corridor conflict resolution in the subsequent second implementation scheme, constituting a core evaluation indicator throughout the entire scheme process.

[0096] In some embodiments, the calculation of the weighted distance sum in S105 can be further extended to a composite weighted form that simultaneously satisfies the three optimization principles of maximum population coverage, maximum GDP coverage, and maximum area coverage. The maximum population coverage principle requires site selection to prioritize densely populated areas to ensure maximum service coverage; the maximum GDP coverage principle requires site selection to focus on the most economically active areas to drive regional economic development, prioritizing locations with high GDP contributions such as industrial parks, ports, and transportation hubs; and the maximum area coverage principle requires site selection to achieve maximum coverage of physical space and functional services within the study area, ensuring that the takeoff and landing field has sufficient expansion potential and high-throughput operational capacity. Correspondingly, for candidate center points... In its cluster The weighted sum of distances within the range can be expressed as: ;in , , Points The population, GDP, and effective coverage area weights of the grid cell. , , The weighting coefficients of the three principles set manually and When focusing on only a single dimension, the remaining coefficients can be set to zero. The optimal center point is determined by... The three points with the smallest weighted distance sum are identified and recorded as candidate center points to enhance the adjustability when implementing the project in practice.

[0097] S106. Select the candidate center point with the smallest weighted cost value from the candidate center point subset as the take-off and landing anchor point of the corresponding cluster to obtain the anchor point set; send the anchor point set to the UAV traffic management system as the take-off and landing anchor point parameters for route planning in the UAV traffic management system.

[0098] Among them, the drone traffic management system refers to a management system that uniformly registers, monitors, approves routes, and issues conflict warnings for drone operations within the target airspace.

[0099] Specifically, the system first selects the candidate center point with the smallest weighted cost value within each cluster as the main center point and records several points with the second smallest weighted cost value as alternative center points. Then, it encapsulates the main center point and alternative center points into take-off and landing field configuration data according to the preset aviation data exchange format. Next, it pushes the configuration data to the UAV traffic management system through the communication module. Finally, the UAV traffic management system registers the main center point as a take-off and landing anchor point for the route planning module to call.

[0100] In some embodiments, the system will decouple the deployment requirements by scenario before acquiring multi-source geographic information. That is, the system will divide the deployment requirements into urban last-mile delivery scenario, medical supply delivery scenario and short-distance logistics transportation scenario according to the business type, set the maximum service radius, candidate point source category and demand weight grid calculation method for each scenario to form a scenario parameter set, and use each scenario parameter set as input parameters to generate the take-off and landing field configuration data corresponding to each scenario.

[0101] The scenario parameter set refers to a set of parameters that are fixed for a specific business scenario and cover the maximum service radius, candidate point source category, and demand weight grid calculation method.

[0102] Specifically, the system first breaks down the input tasks into three deployment scenarios based on business type: urban last-mile delivery, medical supply delivery, and short-distance logistics transportation. Then, it sets a relatively lenient maximum service radius for the urban last-mile delivery scenario, using large commercial complexes as candidate source points and taking food delivery order density as the weighting criterion. For the medical supply delivery scenario, it sets a stricter maximum service radius than the last-mile scenario, using health service stations, hospitals, blood banks, and transportation hubs as candidate source points and taking emergency call density as the weighting criterion. For the short-distance logistics transportation scenario, it divides the take-off and landing sites into sending ends sourced from logistics parks and express delivery centers and receiving ends sourced from residential areas and villa areas. Next, it executes the entire process from S101 to S106 with the parameter sets of each scenario. Finally, it outputs the take-off and landing site configuration data corresponding to each deployment scenario.

[0103] In the above embodiments, by fusing multi-source geographic information data, pruning spatial topology under six strong constraints, fully connected hierarchical clustering with the maximum service radius as the termination threshold, and calculating weighted cost values ​​based on demand weight grids, the system can adaptively determine the number of clusters and find the optimal central anchor point for demand coverage within each cluster, while satisfying multiple hard constraints such as airspace compliance, ecological compliance, electromagnetic compliance, land use compliance, route reference, and ground safety. This provides static take-off and landing anchor point parameters with engineering feasibility for the UAV traffic management system.

[0104] However, the weighted cost value output by the above embodiments only evaluates candidate center points from a two-dimensional plane dimension, without fully considering the mutual interference relationship between the take-off and landing corridor envelopes of each anchor point in the vertical dimension in three-dimensional space. When the main center points of adjacent clusters are close together or the prevailing wind direction causes the corridor to deflect laterally, the take-off and landing corridors corresponding to different anchor points are prone to overlap in space, which can lead to flight path conflicts during actual operation. At the same time, the take-off and landing anchor point parameters output by the above embodiments are static results generated in one go, which are difficult to respond to local airspace failures caused by temporary airspace control updates, and are also difficult to rebalance the throughput pressure of hotspot areas online based on the actual take-off and landing counts and average queuing time returned by the UAV traffic management system. Ultimately, this may result in operational imbalance problems such as some anchor points being overloaded for a long time while the utilization rate of adjacent anchor points is low. In order to further resolve the spatial conflicts between three-dimensional take-off and landing corridors and enable the take-off and landing anchor point parameters to have dynamic adaptive capabilities in response to temporary airspace control and operational load feedback, this application also provides another method for generating UAV take-off and landing site deployment points. The following is combined with Figure 2 Another method for generating UAV take-off and landing site deployment locations in this application embodiment is described below:

[0105] Please see Figure 2 This is another flowchart illustrating a method for generating UAV take-off and landing site deployment locations in an embodiment of this application.

[0106] After calculating the weighted cost value by calculating the sum of weighted distances from each point in the candidate centroid subset to all points in its cluster based on the demand weight grid, the method further includes:

[0107] S201. Within the subset of candidate centroids of each initial cluster, sort them in ascending order of weighted cost value to obtain the sequence of candidate centroids for each cluster.

[0108] S202. Select the first candidate center point of each candidate center point sequence as the current anchor point of the corresponding cluster. Based on the preset take-off and landing corridor geometric parameters and the prevailing wind direction data at the current anchor point location, construct the three-dimensional take-off and landing corridor envelope corresponding to the current anchor point.

[0109] Among them, the geometric parameters of the take-off and landing corridor refer to the set of parameters describing the spatial envelope of the UAV during the take-off and landing phase, including the horizontal angle of the take-off and landing corridor, the vertical climb slope, the maximum climb height, and the radius of the corridor bottom surface, etc. The three-dimensional take-off and landing corridor envelope refers to the three-dimensional closed geometry generated with the anchor point as the geometric origin, based on the geometric parameters of the take-off and landing corridor, and combined with the prevailing wind direction offset.

[0110] S203. Perform pairwise spatial intersection detection on the three-dimensional take-off and landing corridor envelopes corresponding to different clusters to obtain conflicting cluster pairs.

[0111] Among them, conflict clustering pairs refer to clustering combinations in which the three-dimensional take-off and landing corridor envelopes of two clusters intersect in three-dimensional space.

[0112] S204. For each conflicting cluster pair, within the clusters where the weighted cost value is greater than a preset threshold, select the next candidate center point as the updated current anchor point in the order of the candidate center point sequence, and re-execute the construction of the 3D take-off and landing corridor envelope and spatial intersection detection until the 3D take-off and landing corridor envelopes do not intersect with each other, thus obtaining the optimized anchor point set.

[0113] Understandably, if the weighted cost values ​​of both parties are not greater than the preset threshold, the replacement will be performed on the party with the larger weighted cost value.

[0114] S205. The optimized anchor point set is sent to the UAV traffic management system as the updated anchor point set, and used as the take-off and landing anchor point parameters.

[0115] Specifically, the system first sorts the candidate centroids in the subset of each cluster by weighted cost value from smallest to largest to form a sequence of candidate centroids. Then, it takes the first candidate centroid of each sequence as the current anchor point and uses prevailing wind direction data to perform azimuth offset on the centerline of the takeoff and landing corridor to generate the 3D takeoff and landing corridor envelope of the current anchor point. Next, it traverses all anchor point pairs and performs 3D spatial intersection detection, recording the cluster combinations of anchor points that intersect as conflicting cluster pairs. Finally, for each conflicting cluster pair, it selects the next candidate centroid from the candidate centroid sequence of the pair with the larger weighted cost value that is greater than a preset threshold as the updated current anchor point, reconstructs the envelope, and re-detects. This process is repeated until all 3D takeoff and landing corridor envelopes are mutually non-intersecting, resulting in an optimized set of anchor points.

[0116] It should be noted that traditional facility site selection methods only guarantee the anchor point spacing on a two-dimensional plane, failing to detect the three-dimensional corridor geometric conflicts formed in the air during drone takeoff and landing. This step, by coupling the takeoff and landing corridor geometric parameters with the field location and prevailing wind direction to model a three-dimensional envelope, allows for precise determination of the airspace coexistence between anchor points. The candidate center point sequence provides an iterative replacement candidate pool for conflict resolution, and its sorting is directly based on the weighted cost value output by S105, ensuring that each replacement prioritizes the principle of minimizing demand coverage cost. The replacement strategy adopts the "larger one yields" principle, that is, only the one with the larger weighted cost value in the conflict cluster pair that exceeds a preset threshold is replaced, avoiding repeated disturbance of the low-cost optimal solution. The value of the preset threshold should be determined in conjunction with the business's tolerance for cost. For example, a smaller threshold can be used for medical emergency scenarios to ensure timeliness, while a larger threshold can be used for last-mile delivery scenarios to expand the replacement space.

[0117] S206. Receive the temporary airspace control update pushed by the UAV traffic management system and obtain the dynamic control polygon and the corresponding effective time window of the dynamic control polygon.

[0118] Among them, the dynamic control polygon refers to the geometric boundary of the prohibited or restricted airspace area that is triggered by a temporary airspace control event and has a limited effective time.

[0119] S207. Based on the preset airspace buffer distance, perform buffer expansion on the dynamic control polygon to obtain the affected range polygon.

[0120] Among them, the affected area polygon refers to the affected spatial area obtained by expanding the dynamic control polygon outward according to the preset airspace buffer distance.

[0121] S208. In the optimized set of anchor points, filter out the anchor points that fall within the polygon of the influence range to obtain the set of affected anchor points.

[0122] S209. Using the initial cluster corresponding to the affected anchor point set as the local reconstruction range, the dynamic control polygon is incorporated into the spatial control data. Spatial topology pruning, fully connected hierarchical clustering, and weighted cost value calculation are re-executed for the points within the local reconstruction range to obtain the local replacement anchor point set within the effective time window.

[0123] Understandably, if the local replacement anchor point set is empty, the local reconstruction range will be expanded to the adjacency cluster by a preset expansion coefficient, and the above steps will be executed again.

[0124] S210. Within the effective time window, replace the affected anchor point set with the local replacement anchor point set, and use it as the take-off and landing anchor point parameters to be issued to the UAV traffic management system after the update.

[0125] Specifically, the system first continuously monitors the temporary airspace control updates pushed by the UAV traffic management system and parses them to obtain the dynamic control polygon and its effective time window; then, it performs buffer expansion on the dynamic control polygon according to the preset airspace buffer distance to obtain the affected range polygon; next, it selects anchor points that fall within the affected range polygon from the optimized anchor point set to form the affected anchor point set, and uses the initial cluster corresponding to the affected anchor points as the local reconstruction range, after incorporating the dynamic control polygon into the airspace control data, it only re-performs spatial topology pruning, fully connected hierarchical clustering and weighted cost value calculation on this range to obtain the locally replaced anchor point set; finally, within the effective time window, it replaces the affected anchor points with the locally replaced anchor point set and issues the updated take-off and landing anchor point parameters, and restores the original anchor points after the effective time window ends.

[0126] It should be noted that dynamic airspace control events (such as temporary no-fly zones for major events or hazardous areas caused by sudden weather) are characterized by short time windows and limited spatial scope. Triggering a full-domain recalculation for every event would waste computing power and cause unnecessary disturbance to unaffected areas. This step employs a five-level mechanism of "event—buffer—filter—local—time window" to achieve precise self-healing. It only re-executes the clustering and cost calculation process on the spatially affected local area, while ensuring that the anchor point layout can be automatically restored after the event ends through an effective time window constraint. The airspace buffer distance is determined with reference to the commonly used cruising altitude of UAVs and typical approach and departure deviations, avoiding both an excessively narrow impact range that might miss potential conflicts and an excessively wide range that would lead to unnecessary recalculation.

[0127] In some embodiments, the system will continuously perform clustering adaptive expansion based on actual throughput data feedback during operation. That is, the system will receive the actual take-off and landing times and average queuing time of each anchor point within a preset statistical period from the UAV traffic management system to obtain the operating load index. The system will screen overloaded anchor points whose operating load index exceeds the preset throughput threshold, divide the cluster into two sub-clusters according to the weight distribution of the demand weight grid in the cluster where the overloaded anchor point is located, and re-perform candidate center point subset screening, weighted cost value calculation and three-dimensional take-off and landing corridor envelope construction for each sub-cluster to obtain split anchor points. The split anchor points will replace the overloaded anchor points and the updated and optimized anchor point set will be sent back.

[0128] Among them, the operating load index refers to a quantitative indicator that comprehensively represents the operating pressure of the anchor point by combining the actual number of takeoffs and landings and the average queuing time. The overload anchor point refers to the takeoff and landing anchor point whose operating load index exceeds the preset throughput threshold.

[0129] Specifically, the system first receives data from the UAV traffic management system via a communication module, including the actual number of takeoffs and landings and the average queuing time for each anchor point within a preset statistical period, and synthesizes these data to form an operational load index. Then, based on a preset throughput threshold, it filters out overloaded anchor points whose operational load index exceeds the threshold, forming an overloaded anchor point set. Next, for each overloaded anchor point, based on the weight distribution of the demand weight grid within the initial cluster where the overloaded anchor point is located, the cluster is divided into two sub-clusters along the weight centroid direction, ensuring that the weight sum within each sub-cluster is nearly balanced. Finally, for each sub-cluster, the system re-executes the candidate center point subset selection, weighted cost value calculation, and three-dimensional takeoff and landing corridor envelope construction to obtain the split anchor points for that sub-cluster. The optimized anchor point set, replacing the original overloaded anchor points with the two split anchor points, is then sent back to the UAV traffic management system. For example, using the demand weight of each point within a cluster as the mass, the weighted geometric center is calculated. ; along the principal axis corresponding to the largest eigenvalue of the weight covariance matrix Create a splitting hyperplane to divide the clusters into two. If the subset of candidate center points for any sub-cluster is empty after the split, then use the k-medoids algorithm (k=2) to re-divide the clusters.

[0130] It should be noted that static site selection is difficult to detect traffic fluctuations during operation. Some anchor points may experience queuing congestion during peak promotional periods or emergency events, while neighboring anchor points remain under low load. This step uses the actual operational data transmitted back from the UAV traffic management system as the trigger signal for cluster splitting, achieving hotspot expansion through a closed loop of "load acquisition—overload screening—weight-driven splitting—sub-cluster re-solution". The weight distribution, rather than a simple geometric division, serves as the basis for splitting, ensuring a relatively balanced demand load between the two sub-clusters after splitting, avoiding "pseudo-expansion". The new anchor points generated after splitting still need to undergo candidate center point screening and 3D corridor detection to ensure the continuity of radius hard constraints and airspace safety.

[0131] In the above embodiments, a spatial conflict resolution mechanism for the three-dimensional take-off and landing corridor envelope, a local reconstruction mechanism for temporary airspace control events, and a clustering adaptive expansion mechanism for operational load feedback are further introduced. The three mechanisms work together to continuously output take-off and landing anchor point parameters that are feasible for engineering implementation, compatible with airspace, responsive to events, and balanced in load. This provides more complete and robust technical support for the planning and dynamic management of UAV take-off and landing field networks under the large-scale operation of urban low-altitude economy.

[0132] The following describes the UAV take-off and landing site deployment point generation system in the embodiments of this application from the perspective of hardware processing. Please refer to [link to relevant documentation]. Figure 3 This is a schematic diagram of the physical device structure of the UAV take-off and landing site deployment point generation system in this application embodiment.

[0133] It should be noted that, Figure 3The structure of the drone take-off and landing site deployment point generation system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0134] like Figure 3 As shown, the system includes a CPU, which can perform various appropriate actions and processes based on a program stored in the ROM or a program loaded into the RAM from a storage portion, such as executing the methods described in the above embodiments. The RAM also stores various programs and data required for system operation. The CPU, ROM, and RAM are interconnected via a bus. I / O interfaces are also connected to the bus.

[0135] The following components are connected to the I / O interface: input sections including cameras, infrared sensors, etc.; output sections including liquid crystal displays (LCDs) and speakers, etc.; storage sections including hard drives, etc.; and communication sections including network interface cards such as LAN (Local Area Network) cards and modems, etc. The communication section performs communication processing via a network such as the Internet. Drives are also connected to the I / O interface as needed. Removable media, such as disks, optical disks, magneto-optical disks, semiconductor memories, etc., are installed on the drive as needed so that computer programs read from them can be installed into the storage section as needed.

[0136] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by the CPU, it performs the various functions defined in this application.

[0137] 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 this application. Each block in a flowchart or block diagram may represent a module, 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.

[0138] Specifically, the system in 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 method provided in the above embodiment.

[0139] In another aspect, this application also provides a computer-readable storage medium, which may be included in the system described in the above embodiments; or it may exist independently and not assembled into the system. The storage medium carries one or more computer programs that, when executed by a processor of the system, cause the system to implement the methods provided in the above embodiments.

[0140] 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.

[0141] 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)".

Claims

1. A method for generating deployment sites for unmanned aerial vehicle (UAV) take-off and landing zones, characterized in that, The method includes: Multi-source geographic information data of the target area is acquired to obtain point of interest vector layers, airspace control data, land use classification raster, and demand weight raster; the airspace control data includes polygonal boundaries of restricted areas, restricted areas, and danger zones; the demand weight raster is generated from the spatiotemporal distribution density of historical orders and is used to characterize the demand intensity of each geographic unit. Based on the airspace control data and the land use classification raster, spatial topology clipping is performed on the point of interest vector layer to obtain a candidate point set. Calculate the spherical distance between any two points in the candidate point set to obtain a distance matrix. Use the preset maximum service radius as the inter-cluster distance termination threshold, and perform fully connected hierarchical clustering on the distance matrix to obtain several initial clusters. Within each of the initial clusters, points whose distance to all other points within the cluster does not exceed the maximum service radius are selected to obtain a subset of candidate center points; Based on the demand weight grid, the weighted total distance from each point in the candidate center point subset to all points in its cluster is calculated to obtain the weighted cost value. The candidate center point with the smallest weighted cost value in the candidate center point subset is taken as the take-off and landing anchor point of the corresponding cluster, thus obtaining the anchor point set; the anchor point set is sent to the UAV traffic management system as the take-off and landing anchor point parameters for route planning in the UAV traffic management system.

2. The method according to claim 1, characterized in that, The method further includes preloading an ecological protection red line layer and an electromagnetic field strength raster; based on the airspace control data and the land use classification raster, performing spatial topology clipping on the point of interest vector layer to obtain a candidate point set, including the following steps: Perform a spatial difference operation between the ecological protection red line layer and the point of interest vector layer, and remove the points in the point of interest vector layer that fall into the ecological protection red line layer to obtain the first set of filtered points. Based on the polygonal boundaries of restricted areas, prohibited areas, and dangerous areas in the airspace control data, a point-to-surface inclusion judgment is performed on the first set of filtered points, and points in the first set of filtered points that fall within the polygonal boundaries are removed to obtain the second set of filtered points. Based on a preset electromagnetic threshold, points in the second set of filter points that fall within the electromagnetic field strength grid and whose pixel intensity exceeds the electromagnetic threshold are removed, resulting in a third set of filter points. Based on the land use classification grid, preset land use types are retained in the third filter point set to obtain the candidate point set. The preset land use types include commercial land, industrial land, transportation land, residential land, and public management and service land.

3. The method according to claim 2, characterized in that, The method also includes preloading digital elevation raster layers and building height data; After obtaining the candidate point set by retaining the preset land use type in the third filter point set based on the land use classification raster, the method further includes: Based on the digital elevation raster layer, the terrain slope of each candidate point in the candidate point set is calculated to obtain the slope value corresponding to each candidate point. Based on a preset slope threshold, candidate points in the candidate point set whose slope values ​​are greater than the preset slope threshold are removed to obtain a first set of filtered points; Based on the building height data, detect whether there are buildings with heights exceeding a preset occlusion threshold within a preset clearance range above each candidate point in the first set of filter points, and remove candidate points containing such buildings to obtain the second set of filter points; The second set of filtered points is used as the updated set of candidate points.

4. The method according to claim 1, characterized in that, Prior to the step of acquiring multi-source geographic information data of the target area, the method further includes: Based on the business type, the deployment requirements are divided into urban last-mile delivery scenarios, medical supply delivery scenarios, and short-distance logistics transportation scenarios, resulting in multiple deployment scenarios; For each of the multiple deployment scenarios, the corresponding maximum service radius, candidate point source category, and demand weight grid calculation method are set to obtain the scenario parameter set corresponding to each deployment scenario; Based on the scenario parameter set, for each deployment scenario, the scenario parameter set corresponding to the deployment scenario is used as the input parameter to obtain the take-off and landing field configuration data corresponding to each deployment scenario.

5. The method according to claim 1, characterized in that, After calculating the weighted sum of distances from each point in the candidate centroid subset to all points within its cluster based on the demand weight grid, and obtaining the weighted cost value, the method further includes: Within the subset of candidate centroids of each initial cluster, the candidate centroids are sorted in ascending order of the weighted cost value to obtain the sequence of candidate centroids for each cluster. The first candidate center point of each candidate center point sequence is selected as the current anchor point of the corresponding cluster. Based on the preset take-off and landing corridor geometric parameters and the prevailing wind direction data at the location of the current anchor point, a three-dimensional take-off and landing corridor envelope corresponding to the current anchor point is constructed. Perform pairwise spatial intersection detection on the three-dimensional take-off and landing corridor envelopes corresponding to different clusters to obtain conflicting cluster pairs; For each conflict cluster pair, within the cluster where the weighted cost value is greater than a preset threshold, the next candidate center point is selected sequentially as the updated current anchor point according to the order of the candidate center point sequence. The construction of the three-dimensional take-off and landing corridor envelope and the spatial intersection detection are re-executed until the three-dimensional take-off and landing corridor envelopes do not intersect with each other, thus obtaining the optimized anchor point set. The optimized set of anchor points is sent to the UAV traffic management system as the updated set of anchor points, serving as the take-off and landing anchor point parameters.

6. The method according to claim 5, characterized in that, After replacing the anchor point result corresponding to the weighted cost value with the optimized anchor point set as the take-off and landing anchor point parameter sent to the UAV traffic management system, the method further includes: Receive temporary airspace control updates pushed by the UAV traffic management system, and obtain the dynamic control polygon and the effective time window corresponding to the dynamic control polygon; Based on the preset spatial buffer distance, the dynamic control polygon is buffered and expanded to obtain the influence range polygon. In the optimized set of anchor points, anchor points that fall within the polygon of the influence range are selected to obtain the set of affected anchor points. Using the initial cluster corresponding to the affected anchor point set as the local reconstruction range, the dynamic control polygon is incorporated into the spatial control data, and the spatial topology clipping, the fully connected hierarchical clustering, and the weighted cost value are recalculated for the points within the local reconstruction range to obtain the local replacement anchor point set within the effective time window; Within the effective time window, the local replacement anchor point set replaces the affected anchor point set, which is then used as the updated take-off and landing anchor point parameters sent to the UAV traffic management system.

7. The method according to claim 5, characterized in that, The method further includes: The system receives the actual number of takeoffs and landings and the average queuing time of each anchor point within a preset statistical period from the UAV traffic management system, and obtains the corresponding operational load index for each anchor point. Based on a preset throughput threshold, anchor points where the operating load index exceeds the throughput threshold are selected to obtain an overload anchor point set. For each overload anchor point in the set of overload anchor points, the cluster in which the overload anchor point is located is divided into two sub-clusters based on the weight distribution of the demand weight grid within the cluster in which the overload anchor point is located, thus obtaining a set of secondary split clusters. For each sub-cluster in the secondary split cluster set, the screening of the candidate centroid subset, the calculation of the weighted cost value, and the construction of the three-dimensional take-off and landing corridor envelope are re-executed to obtain the split anchor point corresponding to the sub-cluster. The overload anchor point is replaced with the split anchor point, the optimized anchor point set is updated, and the updated optimized anchor point set is sent to the UAV traffic management system.

8. A system for generating deployment sites for unmanned aerial vehicle (UAV) take-off and landing areas, characterized in that, include: One or more processors and a memory; the memory is coupled to the one or more processors, the memory being used to store computer program code, the computer program code including computer instructions, the one or more processors invoking the computer instructions to cause the 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 system, the 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 system, the system performs the method as described in any one of claims 1-7.

Images