Route planning method and device of hovercar, electronic equipment and storage medium

Abstract

The invention discloses an aerocar route planning method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a plurality of independent point cloud data sets and the point cloud position of each point cloud data in the plurality of independent point cloud data sets in response to a planning instruction; performing data registration based on the point cloud position of each point cloud data to obtain a plurality of target point cloud data sets under the three-dimensional coordinate system; further, performing data processing on the plurality of target point cloud data sets to obtain a three-dimensional vector map; determining a safe flight area based on the three-dimensional vector map and a preset buffer rule; and based on the safe flight area and the three-dimensional vector map, performing three-dimensional space path planning processing on the hovercar, and determining the target flight route, through the technical scheme of the invention, the calculation complexity is simplified, the planning efficiency, the planning accuracy and the response real-time performance are improved, the decision delay is reduced, and the flight safety is improved.

Description

Technical Field

[0001] This invention relates to the field of flying car technology, and more specifically, to a route planning method, device, electronic equipment, and storage medium for a flying car. Background Technology

[0002] The low-altitude economy is a comprehensive economic form that drives the integrated development of related fields through various low-altitude flight activities of manned and unmanned aircraft. Its related products mainly include drones, eVTOL (electric vertical takeoff and landing) aircraft, helicopters, and traditional fixed-wing aircraft, involving both residential consumption and industrial applications. However, the development of the low-altitude economy also faces various challenges and risks. Low-altitude flight requires overcoming factors such as collisions and severe weather, making flight safety a major issue for the low-altitude economy. Although drone technology is becoming increasingly mature, with most drones possessing autonomous obstacle avoidance capabilities, these functions are not effective in complex environments such as trees and building clusters. When multiple drones fly in swarms, the obstacle avoidance system struggles to react quickly. Furthermore, as the low-altitude economy gradually develops, low-altitude logistics and low-altitude transportation will also encounter similar problems.

[0003] Regarding flight safety assurance for flying cars, current perception algorithms have limited capabilities and a high risk factor, with some objects such as power lines having low recognition rates or difficulty in timely response. Based on this, this patent proposes a route planning method for flying cars to solve the problem of low flight safety caused by cumbersome and inaccurate path planning in existing flying car systems. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a route planning method, device, electronic device and storage medium for flying cars, in order to solve the problem of low flight safety caused by cumbersome and inaccurate path planning in existing flying cars.

[0005] One aspect of the present invention provides a route planning method for a flying car, the method comprising the following steps: In response to planning instructions, acquire multiple independent point cloud datasets and the point cloud position of each point cloud data in the multiple independent point cloud datasets; Based on the point cloud position of each point cloud data, data registration is performed to obtain multiple target point cloud datasets in a three-dimensional coordinate system; Data processing is performed on the multiple target point cloud datasets to obtain a three-dimensional vector map; Based on the aforementioned three-dimensional vector map and preset buffer rules, a safe flight zone is determined; Based on the safe flight area and the three-dimensional vector map, the flying car is subjected to three-dimensional spatial path planning to determine the target flight route.

[0006] Another aspect of the present invention provides a route planning device for a flying car, the device comprising: The acquisition module is used to acquire, in response to planning instructions, multiple independent point cloud datasets and the point cloud position of each point cloud data in the multiple independent point cloud datasets; The registration module is used to perform data registration based on the point cloud position of each point cloud data to obtain multiple target point cloud datasets in a three-dimensional coordinate system. The processing module is used to process the multiple target point cloud datasets to obtain a three-dimensional vector map; The first determining module is used to determine a safe flight area based on the three-dimensional vector map and preset buffer rules; The second determining module is used to perform three-dimensional spatial path planning processing on the flying car based on the safe flight area and the three-dimensional vector map, and determine the target flight route.

[0007] Another aspect of the present invention provides an electronic device, comprising: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the route planning method for the flying car described in any one of the above descriptions.

[0008] Another aspect of the present invention provides a computer-readable storage medium that, when the instructions in the storage medium are executed by a processor of an electronic device, enables the electronic device to perform the route planning method for a flying car described above.

[0009] This invention provides a route planning method, device, electronic device, and storage medium for a flying car. Responding to a planning command, it acquires multiple independent point cloud datasets and the point cloud position of each point cloud data in the datasets. Based on the point cloud position of each data, it performs data registration to obtain multiple target point cloud datasets in a three-dimensional coordinate system. It converts two-dimensional point cloud data into three-dimensional point cloud data, providing the flying car with prior knowledge of the global environment, improving its avoidance capabilities, and reducing the probability of collisions. Furthermore, it processes the multiple target point cloud datasets to obtain a three-dimensional vector map. Based on the three-dimensional vector map and preset buffer rules, it determines a safe flight area. Based on the safe flight area and the three-dimensional vector map, it performs three-dimensional spatial path planning for the flying car to determine the target flight route. By transforming the complex real-time environment modeling and path search problem into an optimization problem within a predefined "safe flight area," it simplifies computational complexity, reduces the computational load and memory consumption of the path planning algorithm, thereby improving planning efficiency, accuracy, and real-time response, reducing decision delay, and enhancing flight safety. Attached Figure Description

[0010] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a flowchart illustrating a route planning method for a flying car according to an exemplary embodiment. Figure 2 This is a flowchart illustrating a process for determining a dataset of multiple target point clouds in a three-dimensional coordinate system, according to an exemplary embodiment. Figure 3 This is a schematic diagram of a process for determining a target point cloud dataset according to an exemplary embodiment; Figure 4 This is a schematic diagram of a process for processing multiple target point cloud datasets to obtain a three-dimensional vector map, according to an exemplary embodiment. Figure 5 This is a flowchart illustrating a process for determining a point cloud scene model according to an exemplary embodiment; Figure 6 This is a schematic diagram of a process for modeling a point cloud scene model to obtain a three-dimensional vector map, according to an exemplary embodiment. Figure 7This is a schematic diagram of a process for determining a 3D vector map based on manually labeled objects, according to an exemplary embodiment. Figure 8 This is a schematic diagram of the structure of a route planning device for a flying car according to an exemplary embodiment. Detailed Implementation

[0012] 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.

[0013] It should be noted that the terms "first," "second," etc., in the specification and drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0014] Figure 1 This is a flowchart illustrating a route planning method for a flying car according to an exemplary embodiment. This specification provides method operation steps as shown in the embodiments or flowcharts, but based on conventional or non-inventive labor, more or fewer operation steps may be included. The order of steps listed in the embodiments is merely one possible execution order among many and does not represent the only possible execution order. In actual system or server product execution, the method can be executed sequentially according to the embodiments or drawings, or in parallel (e.g., in a parallel processor or multi-threaded processing environment). Specifically, as shown... Figure 1 As shown, taking the route planning system of a flying car as the execution entity, this application introduces an embodiment of a route planning method for a flying car. The method may include: S101: In response to the planning instruction, obtain multiple independent point cloud datasets and the point cloud position of each point cloud data in the multiple independent point cloud datasets.

[0015] In one specific embodiment, the planning command is used to instruct the route planning system to begin path planning for the flying car. Optionally, the planning command is issued by the flying car control system, ground control station, or operation management platform to trigger a path planning task for a specific take-off and landing site. The point cloud data comes from pre-deployed UAV professional collection or collection by the lidar sensor carried by the flying car during operation. The collection scheme adopts a multi-altitude layering strategy, such as collection at different altitude layers such as 50 meters, 100 meters, and 200 meters, to ensure comprehensive coverage of the low-altitude three-dimensional space. Each point cloud data contains two-dimensional coordinate information (X, Y) and corresponding POS data (high-precision position and attitude information provided by the RTK and IMU systems). The independent point cloud dataset can be a collection of point cloud data collected by each lidar sensor. Specifically, when a planning task is triggered, the flying car's route planning system responds to the planning command, activates the lidar sensor to collect point cloud data, and further obtains multiple independent point cloud datasets and the point cloud position of each point cloud data in the multiple point cloud datasets.

[0016] S103: Based on the point cloud position of each point cloud data, perform data registration to obtain multiple target point cloud datasets in a three-dimensional coordinate system.

[0017] In a specific embodiment, the target point cloud dataset can be an independent point cloud dataset in a three-dimensional coordinate system determined by registration of multiple independent point cloud datasets. Specifically, when multiple independent point cloud datasets are obtained, the multiple independent point cloud datasets from different flights or different times are precisely integrated into a unified three-dimensional coordinate system, thereby obtaining multiple target point cloud datasets in the three-dimensional coordinate system.

[0018] Figure 2 This is a schematic diagram illustrating a process for determining a dataset of multiple target point clouds in a three-dimensional coordinate system, according to an exemplary embodiment. In an optional embodiment, such as... Figure 2 As shown, the above-mentioned data registration based on the point cloud position of each point cloud data to obtain multiple target point cloud datasets in a three-dimensional coordinate system includes: S201: Obtain the coordinate transformation formula for each independent point cloud dataset; S203: Based on the point cloud position of each point cloud data in each independent point cloud dataset and the coordinate transformation formula corresponding to each independent point cloud dataset, perform coordinate transformation to determine the corrected point cloud position corresponding to each independent point cloud dataset. S205: Based on the corrected point cloud position corresponding to each independent point cloud dataset, perform distance registration to determine the target point cloud dataset.

[0019] In a specific embodiment, the coordinate transformation formula can be a transformation formula from a two-dimensional coordinate system to a three-dimensional coordinate system corresponding to each independent point cloud dataset; optionally, each independent point cloud dataset is equipped with a corresponding coordinate transformation formula; the corrected point cloud position can be the point cloud position after coordinate transformation; specifically, when it is necessary to perform coordinate transformation on each independent point cloud dataset, the coordinate transformation formula corresponding to each independent point cloud dataset is obtained. Optionally, the coordinate transformation formula can be set by the administrator during the initial use of the system. Further, based on the point cloud position of each point cloud data in each independent point cloud dataset and the coordinate transformation formula corresponding to each independent point cloud dataset, the point cloud position is transformed, that is, the point cloud position of the point cloud data is transformed from two-dimensional to three-dimensional, thereby obtaining the corrected point cloud position corresponding to each independent point cloud dataset. Further, distance registration is performed on the corrected point cloud position corresponding to each independent point cloud dataset to determine the target point cloud dataset.

[0020] Figure 3 This is a schematic diagram of a process for determining a target point cloud dataset according to an exemplary embodiment. In an optional embodiment, such as... Figure 3 As shown, the above-mentioned distance registration based on the corrected point cloud position corresponding to each independent point cloud dataset, to determine the target point cloud dataset includes: S301: Based on the corrected point cloud position, perform position registration for each point cloud data to determine the data point pair corresponding to each independent point cloud dataset; S303: Calculate the distance between each pair of data points to obtain the pair distance; S305: Based on the point-to-point distance and a preset distance threshold, perform a distance comparison to determine the first distance comparison result; and take the point cloud data in each independent point cloud dataset whose point-to-point distance indicated by the first distance comparison result is not greater than the preset distance threshold as the target point cloud dataset.

[0021] In a specific embodiment, a data point pair can be a data pair formed by point cloud data in each independent point cloud dataset and point cloud data corresponding to the overlapping area of ​​the point cloud dataset; the point pair distance can be the distance between the point cloud data in two-dimensional coordinates and the point cloud data in the corresponding three-dimensional coordinates; the first distance comparison result is used to indicate whether the point pair distance is greater than a preset distance threshold; the preset distance threshold can be a distance threshold for filtering target point cloud data; Specifically, based on the acquired corrected point cloud positions, position registration is performed for each point cloud data in each independent point cloud dataset. Optionally, corresponding point pairs are found in the overlapping area of ​​the point clouds to determine the corresponding data point pairs for each independent point cloud dataset. Further, the distance between each data point pair is calculated, optionally using coordinates. The point pair distance is then compared with a preset distance threshold to obtain the first distance comparison result. Further, point cloud data in each independent point cloud dataset whose first distance comparison result indicates a point pair distance not greater than the preset distance threshold are combined as the target point cloud dataset. Optionally, taking two point cloud datasets A and B as an example: First, in the overlapping area of ​​A and B (such as the wall of a building), the nearest neighbor point in B is found for each point in A, forming an initial set of point pairs. Second, point pairs with excessively large or obviously unreasonable distances (such as pairing ground points with roof points) are removed. Then, a rigid body transformation matrix (including rotation and translation) is calculated that minimizes the sum of distances between the remaining valid point pairs. Finally, this transformation matrix is ​​applied to point cloud A, and the above process is iterated until the average distance change is less than a preset threshold (e.g., 0.01 meters), at which point the registration is considered converged. This process is performed pairwise or globally on all point cloud datasets to ultimately obtain a set of target point cloud datasets that are precisely aligned in a unified coordinate system.

[0022] In the above embodiments, by calibrating, pairing and calculating distances between point cloud datasets, the target point cloud dataset is determined, achieving accurate matching between point clouds and thus improving the accuracy of path planning.

[0023] S105: Process multiple target point cloud datasets to obtain a 3D vector map.

[0024] In one specific embodiment, the three-dimensional vector map can be a map formed by a target point cloud dataset in a three-dimensional coordinate system; specifically, when multiple target point cloud datasets are obtained through the aforementioned steps, a series of data processing steps are performed on the obtained multiple target point cloud datasets to obtain a three-dimensional vector map.

[0025] Figure 4 This is a schematic diagram illustrating a process for processing multiple target point cloud datasets to obtain a 3D vector map, according to an exemplary embodiment. In an optional embodiment, such as... Figure 4 As shown, the above-described data processing of the multiple target point cloud datasets to obtain a 3D vector map includes: S401: Delete data from the target point cloud dataset to obtain a point cloud scene model; S403: Model the point cloud scene model to obtain a 3D vector map.

[0026] In one specific embodiment, the point cloud scene model can be generated by using algorithms such as Poisson reconstruction to create a continuous, watertight triangular mesh surface model from multiple target point cloud datasets. Optionally, the multiple target point cloud datasets can be refined to remove unnecessary data, retain key geometric features of the scene, and eliminate redundancy, noise, and unnecessary data details. Specifically, data deletion can be performed on the target point cloud datasets to obtain the point cloud scene model, and then the point cloud scene model can be modeled to obtain a 3D vector map.

[0027] Figure 5 This is a schematic diagram of a process for determining a point cloud scene model according to an exemplary embodiment. In an optional embodiment, such as Figure 5 As shown, the above-mentioned data deletion process for the target point cloud dataset to obtain the point cloud scene model includes: S501: Filter the overlapping regions in the target point cloud dataset to obtain the overlapping dataset; S503: Perform fusion processing on the target point cloud data in the overlapping dataset to determine the fused point cloud dataset; S505: Use the target algorithm to establish a surface model on the fused point cloud dataset to obtain a point cloud scene model.

[0028] In one specific embodiment, a spatial data structure such as a voxelized mesh or octree is used to divide the entire registered point cloud space, dividing the 3D space into countless tiny cubes (voxels). Within each voxel, if data points from multiple point clouds from different sorties exist, the region is determined to be an overlapping region. For each voxel containing multiple points, a thinning operation is performed. Optionally, the centroid of all points within the voxel is calculated, and only the centroid is retained while the other points are discarded; or the geometric information of all points within the voxel is retained, but it is treated as a uniformly distributed region during subsequent modeling, thus obtaining a relatively uniform density overlapping dataset that eliminates high redundancy. Furthermore, the point cloud after thinning is still composed of discrete points. The collection of points lacks topological relationships between them, making it impossible to form a continuous surface. Therefore, the target point cloud data in the overlapping dataset is fused to determine the fused point cloud dataset. The processed point cloud set is regarded as a unified and holistic data source. Point clouds from different sources that are now spatially aligned and have uniform density are merged into a single point cloud data file, and a unified spatial index is established. Furthermore, a normal vector is calculated for each point in the fused point cloud dataset, usually by analyzing the k-nearest neighbors of each point through principal component analysis (PCA). The point cloud space is then adaptively partitioned into an octree structure, and the point cloud with outward-pointing normal vectors is transformed into a continuous vector field. Then, the corresponding triangular mesh surface is extracted from the scalar function through marching cubes or other isosurface extraction algorithms, and finally, the point cloud scene model is output.

[0029] Figure 6 This is a schematic diagram illustrating a process for modeling a point cloud scene model to obtain a 3D vector map, according to an exemplary embodiment. In an optional embodiment, such as... Figure 6 As shown, the above-described modeling process for the point cloud scene model to obtain the 3D vector map includes: S601: Obtain the preset learning model; S603: Use a preset learning model to annotate the target point cloud data in the point cloud scene model to obtain the annotated object and the annotated shape corresponding to the annotated object; S605: Determine the regularity of the labeled object based on the labeled shape corresponding to the labeled object; S607: When the regularity of the object labeling is indicated by the regularity of the object, fit the regular labeled objects to obtain a three-dimensional vector map.

[0030] In a specific embodiment, the preset learning model can be a pre-set model capable of object annotation in a three-dimensional coordinate system. Specifically, the preset learning model is acquired and preprocessed to extract vertex information from the point cloud scene model, or to uniformly sample the mesh surface to generate a dense point set. These points contain three-dimensional coordinates (X, Y, Z) and may optionally contain information such as reflection intensity. This point set is input into the pre-loaded preset learning model, which performs forward propagation calculations on each point and outputs the probability of each point belonging to each preset category. The category with the highest probability is taken as the predicted label for that point. Further, it is necessary to distinguish different individuals, such as distinguishing two adjacent buildings, and then perform instance segmentation. Optionally, a method based on connected component analysis is usually used. Specifically, the predicted category of the points is filtered to retain only the obstacle categories we are interested in (such as buildings, trees, etc.). Then, in three-dimensional space, points belonging to the same category are clustered.Clustering groups spatially connected points into a single, independent object. Common clustering algorithms include Euclidean clustering. This results in several labeled objects, each a point cloud representing an independent physical entity (e.g., "Building A," "Tree B," "Utility Pole C"). The 3D shape formed by this point cloud is the labeled shape of the object. Further, based on the labeled shapes, the regularity of the labeled objects is determined. Specifically, simple geometric models can be used to fit the point cloud of the labeled shapes. For objects identified as "buildings," minimum bounding box algorithms or plane fitting algorithms are used to calculate the best-fit cube; for objects identified as "utility poles," cylinder fitting algorithms (e.g., RANSAC-based fitting) are used to calculate the best-fit cylinder. Optionally, fitting error can be calculated, i.e., the average distance from the original point cloud (labeled shape) to the surface of the fitted geometric model is calculated, and an error threshold (e.g., 0.5 meters) is set. If the fit is... If the error is less than the threshold, the labeled object is considered to have a regular geometric shape; otherwise, it is judged as irregular. Further, regular labeled objects are converted into lightweight, semantically meaningful vector elements. Specifically, for a "building" judged as regular, the system directly extracts the parameters (center point coordinates, length, width, height, rotation angle) of the previously calculated best-fit cube as a vector element. For a "telephone pole" judged as regular, the system extracts the parameters (base center coordinates, height, radius) of the best-fit cylinder. A corresponding attribute record is then created for each generated vector element. The attribute table typically includes: Element ID: a unique identifier; Type: inherited from AI recognition results, such as "building" or "telephone pole"; Geometric parameters: such as height, width, radius, etc.; Spatial location: center point latitude and longitude, elevation; Hazard level: automatically assigned according to type (e.g., "extremely high" for high-voltage power lines, "medium" for trees); Other metadata: such as data source, confidence level, timestamp, etc. The vector elements generated through this automated process construct a 3D vector map.

[0031] In the above embodiments, the complex real-world scene is efficiently and accurately transformed into a three-dimensional vector map that the flying car can directly understand and utilize through an automated production line. This lays a solid data foundation for subsequent safe area calculation and path planning, improves the practicality and economy of the entire system, and thus achieves highly efficient and automated path planning.

[0032] Figure 7 This is a schematic diagram illustrating a process for determining a 3D vector map based on manually labeled objects, according to an exemplary embodiment. In an optional embodiment, such as... Figure 7 As shown, the above method also includes: S701: When the regularity of the object indicates that the labeled object is irregular, the irregular labeled object is sent to the display interface so that the user can label the irregular labeled object and obtain a manually labeled object. S703: Receives manually labeled objects from the display interface, fits the manually labeled objects, and obtains a 3D vector map.

[0033] In one specific embodiment, when the object with regularity is indicated as an irregular object, the irregularly labeled object is sent to the display interface so that the user can annotate it manually. Specifically, irregularities in the object with regularity can include situations where the shape of the object cannot be fitted with a simple parametric geometric model such as a cube or cylinder with low error. Examples include historical buildings and art galleries with complex roof structures, ancient trees or shrubs with extremely irregular canopy shapes, large, non-standard industrial facilities, or situations where the AI ​​recognition has low confidence or is erroneous, and the pre-learning... The model may fail to classify certain objects with high confidence; it may also result in obvious classification errors, such as misclassifying a dense bundle of wires as vegetation or failing to recognize a strangely shaped sculpture; it may also encounter situations where the point cloud quality is poor, due to occlusion, excessive acquisition distance, or other reasons, resulting in a sparse, incomplete, or noisy point cloud for the object, making automated processing and judgment unreliable. In such cases, when the regularity score of an object is below a threshold, or the AI ​​classification confidence is below a threshold, the system will automatically mark the object as "irregular," thus triggering the sending of the irregular object to the display interface. Optional operation: When irregular objects are displayed, they are not shown in isolation, but rather placed within the background of the surrounding point cloud or mesh model. Furthermore, the software interface provides users with a complete set of 3D editing and annotation tools, including classification correction tools, vector drawing tools, and attribute entry panels. Specifically, users can use the classification correction tool to correct AI recognition errors by selecting the correct object category from the drop-down menu (e.g., correcting "unknown" to "high-voltage power line tower"). Users can use the vector drawing tool to manually select key angles on the point cloud for large, irregular objects (such as complex buildings). The system generates polygonal bases in real time, and users can then drag and specify their height to construct a prism that accurately fits the building's outline. For linear objects (such as electrical wires), users can draw the wires' direction point by point, and the system connects them into a 3D line vector, assigning attributes such as diameter. For complex curved objects, users can use a polyhedron fitting tool to approximate the object's complex shape with a series of connected triangular faces. After completing the drawing of the geometry, users need to fill in or confirm the object's detailed attributes, such as material, owner, precise height, and hazard level, in the associated attribute table displayed in the attribute entry panel.Furthermore, after the operator completes the creation of the geometric model and assigns attribute values, they click "Confirm" or "Save" to generate manually labeled objects. These manually labeled objects are then input into the system. The system's backend service receives the data package of manually labeled objects submitted from the front-end interface. The system automatically verifies the integrity and format of the data. After successful verification, automatic fitting is performed, i.e., data integration and fusion. The system seamlessly integrates the manually labeled objects into the 3D vector map database being built. Optionally, during the fitting process, the system automatically checks whether newly added manually labeled objects conflict with existing automated features in terms of spatial location (e.g., overlap). If a conflict is found, the manually labeled objects are usually given higher priority, and the user may be prompted for final confirmation, or the automated features with lower confidence levels may be automatically covered / adjusted. Finally, once all objects marked as "irregular" have been manually processed and successfully integrated, a 3D vector map is formed.

[0034] In the above embodiments, by integrating manually labeled objects into the three-dimensional vector map and introducing a structured human-machine collaboration mechanism, the efficiency of machines and the wisdom of humans are cleverly combined to jointly create a solid and reliable three-dimensional vector map on which flying cars rely for safe take-off and landing, and ensure the reliability of the map in key details, thereby improving planning efficiency and planning accuracy in complex scenarios.

[0035] S107: Determine safe flight areas based on 3D vector maps and preset buffer rules.

[0036] In one specific embodiment, the safe flight zone can be a safe isolation space formed by expanding outward a certain distance from the outline of each obstacle (such as a building, power line, or tree) in a 3D vector map. The preset buffer rules can be rules that dynamically calculate the size and shape of the safe buffer zone around each obstacle. Specifically, the system traverses every obstacle element in the 3D vector map. For each obstacle, the system calls the corresponding algorithm to calculate its buffer distance (D) based on its attributes (type, height, etc.) and the preset buffer rules. Based on the obstacle's geometry (such as the polygonal base of a building), it expands outward by a distance D in 3D space to generate a "danger zone". The "no-fly zone" is optional. For buildings, it expands their cubic model outward by D meters to form a larger cube; for utility poles, it expands their cylindrical model outward by D meters to form a thicker cylinder; and for power lines, it expands their wire model outward by D meters to form a tubular buffer zone. Furthermore, the system merges all the three-dimensional hazard isolation zones generated by obstacles to form a unified three-dimensional "no-fly zone" model. An initial flight mission airspace is defined (usually a large three-dimensional bounding box containing take-off and landing points and flight paths). Through three-dimensional Boolean difference operations, the entire no-fly zone model is "subtracted" from the initial mission airspace. The remaining three-dimensional space after the Boolean operation is the safe flight area.

[0037] S109: Based on the safe flight area and three-dimensional vector map, perform three-dimensional spatial path planning processing on the flying car to determine the target flight route.

[0038] In one specific embodiment, the target flight path can be the final flight path of the flying car. Specifically, based on a pre-calculated safe flight area and a semantically rich 3D vector map, the flying car path planning system will perform intelligent path search within a strictly defined 3D safe airspace. By invoking advanced algorithms such as Algorithm A and Restricted Response Time (RRT), the system generates an initial 3D spatial path from the takeoff point to the landing point, ensuring the path completely avoids all static obstacle buffer zones while comprehensively considering multiple optimization objectives such as path length, flight time, and energy consumption. Furthermore, after generating the initial path, the system further refines and verifies it by incorporating semantic layer information from the 3D vector map. By setting path cost weights for high-risk areas, the system guides the flight path away from specific threats such as high-voltage cables; simultaneously, it performs smooth optimization of the path based on the aircraft dynamics model to ensure it meets flight control requirements; the final output target flight path is directly loaded into the flight control system for execution.

[0039] This invention provides a route planning method for flying cars, which constructs a complete technical closed loop from environmental perception to safety decision-making.

[0040] First, at the environmental perception and map building level, the solution innovatively adopts a collaborative data collection approach using drones and flying cars, acquiring low-altitude 3D point cloud data through a multi-altitude layering strategy. Subsequently, after coarse and fine registration data processing, the multi-source point clouds are accurately fused, and vector modeling technology combining AI recognition and human assistance is used to generate a 3D vector map rich in semantic information. This map not only accurately describes the geometry of obstacles but also labels their type, altitude, and other key attributes.

[0041] Secondly, at the level of safety decision-making and path planning, the solution is based on a 3D vector map and dynamically calculates the 3D safety buffer zone around different obstacles (such as high-voltage power lines, buildings, and trees) through a set of preset intelligent buffer rules. After eliminating all dangerous areas through spatial Boolean operations, a clear 3D safe flight zone is generated. Finally, the path planning algorithm is strictly restricted to operate within this safe zone, efficiently generating a target flight route that is both safe (avoiding all static obstacles) and of high quality (considering path length, energy consumption, and risk).

[0042] In summary, this solution, through a systematic approach of "collaborative data collection → precise modeling → intelligent buffering → safety planning," transforms the complex low-altitude environment into structured information that flying cars can understand and utilize. This provides reliable full-stack safety assurance for flying cars during the most critical take-off and landing phase, effectively addressing the core pain points of existing technologies, such as incomplete environmental perception, low reliability of path planning, inaccurate path planning, cumbersome path planning, and low flight safety.

[0043] Figure 8 This is a schematic diagram of a route planning device for a flying car according to an exemplary embodiment. The following describes an embodiment of the route planning device for a flying car according to this application. Specifically, as shown... Figure 8 As shown, the device includes: The acquisition module 801 is used to acquire, in response to the planning instruction, multiple independent point cloud datasets and the point cloud position of each point cloud data in the multiple independent point cloud datasets; The registration module 803 is used to perform data registration based on the point cloud position of each point cloud data to obtain multiple target point cloud datasets in a three-dimensional coordinate system. Processing module 805 is used to process the multiple target point cloud datasets to obtain a three-dimensional vector map; The first determining module 807 is used to determine a safe flight area based on the three-dimensional vector map and preset buffer rules; The second determining module 809 is used to perform three-dimensional spatial path planning processing on the flying car based on the safe flight area and the three-dimensional vector map, and determine the target flight route.

[0044] In an optional embodiment, the registration module 803 includes: The transformation acquisition unit is used to obtain the coordinate transformation formula corresponding to each independent point cloud dataset. The correction determination unit is used to perform coordinate transformation based on the point cloud position of each point cloud data in each independent point cloud dataset and the coordinate transformation formula corresponding to each independent point cloud dataset, and determine the corrected point cloud position corresponding to each independent point cloud dataset. The distance registration unit is used to perform distance registration based on the corrected point cloud position corresponding to each independent point cloud dataset to determine the target point cloud dataset.

[0045] In an optional embodiment, the distance registration unit includes: The data point pair determination subunit is used to perform position registration of each point cloud data based on the corrected point cloud position, and determine the data point pair corresponding to each independent point cloud dataset. The point-to-point distance determination sub-unit is used to calculate the distance between each pair of data points to obtain the point-to-point distance. The comparison subunit is used to perform a distance comparison based on the point pair distance and a preset distance threshold to determine a first distance comparison result; and to take the point cloud data in each independent point cloud dataset whose point pair distance is not greater than the preset distance threshold as the target point cloud dataset.

[0046] In an optional embodiment, the processing module 805 includes: The model determination unit is used to delete data from the target point cloud dataset to obtain a point cloud scene model. The map determination unit is used to model the point cloud scene model to obtain the three-dimensional vector map.

[0047] In an optional embodiment, the model determination unit includes: The filtering sub-unit is used to filter overlapping regions in the target point cloud dataset to obtain the overlapping dataset. The fusion subunit is used to fuse the target point cloud data in the overlapping dataset to determine the fused point cloud dataset. The model building subunit is used to build a surface model of the fused point cloud dataset using the target algorithm, thereby obtaining the point cloud scene model.

[0048] In an optional embodiment, the map determination unit includes: The learning model acquisition sub-unit is used to acquire the preset learning model; The annotation subunit is used to annotate the target point cloud data in the point cloud scene model using the preset learning model, so as to obtain the annotated object and the annotation shape corresponding to the annotated object; The regularity determination subunit is used to determine the regularity of the labeled object based on the labeled shape corresponding to the labeled object; The fitting subunit is used to fit the regularity of the labeled objects to obtain the three-dimensional vector map, provided that the regularity of the objects indicates the regularity of the labeled objects.

[0049] In an optional embodiment, the above-described apparatus further includes: The manual annotation unit is used to send the irregularly annotated object to the display interface when the regularity of the object indicates that the annotated object is irregular, so that the user can annotate the irregularly annotated object to obtain the manually annotated object; The manual annotation fitting unit is used to receive manually annotated objects received by the display interface and fit the manually annotated objects to obtain the three-dimensional vector map.

[0050] In an exemplary embodiment, an electronic device is also provided, including: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement a route planning method for a flying car as described in the embodiments of this disclosure.

[0051] In an exemplary embodiment, a computer-readable storage medium is also provided, wherein when the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the route planning method for a flying car according to the embodiments of the present disclosure. In an exemplary embodiment, a computer program product or computer program is also provided, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the route planning method for a flying car provided in the various alternative implementations described above.

[0052] It is understood that in the specific embodiments of the present invention, user-related data is involved. When the above embodiments of the present invention are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0053] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0054] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

[0055] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A route planning method for a flying car, characterized in that, The method includes: In response to planning instructions, acquire multiple independent point cloud datasets and the point cloud position of each point cloud data in the multiple independent point cloud datasets; Based on the point cloud position of each point cloud data, data registration is performed to obtain multiple target point cloud datasets in a three-dimensional coordinate system; Data processing is performed on the multiple target point cloud datasets to obtain a three-dimensional vector map; Based on the aforementioned three-dimensional vector map and preset buffer rules, a safe flight zone is determined; Based on the safe flight area and the three-dimensional vector map, the flying car is subjected to three-dimensional spatial path planning to determine the target flight route.

2. The method according to claim 1, characterized in that, The data registration based on the point cloud position of each point cloud data to obtain multiple target point cloud datasets in a three-dimensional coordinate system includes: Obtain the coordinate transformation formula for each independent point cloud dataset; Based on the point cloud position of each point cloud data in each independent point cloud dataset and the coordinate transformation formula corresponding to each independent point cloud dataset, coordinate transformation is performed to determine the corrected point cloud position corresponding to each independent point cloud dataset. Based on the corrected point cloud position corresponding to each independent point cloud dataset, distance registration is performed to determine the target point cloud dataset.

3. The method according to claim 2, characterized in that, The step of determining the target point cloud dataset by performing distance registration based on the corrected point cloud position corresponding to each independent point cloud dataset includes: Based on the corrected point cloud position, position registration is performed on each point cloud data to determine the data point pair corresponding to each independent point cloud dataset; Calculate the distance between each pair of data points to obtain the point-to-point distance; Based on the point-to-point distance and the preset distance threshold, a distance comparison is performed to determine the first distance comparison result; and the point cloud data in each independent point cloud dataset whose point-to-point distance is not greater than the preset distance threshold, as indicated by the first distance comparison result, is taken as the target point cloud dataset.

4. The method according to claim 1, characterized in that, The process of processing the multiple target point cloud datasets to obtain a 3D vector map includes: Data deletion is performed on the target point cloud dataset to obtain a point cloud scene model; The point cloud scene model is modeled to obtain the three-dimensional vector map.

5. The method according to claim 4, characterized in that, The step of deleting data from the target point cloud dataset to obtain the point cloud scene model includes: Filter the overlapping regions in the target point cloud dataset to obtain the overlapping dataset; The target point cloud data in the overlapping dataset are fused to determine the fused point cloud dataset. The target algorithm is used to establish a surface model of the fused point cloud dataset to obtain the point cloud scene model.

6. The method according to claim 4, characterized in that, The process of modeling the point cloud scene model to obtain the 3D vector map includes: Obtain the preset learning model; The preset learning model is used to annotate the target point cloud data in the point cloud scene model to obtain the annotated object and the annotated shape corresponding to the annotated object. Based on the annotation shape corresponding to the annotated object, the object regularity of the annotated object is determined; When the regularity of the object indicates the regularity of the labeled object, the regular labeled object is fitted to obtain the three-dimensional vector map.

7. The method according to claim 6, characterized in that, The method further includes: When the regularity of the object indicates that the labeled object is irregular, the irregular labeled object is sent to the display interface so that the user can annotate the irregular labeled object to obtain a manually labeled object; The system receives manually labeled objects from the display interface and fits these objects to obtain the three-dimensional vector map.

8. A route planning device for a flying car, characterized in that, The device includes: The acquisition module is used to acquire, in response to planning instructions, multiple independent point cloud datasets and the point cloud position of each point cloud data in the multiple independent point cloud datasets; The registration module is used to perform data registration based on the point cloud position of each point cloud data to obtain multiple target point cloud datasets in a three-dimensional coordinate system. The processing module is used to process the multiple target point cloud datasets to obtain a three-dimensional vector map; The first determining module is used to determine a safe flight area based on the three-dimensional vector map and preset buffer rules; The second determining module is used to perform three-dimensional spatial path planning processing on the flying car based on the safe flight area and the three-dimensional vector map, and determine the target flight route.

9. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the route planning method for a flying car as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, When the instructions in the storage medium are executed by the processor of the electronic device, the electronic device is able to perform the route planning method for the flying car as described in any one of claims 1 to 7.

Images