Unmanned aerial vehicle inspection control system and method based on data analysis

Abstract

The invention relates to the technical field of unmanned aerial vehicle control, and discloses an unmanned aerial vehicle inspection control system and method based on data analysis, and the method comprises the steps: receiving environment feature data of a target region, and constructing a static environment model; obtaining kinetic parameters of the unmanned aerial vehicle, and calculating a feasible flight envelope in combination with the static environment model; after taking off, the unmanned aerial vehicle collects environment data in real time, and a multi-level environment topology model is constructed; acquiring state data in real time, and judging an obstacle avoidance level; if yes, local path planning is carried out, and an optimal flight path is generated; acquiring trajectory deviation and energy consumption data, dynamically calibrating the flight envelope, and optimizing the flight envelope; according to the method, environment perception, trajectory optimization and online learning are combined, the method is adaptive to a complex environment, and the intelligent level of unmanned aerial vehicle inspection is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle control, and more specifically, to an unmanned aerial vehicle inspection control system and method based on data analysis. Background Art

[0002] With the rapid development of drone technology, drones have been widely used in many fields such as power inspection, traffic monitoring, geological exploration, etc. Drones can replace manual tasks in complex environments, which can significantly improve work efficiency and reduce labor costs. However, drones still face many technical challenges when performing autonomous inspection tasks.

[0003] The Chinese patent with publication number CN112214032A discloses a drone inspection system and a drone inspection method. The method uses drones to perform layered inspections on target areas, detect abnormal points through thermal imaging, and collect high-definition images of abnormal targets. However, the patent does not consider the problem of flight obstacle avoidance in complex environments, and drones are prone to collisions, affecting inspection efficiency and safety.

[0004] The Chinese patent with publication number CN118672279A discloses a method and system for avoiding obstacles in low-altitude inspection drone flight. This method adopts different obstacle avoidance strategies according to the obstacle type, and can achieve a certain degree of autonomous obstacle avoidance. However, this method only considers the spatial relationship between the drone and the obstacle, and fails to make full use of environmental information and drone status data. The real-time and accuracy of obstacle avoidance decisions need to be improved.

[0005] A Chinese patent with the authorization announcement number CN117519278B discloses a drone obstacle avoidance method for bridge inspection. The method detects obstacles through binocular vision and plans obstacle avoidance paths based on the movement information of the obstacles. However, the method is mainly aimed at bridge scenes and lacks versatility. In addition, the method relies on visual sensors to collect data, and the obstacle avoidance performance will be greatly reduced in low visibility or poor lighting conditions.

[0006] In summary, the existing UAV inspection control technology has insufficient environmental perception capabilities, and it is difficult for UAVs to accurately obtain complex environmental information, resulting in low accuracy in obstacle avoidance decisions and path planning, and collisions are very likely to occur. Especially in scenes such as urban areas with tall buildings and rugged mountains, UAVs face greater obstacle avoidance pressure; and flight constraints are not fully considered. Most existing methods do not consider the dynamic characteristics of the UAV itself, as well as factors such as energy consumption and trajectory deviation during actual flight, resulting in the generated flight trajectory being infeasible or having poor execution effects. Summary of the invention

[0007] In order to overcome the above-mentioned defects of the prior art, the present invention provides a UAV inspection control system and method based on data analysis. Through key technologies such as environmental modeling, multi-level topological expression, online obstacle avoidance decision-making and trajectory optimization, it dynamically perceives the inspection environment and accurately analyzes the flight status of the UAV. On the premise of meeting various constraints, it adaptively performs path planning and obstacle avoidance control, and ultimately realizes efficient and safe inspection of UAVs in complex scenarios.

[0008] To achieve the above object, the present invention provides the following technical solutions: A UAV inspection control method based on data analysis, comprising: Receive environmental feature data of the target inspection area, and build a static environment model of the target inspection area based on the environmental feature data; obtain dynamic parameters of the UAV, and calculate a feasible flight envelope of the UAV by combining the dynamic parameters of the UAV and the static environment model; After the drone takes off, it collects real-time environmental data of the flight area in real time, and builds a multi-level environmental topology model based on the real-time environmental data and the static environmental model; during the inspection process, it obtains the status data of the drone in real time, and dynamically determines the obstacle avoidance level of the drone based on the status data of the drone and the multi-level environmental topology model. The obstacle avoidance level is divided into three levels: safe, warning, and dangerous; if the obstacle avoidance level is warning or dangerous, local path planning is triggered to generate a local optimal flight trajectory; The actual trajectory deviation data and energy consumption data of the UAV when it moves along the local optimal flight trajectory are obtained; based on the actual trajectory deviation data and the energy consumption data, the feasible flight envelope is dynamically calibrated to obtain an optimized flight envelope.

[0009] Furthermore, the construction of a static environment model of the target inspection area based on the environmental feature data includes: Preprocess the received environmental feature data, perform spatial division on the preprocessed environmental feature data, and divide the target inspection area into regular three-dimensional grids; extract the feature vector of each grid and mark it as a grid feature vector; Based on the extracted grid feature vectors, a static environment model of the target inspection area is constructed, and the static environment model is a three-dimensional grid map.

[0010] Further, the dynamic parameters include maximum speed, maximum acceleration and maximum climb rate; The feasible flight envelope of the computational drone includes: Combining the dynamic parameters of the UAV and the static environment model, calculating the initial flight envelope of the UAV, wherein the initial flight envelope is represented by a hexahedron; The initial flight envelope is subjected to collision detection based on the static environment model, and the envelope area where the collision occurs is clipped to obtain the final feasible flight envelope.

[0011] Further, the calculating the initial flight envelope of the UAV includes: Calculate a control cycle based on the maximum speed constraint in the UAV dynamics parameters The upper and lower bounds of the internal drone speed; Calculate a control cycle based on the maximum acceleration constraint in the UAV dynamics parameters The upper and lower bounds of the UAV’s acceleration are calculated; based on the upper and lower bounds of the velocity and the upper and lower bounds of the acceleration, the upper and lower bounds of the UAV’s displacement within a control cycle are calculated; The upper and lower bounds of the displacement are mapped into the Cartesian coordinate system to obtain the vertex coordinates of the flight envelope hexahedron and generate the initial flight envelope of the UAV.

[0012] Furthermore, the step of clipping the envelope region where the collision occurs to obtain the final feasible flight envelope includes: The vertex coordinate set of the initial flight envelope hexahedron is recorded as ,in, Indicates The coordinates of the vertices, 1≤i≤8; at the same time, the obstacle information is extracted from the static environment model, recorded as ,in, is a set of obstacles, Indicates obstacle information, m is the total number of obstacles, 1≤j≤m; For each vertex , traverse all obstacles, check the vertex according to the obstacle set O Whether it is located inside an obstacle; If the vertex If it is inside the obstacle, From the vertex coordinates set Remove it, and after traversal and detection, get the updated vertex set , where k is the number of vertices after the initial flight envelope is updated, ; With the updated vertex set Based on this, the flight envelope polyhedron is reconstructed to obtain the final feasible flight envelope.

[0013] Furthermore, the checking vertex Whether it is inside the obstacle includes: calculation Directed distance to each obstacle surface , if there is a distance , then it means Outside the jth obstacle; if there is a distance , then it means Inside the jth obstacle.

[0014] Furthermore, the real-time environmental data includes laser radar point cloud data and visual image data; The construction of a multi-level environment topology model based on real-time environment data and a static environment model includes: Splice the LiDAR point cloud data into a local point cloud map in a unified coordinate system; extract obstacle semantic information from visual image data; The local point cloud map is integrated with the obstacle semantic information to construct a semantic point cloud map; The semantic point cloud map is registered with the static environment model, and the registered semantic point cloud map is superimposed with the static environment model to form a multi-level environment topology model.

[0015] Furthermore, the dynamic determination of the obstacle avoidance level of the drone includes: Acquire the status data of the drone at a preset frequency, wherein the status data includes position information, speed information and attitude information, and the attitude information includes attitude angle; Calculate the minimum distance between the drone and obstacles in the environment based on the drone's status data and the multi-level environment topology model , and calculate the relative speed between the drone and the nearest obstacle and relative acceleration ; Combined minimum distance , relative speed and relative acceleration , dynamically determine the current obstacle avoidance level of the drone, which is divided into three levels: safe, warning, and dangerous; The dynamic judgment of the current obstacle avoidance level of the drone includes: Preset the first safety threshold , the second safety threshold , the first speed threshold , the second speed threshold , the first acceleration threshold and the second acceleration threshold ;in ; like and and , then the obstacle avoidance level is safe; like ,or ,or , then the obstacle avoidance level is warning; like ,or ,or , the obstacle avoidance level is dangerous.

[0016] Further, generating a local optimal flight trajectory includes: Get the preset global trajectory of the drone. There are n1 waypoints in the global trajectory. Take the current position of the drone as the starting point and the waypoint closest to the current position on the global trajectory as the end point to plan the local optimal flight trajectory. The feasible flight envelope is used as the envelope constraint of the UAV motion, the multi-level environment topology model is used as the flight environment constraint, and the maximum speed of the UAV is and maximum acceleration As dynamic constraints, the overall optimization goal is established according to the total flight distance, total energy consumption and flight time, and the local trajectory optimization model is constructed; A heuristic search algorithm is used to search for feasible flight trajectories from the starting point to the end point within the feasible flight envelope. For the searched feasible flight trajectories, a numerical optimization algorithm is used to solve the local trajectory optimization model and obtain the local optimal flight trajectory.

[0017] Furthermore, the dynamically calibrating the feasible flight envelope includes: Acquire actual flight trajectory data when the UAV moves along the local optimal flight trajectory, compare the actual flight trajectory data with the local optimal flight trajectory, and calculate trajectory deviation data, wherein the trajectory deviation data includes heading deviation, altitude deviation, and speed deviation; Acquire actual energy consumption data of the UAV when it moves along the local optimal flight trajectory, wherein the actual energy consumption data includes battery voltage, current and power; and calculate energy consumption distribution in different areas of the feasible flight envelope based on the actual energy consumption data; Based on the trajectory deviation data and energy consumption distribution, the feasible flight envelope is adaptively adjusted to obtain the adjusted flight envelope vertex coordinates; based on the adjusted flight envelope vertex coordinates, the flight envelope polyhedron is reconstructed to obtain the optimized flight envelope.

[0018] A drone inspection control system based on data analysis, which is used to implement the above-mentioned drone inspection control method based on data analysis, the system comprises: The flight envelope generation module is used to receive the environmental characteristic data of the target inspection area, and build a static environment model of the target inspection area based on the environmental characteristic data; obtain the dynamic parameters of the UAV, and calculate the feasible flight envelope of the UAV by combining the dynamic parameters of the UAV and the static environment model; Trajectory optimization module: After the drone takes off, it collects real-time environmental data of the flight area in real time, and builds a multi-level environmental topology model based on the real-time environmental data and the static environmental model; during the inspection process, it obtains the status data of the drone in real time, and dynamically determines the obstacle avoidance level of the drone based on the status data of the drone and the multi-level environmental topology model. The obstacle avoidance level is divided into three levels: safe, warning, and dangerous; if the obstacle avoidance level is warning or dangerous, local path planning is triggered to generate a local optimal flight trajectory; Flight envelope optimization module: used to obtain the actual trajectory deviation data and energy consumption data of the UAV when it moves along the local optimal flight trajectory; based on the actual trajectory deviation data and the energy consumption data, dynamically calibrate the feasible flight envelope to obtain the optimized flight envelope.

[0019] An electronic device includes a memory, a central processing unit, and a computer program stored in the memory and executable on the central processing unit. When the central processing unit executes the computer program, the above-mentioned unmanned aerial vehicle inspection control method based on data analysis is implemented.

[0020] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed, implements the above-mentioned unmanned aerial vehicle inspection control method based on data analysis.

[0021] Compared with the prior art, the present invention has the following beneficial effects: The present invention constructs a static environment model, which can provide prior constraints for autonomous obstacle avoidance and trajectory optimization of UAVs, make up for dynamic perception blind spots, and improve flight safety. Through spatial data mining and three-dimensional reconstruction technology, the fusion expression of multi-source heterogeneous environmental data is realized, and the perception foundation of intelligent inspection is consolidated. The concept of feasible flight envelope is proposed, and the achievable flight space of the UAV in the current state is described by considering the UAV dynamic constraints and environmental collision constraints. The envelope is used for trajectory search optimization to reduce the computational complexity and improve real-time performance. A multi-level environmental topology model is constructed, and the global static map and local dynamic perception are integrated to form a unified hierarchical environmental expression. Different from traditional static modeling or dynamic mapping, the model takes into account global priors and local observations, and has stronger adaptability to environmental changes. A state-based hierarchical obstacle avoidance strategy is proposed, and the obstacle avoidance level is divided into safety, warning and danger according to the distance and relative speed between the UAV and the obstacle. The strategy makes full use of environmental perception information, quantifies the collision risk, makes adaptive obstacle avoidance decisions, and overcomes the limitations of a single obstacle avoidance standard. When the obstacle avoidance level is warning or dangerous, local path planning is triggered, and the local optimal flight trajectory is searched under the premise of ensuring safety. Different from global offline planning, this method has stronger real-time and flexibility and can cope with dynamic unknown environments. The actual flight data of the UAV, including trajectory deviation and energy consumption distribution, is obtained for online calibration of the flight envelope. By adaptively adjusting the envelope shape through actual flight data, the influence of factors such as environmental interference and model error can be overcome, so that planning and control form a closed loop and continuously iterate and optimize. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0023] Figure 1 This is a principle flow chart of a UAV inspection control method based on data analysis in the present invention; Figure 2 A flowchart of a method for constructing a static environment model of a target inspection area based on environmental feature data in a data analysis-based unmanned aerial vehicle inspection control method of the present invention; Figure 3 A flowchart of a method for calculating a feasible flight envelope of a UAV in a UAV inspection control method based on data analysis of the present invention; Figure 4 A flow chart of a method for calculating the initial flight envelope of a UAV in a UAV inspection control method based on data analysis of the present invention; Figure 5 A flowchart of a method for constructing a multi-level environmental topology model in a UAV inspection control method based on data analysis of the present invention; Figure 6 A flowchart of a method for dynamically judging the obstacle avoidance level of a drone in a drone inspection control method based on data analysis of the present invention; Figure 7 A flowchart of a method for dynamically calibrating a feasible flight envelope in a UAV inspection control method based on data analysis of the present invention; Figure 8 This is a functional module diagram of a UAV inspection control system based on data analysis in the present invention. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] Example 1

[0026] See also Figure 1 As shown, this embodiment provides a drone inspection control method based on data analysis, including: Step S1000, receiving environmental characteristic data of the target inspection area, and constructing a static environmental model of the target inspection area based on the environmental characteristic data; obtaining dynamic parameters of the UAV, and calculating a feasible flight envelope of the UAV by combining the dynamic parameters of the UAV and the static environmental model; Furthermore, step S1000 includes: Step S1100, receiving environmental feature data of a target inspection area, and constructing a static environmental model of the target inspection area based on the environmental feature data; Furthermore, if Figure 2 As shown, step S1100 includes: Step S1110, receiving environmental characteristic data of a target inspection area, wherein the environmental characteristic data includes terrain complexity, building density, and risk area distribution; Step S1120, preprocessing the received environmental characteristic data, performing spatial division on the preprocessed environmental characteristic data, and dividing the target inspection area into a regular three-dimensional grid; Step S1130, extracting a feature vector of each grid and marking it as a grid feature vector, wherein the grid feature vector includes terrain height, building proportion and risk level; Step S1140: constructing a static environment model of the target inspection area based on the extracted grid feature vectors, wherein the static environment model is a three-dimensional grid map.

[0027] Specifically, step S1100 first receives environmental feature data of the inspection area through multi-source heterogeneous sensors. The terrain complexity can be obtained through the digital elevation model (DEM), the building density can be obtained through remote sensing image segmentation, and the risk area distribution can be given by expert annotation. The obtained heterogeneous data is preprocessed by data cleaning, feature normalization, etc. to improve the accuracy and consistency of the data. Then, the entire inspection area is divided into a regular three-dimensional grid using space partitioning algorithms such as octree and BSP tree. The octree recursively divides the space into eight sub-areas, which is suitable for expressing sparse environments; the BSP tree recursively divides the plane, which is more suitable for dense environments. The grid size can balance the fineness of environmental expression and the computational storage overhead. Then, the feature vector of each grid is extracted, and the discrete environmental information such as DEM elevation value, building pixel ratio, and risk area label is mapped to the feature space. Finally, based on the grid and its feature vector, a static environment model is constructed, which is usually represented by a three-dimensional grid map such as octree and BSP tree. The static environment model describes the inherent geographical environment attributes of the inspection area and is the prior knowledge for autonomous obstacle avoidance and trajectory optimization of the drone. The introduction of static environment models can provide constraints for inspection decisions, avoid dynamic perception blind spots, and improve flight safety. At the same time, the hierarchical process of environmental modeling from data acquisition, preprocessing to spatial expression can gradually simplify complex environments, reduce the dimensions of optimization problems, and improve computational efficiency. This step uses spatial data mining and 3D reconstruction technology to fuse multi-source heterogeneous data into a unified environmental representation, consolidating the environmental perception foundation of intelligent inspection.

[0028] Step S1200, obtaining the dynamic parameters of the UAV, and calculating the feasible flight envelope of the UAV by combining the dynamic parameters of the UAV and the static environment model; Furthermore, if Figure 3 As shown, step S1200 includes: Step S1210, obtaining the dynamic parameters of the UAV, wherein the dynamic parameters include maximum speed, maximum acceleration and maximum climb rate; Step S1220, combining the dynamic parameters of the UAV and the static environment model, calculating the initial flight envelope of the UAV, wherein the initial flight envelope is represented by a hexahedron; Furthermore, if Figure 4 As shown, step S1220 includes: Step S1221, based on the maximum speed constraint in the UAV dynamics parameters, calculate a control cycle The upper and lower bounds of the internal drone speed; Step S1222, based on the maximum acceleration constraint in the UAV dynamics parameters, calculate a control cycle The upper and lower bounds of the UAV’s acceleration are calculated; based on the upper and lower bounds of the velocity and the upper and lower bounds of the acceleration, the upper and lower bounds of the UAV’s displacement within a control cycle are calculated; Step S1223, mapping the upper and lower bounds of the displacement into a Cartesian coordinate system, obtaining the vertex coordinates of the flight envelope hexahedron, and generating the initial flight envelope of the UAV.

[0029] The upper and lower bounds of the drone’s speed are calculated as follows: Assume that the current speed of the drone is , the maximum speed is , then the upper bound of the velocity is: ; The lower bound is: ; in is the maximum acceleration of the drone, is a control cycle step. This calculation combines the current speed of the drone with its dynamic performance constraints to obtain the theoretical range of speed changes in the next cycle. The upper and lower bounds of the speed are respectively limited by the current speed and the maximum acceleration. and The function performs logical judgment.

[0030] Based on the maximum acceleration constraint in the UAV dynamics parameters, the upper and lower bounds of acceleration are obtained: the upper bound of acceleration is , the lower bound of the acceleration is ; Based on the velocity and acceleration boundaries, the upper and lower bounds of displacement can be further calculated ,in: ; This formula is based on the uniform acceleration motion model in physics and reflects the constraint relationship between velocity, acceleration and displacement. The displacement calculation is based on the reference point (the current position of the drone), so the result is a relative displacement boundary.

[0031] Map the upper and lower bounds of velocity, acceleration, and displacement to the Cartesian coordinate system. Suppose the current position of the drone is , the displacement boundary in the Cartesian coordinate system is , , then the vertex coordinates of the flight envelope hexahedron are: ; After this step is completed, a complete initial flight envelope model can be obtained, which describes the achievable flight area of ​​the UAV within a control cycle in the current state.

[0032] Step S1230, performing collision detection on the initial flight envelope based on the static environment model, clipping the envelope area where the collision occurs, and obtaining a final feasible flight envelope.

[0033] Further, step S1230 includes: Step S1231, record the vertex coordinate set of the initial flight envelope hexahedron as ,in, Indicates The coordinates of the vertices, 1≤i≤8; at the same time, the obstacle information is extracted from the static environment model, recorded as ,in, is a set of obstacles, Indicates obstacle information, m is the total number of obstacles, 1≤j≤m; Step S1232, for each vertex , traverse all obstacles, check the vertex according to the obstacle set O Whether it is inside an obstacle.

[0034] calculate Directed distance to each obstacle surface , if there is a distance , then it means Outside the jth obstacle; if there is a distance , then it means Inside the jth obstacle.

[0035] Step S1233, if If it is inside the obstacle, From the vertex coordinates set Remove it, and after traversal and detection, get the updated vertex set , where k is the number of vertices after the initial flight envelope is updated, ; Step S1234, using the updated vertex set Based on this, the flight envelope polyhedron is reconstructed to obtain the final feasible flight envelope.

[0036] It should be noted that due to collision clipping, the flight envelope may degenerate into an irregular polyhedron. In this case, it is necessary to generate a new polyhedron topology structure based on the actual connection relationship of the vertices in V' to represent the clipped flight envelope. Common degeneration situations include: Tetrahedron: When only 4 vertices remain after pruning, the flight envelope degenerates into a tetrahedron. The vertex connectivity of a tetrahedron can be directly enumerated.

[0037] Pentahedron: When only 5 vertices are left after clipping, the flight envelope degenerates into a pentahedron. The topology of the pentahedron can be generated by the Delaunay triangulation algorithm.

[0038] Hexahedron: When the number of vertices after clipping is still 6, the flight envelope degenerates into an irregular hexahedron. The topological structure of the hexahedron needs to be determined based on the positional relationship of the vertices.

[0039] Heptahedron: When the number of vertices after clipping is 7, the flight envelope degenerates into a heptahedron. Heptahedron topology can be generated by the alpha shape algorithm.

[0040] Through the above steps, the final flight envelope after collision detection and clipping can be obtained. This envelope eliminates the area occupied by obstacles and provides the space where the drone can actually fly safely.

[0041] Alpha shape and Delaunay triangulation are commonly used polyhedral surface reconstruction algorithms in computational geometry, which can generate a watertight polyhedral topological structure based on scattered point clouds. Combining these two methods can achieve dynamic clipping of the flight envelope.

[0042] The introduction of collision detection and envelope clipping can effectively improve the environmental adaptability of drones. The clipped flight envelope excludes the area occupied by obstacles and only retains the safe airspace that the drone can actually reach, avoiding the blind search for infeasible paths and reducing the invalid calculation of trajectory planning. At the same time, envelope clipping can also be regarded as an incremental motion constraint update, which enables drones to have a certain degree of autonomous decision-making ability in a dynamic environment. When a new obstacle enters the detection range, the envelope can shrink in real time to ensure safety; when the obstacle leaves, the envelope can automatically expand to gain greater freedom of movement. In addition, envelope clipping limits the pathfinding problem to a convex polyhedron, which can effectively reduce the time complexity of subsequent trajectory search.

[0043] Step S1200 further obtains the dynamic parameters of the drone itself based on the environmental modeling. Through ground experiments or simulated flight tests, the dynamic limits of the drone, such as the maximum speed, acceleration, and climb rate, are obtained as hard constraints of the flight envelope. On the basis of the static environmental model, combined with the dynamic constraints, the feasible flight envelope of the drone at the current position can be calculated. The flight envelope represents the accessible flight area of ​​the drone under the obstacle avoidance constraint. The flight envelope is dynamically updated as the state of the drone, such as the position and speed, changes. Reasonable setting of the flight envelope size and narrowing the drone motion search space under the obstacle avoidance constraint can significantly reduce the time complexity of trajectory planning. It should be noted that the feasible flight envelope is calculated based on the static environment and the drone's extreme performance. With the emergence of dynamic obstacles during the flight process, the envelope size needs to be corrected in real time to ensure obstacle avoidance safety and trajectory reachability. This step fully considers the heterogeneous models of drones and environments. Under the premise of meeting the obstacle avoidance constraint, the drone's motion search interval is compressed from the two dimensions of state space and environment space, laying the foundation for rapid trajectory optimization. At the same time, the envelope algorithm overcomes the unstructured constraints of terrain and obstacles, transforms the optimization problem into a convex optimization solution, and can significantly reduce the computational complexity. As an adaptive boundary idea, the flight envelope gives the UAV a certain degree of autonomous decision-making ability, enabling it to autonomously adjust the range of motion in a dynamic unknown environment, improving environmental adaptability and robustness.

[0044] Step S2000: After the UAV takes off, real-time environmental data of the flight area is collected in real time, and a multi-level environmental topology model is constructed based on the real-time environmental data and the static environmental model; during the inspection process, the status data of the UAV is obtained in real time, and the obstacle avoidance level of the UAV is dynamically determined based on the status data of the UAV and the multi-level environmental topology model. The obstacle avoidance level is divided into three levels: safety, warning, and danger; if the obstacle avoidance level is warning or danger, local path planning is triggered to generate a local optimal flight trajectory; actual trajectory deviation data and energy consumption data of the UAV when moving along the local optimal flight trajectory are obtained; based on the actual trajectory deviation data and the energy consumption data, a feasible flight envelope is dynamically calibrated to obtain an optimized flight envelope.

[0045] Furthermore, step S2000 includes: Step S2100: After the UAV takes off, real-time environmental data of the flight area is collected in real time, and a multi-level environmental topology model is constructed based on the real-time environmental data and the static environmental model; Furthermore, if Figure 5 As shown, step S2100 includes: Step S2110, after the drone takes off, real-time environmental data of the flight area is collected in real time, and the real-time environmental data includes laser radar point cloud data and visual image data; Step S2120, stitching the laser radar point cloud data into a local point cloud map in a unified coordinate system; extracting obstacle semantic information from the visual image data; Step S2130, fusing the local point cloud map with the obstacle semantic information to construct a semantic point cloud map; Step S2140, registering the semantic point cloud map with the static environment model, and superimposing the registered semantic point cloud map with the static environment model to form a multi-level environment topology model.

[0046] Specifically, step S2100 first obtains real-time environmental data of the area around the drone through sensors such as airborne laser radar and visible light camera. The laser radar can provide accurate point cloud data to reflect the geometric information such as the spatial position and shape of obstacles; the visible light camera can provide rich visual information such as texture and color to reflect the category and semantic attributes of obstacles. The raw data obtained by the sensor often has quality problems such as noise and outliers, and necessary data preprocessing is required, such as point cloud denoising, downsampling, image distortion correction, contrast enhancement, etc., to improve the signal-to-noise ratio and availability of the data.

[0047] On the basis of data preprocessing, further information extraction is performed on the laser point cloud and visual image. For the laser point cloud, classic point cloud registration algorithms such as ICP (Iterative Closest Point) and NDT (Normal Distributions Transform) are used to stitch the local point clouds obtained at different times and different perspectives into a complete point cloud map in a unified coordinate system. The stitching process needs to optimize the rigid body transformation matrix between point clouds and minimize the distance error between different point clouds. The stitched point cloud map has a larger perception range and higher resolution, which can accurately reflect the spatial distribution of dynamic obstacles in the near range of the drone and provide a geometric reference for real-time obstacle avoidance. For visual images, semantic segmentation algorithms based on deep learning such as PSPNet (Pyramid Scene Parsing Network) and Mask R-CNN (Regions with Convolutional Neural Network) are used to identify the pixel-level position and contour of different categories of obstacles (such as buildings, trees, poles, etc.) from the image. PSPNet uses a pyramid pooling module to extract image features at different scales, perform upsampling fusion, and finally achieve pixel-level classification. Based on the object detection framework Faster R-CNN, Mask R-CNN adds a segmentation branch parallel to bounding box regression and classification to generate pixel-by-pixel masks and achieve instance-level semantic segmentation. The semantic segmentation results contain rich obstacle categories and contour information, which can be used to distinguish dynamic obstacles from static backgrounds, analyze the movement trends of obstacles, such as the future trajectories of moving obstacles such as pedestrians and vehicles, and provide semantic basis for obstacle avoidance decisions.

[0048] By fusing the laser point cloud with the visual semantic segmentation results, a semantic point cloud map can be constructed. Based on the visual semantic segmentation results, each point in the laser point cloud is marked with an obstacle category to give the point cloud additional semantic attributes. The fusion process requires the data of the lidar and camera to be synchronized in time and space and aligned in coordinates. Usually, the external parameters of the camera-lidar are calibrated to achieve coordinate unification of data from different sensors. The semantic point cloud map associates the geometric information of the laser point cloud with the semantic information of the visual image. It contains both the precise spatial position of dynamic obstacles and the attribute descriptions of the obstacle category, contour, etc. It is a hybrid representation of tightly coupled geometric and semantic information. This multimodal, multi-level fusion perception can provide drones with a more comprehensive and reliable understanding of dynamic environments.

[0049] In order to achieve global positioning and obstacle avoidance, it is necessary to register and fuse the local semantic point cloud map with the static environment model constructed in step S1100. The static environment model is a priori global map, which provides a background reference for the global environment; while the semantic point cloud map is a local dynamic map, which provides real-time updates of dynamic obstacles in the vicinity of the drone. Using the static environment model as the global coordinate system, the local semantic point cloud is registered to achieve the conversion from the local map to the global map. This process requires estimating the real-time pose of the drone in the static environment model and solving the rigid body transformation from the local map to the global map. Commonly used positioning algorithms include laser SLAM (Simultaneous Localization and Mapping), visual SLAM, etc., which achieve high-precision positioning and mapping by optimizing the constraints of pose and map. The semantic point cloud and the static environment model after registration and fusion are in the same global unified coordinate framework, and the local dynamic information is organically embedded in the global environment background, which expands the perception scale of the drone.

[0050] In order to improve the efficiency of data query and management, it is necessary to further optimize the data structure of the registered and fused environmental map. Commonly used spatial data structures include octrees, Kdtrees, etc. Through recursive space partitioning, large-scale point clouds and voxels are organized in an orderly manner to achieve fast neighbor search, collision detection, etc. The standardized local semantic point cloud is inserted into the octree node corresponding to the static environmental model to achieve seamless integration of global and local maps. The multi-level environmental topological model finally constructed contains both global-scale environmental background information and real-time update information of local dynamic obstacles. It integrates multi-source heterogeneous data of prior and perception, static and dynamic, and comprehensively describes the flight environment of the drone.

[0051] Multi-level environmental perception can significantly expand the UAV's ability to understand unknown dynamic environments. Traditional UAV environmental perception is mostly limited to local perception or static priors, and it is difficult to cope with complex dynamic environments. The multi-sensor collaborative perception and multi-data fusion modeling proposed in this step can achieve multi-dimensional environmental representation from local to global, from static to dynamic. The introduction of semantic elements not only enhances the UAV's interpretation of the environment, but also lays the foundation for subsequent semantic navigation and mission planning. The construction of a multi-level environmental topology model inherits the management characteristics of the octree for large-scale point clouds, while taking into account the flexibility of dynamic updates, providing a hierarchical environmental organization method that takes into account both global information and local details. This dynamic perception-oriented environmental understanding provides key information support for subsequent autonomous decision-making and planning of UAVs.

[0052] Step S2200, during the inspection process, the state data of the drone is obtained in real time, and the obstacle avoidance level of the drone is dynamically determined based on the state data of the drone and the multi-level environmental topology model; Furthermore, if Figure 6 As shown, step S2200 includes: Step S2210, obtaining status data of the drone at a preset frequency, wherein the status data includes position information, speed information and attitude information, and the attitude information includes an attitude angle; Step S2220, calculating the minimum distance between the drone and obstacles in the environment based on the drone's status data and the multi-level environment topology model , and calculate the relative speed between the drone and the nearest obstacle and relative acceleration ; Step S2230, combining the minimum distance , relative speed and relative acceleration , dynamically determine the current obstacle avoidance level of the drone, which is divided into three levels: safety, warning, and danger.

[0053] The dynamic judgment of the current obstacle avoidance level of the drone includes: Preset the first safety threshold , the second safety threshold , the first speed threshold , the second speed threshold , the first acceleration threshold and the second acceleration threshold ;in ; like and and , then the obstacle avoidance level is safe; ,or ,or , then the obstacle avoidance level is warning; like ,or ,or , the obstacle avoidance level is dangerous.

[0054] Specifically, step S2200 first obtains the status data of the drone in real time at a certain frequency through sensors such as GPS and IMU. GPS can provide real-time location information (such as longitude and latitude, altitude, etc.) of the drone in the global geographic coordinate system, and IMU can provide real-time speed, acceleration, attitude angle and other motion information of the drone in the local coordinate system. The drone status data and the multi-level environmental topological model need to be spatiotemporally aligned to establish a mapping relationship between the two. Time alignment can be achieved by aligning timestamps, and space alignment requires coordinate transformation. The rigid body transformation matrix of the drone relative to the multi-level environmental topological model can be estimated through pose estimation algorithms, such as extended Kalman filtering (EKF), nonlinear optimization, etc., to achieve the conversion from the local coordinate system to the global coordinate system, and then obtain the relative position relationship between the drone and the obstacle. After alignment, the multi-level environmental topological model can be transformed to align the model to the local coordinate system with the drone as the origin, thereby simplifying the calculation of subsequent parameters such as distance and speed.

[0055] Then, based on the state and environment registration, the collision risk between the drone and the obstacle is analyzed. By traversing the obstacles in the multi-level topological model, the distance between the drone and each obstacle is calculated, and the minimum distance is found. , as a safety margin. At the same time, the relative speed between the drone and the nearest obstacle is estimated based on the relative position changes between the two consecutive frames. and relative acceleration . The distance reflects the possibility of collision, the relative speed reflects the urgency of the collision, and the relative acceleration reflects the changing trend of the threat. With reference to manual driving experience, multiple safety thresholds such as minimum distance, relative speed, acceleration, etc. are set to classify the collision risk. When all indicators are better than the safety threshold, the collision risk is the lowest and the drone is in a safe flight state. When an indicator exceeds the safety threshold but does not reach the danger threshold, the collision risk is high, the drone is in an alert state, and close attention should be paid to the dynamics of obstacles. When any indicator exceeds the danger threshold, the collision risk is extremely high, the drone is in a dangerous state, and avoidance measures must be taken immediately.

[0056] Compared with a single distance threshold, multi-parameter obstacle avoidance criteria can more comprehensively assess collision risks, especially for scenarios with high-speed moving obstacles. Distance alone cannot accurately determine the urgency of a collision, and the relative movement trend of both parties must also be considered. The introduction of relative speed can determine the approach rate of obstacles; the introduction of relative acceleration can determine the evolution trend of threats and predict collision risks in advance. Multi-parameter obstacle avoidance criteria form a set of adaptive obstacle avoidance strategies. For different risk levels, drones can maintain their current status, increase vigilance, and actively avoid obstacles, taking into account both flight safety and mission efficiency. Compared with fixed obstacle avoidance strategies, adaptive strategies can dynamically adjust the obstacle avoidance behavior of drones according to the complexity of the flight environment to improve environmental adaptability.

[0057] In summary, step S2200 proposes a multi-parameter obstacle avoidance criterion, introduces multiple measurement indicators such as distance, speed, acceleration, etc., performs graded assessment and dynamic warning on the collision risk of the UAV, and forms an adaptive obstacle avoidance strategy. This strategy makes up for the shortcomings of a single distance criterion, can capture the movement trend of obstacles, and predict future collision risks. By analyzing the spatial relationship and collision risk between the UAV and obstacles in real time, it can provide a basis for the flight decision of the UAV, trigger active obstacle avoidance in dangerous situations, and ensure flight safety. At the same time, through the classification of obstacle avoidance levels, the UAV is prevented from being overly sensitive, unnecessary obstacle avoidance behaviors are reduced, and flight stability and mission continuity are guaranteed under the premise of safety. This step makes full use of multi-source heterogeneous sensor information, improves the perception ability of the UAV, enhances its survivability in complex environments, and lays the foundation for subsequent trajectory optimization.

[0058] Step S2300: If the obstacle avoidance level is warning or danger, local path planning is triggered to generate a local optimal flight trajectory; Further, step S2300 includes: Step S2310, obtaining the preset global trajectory of the drone, wherein the global trajectory is provided with n1 waypoints, taking the current position of the drone as the starting point and the waypoint closest to the current position on the global trajectory as the end point, and planning the local optimal flight trajectory; Step S2320: taking the feasible flight envelope as the envelope constraint of the UAV motion, taking the multi-level environment topology model as the flight environment constraint, and taking the maximum speed of the UAV as the and maximum acceleration As dynamic constraints, the overall optimization goal is established according to the total flight distance, total energy consumption and flight time, and the local trajectory optimization model is constructed; The local trajectory optimization model is: Objective function: ; Envelope constraints: ; Flight environment constraints: ; Dynamic constraints: ; in: : Indicates the time on the trajectory The battery voltage at the moment; : Indicates time The current at the moment; : Drag coefficient, which indicates the influence of air resistance; : The speed of the drone at time t; : Total flight distance, which indicates the length of the trajectory from the starting point to the end point. The shorter the distance, the better the trajectory and the less time it takes. In complex environments, distance optimization can reduce flight detours. : Total energy consumption, which represents the power consumption of the drone in the entire local trajectory; lower energy consumption means a more economical flight path, which is especially important for long-distance flight missions. Optimizing energy consumption can extend the drone's flight time and reduce the number of flights; : Flight time, which indicates the time required for the drone to fly along the trajectory; is the weight coefficient of the total flight distance, is the weight coefficient of total energy consumption, is the weight coefficient of flight time and satisfies ; : feasible flight envelope area; : The position of each point on the UAV trajectory; : The set of obstacles in the environment; : The acceleration of the UAV at time t.

[0059] As the total flight distance increases, the objective function As the total energy consumption increases, the objective function As the flight time increases, the optimization model tends to reduce energy consumption by reducing current or speed to achieve the optimal energy consumption. Increase, the objective function Increase, prompting the model to choose a faster path, suitable for optimizing flight paths in time-sensitive scenarios. The model combines the three indicators of total flight distance, total energy consumption, and flight time to ensure that the drone achieves the optimal balance of flight paths while meeting flight safety. By weighted summation of distance, energy consumption, and time, the trajectory optimization model achieves efficient and economical flight path selection.

[0060] Step S2330, using a heuristic search algorithm to search for a feasible flight trajectory from the starting point to the end point within the feasible flight envelope; for the searched feasible flight trajectory, using a numerical optimization algorithm to solve the local trajectory optimization model to obtain the local optimal flight trajectory.

[0061] Specifically, step S2300 further triggers local path planning based on the obstacle avoidance decision to guide the drone to safely and efficiently fly around obstacles. When the obstacle avoidance decision result is a warning or danger level, it indicates that the risk of collision between the drone and the obstacle is high. At this time, the global trajectory is likely no longer safe and it is necessary to make local adjustments to the trajectory in a timely manner. Local path planning starts from the current position of the drone and ends at the waypoint closest to the current position on the global trajectory, searching for the local optimal flight path between these two points. The nearest waypoint is selected as the end point to make the local trajectory as consistent as possible with the original global trajectory, reduce the cost of flying around, and improve flight efficiency.

[0062] Local path planning uses the feasible flight envelope as the feasible motion space of the drone. The feasible flight envelope excludes the area occupied by obstacles, thereby constraining the movement of the drone within a safe airspace and avoiding the risk of collision with obstacles. The multi-level environmental topology model reflects the static terrain and dynamic obstacle distribution around the drone, providing the necessary environmental information for path planning. Local path planning uses heuristic search algorithms, such as Dijkstra's shortest path algorithm, Heuristic search, etc., searches for feasible paths from the starting point to the end point within the flight envelope. Compared with blind searches such as random sampling, heuristic search uses prior knowledge of the problem to guide the search direction, such as The algorithm uses an estimated cost function to prioritize expanding nodes with lower costs and the potential to reach the target first, thereby significantly improving search efficiency.

[0063] In order to further improve the flight performance of the local trajectory, on the basis of searching for a feasible flight trajectory, it is also necessary to optimize and solve the trajectory. The local trajectory optimization takes the total flight distance, total energy consumption, flight time and other indicators as the optimization targets, and establishes a local trajectory optimization model. The local trajectory optimization model also takes into account factors such as the UAV's dynamic characteristics and maneuver constraints to be closer to the actual flight characteristics of the UAV. The trajectory optimization problem can be transformed into a nonlinear programming problem and solved using numerical optimization algorithms such as the gradient descent method, Newton's method, and the conjugate gradient method. The optimized local optimal flight trajectory can not only avoid obstacles, but also has better performance indicators such as range and energy consumption.

[0064] After the local trajectory flight ends, the drone will switch back to the global trajectory flight mode. To achieve a smooth switch, the local trajectory needs to be spliced ​​with the original global trajectory. When splicing, attention should be paid to the continuity and smoothness of the trajectory, avoiding sharp corners or sudden changes to ensure the stability of the drone's movement. A common splicing strategy is to interpolate between the end point of the local trajectory and the nearest waypoint of the global trajectory to generate a smooth transition trajectory.

[0065] In general, local path planning is an important mechanism for dealing with dynamic environments. The introduction of local planning gives the drone a certain degree of local intelligence, enabling it to autonomously adjust its trajectory based on real-time perception results and flexibly avoid sudden obstacles. At the same time, the local trajectory can be seamlessly connected with the global trajectory, ensuring safety while taking into account flight efficiency. Local path planning is a bottom-up planning approach that gradually approaches the global optimum through local optimality, providing strong support for drones' autonomous decision-making in complex environments.

[0066] Step S2400, obtaining actual trajectory deviation data and energy consumption data when the UAV moves along the local optimal flight trajectory; based on the actual trajectory deviation data and the energy consumption data, dynamically calibrating the feasible flight envelope calculated in step S1200 to obtain an optimized flight envelope; Furthermore, if Figure 7 As shown, step S2400 includes: Step S2410, obtaining actual flight trajectory data when the UAV moves along the local optimal flight trajectory, comparing the actual flight trajectory data with the local optimal flight trajectory, and calculating trajectory deviation data, wherein the trajectory deviation data includes heading deviation, altitude deviation, and speed deviation; Step S2420, obtaining actual energy consumption data of the UAV when it moves along the local optimal flight trajectory, wherein the actual energy consumption data includes battery voltage, current and power; based on the actual energy consumption data, calculating energy consumption distribution in different areas of the feasible flight envelope; Step S2430, based on the trajectory deviation data and the energy consumption distribution, adaptively adjust the feasible flight envelope to obtain the adjusted flight envelope vertex coordinates; Step S2440, based on the adjusted flight envelope vertex coordinates, reconstruct the flight envelope polyhedron to obtain an optimized flight envelope.

[0067] Specifically, step S2400 is a closed-loop feedback optimization process, which evaluates the trajectory tracking accuracy and energy efficiency by obtaining the actual flight data of the UAV, and feedback is used to adjust the flight envelope, so as to realize the adaptive correction of the optimal flight envelope in a dynamic environment. Among them, the trajectory deviation data reflects the accuracy of the UAV flight control and the degree of environmental interference. The real-time posture of the UAV is obtained by navigation equipment such as the airborne inertial measurement unit (IMU) and the differential global positioning system (DGPS), and compared with the planned optimal trajectory, and the trajectory tracking error indicators such as heading deviation, height deviation and speed deviation can be calculated. The trajectory deviation data is affected by both the dynamic characteristics of the UAV (such as the maximum tilt angle, the maximum acceleration, etc.) and the external interference such as the environmental wind field and air pressure. The flight envelope area with large deviation often corresponds to the state space where the UAV is difficult to accurately control or the environmental interference is large, and the envelope size needs to be appropriately reduced to ensure flight safety; while the area with small deviation indicates that the UAV motion is more controllable in this state space, and the envelope constraint can be appropriately relaxed to improve the motion flexibility.

[0068] Energy consumption data reflects the energy efficiency of trajectory planning. By collecting data such as battery voltage, current, and power during flight through the onboard power management system (BMS), the energy consumption distribution of the drone under different states can be analyzed. Energy consumption data is closely related to the aerodynamic characteristics (such as drag coefficient, lift coefficient, etc.) and flight attitude (such as angle of attack, sideslip angle, etc.) of the drone. The areas with higher energy consumption in the envelope usually correspond to high-speed, high angle of attack and other high-drag states of the drone, while the areas with lower energy consumption correspond to low-drag states such as cruising and gliding of the drone. When planning the trajectory, the drone should be kept away from the high-energy consumption area for a long time to extend the flight time. Therefore, the envelope size of the high-energy consumption area can be appropriately reduced to guide the drone to fly to the low-energy consumption area; at the same time, the envelope size of the low-energy consumption area can be appropriately expanded to provide the drone with more energy-saving path options.

[0069] By comprehensively considering the trajectory deviation and energy consumption distribution, the shape and size of the original feasible flight envelope can be adaptively adjusted. The adjusted envelope achieves a compromise between maneuverability and energy efficiency under the obstacle avoidance constraint, which not only ensures the obstacle avoidance safety of the drone, but also takes into account the trajectory tracking accuracy and energy utilization efficiency. The adjusted envelope vertex coordinates can be used to reconstruct the convex polyhedron to form an optimized flight envelope. The optimized envelope, as a priori knowledge, can be used to guide the trajectory search of the next control cycle, so that path planning can achieve closed-loop adaptive optimization in a dynamic environment.

[0070] It should be noted that envelope optimization is an incremental learning method that requires continuous accumulation of measured data and continuous updating of the envelope model during flight. In practical applications, an envelope sliding window can be designed to maintain trajectory deviation and energy consumption data within the last several control cycles for dynamic envelope adjustment. At the same time, a forgetting factor can also be introduced to give new data a higher weight, reduce the impact of outdated data, and improve the real-time performance of optimization.

[0071] Flight envelope optimization is an important manifestation of autonomous intelligence of drones, giving drones certain self-learning and self-adaptive capabilities. Through trajectory deviation feedback, drones can evaluate their own motion control capabilities, actively avoid difficult maneuvers, and improve control robustness; through energy consumption distribution feedback, drones can perceive their own energy consumption characteristics, actively choose energy-saving paths, and extend flight time. The optimized flight envelope comprehensively considers the constraints of multiple factors such as drones, environments, and tasks, and strikes a balance between obstacle avoidance, trajectory tracking, and energy consumption, thereby achieving adaptive evolution of the local optimal flight path. With the increase of flight time and the accumulation of empirical data, the optimized envelope will gradually converge to the global optimum, which will continuously improve the intelligence level of drones. At the same time, the optimization framework has a wide range of environmental adaptability, and can dynamically adjust the constraint threshold according to different terrains, meteorological conditions, etc. to meet the needs of multi-scenario applications.

[0072] In short, flight envelope optimization is a drone optimization control strategy based on actual flight big data analysis. It achieves dynamic adaptation of the optimization envelope through trajectory deviation and energy consumption feedback, balances control performance, energy efficiency and environmental adaptability under obstacle avoidance constraints, and continuously improves the intelligence level of drones, providing key technical support for autonomous, long-flight, multi-scenario drone inspections.

[0073] The method for adaptively adjusting the feasible flight envelope based on trajectory deviation data and energy consumption distribution includes: The coordinates of the vertices of the feasible flight envelope are: ,in Indicates the number of the envelope vertex, ; Adjusted vertex coordinates The calculation formula is: ; in: Adjusts the scale factor for the envelope, which determines the basic amplitude of the envelope scaling.

[0074] Vertex The normal direction of the envelope at .

[0075] A data item representing a heading deviation; A data item representing height deviation; A data item representing speed deviation.

[0076] Indicates the real-time value of battery voltage; Indicates the real-time value of current; Indicates the real-time value of power.

[0077] is the weight coefficient of the heading deviation, reflecting the influence of the heading deviation on the envelope adjustment; is the weight coefficient of altitude deviation, reflecting the influence of altitude deviation on envelope adjustment; It is the weight coefficient of speed deviation, reflecting the influence of speed deviation on envelope adjustment.

[0078] is the weight coefficient of the battery voltage, reflecting the influence of voltage on envelope adjustment; is the weight coefficient of current, reflecting the influence of current on envelope adjustment; is the power weight coefficient, reflecting the influence of power on envelope adjustment.

[0079] The relative influence of the setting deviation in the envelope adjustment is determined by technicians in the field through actual flight debugging; Adjusted by technicians in this field based on comparative analysis of different energy consumption items; trajectory deviation Indicates the deviation between the current actual trajectory of the drone and the optimal trajectory. The larger the deviation, the greater the control difficulty and the lower the safety. It indicates the energy consumption characteristics of the drone during flight. The high energy consumption area indicates a high consumption state that is difficult to maintain flight.

[0080] When the trajectory deviation is large or the energy consumption is small, the ratio term in the formula will increase, causing the vertices to adjust inward, shrink the envelope, and avoid high-risk areas.

[0081] When the trajectory deviation and energy consumption are small, the ratio term decreases, causing the apex to expand outward and the envelope to increase, thereby increasing the flight space and improving flexibility.

[0082] This formula combines trajectory deviation and energy consumption distribution to dynamically adjust the flight envelope vertex, allowing the drone to flexibly change the envelope size in complex environments. By reducing the envelope of high energy consumption and high deviation areas and expanding the envelope of low energy consumption and low deviation areas, it can effectively ensure flight safety and improve energy efficiency.

[0083] Step S2500: During the inspection process, steps S2100-S2400 are executed cyclically to achieve autonomous obstacle avoidance and trajectory optimization of the UAV in a complex environment.

[0084] Specifically, during the inspection process, the drone repeatedly performs obstacle avoidance and trajectory optimization at a certain frequency. The frequency can be dynamically adjusted according to the complexity of the environment to balance the computing efficiency and real-time control. In each cycle, the key control parameters of the drone are monitored in real time, including flight status, environmental perception results, and trajectory execution deviations to ensure flight safety and stability. Once an abnormal situation is detected, an early warning is initiated and measures are taken (such as return or emergency landing). Flight logs and obstacle avoidance data can also be recorded on a longer time scale to generate statistical reports for evaluating and optimizing the control performance of the drone. Based on the analysis results of the flight data, engineers can optimize the control parameters and algorithms to gradually improve the autonomous obstacle avoidance and trajectory optimization capabilities of the drone. The entire step forms a complete closed-loop control process that combines perception, decision-making, execution, and feedback to support the dynamic adaptation and continuous autonomous operation of drones in complex environments, ensuring safety and improving mission efficiency.

[0085] Example 2 This embodiment provides a drone inspection control system based on data analysis on the basis of embodiment 1. Figure 8 As shown, including: The flight envelope generation module is used to receive the environmental characteristic data of the target inspection area, and build a static environment model of the target inspection area based on the environmental characteristic data; obtain the dynamic parameters of the UAV, and calculate the feasible flight envelope of the UAV by combining the dynamic parameters of the UAV and the static environment model; Trajectory optimization module: After the drone takes off, it collects real-time environmental data of the flight area in real time, and builds a multi-level environmental topology model based on the real-time environmental data and the static environmental model; during the inspection process, it obtains the status data of the drone in real time, and dynamically determines the obstacle avoidance level of the drone based on the status data of the drone and the multi-level environmental topology model. The obstacle avoidance level is divided into three levels: safe, warning, and dangerous; if the obstacle avoidance level is warning or dangerous, local path planning is triggered to generate a local optimal flight trajectory; Flight envelope optimization module: used to obtain the actual trajectory deviation data and energy consumption data of the UAV when it moves along the local optimal flight trajectory; based on the actual trajectory deviation data and the energy consumption data, dynamically calibrate the feasible flight envelope to obtain the optimized flight envelope.

[0086] In the flight envelope generation module, the static environment model of the target inspection area is constructed based on the environmental feature data, including: Step S1110, receiving environmental characteristic data of a target inspection area, wherein the environmental characteristic data includes terrain complexity, building density, and risk area distribution; Step S1120, preprocessing the received environmental characteristic data, performing spatial division on the preprocessed environmental characteristic data, and dividing the target inspection area into a regular three-dimensional grid; Step S1130, extracting a feature vector of each grid and marking it as a grid feature vector, wherein the grid feature vector includes terrain height, building proportion and risk level; Step S1140: constructing a static environment model of the target inspection area based on the extracted grid feature vectors, wherein the static environment model is a three-dimensional grid map.

[0087] In the flight envelope generation module, the calculation of the feasible flight envelope of the UAV by combining the dynamic parameters of the UAV and the static environment model includes: Step S1210, obtaining the dynamic parameters of the UAV, wherein the dynamic parameters include maximum speed, maximum acceleration and maximum climb rate; Step S1220, combining the dynamic parameters of the UAV and the static environment model, calculating the initial flight envelope of the UAV, wherein the initial flight envelope is represented by a hexahedron; Step S1230, performing collision detection on the initial flight envelope based on the static environment model, clipping the envelope area where the collision occurs, and obtaining a final feasible flight envelope.

[0088] The step S1220 includes: Step S1221, based on the maximum speed constraint in the UAV dynamics parameters, calculate a control cycle The upper and lower bounds of the internal drone speed; Step S1222, based on the maximum acceleration constraint in the UAV dynamics parameters, calculate a control cycle The upper and lower bounds of the UAV’s acceleration are calculated; based on the upper and lower bounds of the velocity and the upper and lower bounds of the acceleration, the upper and lower bounds of the UAV’s displacement within a control cycle are calculated; Step S1223, mapping the upper and lower bounds of the displacement into a Cartesian coordinate system, obtaining the vertex coordinates of the flight envelope hexahedron, and generating the initial flight envelope of the UAV.

[0089] The step S1230 includes: Step S1231, record the vertex coordinate set of the initial flight envelope hexahedron as ,in, Indicates The coordinates of the vertices, 1≤i≤8; at the same time, the obstacle information is extracted from the static environment model, recorded as ,in, is a set of obstacles, Indicates obstacle information, m is the total number of obstacles, 1≤j≤m; Step S1232, for each vertex , traverse all obstacles, check the vertex according to the obstacle set O Whether it is inside an obstacle.

[0090] calculate Directed distance to each obstacle surface , if there is a distance , then it means Outside the jth obstacle; if there is a distance , then it means Inside the jth obstacle.

[0091] Step S1233, if If it is inside the obstacle, From the vertex coordinate set Remove it, and after traversal and detection, get the updated vertex set , where k is the number of vertices after the initial flight envelope is updated, ; Step S1234, using the updated vertex set Based on this, the flight envelope polyhedron is reconstructed to obtain the final feasible flight envelope.

[0092] In the trajectory optimization module, the construction of a multi-level environment topology model includes: Step S2110, after the drone takes off, real-time environmental data of the flight area is collected in real time, and the real-time environmental data includes laser radar point cloud data and visual image data; Step S2120, stitching the laser radar point cloud data into a local point cloud map in a unified coordinate system; extracting obstacle semantic information from the visual image data; Step S2130, fusing the local point cloud map with the obstacle semantic information to construct a semantic point cloud map; Step S2140, registering the semantic point cloud map with the static environment model, and superimposing the registered semantic point cloud map with the static environment model to form a multi-level environment topology model.

[0093] In the trajectory optimization module, the dynamic judgment of the obstacle avoidance level of the drone includes: Step S2210, obtaining status data of the drone at a preset frequency, wherein the status data includes position information, speed information and attitude information, and the attitude information includes an attitude angle; Step S2220, calculating the minimum distance between the drone and obstacles in the environment based on the drone's status data and the multi-level environment topology model , and calculate the relative speed between the drone and the nearest obstacle and relative acceleration ; Step S2230, combining the minimum distance , relative speed and relative acceleration , dynamically determine the current obstacle avoidance level of the drone, which is divided into three levels: safety, warning, and danger.

[0094] The dynamic judgment of the current obstacle avoidance level of the drone includes: Preset the first safety threshold , the second safety threshold , the first speed threshold , the second speed threshold , the first acceleration threshold and the second acceleration threshold ;in ; like and and , then the obstacle avoidance level is safe; ,or ,or , then the obstacle avoidance level is warning; like ,or ,or , the obstacle avoidance level is dangerous.

[0095] In the trajectory optimization module, generating the local optimal flight trajectory includes: Step S2310, obtaining the preset global trajectory of the drone, wherein the global trajectory is provided with n1 waypoints, taking the current position of the drone as the starting point and the waypoint closest to the current position on the global trajectory as the end point, and planning the local optimal flight trajectory; Step S2320: taking the feasible flight envelope as the envelope constraint of the UAV motion, taking the multi-level environment topology model as the flight environment constraint, and taking the maximum speed of the UAV as the and maximum acceleration As dynamic constraints, the overall optimization goal is established according to the total flight distance, total energy consumption and flight time, and the local trajectory optimization model is constructed; Step S2330, using a heuristic search algorithm to search for a feasible flight trajectory from the starting point to the end point within the feasible flight envelope; for the searched feasible flight trajectory, using a numerical optimization algorithm to solve the local trajectory optimization model to obtain the local optimal flight trajectory.

[0096] The flight envelope optimization module, wherein the dynamic calibration of the feasible flight envelope comprises: Step S2410, obtaining actual flight trajectory data when the UAV moves along the local optimal flight trajectory, comparing the actual flight trajectory data with the local optimal flight trajectory, and calculating trajectory deviation data, wherein the trajectory deviation data includes heading deviation, altitude deviation, and speed deviation; Step S2420, obtaining actual energy consumption data of the UAV when it moves along the local optimal flight trajectory, wherein the actual energy consumption data includes battery voltage, current and power; based on the actual energy consumption data, calculating energy consumption distribution in different areas of the feasible flight envelope; Step S2430, based on the trajectory deviation data and the energy consumption distribution, adaptively adjust the feasible flight envelope to obtain the adjusted flight envelope vertex coordinates; Step S2440, based on the adjusted flight envelope vertex coordinates, reconstruct the flight envelope polyhedron to obtain an optimized flight envelope.

[0097] Example 3

[0098] This embodiment discloses an electronic device, which may include one or more processors and one or more memories. The memories store computer-readable codes, which, when executed by the one or more processors, may execute the above-mentioned drone inspection control method based on data analysis.

[0099] The method or system according to the embodiment of the present application can also be implemented with the help of the architecture of an electronic device. The electronic device may include a bus, one or more CPUs, a read-only memory (ROM), a random access memory (RAM), a communication port connected to a network, an input / output component, a hard disk, etc. A storage device in the electronic device, such as a ROM or a hard disk, can store a drone inspection control method based on data analysis provided in the present application. A method for controlling an unmanned aerial vehicle inspection based on data analysis may, for example, include: receiving environmental characteristic data of a target inspection area, and constructing a static environmental model of the target inspection area based on the environmental characteristic data; obtaining the dynamic parameters of the unmanned aerial vehicle, and calculating the feasible flight envelope of the unmanned aerial vehicle by combining the dynamic parameters of the unmanned aerial vehicle and the static environmental model; after the unmanned aerial vehicle takes off, real-time environmental data of the flight area is collected in real time, and a multi-level environmental topological model is constructed based on the real-time environmental data and the static environmental model; during the inspection process, the status data of the unmanned aerial vehicle is obtained in real time, and the obstacle avoidance level of the unmanned aerial vehicle is dynamically determined based on the status data of the unmanned aerial vehicle and the multi-level environmental topological model, and the obstacle avoidance level is divided into three levels: safe, warning, and dangerous; if the obstacle avoidance level is warning or dangerous, local path planning is triggered to generate a local optimal flight trajectory; actual trajectory deviation data and energy consumption data of the unmanned aerial vehicle when it moves along the local optimal flight trajectory are obtained; based on the actual trajectory deviation data and the energy consumption data, the feasible flight envelope is dynamically calibrated to obtain an optimized flight envelope.

[0100] Furthermore, the electronic device may also include a user interface. Of course, the architecture disclosed in the present invention is only exemplary, and when implementing different devices, one or more components in the electronic device disclosed in the present invention may be omitted according to actual needs.

[0101] Example 4

[0102] This embodiment discloses a computer-readable storage medium, on which computer-readable instructions are stored. When the computer-readable instructions are executed by a processor, a drone inspection control method based on data analysis in an embodiment of the present application can be executed. The storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache). Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0103] In addition, according to an embodiment of the present application, the process described above with reference to the flowchart may be implemented as a computer software program. For example, the present application provides a non-temporary machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided in the present application, for example: receiving environmental feature data of a target inspection area, and constructing a static environment model of the target inspection area based on the environmental feature data; obtaining the dynamic parameters of the drone, and calculating the feasible flight envelope of the drone by combining the dynamic parameters of the drone and the static environment model; after the drone takes off, real-time environmental data of the flight area is collected in real time, and a multi-level environmental topology model is constructed based on the real-time environmental data and the static environment model; during the inspection process, the status data of the drone is obtained in real time, and the obstacle avoidance level of the drone is dynamically determined based on the status data of the drone and the multi-level environmental topology model, and the obstacle avoidance level is divided into three levels: safe, warning, and dangerous; if the obstacle avoidance level is warning or dangerous, local path planning is triggered to generate a local optimal flight trajectory; actual trajectory deviation data and energy consumption data of the drone when it moves along the local optimal flight trajectory are obtained; based on the actual trajectory deviation data and the energy consumption data, the feasible flight envelope is dynamically calibrated to obtain an optimized flight envelope. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are performed.

[0104] The methods, systems, and devices of the present application may be implemented in many ways. For example, the methods, systems, and devices of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present application may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers recording media storing programs for executing the method according to the present application.

[0105] In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0106] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A drone inspection control method based on data analysis, characterized in that: The method comprises: Receive environmental feature data of the target inspection area, and build a static environment model of the target inspection area based on the environmental feature data; obtain dynamic parameters of the UAV, and calculate a feasible flight envelope of the UAV by combining the dynamic parameters of the UAV and the static environment model; After the drone takes off, it collects real-time environmental data of the flight area in real time, and builds a multi-level environmental topology model based on the real-time environmental data and the static environmental model; during the inspection process, it obtains the status data of the drone in real time, and dynamically determines the obstacle avoidance level of the drone based on the status data of the drone and the multi-level environmental topology model. The obstacle avoidance level is divided into three levels: safe, warning, and dangerous; if the obstacle avoidance level is warning or dangerous, local path planning is triggered to generate a local optimal flight trajectory; The actual trajectory deviation data and energy consumption data of the UAV when it moves along the local optimal flight trajectory are obtained; based on the actual trajectory deviation data and the energy consumption data, the feasible flight envelope is dynamically calibrated to obtain an optimized flight envelope.

2. The method for controlling an unmanned aerial vehicle inspection based on data analysis according to claim 1, characterized in that: The static environment model of the target inspection area is constructed based on the environmental feature data, including: Preprocess the received environmental feature data, perform spatial division on the preprocessed environmental feature data, and divide the target inspection area into regular three-dimensional grids; extract the feature vector of each grid and mark it as a grid feature vector; Based on the extracted grid feature vectors, a static environment model of the target inspection area is constructed, and the static environment model is a three-dimensional grid map.

3. The method for controlling unmanned aerial vehicle inspection based on data analysis according to claim 1 is characterized in that: The dynamic parameters include maximum speed, maximum acceleration and maximum climb rate; The feasible flight envelope of the computational drone includes: Combining the dynamic parameters of the UAV and the static environment model, calculating the initial flight envelope of the UAV, wherein the initial flight envelope is represented by a hexahedron; The initial flight envelope is subjected to collision detection based on the static environment model, and the envelope area where the collision occurs is clipped to obtain the final feasible flight envelope.

4. The method for controlling unmanned aerial vehicle inspection based on data analysis according to claim 3 is characterized in that: The initial flight envelope of the computational drone includes: Calculate a control cycle based on the maximum speed constraint in the UAV dynamics parameters The upper and lower bounds of the internal drone speed; Calculate a control cycle based on the maximum acceleration constraint in the UAV dynamics parameters The upper and lower bounds of the UAV’s acceleration are calculated; based on the upper and lower bounds of the velocity and the upper and lower bounds of the acceleration, the upper and lower bounds of the UAV’s displacement within a control cycle are calculated; The upper and lower bounds of the displacement are mapped into the Cartesian coordinate system to obtain the vertex coordinates of the flight envelope hexahedron and generate the initial flight envelope of the UAV.

5. The method for controlling unmanned aerial vehicle inspection based on data analysis according to claim 3 is characterized in that: The envelope region where the collision occurs is clipped to obtain the final feasible flight envelope including: The vertex coordinate set of the initial flight envelope hexahedron is recorded as ,in, Indicates The coordinates of the vertices, 1≤i≤8; at the same time, the obstacle information is extracted from the static environment model, recorded as ,in, is a set of obstacles, Indicates obstacle information, m is the total number of obstacles, 1≤j≤m; For each vertex , traverse all obstacles, check the vertex according to the obstacle set O Whether it is located inside an obstacle; If the vertex If it is inside the obstacle, From the vertex coordinates set Remove it, and after traversal and detection, get the updated vertex set , where k is the number of vertices after the initial flight envelope is updated, ; With the updated vertex set Based on this, the flight envelope polyhedron is reconstructed to obtain the final feasible flight envelope.

6. The method for controlling unmanned aerial vehicle inspection based on data analysis according to claim 5 is characterized in that: The inspection vertex Whether it is inside the obstacle includes: calculation Directed distance to each obstacle surface , if there is a distance , then it means Outside the jth obstacle; if there is a distance , then it means Inside the jth obstacle.

7. The method for controlling unmanned aerial vehicle inspection based on data analysis according to claim 1, characterized in that: The real-time environmental data includes laser radar point cloud data and visual image data; The construction of a multi-level environment topology model based on real-time environment data and a static environment model includes: Splice the LiDAR point cloud data into a local point cloud map in a unified coordinate system; extract obstacle semantic information from visual image data; The local point cloud map is integrated with the obstacle semantic information to construct a semantic point cloud map; The semantic point cloud map is registered with the static environment model, and the registered semantic point cloud map is superimposed with the static environment model to form a multi-level environment topology model.

8. The method for controlling unmanned aerial vehicle inspection based on data analysis according to claim 1, characterized in that: The dynamic judgment of the obstacle avoidance level of the drone includes: Acquire the status data of the drone at a preset frequency, wherein the status data includes position information, speed information and attitude information, and the attitude information includes attitude angle; Calculate the minimum distance between the drone and obstacles in the environment based on the drone's status data and the multi-level environment topology model , and calculate the relative speed between the drone and the nearest obstacle and relative acceleration ; Combined minimum distance , relative speed and relative acceleration , dynamically determine the current obstacle avoidance level of the drone, which is divided into three levels: safe, warning, and dangerous; The dynamic judgment of the current obstacle avoidance level of the drone includes: Preset the first safety threshold , the second safety threshold , the first speed threshold , the second speed threshold , the first acceleration threshold and the second acceleration threshold ;in ; like and and , then the obstacle avoidance level is safe; ,or ,or , then the obstacle avoidance level is warning; like ,or ,or , the obstacle avoidance level is dangerous.

9. The method for controlling unmanned aerial vehicle inspection based on data analysis according to claim 1, characterized in that: Generating a local optimal flight trajectory comprises: Get the preset global trajectory of the drone. There are n1 waypoints in the global trajectory. Take the current position of the drone as the starting point and the waypoint closest to the current position on the global trajectory as the end point to plan the local optimal flight trajectory. The feasible flight envelope is used as the envelope constraint of the UAV motion, the multi-level environment topology model is used as the flight environment constraint, and the maximum speed of the UAV is and maximum acceleration As dynamic constraints, the overall optimization goal is established according to the total flight distance, total energy consumption and flight time, and the local trajectory optimization model is constructed; A heuristic search algorithm is used to search for feasible flight trajectories from the starting point to the end point within the feasible flight envelope. For the searched feasible flight trajectories, a numerical optimization algorithm is used to solve the local trajectory optimization model and obtain the local optimal flight trajectory.

10. The method for controlling unmanned aerial vehicle inspection based on data analysis according to claim 1, characterized in that: The dynamic calibration of the feasible flight envelope comprises: Acquire actual flight trajectory data when the UAV moves along the local optimal flight trajectory, compare the actual flight trajectory data with the local optimal flight trajectory, and calculate trajectory deviation data, wherein the trajectory deviation data includes heading deviation, altitude deviation, and speed deviation; Acquire actual energy consumption data of the UAV when it moves along the local optimal flight trajectory, wherein the actual energy consumption data includes battery voltage, current and power; and calculate energy consumption distribution in different areas of the feasible flight envelope based on the actual energy consumption data; Based on the trajectory deviation data and energy consumption distribution, the feasible flight envelope is adaptively adjusted to obtain the adjusted flight envelope vertex coordinates; based on the adjusted flight envelope vertex coordinates, the flight envelope polyhedron is reconstructed to obtain the optimized flight envelope.

11. A drone inspection control system based on data analysis, which is used to implement a drone inspection control method based on data analysis as claimed in any one of claims 1 to 10, characterized in that: The system comprises: The flight envelope generation module is used to receive the environmental characteristic data of the target inspection area, and build a static environment model of the target inspection area based on the environmental characteristic data; obtain the dynamic parameters of the UAV, and calculate the feasible flight envelope of the UAV by combining the dynamic parameters of the UAV and the static environment model; Trajectory optimization module: After the drone takes off, it collects real-time environmental data of the flight area in real time, and builds a multi-level environmental topology model based on the real-time environmental data and the static environmental model; during the inspection process, it obtains the status data of the drone in real time, and dynamically determines the obstacle avoidance level of the drone based on the status data of the drone and the multi-level environmental topology model. The obstacle avoidance level is divided into three levels: safe, warning, and dangerous; if the obstacle avoidance level is warning or dangerous, local path planning is triggered to generate a local optimal flight trajectory; Flight envelope optimization module: used to obtain the actual trajectory deviation data and energy consumption data of the UAV when it moves along the local optimal flight trajectory; based on the actual trajectory deviation data and the energy consumption data, dynamically calibrate the feasible flight envelope to obtain the optimized flight envelope.

Images