Method and system for adjusting dynamic landing point of UAV helipad based on environmental perception

Abstract

The present invention relates to the field of drone technology, and discloses a method and system for adjusting the dynamic landing point of a drone helipad based on environmental perception. The method comprises the following steps: acquiring environmental data and flight parameters, constructing a three-dimensional terrain grid model and analyzing stability; loading obstacle data and path constraints to generate a comprehensive evaluation model, and simulating control data; training the model to generate initial adjustment parameters, and combining the parameters with the model to obtain dynamic adjustment simulation information; constructing a multi-path optimization model to obtain optimal parameters; and combining real-time perception to generate safety probabilities and adjust landing strategies. The system includes modules such as data acquisition and terrain modeling. This solution achieves intelligent optimization and real-time adjustment of landing points through multi-dimensional environmental perception and dynamic modeling, thereby improving the safety and reliability of drone landing in complex environments, and is suitable for drone intelligent landing scenarios.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicles (UAVs), and in particular to a method and system for adjusting the dynamic landing point of an UAV helipad based on environmental perception. Background Art

[0002] With the widespread application of drone technology, their use in logistics and distribution, power inspections, emergency rescue, and other fields is becoming increasingly prevalent. Safe and stable drone landings are crucial for successful mission completion. As crucial infrastructure for drone landing, the rationality and adaptability of drone landing pads directly impact the safety and reliability of drone landings.

[0003] In traditional drone landing processes, landing point selection is typically based on a fixed, pre-set location, lacking the ability to perceive and dynamically adjust to environmental changes in real time. However, the environment in which drone landing pads are located is often complex and changeable, with factors such as undulating terrain, varying obstacle distribution, and changing weather conditions, all of which can adversely affect drone landings.

[0004] Specifically, the undulating terrain may cause the landing surface to be uneven, increasing the impact and shaking of the drone during landing, and even causing the drone to overturn; the appearance of dynamic obstacles, such as suddenly flying birds and temporary facilities, may hinder the landing path of the drone and cause collision accidents; different weather conditions, such as strong winds, heavy rain, heavy snow, etc., will affect the flight status and control performance of the drone, and thus affect the accuracy and safety of the landing.

[0005] Furthermore, the drone's flight parameters, such as speed, altitude, and attitude, also need to be comprehensively considered during the landing point adjustment process. Traditional fixed landing point strategies cannot dynamically adjust to these changing parameters, resulting in a low landing success rate in complex environments and posing a significant safety risk.

[0006] Although some existing methods for adjusting drone landing points can take into account the influence of environmental factors to a certain extent, most of them have the following shortcomings: First, the environmental perception capability is limited, and it is impossible to realize multi-dimensional and real-time environmental data collection and analysis of the apron area; second, there is a lack of dynamic assessment of terrain stability, and it is impossible to accurately judge the landing safety in different areas; third, the path planning and obstacle avoidance strategies are not flexible enough, and it is difficult to cope with the complex and changeable obstacle distribution and changes in the drone's flight status; fourth, the system's adaptability and intelligence are low, and it is impossible to realize dynamic optimization and real-time adjustment of the landing point.

[0007] Therefore, how to provide a landing point adjustment method and system that can perceive environmental changes in real time, dynamically evaluate terrain stability, flexibly plan landing paths, and make real-time adjustments based on UAV flight status parameters has become a technical problem that needs to be urgently solved in the current UAV field. Summary of the Invention

[0008] The purpose of the present invention is to provide a method and system for adjusting the dynamic landing point of an unmanned aerial vehicle apron based on environmental perception, so as to solve the problems raised in the above-mentioned background technology.

[0009] To achieve the above objectives, the present invention provides the following technical solutions: a method and system for adjusting the dynamic landing point of a drone helipad based on environmental perception, the method comprising:

[0010] Obtain environmental perception data of the drone landing area and drone flight status parameters;

[0011] Conduct multi-dimensional terrain scanning of the apron area and construct a three-dimensional terrain grid model. Combined with environmental perception data, perform dynamic terrain stability analysis on the three-dimensional terrain grid model to generate an initial terrain stability assessment model.

[0012] The initial terrain stability assessment model is loaded with dynamic obstacle distribution data and UAV path constraints to generate a comprehensive terrain stability assessment model. The landing point dynamic simulation is performed in combination with the flight state parameters to obtain the landing point control simulation data.

[0013] A dynamic obstacle prediction model is constructed and trained using landing point control simulation data to obtain a landing point adjustment model, thereby generating initial adjustment parameters. The initial adjustment parameters include at least the initial landing point position, the initial obstacle avoidance path, and the initial adjustment response parameters.

[0014] Based on the initial adjustment parameters, the comprehensive terrain stability assessment model and the flight state parameters, dynamic adjustment simulation information is obtained;

[0015] By combining dynamic adjustment simulation information with apron design parameters, a multi-path optimization model is constructed and the design parameters are iteratively optimized to obtain the optimal path planning parameters.

[0016] Based on the real-time environment perception model, the path planning parameters of the current period are inferred to generate the safe landing probability of the current period. The optimal path planning parameters of the current period are combined with the actual path parameters to adjust the real-time landing strategy of the UAV to achieve the dynamic landing point matching goal.

[0017] Preferably, the landing point control simulation data includes at least terrain undulation distribution, obstacle density distribution, safe landing area set and dynamic adjustment priority sequence;

[0018] The dynamic adjustment simulation information at least includes a landing point offset, a path obstacle avoidance success rate, and a stability evaluation index.

[0019] Preferably, the multi-dimensional terrain scanning of the apron area and the construction of a three-dimensional terrain grid model, the dynamic terrain stability analysis of the three-dimensional terrain grid model in combination with the environmental perception data, and the generation of an initial terrain stability assessment model include the following steps:

[0020] Conduct a laser radar and vision fusion scan of the apron area to obtain a three-dimensional terrain image sequence of the surface structure and obstacle distribution;

[0021] Preprocessing the stereo terrain image sequence to obtain standardized terrain grid data, wherein the preprocessing includes one or more of point cloud denoising, coordinate registration, feature extraction, grid subdivision, and data alignment;

[0022] Generate a 3D terrain grid model based on standardized terrain grid data and combined with 3D surface reconstruction technology;

[0023] A dynamic terrain stability analysis is performed on the three-dimensional terrain grid model to generate an initial terrain stability assessment model, wherein the dynamic terrain stability analysis includes at least grid weight allocation and stability parameter mapping, the terrain safety level is divided by grid weight allocation, and corresponding environmental perception data is assigned to each level area.

[0024] Preferably, the initial terrain stability assessment model is loaded with dynamic obstacle distribution data and UAV path constraints to generate a comprehensive terrain stability assessment model, and a landing point dynamic simulation is performed in combination with flight state parameters to obtain landing point control simulation data, including the following steps:

[0025] Loading dynamic obstacle distribution data into the initial terrain stability assessment model to simulate real-time obstacle changes, thereby generating a first terrain stability assessment model;

[0026] The UAV path constraints are added to the first terrain stability assessment model to generate a comprehensive terrain stability assessment model, where the UAV path constraints include the maximum pitch angle limit, the minimum turning radius, and the power system redundancy parameters;

[0027] Based on the comprehensive terrain stability assessment model and flight state parameters, dynamic simulation of the landing point is performed. The specific process includes:

[0028] Construct a multi-constraint path planning equation, which includes at least an obstacle avoidance distance equation, a trajectory smoothness equation, and a power consumption equation. Combined with a comprehensive terrain stability assessment model, the multi-constraint path planning equation is numerically solved using a piecewise linearization method to obtain terrain undulation distribution, obstacle density distribution, a set of safe landing areas, and a dynamically adjusted priority sequence.

[0029] Obtaining a safe landing area set includes the following steps:

[0030] Based on the obstacle density distribution, the obstacle coverage index of each grid cell is calculated;

[0031] Identify the areas where the obstacle coverage index in the comprehensive terrain stability assessment model is not greater than the preset threshold, and generate a set of safe landing areas, which is expressed as follows:

[0032] The safe landing area set is a spatial distribution mapping result of an obstacle coverage index and a preset threshold.

[0033] Preferably, the construction of the dynamic obstacle prediction model and training it with landing point control simulation data to obtain a landing point adjustment model, and then generating initial adjustment parameters, includes the following steps:

[0034] Construct a dynamic obstacle prediction model based on three-dimensional terrain grid features;

[0035] The dynamic obstacle prediction model is trained and verified using landing point control simulation data to obtain a landing point adjustment model.

[0036] Input the real-time flight status parameters into the landing point adjustment model to predict the set of safe landing areas;

[0037] Based on the predicted safe landing area set, initial adjustment parameters are generated, wherein the initial adjustment parameters at least include an initial landing point position, an initial obstacle avoidance path, and an initial adjustment response parameter.

[0038] Preferably, the method further includes reconstructing the landing point control simulation data, specifically:

[0039] Construct an initial multi-dimensional path input tensor based on terrain undulation distribution, obstacle density distribution, and safe landing area set;

[0040] Normalize and perform feature layering on the initial multi-dimensional path input tensor to generate the final multi-dimensional path input tensor;

[0041] Based on the dynamic adjustment priority sequence, a landing point adjustment label tensor is constructed;

[0042] The final multi-dimensional path input tensor and the landing point adjustment label tensor are combined to form a training sample set.

[0043] Preferably, generating the initial adjustment parameters based on the predicted safe landing area set includes the following steps:

[0044] Extract the grid cell with the lowest obstacle density in the predicted safe landing area set to generate the initial landing point location;

[0045] According to the spatial connectivity of the predicted safe landing area set, the feasible topology of the initial obstacle avoidance path is fitted to generate the initial adjustment response parameters;

[0046] The terrain gradient direction of the predicted safe landing area set is calculated and normalized into a path reference vector, which is the initial obstacle avoidance path direction.

[0047] Preferably, obtaining dynamic adjustment simulation information based on the initial adjustment parameters, the comprehensive terrain stability assessment model and the flight state parameters comprises the following steps:

[0048] The initial landing point position is mapped to the comprehensive terrain stability assessment model, the initial obstacle avoidance path is matched with the initial adjustment response parameters, and the path nodes in the adjustment area are encrypted. The comprehensive terrain stability assessment model is updated, and the dynamic adjustment trigger conditions, path update rules, and response parameter increments are defined to generate a dynamic adjustment simulation model.

[0049] Based on the dynamic adjustment simulation model, the multi-constraint path planning equation is iteratively solved by the piecewise linearization method to obtain dynamic adjustment simulation information, including:

[0050] When the dynamic adjustment trigger conditions are met, the landing point offset, path obstacle avoidance success rate, and stability evaluation index are updated, and the multi-constraint path planning equation is re-solved until the simulation termination conditions are met;

[0051] The dynamic adjustment trigger condition includes triggering a response parameter update when the current path obstacle avoidance success rate is no greater than a preset success rate threshold; the path update rule includes adjusting the initial obstacle avoidance path based on the terrain gradient direction; and the response parameter increment is in a piecewise linear relationship with the current stability evaluation index.

[0052] Preferably, the construction of the multi-path optimization model and iterative optimization of the design parameters to obtain the optimal path planning parameters includes the following steps:

[0053] A multi-path optimization model was constructed, where the optimization variables included path node density, turning angle threshold, and power distribution ratio. The optimization objectives included minimizing path length and maximizing adjustment response speed. The constraints included the UAV power limit and sensor accuracy threshold.

[0054] The multi-path optimization model is preliminarily solved using the A* algorithm to generate an initial optimized path population;

[0055] Based on the initial optimized path population, the Dijkstra algorithm is used for global optimization to generate the optimal path planning parameters.

[0056] Preferably, the present invention further includes a system for adjusting the dynamic landing point of a drone helipad based on environmental perception, the system comprising:

[0057] Data acquisition module, used to obtain environmental perception data of the drone landing area and drone flight status parameters;

[0058] The terrain modeling and analysis module is used to perform multi-dimensional terrain scanning of the apron area and construct a three-dimensional terrain grid model. It then combines environmental perception data to perform dynamic terrain stability analysis on the three-dimensional terrain grid model and generate an initial terrain stability assessment model.

[0059] The comprehensive assessment and simulation module is used to load dynamic obstacle distribution data and UAV path constraints into the initial terrain stability assessment model to generate a comprehensive terrain stability assessment model. It then combines flight state parameters to perform dynamic simulation of the landing point and obtain landing point control simulation data.

[0060] A parameter generation module is used to construct a dynamic obstacle prediction model and train it using landing point control simulation data to obtain a landing point adjustment model, thereby generating initial adjustment parameters. The initial adjustment parameters include at least the initial landing point position, the initial obstacle avoidance path, and the initial adjustment response parameters.

[0061] A dynamic adjustment simulation module is used to obtain dynamic adjustment simulation information based on initial adjustment parameters, a comprehensive terrain stability assessment model, and flight state parameters;

[0062] The path optimization module is used to combine dynamic adjustment simulation information with apron design parameters to build a multi-path optimization model and iteratively optimize the design parameters to obtain the optimal path planning parameters;

[0063] The strategy adjustment module is used to infer the path planning parameters of the current period based on the real-time environment perception model, generate the safe landing probability of the current period, and adjust the real-time landing strategy of the UAV by combining the optimal path planning parameters of the current period with the actual path parameters to achieve the dynamic landing point matching goal.

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] In terms of environmental perception and terrain modeling, a 3D terrain image sequence is acquired through LiDAR and visual fusion scanning. Preprocessing such as point cloud denoising and coordinate registration is then performed, and a high-precision 3D terrain mesh model is generated using 3D surface reconstruction technology. Furthermore, dynamic terrain stability analysis enables the classification of terrain safety levels and the assignment of environmental perception data. This accurately reflects the terrain characteristics and stability of the apron area, providing a solid foundation for landing point selection. This multi-dimensional, high-precision terrain modeling approach overcomes the incomplete and inaccurate acquisition of terrain information by traditional methods, enabling a more realistic simulation of actual terrain environments and improving the system's adaptability to complex terrain.

[0066] In terms of comprehensive evaluation and dynamic simulation, by loading dynamic obstacle distribution data and UAV path constraints, a comprehensive terrain stability assessment model is generated, and dynamic simulation of the landing point is performed in combination with flight state parameters. The constructed multi-constraint path planning equation covers multiple aspects such as obstacle avoidance distance, trajectory smoothness, and power consumption. Through numerical solution using piecewise linearization methods, detailed landing point control simulation data such as terrain undulation distribution, obstacle density distribution, and safe landing area set can be obtained. This enables the system to comprehensively evaluate the feasibility and safety of different landing points while considering multiple factors, predict possible problems in advance, and provide a scientific basis for adjusting the landing point. Compared with traditional methods, this comprehensive evaluation and dynamic simulation method is more comprehensive and accurate, and can effectively reduce landing risks caused by environmental factors and changes in the UAV's flight state.

[0067] In terms of parameter generation and dynamic adjustment, a dynamic obstacle prediction model is constructed and trained using landing point control simulation data to generate a landing point adjustment model. This model can predict the set of safe landing areas based on real-time flight status parameters and generate initial adjustment parameters. At the same time, by dynamically adjusting the simulation model, dynamic adjustment simulation information such as the landing point offset and path obstacle avoidance success rate are updated in real time when trigger conditions are met, achieving dynamic optimization of the landing point and obstacle avoidance path. This parameter generation and dynamic adjustment mechanism based on data-driven and model training makes the system highly adaptable and intelligent, enabling timely adjustments based on real-time changes in the environment and drone status, improving the flexibility and reliability of the landing process.

[0068] In terms of path optimization and policy adjustment, a multi-path optimization model was constructed with the goal of minimizing path length and maximizing adjustment response speed. Iterative optimization using the A* and Dijkstra algorithms yielded optimal path planning parameters. Simultaneously, a safe landing probability was generated based on a real-time environmental perception model. The real-time landing strategy was adjusted based on the optimal path planning parameters and the actual path parameters, achieving dynamic landing point matching. This multi-path optimization and policy adjustment approach optimizes the drone's landing path while ensuring landing safety, improving landing efficiency and accuracy, reducing power consumption, and extending the drone's flight time.

[0069] In addition, the entire system forms a complete closed loop from data collection, modeling analysis, simulation evaluation to parameter generation, path optimization, and strategy adjustment through the collaborative work of various modules, realizing the full process automation and intelligence of UAV landing point adjustment. This system can not only cope with complex and changeable environmental conditions and changes in UAV flight status, but also continuously improve its own performance through data training and model optimization, and has strong scalability and sustainability. In summary, the present invention effectively solves the shortcomings of traditional UAV landing point adjustment methods in terms of environmental adaptability, evaluation accuracy, adjustment flexibility and intelligence level, significantly improves the safety and reliability of UAV landing, and has broad application prospects and important technical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 This is a working principle diagram of the method for adjusting the dynamic landing point of a UAV helipad based on environmental perception according to the present invention;

[0071] Figure 2 Flowchart for the construction of a three-dimensional terrain grid model and an initial terrain stability assessment model;

[0072] Figure 3 Flowchart for dynamic obstacle prediction model training and initial parameter adjustment;

[0073] Figure 4 Flowchart for dynamically adjusting simulation model construction and simulation information generation. DETAILED DESCRIPTION

[0074] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0075] See also Figures 1-4The present invention relates to a method and system for adjusting the dynamic landing point of a drone helipad based on environmental perception, and the specific implementation steps are as follows:

[0076] By deploying laser radar, visual sensors and other equipment in the drone landing area, the environmental perception data of the area is collected in real time, including terrain features, obstacle distribution, meteorological parameters, etc.; at the same time, the flight status parameters of the drone are obtained, such as flight altitude, speed, attitude angle, power system status, etc.

[0077] A multi-dimensional terrain scan of the apron area was conducted, using LiDAR and vision fusion technology to acquire a 3D terrain image sequence. After pre-processing such as point cloud denoising and coordinate registration, a 3D terrain mesh model was constructed using 3D surface reconstruction technology. Combined with environmental perception data, dynamic terrain stability analysis was performed on the mesh model through mesh weight assignment and stability parameter mapping. The terrain safety level was then classified and an initial terrain stability assessment model was generated.

[0078] Dynamic obstacle distribution data and drone path constraints (such as maximum pitch angle limit and minimum turning radius) are loaded into the initial terrain stability assessment model to generate a comprehensive terrain stability assessment model. Based on this model and flight state parameters, multi-constrained path planning equations (including equations for obstacle avoidance distance, trajectory smoothness, and power consumption) are constructed and numerically solved using piecewise linearization methods. Dynamic landing point simulation is then performed to obtain landing point control simulation data, including terrain undulation distribution, obstacle density distribution, safe landing area set, and dynamically adjusted priority sequence.

[0079] A dynamic obstacle prediction model was constructed based on 3D terrain grid features. This model was trained and validated using landing point control simulation data to generate a landing point adjustment model. Real-time flight state parameters were input into the model to predict a set of safe landing areas and generate initial adjustment parameters, including the initial landing point location, initial obstacle avoidance path, and initial adjustment response parameters.

[0080] The initial adjustment parameters are mapped to a comprehensive terrain stability assessment model. A dynamic adjustment simulation model is generated by encrypting path nodes, defining dynamic adjustment trigger conditions (such as when the path obstacle avoidance success rate falls below a preset threshold), and defining path update rules (such as adjusting the path based on terrain gradient). By iteratively solving the multi-constrained path planning equations, dynamic adjustment simulation information such as landing point offset, path obstacle avoidance success rate, and stability assessment indicators is obtained.

[0081] Combining dynamic adjustment simulation information with apron design parameters, a multi-path optimization model is constructed with path node density, turning angle threshold, etc. as optimization variables, and the goals of minimizing path length and maximizing adjustment response speed. The optimal path planning parameters are obtained through iterative optimization using the A* algorithm and the Dijkstra algorithm.

[0082] Based on the real-time environment perception model, the path planning parameters of the current period are inferred to generate the probability of safe landing. By combining the optimal path planning parameters with the actual path parameters, the UAV landing strategy is dynamically adjusted to achieve landing point matching.

[0083] The present invention will be further described below in conjunction with Examples 1 to 5:

[0084] Example 1:

[0085] First, a hybrid perception system consisting of lidar and vision sensors is used to simultaneously scan the drone landing area. The lidar emits laser light in pulses, measuring the time difference between reflected signals to determine the target's distance. This generates high-density point cloud data that accurately depicts terrain undulations and obstacle outlines. The vision sensor collects texture and color information within the area. The two are then combined to create a 3D terrain image sequence that includes surface structure, obstacle locations, and morphological features, providing multi-dimensional raw data for subsequent modeling.

[0086] After acquiring a sequence of stereo terrain images, preprocessing is required to improve data quality. The preprocessing process begins with point cloud denoising, using a statistical filtering algorithm to remove outliers. This involves calculating the distance distribution between each point and its k-neighboring points, setting a distance threshold (e.g., the mean plus three standard deviations), and removing points exceeding the threshold as noise, ensuring that the point cloud data more accurately reflects the terrain. Next, coordinate registration is performed, using the Iterative Closest Point (ICP) algorithm to unify the multi-sensor coordinate systems. Identical feature points are selected from the lidar point cloud and visual imagery, and translation and rotation matrices are iteratively calculated to fully align the two sets of data in space, ensuring spatial consistency for subsequent modeling.

[0087] The feature extraction stage uses the Harris corner detection algorithm. By calculating the grayscale gradient matrix of local image regions, it extracts feature points with significant edges and corners. These points correspond to key locations such as protruding parts in the terrain and vertices of obstacle outlines. They are used in subsequent mesh subdivision and model construction to enhance the model's ability to depict terrain details. Mesh subdivision encrypts the original mesh based on the feature points and recursively splits the mesh using the Loop subdivision algorithm. New vertices are inserted at edge midpoints and vertices to generate a denser mesh structure. This improves the model's accuracy in depicting subtle terrain fluctuations (such as small hills and depressions) while maintaining a smooth and continuous mesh.

[0088] Data alignment uses translation, rotation, and scaling to unify pre-processed terrain data into a global coordinate system and rescale it to actual physical units (e.g., meters). This ensures seamless integration of scanned data from different time periods and regions, creating standardized terrain grid data. Each grid cell contains unique coordinates (x, y, z) as well as attribute parameters such as altitude, slope, and roughness, laying the foundation for 3D modeling.

[0089] Based on standardized data, a three-dimensional terrain grid model was constructed using the moving least squares (MLS) method. This method fits a quadratic polynomial surface within the neighborhood of each grid cell, smoothly connecting the height values of adjacent cells to generate a continuous terrain surface. During the fitting process, weights are automatically adjusted based on the density of the grid cell's feature points. This improves fitting accuracy in areas with complex terrain (such as those around obstacles) and simplifies calculations in flat areas, balancing model accuracy and computational efficiency. After surface fitting, triangulation is used to discretize the surface into triangular cells. The vertex coordinates and normal vector of each cell record the terrain geometry, facilitating computer processing and stability analysis.

[0090] Dynamic terrain stability analysis is performed by assigning weights to grid cells. Weighting factors include terrain slope, surface roughness, and historical settlement data. Terrain slope is calculated by the elevation difference and horizontal distance between adjacent grid cells. Areas with a slope ≤5° are assigned a high weight (0.8-1.0), indicating stability; slopes 5° < 15° are assigned a medium weight (0.5-0.8); and slopes >15° are assigned a low weight (0.1-0.5). Surface roughness is measured by the change in point cloud density, with smooth areas (such as concrete surfaces) receiving a higher weight and rough areas (such as gravel surfaces) receiving a lower weight. Historical settlement data are obtained from geological reports, and settlement risk areas are weighted down based on the magnitude of settlement (e.g., areas with annual settlement >10 mm are assigned a weight ≤0.3).

[0091] Terrain is divided into three levels based on weights: safe zone (weight 0.8-1.0), sub-safe zone (0.5-0.8), and dangerous zone (<0.5). Safe zones, characterized by flat and stable terrain, are preferred landing areas; sub-safe zones require real-time environmental assessments; and landings in dangerous zones are prohibited. Environmental sensing data is associated with each level. For example, when wind speeds are ≥ Level 5, the safe zone weight is reduced by 0.1-0.2; when the probability of precipitation is ≥ 60%, the sub-safe zone weight is reduced by 0.1, enabling dynamic updates of stability assessments.

[0092] During the model validation phase, actual measurement data is compared with the model output, and the root mean square error (RMSE) and mean absolute error (MAE) are calculated. If the RMSE exceeds 0.2 meters or the MAE exceeds 0.15 meters, the grid density (e.g., from 1 meter to 0.5 meters) or the surface fitting parameters (e.g., increasing the polynomial degree) are adjusted, and the model is rebuilt until the error meets the required standards. Through iterative optimization, the initial terrain stability assessment model accurately reflects the terrain conditions, providing a reliable basis for subsequent landing site selection.

[0093] Throughout the entire process, technical details such as synchronized triggering of LiDAR and vision sensors, data transmission rates (e.g., 10Gbps fiber optic transmission), and parallel computing of point cloud processing (using GPU acceleration) ensure real-time and accurate modeling. Standardized data formats (e.g., PLY format for point clouds and OBJ format for mesh models) ensure smooth data exchange between modules, providing fundamental support for dynamic landing point adjustment of the drone.

[0094] Example 2:

[0095] The specific implementation of the comprehensive assessment model and landing point dynamic simulation is as follows: First, dynamic obstacle distribution data is loaded onto the initial terrain stability assessment model. This data is collected in real time by millimeter-wave radar and vision sensors deployed around the UAV landing area. Millimeter-wave radar has strong penetration and is unaffected by lighting conditions, enabling it to accurately detect the distance, speed, and direction of obstacles even in adverse weather conditions such as rain and fog. Vision sensors provide detailed features such as the color, shape, and texture of obstacles. After time synchronization and spatial registration, the data from these two sensor types is fused to generate a dynamic data stream containing information such as obstacle location, size, and trajectory.

[0096] The dynamic obstacle distribution data is superimposed on the three-dimensional terrain grid of the initial terrain stability assessment model through a grid mapping algorithm. Specifically, the spatial coordinates of each obstacle are first mapped to the corresponding grid cell, and the area or volume ratio occupied by the obstacle within the grid cell is calculated based on the obstacle's geometric shape (such as a sphere, cube, etc.). For moving obstacles, their positions at multiple moments in the future are predicted according to their motion trajectory and mapped to the corresponding grid cells in sequence, thereby simulating the dynamic changes of real-time obstacles in the three-dimensional terrain grid model and generating the first terrain stability assessment model. This model clearly identifies the grid cells currently occupied by the obstacle and the areas that may be affected in the future, providing real-time obstacle constraint information for subsequent path planning.

[0097] Furthermore, the UAV path constraints are loaded into the first terrain stability assessment model to generate a comprehensive terrain stability assessment model. The UAV path constraints mainly include the maximum pitch angle limit, the minimum turning radius, and the power system redundancy parameter. The maximum pitch angle limit is determined by the UAV's aerodynamic design and power performance. For example, the maximum allowable pitch angle of a certain type of UAV is ±30°. Exceeding this angle may cause the UAV to stall or lose attitude control. The minimum turning radius is determined by the UAV's wheelbase and motor thrust. Assuming a UAV with a wheelbase of 1.5 meters, its minimum turning radius is set to 2 meters to ensure the UAV maintains flight stability during turns. The power system redundancy parameter requires that the UAV's remaining battery power must meet 1.5 times the energy consumption required to complete the current path planning to cope with the additional energy consumption caused by emergencies.

[0098] After converting the above path constraints into mathematical expressions, they are quantitatively evaluated using the attribute parameters of the grid cells. For example, for each grid cell, its pitch angle relative to the current position of the drone is calculated. If it exceeds the maximum pitch angle limit, the grid cell is marked as an impassable area. Based on the connection between the grid cells, the curvature radius of the path when turning is calculated. If it is less than the minimum turning radius, the location of the path inflection point is adjusted to meet the constraints. At the same time, the energy consumption of each path segment is calculated based on the drone's flight speed and path length, and compared with the remaining power to ensure that the power system redundancy parameters are met. Through the above processing, the comprehensive terrain stability assessment model organically combines terrain stability, obstacle distribution, and the physical limitations of the drone to form a complete path planning feasible domain constraint.

[0099] Based on a comprehensive terrain stability assessment model and the UAV's flight state parameters, dynamic simulation of the landing point was carried out. First, a multi-constraint path planning equation was constructed, which contains multiple sub-equations such as the obstacle avoidance distance equation, the trajectory smoothness equation, and the power consumption equation. The obstacle avoidance distance equation requires that the UAV maintain a horizontal distance of no less than 1.2 meters and a vertical distance of no less than 0.8 meters from obstacles during flight to avoid collision risk. By calculating the spatial distance between the UAV's flight path and the grid cell where the obstacle is located, it is determined whether the obstacle avoidance requirements are met. If not, the path is replanned.

[0100] The trajectory smoothness equation constrains the curvature of the path through the radius of curvature, preventing the drone from making sharp turns during flight. Specifically, for each inflection point on the path, the radius of curvature of its adjacent line segments is calculated. If the radius of curvature is less than a preset threshold (such as 5 meters), intermediate inflection points are added through interpolation to smooth the path curve and ensure the drone can fly smoothly. The power consumption equation calculates the energy consumption of each path segment based on the drone's flight resistance model. The flight resistance model takes into account factors such as air density, the drone's angle of attack, and flight speed. The drag coefficient is determined through empirical formulas or wind tunnel test data, and the energy consumption per unit distance is then calculated, providing a power optimization basis for path planning.

[0101] The multi-constrained path planning equations are numerically solved using a piecewise linearization method. First, the continuous flight path is discretized into a piecewise linear path consisting of multiple line segments, where each line segment corresponds to movement between grid cells. By setting an appropriate discretization step size (e.g., 0.5 meters), the computational complexity is controlled while ensuring accuracy. For each discretized path segment, the obstacle avoidance distance equation, trajectory smoothness equation, and power consumption equation are solved in sequence to determine whether all constraints are met. If not, the direction or length of the path segment is adjusted and the solution is repeated until all constraints are met.

[0102] Through the above numerical solution process, landing point control simulation data such as terrain undulation distribution, obstacle density distribution, safe landing area set, and dynamically adjusted priority sequence are obtained. The terrain undulation distribution is generated by the altitude data of each grid cell, and the terrain height changes are visually displayed in the form of contour maps or three-dimensional surface maps, helping to identify areas with complex terrain. The obstacle density distribution calculates the number of grid cells occupied by obstacles per unit area. The higher the density value, the denser the obstacles in the area, and the greater the landing risk.

[0103] The safe landing area set is generated as follows: First, the obstacle coverage index (OCI) is calculated for each grid cell. This index is defined as the ratio of the area occupied by obstacles to the total area of the grid cell. A preset threshold (e.g., 0.3) is set to identify grid cells with an OCI no greater than that threshold as safe landing areas. The spatial location information of these areas is then integrated to generate a safe landing area set. This set is stored as a list of grid cells, each containing parameters such as coordinates, altitude, and stability weight, providing a clear range of feasible areas for subsequent landing point selection.

[0104] The dynamically adjusted priority sequence is determined based on factors such as the grid cell's stability weight and obstacle density. Grid cells within the safe landing area are first sorted from high to low by stability weight. Grid cells with the same weight are then sorted from low to high by obstacle density, forming a dynamically adjusted priority sequence. Grid cells with higher priorities are prioritized as candidate landing sites, ensuring the drone lands in areas with high stability and few obstacles, minimizing landing risk.

[0105] During the dynamic landing point simulation process, the UAV's real-time flight parameters, such as altitude, speed, and attitude angle, must also be considered. These parameters are acquired in real time by sensors such as the UAV's inertial measurement unit (IMU) and global positioning system (GPS) and transmitted to the simulation system. Based on these parameters, the simulation system adjusts the initial and boundary conditions of the path planning equations in real time, ensuring that the simulation results accurately reflect the UAV's actual flight state and improving the reliability and practicality of the landing point control simulation data.

[0106] To verify the effectiveness of the comprehensive terrain stability assessment model and landing point dynamic simulation method, multiple simulation experiments and parameter adjustments were required. These experiments simulated various environmental conditions and obstacle distributions, observed the changing patterns of the safe landing area set and the dynamically adjusted priority sequence, and analyzed the efficiency and accuracy of solving the multi-constrained path planning equations. By adjusting parameters such as the discretization step size and constraint thresholds, the simulation model was optimized to ensure accurate and efficient generation of landing point control simulation data in a variety of complex scenarios, providing a reliable basis for subsequent landing point adjustments and path optimization.

[0107] Example 3:

[0108] The specific implementation of the landing point adjustment model and the generation of initial adjustment parameters is as follows: First, a dynamic obstacle prediction model is constructed based on the geometric features of the three-dimensional terrain grid model and time series data. The geometric features of the three-dimensional terrain grid model include spatial information such as the coordinates (x, y, z) of each grid cell, slope, roughness, and the historical location of obstacles. The time series data includes temporal information such as the obstacle's motion trajectory and speed changes over the previous N moments. Considering the temporal dependence of obstacle motion trajectories and the spatial correlation of terrain features, a long short-term memory network (LSTM) is used as the model architecture. This network, through a gating mechanism, effectively captures long-term dependencies in time series, making it suitable for handling dynamic obstacle prediction problems.

[0109] The input layer of the dynamic obstacle prediction model is designed as a multidimensional feature vector, containing the geometric features of the current grid cell (such as coordinates, slope, and roughness) as well as obstacle position coordinates, motion speed, and azimuth data from the previous N moments. For example, when N = 5, the input vector will contain the current grid features and obstacle status data from the past five moments, totaling (3+3+1)×(5+1)=42 dimensions (assuming the grid features are 3-dimensional coordinates and 1-dimensional slope, and the obstacle status is 3-dimensional position and 1-dimensional speed). The input layer transmits the feature vector to the hidden layer through a fully connected method. The hidden layer contains multiple LSTM units. Each unit processes the input sequence layer by layer through the coordinated action of forget gates, input gates, and output gates, extracting the underlying patterns of obstacle movement and the influence of terrain characteristics on it.

[0110] The model training process is based on the landing point control simulation data. First, the simulation data needs to be reconstructed to construct input and output samples suitable for model training. The specific steps are as follows: Based on the terrain undulation distribution, obstacle density distribution, and the set of safe landing areas, an initial multi-dimensional path input tensor is constructed. The dimensions of this tensor are designed to be [number of samples, time steps, number of features], where the number of samples corresponds to the number of simulation scenarios, the time step corresponds to the length of the time series of obstacle motion in each scenario, and the number of features includes terrain features (such as grid coordinates, slope, roughness) and obstacle features (such as position, speed, density). For example, for a sample set containing 1000 simulation scenarios, each of which records 20 time steps of data, the dimensions of the initial input tensor are [1000, 20, 10] (assuming 6-dimensional terrain features and 4-dimensional obstacle features).

[0111] The initial multi-dimensional path input tensor is normalized and feature-stratified. The Z-score standardization method is used to normalize the feature values of each dimension to a distribution with a mean of 0 and a standard deviation of 1, thereby eliminating the impact of different feature dimensions on model training. Feature stratification maps terrain features and obstacle features to different channels, for example, the first six channels store terrain features, and the last four channels store obstacle features. This allows the model to learn the relationship between spatial features and motion features separately. After processing, the final multi-dimensional path input tensor is generated. Its dimensions remain unchanged, but the feature values are now in a standardized range and have a clear structure.

[0112] A landing point adjustment label tensor is constructed based on a dynamically adjusted priority sequence. Each element of the label tensor corresponds to the expected output of a sample, including information such as the recommended landing point coordinates (x, y, z) and the obstacle avoidance path angle (θ, φ). The recommended landing point coordinates are selected from the center coordinates of the grid cell with the lowest obstacle density in the set of safe landing areas, and the obstacle avoidance path angle is determined based on the terrain gradient direction and obstacle motion trends. The label tensor has a dimension of [number of samples, number of output features]. For example, if the output features include 3D coordinates and 2D angles, the dimension is [1000, 5].

[0113] The final multi-dimensional path input tensor and the landing point adjustment label tensor are combined to form a training sample set. The dynamic obstacle prediction model is trained using supervised learning. The mean squared error (MSE) loss function is used during training, and the model weight parameters are optimized via backpropagation. To prevent overfitting, a dropout layer is added after the hidden layer to randomly drop a certain percentage (e.g., 20%) of neuron connections to enhance the model's generalization ability. The training process is divided into multiple epochs. In each epoch, the sample set is randomly shuffled and fed into the model in batches of 32 until the loss function converges to a preset threshold or the maximum number of training epochs (e.g., 200) is reached.

[0114] After model training is complete, it is validated using an independent validation set. During validation, simulated data not used in training is fed into the model. Metrics such as the overlap between the predicted safe landing area set and the actual safe area, as well as the error in the recommended landing point coordinates, are calculated to assess the model's prediction accuracy. If validation results do not meet requirements (e.g., the overlap rate is less than 70%), the model architecture is adjusted (e.g., increasing the number of LSTM layers or changing the number of hidden units) or training parameters are optimized (e.g., adjusting the learning rate or dropout ratio), and training is repeated until the model performance meets the requirements.

[0115] The trained and validated dynamic obstacle prediction model is the landing point adjustment model. It feeds the drone's real-time flight status parameters (such as current location, flight speed, remaining battery power, attitude angle, etc.) and the latest environmental perception data (such as real-time obstacle distribution and terrain stability weight) into the model. Through forward propagation, the model outputs a prediction of a set of safe landing areas for a future timeframe (e.g., the next 10 seconds). This prediction is presented as a list of grid cells. Each grid cell contains parameters such as the obstacle coverage index and stability weight correction at the time of prediction, reflecting the safety and suitability of the area in the future.

[0116] The process for generating initial adjustment parameters based on the predicted safe landing area set is as follows: First, the grid cell with the lowest obstacle coverage index in the set is extracted, the geometric center coordinates of this cell are calculated, and this coordinate is determined as the initial landing point location. The obstacle coverage index is calculated as the ratio of the predicted area occupied by obstacles in the grid cell to the total area of the cell. The lower the value, the fewer obstacles there are in the area, and the safer the landing. For example, in the predicted safe landing area set, the obstacle coverage index of a grid cell is 0.1, which is significantly below the average level. The center coordinates of this grid cell (x0, y0, z0) are used as the initial landing point location.

[0117] An initial obstacle avoidance path is generated based on the spatial connectivity of the predicted set of safe landing areas. Spatial connectivity analysis is implemented using the adjacency matrix method from graph theory. Each safe grid cell is considered a node in the graph. If two nodes are spatially adjacent (i.e., grid cells share an edge or face), an edge connection is established between the nodes. A minimum spanning tree algorithm (such as the Kruskal algorithm) is used to generate a tree structure containing all nodes in the connected graph. The edge set of this tree structure constitutes the feasible topology of the initial obstacle avoidance path. The initial obstacle avoidance path is composed of key edges in the tree structure. By extracting the endpoint coordinates of each edge, a path sequence containing a series of inflection points is formed. For example, from the current position, through node A, node B, and node C, to the initial landing point, the coordinates of each inflection point constitute the key parameters of the initial obstacle avoidance path.

[0118] At the same time, the terrain gradient direction of the predicted safe landing area set is calculated to determine the initial obstacle avoidance path direction. The terrain gradient is obtained by spatially differencing the altitude data of the grid cells. Specifically, for each grid cell, its altitude change rate in the x-axis and y-axis directions (∂z / ∂x, ∂z / ∂y) is calculated to form a gradient vector (gx, gy), the direction of which is the direction of the maximum terrain undulation. The gradient vector is normalized to a unit vector to obtain the path reference vector (gx_normalized, gy_normalized, 0), which is the initial obstacle avoidance path direction. For example, if the gradient vector is (2, -3), the normalized path reference vector is (2 / √13, -3 / √13, 0), indicating that the obstacle avoidance path should be planned preferentially along the downhill direction of the terrain to reduce flight resistance and energy consumption.

[0119] The generation of initial adjustment response parameters combines terrain stability assessment results with the UAV's power state. Terrain stability assessment results are obtained using a comprehensive terrain stability assessment model. Each safety grid cell is assigned a stability weight (e.g., 0.8, 0.7, etc.). Higher weights indicate more stable terrain and allow for larger path adjustment steps. UAV power state parameters include remaining battery power and motor thrust output. When the remaining battery power is sufficient, larger adjustment steps are permitted; otherwise, smaller steps are required to conserve energy. Initial adjustment response parameters specifically include the path adjustment step size (e.g., 0.5 meter per step) and the speed correction factor (e.g., adjusted between 0.9 and 1.1 based on the stability weight). These parameters are linearly linked to the stability weight and power state parameters. For example, path adjustment step size = base step size × stability weight × power factor. The power factor is determined by the remaining battery power percentage. When the remaining battery power is 80%, the power factor is 1.0, and decreases by 0.1 for every 10% decrease.

[0120] Throughout the implementation process, the real-time and accurate data transmission must be ensured. The drone transmits real-time flight status parameters to the ground control system via a wireless communication module. The sensor network within the ground control system simultaneously collects environmental perception data and integrates it with simulation data and model parameters. The data processing module utilizes a distributed computing architecture and GPU acceleration technology to improve the efficiency of model inference and parameter calculation, ensuring that initial adjustment parameters can be generated within milliseconds to meet the requirements of real-time drone control. Furthermore, a data verification mechanism is established to verify the validity of real-time data input to the model, eliminate outliers and erroneous data, and ensure the reliability of the model's output results.

[0121] Example 4:

[0122] In the process of generating dynamic adjustment simulation information and path optimization, the specific implementation method is as follows: First, the initial landing point position is mapped to the corresponding grid cell of the comprehensive terrain stability assessment model. The initial landing point position is determined by the previous step, such as the center coordinates of the grid cell with the lowest obstacle density selected from the safe landing area set. The position parameters are mapped to the coordinate system of the three-dimensional terrain grid model through a coordinate conversion algorithm to ensure that they correspond one-to-one with the grid cells in the model. Next, the initial obstacle avoidance path is matched with the initial adjustment response parameters. The initial obstacle avoidance path is generated by fitting the spatial connectivity of the safe landing area and contains a series of inflection point coordinates. These inflection point coordinates are matched with the grid cell path in the comprehensive model to check whether the path is completely within the safe landing area and meets the drone path constraints (such as maximum pitch angle, minimum turning radius, etc.).

[0123] Path nodes in the adjusted area are densified to improve simulation accuracy. Densification rules are determined based on terrain complexity and obstacle density. For example, in areas with high terrain volume or dense obstacles, the path node spacing is shortened from 1 meter to 0.5 meters. New nodes are inserted between the original nodes using linear interpolation to generate a denser path node sequence. After node densification, the path constraints in the comprehensive terrain stability assessment model are updated, including parameters such as pitch angle, turning radius, and energy consumption at each node, to ensure that the model more accurately reflects the flight status of the drone on the densified path.

[0124] Define dynamic adjustment trigger conditions, path update rules, and response parameter increments to generate a dynamic adjustment simulation model. The dynamic adjustment trigger condition is set to ensure that the current path's obstacle avoidance success rate is no greater than a preset success rate threshold. This threshold is determined based on the drone's safety level requirements, for example, 85%. If the path's obstacle avoidance success rate falls below this threshold during the simulation, the response parameter update mechanism is triggered. The path update rule is based on the direction of the terrain's undulating gradient. Specifically, the current obstacle avoidance path is adjusted to within ±15° of the descending terrain gradient to leverage the terrain's advantages and reduce flight drag and obstacle impacts. The response parameter increment exhibits a piecewise linear relationship with the current stability assessment index, which is output by the integrated terrain stability assessment model and ranges from 0 to 1. For example, when the index is ≥0.7, the response parameter increment is 0.1 times the current value; when 0.4 < the index < 0.7, the increment is 0.05 times the current value; and when the index is ≤0.4, the increment is 0. This approach achieves dynamic coupling between the response parameter and terrain stability.

[0125] Based on a dynamically adjusted simulation model, a piecewise linearization method is used to iteratively solve the multi-constrained path planning equations. These equations include the obstacle avoidance distance equation, trajectory smoothness equation, and power consumption equation. The piecewise linearization method discretizes the continuous flight path into multiple linear segments, each corresponding to a simulation step (e.g., 0.1 seconds). Within each simulation step, the obstacle avoidance success rate of the current path is first calculated. This rate is determined by counting the ratio of the spatial distances between the path segments and the obstacle grid cells that meet safety thresholds (e.g., horizontal distance ≥ 1.2 meters, vertical distance ≥ 0.8 meters). If the success rate falls below a preset threshold, the obstacle avoidance path direction is adjusted according to a path update rule. For example, a new path direction vector is obtained by taking a weighted average of the original path direction vector and the terrain gradient descent direction vector, and the path node coordinates are updated. Simultaneously, a response parameter increment is calculated based on the stability evaluation index, adjusting the UAV's flight speed, acceleration, and other parameters. For example, when the stability index is 0.6, the speed correction factor is 0.95, which is multiplied by the current flight speed to reduce the speed and improve flight stability.

[0126] Resolve the multi-constraint path planning equations to calculate whether the adjusted path segment satisfies all constraints. If not, continue adjusting the path direction and response parameters until the obstacle avoidance success rate exceeds the preset threshold or the maximum number of iterations (e.g., 5) is reached. After each iteration, update the landing point offset, path obstacle avoidance success rate, and stability evaluation index. The landing point offset is centered on the initial landing point, with a maximum translation distance of 5 meters within the safe landing area, represented by the coordinate offset (Δx, Δy, Δz). The stability evaluation index is calculated based on the average stability weight of the grid cells passed through by the adjusted path, reflecting the overall stability level of the path.

[0127] The simulation termination condition is set to reach the preset maximum simulation duration (e.g., 30 seconds) or when the landing point offset is less than 0.1 meter. At this point, the path adjustment is considered stable and the final dynamic adjustment simulation information is generated, including the landing point offset sequence, the path obstacle avoidance success rate change curve, and the stability evaluation index time series. This information is stored in a structured data format and can be used by subsequent path optimization modules.

[0128] In the path optimization phase, a multi-path optimization model is first constructed. The optimization variables include path node density, turning angle threshold, and power distribution ratio. Path node density is defined as the number of nodes per meter of path, ranging from 5 to 15. A higher node density results in a more refined path but also increases computational complexity. The turning angle threshold is the maximum allowable turning angle for the drone at a path turning point, ranging from 30° to 90°. A higher threshold increases path flexibility but also requires higher maneuverability. The power distribution ratio is the ratio of power output between horizontal and vertical flight, ranging from 0.3 to 0.7, which affects the drone's flight attitude and energy consumption.

[0129] Optimization objectives include minimizing path length and maximizing adjustment response speed. Minimizing path length uses the Euclidean distance formula to calculate the straight-line distance between the path's starting and ending points, combined with detour distances at path inflection points, to minimize the total path length. Maximizing adjustment response speed is measured by the number of path adjustments per unit time; a greater number of adjustments indicates a more sensitive system to environmental changes. Constraints include the drone's power limit (e.g., maximum motor output power) and sensor accuracy thresholds (e.g., GPS positioning accuracy of ±0.1 meter), ensuring that the optimized path is feasible within the drone's physical capabilities and sensor accuracy.

[0130] The A* algorithm is used to perform a preliminary solution to the multi-path optimization model and generate an initial population of optimized paths. The A* algorithm is a heuristic search algorithm that prioritizes lower-cost paths by combining the actual cost from the current node to the starting point with the estimated cost to the end point (a heuristic function). During implementation, the three-dimensional terrain grid model is converted into a cost grid. The cost of each grid cell is determined by factors such as terrain stability weight, obstacle density, and path node density. Grid cells with higher stability weights and lower obstacle density have lower costs. The A* algorithm is used to search the cost grid for the shortest path from the current location to the initial landing point, generating an initial population of optimized paths containing 10 candidate paths. Each path includes parameters such as node coordinates, turning angles, and power distribution ratios.

[0131] Based on the initial optimized path population, the Dijkstra algorithm is used for global optimization. The Dijkstra algorithm is a single-source shortest path algorithm suitable for finding the shortest path from a starting point to all other nodes in a weighted graph. Each path in the initial optimized path population is used as a starting point to construct a weighted graph containing dimensions such as path nodes, turning angles, and power distribution ratios. The weight values are calculated according to the optimization objective function. For example, the path length weight is 0.6, and the adjustment response speed weight is 0.4. The Dijkstra algorithm is used to search for the global optimal path in the weighted graph, comprehensively considering the path length and response speed to generate the optimal path planning parameters. For example, the optimal parameters obtained after optimization are a path node density of 10 / m, a turning angle threshold of 60°, and a power distribution ratio of 0.5. This set of parameters strikes a balance between path length and response speed, while also meeting the UAV's power limit and sensor accuracy constraints.

[0132] During the path optimization process, feasibility verification of the optimization results is required. The optimal path planning parameters are substituted into a comprehensive terrain stability assessment model to verify that the path completely avoids hazardous areas and satisfies all UAV path constraints. Simulated flights are then conducted by dynamically adjusting the simulation model to observe whether the UAV's obstacle avoidance success rate, stability indicators, and other parameters on the optimized path meet safety requirements. If a path is found to contain local infeasible areas (e.g., a pitch angle exceeding the maximum allowable value on a certain path segment), the algorithm returns to the A* algorithm stage, adjusts the heuristic function parameters or the cost grid weights, and regenerates candidate paths until the optimization result satisfies all constraints and safety indicators.

[0133] Throughout the implementation process, data processing and algorithm calculations were performed on a high-performance computing platform. Parallel computing technology was used to accelerate the solution of the A* and Dijkstra algorithms, ensuring that path optimization could be completed within the limited time before the drone landed. Furthermore, a parameter storage and backtracking mechanism was established to save the intermediate results and final parameters of each optimization. This allowed for a rapid rollback to the previous feasible solution in the event of sudden changes in environmental conditions, ensuring drone flight safety.

[0134] Example 5:

[0135] The specific implementation of adjusting the real-time landing strategy and system module coordination is as follows: The strategy adjustment module infers the path planning parameters for the current time period based on the real-time environmental perception model. This real-time environmental perception model is constructed by integrating the latest lidar point cloud data, visual image data, and meteorological sensor data (such as wind speed, temperature, and humidity). It can update parameters such as obstacle distribution and terrain stability weights in the three-dimensional terrain grid model in real time. For example, when the lidar detects a new static obstacle in the apron area, the model immediately maps the obstacle's location and size to the corresponding grid cell, updates the obstacle coverage index, and recalculates the set of safe landing areas.

[0136] The probability of a safe landing during the current period is generated using a Bayesian network model. The nodes of the Bayesian network include environmental variables (wind speed, visibility, obstacle density), terrain variables (stability weight, slope), and landing safety status (safe / dangerous). The edges between nodes represent the probabilistic dependencies between variables. The calculation expression for the safe landing probability is:

[0137]

[0138] in, Indicates that in a given environment variable and terrain variables The probability of a safe landing under is the probability distribution of environmental variables in a safe state, is the probability distribution of terrain variables in a safe state, is the prior probability of the safe state, and are the marginal probabilities of environmental variables and terrain variables, respectively. This formula integrates multi-source data through Bayesian theorem to quantify the combined impact of environment and terrain on landing safety.

[0139] The real-time landing strategy of the UAV is adjusted by combining the optimal path planning parameters and the actual path parameters of the current period. The optimal path planning parameters are output by the path optimization module, including the coordinates of the path nodes, the turning angle threshold, the power distribution ratio, etc. The actual path parameters are obtained in real time through the inertial navigation system (INS) and the global navigation satellite system (GNSS) carried by the UAV, including the actual position, speed, heading angle, etc. of the UAV. When the probability of a safe landing is lower than the preset threshold (such as 0.7), the adjustment instruction is calculated by the proportional-integral-differential (PID) controller. The input of the controller is the path deviation (the coordinate difference between the optimal path and the actual path), and the output is the landing point position correction. , Flight speed adjustment value and the steering angle of the obstacle avoidance path .

[0140] The coordinated operation of the system's various modules is scheduled through a distributed real-time operating system (RTOS). The data acquisition module, comprised of LiDAR, visual sensors, and IMUs, continuously collects environmental perception data and drone flight parameters at a 50Hz frequency, transmitting them to the terrain modeling and analysis module via an Ethernet bus. The module performs preprocessing on the received data, including point cloud denoising and coordinate registration. It updates the 3D terrain mesh model and initial terrain stability assessment model every second and sends the updated model data to the comprehensive evaluation and simulation module.

[0141] The comprehensive assessment and simulation module loads the dynamic obstacle distribution and UAV path constraints based on the latest model data to generate a comprehensive terrain stability assessment model. It then performs a dynamic landing point simulation every 0.5 seconds, outputting landing point control simulation data, such as terrain undulation distribution and the set of safe landing areas, and transmits this data to the parameter generation module. The parameter generation module uses this landing point control simulation data to train a dynamic obstacle prediction model, updating the landing point adjustment model every 2 seconds. It then generates initial adjustment parameters (such as initial landing point location and initial obstacle avoidance path) based on real-time flight state parameters and sends them to the dynamic adjustment simulation module.

[0142] The dynamic adjustment simulation module combines initial adjustment parameters with a comprehensive terrain stability assessment model, performing dynamic adjustment simulations in 0.1-second steps. This generates dynamic adjustment simulation information, such as landing point offset and path obstacle avoidance success rate, and feeds it back to the path optimization module every 0.2 seconds. Based on this dynamic adjustment simulation information and apron design parameters, the path optimization module iteratively optimizes using the A* and Dijkstra algorithms, outputting the optimal path planning parameters every 1 second and transmitting them to the strategy adjustment module. Based on the optimal path planning parameters, actual path parameters, and safe landing probability, the strategy adjustment module generates landing strategy adjustment commands in real time. These commands are sent to the drone via a wireless communication link (such as 5G) at a frequency of 10 Hz, ensuring that the drone can respond promptly to environmental changes.

[0143] During inter-module data transmission, TCP / IP is used to ensure data reliability. Sensitive information (such as drone location and path planning parameters) is encrypted and transmitted using data encryption algorithms (such as AES-128) to prevent data leakage or malicious tampering. A heartbeat detection mechanism is also established, with each module regularly sending status information to the main controller. If the main controller does not receive a module's heartbeat signal within a specified time, a fault alarm is triggered and operation is switched to the backup module, ensuring high system availability and fault tolerance.

[0144] The entire real-time adjustment process forms a closed-loop control loop: from environmental data collection to model updates, path planning, and strategy adjustments, each link is tightly linked and strictly synchronized, ensuring that the drone can continuously perceive its surroundings in complex and dynamic environments, dynamically optimizing its landing point and flight path, ultimately achieving a safe and precise landing. Through its modular design and standardized interfaces, the system supports flexible expansion of sensor devices and algorithm models to accommodate different drone models and diverse landing scenarios.

[0145] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0146] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for adjusting the dynamic landing point of a UAV helipad based on environmental perception, characterized in that: The following steps are involved: Obtain environmental perception data of the drone landing area and drone flight status parameters; Conduct multi-dimensional terrain scanning of the apron area and construct a three-dimensional terrain grid model. Combined with environmental perception data, perform dynamic terrain stability analysis on the three-dimensional terrain grid model to generate an initial terrain stability assessment model. The initial terrain stability assessment model is loaded with dynamic obstacle distribution data and UAV path constraints to generate a comprehensive terrain stability assessment model. The landing point dynamic simulation is performed in combination with the flight state parameters to obtain the landing point control simulation data. A dynamic obstacle prediction model is constructed and trained using landing point control simulation data to obtain a landing point adjustment model, thereby generating initial adjustment parameters. The initial adjustment parameters include at least the initial landing point position, the initial obstacle avoidance path, and the initial adjustment response parameters. Based on the initial adjustment parameters, the comprehensive terrain stability assessment model and the flight state parameters, dynamic adjustment simulation information is obtained; By combining dynamic adjustment simulation information with apron design parameters, a multi-path optimization model is constructed and the design parameters are iteratively optimized to obtain the optimal path planning parameters. Based on the real-time environment perception model, the path planning parameters of the current period are inferred to generate the safe landing probability of the current period. The optimal path planning parameters of the current period are combined with the actual path parameters to adjust the real-time landing strategy of the UAV to achieve the dynamic landing point matching goal.

2. The method for adjusting the dynamic landing point of a UAV helipad based on environmental perception according to claim 1 is characterized in that: The landing point control simulation data at least includes terrain undulation distribution, obstacle density distribution, safe landing area set and dynamic adjustment priority sequence; The dynamic adjustment simulation information at least includes a landing point offset, a path obstacle avoidance success rate, and a stability evaluation index.

3. The method for adjusting the dynamic landing point of a UAV helipad based on environmental perception according to claim 1 is characterized in that: The multi-dimensional terrain scanning of the apron area and the construction of a three-dimensional terrain grid model are performed, and the dynamic terrain stability analysis of the three-dimensional terrain grid model is performed in combination with the environmental perception data to generate an initial terrain stability assessment model, including the following steps: Conduct a laser radar and vision fusion scan of the apron area to obtain a three-dimensional terrain image sequence of the surface structure and obstacle distribution; Preprocessing the stereo terrain image sequence to obtain standardized terrain grid data, wherein the preprocessing includes one or more of point cloud denoising, coordinate registration, feature extraction, grid subdivision, and data alignment; Generate a 3D terrain grid model based on standardized terrain grid data and combined with 3D surface reconstruction technology; A dynamic terrain stability analysis is performed on the three-dimensional terrain grid model to generate an initial terrain stability assessment model, wherein the dynamic terrain stability analysis includes at least grid weight allocation and stability parameter mapping, the terrain safety level is divided by grid weight allocation, and corresponding environmental perception data is assigned to each level area.

4. The method for adjusting the dynamic landing point of a UAV helipad based on environmental perception according to claim 1, characterized in that: The initial terrain stability assessment model is loaded with dynamic obstacle distribution data and UAV path constraint conditions to generate a comprehensive terrain stability assessment model, and a landing point dynamic simulation is performed in combination with flight state parameters to obtain landing point control simulation data, including the following steps: Loading dynamic obstacle distribution data into the initial terrain stability assessment model to simulate real-time obstacle changes, thereby generating a first terrain stability assessment model; The UAV path constraints are added to the first terrain stability assessment model to generate a comprehensive terrain stability assessment model, where the UAV path constraints include the maximum pitch angle limit, the minimum turning radius, and the power system redundancy parameters; Based on the comprehensive terrain stability assessment model and flight state parameters, dynamic simulation of the landing point is performed. The specific process includes: Construct a multi-constraint path planning equation, which includes at least an obstacle avoidance distance equation, a trajectory smoothness equation, and a power consumption equation. Combined with a comprehensive terrain stability assessment model, the multi-constraint path planning equation is numerically solved using a piecewise linearization method to obtain terrain undulation distribution, obstacle density distribution, a set of safe landing areas, and a dynamically adjusted priority sequence. Obtaining a safe landing area set includes the following steps: Based on the obstacle density distribution, the obstacle coverage index of each grid cell is calculated; Identify the areas where the obstacle coverage index in the comprehensive terrain stability assessment model is not greater than the preset threshold, and generate a set of safe landing areas, which is expressed as follows: The safe landing area set is a spatial distribution mapping result of an obstacle coverage index and a preset threshold.

5. The method for adjusting the dynamic landing point of a UAV helipad based on environmental perception according to claim 1 is characterized in that: The method of constructing a dynamic obstacle prediction model and training it with landing point control simulation data to obtain a landing point adjustment model and then generate initial adjustment parameters includes the following steps: Construct a dynamic obstacle prediction model based on three-dimensional terrain grid features; The dynamic obstacle prediction model is trained and verified using landing point control simulation data to obtain a landing point adjustment model. Input the real-time flight status parameters into the landing point adjustment model to predict the set of safe landing areas; Based on the predicted safe landing area set, initial adjustment parameters are generated, wherein the initial adjustment parameters at least include an initial landing point position, an initial obstacle avoidance path, and an initial adjustment response parameter.

6. The method for adjusting the dynamic landing point of a UAV helipad based on environmental perception according to claim 5 is characterized in that: It also includes data reconstruction of landing point control simulation data, specifically: Construct an initial multi-dimensional path input tensor based on terrain undulation distribution, obstacle density distribution, and safe landing area set; Normalize and perform feature layering on the initial multi-dimensional path input tensor to generate the final multi-dimensional path input tensor; Based on the dynamic adjustment priority sequence, a landing point adjustment label tensor is constructed; The final multi-dimensional path input tensor and the landing point adjustment label tensor are combined to form a training sample set.

7. The method for adjusting the dynamic landing point of a UAV helipad based on environmental perception according to claim 6 is characterized in that: Generating initial adjustment parameters based on the predicted safe landing area set includes the following steps: Extract the grid cell with the lowest obstacle density in the predicted safe landing area set to generate the initial landing point location; According to the spatial connectivity of the predicted safe landing area set, the feasible topology of the initial obstacle avoidance path is fitted to generate the initial adjustment response parameters; The terrain gradient direction of the predicted safe landing area set is calculated and normalized into a path reference vector, which is the initial obstacle avoidance path direction.

8. The method for adjusting the dynamic landing point of a UAV helipad based on environmental perception according to claim 1 is characterized in that: The method of obtaining dynamic adjustment simulation information based on the initial adjustment parameters, the comprehensive terrain stability assessment model, and the flight state parameters includes the following steps: The initial landing point position is mapped to the comprehensive terrain stability assessment model, the initial obstacle avoidance path is matched with the initial adjustment response parameters, and the path nodes in the adjustment area are encrypted. The comprehensive terrain stability assessment model is updated, and the dynamic adjustment trigger conditions, path update rules, and response parameter increments are defined to generate a dynamic adjustment simulation model. Based on the dynamic adjustment simulation model, the multi-constraint path planning equation is iteratively solved by the piecewise linearization method to obtain dynamic adjustment simulation information, including: When the dynamic adjustment trigger conditions are met, the landing point offset, path obstacle avoidance success rate, and stability evaluation index are updated, and the multi-constraint path planning equation is re-solved until the simulation termination conditions are met; The dynamic adjustment trigger condition includes triggering a response parameter update when the current path obstacle avoidance success rate is no greater than a preset success rate threshold; the path update rule includes adjusting the initial obstacle avoidance path based on the terrain gradient direction; and the response parameter increment is in a piecewise linear relationship with the current stability evaluation index.

9. The method for adjusting the dynamic landing point of a UAV helipad based on environmental perception according to claim 1, characterized in that: The multi-path optimization model is constructed and the design parameters are iteratively optimized to obtain the optimal path planning parameters, including the following steps: A multi-path optimization model was constructed, where the optimization variables included path node density, turning angle threshold, and power distribution ratio. The optimization objectives included minimizing path length and maximizing adjustment response speed. The constraints included the UAV power limit and sensor accuracy threshold. The multi-path optimization model is preliminarily solved using the A* algorithm to generate an initial optimized path population; Based on the initial optimized path population, the Dijkstra algorithm is used for global optimization to generate the optimal path planning parameters.

10. The UAV landing pad dynamic landing point adjustment system based on environmental perception is characterized by: include: Data acquisition module, used to obtain environmental perception data of the drone landing area and drone flight status parameters; The terrain modeling and analysis module is used to perform multi-dimensional terrain scanning of the apron area and construct a three-dimensional terrain grid model. It then combines environmental perception data to perform dynamic terrain stability analysis on the three-dimensional terrain grid model and generate an initial terrain stability assessment model. The comprehensive assessment and simulation module is used to load dynamic obstacle distribution data and UAV path constraints into the initial terrain stability assessment model to generate a comprehensive terrain stability assessment model. It then combines flight state parameters to perform dynamic simulation of the landing point and obtain landing point control simulation data. A parameter generation module is used to construct a dynamic obstacle prediction model and train it using landing point control simulation data to obtain a landing point adjustment model, thereby generating initial adjustment parameters. The initial adjustment parameters include at least the initial landing point position, the initial obstacle avoidance path, and the initial adjustment response parameters. A dynamic adjustment simulation module is used to obtain dynamic adjustment simulation information based on initial adjustment parameters, a comprehensive terrain stability assessment model, and flight state parameters; The path optimization module is used to combine dynamic adjustment simulation information with apron design parameters to build a multi-path optimization model and iteratively optimize the design parameters to obtain the optimal path planning parameters; The strategy adjustment module is used to infer the path planning parameters of the current period based on the real-time environment perception model, generate the safe landing probability of the current period, and adjust the real-time landing strategy of the UAV by combining the optimal path planning parameters of the current period with the actual path parameters to achieve the dynamic landing point matching goal.

Images