A method, system, and electronic device for dynamic path planning of unmanned aerial vehicles (UAVs) based on terrain priority classification for curves.

Abstract

This invention relates to the field of autonomous flight control technology for unmanned aerial vehicles (UAVs), and provides a UAV dynamic path planning method, system, and electronic equipment based on cornering terrain priority classification. The method includes: acquiring environmental image data for target detection and terrain segmentation to identify cornering terrain areas; constructing an environmental risk map and generating a basic flight path; calculating the cornering angle and determining the cornering priority; determining the rotation angle and correcting the safety boundary based on the heading angle difference; calculating wind resistance in response to crosswinds and adjusting the cornering speed and angle in conjunction with the priority; generating a corrected flight path based on a multi-cost path planning cost function and performing dynamic replanning. This invention achieves adaptive correction of the cornering safety boundary and dynamic adjustment of cornering motion parameters, improving the safety and stability of UAV cornering flight in complex environments.

Description

Technical Field

[0001] This invention relates to the field of autonomous flight control technology for unmanned aerial vehicles (UAVs), specifically to a dynamic path planning method and system for UAVs based on terrain priority classification for curves, and electronic equipment. Background Technology

[0002] Currently, in the field of autonomous flight control for unmanned aerial vehicles (UAVs), path planning is typically based on static obstacle distribution or generating flight trajectories using the global shortest distance. However, in complex environments such as urban canyons and densely built-up areas, UAVs not only need to avoid static obstacles but also frequently perform cornering maneuvers. Existing technologies often simplify the UAV into a point mass model or use fixed safety buffer boundaries when planning cornering paths, failing to fully consider the space requirements for airframe rotation caused by changes in heading during cornering. This results in the actual physical space occupied exceeding the planned boundaries during sharp turns, increasing the risk of collisions. Furthermore, existing solutions lack a coupled consideration of crosswind disturbances and the UAV's power margin during cornering, failing to dynamically adjust cornering speed and turning radius based on real-time wind resistance. This makes them prone to attitude instability, trajectory deviation, or energy surges under crosswind conditions. Therefore, there is an urgent need for a dynamic path planning method for UAVs that can balance physical space safety during cornering with dynamic stability under complex weather conditions. Summary of the Invention

[0003] To address the problem that existing UAV cornering path planning does not fully consider the airframe's rotation space requirements and crosswind dynamic constraints, resulting in insufficient flight safety and stability, this application proposes a UAV dynamic path planning method, system, and electronic equipment based on cornering terrain priority classification to achieve adaptive correction of cornering safety boundaries and dynamic adjustment of cornering motion parameters.

[0004] To achieve the above objectives, the present invention adopts the following technical solution:

[0005] The system acquires environmental image data of the flight area, performs target detection and terrain segmentation on the environmental image data, and identifies wingable areas, obstacle areas, target areas, and cornering terrain areas. An environmental risk map is constructed based on the state parameters of the target areas, and a basic flight path is generated. The cornering angle of each cornering terrain area in the basic flight path is calculated, and cornering priority is determined accordingly. The autorotation angle is determined based on the difference in heading angle between the UAV entering and leaving the cornering terrain area, and the safety boundary of the cornering terrain area is corrected based on the autorotation angle. In response to crosswind detection, wind resistance is calculated based on the UAV's power parameters and crosswind parameters, and the cornering speed and cornering angle are adjusted in conjunction with the cornering priority and wind resistance. A corrected flight path is generated based on a path planning cost function that includes environmental risk, path distance, cornering priority, crosswind resistance, and energy consumption costs, and dynamic replanning is performed during flight in response to changes in environmental conditions.

[0006] The above scheme combines the geometric features of the curve terrain with the kinematic state of the UAV, realizing the transformation from static geometric obstacle avoidance to dynamic spatial safety planning, and ensuring the dynamic feasibility of curve maneuvering under crosswind interference.

[0007] Optionally, the step of correcting the safety boundary of the curve terrain region based on the rotation angle includes: when the rotation angle is greater than a preset rotation angle threshold, expanding the safety boundary of the curve terrain region in the terrain segmentation result.

[0008] Optionally, the formula for calculating the rotation angle is:

[0009] ;

[0010] in, This represents the drone's rotation angle corresponding to the i-th curve terrain region. This indicates the heading angle of the drone before entering the curved terrain area. This indicates the heading angle of the drone after it leaves the curved terrain area.

[0011] Optionally, the step of calculating wind resistance based on the UAV's power parameters and crosswind parameters, and adjusting the turning speed and turning angle in combination with the turning priority and the wind resistance, includes: the wind resistance is determined by subtracting the thrust required by the load and self-weight, flight drag, and crosswind force from the maximum thrust that the UAV can provide; when the wind resistance is less than a preset wind resistance threshold, the UAV's turning speed is reduced and the turning radius is increased.

[0012] Optionally, the formula for calculating the wind resistance is:

[0013] ;

[0014] in, Indicates wind resistance. This indicates the maximum thrust that the drone can provide. This indicates the thrust required to account for the load and its own weight. Indicates flight resistance, The method of calculating the cornering angle of each cornering terrain region in the basic flight path and determining the cornering priority accordingly includes: dividing the cornering terrain region into large-angle cornering region, medium-angle cornering region and small-angle cornering region according to the size of the cornering angle, and prioritizing the speed and angle adjustment of the large-angle cornering region during path correction.

[0015] Optionally, the steps of adjusting the cornering speed and cornering angle include: constructing a cornering energy consumption evaluation function, which integrates cornering power consumption, rotation angle change, crosswind force, and cornering time; and, under the condition of satisfying wind resistance constraints and safety boundary constraints, selecting the cornering speed and cornering angle with the smallest evaluation value of the cornering energy consumption evaluation function as the UAV cornering control parameters.

[0016] Optionally, the calculation formula for the path planning cost function is as follows:

[0017] ;

[0018] in, This represents the total cost of path planning. Indicates the path distance cost. Indicates the environmental risk cost, Indicates the priority cost of cornering angle. This indicates the cost of resisting crosswinds. This indicates the energy consumption evaluation value for cornering. , , , These are the corresponding weighting coefficients; the calculation formula for the bending energy consumption evaluation function is:

[0019] ;

[0020] in, This represents the power consumption of the drone when it turns at speed v. This represents the change in the rotation angle. Indicates the force of crosswind. Indicates the time taken to turn. , , , These are the corresponding weighting coefficients.

[0021] Optionally, the formula for calculating the risk value of the environmental risk map is:

[0022] ;

[0023] in, Indicates location point Environmental risk value at the location, Represents the target density term. Indicates the effect of target speed. The reciprocal term representing the distance between the target and the drone. , , These are the corresponding weighting coefficients.

[0024] Furthermore, this invention also provides a UAV dynamic path planning system based on cornering terrain priority classification, comprising: an image acquisition and terrain segmentation module for acquiring environmental image data of the flight area and performing terrain segmentation to identify wingable areas, obstacle areas, target areas, and cornering terrain areas; a basic path generation module for constructing an environmental risk map based on the state parameters of the target area and generating a basic flight path; a cornering priority classification module for calculating the cornering angle of each cornering terrain area in the basic flight path and determining the cornering priority accordingly; and a rotation angle correction module for adjusting the UAV's rotation angle based on the UAV's rotation angle. The difference in heading angle between entering and leaving the cornering terrain area determines the rotation angle, and the safety boundary of the cornering terrain area is corrected based on the rotation angle; the crosswind resistance calculation and cornering adjustment module is used to calculate the wind resistance capability based on the UAV's power parameters and crosswind parameters in response to crosswind detection, and adjust the cornering speed and cornering angle in combination with the cornering priority and the wind resistance capability; the path correction and dynamic replanning module is used to generate a corrected flight path based on a path planning cost function that includes environmental risk, path distance, cornering priority, crosswind resistance and energy consumption cost, and to perform dynamic replanning in response to changes in environmental conditions during flight.

[0025] In addition, the present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described above.

[0026] Beneficial effects:

[0027] This invention determines the rotation angle by calculating the difference in heading angle when the UAV enters and leaves the curve terrain area, and dynamically corrects the safety boundary accordingly. The principle is to restore the UAV's curve process from point mass motion to rigid body motion with attitude rotation characteristics, so that the safety boundary can adaptively expand as the heading change increases, thereby reserving a real physical anti-collision space for the body rotation, effectively solving the collision risk caused by insufficient space margin when the fixed boundary is turned at a large angle.

[0028] This invention calculates the wind resistance capability of a UAV under crosswind conditions in real time (i.e., the power margin after deducting various loads from the maximum thrust) and dynamically adjusts the cornering speed and turning radius based on cornering priority. The principle is to establish a quantitative mapping relationship between environmental disturbances and the dynamic performance of the UAV. When the power margin is insufficient to resist the current crosswind, the UAV actively reduces the cornering speed and increases the turning radius to reduce the lateral acceleration requirement. This avoids trajectory deviation or loss of control caused by forced cornering while ensuring attitude stability. At the same time, it combines energy consumption evaluation function to find the optimal solution within safety constraints, achieving a comprehensive balance between cornering safety, stability and energy efficiency under complex weather conditions. Attached Figure Description

[0029] Figure 1 This is a flowchart of the UAV dynamic path planning method based on terrain priority classification for curves, according to an embodiment of the present invention.

[0030] Figure 2 This is a flowchart of environmental image acquisition, target detection, and terrain segmentation according to an embodiment of the present invention;

[0031] Figure 3 This is a schematic diagram of environmental risk map construction and basic flight path generation according to an embodiment of the present invention;

[0032] Figure 4 This is a schematic diagram illustrating the identification of terrain areas around curves, calculation of curve angles, and priority classification in an embodiment of the present invention.

[0033] Figure 5 This is a comparative schematic diagram of the calculation of the UAV's autorotation angle and the correction of the cornering safety boundary according to an embodiment of the present invention;

[0034] Figure 6 This is a curve showing the calculation of the wind resistance capability of the UAV under crosswind and the adjustment of the cornering parameters according to an embodiment of the present invention;

[0035] Figure 7 This is a bar chart comparing the total energy consumption of different path planning methods in this invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0037] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0038] Example 1

[0039] like Figure 1 As shown, this embodiment provides a UAV dynamic path planning method based on terrain priority classification for cornering. This method is applied to the UAV's onboard computer or ground control station to achieve autonomous flight that balances spatial safety and dynamic stability in complex urban environments or wilderness terrain. It should be understood that the steps described in this embodiment are real-time action sequences executed by the UAV during flight missions, rather than offline data processing or model training processes. The method specifically includes the following steps:

[0040] Step S1: Acquire environmental image data of the flight area, perform target detection and terrain segmentation on the environmental image data, and identify wingable areas, obstacle areas, target areas, and cornering terrain areas. Specifically, the UAV uses its onboard visual sensors (such as a monocular camera, a binocular stereo camera, or an RGB-D camera) to collect real-time image streams of the foreground and surrounding environment. In this embodiment, "cornering terrain area" is a broad functional definition, not just referring to geographical road bends or canyon turns, but generally referring to any three-dimensional spatial area along the flight path where the UAV needs to change its flight direction, perform yaw maneuvers, or adjust its attitude, i.e., a "turning area." Through real-time processing of environmental image data, the system can discretize the continuous flight space into a set of regions with semantic attributes, where wingable areas are characterized as passable spaces without collision risk, obstacle areas are characterized as static or dynamic physical barriers, target areas are characterized as mission-related points of interest or threat sources, and cornering terrain areas are separately identified for subsequent specialized processing. This area recognition method enables path planning to no longer rely solely on a pre-set two-dimensional map, but to adaptively adjust based on real-time perceived three-dimensional environmental semantics, providing a data foundation for subsequent refined cornering control.

[0041] Step S2 involves constructing an environmental risk map based on the state parameters of the target area and generating a basic flight path. Specifically, after identifying various regions, the system does not directly generate the final trajectory. Instead, it first maps the dynamic attributes of the target area (such as location, speed, category confidence, etc.) into a spatialized risk field. The environmental risk map is essentially a multi-dimensional cost grid map, where the value of each grid cell reflects the overall threat level at that location. Based on this, a global path search algorithm (such as the A algorithm, Dijkstra's algorithm, or RRT algorithm) is used to search for a connected path from the starting point to the endpoint in the risk map, serving as the basic flight path. While this basic flight path meets basic obstacle avoidance and arrival requirements, it typically consists of a series of discrete waypoints or straight segments. In curved terrain areas, it often manifests as sharp broken lines or geometric curves that do not consider dynamic constraints, lacking the smoothness and safety required for direct execution. Therefore, subsequent steps are needed for targeted correction.

[0042] Step S3: Calculate the cornering angle of each cornering terrain region in the basic flight path and determine the cornering priority accordingly. Specifically, the system traverses the segments within the cornering terrain region of the basic flight path and quantifies the cornering angle by analyzing the angle between the direction vectors of adjacent segments. The cornering angle is a key geometric indicator for measuring the severity of turning; a larger angle means the UAV needs to complete a larger heading change in a shorter distance, placing higher demands on the response speed and stability of the attitude control system. Based on this, the system uses the cornering angle as the basis for priority ranking, for example, marking large-angle turning areas as high priority and small-angle turning areas as low priority. This priority classification mechanism ensures that, under conditions of limited computing resources or time constraints, the system can prioritize sharp turn nodes that have the greatest impact on flight safety, avoiding local risk loss of control caused by average force application.

[0043] Step S4: Determine the rotation angle based on the difference in heading angle between the UAV entering and leaving the curve terrain area, and correct the safety boundary of the curve terrain area based on the rotation angle. This is one of the core steps that distinguishes this embodiment from traditional point mass model path planning. Specifically, the UAV is not a point mass without volume, but a rigid body with a certain wingspan or fuselage size. When performing a turning maneuver, the body rotates around its own vertical axis, sweeping through additional physical space. If the safety boundary is planned only according to the geometric center trajectory, the UAV's arms or propellers are very likely to intrude into the obstacle range during large-angle turns. Therefore, this embodiment evaluates the space margin required for the body rotation by calculating in real time the difference between the heading angle before entering the curve area and the heading angle after leaving (i.e., the rotation angle). When the rotation angle is large, the system will dynamically expand the safety boundary of the curve terrain area outward based on the original terrain segmentation result (e.g., increase the buffer distance or expand the radius of the no-fly zone). This correction is performed directly at the perception level, which is equivalent to reserving real physical collision avoidance space for the drone's attitude adjustment, thus effectively solving the problem of insufficient space margin at fixed boundaries in sharp turning scenarios.

[0044] Step S5: In response to the detection of crosswind, the system calculates wind resistance capability based on the UAV's power parameters and crosswind parameters, and adjusts the cornering speed and cornering angle in conjunction with the cornering priority and wind resistance capability. Specifically, the cornering process is not only about following a geometric path, but also a process of dynamic balancing. Especially when crosswind interference is present, the UAV needs to provide additional lateral thrust to resist the wind force and maintain centripetal acceleration. The system monitors crosswind parameters (such as wind speed and direction) and the UAV's own power status (such as maximum thrust, current load, battery discharge capacity, etc.) in real time, and calculates the thrust margin available to resist lateral disturbances at the current moment, i.e., wind resistance capability. When the wind resistance capability is insufficient to support the currently planned cornering maneuver, the system will not forcibly execute the original plan, but will actively reduce the cornering speed to reduce the centripetal acceleration requirement based on the cornering priority, while increasing the cornering radius to smooth the rate of change of heading. This "speed reduction + radius increase" linkage adjustment strategy establishes a quantitative mapping between environmental disturbances and the dynamic performance of the UAV, ensuring the feasibility and attitude stability of cornering maneuvers under adverse weather conditions, and preventing sideslip or loss of control due to power saturation.

[0045] Step S6 involves generating a corrected flight path based on a path planning cost function that includes environmental risk, path distance, cornering priority, crosswind resistance, and energy consumption costs. The system then dynamically replans the flight path in response to changes in environmental conditions. Specifically, after completing the aforementioned safety boundary correction and motion parameter adjustment, the system transforms all constraints into a unified cost, constructing a comprehensive path planning cost function. This function not only considers traditional path length and environmental risk but also incorporates the difficulty (priority) of cornering, additional energy consumption under crosswinds, and the cost of attitude adjustment into the optimization objective. The final corrected flight path is generated by searching for the minimum value of this cost function within the feasible region that satisfies the safety boundary and wind resistance constraints. Furthermore, since the flight environment is dynamically changing (e.g., sudden gusts, intrusion of moving obstacles), the system continuously monitors the environmental state during flight. Once key parameters exceed preset thresholds or undergo significant changes, it immediately triggers local or global dynamic replanning, updating the corrected flight path in real time. This closed-loop dynamic planning mechanism ensures that the UAV remains in an optimal safety and energy efficiency state throughout the entire mission cycle.

[0046] Through the coordinated execution of steps S1 to S6, this embodiment achieves a leap from static geometric obstacle avoidance to dynamic spatial safety planning. By uniformly defining the curve terrain area as the turning area and introducing a rotation angle correction safety boundary, the technical problem of neglecting the rigid body rotation space requirement is solved. By establishing a linkage adjustment mechanism between wind resistance and curve parameters, the problem of missing dynamic constraints under complex weather conditions is solved. The entire method is based entirely on real-time acquisition and calculation of runtime data, without relying on specific offline trained models, and has good versatility and engineering applicability, providing reliable assurance for the safe flight of UAVs in complex scenarios such as urban canyons and bridge inspections.

[0047] Example 2

[0048] Building upon Example 1, this example further details the dynamic correction logic for the safety boundary. Specifically, when the rotation angle exceeds a preset rotation angle threshold, the safety boundary of the cornering terrain area in the terrain segmentation result is expanded. The core of this feature lies in addressing the spatial safety hazards arising from simplifying the UAV as a point mass or using a fixed safety margin in traditional path planning. It should be understood that as a rigid body with actual physical dimensions, the rotation of the UAV around its vertical axis during cornering maneuvers causes significant changes in the projection envelope of the arms, propellers, or loads on the horizontal plane. Especially in large-angle turning scenarios, the physical space swept by the body rotation often far exceeds the area occupied during straight flight. If only static terrain segmentation boundaries or fixed buffer distances are relied upon, insufficient attitude adjustment space at sharp turns can easily lead to the arms scraping obstacles or triggering obstacle avoidance emergency stops. Therefore, this example establishes a positive correlation mapping mechanism between the rotation angle and the safety boundary width, enabling the safety boundary to adaptively expand with increasing turning intensity, thereby reserving a realistic physical collision avoidance space for rigid body rotation.

[0049] Regarding the specific implementation of expanding the safety boundary, this embodiment does not involve re-acquiring environmental images or re-running the semantic segmentation network. Instead, it performs post-processing operations based on existing terrain segmentation results (such as binarized masks, occupied grid maps, or vector boundaries) to ensure the system's real-time response performance. For example, a morphological dilation algorithm can be used to dynamically calculate the radius of the dilation kernel based on the rotation angle, expanding the passable area boundary outwards at the pixel level within the curve terrain region; alternatively, in the vector map, a preset physical buffer distance can be directly added along the normal direction of the curve terrain region. The advantage of this approach is that it reuses the computational results of front-end perception, updating the safety domain only through low-overhead geometric operations. This avoids the waste of computational power caused by repeated perception and ensures strict alignment between boundary correction and the current environmental state. Furthermore, the expanded boundary not only includes the avoidance distance for static obstacles but also implicitly compensates for dynamic errors such as sideslip and overshoot that may occur during UAV turning, making the planned path physically more robust.

[0050] The preset rotation angle threshold is not an arbitrary empirical value, but an engineering parameter determined based on the characteristics of the UAV platform and safety redundancy. In this embodiment, the threshold is preferably set to 30°. The setting is based on two main aspects: First, from a geometric perspective, when the yaw angle of a multi-rotor UAV changes by more than 30°, the lateral displacement increment of the arm tip in the diagonal direction relative to the fuselage center begins to significantly exceed 10% of the fuselage radius. At this point, the safety margin of the fixed boundary is insufficient to cover the rotation envelope. Second, from a control perspective, 30° is usually the critical point between smooth steering and rapid response for the flight control system. When the angle is less than this, the attitude adjustment is relatively gentle and the position deviation is controllable. However, when the angle is greater than this, in order to overcome the gyro effect and wind resistance, the differential output of the motors increases, resulting in a decrease in position holding accuracy and requiring greater external space tolerance. Of course, in other embodiments, the threshold can also be dynamically configured according to the drone's wheelbase, maximum angular velocity, or the risk level of the mission scenario. For example, in a narrow indoor environment, the threshold can be lowered to 15° to improve safety, while in an open field inspection, it can be raised to 45° to improve passage efficiency. This invention does not limit it to a single value.

[0051] To more intuitively demonstrate the technical effects of this embodiment, please refer to... Figure 5 To understand. For example Figure 5 As shown in the diagram on the left, the fixed safety boundary is not considered when the rotation angle is not taken into account. It can be seen that at the turning point, the radius of the safety boundary (the cyan circle in the diagram) is constant and small. When the UAV performs a large-angle turn, its actual rotation envelope approaches or even exceeds this fixed boundary, posing a significant collision risk. In contrast, the diagram on the right shows the safety boundary corrected using the method of this embodiment. At the same turning point, the system detects that the rotation angle exceeds the threshold and automatically expands the range of the safety boundary in the terrain segmentation result. This corrected boundary (the expanded cyan area in the diagram) completely encompasses the maximum physical space occupied by the UAV at that turning angle. This comparison clearly demonstrates that by introducing rigid body kinematic constraints into the boundary definition of the perception layer, this invention effectively eliminates the gap between geometric planning and physical execution, fundamentally improving the intrinsic safety level of turning flight.

[0052] Example 3

[0053] Based on Example 2, this example further details the specific calculation method for the rotation angle.

[0054] The formula for calculating the rotation angle is:

[0055] ;

[0056] in, This represents the drone's rotation angle corresponding to the i-th curve terrain region. This indicates the heading angle of the drone before entering the curved terrain area. This indicates the heading angle of the drone after it leaves the curved terrain area.

[0057] It is important to note that this formula is not simply a mathematical operation rule, but a technical means of measuring physical quantities based on attitude data collected in real time by the UAV's onboard sensors. Specifically, the heading angles ψ_i and ψ_{i-1} in the formula are the yaw angle obtained by fusing the gyroscope and magnetometer in the UAV's inertial measurement unit (IMU), or the body orientation angle estimated in real time by visual odometry (VIO) in GNSS signal-constrained environments. These data reflect the true angle between the UAV's body coordinate system Z-axis and the geographic north or world coordinate system reference direction, representing the physical state quantities characterizing the UAV's rigid body attitude, rather than the virtually generated geometric parameters in the path planning algorithm. By binding the heading angle acquisition process to specific hardware sensing and state estimation modules, the technical attributes and engineering feasibility of the rotation angle calculation are ensured.

[0058] From a physical mechanism perspective, the rotation angle Δψ_i described in this embodiment represents the absolute amount of rotation of the UAV around its vertical axis during cornering maneuvers. This is fundamentally different from the change in geometric curvature or tangential direction of the basic flight path. Path curvature only describes the degree of curvature of the trajectory curve and cannot fully reflect the attitude adjustment requirements of the UAV. For example, in certain specific mission scenarios, the UAV may need to fly along a straight trajectory, but in order to keep the payload (such as the camera) always aligned with the side target, the body still needs to perform a large yaw rotation; conversely, when flying along a circular trajectory, if a coordinated turning control strategy is adopted, the body may only need a small relative attitude adjustment. Therefore, directly using the heading angle difference as a metric for the rotation angle can more accurately characterize the physical space occupied by the rigid body rotation of the UAV than simply relying on path geometric features, thus providing a decision-making basis that conforms to the rigid body kinematics characteristics for the dynamic correction of the safety boundary in Embodiment 2.

[0059] To more clearly illustrate the above calculation process and its technical effects, a specific cornering scenario is used as an example. Assume the drone is performing an urban patrol mission and identifies a right-angle turn in the terrain ahead. When the drone reaches the trigger point at the entrance of this area, the system reads the current IMU output yaw angle as 0° (due north), denoted as the forward heading angle ψ_{i-1}; simultaneously, based on the basic flight path, the target heading after leaving this area is 90° (due east), denoted as the departure heading angle ψ_i. Substituting these values ​​into the above formula, we can calculate that the rotation angle Δψ_i corresponding to this cornering terrain area is |90° - 0°| = 90°. Since 90° is significantly greater than the preset rotation angle threshold set in Example 2 (e.g., 30°), the system will immediately trigger a safety boundary correction mechanism, extending the passable boundary of the right-angle turn area outward by a corresponding buffer distance in the terrain segmentation result. This process is completed entirely in real-time the instant the drone reaches the entrance of the cornering area, ensuring strict synchronization between boundary correction and the current flight state. It should be understood that although this example uses a 90° right-angle turn, in practical applications, this calculation formula applies to sharp turns, obtuse turns, and continuous S-shaped curves, as long as there is a change in heading. Furthermore, for cases where the heading angle crosses the 0° / 36° boundary (e.g., from 350° to 10°), the system automatically performs modulo 360° processing when calculating the difference to obtain the minimum rotation angle, ensuring the physical correctness of the calculation results. Through this calculation method based on real-time attitude feedback, the present invention effectively avoids attitude space prediction deviations caused by path geometry simplification, further improving the safety of cornering flight. Example 4

[0060] Building upon Example 1, this example further details the calculation logic of wind resistance capability and its linkage adjustment mechanism with cornering parameters. Specifically, the wind resistance capability is determined by subtracting the thrust required for the payload and its own weight, flight drag, and crosswind force from the maximum thrust the UAV can provide. In this example, wind resistance capability is not an abstract performance indicator, but rather represents the remaining thrust margin of the UAV in its current flight state, after deducting the thrust required to maintain basic hovering, overcome aerodynamic drag, and balance the current environmental wind force. This real-time quantification of the physical quantity transforms the UAV's dynamic boundary from a static parameter table into a dynamic environmental adaptability metric. For example, even if the UAV's nominal maximum thrust is large, its actual usable wind resistance thrust margin may be close to the critical value under conditions of full load, high-speed flight, and strong crosswinds. Through this subtraction calculation, the system can accurately perceive the dynamic safety margin at the current moment, providing a real physical constraint basis for subsequent cornering decisions, rather than relying solely on empirical fixed speed limits.

[0061] Furthermore, when the wind resistance capability is less than a preset wind resistance capability threshold, the UAV's cornering speed is reduced and the cornering radius is increased. This coordinated adjustment strategy stems from a deep consideration of the cornering dynamics of multi-rotor UAVs. According to the physical laws of circular motion, the centripetal acceleration required for a UAV to corner is directly proportional to the square of its speed and inversely proportional to the turning radius. When insufficient wind resistance is detected (i.e., the thrust margin is lower than the minimum lateral force margin required to maintain attitude stability), simply reducing the cornering speed, while theoretically reducing the centripetal force requirement, may lead to excessively low airspeed, weakening the aerodynamic control surface effect or propeller control response, and increasing the risk of instability. Conversely, simply increasing the cornering radius, without reducing the speed, may not be sufficient to compensate for the thrust shortfall, still leading to sideslip or trajectory deviation. Therefore, this embodiment forcibly adopts a coordinated adjustment mode of "speed reduction" and "radius increase": on the one hand, reducing the speed directly reduces the demand for the square term of centripetal force; on the other hand, increasing the radius smooths the rate of change of heading to reduce instantaneous lateral overload. This dual adjustment ensures that the UAV remains within the dynamically feasible region with a limited thrust margin, effectively defending against control failures or evasion designs that might result from adjusting only a single parameter.

[0062] The physical meaning of setting the preset wind resistance threshold lies in the minimum lateral force margin required to maintain the drone's attitude stability during cornering. This threshold is not a fixed empirical constant, but rather an engineering parameter that can be configured based on the drone's inertia characteristics, motor response bandwidth, and mission safety level. For example, for drones with a high center of gravity or large moment of inertia, a higher wind resistance threshold should be set to reserve more buffer space due to their longer recovery time under lateral disturbances; while for highly maneuverable small drones, the threshold can be appropriately lowered to improve passage efficiency. In actual operation, this threshold can also be dynamically adjusted based on the remaining battery power. When the battery voltage drops, causing a decrease in maximum available thrust, the system automatically increases the proportion of the wind resistance threshold to prevent cornering accidents caused by a sudden drop in power.

[0063] To more intuitively demonstrate the performance of the above-mentioned linkage adjustment mechanism under different wind conditions, please refer to... Figure 6 To understand. For example Figure 6As shown in the graph, this curve reveals the nonlinear mapping relationship between crosswind speed and optimized cornering parameters. In the low crosswind speed range (e.g., 0 to 5 m / s), the UAV has ample wind resistance, and the system maintains a baseline cornering speed and minimum turning radius to ensure passage efficiency. When the crosswind speed enters the medium intensity range (e.g., 5 to 7 m / s), the wind resistance gradually approaches the threshold, and the curve shows that the turning radius begins to increase significantly in an approximately linear manner, while the cornering speed remains relatively stable or decreases slightly. This indicates that the system prioritizes absorbing lateral force demands by smoothing the geometric path. However, when the crosswind speed further increases to the high intensity range (e.g., above 7 m / s), the wind resistance has reached the safety threshold. At this point, the turning radius reaches its upper limit saturation, and the system then significantly reduces the cornering speed, sacrificing time for dynamic stability. This phased nonlinear adjustment strategy fully embodies the technical concept of this invention, which aims to balance safety with efficiency, avoiding the performance waste or unnecessary safety risks caused by the "one-size-fits-all" speed limit in traditional methods. It should be understood that... Figure 6 The specific values ​​shown are for illustrative purposes only. In practical applications, the inflection points and slopes of each parameter curve will be calibrated according to the aerodynamic parameters and power configuration of the specific model. As long as the core logic of "synchronously reducing speed and increasing radius when wind resistance is insufficient" is followed, it falls within the protection scope of this invention.

[0064] Example 5

[0065] Building upon Example 4, this example further details the specific calculation formula for wind resistance, the physical method for obtaining its parameters, and the grading strategy for cornering priority. The calculation formula for wind resistance is as follows:

[0066] ;

[0067] in, Indicates wind resistance. This indicates the maximum thrust that the drone can provide. This indicates the thrust required to account for the load and its own weight. Indicates flight resistance, This represents the crosswind force. The formula deconstructs the abstract concept of "wind resistance" into the difference of four directly measurable or estimable physical quantities, providing a clear engineering implementation path for assessing power margin. Specifically, F_max is not a constant nominal value, but rather the real-time maximum available thrust obtained by dynamically querying a pre-set motor-electromechanical controller (EMC) performance calibration table based on the current battery voltage, ambient temperature, and motor health status. For example, for a quadcopter drone with a 450mm wheelbase, F_max might be 200N when fully charged, but when the battery level drops to 20% and the voltage drops, the F_max obtained from the table might decrease to 160N. F_load can be obtained in various ways. In a preferred embodiment, it can be directly measured using a miniature weighing sensor integrated into the landing gear; in another embodiment, it can be indirectly estimated by monitoring the motor current during hovering and combining it with the remaining battery capacity (SOC), to accommodate lightweight platforms without dedicated sensors. F_air is typically measured directly by an airborne pitot tube, or, in the absence of a pitot tube, calculated in real time based on the UAV's current velocity vector, attitude angle, and a pre-identified aerodynamic drag coefficient model. F_wind can be directly sensed by an airborne three-dimensional anemometer, or, in the absence of dedicated wind measurement equipment, can be derived by fusing data from an inertial measurement unit (IMU) and a global navigation satellite system (GNSS) through a state observer to infer the external disturbance forces acting on the aircraft. Through the real-time fusion of the above multi-source heterogeneous data, each term in the formula has a solid physical origin, ensuring that the W_c calculation result can accurately reflect the dynamic boundaries of the UAV under the current operating conditions.

[0068] Regarding the determination of cornering priority, the step of calculating the cornering angle of each cornering terrain region in the basic flight path and determining the cornering priority accordingly includes: dividing the cornering terrain region into large-angle cornering regions, medium-angle cornering regions, and small-angle cornering regions based on the size of the cornering angle, and prioritizing speed and angle adjustments for the large-angle cornering regions during path correction. This angle-range-based hierarchical method avoids the potential limitation of protection range caused by using fixed hierarchical labels such as "Level 1, Level 2, Level 3," while clearly expressing the positive correlation between priority and the degree of steering aggression. In this embodiment, a large-angle cornering region preferably refers to a region with a cornering angle greater than 60°, a medium-angle cornering region refers to a region with a cornering angle between 30° and 60°, and a small-angle cornering region refers to a region with a cornering angle less than 30°. It should be understood that the aforementioned thresholds of 30° and 60° are merely exemplary values ​​adapted to urban inspection scenarios. In practical applications, these thresholds can be dynamically adjusted based on the drone's maneuverability, mission urgency, or environmental complexity. For example, in high-speed logistics delivery scenarios, the large-angle threshold can be raised to 75° to improve traffic efficiency, while in precision mapping scenarios, it can be lowered to 45° to ensure imaging stability. The engineering significance of prioritizing large-angle cornering areas lies in the fact that large-angle turns not only have the most significant impact on the aircraft's attitude stability but are also the main source of triggering safety boundary corrections and energy consumption peaks. By prioritizing the allocation of limited computing resources and control bandwidth to such high-risk nodes, the system can achieve Pareto optimization of overall flight performance while ensuring core safety. For example, when a drone passes through a 90° sharp turn and a 15° gentle turn consecutively, the system will first perform the speed reduction and radius increase adjustment and safety boundary expansion described in Example 4 for the 90° sharp turn. After the parameters in this area converge, the system will then process the 15° gentle turn, thereby avoiding safety hazards caused by response lag at sharp turns due to the average allocation of computing power. Through this hierarchical processing mechanism, the present invention effectively improves the real-time performance and robustness of path planning while ensuring safety when cornering.

[0069] Example 6

[0070] Based on step S7 of Example 1, this example further details the energy consumption evaluation mechanism and parameter optimization selection logic during cornering. Specifically, the steps of adjusting the cornering speed and cornering angle include: constructing a cornering energy consumption evaluation function, which integrates cornering power consumption, rotation angle change, crosswind force, and cornering time. The construction of this evaluation function is not a simple numerical weighting, but a multi-dimensional energy consumption model based on the cornering dynamics of the UAV. Among them, the cornering power consumption P(v) characterizes the mechanical work done by the motor propulsion system when the UAV maintains its cornering trajectory at a speed v. It is usually proportional to the cube or square of the speed, reflecting the coupling relationship between aerodynamic drag and propulsion efficiency. The change in rotation angle |Δψ| characterizes the energy consumed by the UAV to overcome rotational inertia and gyroscopic effects when rotating around the vertical axis. This part of the energy consumption is closely related to the turning rate and the mass distribution of the airframe, and is a unique energy consumption item that distinguishes it from straight flight. The energy consumption item corresponding to the crosswind force F_wind characterizes the additional induced drag and attitude correction work done by the UAV to maintain the predetermined trajectory under crosswind disturbance, reflecting the penalty mechanism of environmental disturbance on energy efficiency. The cornering time T is related to the basic static power consumption of airborne electronic equipment, sensors and flight control system, reflecting the trade-off between mission timeliness and total energy consumption. By incorporating the above four physical quantities into a unified evaluation system, the system can comprehensively quantify the comprehensive energy cost under different cornering strategies, avoiding the one-sidedness of only considering propulsion power consumption while ignoring the cost of attitude adjustment or environmental countermeasures.

[0071] More importantly, this embodiment establishes a strict safety-first optimization hierarchy. Specifically, under the constraints of wind resistance and safety boundaries, the turning speed and turning angle that minimize the evaluation value of the turning energy consumption evaluation function are selected as the UAV's turning control parameters. This means that the minimization search of the energy consumption evaluation function is strictly limited to the "dynamic-geometric dual feasible region" enclosed by the wind resistance threshold described in Embodiment 4 and the safety boundary described in Embodiment 2. In practice, the system first determines the upper limit of the allowable turning speed and the lower limit of the turning radius based on the current crosswind state and terrain segmentation results, eliminating all parameter combinations that may lead to thrust saturation or collision risks. Subsequently, it only traverses or performs a gradient search in the remaining safety parameter space to find the (v, Δψ) combination that minimizes the energy consumption evaluation value E. This "constraint-first, energy efficiency optimization" strategy fundamentally defends against dangerous schemes that may result from unconstrained minimization, such as choosing extremely low speeds to avoid high energy consumption leading to wind instability, or compressing the safety boundary to reduce turning energy consumption, leading to the risk of scraping. It should be understood that "minimum" here refers to the relative optimality under the premise of safety, rather than the global mathematical extreme value, ensuring that every parameter selection has physical feasibility and safety.

[0072] Regarding the determination of the weight coefficients k1 to k4 in the cornering energy consumption evaluation function, this embodiment adopts an engineering calibration method based on measured data, rather than relying on subjective experience. Specifically, multiple sets of cornering flight tests under different speeds, turning angles, and wind speeds can be performed in a standard test site, simultaneously recording battery discharge power, IMU attitude data, and airspeed indicator readings. Using multiple linear regression or nonlinear fitting algorithms, a mapping relationship between each physical component and the actual total energy consumption can be established, thereby reversing the process to find the weight coefficients that best reflect the energy consumption characteristics of the current aircraft model. For example, for heavy-load UAVs with large rotational inertia, the weight k2 of the |Δψ| term will be automatically calibrated to a higher value to accurately reflect the significant contribution of attitude adjustment to the total energy consumption; while for high-speed UAVs with good aerodynamic streamlines, the weight k1 of the P(v) term may dominate. Furthermore, these weighting coefficients can be dynamically configured according to the mission mode. When performing long-endurance inspection missions, the weight of the time-related term k4 can be appropriately increased to maximize endurance, while the weight of k4 can be decreased for rapid response missions to improve traffic efficiency. This data-driven calibration method enables the energy consumption evaluation function to adaptively match the characteristics of different platforms and mission requirements, improving the versatility and accuracy of the technical solution.

[0073] It is important to emphasize that the cornering energy consumption evaluation function described in this embodiment is an online decision-making tool for real-time parameter selection during flight, not a loss function or training objective for offline training of artificial intelligence models. In actual operation, whenever the UAV approaches a cornering terrain area, the onboard processor calls this function, combining real-time perceived crosswind parameters and its own state, to calculate the optimal cornering speed and angle command under the current conditions within milliseconds, and directly sends it to the underlying flight control for execution. This process does not involve backpropagation updates of model parameters, nor does it rely on iterative training of large-scale historical datasets. Instead, it is based on a defined physical model and real-time constraints for analytical solutions or lookup table optimization. This online application attribute ensures that the technical solution meets the requirements of patent law for solving technical problems, distinguishing it from purely mathematical operations or rules of intellectual activity. It also ensures that the system has the ability to respond instantly to sudden environmental changes and will not cause actual control failure due to generalization errors of offline models.

[0074] Example 7

[0075] Building upon Example 6, this example further details the specific composition of the total cost function for path planning and its synergistic relationship with the cornering energy consumption evaluation function. The calculation formula for the path planning cost function is as follows:

[0076] ;

[0077] in, This represents the total cost of path planning. Indicates the path distance cost. Indicates the environmental risk cost, Indicates the priority cost of cornering angle. This indicates the cost of resisting crosswinds. This indicates the energy consumption evaluation value for cornering. , , , These are the corresponding weight coefficients. This total cost function constitutes the core decision criterion for dynamic path planning in this invention. Essentially, it maps the evaluation indicators of five heterogeneous dimensions—geometric distance, environmental safety, maneuvering difficulty, weather disturbance, and energy consumption—to a unified scalar cost space for search and optimization. Specifically, C_d reflects the efficiency benchmark of the flight mission, usually obtained by summing the Euclidean or Manhattan distances of the path nodes; Risk(x,y) characterizes the comprehensive threat level of static obstacles and dynamic targets within the flight area, and its calculation method can refer to the description of the environmental risk map in the aforementioned embodiment; P_turn quantifies the geometric difficulty of navigating a cornering terrain area. The larger the cornering angle and the higher the priority, the greater the cost of this item, so as to guide the planner to avoid choosing overly sharp turning paths when unnecessary; W_cross is specifically introduced to address crosswind interference. When the UAV traverses an area with significant crosswinds, even if there are no obstacles in the area, it will incur additional crossing costs due to the difficulty in maintaining attitude, thus prompting the path to appropriately detour or adjust the entry angle when conditions permit; E is the cornering energy consumption evaluation value detailed in Embodiment 6, representing the energy economy of performing cornering maneuvers under the current dynamic constraints. Through the weighted summation of the above five items, the system can automatically weigh the optimal corrected flight path with comprehensive performance while ensuring safety and stability, rather than simply pursuing the shortest distance or the lowest energy consumption.

[0078] Regarding the cornering energy consumption evaluation function, this embodiment reiterates its calculation formula as follows:

[0079] ;

[0080] in, This represents the power consumption of the drone when it turns at speed v. This represents the change in the rotation angle. Indicates the force of crosswind. Indicates the time taken to turn. , , , These are the corresponding weighting coefficients. It is particularly important to emphasize that the symbols for each variable in this formula remain strictly consistent with those in the aforementioned embodiments: |Δψ| is the absolute value of the rotation angle calculated based on the difference in heading angle in Embodiment 3; F_wind is the crosswind force involved in the wind resistance calculation in Embodiments 4 and 5; and P(v) and T correspond to the propulsion power consumption and time cost described in Embodiment 6. This global consistency between the symbols and the physical definitions ensures that the total cost function Cost can seamlessly inherit the state estimation and constraint verification results completed in the preceding steps when calling the energy consumption sub-item E, avoiding decision conflicts caused by ambiguity of dependent variables. For example, when Embodiment 4 determines that the current crosswind force F_wind is too large, resulting in insufficient wind resistance, this increased F_wind value will not only trigger the speed reduction and radius increase adjustment action, but will also be simultaneously passed to the total cost function Cost through the energy consumption formula E. This allows paths containing strong crosswind sections to be given higher cost value in the global search, thereby achieving deep coupling between local control strategies and global planning objectives.

[0081] For the weighting coefficients λ, μ, η, and ρ in the total cost function, and k_1 to k_4 in the energy consumption formula, this invention does not limit them to fixed constants, but designs them as adjustable parameters that can be dynamically configured according to the task mode and environmental conditions. This design aims to prevent the technical solution from being misunderstood as a rigid mathematical model and to give the system the ability to adapt to diverse operational needs. Specifically, when performing emergency medical supply delivery tasks, the system can automatically load the "time-priority" configuration set, appropriately reduce the energy consumption weight ρ and the environmental risk weight λ (above the safety baseline), and at the same time increase the implicit weight of the distance cost C_d in exchange for faster arrival speed; while performing high-precision power line inspection or long-endurance monitoring tasks, it switches to the "steady-state energy saving" configuration set, significantly increasing the cornering priority cost weight μ and the crosswind resistance cost weight η, forcing the planner to generate a smoother, wind-avoiding, and lower-energy-consumption trajectory, even if this means an increase in path length. In addition, these weights can also be adaptively fine-tuned according to the real-time status of the UAV. For example, when the battery level is lower than a preset threshold, the system automatically linearly increases the ρ value to maximize the utilization efficiency of the remaining power. It should be understood that although this embodiment lists specific task modes as examples, in actual applications, the weight adjustment strategy can be continuous, segmented, or manually set by the operator. As long as it follows the basic principle of "multi-objective weighted optimization", it falls within the protection scope of this invention.

[0082] At the engineering implementation level, since the physical dimensions of C_d, Risk, P_turn, W_cross, and E are different (length, dimensionless risk value, angle, force, and energy, respectively), direct addition would cause the numerical magnitude differences to obscure certain important features. Therefore, this embodiment normalizes each component before calculating the total cost. Typical normalization methods include min-max normalization or Z-score standardization, which maps the values ​​of each component to the interval [0, 1] or [-1, 1] to make them comparable. Based on this, the typical value range of each weight coefficient can be set as follows: λ∈[0.1, 0.5], μ∈[0.2, 0.6], η∈[0.1, 0.4], ρ∈[0.1, 0.3]. For example, in a complex urban environment inspection scenario, a set of calibrated optimal weight configurations is λ=0.3, μ=0.4, η=0.2, and ρ=0.1. This configuration emphasizes the consideration of cornering difficulty and environmental risk, while downplaying the influence of pure distance and energy consumption. Real-world testing shows that the path generated under this configuration exhibits significantly better smoothness and safety at sharp turns than the traditional A* algorithm. Of course, the above numerical ranges are merely illustrative and not intended to limit the invention. Those skilled in the art can determine the most suitable weight combination and normalization parameters based on specific aircraft characteristics, sensor accuracy, and operational requirements through simulation testing or regression analysis of flight data. As long as the core logic is still based on the multi-dimensional cost fusion framework described in this invention, it should be considered within the scope of protection of this invention. Example 8

[0083] Building upon step S2 of Example 1, which constructs an environmental risk map, this example further details the method for quantifying environmental risk values ​​and its application logic in route planning. The formula for calculating the risk value of the environmental risk map is as follows:

[0084] ;

[0085] in, Indicates location point Environmental risk value at the location, Represents the target density term. Indicates the effect of target speed. The reciprocal term representing the distance between the target and the drone. , , These are the corresponding weighting coefficients. This formula deconstructs complex environmental threats into three physically calculable dimensions, aiming to provide a refined cost assessment basis for the generation of basic flight paths.

[0086] Specifically, Density(x,y) represents the ratio of the number of targets detected within a preset neighborhood window centered at location (x,y) to the area of ​​that window, i.e., the target density per unit area. This term reflects the static congestion level or target aggregation characteristics of the environment. For example, in pedestrian-dense areas such as squares and intersections, the Density value will increase significantly, indicating to the planner that there is a high potential collision probability or traffic efficiency bottleneck in that area. Velocity(x,y) represents the magnitude of the average velocity vector of all detected targets within the neighborhood, used to quantify the dynamic uncertainty of the environment. High-speed moving targets (such as moving vehicles) not only occupy space that changes rapidly over time, but their trajectory prediction errors are also much greater than those of stationary or low-speed targets. Therefore, the Velocity term can effectively penalize path options that cross high-speed traffic flows. Distance⁻¹(x,y) represents the reciprocal of the Euclidean distance from location (x,y) to the nearest detected target. The reciprocal form, rather than the nonlinear distance, is used because environmental threats have a significant near-field nonlinear amplification effect: when the drone is far from an obstacle, small changes in distance have a weak impact on safety; however, when the distance shortens to a critical range, the risk value should increase exponentially to force avoidance behavior. Through this nonlinear mapping, the system can maintain extremely high sensitivity to near-field threats while ensuring long-distance passage efficiency.

[0087] Regarding the configuration of weighting coefficients α, β, and γ, this invention does not limit them to globally fixed constants, but rather designs them as parameters that can be dynamically adjusted according to the task scenario and environmental semantics to reflect risk preferences under different operational modes. For example, in scenarios with dense pedestrian traffic but slow movement, such as urban pedestrian streets or parks, the system can automatically increase the density weight α and decrease the speed weight β, making path planning more inclined to avoid crowded areas, even if this means increased detour distance. In scenarios such as highway inspection or industrial areas with frequent vehicle traffic, the speed weight β should be significantly increased, forcing the UAV to prioritize airspace away from high-speed moving targets, even if there are slightly more static obstacles in that airspace, but they are relatively controllable. In addition, in tasks with extremely high safety margin requirements, such as precision equipment inspection or operations in confined spaces, the weight γ of the inverse distance term can be appropriately increased to ensure that the UAV always maintains a larger safety distance. This scenario-adaptive weight adjustment mechanism allows the same risk assessment model to flexibly adapt to diverse operational needs, avoiding the problem of a single fixed parameter being inadequate in different scenarios.

[0088] At the data structure level, the environmental risk map described in this embodiment is a rasterized two-dimensional field that is strictly aligned with the terrain segmentation results in Embodiment 1. Specifically, the resolution, coordinate system origin, and grid size of the risk map are consistent with the semantically occupied grid of the terrain segmentation output, for example, both using a grid precision of 0.1m × 0.1m. This data alignment design ensures that risk values ​​can be directly superimposed on the geometric boundaries of the flyable area without additional coordinate transformations or interpolation calculations, thus guaranteeing computational efficiency and avoiding local risk omissions or misjudgments due to resolution mismatch. The Risk value of each grid is updated in real time driven by the output results of the target detection and tracking module. When a new target is detected entering or the state of an existing target changes, the system only performs incremental recalculation on the affected local grid area, rather than refreshing the entire map, thereby meeting the real-time requirements of the UAV's onboard processor within millisecond cycles.

[0089] It is particularly important to emphasize that the environmental risk value Risk(x,y) calculated in this embodiment is mainly used for cost assessment in the basic flight path generation stage, that is, as a component of the heuristic cost or edge weight of the global path search algorithm (such as A* or Dijkstra) in steps S2 and S3 of Embodiment 1. Its role is to guide the path to avoid high-risk areas at the macro level, forming an initial, geometrically feasible connected trajectory. However, this risk value is not the sole basis or decisive factor for the final modified flight path generation. As described in Embodiments 6 and 7 above, the final modified flight path also needs to comprehensively consider multiple constraints such as cornering angle priority, crosswind resistance capability, energy consumption evaluation, and dynamic safety boundaries. In other words, the environmental risk map provides the "base color" or "background field" for path planning, while subsequent steps such as cornering correction and wind resistance adjustment are fine-tuning and "refining" on this base color. This hierarchical and progressive planning architecture ensures both the basic path's global awareness of environmental threats and avoids the problem of unfeasible or suboptimal paths under complex dynamic constraints due to over-reliance on a single risk indicator, thus achieving an organic balance between environmental safety and flight feasibility. Example 9

[0090] like Figure 1 As shown, this embodiment provides a UAV dynamic path planning system based on terrain priority classification for cornering. This system, serving as a virtual device carrier in the aforementioned method embodiment, achieves intelligent control of the entire UAV cornering flight process through a modular architecture. It should be understood that each module described in this embodiment is a software functional unit, which can be stored in memory as computer program instructions and executed by a processor, or embedded in application-specific integrated circuits such as FPGAs or ASICs, or deployed as a cloud service or edge computing node. This invention does not impose a unique limitation on the specific physical implementation of the modules. The system specifically includes the following functional modules:

[0091] The image acquisition and terrain segmentation module is used to acquire environmental image data of the flight area and perform terrain segmentation, identifying wingable areas, obstacle areas, target areas, and curve terrain areas. Specifically, this module is the system's perception front end. It receives real-time video streams or image frames from the UAV's onboard visual sensors (such as monocular, binocular, or RGB-D cameras) and transforms unstructured pixel data into structured environmental semantic information by running target detection and semantic segmentation algorithms. In terms of data flow, the segmentation result map or occupancy grid map containing various region identifiers output by this module directly serves as the input basis for downstream modules. For example, this module can output the identified curve terrain areas as a specific mask or vector boundary for subsequent modules to process. Through this standardized perception interface design, the system can be compatible with different types of visual sensors and perception algorithms, as long as their output conforms to the predefined semantic format, thereby improving the system's hardware adaptability.

[0092] The basic path generation module is used to construct an environmental risk map based on the state parameters of the target area and generate a basic flight path. Specifically, this module receives environmental semantic data from the image acquisition and terrain segmentation module 10 and calculates the risk value of each grid or node by combining the dynamic attributes of the target area (such as location, speed, category, etc.) to form an environmental risk map. Based on this, the module uses a global path search algorithm (such as A*, Dijkstra, or RRT*) to plan an initial connected path from the starting point to the ending point in the risk map. Although this basic flight path meets the basic obstacle avoidance and arrival requirements, it usually only considers geometric distance and static risk and has not yet been optimized for cornering dynamics. The output of this module is a basic flight path sequence, which includes the coordinate information of the path nodes and the connection relationship between adjacent nodes, providing an original trajectory reference for subsequent fine-tuning.

[0093] The cornering priority classification module calculates the cornering angles of each cornering terrain region in the basic flight path and determines the cornering priority accordingly. Specifically, this module traverses the path sequence output by the basic path generation module 20, locates the segments within the cornering terrain regions, and quantifies the cornering angle by calculating the angle between the direction vectors of adjacent segments. Based on the magnitude of the cornering angle, this module divides each cornering terrain region into different priority levels (e.g., large angle, medium angle, small angle) and attaches priority labels to the corresponding path node or segment metadata. This classification result provides a basis for differentiated processing by subsequent modules, ensuring that the system can prioritize the allocation of limited computing resources to sharp turn nodes that have the greatest impact on flight safety. For example, when the system's computing power is strained, complex wind resistance adjustments and boundary corrections can be performed only on high-priority areas, while simplified strategies can be used for low-priority areas, thereby achieving a balance between real-time performance and safety.

[0094] The rotation angle correction module determines the rotation angle based on the difference in heading angle between the UAV entering and leaving the curve terrain area, and corrects the safety boundary of the curve terrain area based on the rotation angle. Specifically, this module is the core unit for ensuring safety in the curve space. It reads the UAV's current attitude information (such as IMU yaw angle) or the expected heading angle in the path planning, and calculates the change in heading when entering and leaving the curve area, i.e., the rotation angle. When the rotation angle exceeds a preset threshold, the module dynamically expands the original safety boundary output by the image acquisition and terrain segmentation module 10, for example, by increasing the buffer distance through morphological dilation or normal offset. The corrected safety boundary data is passed to the path correction and dynamic replanning module 60 as a constraint condition in the generation of the final path. This mechanism ensures that the planned path not only avoids static obstacles, but also reserves sufficient physical space for the rigid body rotation of the UAV, effectively preventing scraping accidents caused by attitude adjustments.

[0095] The crosswind resistance calculation and cornering adjustment module is used to respond to the detection of crosswinds, calculate the wind resistance capability based on the UAV's power parameters and crosswind parameters, and adjust the cornering speed and cornering angle in conjunction with the cornering priority and the wind resistance capability. Specifically, this module monitors the ambient wind speed and direction and the UAV's own power status (such as maximum thrust, payload, battery charge, etc.) in real time, quantifying the current thrust margin, i.e., the wind resistance capability. When the wind resistance capability is insufficient to support the original cornering maneuver, this module adaptively reduces the cornering speed and increases the cornering radius based on the priority information provided by the cornering priority classification module 30, generating a set of recommended control parameters that meet dynamic constraints. These parameters are also passed to the path correction and dynamic replanning module 60. Through this linkage adjustment, the system tightly couples environmental disturbances with the airframe performance, ensuring the stability and feasibility of cornering flight under complex weather conditions and avoiding the risk of loss of control due to power saturation.

[0096] The path correction and dynamic replanning module generates a corrected flight path based on a path planning cost function that includes environmental risk, path distance, cornering priority, crosswind resistance, and energy consumption. It then dynamically replans the flight path in response to changes in environmental conditions during flight. Specifically, this module is the system's decision-making center. It integrates multi-source information from the aforementioned modules, including environmental semantics, basic path, priority labels, corrected safety boundaries, and recommended control parameters, to construct a multi-dimensional path planning cost function. Under the dual constraints of safety boundaries and wind resistance, this module generates the final corrected flight path through an optimized search algorithm and smooths it to meet flight control execution requirements. More importantly, this module possesses closed-loop feedback capability. During flight, it continuously monitors changes in environmental and its own states. Once it detects that key parameters exceed thresholds or undergo significant changes, it immediately triggers a local or global replanning process, updating the flight path in real time. This dynamic response mechanism ensures that the UAV remains in an optimal safety and energy efficiency state throughout the entire mission cycle.

[0097] Through the collaborative work of the aforementioned modules, the system provided in this embodiment fully replicates the technical solution of the aforementioned method embodiment, realizing a closed-loop control across the entire link, from environmental perception, risk assessment, cornering classification, boundary correction, wind resistance adjustment to path optimization. It should be understood that although each module in this embodiment is described as an independent functional unit, in actual engineering implementation, they can be flexibly deployed and combined according to computing power distribution and communication bandwidth. For example, in scenarios with extremely high real-time requirements, all modules can be integrated into the UAV's onboard computer to achieve autonomous decision-making; while in scenarios with good communication conditions and limited onboard computing power, the image acquisition and terrain segmentation module 10 can be deployed on the airborne end, while the basic path generation module 20 to the path correction and dynamic replanning module 60 can be deployed on a ground control station or cloud server, interacting via wireless links. This modular architecture design not only facilitates independent system upgrades and maintenance but also endows the technical solution with broad applicability and commercial implementation value, effectively supporting the rights protection needs of product claims.

[0098] Example 10

[0099] like Figure 1As shown, this embodiment provides an electronic device that serves as the hardware execution carrier for the aforementioned method embodiments, used to implement a UAV dynamic path planning function based on curve terrain priority classification at the physical level. Specifically, the electronic device includes a processor 100 and a memory 200, wherein the memory 200 stores a computer program, and the processor 100 executes the computer program to implement the method as described in any one of embodiments 1 to 8 above. It should be understood that the "computer program" here refers to a sequence of instructions that can be read and executed by the processor 100, which encodes the entire logical process from environmental image acquisition, terrain segmentation, risk map construction, curve priority classification, rotation angle correction, wind resistance calculation to path dynamic replanning. When the processor 100 calls and runs these instructions, the electronic device is transformed from a general-purpose computing device into a dedicated technical device with specific UAV path planning capabilities, thereby solidifying the technical solution of the present invention at the hardware level.

[0100] In terms of specific hardware architecture, the processor 100 and memory 200 are typically electrically connected via a high-speed bus or dedicated communication interface to enable rapid transmission of instructions and data. The processor 100 can be one or more combinations of a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), a field-programmable gate array (FPGA), or an application-specific integrated circuit (ASIC). For example, in UAV onboard scenarios with extremely high real-time requirements, the processor 100 preferably uses an embedded SoC chip with an integrated AI acceleration unit to process image segmentation and path search tasks in parallel; while in ground station or cloud scenarios with sufficient computing power, the processor 100 can be composed of a high-performance server cluster. The memory 200 includes volatile memory (such as RAM) and non-volatile memory (such as ROM, Flash, hard disk, etc.), where the non-volatile memory is used to persistently store the aforementioned computer program and preset parameter tables (such as motor thrust calibration tables, weight coefficient configuration sets, etc.), while the volatile memory temporarily stores environmental image data, intermediate calculation results, and real-time flight status during runtime. This hardware and software co-engineering architecture ensures that complex path planning algorithms can be solved within milliseconds, meeting the timeliness requirements of UAV dynamic obstacle avoidance and attitude control.

[0101] Regarding the deployment form of this electronic device, this invention does not limit it to a single physical entity, but rather encompasses all computing platforms capable of executing the aforementioned methods. In one typical embodiment, the electronic device is an onboard computer housed within the UAV's fuselage. It directly receives visual sensor data and outputs flight control commands, achieving fully autonomous edge-side path planning, suitable for scenarios with limited communication or high dynamic response. In another embodiment, the electronic device is a ground-based control station or remote command center. It receives image and status data transmitted back from the UAV via a wireless link, completes the planning, and uploads the corrected path back to the UAV, suitable for tasks requiring high computing power or manual supervision. In yet another embodiment, the electronic device can also be a cloud server, utilizing cloud computing resources to provide centralized path planning services and historical data learning for multiple UAVs. It should be understood that regardless of the specific form of the electronic device, as long as it contains a memory storing the program of this invention and a processor executing the program, it falls within the protection scope of this invention. This multi-form coverage strategy not only satisfies the patent law's requirement for the integrity of the technical solution but also provides a solid legal basis for subsequent rights protection against different types of hardware manufacturers or software service providers.

[0102] By combining the aforementioned hardware architecture with program instructions, the electronic device provided in this embodiment transforms the abstract logical steps in the aforementioned method embodiments into specific physical signal processing procedures. When the processor 100 executes the program, it essentially drives the underlying hardware circuitry to transform and calculate electrical signals representing physical quantities such as environmental risk, bending angle, and wind resistance, ultimately generating drive signals to control the drone's movement. This "structure + function" constraint effectively avoids the technical solution being deemed a purely intellectual activity rule or mathematical algorithm, establishing its legal attributes as a technological product. Simultaneously, since the electronic device's claims directly target the hardware product itself, there is no need to prove the defendant used a specific method or process in infringement determination; it is only necessary to confirm the product possesses the corresponding function through disassembly analysis or black-box testing to assert rights, greatly reducing the difficulty of proving rights and providing a comprehensive intellectual property protection barrier for the commercial implementation of this invention.

[0103] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. The specific implementation details of the UAV dynamic path planning method, system, and electronic equipment based on curve terrain priority classification described in Embodiments 1 to 10 above, including but not limited to the calculation method of the rotation angle, the correction strategy of the safety boundary, the quantitative model of wind resistance, the construction of the curve energy consumption evaluation function, and the weight configuration of the path planning cost function, are all intended to provide illustrative examples of the technical solutions of the present invention and not to impose restrictive limitations. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention, such as using an equivalent sensor fusion scheme to replace single visual perception, using different optimization algorithms to minimize the cost function, or adjusting the threshold parameters according to the specific characteristics of the aircraft model, should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A dynamic path planning method for unmanned aerial vehicles (UAVs) based on terrain priority classification for cornering, characterized in that, include: The system acquires environmental image data of the flight area, performs target detection and terrain segmentation on the environmental image data, and identifies the flyable area, obstacle area, target area, and curve terrain area. An environmental risk map is constructed based on the state parameters of the target area, and a basic flight path is generated. Calculate the cornering angle of each cornering terrain area in the basic flight path, and determine the cornering priority accordingly; The rotation angle is determined based on the difference in heading angle between the UAV entering and leaving the curve terrain area, and the safety boundary of the curve terrain area is corrected based on the rotation angle. In response to the detection of crosswind, the wind resistance capability is calculated based on the UAV's power parameters and crosswind parameters, and the turning speed and turning angle are adjusted in combination with the turning priority and the wind resistance capability. A modified flight path is generated based on a path planning cost function that includes environmental risk, path distance, cornering priority, crosswind resistance, and energy consumption costs, and dynamic replanning is performed in response to changes in environmental conditions during flight.

2. The method according to claim 1, characterized in that, The step of correcting the safety boundary of the curve terrain area based on the rotation angle includes: When the rotation angle is greater than a preset rotation angle threshold, the safety boundary of the curve terrain area in the terrain segmentation result is expanded.

3. The method according to claim 2, characterized in that, The formula for calculating the rotation angle is: ； in, This represents the drone's rotation angle corresponding to the i-th curve terrain region. This indicates the heading angle of the drone before entering the curved terrain area. This indicates the heading angle of the drone after it leaves the curved terrain area.

4. The method according to claim 1, characterized in that, The step of calculating wind resistance based on the UAV's power parameters and crosswind parameters, and adjusting the turning speed and turning angle in combination with the turning priority and the wind resistance, includes: the wind resistance is determined by the maximum thrust that the UAV can provide minus the thrust required by the load and its own weight, flight drag and crosswind force; When the wind resistance is less than a preset wind resistance threshold, the turning speed of the drone is reduced and the turning radius is increased.

5. The method according to claim 4, characterized in that, The formula for calculating the wind resistance is: ； in, Indicates wind resistance. This indicates the maximum thrust that the drone can provide. This indicates the thrust required to handle the load and its own weight. Indicates flight resistance, Indicates the force of crosswinds; The step of calculating the cornering angle of each cornering terrain region in the basic flight path and determining the cornering priority accordingly includes: dividing the cornering terrain region into large-angle cornering region, medium-angle cornering region and small-angle cornering region according to the size of the cornering angle, and prioritizing the speed and angle adjustment of the large-angle cornering region during path correction.

6. The method according to claim 1, characterized in that, The steps for adjusting cornering speed and cornering angle include: A cornering energy consumption evaluation function is constructed, which integrates cornering power consumption, rotation angle change, crosswind force, and cornering time. Under the conditions of satisfying wind resistance and safety boundary constraints, the turning speed and turning angle with the minimum evaluation value of the turning energy consumption evaluation function are selected as the turning control parameters of the UAV.

7. The method according to claim 6, characterized in that, The formula for calculating the path planning cost function is as follows: ； in, This represents the total cost of path planning. Indicates the path distance cost. Indicates the environmental risk cost, Indicates the priority cost of cornering angle. This indicates the cost of resisting crosswinds. This indicates the energy consumption evaluation value for cornering. , , , These are the corresponding weight coefficients; The calculation formula for the cornering energy consumption evaluation function is as follows: ； in, This represents the power consumption of the drone when it turns at speed v. This represents the change in the rotation angle. Indicates the force of crosswind. Indicates the time taken to turn. , , , These are the corresponding weighting coefficients.

8. The method according to claim 1, characterized in that, The formula for calculating the risk value of the environmental risk map is as follows: ； in, Indicates location point Environmental risk value at the location, Represents the target density term. Indicates the effect of target speed. The reciprocal term representing the distance between the target and the drone. , , These are the corresponding weighting coefficients.

9. A UAV dynamic path planning system based on terrain priority classification for curves, characterized in that, include: The image acquisition and terrain segmentation module is used to acquire environmental image data of the flight area and perform terrain segmentation to identify the flyable area, obstacle area, target area and curve terrain area; The basic path generation module is used to construct an environmental risk map based on the state parameters of the target area and generate a basic flight path; The cornering priority classification module is used to calculate the cornering angle of each cornering terrain area in the basic flight path and determine the cornering priority accordingly. The autorotation angle correction module is used to determine the autorotation angle based on the difference in heading angle between the UAV entering and leaving the curve terrain area, and to correct the safety boundary of the curve terrain area based on the autorotation angle. The crosswind resistance calculation and cornering adjustment module is used to calculate the wind resistance capability based on the UAV's power parameters and crosswind parameters in response to the detection of crosswind, and adjust the cornering speed and cornering angle in combination with the cornering priority and the wind resistance capability. The path correction and dynamic replanning module is used to generate a corrected flight path based on a path planning cost function that includes environmental risk, path distance, cornering priority, crosswind resistance, and energy consumption cost, and to dynamically replan the flight path in response to changes in environmental conditions.

10. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 8.

Images