An all-covering unmanned aerial vehicle surveying and mapping path planning method and system

Abstract

The application discloses a full-coverage unmanned aerial vehicle surveying and mapping path planning method and system, relates to the technical field of surveying and mapping path planning, and comprises the following steps: constructing a neural network energy model after carrying out inflation processing on original obstacles, rasterizing a surveying and mapping area, calculating total energy values of each grid center point, forming a three-dimensional energy field matrix, generating an initial skeleton path node sequence based on surveying and mapping start point coordinates, surveying and mapping end point coordinates and a preset segmentation number of the surveying and mapping area, adjusting the positions of internal nodes in the initial skeleton path node sequence, obtaining an optimized skeleton path node sequence, carrying out encryption processing, generating a dense flight point sequence, constructing a three-dimensional unmanned aerial vehicle motion model, constructing a feasible speed space, generating a candidate speed combination set, selecting optimal control instructions from the candidate speed combination set, tracking local target points through the optimal control instructions, and generating a final flight trajectory by complementing uncovered areas through iteration.

Description

Technical Field

[0001] This invention relates to the field of surveying and mapping path planning technology, and in particular to a method and system for full-coverage UAV surveying and mapping path planning. Background Technology

[0002] In recent years, with the rapid development of UAV technology and the increasing complexity of surveying and mapping needs, multi-source data fusion and intelligent path planning have become core technologies for improving the efficiency and safety of surveying and mapping operations. Especially in complex environments including dense obstacles, how to comprehensively utilize obstacle geometry information, UAV kinematic constraints, and real-time environmental perception data to conduct global and local collaborative path planning has become the key to achieving high-quality full-coverage surveying and mapping. Multi-source data tracing technology has emerged in this context. Its core lies in the organic fusion of different types of prior environmental data and real-time sensor data to construct a high-precision environmental representation model. Based on this model, the planning algorithm is driven to achieve dynamic, safe, and complete surveying trajectory generation, which is of great significance for ensuring the autonomous operation capability of UAVs in complex scenarios.

[0003] Existing technologies still have significant shortcomings. On the one hand, traditional methods often simply expand obstacles and then use them directly for path search, lacking fine-grained quantitative modeling of the degree of obstacle influence. This makes it difficult for the generated path to balance safety and path optimality when avoiding obstacles, and it is prone to problems such as local oscillation or excessive detour. On the other hand, most methods separate global path planning from local trajectory tracking. Global paths are usually generated based on fixed grids or sampling points, lacking the ability to continuously optimize node positions. Local planning often uses evaluation functions with fixed weights, making it difficult to dynamically adjust the control strategy according to the real-time distance between the UAV and the target and obstacles. This results in insufficient tracking accuracy and delayed obstacle avoidance response in complex terrain. In addition, existing full coverage strategies mostly adopt an open-loop mode of "plan first, then execute", lacking a closed-loop completion mechanism for areas missing in actual flight. This can easily lead to missed areas due to initial path deviations or environmental changes, affecting the integrity of the mapping. Summary of the Invention

[0004] The technical problem solved by this invention is that existing technologies still have significant shortcomings. On the one hand, traditional methods often simply expand obstacles and then directly use them for path search, lacking fine-grained quantitative modeling of the degree of influence of obstacles. This makes it difficult for the generated path to balance safety and path optimality when avoiding obstacles, and it is prone to local oscillations or excessive detours. On the other hand, most methods separate global path planning from local trajectory tracking. Global paths are usually generated based on fixed grids or sampling points, lacking the ability to continuously optimize node positions. Local planning often uses evaluation functions with fixed weights, making it difficult to dynamically adjust the control strategy according to the real-time distance between the UAV and the target and obstacles. This results in insufficient tracking accuracy and delayed obstacle avoidance response in complex terrain. In addition, existing full coverage strategies mostly adopt an open-loop mode of "plan first, then execute," lacking a closed-loop completion mechanism for areas missing in actual flight. This can easily lead to missed areas due to initial path deviations or environmental changes, affecting the integrity of the mapping.

[0005] To solve the above technical problems, the present invention provides the following technical solution: A full-coverage UAV mapping path planning method includes the following steps: Step S1, after the original obstacles are expanded, a neural network energy model is constructed, and the mapping area is rasterized. The total energy value of the center point of each grid is calculated using the neural network energy model to form a three-dimensional energy field matrix. Step S2: Based on the coordinates of the surveying start point and the surveying end point of the surveying area and the preset number of segments, an initial skeleton path node sequence is generated. Through iterative optimization, the positions of the internal nodes in the initial skeleton path node sequence are adjusted to obtain the optimized skeleton path node sequence. Step S3: Encrypt the optimized skeleton path node sequence according to the preset waypoint spacing to generate a dense waypoint sequence; Step S4: Based on the principle of rigid body kinematics, a three-dimensional UAV motion model is constructed. Based on the UAV kinematic constraints, a feasible velocity space is constructed and a candidate velocity combination set is generated. Using the dynamic window method, the weights are dynamically adjusted through a fuzzy controller. The optimal control command is selected from the candidate velocity combination set. The optimal control command is used to track local target points. The final flight trajectory is generated by iteratively completing the uncovered areas.

[0006] As a preferred embodiment of the full-coverage UAV mapping path planning method described in this invention, step S1 includes steps S101, S102, S103 and S104. Step S101: Obtain the geometric parameters of the original obstacle; The original obstacles include polyhedral obstacles and spherical obstacles; The geometric parameters include the original constraint surfaces of the polyhedral obstacle and the original center coordinates and original radius of the spherical obstacle; Each original obstacle is expanded to obtain the equations of the expanded constraint surface and the expanded sphere. The dilation process specifically includes dilating the original constraint surfaces of the polyhedral obstacle to obtain the corresponding dilated constraint surface equations. The expression for the dilated constraint surface equations is as follows: ; in, This represents the normal coefficient of the constrained surface after expansion. This represents the threshold parameter after dilation. Represents the coordinates of a point in space; Based on the original center and radius of the spherical obstacle, the spherical obstacle is expanded to obtain the equation of the expanded sphere. The expression of the equation of the expanded sphere is as follows: ; in, This represents the coordinates of the center of the spherical obstacle after it expands. This represents the radius of the spherical obstacle after it expands; Step S102: Obtain the coordinates of the spatial point to be calculated, and based on the center coordinates and radius of the sphere corresponding to the equation of the expanded sphere, as well as the normal coefficient and threshold parameter corresponding to the equation of the expanded constraint surface of each polyhedral obstacle, construct a neural network energy model to obtain the total energy value of each spatial point to be calculated. The construction of the neural network energy model specifically includes: constructing a first layer of hidden neurons based on the coordinates of the center of the sphere after expansion and the radius after expansion; and using the function of the first layer of neurons to calculate the distance correlation value of the spherical obstacle. The expression of the function of the first layer of neurons is as follows: ; in, This represents the distance correlation value between the j-th spherical obstacle and the i-th spatial point to be calculated. These represent the X, Y, and Z axes of the i-th spatial point to be calculated in the ground coordinate system, respectively. Let represent the radius of the j-th spherical obstacle after expansion. This represents the square of the radius of the j-th spherical obstacle after it expands. These represent the coordinate components of the center of the j-th spherical obstacle in the X, Y, and Z directions of the ground coordinate system after expansion; Each polyhedral obstacle is formed by multiple constraint surfaces, and each constraint surface corresponds to an expanded constraint surface equation. For the k-th constraint surface, calculate the signed distance from the spatial point to be calculated to the constraint surface, expressed as: ; in, This represents the signed distance from the spatial point to be calculated to the k-th constraint surface. This represents the normal coefficient of the k-th constraint surface. This represents the dilation threshold parameter corresponding to the k-th constraint surface; The maximum value among the signed distances of all constrained surfaces is taken as the distance correlation value of the polyhedral obstacle, denoted as . .

[0007] As a preferred embodiment of the full-coverage UAV mapping path planning method described in this invention, in step S103, the distance correlation value of each original obstacle is input into the second hidden layer neuron, and the distance correlation value is converted into the corresponding energy value using the Sigmoid function. The function expression of the second layer neuron is as follows: ; in, This represents the energy value of the j-th original obstacle. This represents the distance correlation value of the j-th original obstacle. If the original obstacle is a spherical obstacle, then... The value is the distance-related value of the spherical obstacle. ; If the original obstacle is a polyhedral obstacle, then The value is the distance-related value of the polyhedral obstacle. ; This represents the preset baseline threshold for the j-th original obstacle. Indicates the smoothing parameter; The energy values ​​are input to the output layer, and a summation function is used to calculate the sum of the energy values ​​of all original obstacles, thus obtaining the total energy value of the spatial point to be calculated. The expression for the summation function is: ; in, This represents the total energy value of the i-th spatial point to be calculated. This indicates the total number of all original obstacles; Step S104: Obtain the three-dimensional boundary and preset resolution of the survey area; perform rasterization processing on the survey area according to the three-dimensional boundary and resolution to obtain the coordinates of the center points of all grids. The coordinates of the center point of each grid are used as spatial points to be calculated, and the total energy value of each grid center point is obtained by using the neural network energy model. The total energy values ​​of all grid center points are arranged according to their spatial positions to form a three-dimensional energy field matrix.

[0008] As a preferred embodiment of the full-coverage UAV mapping path planning method described in this invention, step S2 includes steps S201, S202, S203 and S204. Step S201: Obtain the coordinates of the starting point and ending point of the surveying area, as well as the preset number of segments. Generate an initial coverage path based on the starting point, ending point, and preset number of segments to obtain the initial skeleton path node sequence. The process of generating the initial skeleton path node sequence involves dividing the straight path between the start point and the end point into m equal segments, resulting in a total of m+1 nodes. Each m+1 node includes the start point, the end point, and m-1 intermediate nodes. The expression for calculating the node position is: ; ; in, This represents the initial position of the i-th node. Indicates the coordinates of the starting point of the surveying area. This represents the coordinates of the endpoint of the survey area. When i=0, When i=m, m represents the number of segments; Step S202: In the t-th iteration, for each internal node in the initial skeleton path node sequence, candidate nodes are generated in the three-dimensional space along six orthogonal directions with a preset search step size. The step size is moved along the positive and negative X-axis, positive and negative Y-axis, and positive and negative Z-axis respectively to obtain six candidate nodes. The preset search step size is denoted as: ; The logic for generating the coordinates of the six candidate nodes is as follows: when moving along the positive X-axis, the coordinates of the candidate nodes are... ; When moving along the negative X-axis, the coordinates of the candidate node are: ; When moving along the positive Y-axis, the coordinates of the candidate node are: ; When moving along the negative Y-axis, the coordinates of the candidate node are: ; When moving along the positive Z-axis, the coordinates of the candidate node are: ; When moving along the negative Z-axis, the coordinates of the candidate node are: ; in, These represent the coordinate components of the i-th internal node in the initial skeleton path node sequence in the X, Y, and Z directions in the ground coordinate system during the t-th iteration. Step S203: For each candidate node, replace the corresponding internal node in the initial skeleton path node sequence with the candidate node to obtain the replaced candidate node. Combine the replaced candidate node with the start point, end point and other internal nodes in the initial skeleton path node sequence that did not participate in the replacement to form a candidate path. The total energy value of each candidate path is calculated by weighting the sum of the energy values ​​of all path nodes constituting the candidate path with the sum of the path lengths, as expressed in the following expression: ; in, This represents the total energy value of the candidate path formed after the i-th internal node is replaced along the j-th candidate direction. This represents the r-th path node in the candidate paths. Represents the energy field matrix. Path node The corresponding energy value in the energy field matrix and This represents the preset weighting coefficients. Denotes the Euclidean norm; For each internal node, select the minimum total energy value from the six corresponding candidate paths, and take the candidate node with the minimum total energy value as the new position of the internal node. Step S204: Repeat steps S202 and S203 for iteration, and determine whether the iteration termination condition is met after each iteration. The iteration termination condition is as follows: if the maximum distance between the new position of all internal nodes and the position in the previous iteration is less than a preset threshold, or the current iteration number has reached the preset maximum iteration number, then the iteration stops and the current skeleton path node sequence is output as the optimized skeleton path node sequence. Otherwise, proceed to the next iteration.

[0009] As a preferred embodiment of the full-coverage UAV mapping path planning method described in this invention, step S3 specifically includes: obtaining an optimized skeleton path node sequence, using B-spline interpolation to densify the paths between adjacent nodes in the optimized skeleton path node sequence, and generating a dense waypoint sequence according to a preset waypoint spacing.

[0010] As a preferred embodiment of the full-coverage UAV mapping path planning method described in this invention, step S4 includes steps S401, S402, S403, S404, S405 and S406. Step S401: Obtain the current state of the UAV, which includes the UAV's position coordinates, attitude angles, body speed, body angular velocity, and time interval. Based on the principles of rigid body kinematics, a three-dimensional UAV motion model is constructed, which consists of position update equations and attitude update equations. The expression for the position update equation is: ; ; in, Indicates time interval, This represents the rotation matrix from the body coordinate system to the ground coordinate system, using the ZYX Euler angle order. Indicates the yaw angle. Indicates pitch angle, Indicates the roll angle. s represents cos, and s represents sin. , , These represent the position coordinate components of the UAV in the X, Y, and Z axes of the ground coordinate system, respectively. , , These represent the linear velocity components of the UAV in the X, Y, and Z axes of the body coordinate system, respectively. The attitude update equation is expressed as follows: ; Wherein, the Euler angle vector is denoted as Euler angular rate is denoted as The angular velocity of the machine body is denoted as ,in, Represents Euler angle vectors. This represents the rate of change of the drone's attitude angle over time. , representing the angular velocity components of the UAV's rotation around each axis of the body coordinate system; ; Among them, Euler angular rate With body angular velocity The relationship is ; These represent the roll angle, pitch angle, and yaw angle at the current moment, respectively. These represent the angular velocity components of the UAV in the X, Y, and Z axes of the body coordinate system, respectively. express The inverse matrix; The position at the next moment is obtained based on the position update equation and the attitude update equation. and the attitude angle corresponding to the next moment .

[0011] As a preferred embodiment of the full-coverage UAV mapping path planning method described in this invention, step S402 involves using a dynamic window method to obtain the current body speed, body angular velocity, preset speed limit, acceleration limit, and real-time obstacle information. Real-time obstacle information includes the location and distance of obstacles detected in real time by airborne sensors; A feasible velocity space is constructed based on the kinematic constraints of the UAV. The feasible velocity space is the intersection of the velocity limit constraint, the acceleration limit constraint, and the braking distance constraint. The feasible velocity space is denoted as ; The expression for the speed limit constraint is: ; in, Indicates the speed limit constraint. Indicates the candidate linear velocity. Indicates the candidate angular velocity, These represent the preset minimum and maximum linear velocities, respectively. These represent the preset minimum and maximum angular velocities, respectively; The expression for the speed limit constraint is: ; in, This indicates a speed limit constraint. These represent the candidate linear velocity and candidate angular velocity after considering acceleration constraints, respectively. and These represent the current velocity and angular velocity of the machine, respectively. and These represent the preset maximum linear acceleration and maximum angular acceleration, respectively. and These represent the preset maximum linear deceleration and maximum angular deceleration, respectively. Indicates a time interval; The expression for the braking distance constraint is: ; in, Indicates braking distance constraint. This represents the candidate speed after considering braking distance constraints. Indicates the current candidate velocity The minimum distance between the predicted trajectory and the obstacle; In feasible velocity space The system performs uniform sampling at a preset resolution to generate a set of candidate velocity combinations.

[0012] As a preferred embodiment of the full-coverage UAV mapping path planning method described in this invention, step S403 involves initializing the current local target point as the first waypoint in the dense waypoint sequence using the dynamic window method. Obtain the distance between the current drone position and the current local target point, and record it as the first distance; obtain the distance between the drone and the nearest obstacle, and record it as the second distance; The first and second distances are input into a preset two-input four-output fuzzy controller for fuzzing to obtain the weight coefficients of the evaluation function; The weighting coefficients correspond to the weights of the yaw angle deviation term, pitch angle deviation term, obstacle distance term, and velocity term of the evaluation function, respectively. Step S404: For each candidate velocity combination in the candidate velocity combination set, predict the motion trajectory under the candidate velocity combination using the three-dimensional UAV motion model, and calculate the evaluation function value, the expression of which is: ; in, This represents the evaluation function value of the candidate velocity combination. and These represent the yaw angle deviation and pitch angle deviation between the UAV's current heading and the direction of the local target point, respectively. This represents the minimum distance between the predicted trajectory and the obstacle. Indicates the magnitude of the resultant velocity; Select the optimal control command from all candidate velocity combinations, choosing the one with the largest evaluation function value, and output it. Step S405: Calculate the first distance in real time. If the first distance is less than the preset distance threshold, select the next waypoint from the dense waypoint sequence as the new local target point and output the updated local target point.

[0013] As a preferred embodiment of the full-coverage UAV mapping path planning method described in this invention, step S406 involves obtaining the effective coverage width of the mapping sensor. The effective coverage width is calculated based on the current flight altitude of the UAV and the sensor's field of view. The calculation process is as follows: calculate the product of the current flight altitude of the UAV and 2, which is recorded as the first product; calculate the tangent of half of it multiplied by the sensor's field of view, which is recorded as the half-angle tangent; and calculate the product of the first product and the half-angle tangent, which is recorded as the second product. The second product is the effective coverage width. Discretize the survey area into grid cells and initialize all grid cells as uncovered. Traverse all adjacent waypoint pairs in the dense waypoint sequence of right step S3. For each pair of adjacent waypoints, calculate the projection area of ​​the sensor's scan band on the ground when the UAV flies along the flight segment between the pair of adjacent waypoints, and mark the grid cells covered by the projection area as covered. Check the coverage status of all grid cells, including uncovered and covered; The inspection process includes inserting new waypoints near the center of the uncovered area if there are uncovered grid cells, and generating smooth path segments between the new waypoints and adjacent waypoints using B-spline interpolation. Step S406 is repeated until all grid cells are marked as covered, and the final dense waypoint sequence is output as the flight trajectory of the UAV mapping.

[0014] A full-coverage UAV mapping path planning system includes a construction module, an optimization module, an encryption module, and a generation module; The module constructs a neural network energy model after inflating the original obstacles, and then rasterizes the survey area. The neural network energy model is used to calculate the total energy value of the center point of each grid, forming a three-dimensional energy field matrix. The optimization module generates an initial skeleton path node sequence based on the coordinates of the surveying start point and the surveying end point of the surveying area and the preset number of segments. Through iterative optimization, the positions of the internal nodes in the initial skeleton path node sequence are adjusted to obtain the optimized skeleton path node sequence. The encryption module encrypts the optimized skeleton path node sequence according to the preset waypoint spacing to generate a dense waypoint sequence. The generation module constructs a 3D UAV motion model based on the principle of rigid body kinematics, builds a feasible velocity space based on UAV kinematic constraints and generates a set of candidate velocity combinations. It adopts the dynamic window method, dynamically adjusts the weights through a fuzzy controller, selects the optimal control command from the set of candidate velocity combinations, tracks local target points through the optimal control command, and generates the final flight trajectory by iteratively completing the uncovered areas.

[0015] The beneficial effects of this invention are as follows: By transforming obstacle constraints into a continuous energy field through a neural network energy model, path optimization can finely balance safety and smoothness. By iteratively optimizing the position of skeleton nodes, the local optimum problem of traditional fixed grid paths is overcome, and the obstacle avoidance quality of the global path is improved. Furthermore, a fuzzy controller is used to dynamically adjust the evaluation weights of yaw angle deviation, pitch angle deviation, obstacle distance and velocity terms, so that the UAV pays more attention to directional accuracy when approaching the target point and emphasizes obstacle avoidance safety when approaching obstacles, thereby significantly enhancing the robustness of trajectory tracking in complex environments. Finally, through the iterative completion mechanism of the uncovered area, a full-coverage mapping trajectory with high integrity, high smoothness and kinematic constraints is achieved. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the steps of a full-coverage UAV mapping path planning method according to an embodiment of the present invention.

[0017] Figure 2 This is a basic flowchart of a full-coverage UAV mapping path planning system provided in one embodiment of the present invention. Detailed Implementation

[0018] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0019] Example 1, referring to Figure 1 As an embodiment of the present invention, a method for full-coverage UAV mapping path planning is provided, comprising the following steps: Step S1: After expanding the original obstacle, construct a neural network energy model, and rasterize the survey area. Calculate the total energy value of the center point of each grid using the neural network energy model to form a three-dimensional energy field matrix. Step S2: Based on the coordinates of the surveying start point and the surveying end point of the surveying area and the preset number of segments, an initial skeleton path node sequence is generated. Through iterative optimization, the positions of the internal nodes in the initial skeleton path node sequence are adjusted to obtain the optimized skeleton path node sequence. Step S3: Encrypt the optimized skeleton path node sequence according to the preset waypoint spacing to generate a dense waypoint sequence; Step S4: Based on the principle of rigid body kinematics, a three-dimensional UAV motion model is constructed. Based on the UAV kinematic constraints, a feasible velocity space is constructed and a candidate velocity combination set is generated. Using the dynamic window method, the weights are dynamically adjusted through a fuzzy controller. The optimal control command is selected from the candidate velocity combination set. The optimal control command is used to track local target points. The final flight trajectory is generated by iteratively completing the uncovered areas.

[0020] In one embodiment, step S1 transforms the original obstacle into a quantifiable potential field influence through expansion processing, addressing the difficulty in continuously assessing spatial safety in traditional geometric obstacle avoidance. Specific steps include: constructing a neural network energy model after expanding the original obstacle; rasterizing the survey area; calculating the total energy value at the center point of each raster using the neural network energy model to form a three-dimensional energy field matrix; and transforming obstacle constraints into a continuous energy distribution, providing a refined environmental cost basis for subsequent path optimization. Step S2 optimizes the global path structure by iteratively adjusting the positions of skeleton path nodes, overcoming the tendency of initial straight paths or fixed grid paths to fall into local optima. Specific steps include: generating an initial skeleton path node sequence based on the survey starting point coordinates, survey ending point coordinates, and a preset number of segments in the survey area; and adjusting the internal node positions in the initial skeleton path node sequence through iterative optimization to obtain an optimized skeleton path node sequence, achieving a balance between obstacle avoidance effectiveness and path smoothness at the global level. Step S3 serves as an intermediate link between global planning and local execution. Specifically, it involves encrypting the optimized skeleton path node sequence according to a preset waypoint spacing to generate a dense waypoint sequence, providing a continuous and appropriately dense target point sequence for subsequent local tracking. Step S4 is used to construct a local control framework that conforms to physical constraints and introduces adaptive weights and a coverage closed-loop mechanism to solve the problems of traditional dynamic window methods where fixed weights are difficult to adapt to dynamic environmental changes and open-loop planning easily missing coverage areas. Specifically, it involves constructing a 3D UAV motion model based on rigid body kinematics principles, constructing a feasible velocity space based on UAV kinematic constraints and generating a candidate velocity combination set, using a dynamic window method, dynamically adjusting weights through a fuzzy controller, selecting the optimal control command from the candidate velocity combination set, tracking local target points through the optimal control command, and generating the final flight trajectory by iteratively completing the uncovered areas. This achieves accurate tracking under kinematic constraints and improves environmental adaptability, while ensuring the integrity of the mapping task through coverage feedback. The entire design framework organically combines environmental energy modeling, global skeleton optimization, local adaptive tracking, and coverage closed-loop completion to form a path planning architecture that extends from global structural planning to local dynamic control and then to mission integrity verification. This significantly improves the safety, smoothness, and mission integrity of the full-coverage mapping path for UAVs in complex obstacle environments.

[0021] Step S1 includes steps S101, S102, S103 and S104; Step S101: Obtain the geometric parameters of the original obstacle; The original obstacles include polyhedral obstacles and spherical obstacles; The geometric parameters include the original constraint surfaces of the polyhedral obstacle and the original center coordinates and original radius of the spherical obstacle; Each original obstacle is expanded to obtain the equations of the expanded constraint surface and the expanded sphere. The dilation process specifically includes dilating the original constraint surfaces of the polyhedral obstacle to obtain the corresponding dilated constraint surface equations. The expression for the dilated constraint surface equations is as follows: ; in, This represents the normal coefficient of the constrained surface after expansion. This represents the threshold parameter after dilation. Represents the coordinates of a point in space; Based on the original center and radius of the spherical obstacle, the spherical obstacle is expanded to obtain the equation of the expanded sphere. The expression of the equation of the expanded sphere is as follows: ; in, This represents the coordinates of the center of the spherical obstacle after it expands. This represents the radius of the spherical obstacle after it expands; Step S102: Obtain the coordinates of the spatial point to be calculated, and based on the center coordinates and radius of the sphere corresponding to the equation of the expanded sphere, as well as the normal coefficient and threshold parameter corresponding to the equation of the expanded constraint surface of each polyhedral obstacle, construct a neural network energy model to obtain the total energy value of each spatial point to be calculated. The construction of the neural network energy model specifically includes: constructing a first layer of hidden neurons based on the coordinates of the center of the sphere after expansion and the radius after expansion; and using the function of the first layer of neurons to calculate the distance correlation value of the spherical obstacle. The expression of the function of the first layer of neurons is as follows: ; in, This represents the distance correlation value between the j-th spherical obstacle and the i-th spatial point to be calculated. These represent the X, Y, and Z axes of the i-th spatial point to be calculated in the ground coordinate system, respectively. Let represent the radius of the j-th spherical obstacle after expansion. This represents the square of the radius of the j-th spherical obstacle after it expands. These represent the coordinate components of the center of the j-th spherical obstacle in the X, Y, and Z directions of the ground coordinate system after expansion; Each polyhedral obstacle is formed by multiple constraint surfaces, and each constraint surface corresponds to an expanded constraint surface equation. For the k-th constraint surface, calculate the signed distance from the point in space to be calculated to the constraint surface, expressed as: ; in, This represents the signed distance from the spatial point to be calculated to the k-th constraint surface. This represents the normal coefficient of the k-th constraint surface. This represents the dilation threshold parameter corresponding to the k-th constraint surface; The maximum value among the signed distances of all constrained surfaces is taken as the distance correlation value of the polyhedral obstacle, denoted as . .

[0022] Step S103: Input the distance correlation value of each original obstacle into the second hidden layer neuron, and use the Sigmoid function to convert the distance correlation value into the corresponding energy value. The function expression of the second layer neuron is: ; in, This represents the energy value of the j-th original obstacle. This represents the distance correlation value of the j-th original obstacle. If the original obstacle is a spherical obstacle, then... The value is the distance-related value of the spherical obstacle. ; If the original obstacle is a polyhedral obstacle, then The value is the distance-related value of the polyhedral obstacle. ; This represents the preset baseline threshold for the j-th original obstacle. Indicates the smoothing parameter; The energy values ​​are input to the output layer, and a summation function is used to calculate the sum of the energy values ​​of all original obstacles, thus obtaining the total energy value of the spatial point to be calculated. The expression for the summation function is: ; in, This represents the total energy value of the i-th spatial point to be calculated. This indicates the total number of all original obstacles; Step S104: Obtain the three-dimensional boundary and preset resolution of the survey area; perform rasterization processing on the survey area according to the three-dimensional boundary and resolution to obtain the coordinates of the center points of all grids. The coordinates of the center point of each grid are used as the spatial point to be calculated. The total energy value of the center point of each grid is obtained by using a neural network energy model. The total energy values ​​of all grid center points are arranged according to their spatial positions to form a three-dimensional energy field matrix.

[0023] In one embodiment, step S101 transforms the original obstacles into a unified mathematical description form through dilation processing, providing standardized geometric boundaries for subsequent energy field construction. Specifically, this includes obtaining the geometric parameters of the original obstacles, which include polyhedral obstacles and spherical obstacles. The geometric parameters include the original constraint surfaces of the polyhedral obstacles and the original center coordinates and original radii of the spherical obstacles. Dilation processing is performed on each original obstacle to obtain the dilated constraint surface equation and the dilated spherical surface equation. The dilation processing specifically includes dilating the polyhedral obstacles based on their original constraint surfaces to obtain the corresponding dilated constraint surface equation. This represents the normal coefficient of the constrained surface after expansion. This represents the threshold parameter after expansion. This threshold parameter is set based on a safety distance, for example, using the maximum circumscribed circle radius of the UAV fuselage as the safety distance. It is calculated by shifting the original constraint surface outward by this distance. value, Represents the coordinates of a point in space; based on the original center and radius of the spherical obstacle, the spherical obstacle is expanded to obtain the equation of the expanded sphere, where... The coordinates of the center of the spherical obstacle after expansion are represented. The expanded radius r is obtained by adding the original radius and the safety distance. The safety distance is taken as the radius of the largest circumcircle of the UAV fuselage. This process makes the expanded geometric boundary equivalent to the safety boundary that allows the UAV's center of mass to move. Step S102 involves transforming the geometric constraints of the expanded obstacle into quantifiable distance-related values, constructing the input layer and the first hidden layer of the neural network energy model, and achieving a continuous representation of the obstacle's influence. Specifically, this includes obtaining the coordinates of the spatial point to be calculated, and based on the coordinates and radius of the sphere's center corresponding to the expanded spherical equation, as well as the normal coefficient and expanded threshold parameter corresponding to the expanded constraint surface equation of each polyhedral obstacle. A neural network energy model is constructed to obtain the total energy value of each spatial point to be calculated. Specifically, the construction of the neural network energy model includes building a first layer of hidden neurons based on the expanded sphere's center coordinates and expanded radius, and using the functions of the first layer of neurons to calculate the distance-related values ​​of the spherical obstacle. This represents the distance correlation value between the j-th spherical obstacle and the i-th spatial point to be calculated. Let X, Y, and Z represent the coordinates of the i-th spatial point to be calculated in the X, Y, and Z directions of the ground coordinate system, respectively. Let represent the radius of the j-th spherical obstacle after expansion. Let X, Y, and Z represent the coordinate components of the expanded center coordinates of the j-th spherical obstacle in the X, Y, and Z axes of the ground coordinate system, respectively. Each polyhedral obstacle is formed by multiple constraint surfaces, and each constraint surface corresponds to an expanded constraint surface equation. For the k-th constraint surface, the signed distance from the point to be calculated to that constraint surface is calculated, where... This represents the signed distance from the spatial point to be calculated to the k-th constraint surface. This represents the normal coefficient of the k-th constraint surface. This represents the dilated threshold parameter corresponding to the k-th constraint surface; the maximum value among the signed distances of all constraint surfaces is taken as the distance correlation value of the polyhedral obstacle, denoted as . This process ensures that as long as a spatial point is located inside any constraint surface, it is determined to be inside a polyhedral obstacle, thus achieving a unified quantitative expression of the internal space of complex polyhedral obstacles. Step S103 involves mapping the distance correlation values ​​to smooth, continuous energy values ​​using the Sigmoid function, and summing the total energy values ​​of the spatial points to ensure a smooth transition of the energy field at obstacle boundaries. This avoids abrupt changes in path search caused by traditional binary obstacle representations. Specifically, this includes inputting the distance correlation value of each original obstacle into the second hidden layer neuron, and using the Sigmoid function to convert the distance correlation value into the corresponding energy value. The function expression for the second layer neuron is: where... This represents the energy value of the j-th original obstacle. This represents the distance correlation value of the j-th original obstacle. If the original obstacle is a spherical obstacle, then... Values If the original obstacle is a polyhedral obstacle, then Values , This represents the preset baseline threshold for the j-th original obstacle. This threshold is set based on the distance-related value at the boundary after the obstacle expands. For example, for a spherical obstacle, it is set to... = So that on the expanded spherical surface =0.5, for polyhedral obstacles, take =0 makes the constraint surface after expansion equal to 0. =0.5, T represents the smoothing parameter, which takes a value between 0.1 and 0.5 to control the width of the transition zone from 0 to 1 in terms of energy value; the energy value is input to the output layer, and the summation function is used to calculate the sum of the energy values ​​of all original obstacles to obtain the total energy value of the spatial point to be calculated, where represents the total energy value of the i-th spatial point to be calculated, and n represents the number of all original obstacles. This summation process allows the effects of multiple obstacles to be superimposed in space, forming a higher energy value in areas with dense obstacles, thereby guiding path planning to prioritize safe areas with lower energy values. Step S104 discretizes the continuous space into a grid energy field matrix, providing a directly indexable environmental cost query structure for subsequent path optimization. Specifically, this includes obtaining the 3D boundary and preset resolution of the surveyed area. The resolution is set based on a balance between surveying accuracy and computational resources; for example, half the size of the UAV fuselage is used as the grid side length. The surveyed area is rasterized according to the 3D boundary and resolution to obtain the coordinates of all grid center points. The coordinates of each grid center point are used as spatial points to be calculated, and a neural network energy model is used to calculate the total energy value of each grid center point. The total energy values ​​of all grid center points are arranged according to their spatial positions to form a 3D energy field matrix. The three-dimensional energy field matrix serves as a discretized representation of environmental cost, enabling subsequent steps to query the obstacle impact level at any spatial point using only matrix indexing and interpolation operations, significantly improving the computational efficiency of the global path optimization process. Step S1, through a combination of dilation processing, neural network energy modeling, and rasterization, transforms complex geometric obstacles into a continuous, smooth, and efficiently queryable three-dimensional energy field matrix, providing a precise environmental cost basis for the iterative optimization of the skeleton path nodes in step S2. This allows the global path to achieve a quantitative trade-off between safety and smoothness, fundamentally solving the problems of fuzzy safety boundaries and difficulty in continuously assessing obstacle impact in traditional obstacle processing.

[0024] Step S2 includes steps S201, S202, S203 and S204; Step S201: Obtain the coordinates of the starting point and ending point of the surveying area, as well as the preset number of segments. Generate an initial coverage path based on the starting point, ending point, and preset number of segments to obtain the initial skeleton path node sequence. The process of generating the initial skeleton path node sequence involves dividing the straight path between the start point and the end point into m equal segments, resulting in a total of m+1 nodes. Each m+1 node includes the start point, the end point, and m-1 intermediate nodes. The expression for calculating the node position is: ; ; in, This represents the initial position of the i-th node. Indicates the coordinates of the starting point of the surveying area. This represents the coordinates of the endpoint of the survey area. When i=0, When i=m, m represents the number of segments; Step S202: In the t-th iteration, for each internal node in the initial skeleton path node sequence, candidate nodes are generated in the three-dimensional space along six orthogonal directions with a preset search step size. The step size is moved along the positive and negative X-axis, positive and negative Y-axis, and positive and negative Z-axis respectively to obtain six candidate nodes. The preset search step size is denoted as: ; The logic for generating the coordinates of the six candidate nodes is as follows: when moving along the positive X-axis, the coordinates of the candidate nodes are... ; When moving along the negative X-axis, the coordinates of the candidate node are: ; When moving along the positive Y-axis, the coordinates of the candidate node are: ; When moving along the negative Y-axis, the coordinates of the candidate node are: ; When moving along the positive Z-axis, the coordinates of the candidate node are: ; When moving along the negative Z-axis, the coordinates of the candidate node are: ; in, These represent the coordinate components of the i-th internal node in the initial skeleton path node sequence in the X, Y, and Z directions in the ground coordinate system during the t-th iteration. Step S203: For each candidate node, replace the corresponding internal node in the initial skeleton path node sequence with the candidate node to obtain the replaced candidate node. Combine the replaced candidate node with the start point, end point and other internal nodes in the initial skeleton path node sequence that did not participate in the replacement to form a candidate path. The total energy value of each candidate path is calculated by weighting the sum of the energy values ​​of all path nodes constituting the candidate path with the sum of the path lengths, as expressed in the following expression: ; in, This represents the total energy value of the candidate path formed after the i-th internal node is replaced along the j-th candidate direction. This represents the r-th path node in the candidate paths. Represents the energy field matrix. Path node The corresponding energy value in the energy field matrix and This represents the preset weighting coefficients. Denotes the Euclidean norm; For each internal node, select the minimum total energy value from the six candidate paths and take the candidate node with the minimum total energy value as the new position of the internal node. Step S204: Repeat steps S202 and S203 for iteration, and determine whether the iteration termination condition is met after each iteration. The iteration termination condition is that if the maximum distance between the new position of all internal nodes and the position in the previous iteration is less than a preset threshold, or the current iteration number has reached the preset maximum iteration number, then the iteration stops and the current skeleton path node sequence is output as the optimized skeleton path node sequence. Otherwise, proceed to the next iteration.

[0025] In one embodiment, step S201 establishes a simple initial reference path from the starting point to the ending point as the basic skeleton for subsequent optimization, avoiding the initial path from falling into a random state due to lack of guidance. Specifically, this includes obtaining the coordinates of the starting point and the ending point of the surveying area, as well as the preset number of segments. An initial coverage path is generated based on the starting point, the ending point, and the preset number of segments to obtain the initial skeleton path node sequence. The number of segments m is preset based on the size of the surveying area and the performance of the UAV. For example, it is obtained by dividing the Euclidean distance between the starting point and the ending point by the maximum safe flight distance of a single path segment of the UAV at a typical cruise speed and then rounding it up to ensure that the spacing between adjacent nodes is moderate. When generating the initial skeleton path node sequence, the straight path between the starting point and the ending point is divided into m segments, resulting in a total of m+1 nodes, including the starting point, the ending point, and m-1 intermediate nodes. Although the initial path formed in this way does not consider obstacles, it provides an adjustable node distribution basis for subsequent iterative optimization. Step S202 provides candidate adjustment directions for each internal node through local neighborhood search, enabling the path to flexibly deform in three-dimensional space to avoid obstacles. Specifically, in the t-th iteration, for each internal node in the initial skeleton path node sequence, candidate nodes are generated in three-dimensional space along six orthogonal directions with a preset search step size. The step size is moved along the positive and negative X-axis, positive and negative Y-axis, and positive and negative Z-axis directions to obtain six candidate nodes. The preset search step size δ is set according to the resolution of the rasterization process in step S1, for example, it is 0.5 to 1 times the raster side length. This ensures that each adjustment can effectively change the path shape, while avoiding the step size being too large and skipping low-energy regions. This step provides a concise set of candidate positions for each internal node that covers all spatial dimensions by enumerating the movement in six orthogonal directions. Step S203 involves quantitatively evaluating candidate paths using an evaluation function, selecting the node positions that optimally combine the total energy value and path length of the path, thus achieving local optimal adjustment for a single node. Specifically, for each candidate node, the corresponding internal node in the initial skeleton path node sequence is replaced with the candidate node to obtain the replaced candidate node. The replaced candidate node, along with the start point, end point, and other internal nodes in the initial skeleton path node sequence that were not replaced, constitute a candidate path. The total energy value of each candidate path is calculated by summing the energy values ​​of all path nodes constituting the candidate path and the sum of the path length. The weights are obtained, where the weighting coefficients α and β are preset based on the balance between safety preference and smoothness preference, for example, α=1 and β=0.5, so that the energy cost and the path length cost are on the same order of magnitude. The energy value is obtained by querying the three-dimensional energy field matrix formed in step S1, and the path length is obtained by summing the Euclidean distances between adjacent nodes. For each internal node, the minimum value is selected from the total energy values ​​of the corresponding six candidate paths, and the candidate node with the minimum total energy value is taken as the new position of the internal node. In this way, the position of each node is gradually optimized through a local greedy strategy, so that the path evolves towards lower energy and shorter length while maintaining the connection with the starting point and the ending point. Step S204 involves using an iterative mechanism to gradually converge the path to a stable state, avoiding the waste of computational resources caused by infinite optimization. Specifically, this includes repeatedly executing steps S202 and S203 for iteration, and determining whether the iteration termination condition is met after each iteration. The iteration termination condition is: if the maximum distance between the new positions of all internal nodes and their positions in the previous iteration is less than a preset threshold, or if the current iteration count has reached the preset maximum iteration count, then the iteration stops, and the current skeleton path node sequence is output as the optimized skeleton path node sequence. The preset threshold is set according to the optimization accuracy requirements; for example, it is set to 0.1 times the grid side length, indicating that the path... The path has been basically stabilized. The maximum number of iterations is preset based on computing resources and convergence speed, for example, 50 to 100 times, to ensure that the algorithm can terminate within a reasonable time. This iterative mechanism ensures the controllability and convergence of the path optimization process. The entire step S2 transforms global path optimization into a series of local adjustment processes of discrete nodes through initial straight path generation, six-directional neighborhood search, comprehensive cost evaluation and iterative convergence control. This allows the three-dimensional energy field provided in step S1 to effectively guide the path to avoid high-energy areas, ensuring safety while taking into account path length. This provides a structurally reasonable and obstacle-avoiding skeleton path for waypoint encryption in step S3.

[0026] Step S3 involves encrypting the skeleton path node sequence into a uniformly dense waypoint sequence to provide continuous target point input for subsequent local dynamic tracking. Specifically, this includes obtaining the optimized skeleton path node sequence, using B-spline interpolation to encrypt the paths between adjacent nodes in the optimized skeleton path node sequence, and generating a dense waypoint sequence according to a preset waypoint spacing. The preset waypoint spacing is set based on the UAV's dynamic response capability and the sampling frequency of the mapping sensor, for example, 0.5 to 1 times the flight distance of the UAV in one control cycle at a typical cruise speed. This ensures that the generated waypoint sequence can be effectively tracked by the UAV motion model without causing redundant calculations due to excessive density. B-spline interpolation ensures the smoothness of the encrypted waypoint sequence and avoids trajectory bends caused by sparse nodes, enabling a continuous and kinematically sound reference trajectory to be obtained during local target point tracking in step S4.

[0027] Step S3 specifically includes obtaining the optimized skeleton path node sequence, using the B-spline interpolation method to encrypt the paths between adjacent nodes in the optimized skeleton path node sequence, and generating a dense waypoint sequence according to the preset waypoint spacing.

[0028] In one embodiment, step S3 transforms the optimized skeleton path node sequence from a sparse node pattern into a dense waypoint sequence with uniform density and continuous curvature. This provides a smooth and kinematically sound target point input for subsequent local dynamic tracking. This step acts as a bridge between global skeleton optimization and local trajectory tracking, enabling the discrete node path generated in step S2 to be transformed into a continuous reference trajectory that can be directly tracked by the dynamic window method in step S4. Specifically, this includes obtaining the optimized skeleton path node sequence output in step S2 and using cubic B-spline interpolation to densify the path between adjacent nodes. Cubic B-spline interpolation ensures that the interpolated curve achieves second-order continuity at the nodes, effectively avoiding... To avoid trajectory bends caused by sparse nodes, a dense waypoint sequence is generated during the encryption process according to a preset waypoint spacing. The preset waypoint spacing is set based on the UAV's dynamic response capability and the sampling frequency of the mapping sensor. For example, it is set to 0.6 times the flight distance of the UAV in one control cycle at a typical cruise speed. This ensures that the generated waypoints are dense enough for the UAV to accurately track the reference trajectory, while avoiding an increase in computational burden due to excessive waypoint density. Through this step, the skeleton path originally consisting of only m+1 nodes is expanded into a dense sequence of hundreds of waypoints. Each waypoint falls on a smooth path defined by a B-spline curve, thus providing a continuous and easily trackable sequence of local target points for step S4.

[0029] Step S4 includes steps S401, S402, S403, S404, S405 and S406. Step S401: Obtain the current state of the UAV, which includes the UAV's position coordinates, attitude angles, body speed, body angular velocity, and time interval. Based on the principles of rigid body kinematics, a three-dimensional UAV motion model is constructed, which consists of position update equations and attitude update equations. The expression for the position update equation is: ; ; in, Indicates time interval, This represents the rotation matrix from the body coordinate system to the ground coordinate system, using the ZYX Euler angle order. Indicates the yaw angle. Indicates pitch angle, Indicates the roll angle. s represents cos, and s represents sin. , , These represent the position coordinate components of the UAV in the X, Y, and Z axes of the ground coordinate system, respectively. , , These represent the linear velocity components of the UAV in the X, Y, and Z axes of the body coordinate system, respectively. The attitude update equation is expressed as follows: ; Wherein, the Euler angle vector is denoted as Euler angular rate is denoted as The angular velocity of the machine body is denoted as ,in, Represents Euler angle vectors. This represents the rate of change of the drone's attitude angle over time. , representing the angular velocity components of the UAV's rotation around each axis of the body coordinate system; ; Among them, Euler angular rate With body angular velocity The relationship is ; These represent the roll angle, pitch angle, and yaw angle at the current moment, respectively. These represent the angular velocity components of the UAV in the X, Y, and Z axes of the body coordinate system, respectively. express The inverse matrix; The position at the next moment is obtained based on the position update equation and the attitude update equation. and the attitude angle corresponding to the next moment .

[0030] Step S402: Use the dynamic window method to obtain the current body velocity, body angular velocity, preset speed limit, acceleration limit and real-time obstacle information; Real-time obstacle information includes the location and distance of obstacles detected in real time by airborne sensors; A feasible velocity space is constructed based on the kinematic constraints of the UAV. The feasible velocity space is the intersection of the velocity limit constraint, the acceleration limit constraint, and the braking distance constraint. The feasible velocity space is denoted as ; The expression for the speed limit constraint is: ; in, Indicates the speed limit constraint. Indicates the candidate linear velocity. Indicates the candidate angular velocity, These represent the preset minimum and maximum linear velocities, respectively. These represent the preset minimum and maximum angular velocities, respectively; The expression for the speed limit constraint is: ; in, This indicates a speed limit constraint. These represent the candidate linear velocity and candidate angular velocity after considering acceleration constraints, respectively. and These represent the current velocity and angular velocity of the machine, respectively. and These represent the preset maximum linear acceleration and maximum angular acceleration, respectively. and These represent the preset maximum linear deceleration and maximum angular deceleration, respectively. Indicates a time interval; The expression for the braking distance constraint is: ; in, Indicates braking distance constraint. This represents the candidate speed after considering braking distance constraints. Indicates the current candidate velocity The minimum distance between the predicted trajectory and the obstacle; In feasible velocity space The system performs uniform sampling at a preset resolution to generate a set of candidate velocity combinations.

[0031] Step S403: Initialize the current local target point as the first waypoint in the dense waypoint sequence using the dynamic window method; Obtain the distance between the current drone position and the current local target point, and record it as the first distance; obtain the distance between the drone and the nearest obstacle, and record it as the second distance; The first and second distances are input into a preset two-input four-output fuzzy controller for fuzzing to obtain the weight coefficients of the evaluation function; The weighting coefficients correspond to the weights of the yaw angle deviation term, pitch angle deviation term, obstacle distance term, and velocity term in the evaluation function, respectively. Step S404: For each candidate velocity combination in the candidate velocity combination set, predict the motion trajectory under the candidate velocity combination using the 3D UAV motion model, and calculate the evaluation function value, the expression of which is: ; in, This represents the evaluation function value of the candidate velocity combination. and These represent the yaw angle deviation and pitch angle deviation between the UAV's current heading and the direction of the local target point, respectively. This represents the minimum distance between the predicted trajectory and the obstacle. Indicates the magnitude of the resultant velocity; Select the optimal control command from all candidate velocity combinations, choosing the one with the largest evaluation function value, and output it. Step S405: Calculate the first distance in real time. If the first distance is less than the preset distance threshold, select the next waypoint from the dense waypoint sequence as the new local target point and output the updated local target point.

[0032] Step S406: Obtain the effective coverage width of the mapping sensor. The effective coverage width is calculated based on the current flight altitude of the UAV and the sensor's field of view. The calculation process is as follows: calculate the product of the current flight altitude of the UAV and 2, which is recorded as the first product; calculate the tangent of half of it multiplied by the sensor's field of view, which is recorded as the half-angle tangent; calculate the product of the first product and the half-angle tangent, which is recorded as the second product. The second product is the effective coverage width. Discretize the survey area into grid cells and initialize all grid cells as uncovered. Traverse all adjacent waypoint pairs in the dense waypoint sequence of right step S3. For each pair of adjacent waypoints, calculate the projection area of ​​the sensor's scan band on the ground when the UAV flies along the flight segment between the pair of adjacent waypoints, and mark the grid cells covered by the projection area as covered. Check the coverage status of all grid cells, including uncovered and covered; The inspection process includes inserting new waypoints near the center of the uncovered area if there are uncovered grid cells, and generating smooth path segments between the new waypoints and adjacent waypoints using B-spline interpolation. Step S406 is repeated until all grid cells are marked as covered, and the final dense waypoint sequence is output as the flight trajectory of the UAV mapping.

[0033] In one embodiment, step S401 establishes the mathematical description of the UAV's motion state, providing an accurate physical model for subsequent local trajectory prediction and control command generation. This step provides a prediction model that matches the actual dynamic characteristics of the UAV for the dynamic window method, ensuring that the candidate trajectory generated in step S404 can accurately reflect the UAV's actual motion capabilities. Specifically, this includes obtaining the UAV's current state, including its position coordinates, attitude angles, body velocity, body angular velocity, and time interval; and constructing a three-dimensional UAV motion model based on rigid body kinematics principles. The three-dimensional UAV motion model consists of position update equations and attitude update equations, where the time interval... Based on real-time requirements, for example, the control cycle of the UAV flight control system is set to 0.02 seconds to 0.05 seconds. The position and attitude of the next moment can be accurately predicted by the three-dimensional UAV motion model based on the current state and the selected control command, providing a reliable recursive relationship for trajectory prediction of candidate velocity combinations. Step S402 involves constructing a feasible velocity space that conforms to the physical constraints of the UAV. By filtering out executable candidate velocity combinations through the intersection of multiple constraints, unfeasible control commands are avoided. This step transforms the UAV's hardware limitations and real-time perceived environmental information into a controllable decision space, providing a legitimate and safe candidate set for the evaluation and selection in step S404. Specifically, this includes using a dynamic window method to obtain the current body velocity, body angular velocity, preset velocity limits, acceleration limits, and real-time obstacle information. Real-time obstacle information includes the location and distance of obstacles detected in real-time by onboard sensors such as lidar or binocular cameras. A feasible velocity space is constructed based on the UAV's kinematic constraints, which is the intersection of velocity limit constraints, acceleration limit constraints, and braking distance constraints. The velocity limit constraints are preset based on the UAV's hardware performance, for example, taking the minimum linear velocity. 0 meters per second, maximum linear velocity The minimum angular velocity is 10 meters per second. -1 radians per second, maximum angular velocity 1 radian per second; maximum linear acceleration in the acceleration constraint. and maximum linear deceleration Based on the drone's dynamic response capabilities, for example, take 5 meters per square second The maximum angular acceleration is 8 meters per square second. and maximum angular deceleration Based on the attitude response capability settings, for example, take 2 radians per square second The time interval is 3 radians per square second. Take from step S401 The same value is used to ensure consistency; the braking distance constraint ensures that the vehicle can safely stop before hitting the obstacle at the current candidate speed, where dist( , The minimum Euclidean distance between the predicted trajectory and the obstacle position is obtained; uniform sampling is performed within the feasible velocity space at a preset resolution. The resolution is set according to the balance between control accuracy and computing resources. For example, the linear velocity sampling interval is 0.5 meters per second and the angular velocity sampling interval is 0.1 radians per second, generating a candidate velocity combination set that covers all physically feasible velocity choices. Step S403: Dynamically adjust the weights of each indicator in the evaluation function based on the real-time distance between the UAV and the target point and obstacles. This makes the UAV prioritize speed and directional efficiency in open areas, emphasize angular accuracy when approaching the target point, and prioritize obstacle avoidance safety when approaching obstacles. This step achieves adaptive adjustment of the control strategy, overcoming the shortcomings of the traditional dynamic window method with fixed weights that are difficult to adapt to environmental changes. Specifically, it includes: initializing the current local target point as the first waypoint in the dense waypoint sequence using the dynamic window method; obtaining the Euclidean distance between the current UAV position and the current local target point and recording it as the first distance; and obtaining the Euclidean distance between the UAV and the nearest obstacle. The distance is denoted as the second distance. The first and second distances are input into a preset two-input four-output fuzzy controller for fuzzy processing to obtain the weight coefficients of the evaluation function. These weight coefficients correspond to the weights of the yaw angle deviation, pitch angle deviation, obstacle distance, and velocity terms of the evaluation function, respectively. For example, the first distance is divided into three fuzzy sets: "near", "medium", and "far", and the second distance is divided into three fuzzy sets: "danger", "warning", and "safe". The output weight coefficients are determined according to the control priority, such that when the first distance is small, the weights of the yaw angle deviation and pitch angle deviation terms are increased to improve pointing accuracy, and when the second distance is small, the weight of the obstacle distance term is increased to prioritize obstacle avoidance. Step S404 involves comprehensively evaluating the performance of candidate velocity combinations in terms of directional accuracy, obstacle avoidance capability, and motion efficiency, selecting the optimal control command for the UAV to execute, and achieving real-time optimized tracking of the local trajectory. Specifically, for each candidate velocity combination in the candidate velocity combination set, the 3D UAV motion model constructed in step S401 is used to predict the motion trajectory under that candidate velocity combination, and the evaluation function value is calculated, where the yaw angle deviation term is included. The pitch deviation term represents the absolute value of the difference between the current heading angle of the UAV and the direction angle pointing to the local target point. The pitch angle of the UAV is represented by the absolute value of the difference between the current pitch angle and the pitch angle pointing towards the local target point; dist represents the minimum distance between the predicted trajectory and the real-time obstacle information in step S402; speed represents the magnitude of the resultant velocity; and the evaluation function value G is derived from the yaw angle deviation term. , The values ​​of speed, dist, and speed are multiplied by the corresponding weight coefficients output in step S403 and then summed. The optimal control command is selected from all candidate speed combinations, and the command with the largest evaluation function value is output. This command can achieve the optimal balance between safety and tracking accuracy in the current environment. Step S405: Dynamically update the local target point based on the proximity of the UAV to the target point to ensure that the UAV continuously advances along the dense waypoint sequence until the entire planned path is completed. Specifically, this includes: calculating the first distance in real time, i.e., the Euclidean distance between the current UAV position and the current local target point; if the distance is less than a preset distance threshold, selecting the next waypoint from the dense waypoint sequence as the new local target point and outputting the updated local target point; wherein the preset distance threshold is set based on the UAV's positioning accuracy and tracking error, for example, taking 0.5 times the radius of the UAV's maximum outer circle, to ensure that the UAV switches to the next target point only after it is sufficiently close to the current target point, avoiding path oscillations caused by frequent switching; Step S406 involves detecting and filling in missing areas through a coverage status feedback mechanism to ensure complete coverage of the surveyed area by the final flight trajectory. This step incorporates the integrity of the surveying task as a closed-loop feedback into the planning process, solving the problem of missed scans caused by initial deviations or environmental changes in traditional open-loop planning. Specifically, this includes obtaining the effective coverage width of the surveying sensor, which is calculated based on the UAV's current flight altitude and the sensor's field of view. For example, when the sensor's field of view is 90 degrees, the effective coverage width is equal to twice the flight altitude. The surveyed area is discretized into grid cells, with the grid cell size set according to the surveying accuracy requirements, for example, 1 / 5 of the UAV's effective coverage width. All grid cells are initialized to an uncovered state. The dense grid cells generated in step S3 are then traversed. For each pair of adjacent waypoints in the waypoint sequence, the projection area of ​​the sensor's scan strip on the ground is calculated when the UAV flies along that flight segment. This projection area is formed by extending half the effective coverage width to both sides of the projection line of the flight trajectory on the ground. The grid cells covered by the projection area are marked as covered. The coverage status of all grid cells is checked. If there are uncovered grid cells, a new waypoint is inserted near the center of the uncovered area. A smooth path segment between the new waypoint and the adjacent waypoint is generated by the B-spline interpolation method. The coverage check and completion operation is repeated until all grid cells are marked as covered. The final dense waypoint sequence is output as the flight trajectory of the UAV mapping. This closed-loop completion mechanism ensures the integrity of the mapping task.

[0034] Example 2, refer to Figure 2 In another embodiment of the present invention, which differs from the first embodiment, a full-coverage UAV mapping path planning system is provided, including a construction module, an optimization module, an encryption module, and a generation module; The module constructs a neural network energy model after inflating the original obstacles, and then rasterizes the survey area. The neural network energy model is used to calculate the total energy value of the center point of each grid, forming a three-dimensional energy field matrix. The optimization module generates an initial skeleton path node sequence based on the coordinates of the surveying start point and the surveying end point of the surveying area and the preset number of segments. Through iterative optimization, the positions of the internal nodes in the initial skeleton path node sequence are adjusted to obtain the optimized skeleton path node sequence. The encryption module encrypts the optimized skeleton path node sequence according to the preset waypoint spacing to generate a dense waypoint sequence. The generation module constructs a 3D UAV motion model based on the principle of rigid body kinematics, builds a feasible velocity space based on UAV kinematic constraints and generates a set of candidate velocity combinations. It adopts the dynamic window method, dynamically adjusts the weights through a fuzzy controller, selects the optimal control command from the set of candidate velocity combinations, tracks local target points through the optimal control command, and generates the final flight trajectory by iteratively completing the uncovered areas.

[0035] In one embodiment, the original obstacle is transformed into a quantifiable potential field influence through expansion processing to address the difficulty in continuously assessing spatial safety in traditional geometric obstacle avoidance. Specific steps include: constructing a neural network energy model after expanding the original obstacle; rasterizing the survey area; calculating the total energy value of each grid center point using the neural network energy model to form a three-dimensional energy field matrix; transforming obstacle constraints into a continuous energy distribution; and optimizing the global path structure by iteratively adjusting the positions of skeleton path nodes to overcome the tendency of initial straight paths or fixed grid paths to fall into local optima. Specific steps include: generating an initial skeleton path node sequence based on the survey starting point coordinates, survey ending point coordinates, and a preset number of segments in the survey area; and iteratively optimizing the internal node positions in the initial skeleton path node sequence to obtain an optimized skeleton path node sequence, achieving a balance between obstacle avoidance effectiveness and path smoothness at the global level. As an intermediate link connecting global planning and local execution, the specific steps include: encrypting the optimized skeleton path node sequence according to the preset waypoint spacing to generate a dense waypoint sequence, providing a continuous and appropriately dense target point sequence for subsequent local tracking; and constructing a local control framework that conforms to physical constraints and introducing adaptive weights and a coverage closed-loop mechanism to solve the problems of traditional dynamic window methods where fixed weights are difficult to adapt to dynamic environmental changes and open-loop planning easily missing coverage areas. Specific steps include: constructing a 3D UAV motion model based on rigid body kinematics principles; constructing a feasible velocity space and generating a candidate velocity combination set based on UAV kinematic constraints; using a dynamic window method, dynamically adjusting weights through a fuzzy controller, selecting the optimal control command from the candidate velocity combination set, performing local target point tracking through the optimal control command, and generating the final flight trajectory by iteratively completing the uncovered areas, thereby achieving accurate tracking under kinematic constraints and improving environmental adaptability, and ensuring the integrity of the mapping task through coverage feedback.

[0036] This invention transforms obstacle constraints into a continuous energy field through a neural network energy model, enabling path optimization to finely balance safety and smoothness. By iteratively optimizing the position of skeleton nodes, it overcomes the local optima problem of traditional fixed grid paths, improving the obstacle avoidance quality of the global path. Furthermore, it employs a fuzzy controller to dynamically adjust the evaluation weights of yaw angle deviation, pitch angle deviation, obstacle distance, and velocity terms, making the UAV prioritize directional accuracy when approaching the target point and obstacle avoidance safety when approaching obstacles. This significantly enhances the robustness of trajectory tracking in complex environments. Finally, through an iterative completion mechanism for uncovered areas, it achieves a fully covered mapping trajectory with high integrity, high smoothness, and kinematic constraints.

[0037] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0038] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the protection scope of the present invention.

Claims

1. A method for full-coverage UAV mapping path planning, characterized in that, Includes the following steps: Step S1: After expanding the original obstacle, construct a neural network energy model, and rasterize the survey area. Calculate the total energy value of the center point of each grid using the neural network energy model to form a three-dimensional energy field matrix. Step S2: Based on the coordinates of the surveying start point and the surveying end point of the surveying area and the preset number of segments, an initial skeleton path node sequence is generated. Through iterative optimization, the positions of the internal nodes in the initial skeleton path node sequence are adjusted to obtain the optimized skeleton path node sequence. Step S3: Encrypt the optimized skeleton path node sequence according to the preset waypoint spacing to generate a dense waypoint sequence; Step S4: Based on the principle of rigid body kinematics, a three-dimensional UAV motion model is constructed. Based on the UAV kinematic constraints, a feasible velocity space is constructed and a candidate velocity combination set is generated. Using the dynamic window method, the weights are dynamically adjusted through a fuzzy controller. The optimal control command is selected from the candidate velocity combination set. The optimal control command is used to track local target points. The final flight trajectory is generated by iteratively completing the uncovered areas.

2. The method for full-coverage UAV mapping path planning as described in claim 1, characterized in that, Step S1 includes steps S101, S102, S103 and S104; Step S101: Obtain the geometric parameters of the original obstacle; The original obstacles include polyhedral obstacles and spherical obstacles; The geometric parameters include the original constraint surfaces of the polyhedral obstacle and the original center coordinates and original radius of the spherical obstacle; Each original obstacle is expanded to obtain the equations of the expanded constraint surface and the expanded sphere. The dilation process specifically includes dilating the original constraint surfaces of the polyhedral obstacle to obtain the corresponding dilated constraint surface equations. The expression for the dilated constraint surface equations is as follows: ； in, This represents the normal coefficient of the constrained surface after expansion. This represents the threshold parameter after dilation. Represents the coordinates of a point in space; Based on the original center and radius of the spherical obstacle, the spherical obstacle is expanded to obtain the equation of the expanded sphere. The expression of the equation of the expanded sphere is as follows: ； in, This represents the coordinates of the center of the spherical obstacle after it expands. This represents the radius of the spherical obstacle after it expands; Step S102: Obtain the coordinates of the spatial point to be calculated, and based on the center coordinates and radius of the sphere corresponding to the equation of the expanded sphere, as well as the normal coefficient and threshold parameter corresponding to the equation of the expanded constraint surface of each polyhedral obstacle, construct a neural network energy model to obtain the total energy value of each spatial point to be calculated. The construction of the neural network energy model specifically includes: constructing a first layer of hidden neurons based on the coordinates of the center of the sphere after expansion and the radius after expansion; and using the function of the first layer of neurons to calculate the distance-related values ​​of the spherical obstacle. The expression of the function of the first layer of neurons is as follows: ； in, This represents the distance correlation value between the j-th spherical obstacle and the i-th spatial point to be calculated. These represent the X, Y, and Z axes of the i-th spatial point to be calculated in the ground coordinate system, respectively. Let represent the radius of the j-th spherical obstacle after expansion. This represents the square of the radius of the j-th spherical obstacle after it expands. These represent the coordinate components of the center of the j-th spherical obstacle in the X, Y, and Z directions of the ground coordinate system after expansion; Each polyhedral obstacle is formed by multiple constraint surfaces, and each constraint surface corresponds to an expanded constraint surface equation. For the k-th constraint surface, calculate the signed distance from the spatial point to be calculated to the constraint surface, expressed as: ； in, This represents the signed distance from the spatial point to be calculated to the k-th constraint surface. This represents the normal coefficient of the k-th constraint surface. This represents the dilation threshold parameter corresponding to the k-th constraint surface; The maximum value among the signed distances of all constrained surfaces is taken as the distance correlation value of the polyhedral obstacle, denoted as . .

3. The method for full-coverage UAV mapping path planning as described in claim 2, characterized in that, Step S103: Input the distance correlation value of each original obstacle into the second hidden layer neuron, and use the Sigmoid function to convert the distance correlation value into the corresponding energy value. The function expression of the second layer neuron is: ； in, This represents the energy value of the j-th original obstacle. This represents the distance correlation value of the j-th original obstacle. If the original obstacle is a spherical obstacle, then... The value is the distance-related value of the spherical obstacle. ; If the original obstacle is a polyhedral obstacle, then The value is the distance-related value of the polyhedral obstacle. ; This represents the preset baseline threshold for the j-th original obstacle. Indicates the smoothing parameter; The energy values ​​are input to the output layer, and a summation function is used to calculate the sum of the energy values ​​of all original obstacles, thus obtaining the total energy value of the spatial point to be calculated. The expression for the summation function is: ； in, This represents the total energy value of the i-th spatial point to be calculated. This indicates the total number of all original obstacles; Step S104: Obtain the three-dimensional boundary and preset resolution of the survey area; perform rasterization processing on the survey area according to the three-dimensional boundary and resolution to obtain the coordinates of the center points of all grids. The coordinates of the center point of each grid are used as spatial points to be calculated, and the total energy value of each grid center point is obtained by using the neural network energy model. The total energy values ​​of all grid center points are arranged according to their spatial positions to form a three-dimensional energy field matrix.

4. The method for full-coverage UAV mapping path planning as described in claim 3, characterized in that, Step S2 includes steps S201, S202, S203 and S204; Step S201: Obtain the coordinates of the starting point and ending point of the surveying area, as well as the preset number of segments. Generate an initial coverage path based on the starting point, ending point, and preset number of segments to obtain the initial skeleton path node sequence. The process of generating the initial skeleton path node sequence involves dividing the straight path between the start point and the end point into m equal segments, resulting in a total of m+1 nodes. Each m+1 node includes the start point, the end point, and m-1 intermediate nodes. The expression for calculating the node position is: ； ； in, This represents the initial position of the i-th node. Indicates the coordinates of the starting point of the surveying area. This represents the coordinates of the endpoint of the survey area. When i=0, When i=m, m represents the number of segments; Step S202: In the t-th iteration, for each internal node in the initial skeleton path node sequence, candidate nodes are generated in the three-dimensional space along six orthogonal directions with a preset search step size. The step size is moved along the positive and negative X-axis, positive and negative Y-axis, and positive and negative Z-axis respectively to obtain six candidate nodes. The preset search step size is denoted as: ; The logic for generating the coordinates of the six candidate nodes is as follows: when moving along the positive X-axis, the coordinates of the candidate nodes are... ; When moving along the negative X-axis, the coordinates of the candidate node are: ; When moving along the positive Y-axis, the coordinates of the candidate node are: ; When moving along the negative Y-axis, the coordinates of the candidate node are: ; When moving along the positive Z-axis, the coordinates of the candidate node are: ; When moving along the negative Z-axis, the coordinates of the candidate node are: ; in, These represent the coordinate components of the i-th internal node in the initial skeleton path node sequence in the X, Y, and Z directions in the ground coordinate system during the t-th iteration. Step S203: For each candidate node, replace the corresponding internal node in the initial skeleton path node sequence with the candidate node to obtain the replaced candidate node. Combine the replaced candidate node with the start point, end point and other internal nodes in the initial skeleton path node sequence that did not participate in the replacement to form a candidate path. The total energy value of each candidate path is calculated by weighting the sum of the energy values ​​of all path nodes constituting the candidate path with the sum of the path lengths, as expressed in the following expression: ； in, This represents the total energy value of the candidate path formed after the i-th internal node is replaced along the j-th candidate direction. This represents the r-th path node in the candidate paths. Represents the energy field matrix. Path node The corresponding energy value in the energy field matrix and This represents the preset weighting coefficients. Denotes the Euclidean norm; For each internal node, select the minimum total energy value from the six candidate paths, and take the candidate node with the minimum total energy value as the new position of the internal node. Step S204: Repeat steps S202 and S203 for iteration, and determine whether the iteration termination condition is met after each iteration. The iteration termination condition is as follows: if the maximum distance between the new position of all internal nodes and the position in the previous iteration is less than a preset threshold, or the current iteration number has reached the preset maximum iteration number, then the iteration stops and the current skeleton path node sequence is output as the optimized skeleton path node sequence. Otherwise, proceed to the next iteration.

5. The method for full-coverage UAV mapping path planning as described in claim 4, characterized in that, Step S3 specifically includes obtaining the optimized skeleton path node sequence, using the B-spline interpolation method to encrypt the paths between adjacent nodes in the optimized skeleton path node sequence, and generating a dense waypoint sequence according to the preset waypoint spacing.

6. The method for full-coverage UAV mapping path planning as described in claim 5, characterized in that, Step S4 includes steps S401, S402, S403, S404, S405 and S406. Step S401: Obtain the current state of the UAV, which includes the UAV's position coordinates, attitude angles, body speed, body angular velocity, and time interval. Based on the principles of rigid body kinematics, a three-dimensional UAV motion model is constructed, which consists of position update equations and attitude update equations. The expression for the position update equation is: ； ； in, Indicates time interval, This represents the rotation matrix from the body coordinate system to the ground coordinate system, using the ZYX Euler angle order. Indicates the yaw angle. Indicates pitch angle, Indicates the roll angle. s represents cos, and s represents sin. , , These represent the position coordinate components of the UAV in the X, Y, and Z axes of the ground coordinate system, respectively. , , These represent the linear velocity components of the UAV in the X, Y, and Z axes of the body coordinate system, respectively. The attitude update equation is expressed as follows: ； Wherein, the Euler angle vector is denoted as Euler angular rate is denoted as The angular velocity of the machine body is denoted as ,in, Represents Euler angle vectors. This represents the rate of change of the drone's attitude angle over time. , representing the angular velocity components of the UAV's rotation around each axis of the body coordinate system; ； Among them, Euler angular rate With body angular velocity The relationship is ; These represent the roll angle, pitch angle, and yaw angle at the current moment, respectively. These represent the angular velocity components of the UAV in the X, Y, and Z axes of the body coordinate system, respectively. express The inverse matrix; The position at the next moment is obtained based on the position update equation and the attitude update equation. and the attitude angle corresponding to the next moment .

7. The method for full-coverage UAV mapping path planning as described in claim 6, characterized in that, Step S402: Use the dynamic window method to obtain the current body velocity, body angular velocity, preset speed limit, acceleration limit and real-time obstacle information; Real-time obstacle information includes the location and distance of obstacles detected in real time by airborne sensors; A feasible velocity space is constructed based on the kinematic constraints of the UAV. The feasible velocity space is the intersection of the velocity limit constraint, the acceleration limit constraint, and the braking distance constraint. The feasible velocity space is denoted as ; The expression for the speed limit constraint is: ； in, Indicates the speed limit constraint. Indicates the candidate linear velocity. Indicates the candidate angular velocity, These represent the preset minimum and maximum linear velocities, respectively. These represent the preset minimum and maximum angular velocities, respectively; The expression for the speed limit constraint is: ； in, This indicates a speed limit constraint. These represent the candidate linear velocity and candidate angular velocity after considering acceleration constraints, respectively. and These represent the current velocity and angular velocity of the machine, respectively. and These represent the preset maximum linear acceleration and maximum angular acceleration, respectively. and These represent the preset maximum linear deceleration and maximum angular deceleration, respectively. Indicates a time interval; The expression for the braking distance constraint is: ； in, Indicates braking distance constraint. This represents the candidate speed after considering braking distance constraints. Indicates the current candidate velocity The minimum distance between the predicted trajectory and the obstacle; In feasible velocity space The system performs uniform sampling at a preset resolution to generate a set of candidate velocity combinations.

8. The method for full-coverage UAV mapping path planning as described in claim 7, characterized in that, Step S403: Initialize the current local target point as the first waypoint in the dense waypoint sequence using the dynamic window method; Obtain the distance between the current drone position and the current local target point, and record it as the first distance; obtain the distance between the drone and the nearest obstacle, and record it as the second distance; The first and second distances are input into a preset two-input four-output fuzzy controller for fuzzing to obtain the weight coefficients of the evaluation function; The weighting coefficients correspond to the weights of the yaw angle deviation term, pitch angle deviation term, obstacle distance term, and velocity term of the evaluation function, respectively. Step S404: For each candidate velocity combination in the candidate velocity combination set, predict the motion trajectory under the candidate velocity combination using the three-dimensional UAV motion model, and calculate the evaluation function value, the expression of which is: ； in, This represents the evaluation function value of the candidate velocity combination. and These represent the yaw angle deviation and pitch angle deviation between the UAV's current heading and the direction of the local target point, respectively. This represents the minimum distance between the predicted trajectory and the obstacle. Indicates the magnitude of the resultant velocity; Select the optimal control command from all candidate velocity combinations, choosing the one with the largest evaluation function value, and output it. Step S405: Calculate the first distance in real time. If the first distance is less than the preset distance threshold, select the next waypoint from the dense waypoint sequence as the new local target point and output the updated local target point.

9. The method for full-coverage UAV mapping path planning as described in claim 8, characterized in that, Step S406: Obtain the effective coverage width of the mapping sensor. The effective coverage width is calculated based on the current flight altitude of the UAV and the sensor's field of view. The calculation process is as follows: calculate the product of the current flight altitude of the UAV and 2, which is recorded as the first product; calculate the tangent of half of it multiplied by the sensor's field of view, which is recorded as the half-angle tangent; calculate the product of the first product and the half-angle tangent, which is recorded as the second product. The second product is the effective coverage width. Discretize the survey area into grid cells and initialize all grid cells as uncovered. Traverse all adjacent waypoint pairs in the dense waypoint sequence of right step S3. For each pair of adjacent waypoints, calculate the projection area of ​​the sensor's scan band on the ground when the UAV flies along the flight segment between the pair of adjacent waypoints, and mark the grid cells covered by the projection area as covered. Check the coverage status of all grid cells, including uncovered and covered; The inspection process includes inserting new waypoints near the center of the uncovered area if there are uncovered grid cells, and generating smooth path segments between the new waypoints and adjacent waypoints using B-spline interpolation. Step S406 is repeated until all grid cells are marked as covered, and the final dense waypoint sequence is output as the flight trajectory of the UAV mapping.

10. A full-coverage UAV mapping path planning system, applied to a full-coverage UAV mapping path planning method as described in any one of claims 1-9, characterized in that, It includes a build module, an optimization module, an encryption module, and a generation module; The module constructs a neural network energy model after inflating the original obstacles, and then rasterizes the survey area. The neural network energy model is used to calculate the total energy value of the center point of each grid, forming a three-dimensional energy field matrix. The optimization module generates an initial skeleton path node sequence based on the coordinates of the surveying start point and the surveying end point of the surveying area and the preset number of segments. Through iterative optimization, the positions of the internal nodes in the initial skeleton path node sequence are adjusted to obtain the optimized skeleton path node sequence. The encryption module encrypts the optimized skeleton path node sequence according to the preset waypoint spacing to generate a dense waypoint sequence. The generation module constructs a 3D UAV motion model based on the principle of rigid body kinematics, builds a feasible velocity space based on UAV kinematic constraints and generates a set of candidate velocity combinations. It adopts the dynamic window method, dynamically adjusts the weights through a fuzzy controller, selects the optimal control command from the set of candidate velocity combinations, tracks local target points through the optimal control command, and generates the final flight trajectory by iteratively completing the uncovered areas.

Images