Unmanned aerial vehicle path planning method and system for building curtain wall cleaning operation

Abstract

The present application belongs to the technical field of unmanned aerial vehicle path planning, in particular to an unmanned aerial vehicle path planning method and system for building curtain wall cleaning operation, comprising: collecting three-dimensional point cloud data of building facade and constructing a three-dimensional grid map, generating a cleaning operation surface by offsetting a safe distance outward along the normal vector with the wall surface as the reference, constraining the path points within the operation surface range, and realizing dimension reduction constraint of three-dimensional search space to two-dimensional surface. In the two-dimensional grid space, an improved A* algorithm is executed using a multi-constraint cost function that integrates safety proximity, attitude smoothness, energy consumption and coverage continuity to obtain a discrete optimal path; the path is smoothed to generate a continuous trajectory executable by the flight control, and local dynamic re-planning is performed for sudden obstacles during the operation. The present application does not require manual preset trajectory, can adapt to complex building facade shape, and synchronously ensures safety, stability, full coverage, low energy consumption and real-time performance of the cleaning operation.

Description

Technical Field

[0001] This invention belongs to the field of drone path planning technology, and in particular refers to a drone path planning method and system for building curtain wall cleaning operations. Background Technology

[0002] With the widespread use of glass curtain walls in high-rise buildings, regular curtain wall cleaning has become a necessary maintenance task. Traditional methods are high-risk and costly. Drone-based aerial cleaning technology offers high safety and flexibility, representing an important development direction for intelligent curtain wall maintenance. The core technical challenge lies in automatically planning safe, efficient, and reliable flight paths for drones on complex three-dimensional building facades.

[0003] Existing drone-based curtain wall cleaning path planning technologies have significant shortcomings: most commercial and research solutions employ manual teaching and preset waypoints, requiring repeated manual operations for different buildings. They cannot automatically adapt to complex facade geometry features such as protruding balconies, recessed windows, and curved corners, resulting in low levels of automation and path optimization. Some studies directly use general path planning algorithms such as A*, Dijkstra, and RRT, which have significant limitations in 3D facade scenarios: firstly, the large number of nodes searched in 3D space easily leads to the curse of dimensionality, resulting in low planning efficiency; secondly, focusing solely on the shortest path fails to consider the unique constraints of high-altitude cleaning operations, making it impossible to guarantee a constant safe operating distance between the drone and the wall, and the lack of a nozzle coverage model can easily lead to missed or repeated cleaning. Furthermore, the failure to consider the drone's dynamic stability results in flight swaying and poor cleaning effects, and the failure to differentiate between vertical and horizontal movement energy consumption leads to unreasonable energy consumption.

[0004] Patent CN119837453A discloses a building curtain wall cleaning system. While this solution can achieve facade scanning, area division, and task scheduling, its flight path still relies on preset trajectories or simple rule generation. It does not construct a spatial model adapted to operational requirements, nor does it employ a multi-objective cost function to collaboratively optimize safety, coverage, stability, and energy consumption. When faced with complex, irregular facades or sudden obstacles, it struggles to automatically generate optimized paths that meet engineering needs.

[0005] In summary, existing technologies lack a dedicated path planning method that fits the constraints of high-altitude cleaning operations and can efficiently and automatically achieve multi-objective optimization on complex three-dimensional surfaces, thus failing to simultaneously meet the comprehensive requirements of safety, full coverage, stable flight, and low energy consumption. Summary of the Invention

[0006] Therefore, the technical problem to be solved by the present invention is: addressing the problems that existing building facade cleaning drone path planning is difficult to adapt to complex building shapes, cannot simultaneously and quantitatively meet multiple constraints such as safe operation, full coverage cleaning, stable attitude and low energy consumption, and has low three-dimensional space search efficiency and difficulty in meeting real-time requirements, the present invention provides a drone path planning method and system for building curtain wall cleaning operations, so as to realize safe, efficient, comprehensive and economical automated cleaning operations for building facades.

[0007] Specifically, the UAV path planning method for building curtain wall cleaning operations includes the following steps:

[0008] Step S1: Acquire 3D point cloud data of the building facade, discretize it into a 3D grid map and mark wall obstacles; take the wall surface as the reference plane, offset outward by a preset safety distance along the normal vector direction of the wall surface to generate a cleaning operation surface that is consistent with and parallel to the geometry of the wall surface, and constrain all path points of the UAV path planning to the cleaning operation surface or its adjacent allowable deviation range.

[0009] Step S2: Based on the cleaning operation surface, construct a corresponding two-dimensional grid search space; within the grid search space, use the reconstructed multi-constraint cost function to execute the improved A* path search algorithm, solve for the optimal path from the preset starting point to the target endpoint, and obtain the discrete grid path; wherein, the multi-constraint cost function includes constraint cost terms for safety proximity, attitude smoothness, energy consumption, and coverage continuity;

[0010] Step S3: Perform trajectory smoothing on the discrete grid path to obtain a continuous executable trajectory suitable for execution by the UAV flight control system; when the UAV performs facade cleaning operation along the continuous smooth trajectory, if a sudden obstacle not included in the original plan is identified, perform local dynamic replanning with the current real-time position of the UAV as a new starting point.

[0011] In one embodiment of the present invention, in step S2, the expression of the multi-constraint cost function is:

[0012] ,

[0013] in, The actual cost from the starting point to the current node n. Let n be the heuristic cost from the current node n to the destination. For the sake of safety and close proximity, For the sake of pose smoothness, For the cost of energy consumption, To cover the cost of continuity; , , , These are the weight coefficients for the corresponding cost terms.

[0014] In one embodiment of the invention, the security proximity cost The calculation method is as follows:

[0015] ,

[0016] in, Let n be the distance from the current node n to the actual wall. To preset the optimal cleaning distance.

[0017] In one embodiment of the present invention, the attitude smoothness cost The calculation method is as follows:

[0018] ,

[0019] in, It represents the minimum change in attitude angle from the previous node p to the current node n.

[0020] In one embodiment of the present invention, the energy consumption cost The calculation method is as follows:

[0021] ,

[0022] in, Energy consumption coefficient This represents the change in altitude between the start and end points of the current flight segment.

[0023] In one embodiment of the invention, the coverage continuity cost The calculation method is as follows:

[0024] Based on the effective working radius R of the cleaning nozzle, the cleaning area of ​​the building facade is marked by a grid, and the spatial status of the covered and uncovered areas is maintained in real time.

[0025] If the current node n is located in an uncovered area, then ;

[0026] If the current node n falls into the covered area, then The specific calculation formula for the preset positive value penalty term is as follows:

[0027] ,

[0028] in, The penalty coefficient is a constant, with a value range of [2,5]. This refers to the building facade area that has been cleaned in the current phase of the operation. The total area of ​​the building facade to be cleaned; This is the basic cost of single-step movement in the grid space.

[0029] In one embodiment of the present invention, the method for performing trajectory smoothing processing on the discrete grid path in step S3 to obtain a continuous executable trajectory suitable for execution by the UAV flight control system is as follows:

[0030] S31: Extract all path nodes on the discrete grid path and obtain the three-dimensional spatial coordinates of each node; perform noise reduction and redundancy removal on the extracted node coordinates, remove abnormal nodes and duplicate nodes caused by grid discretization, and filter to obtain a spatially continuous effective path node sequence.

[0031] S32: Set the order of the cubic B-spline curve to 3. Based on the number of nodes and spatial distribution density of the effective path node sequence, generate a node vector using uniform node interpolation. Initialize the control vertices of the cubic B-spline curve so that the number of control vertices matches the number of nodes in the effective path node sequence, while constraining the spatial position of the control vertices to always be within the cleaning operation surface or its allowable deviation range.

[0032] S33: The preprocessed effective node sequence is used as the fitting target point. The cubic B-spline curve fitting formula is substituted into it, and the optimal solution of the control vertex is solved by the least squares method to keep the deviation between the fitted curve and the discrete nodes within the preset allowable range. During the fitting process, the constrained fitted curve is always located within the preset cleaning operation surface or its allowable deviation range.

[0033] S34: Calculate the curvature, first derivative, and second derivative of each segment of the fitted cubic B-spline curve, and verify whether the curve has inflection points, abrupt changes, or other areas that do not meet the requirements for UAV flight; if there are areas with unsmoothness, adjust the control vertex parameters and refit and iterate until the curvature of each segment of the fitted curve changes uniformly.

[0034] S35: The optimized cubic B-spline continuous curve is uniformly discretized according to the control frequency and accuracy requirements of the UAV flight control system to generate trajectory sampling points; the trajectory sampling points are format converted to finally generate a continuous executable trajectory adapted to the execution of the UAV flight control system.

[0035] In one embodiment of the present invention, in step S3, when the UAV performs facade cleaning operations along the continuous smooth trajectory and a sudden obstacle not included in the original plan is identified, the method for local dynamic replanning with the UAV's current real-time position as a new starting point is as follows:

[0036] S301: Real-time acquisition of building facade environmental information, and comparison of the real-time acquired environmental data with the pre-constructed 3D raster map and the original planned path;

[0037] S302: When an unmarked and unforeseen sudden obstacle is detected in the original planned path, a local dynamic replanning event is triggered. The current trajectory execution is immediately paused, and the current real-time position of the UAV is used as the new starting point for dynamic replanning. Through path tracing and retrieval, the nearest effective path point behind the sudden obstacle in the original planned path is determined as the temporary endpoint of this local replanning.

[0038] S303: Based on the determined spatial coordinate range of the new starting point and the temporary end point, on the pre-constructed overall cleaning operation surface, a local sub-operation surface containing the area of ​​sudden obstacles and the preset buffer range around the new starting point and the temporary end point is cut out, and only the grid data, cleaning operation constraints and safety distance requirements within the local sub-operation surface are retained.

[0039] S304: In the trimmed local sub-work surface, the improved A* path search algorithm described in step S2 is run again. The multi-constraint cost function is used to comprehensively consider safety proximity, posture smoothness, energy consumption and coverage continuity constraints. Under the premise of avoiding sudden obstacles, the optimal local path from the new starting point to the temporary end point is searched.

[0040] S305: The optimal local path generated by replanning is concatenated with the original planned path and smoothed to obtain a smoothly connected path;

[0041] S306: Based on the smooth connection path, perform cleaning operations in local areas. After passing through the obstacle area, restore to the original planned path or continue along the updated path to complete the remaining building facade cleaning operations.

[0042] In one embodiment of the present invention, when a sudden obstacle not included in the original plan is identified during the facade cleaning operation performed by the UAV along the continuous smooth trajectory, the method for local dynamic replanning with the current real-time position of the UAV as a new starting point further includes:

[0043] If the sudden obstacle is a fixed and long-term obstacle, the obstacle information can be updated to the local 3D raster map.

[0044] Based on the same inventive concept, the present invention also provides a UAV path planning system for building curtain wall cleaning operations, used to implement the UAV path planning method for building curtain wall cleaning operations, the system including: a 3D modeling and work surface generation module, a path search module, and a trajectory smoothing and dynamic replanning module;

[0045] The 3D modeling and work surface generation module is used to acquire 3D point cloud data of the building facade, discretize it into a 3D grid map and mark wall obstacles; using the wall surface as a reference plane, it offsets outward by a preset safety distance along the normal vector direction of the wall surface to generate a cleaning work surface that is consistent with and parallel to the geometric shape of the wall surface, and constrains all path points of the UAV path planning to the cleaning work surface or its adjacent allowable deviation range;

[0046] The path search module is used to construct a corresponding two-dimensional grid search space based on the cleaning operation surface; within the grid search space, an improved A* path search algorithm is executed using a reconstructed multi-constraint cost function to solve for the optimal path from the preset starting point to the target endpoint, thereby obtaining a discrete grid path; wherein, the multi-constraint cost function includes constraint cost terms for safety proximity, attitude smoothness, energy consumption, and coverage continuity.

[0047] The trajectory smoothing and dynamic replanning module is used to smooth the discrete grid path to obtain a continuous executable trajectory suitable for the UAV flight control system. When the UAV performs facade cleaning operation along the continuous smooth trajectory, if a sudden obstacle not included in the original plan is identified, local dynamic replanning is performed with the current real-time position of the UAV as a new starting point.

[0048] Compared with the prior art, the above-described technical solution of the present invention has the following advantages:

[0049] This invention reduces traditional three-dimensional spatial path planning to a two-dimensional surface search by constructing a cleaning work surface parallel to the wall surface and with a safe offset distance. This significantly compresses the search space and improves planning efficiency and real-time performance. Simultaneously, by reconstructing and integrating a multi-constraint cost function that considers safety proximity, posture smoothness, energy consumption, and coverage continuity, path planning can simultaneously and quantitatively meet the engineering requirements of safe operation, stable posture, energy consumption optimization, and full cleaning coverage. Furthermore, by combining cubic B-spline trajectory smoothing and a local dynamic replanning mechanism, it can effectively adapt to complex building facade shapes, avoid sudden obstacles, and eliminate the need for manual trajectory pre-setting. Ultimately, it achieves automated, safe, efficient, comprehensive, and low-energy cleaning of building facades, demonstrating strong scene adaptability and practical engineering value. Attached Figure Description

[0050] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0051] Figure 1 This is a flowchart illustrating a drone path planning method for building curtain wall cleaning operations provided in an embodiment of the present invention.

[0052] Figure 2 This is a schematic diagram of a specific process for a drone path planning method for building curtain wall cleaning operations provided in an embodiment of the present invention;

[0053] Figure 3 This is a schematic diagram of a drone path planning system for building curtain wall cleaning operations provided in an embodiment of the present invention.

[0054] Explanation of reference numerals in the accompanying drawings: 100, 3D modeling and work surface generation module; 200, path search module; 300, trajectory smoothing and dynamic replanning module. Detailed Implementation

[0055] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0056] Reference Figure 1 and Figure 2 As shown, this invention provides a drone path planning method for building curtain wall cleaning operations, including the following steps:

[0057] Step S1: Acquire 3D point cloud data of the building facade through LiDAR scanning or BIM model import, discretize it into a 3D raster map and mark wall obstacles; using the wall surface as the reference plane, offset outwards by a preset safety distance along the normal vector direction of the wall surface. This generates a cleaning work surface that is consistent with and parallel to the geometry of the wall surface, and constrains all path points of the UAV path planning to the cleaning work surface or its adjacent allowable deviation range, thereby realizing the dimensionality reduction mapping and constraint from the three-dimensional search space to the two-dimensional curved surface search space.

[0058] Step S2: Based on the cleaning operation surface, construct a corresponding two-dimensional grid search space; within the grid search space, use a reconstructed multi-constraint cost function to execute an improved A* path search algorithm, solve for the optimal path from the preset starting point to the target endpoint, and obtain a discrete grid path; wherein, the multi-constraint cost function includes constraint cost terms for safety proximity, attitude smoothness, energy consumption, and coverage continuity, to achieve multi-objective path optimization for building facade cleaning scenarios;

[0059] Step S3: Perform trajectory smoothing on the discrete grid path to obtain a continuous executable trajectory suitable for execution by the UAV flight control system; when the UAV performs facade cleaning operation along the continuous smooth trajectory, if a sudden obstacle not included in the original plan is identified, perform local dynamic replanning with the current real-time position of the UAV as a new starting point.

[0060] Furthermore, in step S2, for the building facade cleaning drone operation scenario, based on the cost function of the traditional A* path planning algorithm, by introducing a constraint cost term specific to the cleaning operation and configuring adjustable weight coefficients, the cost function of multiple constraints and multiple objectives is reconstructed, forming a comprehensive multi-constraint cost function adapted to building curtain wall cleaning operations. The specific reconstruction steps are as follows:

[0061] The cost function of the traditional A* algorithm is: ,in, The actual cost from the planning starting point to the current node n. This represents the heuristic cost from the current node n to the planned endpoint. This invention retains this basic cost structure as a fundamental term in the multi-constraint cost function, ensuring the reachability and optimality of the algorithm's path search.

[0062] To ensure that drones maintain a safe and optimal operating distance from building walls during cleaning operations, a safe proximity cost is established. ,as follows:

[0063] Using the pre-generated cleaning surface as a reference, calculate the actual distance from the current node n to the actual building wall. Set the optimal cleaning distance This distance is slightly larger than the preset safety offset distance. ;

[0064] The absolute difference between the actual distance from the current node to the wall and the optimal working distance is used as the safety proximity cost, and its expression is: The value of this cost increases as the deviation of the drone from the optimal operating distance increases. The cost penalty mechanism constrains the drone to always fly close to the cleaning surface, avoiding the risk of collision due to being too close and the impact on the cleaning effect due to being too far away.

[0065] To ensure the stability of the UAV's flight attitude and suppress sudden attitude changes and sharp turns during flight, an attitude smoothness cost is constructed. ,as follows:

[0066] Record the change in flight direction of the UAV from the previous node p to the current node n, and calculate the minimum attitude angle change required to achieve this flight segment. ; attitude angle change The absolute value of is used as the pose smoothness cost, and its expression is: This cost term quantifies the penalty for large-angle attitude adjustments, ensuring that the planned path remains smooth in complex areas such as building facade corners and concave-convex structures, thereby improving the flight stability of the drone and the uniformity of the cleaning effect.

[0067] To optimize the energy consumption of drone operations, suppress meaningless vertical take-off and landing movements, and construct an energy cost-benefit analysis system... ,as follows:

[0068] Extract the altitude information of the start and end points of the current flight segment and calculate the altitude change. A preset energy consumption coefficient k related to the drone's gravity-based work characteristics is used, and the product of the altitude change and the energy consumption coefficient is taken as the energy consumption cost, expressed as:

[0069] This cost item quantifies the energy consumption cost of vertical take-off and landing, prioritizes encouraging horizontal movement of drones, optimizes energy consumption for path planning, and extends the endurance of a single cleaning operation.

[0070] To ensure full coverage of the building curtain wall cleaning area and reduce repeated cleaning, a coverage continuity cost is constructed based on the effective operating radius R of the cleaning nozzles. ,as follows:

[0071] Based on the effective operating radius R of the cleaning nozzles, the cleaning area of ​​the building facade is rasterized and marked. During the path search process, the spatial status of covered and uncovered areas is maintained in real time. For the spatial position of the current node n, the coverage continuity cost is defined in two cases:

[0072] If the current node n is located in an uncovered area, then No penalty;

[0073] If the current node n falls into the covered area, then , The specific calculation formula for the preset positive value penalty term is as follows:

[0074] ,

[0075] in, The penalty coefficient is a constant, with a value range of [2,5]. This refers to the building facade area that has been cleaned in the current phase of the operation. The total area of ​​the building facade to be cleaned; The base cost for a single step movement in the grid space (with a base cost of 1 for each horizontal or vertical movement of one grid cell, and a base cost of 1 for each diagonal grid cell movement). ).

[0076] The value of this penalty term increases monotonically with the cleaning coverage ratio. When the building facade cleaning coverage ratio reaches 50% or more, the calculated value of the penalty term is greater than 1.5 times the basic cost of single-step grid movement. This quantitative penalty mechanism forms a hard constraint on the direction of algorithm node expansion, ensuring that the improved A* algorithm prioritizes expanding nodes to uncovered cleaning areas during path search, avoiding repeated cleaning behavior from the cost optimization perspective, and ensuring the full coverage and continuity of cleaning operations.

[0077] The four constraints—safety proximity, attitude smoothness, energy consumption, and coverage continuity—are weighted and fused with the basic cost of the traditional A* algorithm to generate a comprehensive multi-constraint cost function adapted to building curtain wall cleaning operations. The expression is as follows:

[0078] ,

[0079] in, For the safety proximity weighting coefficient, This is the attitude smoothness weighting coefficient. Energy consumption weighting coefficient To cover continuity weighting coefficients, each weighting coefficient can be adaptively adjusted according to actual cleaning operation needs (such as emphasizing operational safety, flight stability, energy consumption control, or cleaning coverage quality), achieving multi-objective collaborative optimization of path planning.

[0080] Optionally, to avoid search imbalance caused by excessive differences in the values ​​of constraint cost terms with different dimensions, each constraint cost term can be normalized to ensure that all costs are on the same order of magnitude. At the same time, it supports dynamic adjustment of weight coefficients based on working conditions such as building facade complexity, wind speed in the working environment, and drone battery power, to ensure that the cost function has stable and reliable path guidance capabilities in different working scenarios.

[0081] Furthermore, in step S3, the method for smoothing the discrete grid path to obtain a continuous executable trajectory suitable for execution by the UAV flight control system is as follows:

[0082] S31: Extract all path nodes on the discrete grid path and obtain the three-dimensional spatial coordinates of each node. This forms the original path node coordinate sequence. ,in For the first The three-dimensional coordinates of each node, This represents the original total number of nodes;

[0083] The original node coordinate sequence is sequentially subjected to denoising and redundancy removal processes, including:

[0084] The neighborhood mean filtering method is used to calculate the average coordinates of the three neighboring nodes of each node. If the deviation between the current node's coordinates and the neighborhood mean exceeds the preset grid resolution threshold (1 / 2 of the grid resolution in this invention, i.e., 0.1m), it is determined to be an abnormal node generated by grid discretization, and its coordinates are replaced with the neighborhood mean.

[0085] Traverse the denoised node sequence. If the three-dimensional coordinate deviation of two or more consecutive nodes is less than the preset allowable deviation (0.05m), it is determined to be a duplicate node. Only the first node is kept and subsequent redundant nodes are removed.

[0086] After denoising and redundancy removal, a sequence of valid path nodes with continuous spatial distribution, no anomalies, and no redundancy was obtained. ,in , Let be the three-dimensional coordinates of the j-th valid node.

[0087] S32: To meet the dynamic requirements of UAV flight control systems for trajectory smoothness and curvature continuity, the order of the cubic B-spline curve is set. The curve basis function adopts the standard cubic B-spline basis function; based on the effective path node sequence Number of nodes To initialize the spatial distribution density, perform the following initialization operations:

[0088] The node vectors of cubic B-spline curves are generated using uniform node interpolation. The node vectors take values ​​in the range [0,1] and satisfy the uniform distribution property. This ensures the spatial uniformity of the curve fitting.

[0089] Initialize the control vertices of the cubic B-spline curve The number of control vertices is related to the sequence of valid path nodes. Number of nodes Strict matching ensures that the fitted curve accurately matches the spatial distribution trend of the effective nodes; simultaneously, each control vertex is verified one by one. The three-dimensional coordinates are used to ensure that the control vertex is always within the allowable deviation range (±0.1m) of the cleaning work surface or its vicinity. If a control vertex exceeds this range, its position is corrected using the normal projection coordinates of the cleaning work surface at that location, thus achieving the work surface spatial constraint of the fitting curve from the source.

[0090] S33: The preprocessed valid node sequence Substituting the target point into the cubic B-spline curve fitting formula:

[0091] ,in For any point in space on the fitted curve, Let t be the k-th order B-spline basis function. ;

[0092] With the goal of minimizing the spatial deviation between the fitted curve and the effective nodes, a least-squares optimization objective function is constructed:

[0093] ,in Valid node The corresponding node vector parameters, It is a Euclidean distance.

[0094] The objective function is solved using numerical methods, and the control vertex sequence is calculated iteratively. The optimal solution is obtained until the spatial deviation between the fitted curve and all effective path nodes is controlled within the preset allowable deviation range, thus meeting the trajectory accuracy requirements of UAV cleaning operations.

[0095] During the fitting process, the spatial position of the fitted curve is continuously verified. If a certain segment of the curve exceeds the allowable deviation range of the cleaning operation surface, the solution is paused, the coordinates of the control vertex of the corresponding area are finely adjusted, and the iteration is restarted to ensure that the fitted curve is constrained to the cleaning operation surface or its adjacent allowable deviation range throughout the entire process.

[0096] S34: Perform geometric feature calculations on the obtained cubic B-spline fitting curve: uniformly take a number of sampling points along the curve (the number of sampling points is 5 to 10 times the number of effective nodes), and calculate the curvature of the curve, the first derivative used to reflect the velocity of the trajectory, and the second derivative used to reflect the acceleration of the trajectory at each sampling point.

[0097] Based on the calculated geometric features, a smoothness threshold is set. If the fitted curve exhibits any of the following conditions, it is determined to be a region where smoothness does not meet the standard:

[0098] Scenario 1: The rate of change of curvature exceeds the preset threshold (5° / m), or the maximum curvature exceeds the maximum allowable curvature of the UAV flight control system (15° / m in this invention), which is a curvature abrupt change segment;

[0099] Scenario 2: If the abrupt change in the first derivative exceeds the preset abrupt change range (2m / s), or the absolute value of the second derivative exceeds 1m / s², it is determined to be a velocity / acceleration abrupt change segment;

[0100] Scenario 3: The fitted curve has non-smooth geometric features such as inflection points and cusps.

[0101] For regions where smoothness is not up to standard, adjust the control vertex parameters: locate the control vertex corresponding to the region where smoothness is not up to standard, and correct its three-dimensional coordinates by fine-tuning with small steps (fine-tuning step size is 0.02~0.05m). The fine-tuning direction is to reduce the curvature of the curve and suppress sudden changes in velocity / acceleration. Substitute the adjusted control vertex sequence back into the fitting formula in step S33, and perform least squares fitting and deviation verification again until the curvature, velocity, and acceleration changes of each segment of the fitted curve remain uniform, without any sudden changes, and fully meet the dynamic smoothness requirements of UAV flight.

[0102] S35: The iteratively optimized cubic B-spline continuous curve is uniformly discretized and converted into a format recognizable by the flight control system to generate a continuous executable trajectory. The specific steps are as follows:

[0103] Based on the control frequency of the UAV flight control system (in this invention, the universal control frequency of industrial-grade UAVs is 10Hz, and the trajectory control accuracy is 0.05m), equally spaced trajectory sampling points are generated along a continuous curve. The spacing between the sampling points satisfies the following conditions: ,in The maximum flight speed for drone cleaning operations is 0.5 m / s in this invention. To set the flight control frequency, ensuring that the flight control system can accurately track the trajectory;

[0104] Extract the three-dimensional spatial coordinates of all trajectory sampling points to form a trajectory sampling point sequence. ,in Let be the three-dimensional coordinates of the p-th sampling point, and k be the total number of sampling points;

[0105] According to the communication protocol and data format requirements of the target UAV flight control system, the three-dimensional coordinates of the sampling points are converted into a digital signal format that the flight control can recognize. At the same time, flight attitude commands (generated based on the tangential direction of the curve at that point) and flight speed commands (generated based on the first derivative of the curve) are added to each sampling point.

[0106] The integrity of the converted trajectory data is verified to ensure that the sampling point sequence is continuous without any breaks and that the command information matches the trajectory coordinates correctly. Finally, a continuous executable trajectory adapted to the UAV flight control system is generated, which can be directly sent to the flight control system to drive the UAV to complete the cleaning operation.

[0107] The aforementioned trajectory smoothing method revolves around the safety, accuracy, and smoothness of UAV cleaning operations. Through multi-stage preprocessing, fitting, verification, and optimization, it ensures the consistency between the continuous trajectory and the original discrete optimal path, meets the dynamic execution requirements of the flight control system, and always constrains the trajectory within the cleaning operation area, effectively avoiding the risk of collision between the UAV and the building facade.

[0108] Furthermore, in step S3, during the facade cleaning operation performed by the UAV along the continuous smooth trajectory, when a sudden obstacle not included in the original plan is identified, the method for local dynamic replanning using the UAV's current real-time position as a new starting point is as follows:

[0109] S301: During the process of the UAV performing building facade cleaning operation along a continuous smooth trajectory, the UAV collects real-time environmental information of the building facade through airborne lidar, visual sensors and ultrasonic sensors. After noise reduction and filtering preprocessing of the collected environmental data, feature comparison and consistency verification are performed with the pre-constructed high-precision three-dimensional facade grid map and the original planned path data.

[0110] S302: When an unmarked or unforeseen sudden obstacle (including but not limited to open windows, temporary components protruding from the wall, and hanging objects) is detected in the original planned path, a local dynamic replanning event is triggered. The current trajectory execution is immediately paused, and the current real-time position of the UAV is used as the new starting point for dynamic replanning. Through path tracing and retrieval, the nearest effective path point behind the sudden obstacle in the original planned path is determined as the temporary endpoint of this local replanning, ensuring that the replanned path can smoothly return to the original operation trajectory and ensuring the continuity of the cleaning operation.

[0111] S303: Based on the determined spatial coordinate range of the new starting point and temporary endpoint, a local sub-operation surface containing the sudden obstacle area and the preset buffer range around the new starting point and temporary endpoint is cut out on the pre-constructed overall cleaning operation surface. Only the raster data, cleaning operation constraints and safety distance requirements within the local sub-operation surface are retained, thereby further compressing the search space and improving the efficiency of replanning.

[0112] S304: In the trimmed local sub-work surface, the improved A* path search algorithm described in step S2 is run again. The multi-constraint cost function is used to comprehensively consider safety proximity, attitude smoothness, energy consumption and coverage continuity constraints. Under the premise of ensuring that the UAV maintains a safe working distance from the wall and avoiding sudden obstacles, the optimal local path from the new starting point to the temporary end point is searched, taking into account both path optimality and work adaptability.

[0113] S305: The optimal local path generated by replanning is spliced ​​with the original planned path. The cubic B-spline smoothing algorithm is used to smooth the path connection, eliminating abrupt changes in direction and attitude at the connection, and generating a smooth connection path to ensure stable UAV flight attitude and uniform cleaning effect.

[0114] S306: The smoothly connected path is sent to the UAV flight control system to control the UAV to perform local area cleaning operations along the path; when the UAV passes through the sudden obstacle area and the flight status becomes stable, the path matching verification is used to restore the original planned path to continue the operation, or the remaining building facade cleaning operation is completed along the updated path to ensure that the cleaning operation is complete and uninterrupted.

[0115] Optionally, during the facade cleaning operation performed by the UAV along the continuous smooth trajectory, when a sudden obstacle not included in the original plan is identified, the method of performing local dynamic replanning with the UAV's current real-time position as a new starting point further includes:

[0116] If the obstacle is a fixed and long-term obstacle, the obstacle information can be updated to the local 3D grid map for path planning during subsequent operations in the same area, avoiding repeated replanning.

[0117] In summary, this invention, through 3D modeling and adaptive planning, can effectively adapt to complex and irregular building facade shapes. By leveraging safety distances and attitude constraints, it significantly reduces the risk of collisions during operations, thereby improving cleaning safety. Path planning based on the nozzle coverage model effectively avoids repeated cleaning and cleaning dead zones, greatly improving cleaning efficiency and coverage integrity. By introducing energy consumption cost optimization, it reduces ineffective maneuvering flights, lowers overall energy consumption and operating costs, and extends the duration of a single operation. Furthermore, the algorithm and hardware platform are highly compatible, possessing excellent engineering feasibility and meeting the practical application needs of large quadcopter UAVs for high-altitude building facade cleaning.

[0118] like Figure 3 As shown, based on the same inventive concept as the above-mentioned UAV path planning method for building curtain wall cleaning operations, the present invention also provides a UAV path planning system for building curtain wall cleaning operations, used to implement the aforementioned UAV path planning method for building curtain wall cleaning operations. The system includes: a 3D modeling and work surface generation module 100, a path search module 200, and a trajectory smoothing and dynamic replanning module 300.

[0119] The 3D modeling and work surface generation module 100 is used to acquire 3D point cloud data of the building facade, discretize it into a 3D grid map and mark wall obstacles; using the wall surface as a reference plane, it offsets outward by a preset safety distance along the normal vector direction of the wall surface to generate a cleaning work surface that is consistent with and parallel to the geometric shape of the wall surface, and constrains all path points of the UAV path planning to the cleaning work surface or its adjacent allowable deviation range;

[0120] The path search module 200 is used to construct a corresponding two-dimensional grid search space based on the cleaning operation surface; within the grid search space, an improved A* path search algorithm is executed using a reconstructed multi-constraint cost function to solve for the optimal path from a preset starting point to the target endpoint, thereby obtaining a discrete grid path; wherein, the multi-constraint cost function includes constraint cost terms for safety proximity, attitude smoothness, energy consumption, and coverage continuity.

[0121] The trajectory smoothing and dynamic replanning module 300 is used to perform trajectory smoothing on the discrete grid path to obtain a continuous executable trajectory suitable for execution by the UAV flight control system; when the UAV performs facade cleaning operation along the continuous smooth trajectory, when a sudden obstacle not included in the original plan is identified, local dynamic replanning is performed with the current real-time position of the UAV as a new starting point.

[0122] This embodiment proposes a UAV path planning system for building curtain wall cleaning operations, which is used to implement the aforementioned UAV path planning method for building curtain wall cleaning operations. Therefore, the specific implementation of the UAV path planning system for building curtain wall cleaning operations can be found in the embodiment section of the aforementioned UAV path planning method for building curtain wall cleaning operations. For example, the 3D modeling and work surface generation module 100, the path search module 200, and the trajectory smoothing and dynamic replanning module 300 are respectively used to implement steps S1 to S3 in the method described in Embodiment 1. Therefore, the specific implementation can be referred to the description of the corresponding embodiments. To avoid redundancy, it will not be repeated here.

[0123] To verify the engineering effectiveness and performance superiority of the UAV path planning method for building curtain wall cleaning proposed in this invention, a dedicated simulation experimental environment was built. Using the traditional A* path planning algorithm as a control, comparative experiments were conducted from five dimensions: overall path planning performance, adaptability to complex facade scenes, energy consumption characteristics, dynamic replanning capability, and statistical stability. The experimental setup and result analysis are detailed below.

[0124] This experiment uses drone-based cleaning of building facades as the application scenario. A MATLAB-based, self-built 3D grid simulation system was constructed as the experimental platform. To ensure the reliability and generalizability of the experimental results, all experimental indicators were calculated by averaging after 10 repeated experiments. The core experimental environment parameters and comparison scheme settings are as follows:

[0125] The building facade measures 40m × 25m, and a facade model was constructed using approximately 2.1 million 3D point cloud data points. The grid resolution was set to 0.2m, with a total of approximately 25,000 grid points. The safety offset distance was 0.8m, the optimal cleaning distance was 1.0m, and the effective coverage radius of the cleaning nozzles was R = 0.5m. The control group used the traditional A* path planning algorithm, while the experimental group used the multi-constraint A* path planning algorithm proposed in this invention.

[0126] Table 1 shows a comparison of the core performance indicators of overall path planning between the traditional A* algorithm and the method of this invention. The improvement of each indicator is calculated using relative change rates. "This indicates an improvement in the indicator value," "Indicates a decrease in the indicator value."

[0127] Table 1 Overall Path Planning Performance Comparison Table

[0128]

[0129] As shown in Table 1, compared with the traditional A* path planning algorithm, the method of this invention achieves significant optimization in the core indicators of job adaptability: the cleaning coverage rate is increased by 6.5%, and the repetitive coverage rate is significantly reduced by 75.9%, effectively solving the problems of missed cleaning areas and repeated cleaning in traditional algorithms; the average close-range distance error is reduced by 80.6%, ensuring that the UAV maintains the optimal operating distance with the building facade and improving operational safety; the average attitude angle change and cumulative height change are reduced by 54.2% and 30.7% respectively, achieving smooth optimization of UAV flight attitude and effective suppression of vertical maneuvering; the total path length is reduced by 5.25%, further improving operational efficiency. The path search time of the method of this invention is slightly increased because the multi-objective optimization calculation of the multi-constraint cost function introduces reasonable algorithm complexity, which is a reasonable cost of performance optimization.

[0130] To verify the adaptability of the method of the present invention to building facades of different complexities, a special comparative experiment was conducted in three typical scenarios: flat facade, facade with concave and convex structure, and facade with bay windows and hanging objects. The core assessment indicators were cleaning coverage rate and repetition coverage rate. The results are shown in Table 2.

[0131] Table 2 Performance Comparison Table for Different Complex Facade Scenarios

[0132]

[0133] Experimental results show that both algorithms exhibit good performance in flat facade scenarios, but the method of this invention still achieves a 4.2 percentage point increase in coverage and a 75.8% decrease in repetition rate. As the complexity of building facades increases, the cleaning coverage of the traditional A* algorithm shows a significant downward trend, while the repetition coverage increases significantly. In complex scenarios with bay windows and hanging objects, the coverage is only 87.5%, and the repetition coverage is as high as 23.7%, which is difficult to meet the requirements of engineering operations. In contrast, the method of this invention maintains a high cleaning coverage of over 97% and a low repetition coverage of less than 7% in various complex scenarios, demonstrating strong adaptability to complex building facades and solving the problem of insufficient adaptability of traditional algorithms in irregular facade scenarios.

[0134] To address the energy-sensitive nature of UAV high-altitude cleaning operations, a quantitative comparison of the flight energy consumption of two types of algorithms was conducted. The energy consumption indicators were divided into horizontal flight energy consumption, vertical take-off and landing energy consumption, and total energy consumption, all in watt-hours (Wh). The results are shown in Table 3.

[0135] Table 3 Comparison of Operational Energy Consumption Characteristics

[0136]

[0137] As shown in Table 3, the method of the present invention achieves comprehensive optimization of the energy consumption of UAV operations. Among them, the energy consumption of vertical take-off and landing is reduced by 32.7%, which is the core optimization item. This is due to the effective suppression of meaningless vertical maneuvers by the energy consumption cost term in the multi-constraint cost function. The energy consumption of horizontal flight is reduced slightly by 5.0%, which is due to the shorter total path length and smoother trajectory after optimization. Finally, the total energy consumption is reduced by 12.2%, which effectively extends the endurance of a single UAV operation and reduces the energy consumption cost of high-altitude cleaning operations.

[0138] In actual drone cleaning operations, the ability to avoid sudden obstacles in real time is the key to ensuring the continuity and safety of the operation. This experiment verifies the dynamic replanning ability of two types of algorithms and evaluates three indicators: average replanning time, increase in replanning path length, and cleaning coverage after replanning. The results are shown in Table 4.

[0139] Table 4. Performance Comparison of Dynamic Replanning

[0140]

[0141] Experimental results show that the average time taken for dynamic replanning using the method of this invention is reduced by 48.4% compared to the traditional A* algorithm, the increase in replanning path length is reduced by 56.0%, and a high cleaning coverage rate of 98.4% is still maintained after replanning. This is because the method of this invention adopts a replanning strategy of local sub-job face pruning, which significantly compresses the search space for replanning. At the same time, the multi-constraint cost function ensures the job adaptability of the replanning path. In contrast, the traditional A* algorithm, due to the lack of local search constraints, needs to traverse a large area of ​​space again during replanning, resulting in long processing time, high path redundancy, and difficulty in guaranteeing the cleaning coverage effect after replanning.

[0142] To verify the performance consistency and robustness of the two types of algorithms in multiple experiments, the standard deviation of the core indicators (coverage, repetition rate, path length) of 10 repeated experiments was calculated. The smaller the standard deviation, the more stable the algorithm performance. The results are shown in Table 5.

[0143] Table 5 Comparison of Statistical Stability of Experimental Results

[0144]

[0145] Statistical results show that the standard deviations of all core indicators of the method of this invention are significantly lower than those of the traditional A* algorithm. Specifically, the standard deviation of coverage rate decreases by 73.1%, the standard deviation of repetition rate decreases by 70.2%, and the standard deviation of path length decreases by 53.8%. This indicates that the multi-constraint cost function and path planning process of the method of this invention have good robustness and can maintain stable operation performance in multiple experiments. In contrast, the traditional A* algorithm, due to the lack of specific constraints for the operation scenario, is easily affected by factors such as grid discretization and changes in scene features, resulting in large performance fluctuations and difficulty in ensuring the stability of engineering operations.

[0146] Through multi-dimensional simulation comparison experiments, the comprehensive performance superiority of the proposed multi-constraint A* UAV cleaning path planning algorithm was fully verified. Compared with the traditional A* algorithm, this method presents the following core technical advantages:

[0147] 1. Significantly improved job adaptability: Achieved a 6.5% increase in cleaning coverage and a 75.9% decrease in duplicate coverage, while maintaining a high coverage rate of over 97% in various complex building facade scenarios, effectively solving the problems of missed cleaning and duplicate cleaning in traditional algorithms;

[0148] 2. Optimization of flight safety and stability: The average close-range distance error decreased by 80.6%, and the average attitude angle change decreased by 54.2%, ensuring that the UAV maintains the optimal operating distance from the building facade, while suppressing drastic changes in flight attitude and improving the stability of the operation process;

[0149] 3. Outstanding energy efficiency: By effectively constraining vertical maneuvering, the total energy consumption of the operation is reduced by 12.2%, the single-operation endurance of the UAV is extended, and the engineering operation cost is reduced;

[0150] 4. Strong adaptability to dynamic environment: The time for dynamic replanning is reduced by nearly 50%, and a high cleaning coverage rate is maintained after replanning. It can quickly and efficiently avoid sudden obstacles in the operation process and ensure the continuity of operation.

[0151] 5. Excellent algorithm robustness: The statistical standard deviation of each core performance indicator is significantly reduced, and the planning effect remains stable in multiple experiments, meeting the operational stability requirements of actual engineering.

[0152] Furthermore, although the path search time of the method of this invention increases slightly, this increase is a reasonable increase in algorithmic complexity due to multi-objective optimization. Compared to the significant improvements in operational efficiency, safety, and adaptability, this cost is engineering-acceptable. Overall, the method of this invention can effectively adapt to the needs of drone cleaning operations on complex building facades, taking into account the safety, full coverage, stability, low energy consumption, and real-time performance of path planning, and has good engineering application value.

[0153] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0154] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0155] 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.

[0156] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0157] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for UAV path planning in building curtain wall cleaning operations, characterized in that, Includes the following steps: Step S1: Acquire 3D point cloud data of the building facade, discretize it into a 3D grid map and mark wall obstacles; take the wall surface as the reference plane, offset outward by a preset safety distance along the normal vector direction of the wall surface to generate a cleaning operation surface that is consistent with and parallel to the geometry of the wall surface, and constrain all path points of the UAV path planning to the cleaning operation surface or its adjacent allowable deviation range. Step S2: Based on the cleaning operation surface, construct a corresponding two-dimensional grid search space; within the grid search space, execute the improved A* path search algorithm using a reconstructed multi-constraint cost function to solve for the optimal path from the preset starting point to the target endpoint, obtaining a discrete grid path; wherein, the multi-constraint cost function includes constraint cost terms for safety proximity, attitude smoothness, energy consumption, and coverage continuity, and the expression of the multi-constraint cost function is: ， in, The actual cost from the starting point to the current node n. Let n be the heuristic cost from the current node n to the destination. For the sake of safety and close proximity, For the sake of pose smoothness, For the cost of energy consumption, To cover the cost of continuity, , , , These are the weighting coefficients for the corresponding cost terms, and the safety proximity cost is... The calculation method is as follows: ， in, Let n be the distance from the current node n to the actual wall. To preset the optimal cleaning operation distance, the posture smoothness cost The calculation method is as follows: ， in, The energy cost is the minimum change in attitude angle from the previous node p to the current node n. The calculation method is as follows: ， in, Energy consumption coefficient The altitude change between the start and end points of the current flight segment, and the coverage continuity cost. The calculation method is as follows: Based on the effective operating radius R of the cleaning nozzles, the cleaning area of ​​the building facade is marked with a grid, and the spatial status of covered and uncovered areas is maintained in real time. If the current node n is located in an uncovered area, then , If the current node n falls into the covered area, then The specific calculation formula for the preset positive value penalty term is as follows: ， in, The penalty coefficient is a constant, with a value range of [2, 5]. This refers to the building facade area that has been cleaned in the current phase of the operation. The total area of ​​the building facade to be cleaned. The basic cost of single-step movement in the grid space; Step S3: Perform trajectory smoothing on the discrete grid path to obtain a continuous executable trajectory suitable for execution by the UAV flight control system; when the UAV performs facade cleaning operation along the continuous smooth trajectory, if a sudden obstacle not included in the original plan is identified, perform local dynamic replanning with the current real-time position of the UAV as a new starting point.

2. The UAV path planning method for building curtain wall cleaning operations according to claim 1, characterized in that: In step S3, the method for smoothing the discrete grid path to obtain a continuous executable trajectory suitable for execution by the UAV flight control system is as follows: S31: Extract all path nodes on the discrete grid path and obtain the three-dimensional spatial coordinates of each node; perform noise reduction and redundancy removal on the extracted node coordinates, remove abnormal nodes and duplicate nodes caused by grid discretization, and filter to obtain a spatially continuous effective path node sequence. S32: Set the order of the cubic B-spline curve to 3. Based on the number of nodes and spatial distribution density of the effective path node sequence, generate a node vector using uniform node interpolation. Initialize the control vertices of the cubic B-spline curve so that the number of control vertices matches the number of nodes in the effective path node sequence, while constraining the spatial position of the control vertices to always be within the cleaning operation surface or its allowable deviation range. S33: The preprocessed effective node sequence is used as the fitting target point. The cubic B-spline curve fitting formula is substituted into it, and the optimal solution of the control vertex is solved by the least squares method to keep the deviation between the fitted curve and the discrete nodes within the preset allowable range. During the fitting process, the constrained fitted curve is always located within the preset cleaning operation surface or its allowable deviation range. S34: Calculate the curvature, first derivative, and second derivative of each segment of the fitted cubic B-spline curve, and verify whether the curve has inflection points, abrupt changes, or other areas that do not meet the requirements for UAV flight; if there are areas with unsmoothness, adjust the control vertex parameters and refit and iterate until the curvature of each segment of the fitted curve changes uniformly. S35: The optimized cubic B-spline continuous curve is uniformly discretized according to the control frequency and accuracy requirements of the UAV flight control system to generate trajectory sampling points; The trajectory sampling points are formatted and converted to generate a continuous executable trajectory that is adapted to the UAV flight control system.

3. The UAV path planning method for building curtain wall cleaning operations according to claim 1, characterized in that: In step S3, during the facade cleaning operation performed by the UAV along the continuous smooth trajectory, when a sudden obstacle not included in the original plan is identified, the method for local dynamic replanning using the UAV's current real-time position as a new starting point is as follows: S301: Real-time acquisition of building facade environmental information, and comparison of the real-time acquired environmental data with the pre-constructed 3D raster map and the original planned path; S302: When an unmarked and unforeseen sudden obstacle is detected in the original planned path, a local dynamic replanning event is triggered. The current trajectory execution is immediately paused, and the current real-time position of the UAV is used as the new starting point for dynamic replanning. Through path tracing and retrieval, the nearest effective path point behind the sudden obstacle in the original planned path is determined as the temporary endpoint of this local replanning. S303: Based on the determined spatial coordinate range of the new starting point and the temporary end point, on the pre-constructed overall cleaning operation surface, a local sub-operation surface containing the area of ​​sudden obstacles and the preset buffer range around the new starting point and the temporary end point is cut out, and only the grid data, cleaning operation constraints and safety distance requirements within the local sub-operation surface are retained. S304: In the trimmed local sub-work surface, the improved A* path search algorithm described in step S2 is run again. The multi-constraint cost function is used to comprehensively consider safety proximity, posture smoothness, energy consumption and coverage continuity constraints. Under the premise of avoiding sudden obstacles, the optimal local path from the new starting point to the temporary end point is searched. S305: The optimal local path generated by replanning is concatenated with the original planned path and smoothed to obtain a smoothly connected path; S306: Based on the smooth connection path, perform cleaning operations in local areas. After passing through the obstacle area, restore to the original planned path or continue along the updated path to complete the remaining building facade cleaning operations.

4. The UAV path planning method for building curtain wall cleaning operations according to claim 1, characterized in that: During the facade cleaning operation performed by the UAV along the continuous smooth trajectory, when a sudden obstacle not included in the original plan is identified, the method of performing local dynamic replanning with the UAV's current real-time position as a new starting point also includes: If the sudden obstacle is a fixed and long-term obstacle, the obstacle information can be updated to the local 3D raster map.

5. A drone path planning system for building curtain wall cleaning operations, characterized in that, The system for implementing the UAV path planning method for building curtain wall cleaning operations as described in any one of claims 1 to 4 includes the following modules: The 3D modeling and work surface generation module is used to acquire 3D point cloud data of the building facade, discretize it into a 3D grid map and mark wall obstacles; using the wall surface as a reference plane, it offsets outward by a preset safety distance along the normal vector direction of the wall surface to generate a cleaning work surface that is consistent with and parallel to the geometry of the wall surface, and constrains all path points of the UAV path planning to the cleaning work surface or its adjacent allowable deviation range. The path search module is used to construct a corresponding two-dimensional grid search space based on the cleaning operation surface; within the grid search space, an improved A* path search algorithm is executed using a reconstructed multi-constraint cost function to solve for the optimal path from a preset starting point to the target endpoint, thereby obtaining a discrete grid path; wherein, the multi-constraint cost function includes constraint cost terms for safety proximity, attitude smoothness, energy consumption, and coverage continuity. The trajectory smoothing and dynamic replanning module is used to smooth the discrete grid path to obtain a continuous executable trajectory suitable for the UAV flight control system. When the UAV performs facade cleaning operation along the continuous smooth trajectory, if a sudden obstacle not included in the original plan is identified, local dynamic replanning is performed with the current real-time position of the UAV as a new starting point.

Images