Multi-unmanned aerial vehicle cluster curtain wall cleaning task partition scheduling and real-time cooperation control method

Abstract

The invention discloses a multi-unmanned aerial vehicle cluster curtain wall cleaning task partition scheduling and real-time cooperation control method. The method comprises the following steps of 1, curtain wall cleaning task demand analysis and three-dimensional environment modeling; 2, carrying out dynamic task partitioning based on curtain wall characteristics and pollution distribution; step 3, multi-unmanned aerial vehicle state real-time sensing and distributed information interaction; 4, constructing a dynamic scheduling model based on the state of the unmanned aerial vehicle and the partition task; 5, executing a multi-unmanned aerial vehicle real-time cooperative control strategy; step 6, performing real-time feedback and optimization on scheduling and cooperation effects; 7, verifying and ending the whole system; according to the invention, through a dynamic task partitioning strategy, in combination with curtain wall pollution level distribution and the real-time state of the unmanned aerial vehicle, efficient and balanced distribution of resources is realized, and the problem of uneven load caused by traditional static partitioning is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicle application, in particular to a multi-unmanned aerial vehicle cluster curtain wall cleaning task partition scheduling and real-time cooperative control method. BACKGROUND

[0002] With the acceleration of urbanization, high-rise buildings, exhibition centers, airport terminals and other large buildings widely use glass curtain walls as the outer facade, and the demand for cleaning and maintenance is growing. Traditional curtain wall cleaning mainly relies on manual suspension operation, which has high risk of high-altitude operation, low efficiency, high cost and is greatly affected by weather; some scenes use single unmanned aerial vehicle for cleaning, which can avoid manual risks, but the operation range of a single device is limited, and the operation cycle is long when facing large-area curtain walls, which is difficult to meet the efficient cleaning demand.

[0003] To improve efficiency, existing technologies try to introduce multi-unmanned aerial vehicle cluster operation, but there are significant limitations:

[0004] Task partition is static: Most methods are based on preset fixed partition of curtain wall initial size, without considering the difference in curtain wall pollution level distribution, real-time operation state of unmanned aerial vehicle (such as power, cleaning tool wear), leading to overloading of some partition unmanned aerial vehicles, idling of some unmanned aerial vehicles, and low resource utilization;

[0005] Lack of real-time cooperative control: The data interaction frequency between unmanned aerial vehicles is low, and only relies on preset path planning, which cannot respond to unexpected situations (such as unmanned aerial vehicle failure, local pollution intensification, sudden airflow), is prone to operation conflict, task fault, and even causes collision risk of equipment;

[0006] Single scheduling model: Existing scheduling only considers the simple matching of "task-unmanned aerial vehicle", without multi-objective optimization of operation efficiency, energy consumption cost, safety constraints, etc., leading to suboptimal overall operation efficiency, and difficulty in adapting to complex curtain wall structures (such as curved and irregular curtain walls).

[0007] Therefore, the present application provides a multi-unmanned aerial vehicle cluster curtain wall cleaning task partition scheduling and real-time cooperative control method, which solves the problems of low efficiency, poor adaptability and insufficient safety of existing technologies through dynamic task partition, real-time scheduling based on unmanned aerial vehicle state and multi-dimensional cooperative control. SUMMARY

[0008] The purpose of the present application is to provide a multi-unmanned aerial vehicle cluster curtain wall cleaning task partition scheduling and real-time cooperative control method to solve the problems raised in the background.

[0009] To achieve the above purpose, the present application provides the following technical solution: a multi-unmanned aerial vehicle cluster curtain wall cleaning task partition scheduling and real-time cooperative control method, comprising the following steps:

[0010] Step one: curtain wall cleaning task demand analysis and three-dimensional environment modeling, collecting curtain wall basic parameters, pollution level distribution and unmanned aerial vehicle initial parameters, constructing a three-dimensional environment model containing curtain wall grid elements;

[0011] Step two: dynamic task partitioning based on curtain wall characteristics and pollution distribution, using an improved K-means algorithm combined with partition evaluation index S to achieve reasonable partitioning and ensure that the partitioning meets the area, pollution uniformity, and job radius constraints;

[0012] Step three: multi-unmanned aerial vehicle state real-time sensing and distributed information interaction, collecting unmanned aerial vehicle position, power, and operation state through multiple sensors, and constructing a three-level communication architecture to realize data interaction and fusion;

[0013] Step four: dynamic scheduling model construction based on unmanned aerial vehicle state and partitioned tasks, taking multi-objective optimization objective function F as the core, formulating scheduling trigger conditions and solution rules, and realizing optimal allocation of unmanned aerial vehicle-partitioned tasks;

[0014] Step five: multi-unmanned aerial vehicle real-time cooperative control strategy execution, including path planning cooperation, task handover cooperation, and cleaning quality cooperation to avoid conflicts and missed cleaning;

[0015] Step six: real-time feedback and optimization of scheduling and cooperation effect, collecting operation progress, quality, and resource utilization rate indicators to adjust partitioning and scheduling parameters;

[0016] Step seven: overall system verification and completion, verifying the feasibility of the method through simulation and field testing, and generating an operation report.

[0017] Preferably, in step one, the curtain wall basic parameters include total area S, structure type, and material, the pollution level is divided into light pollution P1, moderate pollution P2, and heavy pollution P3 through image recognition, and the quantitative standard is P∈[0,1]; the unmanned aerial vehicle initial parameters include maximum operation radius R j , maximum endurance time T imaxj , real-time power E j , cleaning tool wear coefficient L j , the three-dimensional environment model is constructed based on laser radar point cloud data using an improved octree algorithm, and the curtain wall is divided into a plurality of basic grid elements U i , each element parameter includes area A i , center coordinates (x i , y i , z i ), pollution level P i , and material coefficient M i .

[0018] Preferably, in step two, the partitioning constraints include the area S k≤ K x C i , the pollution level difference ΔP in the partition ≤ 0.2, the distance D between the partition center and the initial take-off and landing point of the UAV k ≤ R i ; the calculation formula of the partition evaluation index S of the improved K-means algorithm is:

[0019]

[0020] Wherein, ω1+ω2+ω3=1, A max is the maximum partition area, n k is the number of grid units of partition k, is the average pollution level of partition k, C k is the path cost of partition k, C max is the maximum path cost, and the partition meets the standard when S≥0.8.

[0021] Preferably, in step three, the parameters of UAV state perception include position (x j (t), y j (t), z j (t)), flight speed v j (t), attitude angle (φ j (t), θ j (t), ψ j (t)), real-time power E j (t), cleaning tool wear coefficient L j (t), completed cleaning area S jn (t), distance D jk (t) from surrounding UAVs; the three-level communication architecture includes inter-UAV LoRaWAN communication, UAV-5G communication with edge nodes, and edge node-fiber communication with ground control center, and state data is fused by Kalman filtering algorithm.

[0022] Preferably, in step four, the calculation formula of the multi-objective optimization target function F of dynamic scheduling is:

[0023]

[0024] Wherein, α+β+γ=1, T total is the total operation time, T0 is the preset operation time constraint, E total is the total power consumption, E max is the initial total power, R conflict is the conflict rate; the scheduling trigger conditions include UAV power E j (t)≤0.2, partition pollution level difference ≥0.3, and conflict rate R conflict≥0.05, UAV malfunction, the scheduling scheme is solved using particle swarm optimization algorithm, the closer the F value is to 0, the better the scheduling.

[0025] Preferably, in step five, the path planning collaboration adopts the A* algorithm, and the path must meet the following requirements: covering all grid cells, the distance between UAVs ≥ the safe distance threshold D0, and minimizing the path length; during task handover collaboration, the low-battery UAV sends a request to the edge node, selects the UAV with the lightest load to receive the unfinished task, and hovers at the safe waiting point during handover; the cleaning quality collaboration adopts a unified parameter benchmark: lightly contaminated cleaning tool speed 800 r / min, spray pressure 0.3 MPa, moderately contaminated 1200 r / min, 0.5 MPa, heavily contaminated 1600 r / min, 0.8 MPa, and a rewash is triggered when the cleanliness is <95%.

[0026] Preferably, in step six, the evaluation indicators for real-time feedback include work progress (completed area / total area × 100%), cleaning quality (average cleanliness), resource utilization (actual work time / total time × 100%), and safety indicators (number of conflicts, number of collision risks). The feedback adjustment mechanism includes: adding drones or splitting zones when the progress is lower than the preset value, optimizing cleaning parameters when the quality is substandard, adjusting scheduling trigger conditions when resource utilization is low, and iteratively optimizing the weight parameters of zones and scheduling.

[0027] Preferably, in step seven, system verification includes Unity simulation verification (simulating pollution distribution and fault scenarios, verifying F-value < 0.3 and conflict rate < 0.01) and field testing (selecting a super high-rise curtain wall, deploying 5-8 drones, and recording operation time, quality, and number of conflicts); when the operation is completed, a report containing total operation time, power consumption, quality compliance rate, and safety records is generated to provide parameter reference for subsequent tasks.

[0028] Preferably, in step two, if the zoning result verification does not meet the constraints, an adjustment method of "splitting large zoning and merging small zoning" is adopted. When splitting, the zoning is based on the curtain wall structure. When merging, adjacent zoning with similar pollution levels is merged first. The evaluation index S is recalculated until the standard is met.

[0029] Preferably, in step four, the scheduling instruction includes a list of grid cells for the new partition, path planning update parameters (such as flight speed adjustment values ​​and obstacle avoidance point coordinates), and cooperation requirements (such as handover time windows and rewash area ranges). The instruction is transmitted encrypted through edge nodes to ensure data security.

[0030] Compared with the prior art, the beneficial effects of the present invention are:

[0031] This invention achieves efficient and balanced resource allocation through a dynamic task partitioning strategy, combining the distribution of curtain wall pollution levels with the real-time status of UAVs, thus solving the problem of uneven load caused by traditional static partitioning. Based on a multi-objective optimization dynamic scheduling model, which integrates operational efficiency, energy consumption costs, and safety constraints, the F-value of the scheduling scheme is significantly reduced, and the overall operational efficiency is improved. The three-level communication architecture and real-time collaborative control strategy ensure high-frequency data interaction and rapid response between UAVs, and greatly reduce the risk of task interruption and equipment collision. Attached Figure Description

[0032] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0033] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0034] Please see Figure 1 This invention provides a method for partitioned scheduling and real-time collaborative control of multi-UAV cluster curtain wall cleaning tasks, including the following steps:

[0035] Step 1: Analysis of curtain wall cleaning task requirements and 3D environment modeling, collection of basic curtain wall parameters, pollution level distribution and initial parameters of UAV, and construction of a 3D environment model containing curtain wall mesh units;

[0036] Step 2: Based on the characteristics of the curtain wall and the distribution of pollution, dynamic task zoning is carried out. An improved K-means algorithm is used in combination with the zoning evaluation index S to achieve reasonable zoning and ensure that the zoning meets the constraints of area, pollution uniformity and operation radius.

[0037] Step 3: Real-time perception and distributed information interaction of multiple drone statuses. By collecting drone location, battery level, and operational status through multiple sensors, a three-level communication architecture is constructed to achieve data interaction and fusion.

[0038] Step 4: Construct a dynamic scheduling model based on UAV status and partitioned tasks. With the multi-objective optimization objective function F as the core, formulate scheduling trigger conditions and solution rules to achieve optimal allocation of UAV-partitioned tasks.

[0039] Step 5: Execute real-time collaborative control strategies for multiple drones, including path planning collaboration, task handover collaboration, and cleaning quality collaboration, to avoid conflicts and missed cleaning.

[0040] Step Six: Real-time feedback and optimization of scheduling and collaboration effects; collection of work progress, quality, and resource utilization indicators; and adjustment of partitioning and scheduling parameters.

[0041] Step 7: System overall verification and completion. Verify the feasibility of the method through simulation and field testing, and generate a work report.

[0042] In step one, the basic parameters of the curtain wall include the total area S, structural type, and material. The pollution level is classified into light pollution P1, moderate pollution P2, and heavy pollution P3 through image recognition, with the quantification standard being P∈[0,1]. The initial parameters of the UAV include the maximum operating radius R. j Maximum battery life T imaxj Real-time battery level E j Cleaning tool wear factor L j The 3D environment model is constructed using an improved octree algorithm based on LiDAR point cloud data, and the curtain wall is divided into several basic mesh units U. i Each unit parameter includes area A i , center coordinates (x i ,y i ,z i Pollution level P i Material coefficient M i .

[0043] In step two, the partitioning constraints include the area S of each partition. k ≤K×C i The difference in pollution level within the zone ΔP ≤ 0.2; the distance D between the zone center and the initial take-off and landing point of the UAV. k ≤R i The formula for calculating the partition evaluation index S of the improved K-means algorithm is as follows:

[0044]

[0045] Where ω1+ω2+ω3=1, A max For the maximum partition area, n k The number of grid cells in partition k. C represents the average pollution level of zone k. k For the path cost of partition k, C max For the maximum path cost, the partition meets the standard when S≥0.8.

[0046] In step three, the parameters for the UAV's state perception include position (x... j (t),y j (t),z j (t)), flight speed v j (t), attitude angle (φ) j (t),θj (t),ψ j (t)), Real-time power E j (t), Cleaning tool wear coefficient L j (t), Area S that has been cleaned jn (t), Distance D from surrounding drones jk (t); The three-level communication architecture includes LoRaWAN communication between UAVs, 5G communication between UAVs and edge nodes, and fiber optic communication between edge nodes and ground control centers. The status data is fused using the Kalman filter algorithm.

[0047] In step four, the formula for calculating the multi-objective optimization objective function F for dynamic scheduling is:

[0048]

[0049] Where α+β+γ=1, T total The total operation time is T0, where T0 is the preset operation time constraint, and E is the total operation time. total E represents the total power consumption. max R is the initial total energy. conflict The conflict rate; scheduling trigger conditions include the drone's battery level E. j (t)≤0.2, Difference in pollution level between zones≥0.3, Conflict rate R conflict ≥0.05, UAV malfunction, the scheduling scheme is solved using particle swarm optimization algorithm, the closer the F value is to 0, the better the scheduling.

[0050] In step five, the path planning collaboration adopts the A* algorithm. The path must meet the following requirements: covering all grid cells, the distance between drones ≥ the safe distance threshold D0, and minimizing the path length. During task handover collaboration, low-battery drones send requests to edge nodes, selecting the drone with the lightest load to receive the unfinished task, and hovering at a safe waiting point during handover. The cleaning quality collaboration adopts a unified parameter benchmark: lightly contaminated cleaning tool speed 800 r / min, spray pressure 0.3 MPa; moderately contaminated 1200 r / min, 0.5 MPa; heavily contaminated 1600 r / min, 0.8 MPa; and a rewash is triggered when the cleanliness is <95%.

[0051] In step six, the evaluation indicators for real-time feedback include work progress (completed area / total area × 100%), cleaning quality (average cleanliness), resource utilization (actual work time / total time × 100%), and safety indicators (number of conflicts, number of collision risks). The feedback adjustment mechanism includes: adding drones or splitting zones when the progress is lower than the preset value, optimizing cleaning parameters when the quality is substandard, adjusting scheduling trigger conditions when resource utilization is low, and iteratively optimizing the weight parameters of zones and scheduling.

[0052] In step seven, system verification includes Unity simulation verification (simulating pollution distribution and fault scenarios to verify F-value < 0.3 and conflict rate < 0.01) and field testing (selecting a super high-rise curtain wall, deploying 5-8 drones, and recording operation time, quality, and number of conflicts); when the operation is completed, a report containing total operation time, power consumption, quality compliance rate, and safety records is generated to provide parameter reference for subsequent tasks.

[0053] In step two, if the zoning results do not meet the constraints, the adjustment method of "splitting large zoning and merging small zoning" is adopted. When splitting, the zoning is based on the curtain wall structure. When merging, adjacent zoning with similar pollution levels is prioritized. The evaluation index S is recalculated until the standard is met.

[0054] In step four, the scheduling instructions include a list of grid cells for the new partition, path planning update parameters (such as flight speed adjustment values ​​and obstacle avoidance point coordinates), and cooperation requirements (such as handover time windows and rewash area ranges). The instructions are transmitted encrypted through edge nodes to ensure data security.

[0055] Example:

[0056] Taking the curtain wall cleaning task of a 30-story super high-rise building as an example, the total area of ​​the curtain wall is 5,000 square meters, and the structure is a mixture of glass curtain wall and aluminum panel curtain wall, with a material coefficient between 0.8 and 1.2. Using image recognition technology, the curtain wall pollution level was classified into light pollution P1 (30%), moderate pollution P2 (50%), and heavy pollution P3 (20%). Six drones were deployed, with initial parameters of a maximum operating radius of 15 meters, a maximum flight time of 45 minutes, initial battery level of 100%, and a cleaning tool wear coefficient of 0.05.

[0057] In step one, an improved octree algorithm is used to construct a three-dimensional environment model based on lidar point cloud data, dividing the curtain wall into 2000 basic grid units, each with an area of ​​2.5 square meters, including parameters such as center coordinates, pollution level, and material coefficient.

[0058] In step two, based on the zoning constraints (each zone area ≤ 200 square meters, pollution level difference within a zone ≤ 0.2, and distance between the zone center and the initial take-off and landing point of the UAV ≤ 15 meters), an improved K-means algorithm is used for dynamic task zoning. After initial zoning, by calculating the zoning evaluation index S, it was found that some zones did not meet the compliance condition of S ≥ 0.8. An adjustment method of "splitting large zones and merging small zones" was adopted, ultimately resulting in 12 compliant zones.

[0059] In step three, real-time status perception and distributed information exchange among multiple drones proceed normally. The drones collect parameters such as location, battery level, and operational status through multiple sensors, and achieve data exchange and fusion through a three-tier communication architecture. The Kalman filter algorithm effectively fuses the status data, improving data accuracy.

[0060] In step four, based on the UAV status and the assigned tasks, a dynamic scheduling model is constructed and the multi-objective optimization objective function F is solved. During the operation, multiple scheduling events are triggered, including situations where the UAV battery level is below 20% or the difference in pollution levels between zones exceeds 0.3. A particle swarm optimization algorithm is used to solve the scheduling scheme, and the F value gradually approaches 0, achieving the optimal allocation of UAVs to assigned tasks.

[0061] In step five, the multi-UAV real-time collaborative control strategy was executed smoothly. Path planning collaboration employed the A* algorithm to ensure that the path covered all grid cells, the distance between UAVs was greater than or equal to the safe distance threshold, and the path length was minimized. During task handover collaboration, low-battery UAVs promptly sent requests to edge nodes, selecting the lightest-loaded UAV to receive unfinished tasks. Cleaning quality collaboration used a unified parameter benchmark, adjusting the cleaning tool's rotation speed and spray pressure according to the contamination level, achieving an average cleanliness level of 96%.

[0062] In step six, the real-time feedback and optimization mechanism for scheduling and collaboration is effectively implemented. Indicators such as operation progress, quality, and resource utilization are collected. When the operation progress falls below the preset value, two more drones are added and some zones are split. When quality is substandard, cleaning parameters are optimized to improve cleanliness. When resource utilization is low, scheduling trigger conditions are adjusted to reduce ineffective scheduling.

[0063] In step seven, the overall system verification and finalization were successfully completed. Unity simulation was used to verify that pollution distribution and fault scenarios were simulated, confirming an F-value < 0.3 and a conflict rate < 0.01. Field testing was conducted on the super high-rise curtain wall, deploying six drones to record 420 minutes of operation time, a 98% quality compliance rate, and two conflicts. At the end of the operation, a report was generated including total operation time, power consumption, quality compliance rate, and safety records, providing parameter references for subsequent tasks.

[0064] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for zoned scheduling and real-time collaborative control of multi-UAV swarm curtain wall cleaning tasks, characterized by: Includes the following steps: Step 1: Analysis of curtain wall cleaning task requirements and 3D environment modeling, collection of basic curtain wall parameters, pollution level distribution and initial parameters of UAV, and construction of a 3D environment model containing curtain wall mesh units; Step 2: Based on the characteristics of the curtain wall and the distribution of pollution, dynamic task zoning is carried out. An improved K-means algorithm is used in combination with the zoning evaluation index S to achieve reasonable zoning and ensure that the zoning meets the constraints of area, pollution uniformity and operation radius. Step 3: Real-time perception and distributed information interaction of multiple drone statuses. By collecting drone location, battery level, and operational status through multiple sensors, a three-level communication architecture is constructed to achieve data interaction and fusion. Step 4: Construct a dynamic scheduling model based on UAV status and partitioned tasks. With the multi-objective optimization objective function F as the core, formulate scheduling trigger conditions and solution rules to achieve optimal allocation of UAV-partitioned tasks. Step 5: Execute real-time collaborative control strategies for multiple drones, including path planning collaboration, task handover collaboration, and cleaning quality collaboration, to avoid conflicts and missed cleaning. Step Six: Real-time feedback and optimization of scheduling and collaboration effects; collection of work progress, quality, and resource utilization indicators; and adjustment of partitioning and scheduling parameters. Step 7: System overall verification and completion. Verify the feasibility of the method through simulation and field testing, and generate a work report.

2. The method for zoned scheduling and real-time collaborative control of multi-UAV cluster curtain wall cleaning tasks according to claim 1, characterized in that: In step one, the basic parameters of the curtain wall include the total area S, structural type, and material. The pollution level is classified into light pollution P1, moderate pollution P2, and heavy pollution P3 through image recognition, with the quantification standard being P∈[0,1]. The initial parameters of the UAV include the maximum operating radius R. j Maximum battery life T imaxj Real-time battery level E j Cleaning tool wear factor L j The 3D environment model is constructed using an improved octree algorithm based on LiDAR point cloud data, and the curtain wall is divided into several basic mesh units U. i Each unit parameter includes area A i , center coordinates (x i ,y i ,z i Pollution level P i Material coefficient M i .

3. The method for zoned scheduling and real-time collaborative control of multi-UAV cluster curtain wall cleaning tasks according to claim 1, characterized in that: In step two, the partitioning constraints include the area S of each partition. k ≤K×C i The difference in pollution level within the zone ΔP ≤ 0.2; the distance D between the zone center and the initial take-off and landing point of the UAV. k ≤R i The formula for calculating the partition evaluation index S of the improved K-means algorithm is as follows: Where ω1+ω2+ω3=1, A max For the maximum partition area, n k The number of grid cells in partition k. C represents the average pollution level of zone k. k For the path cost of partition k, C max For the maximum path cost, the partition meets the standard when S≥0.

8.

4. The method for zoned scheduling and real-time collaborative control of multi-UAV cluster curtain wall cleaning tasks according to claim 1, characterized in that: In step three, the parameters for UAV state perception include position (x) j (t),y j (t),z j (t)), flight speed v j (t), attitude angle (φ) j (t),θ j (t),ψ j (t)), Real-time power E j (t), Cleaning tool wear coefficient L j (t), Area S that has been cleaned jn (t), Distance D from surrounding drones jk (t); The three-level communication architecture includes LoRaWAN communication between UAVs, 5G communication between UAVs and edge nodes, and fiber optic communication between edge nodes and ground control centers. The status data is fused using the Kalman filter algorithm.

5. The method for zoned scheduling and real-time collaborative control of multi-UAV cluster curtain wall cleaning tasks according to claim 1, characterized in that: In step four, the formula for calculating the multi-objective optimization objective function F for dynamic scheduling is: Where α+β+γ=1, T total The total operation time is T0, where T0 is the preset operation time constraint, and E is the total operation time. total E represents the total power consumption. max R is the initial total energy. conflict The conflict rate; scheduling trigger conditions include the drone's battery level E. j (t)≤0.2, Difference in pollution level between zones≥0.3, Conflict rate R conflict ≥0.05, UAV malfunction, the scheduling scheme is solved using particle swarm optimization algorithm, the closer the F value is to 0, the better the scheduling.

6. The method for zoned scheduling and real-time collaborative control of multi-UAV cluster curtain wall cleaning tasks according to claim 1, characterized in that: In step five, the path planning collaboration adopts the A* algorithm. The path must meet the following requirements: covering all grid cells, the distance between drones ≥ the safe distance threshold D0, and minimizing the path length. During task handover collaboration, low-battery drones send requests to edge nodes, selecting the drone with the lightest load to receive the unfinished task, and hovering at a safe waiting point during handover. The cleaning quality collaboration adopts a unified parameter benchmark: lightly contaminated cleaning tool speed 800 r / min, spray pressure 0.3 MPa; moderately contaminated 1200 r / min, 0.5 MPa; heavily contaminated 1600 r / min, 0.8 MPa; and a rewash is triggered when the cleanliness is <95%.

7. The method for zoned scheduling and real-time collaborative control of multi-UAV cluster curtain wall cleaning tasks according to claim 1, characterized in that: In step six, the evaluation indicators for real-time feedback include work progress, cleaning quality, resource utilization, and safety indicators. The feedback adjustment mechanism includes: adding drones or splitting the partition when the progress is lower than the preset value; optimizing cleaning parameters when the quality is substandard; adjusting scheduling trigger conditions when resource utilization is low; and iteratively optimizing the weight parameters of partitioning and scheduling.

8. The method for zoned scheduling and real-time collaborative control of multi-UAV cluster curtain wall cleaning tasks according to claim 1, characterized in that: In step seven, system verification includes Unity simulation verification and field testing; when the job is completed, a report is generated that includes total job time, power consumption, quality compliance rate, and safety records, providing parameter references for subsequent tasks.

9. The method for zoned scheduling and real-time collaborative control of multi-UAV cluster curtain wall cleaning tasks according to claim 1, characterized in that: In step two, if the zoning results do not meet the constraints, an adjustment method of "splitting large zoning and merging small zoning" is adopted. When splitting, the zoning is based on the curtain wall structure. When merging, adjacent zoning with similar pollution levels is merged first. The evaluation index S is recalculated until the standard is met.

10. The method for zoned scheduling and real-time collaborative control of multi-UAV cluster curtain wall cleaning tasks according to claim 1, characterized in that: In step four, the scheduling instruction includes a list of grid cells for the new partition, path planning update parameters, and collaboration requirements. The instruction is transmitted encrypted through edge nodes to ensure data security.

Images