Automatic cleaning control method of curtain wall cleaning unmanned aerial vehicle, unmanned aerial vehicle and medium

Abstract

The invention relates to an automatic cleaning control method of a curtain wall cleaning unmanned aerial vehicle, the unmanned aerial vehicle and a medium in the technical field of unmanned aerial vehicle application. The automatic cleaning control method of the curtain wall cleaning unmanned aerial vehicle comprises the steps that coverage units are divided according to curtain wall image data I (t), and a coverage cleaning path P is generated with the coverage units as granularity. And calculating a cleanliness score Sk according to the change of the P before and after cleaning. And when Sk is less than a threshold Sth, carrying out supplementary washing, adding unqualified products after supplementary washing into a return visit queue, generating a return visit path according to the return visit queue, and carrying out return visit supplementary washing according to the return visit path. And when all the covering units meet the condition that Sk is larger than or equal to Sth, cleaning is ended. According to the curtain wall cleaning method, the covering cleaning path is planned for primary cleaning through the curtain wall local image, meanwhile, whether the cleaned curtain wall is clean or not can be automatically detected, and the return visit path is planned for the curtain wall area where residual stains are detected for supplementary cleaning, so that the operation burden is reduced and the working efficiency is improved while the curtain wall cleaning effect is ensured.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) application technology, and in particular to an automatic cleaning control method for a curtain wall cleaning UAV, the UAV itself, and the cleaning medium. Background Technology

[0002] With the large-scale construction of high-rise buildings, large commercial complexes, and glass curtain wall buildings in cities, the demand for cleaning building exteriors, especially glass curtain walls, is increasing. Traditional curtain wall cleaning mainly relies on suspended platforms, window cleaning machines, or high-altitude rope operations, with manual labor carrying cleaning tools at high altitudes. This method has the following problems: First, high-altitude operations are extremely risky, greatly affected by wind speed, weather, and operator skills, resulting in a high accident rate; second, manual cleaning is inefficient, requiring repeated movement of the suspended platform or suspension ropes for large curtain walls, leading to long operation cycles and high labor costs; third, human judgment of cleaning quality is highly subjective, making it difficult to ensure uniform cleaning quality. In recent years, solutions using drones for exterior wall cleaning have gradually emerged. Some solutions use nozzles to spray high-pressure water onto the curtain wall or install simple brushes, sponges, and other cleaning components on the drone. However, existing technologies still have the following shortcomings: 1. Most solutions only use a single sensor (such as ultrasonic or simple ranging modules) to control the distance between the drone and the wall, resulting in poor anti-interference capabilities and difficulty in achieving stable and safe distance control in complex wind fields and curtain wall reflection environments. 2. Cleaning equipment is mostly fixed, lacking a controllable pre-pressure adjustment mechanism. Uneven contact between the cleaning rollers and the curtain wall can easily lead to incomplete cleaning in certain areas or scratches on the curtain wall surface. 3. Cleaning routes rely on manual planning, and the cleaning process lacks a visual inspection loop. It depends solely on operators' remote observation or experience-based judgment, making it impossible to assess cleaning effectiveness in real time or automatically plan the next work area. 4. Cleaning methods such as water spraying and brushing are often used alone, without precise control over the ratio and flow rate of detergent and water. This results in low cleaning efficiency and significant waste of water and detergent. Summary of the Invention

[0003] To address the technical problem that existing curtain wall cleaning drones rely on manual planning for cleaning routes and depend solely on operators' remote observation or experience for cleaning effectiveness, resulting in poor curtain wall cleaning results, this invention provides an automatic cleaning control method for curtain wall cleaning drones, the drone itself, and the cleaning medium.

[0004] In a first aspect, the present invention proposes an automatic cleaning control method for a curtain wall cleaning drone, comprising: based on curtain wall image data captured by the drone... I ( t The effective working area of ​​the curtain wall is identified and divided into... K Divide the coverage area into equal units; within the effective operating area, determine the effective coverage width of the drone cleaning. WGenerate coverage cleaning path with path overlap coefficient P ; along the drone P During the cleaning process, based on before and after cleaning I ( t ) calculation of the change of the first k Cleanliness rating of each coverage unit S k ;when S k <threshold S th At that time, for the first k Each coverage unit is preset with a maximum number of times. N Re-washing; if N Even after the second wash S k < S th , will the k Each coverage unit is added to the return visit queue; k ∈[1, K ];current P After traversal, a return path is generated based on the return queue, and return shuffling is performed based on the return path: the coordinates of the center points corresponding to all covered units in the return queue constitute the set of return points Ω: Ω = { p k,x,y}, and record the first k Number of washes per coverage unit n k Starting from the drone's current location, based on the current distance between the drone and the revisit point and... n k For the innermost Ω k Calculation priority of each coverage unit P k ;based on P k Select the next visit point according to the nearest neighbor greedy strategy. p next Perform follow-up cleaning; when all coverage units meet the requirements. S k ≥ S th The cleaning process ends at that time.

[0005] Secondly, this invention also proposes a curtain wall cleaning drone, which includes: a body, a sensing system, a cleaning execution system, and an edge processing system. The sensing system is mounted on the body and is used to acquire... I ( t )and d ( tThe cleaning execution system is mounted on the airframe for cleaning the curtain wall. The edge processing system is also mounted on the airframe and uses the automatic cleaning control method of the curtain wall cleaning UAV described in the first aspect to control the airframe along... P The system moves along the return path and controls the cleaning execution system to perform a cleaning and rewash operation.

[0006] Thirdly, the present invention also proposes a computer-readable storage medium. This computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the automatic cleaning control method for the curtain wall cleaning drone described in the first aspect.

[0007] The beneficial effects of this invention are as follows: This invention acquires partial images of the curtain wall using a visual camera and plans a comprehensive cleaning path for a single cleaning operation. It can also automatically detect whether the curtain wall is clean after cleaning and plan a return path for re-cleaning areas where residual stains are detected. This eliminates the need for manual planning of the trajectory segment by segment, ensuring the cleaning effect of the curtain wall while reducing the operational burden and improving work efficiency. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a structural schematic diagram of a curtain wall drone; Figure 2 This is a schematic diagram showing the state of the curtain wall drone when its telescopic mechanism is extended; Figure 3 This is a schematic diagram of the retractable mechanism of the curtain wall drone when it is in the retracted state. Figure 4 This is a flowchart of the automatic cleaning control method for curtain wall cleaning drones; Figure 5 This is a schematic diagram of the planning for covering the cleaning path; Figure 6 This is a flowchart for follow-up visits and re-washing based on the follow-up path; Figure 7 This is a schematic diagram of the follow-up visit route.

[0010] In the diagram: 1-1 is the duckbill nozzle, 1-2 is the roller, 1-3 is the millimeter-wave radar, 1-4 is the water pipe, 1-5 is the support pipe, 1-6 is the scissor telescopic rod assembly, 1-7 is the camera, 1-8 is the guide rail, 1-9 is the hydraulic push rod, and 1-10 is the solenoid valve. Detailed Implementation

[0011] The technical solutions in the embodiments of the present invention will be clearly and completely described below. 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 skilled in the art without creative effort are within the scope of protection of the present invention.

[0012] It should be noted that when a component is said to be "installed on" another component, it can be directly on the other component or it may be in a component that is centered on it. When a component is said to be "set on" another component, it can be directly set on the other component or it may also be in a component that is centered on it. When a component is said to be "fixed to" another component, it can be directly fixed to the other component or it may also be in a component that is centered on it.

[0013] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the specification of this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "or / and" as used herein includes any and all combinations of one or more of the associated listed items.

[0014] This invention achieves closed-loop control of "maintaining a safe distance—cleanliness evaluation—online replanning" through millimeter-wave radar 1-3 and visual cameras 1-7 carried by the drone itself, and improves cleaning effect and operation efficiency through automatic generation of cleaning paths and a follow-up mechanism. Specifically, such as Figure 1 As shown in the figure, in this embodiment, the curtain wall cleaning drone includes: a fuselage, a sensing system, a cleaning execution system (partially shown in the figure), and an edge processing system (not shown in the figure). The fuselage can be a multi-rotor structure, such as a hexacopter or octocopter. The fuselage is equipped with corresponding arms, a frame, landing gear, a battery compartment, a power module, a flight control module, a power management module, a wireless communication module, and other components. Existing models containing these components can be directly selected for the fuselage, and will not be elaborated further here.

[0015] The perception system of this curtain wall drone includes millimeter-wave radar 1-3 and visual cameras 1-7. Both are connected to the edge processing system via data cables or buses for data transmission. Visual cameras 1-7 are mounted in front of the drone's working face. Millimeter-wave radar 1-3 is mounted at the end of the cleaning execution system, emitting a beam towards the curtain wall plane to measure the distance between the cleaning execution system and the curtain wall in real time during the cleaning process. d ( tThe millimeter-wave radar 1-3 can operate in the 24GHz or 77GHz frequency band and has strong resistance to rain, fog, and glass reflection interference. The visual cameras 1-7 must have a visual range covering at least the curtain wall surface of the cleaning system and the area where it operates, for acquiring curtain wall image data. I ( t The visual camera 1-7 can be a regular RGB camera 1-7 or an industrial camera 1-7 with a wide dynamic range to adapt to outdoor strong light and reflective environments.

[0016] One of the innovations of this invention is the cleaning execution system, which includes a telescopic mechanism and a cleaning mechanism. In this embodiment, the telescopic mechanism includes a scissor-type telescopic rod assembly 1-6, a support pipe 1-5, a guide rail 1-8, and a hydraulic push rod 1-9 (or an electric push rod) for driving the telescopic extension and retraction of the scissor-type telescopic rod assembly 1-6. The cleaning mechanism includes a roller 1-2, a nozzle 1-1, a clean water tank, a solution tank, two water pipes 1-4, two water pumps, and two solenoid valves 1-10. The scissor-type telescopic rod assembly 1-6 consists of multiple sets of cross-arranged connecting rods, with adjacent connecting rods connected by pins, forming a typical "scissor-type" telescopic mechanism. One end of the scissor-type telescopic rod assembly 1-6 is fixed to the machine body, and the other end is connected to the roller 1-2 and the nozzle 1-1. The cylinder of the hydraulic push rod 1-9 is fixed to the machine body, and the piston is hinged to the intersection of the connecting rods in the middle of the scissor-type telescopic rod assembly 1-6. Figure 2 , Figure 3As shown, when the hydraulic push rod 1-9 extends, it drives the scissor-type telescopic rod assembly 1-6 to open, and moves the roller 1-2 and nozzle 1-1 away from the machine body. When the hydraulic push rod 1-9 retracts, it drives the scissor-type telescopic rod assembly 1-6 to fold and retract the roller 1-2 and nozzle 1-1. The scissor-type telescopic rod assembly 1-6 is made of a high-strength, moderately rigid material, such as aluminum alloy or steel. If necessary, rubber pads or elastic elements can be installed on the inner side of the connecting rod to provide a certain degree of elasticity and buffering capacity when subjected to external forces. One end of the support tube 1-5 is slidably connected to the machine body, and the other end extends towards the curtain wall to support the scissor-type telescopic rod assembly 1-6, roller 1-2, and nozzle 1-1. The support tube 1-5 can be made of aluminum alloy or carbon fiber round or rectangular tubes, which have high bending stiffness to reduce the deflection generated after the cleaning mechanism extends. To further improve the guiding accuracy, the guide rail 1-8 is set on the lower side or side wall of the support tube 1-5. The guide rail 1-8 is connected to the support tube 1-5 via a slider. The slider slides back and forth along the guide rail 1-8, stabilizing the movement trajectory of the support tube 1-5 during extension and retraction. In the cleaning mechanism, the nozzle 1-1 can be a duckbill nozzle. The roller 1-2 is arranged horizontally along the curtain wall. Furthermore, the roller 1-2 has an internal roller 1-2 shaft, with both ends connected to the ends of the scissor-type telescopic rod assembly 1-6 via bearings and brackets, allowing the roller 1-2 to rotate freely relative to the scissor-type telescopic rod assembly 1-6. The outer surface of the roller 1-2 is covered with an elastic wear-resistant layer, such as a rubber layer or a fiber bristle layer. Circumferential or spiral grooves can also be machined on the surface of the roller 1-2 to increase friction against stains and store detergent solution. The nozzle 1-1 is fixed in front of or above the roller 1-2, its spray direction pointing towards the curtain wall surface, forming a fan-shaped water curtain of a certain width to rinse the area cleaned by the roller 1-2. Both the clean water tank and the solution tank are mounted on the main body. Two water pipes 1-4 are arranged along the support pipe 1-5. One water pipe 1-4 serves as the clean water line, connecting one end to the clean water tank and water pump, and the other end to the nozzle 1-1 for spraying clean water onto the curtain wall. The other water pipe 1-4 serves as the detergent line, connecting one end to the solution tank and water pump, and the other end to the roller 1-2, allowing detergent to seep onto the surface of the roller 1-2. Two solenoid valves 1-10 are respectively installed on the two water pipes 1-4. The solenoid valves 1-10 can be either proportional or multi-stage adjustable, allowing for on-demand control of the clean water flow and detergent solution flow by adjusting the opening degree of the solenoid valves 1-10 according to actual conditions. In another embodiment, for lower curtain walls, the unit may not carry a water tank or solution tank, but can be placed on the ground. In this case, the required length of water pipe 1-4 is within an acceptable range, and the weight of the water tank and solution tank is removed from the unit, greatly improving the endurance.

[0017] The hardware portion of the edge processing system can be implemented using a microcontroller, embedded industrial computer, or automotive-grade computing platform. Its internally designed control logic can autonomously plan the drone's route and form a closed-loop control. This control logic is the automatic cleaning control method for the curtain wall cleaning drone in this invention, representing another innovation of the invention. Specifically, as... Figure 4 As shown, the automatic cleaning control method for curtain wall cleaning drones includes: Based on drone-captured images of the curtain wall I ( t The effective working area of ​​the curtain wall is identified. The method for identifying the effective working area of ​​the curtain wall includes: firstly... I ( t The curtain wall boundary is obtained by extracting straight line features and / or corner mesh features. Then, the variance of the distance between the drone and the curtain wall is calculated. If the variance is less than the distance stability threshold, the current... I ( t The working surface in the diagram is fitted to the curtain wall plane within the curtain wall boundary. The curtain wall boundary and the curtain wall plane constitute the effective working area. Of course, other algorithms can also be used for identification. The identified effective working area is then divided into... K The system is divided into several equal-sized covering units. The size of each covering unit can be customized, and it must completely fill the effective working area. All subsequent calculations will be performed at the covering unit level.

[0018] After determining the effective operating area and coverage unit, the effective coverage width of the drone cleaning is determined within the effective operating area. W Generate coverage cleaning path with path overlap coefficient P .like Figure 5 As shown, in this embodiment, a preset starting point is used as the beginning. P It is set to a serpentine path that moves back and forth along the horizontal direction of the curtain wall. Among them, P Lateral step size Δ x for: Δ x = W · α .

[0019] α This represents the horizontal path overlap coefficient, with a value ranging from 0.70 to 0.85. When layered cleaning is required, P longitudinal step size Δ y for: Δ y = W · β .

[0020] βThis represents the longitudinal path overlap coefficient, with a value ranging from 0.70 to 0.85. Furthermore, when cleaning reaches the curtain wall boundary, the UAV's turning and changing direction are completed within the safe zone after the curtain wall boundary retracts inward.

[0021] drone along P During the cleaning process, using the covered unit as the particle size, first use detergent and rollers 1-2 to roll and scrub the curtain wall within the covered unit area one to two times, then use clean water and spray nozzles 1-1 to rinse away any remaining dirt and detergent. Simultaneously, based on the before and after cleaning... I ( t Calculate the changes before and after ) k Cleanliness rating of each coverage unit S k and update the number k Cleanliness grid information for each coverage unit G k .Should G k Similar to the label of an overlay unit, it contains: the first k The number of times the coverage unit is cleaned, the first k The most recent coverage unit S k and the k Information on whether each coverage unit has passed cleaning. Specifically, S k The calculation methods include: according to I ( t ) Calculate the first k The average image brightness change of each coverage unit before and after cleaning Δ μ Image brightness standard deviation change Δ σ Image texture / edge energy change Δ E Among them, for those in I ( t Brightness and texture are identified in the image. Existing methods such as average pixel value method, HSV color space V channel method, histogram analysis, and gray-level co-occurrence matrix can be used to obtain the required brightness and texture. After obtaining the desired brightness and texture, the values ​​before and after cleaning are subtracted from the values ​​after cleaning to obtain the required data. Then, Δ... μ Δ σ Δ E After weighted fusion, the result is S k : S k = W 1·Δ μ + W 2·Δ σ + W 3·Δ E .

[0022] In the formula, W 1. W 2. W 3 are the weights of the corresponding items, and satisfy: W 1+ W 2+ W 3 = 1. When S k <threshold S th At that time, the judgment of the first k The first coverage unit failed the cleaning test, and the second... k Each coverage unit is preset with a maximum number of times. N The re-washing. If N If it still fails to meet the standards after the second wash, then the third wash will be... k Each coverage unit is added to the return visit queue. k ∈[1, K ].when S k ≥ S th When, then determine the first k All coverage units passed the cleaning test. N Cover units that pass the second cleaning are also deemed to have passed the cleaning. All of the above information is recorded in... G k middle.

[0023] In the present P After traversal, all coverage cells that are still unqualified after re-washing are in the revisit queue. At this point, a revisit path is generated based on the revisit queue, and revisit washing is performed according to the revisit path. For example... Figure 6 As shown, the specific process is as follows: the coordinates of the center points corresponding to all coverage units in the revisit queue constitute the revisit point set Ω: Ω = { p k,x,y} p k,x,y Indicates the first k The coordinates of the center point of each covering unit (the curtain wall surface can be considered as a two-dimensional plane) x , y (representing the x and y coordinates of a two-dimensional coordinate system, respectively), and simultaneously record the first... k Number of washes per coverage unit n k .Should n k For the first k The number of times a coverage unit is re-washed during a follow-up visit can be determined by... S k Determined after linear mapping and rounding, for example: n k =round( γ / S k `round(·)` represents the rounding function. When the current condition is complete... P Or, after other return visit trigger conditions are met, starting from the drone's current location, based on the current distance between the drone and the return visit point and... n k For the innermost Ω k Calculation priority of each coverage unit P k : P k = W s ·( S th - S k )+ W d · d ( p cur , p k,x,y )+ W n · n k .

[0024] In the formula, ( S th - S k This item indicates the degree to which the cleaning did not meet the standards. d ( p cur , p k,x,y (This refers to the current location of the drone) p cur To the k The distance between the center points of each coverage unit W s , W d , W n These are the weight coefficients for the corresponding terms. Based on P k Select the next visit point in the visit queue using the nearest neighbor greedy strategy. p next Perform a revisit and rewash, which means selecting the highest priority coverage unit among the remaining coverage units in the revisit queue. p k,x,y As p next The final follow-up path is as follows: Figure 7 As shown. Furthermore, during the follow-up cleaning process, the covered units being visited are still subjected to maximum [efforts / treatments].N The next wash. If N The cover unit still after the second cleaning S k < S th If the coverage unit fails to complete the coverage cleaning process, it will be added to the next revisit queue and the revisit and cleaning process will be repeated. Thus, the coverage cleaning path described above is completed. P The system includes a follow-up cleaning path to achieve comprehensive cleaning of the curtain wall. Each covered unit undergoes a process of "detergent brushing - water rinsing - cleanliness evaluation" to form a unit-level closed loop, ensuring the integrity and consistency of cleaning coverage.

[0025] It should be further noted that when the drone in this invention performs automatic cleaning operations, it preferably cleans layer by layer from top to bottom to avoid secondary pollution during the upper cleaning process. When bottom-up cleaning is required due to site / take-off / landing limitations, backflow pollution is reduced through water control and recovery / scraping structures, such as limiting the clean water flow, shortening the single rinsing time, and installing scraping strips or recovery covers / guide channels below the nozzles 1-1. In addition, if the area of ​​a single curtain wall is too large, a single curtain wall can be divided into several coverage cleaning paths. P Single item P Each coverage unit within the system executes a "detergent scrubbing—water rinsing—cleanliness evaluation" process, forming a unit-level closed loop. Currently... P Once completed, proceed to the next step. P The assignments, until all P Finish.

[0026] On the other hand, during the automatic cleaning process performed by the drone using the aforementioned method, the distance data between the cleaning system and the curtain wall is also used as a reference. d ( t Control the drone to approach or move away from the curtain wall. First, preset the safe distance range between the cleaning system and the curtain wall. d min , d max ],.like d ( t (less than minimum distance) d min Then, control the drone to move away from the curtain wall and / or slow down, while simultaneously controlling the telescopic mechanism to retract to reduce contact pressure. If d ( t (Greater than the maximum distance) d maxThe system controls the drone to approach and / or accelerate towards the curtain wall, while simultaneously extending the telescopic mechanism to generate a pre-set pressure between the drone and the curtain wall. This pre-pressure can be controlled between 150 and 300 N via hydraulic push rods 1-9. In an experiment on the exterior wall of a 20-story glass curtain wall building, the cleaning time was reduced by more than 40% compared to manual suspended platform cleaning, and the drone maintained a safe distance from the curtain wall at all times, without any collisions or scratches.

[0027] In another embodiment, the curtain wall cleaning drone has a hexacopter structure with a diagonal wheelbase of approximately 1500mm, a weight of approximately 20kg, and a maximum takeoff weight of approximately 40kg. The power system uses a 12S lithium battery pack as the primary power source, coupled with six brushless motors and folding propellers, providing a total maximum thrust of no less than 60kg to ensure sufficient margin even after mounting the cleaning execution system and cleaning fluid. The scissor-type telescopic boom assembly 1-6 consists of four sets of linked rods connected in series. Each set of rods is 250mm long and made of 6061-T6 aluminum alloy sheet with a thickness of 6mm, hinged at intersections using stainless steel pins. The scissor-type telescopic boom assembly 1-6 has a total length of approximately 450mm when retracted and approximately 900mm when fully extended, providing an effective telescopic stroke of 450mm. Support tube 1-5 is a rectangular cross-section carbon fiber tube with dimensions of 50mm × 30mm, a wall thickness of 2mm, and a length of 1.8m, of which 0.5m is fixed below the frame and 1.3m extends outward from the machine body. The cylinder of hydraulic push rod 1-9 is fixed to the machine body, and the piston rod end is connected to the connecting rod intersection point in the middle of the scissor telescopic rod via a hinge. Hydraulic push rod 1-9 has a rated thrust of 1.5kN and a stroke of 200mm. By extending or retracting the stroke, it drives the scissor telescopic rod from the fully retracted state to the fully extended state. Hydraulic push rod 1-9 is supplied with oil by a small DC hydraulic pump inside the machine body. The hydraulic pump has a rated pressure of 5MPa and a flow rate of 0.5L / min, and the extension / retraction is controlled by an electromagnetic reversing valve. The effective length of roller 1-2 is 600mm. The core material of roller 1-2 is an aluminum alloy cylinder with an outer diameter of 80mm. The outer surface is covered with a 10mm thick foamed rubber layer. The outer surface of the rubber layer has spiral shallow grooves, 2mm deep and spaced 15mm apart, used to store detergent solution and enhance friction. Roller 1-2 is connected to the bracket at both ends via bearings, allowing it to rotate freely in the horizontal direction. The bracket is bolted to the mounting base at the end of the scissor-type telescopic rod assembly 1-6 to allow for replacement of roller 1-2 of different specifications. Spray head 1-1 is installed below roller 1-2. Its nozzle length is approximately 500mm, and the nozzle outlet width is approximately 1mm. It forms a fan-shaped water curtain through an internal flow-dividing cavity, with a spray angle of approximately 30°–45°. The coverage width of the fan-shaped water curtain on the curtain wall surface is slightly greater than the length of roller 1-2, ensuring that the cleaning area of ​​roller 1-2 is completely rinsed. Spray head 1-1 is connected to water pipe 1-4 via stainless steel clamps. Two water pipes 1-4 connect to a clean water tank and a solution tank located above the machine body, both with a volume of 10L. The clean water tank is connected to a high-pressure pump with a rated pressure of 0.8MPa and a flow rate of 6L / min, used to supply water to the duckbill nozzle 1-1. The solution tank is connected to a metering pump with a flow rate of 0.5-1L / min, used to supply detergent solution to the inside of the drum 1-2. In this embodiment, the detergent is a neutral surfactant solution with a mass fraction of 1%-3%.Proportional solenoid valves 1-10 are connected in series on the water supply lines 1-4 and the detergent supply lines, with a control accuracy better than 5%. Millimeter-wave radar 1-3 is fixed on a bracket above roller 1-2, with its antenna pointing perpendicularly to the curtain wall surface. Its ranging range is 0.2–5 m, with a ranging accuracy better than ±3 cm, and a sampling frequency of 50 Hz. Visual camera 1-7 is fixed on a gimbal on the underside of the unit, with a field of view of approximately 90°, a resolution of 1920×1080, and a frame rate of 30 fps. It can operate normally under strong backlight conditions.

[0028] In another embodiment, a computer-readable storage medium is also provided. This computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the automatic cleaning control method for the curtain wall cleaning drone described in the above embodiments. The computer-readable storage medium may include, but is not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above.

[0029] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0030] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. An automatic cleaning control method for a curtain wall cleaning drone, characterized in that, It includes: Based on drone-captured images of the curtain wall I ( t The effective working area of ​​the curtain wall is identified and divided into... K A number of equally divided coverage units; Within the effective operating area, based on the effective coverage width of the drone cleaning W Generate coverage cleaning path with path overlap coefficient P ; Along the drone P During the cleaning process, based on the before and after cleaning I ( t ) change calculation k Cleanliness rating of each coverage unit S k ;when S k <threshold S th At that time, for the first k Each coverage unit is preset with a maximum number of times. N Repair and clean; like N Even after the second wash S k < S th , will the k Each coverage unit is added to the return visit queue; k ∈[1, K ]; current P After traversal, a return path is generated based on the return queue, and return shuffling is performed based on the return path: the coordinates of the center points corresponding to all covered units in the return queue constitute the set of return points Ω: Ω = { p k,x,y }, and record the first k Number of washes per coverage unit n k Starting from the drone's current location, based on the current distance between the drone and the revisit point and... n k For the innermost Ω k Calculation priority of each coverage unit P k ;based on P k Select the next visit point according to the nearest neighbor greedy strategy. p next Conduct follow-up visits and re-washing; When all coverage units satisfy S k ≥ S th The cleaning process ends at that time.

2. The automatic cleaning control method for the curtain wall cleaning drone according to claim 1, characterized in that, Methods for identifying the effective working area of ​​a curtain wall include: right I ( t Extract straight line features and / or corner mesh features to obtain the curtain wall boundary; Calculate the variance of the distance between the drone and the curtain wall; if the variance is less than the distance stability threshold, then... I ( t The working surface in the diagram is fitted to the curtain wall plane within the curtain wall boundary; the curtain wall boundary and the curtain wall plane constitute the effective working area.

3. The automatic cleaning control method for the curtain wall cleaning drone according to claim 1, characterized in that, Coverage cleaning path P The design methods include: Will cover the cleaning path P Set to a serpentine path that moves back and forth along the horizontal direction of the curtain wall; in, P Lateral step size Δ x Longitudinal step size Δ y They are respectively: D x = W · α ; D y = W · β ; In the formula, α Indicates the horizontal path overlap coefficient. β Indicates the longitudinal path overlap coefficient; When the cleaning reaches the edge of the curtain wall, the drone changes course and turns.

4. The automatic cleaning control method for the curtain wall cleaning drone according to claim 1, characterized in that, Cleanliness rating S k The calculation methods include: according to I ( t ) Calculate the first k The average image brightness change of each coverage unit before and after cleaning Δ μ Image brightness standard deviation change Δ σ Image texture changes Δ E ; Δ μ Δ σ Δ E After weighted fusion, the result is S k .

5. The automatic cleaning control method for the curtain wall cleaning drone according to claim 4, characterized in that, when S k ≥ S th At that time, the judgment of the first k Each coverage unit has passed cleaning; when S k < S th At that time, the judgment of the first k One coverage unit failed the cleaning test; No. k Each coverage unit is marked with corresponding cleanliness grid information. G k ; G k Includes: the k Number of cleaning cycles per coverage unit, most recent cleaning cycle S k And information on whether the coverage unit has been cleaned properly.

6. The automatic cleaning control method for the curtain wall cleaning drone according to claim 1, characterized in that, Priority P k The calculation formula is: P k = W s ·( S th － S k )+ W d · d ( p cur , p k,x,y )+ W n · n k ； In the formula, d ( p cur , p k,x,y (This refers to the drone's current location up to the [number]th position) k The distance between the center points of each coverage unit W s , W d , W n These are the weight coefficients for the corresponding items; And / or, during the follow-up cleaning process, maximize the coverage of the covered units being followed up. N The next wash; like N The cover unit still after the second cleaning S k < S th If the corresponding coverage unit is added to the next revisit queue, the revisit and cleanup process will be repeated.

7. The automatic cleaning control method for the curtain wall cleaning drone according to claim 1, characterized in that, The automatic cleaning control method also includes collecting distance data between the drone cleaning execution system and the curtain wall. d ( t Based on the preset safe distance range, d ( t Control the drone to approach or move away from the curtain wall; Among them, if d ( t If the distance is less than the minimum distance within the safe distance range, control the drone to move away from the curtain wall and / or slow down; like d ( t If the distance exceeds the maximum distance of the safe distance range, control the drone to approach the curtain wall and / or accelerate.

8. A curtain wall cleaning drone, characterized in that, It includes: Organism; The sensing system, installed on the machine body, is used to acquire... I ( t )and d ( t ); The cleaning execution system, installed on the machine body, is used to clean the curtain wall; An edge processing system, mounted on the airframe, uses the automatic cleaning control method of the curtain wall cleaning drone as described in any one of claims 1 to 7 to control the airframe along... P The system moves along the return path and controls the cleaning execution system to perform a cleaning and rewash operation.

9. The curtain wall cleaning drone according to claim 8, characterized in that, The cleaning execution system includes: a telescopic mechanism and a cleaning mechanism; One end of the telescopic mechanism is connected to the machine body, and the other end is connected to the cleaning mechanism; The cleaning mechanism includes: a roller, a nozzle, a clean water tank, and a solution tank; the roller and nozzle are installed at the end of the telescopic mechanism away from the machine body; the clean water tank and the solution tank are installed on the machine body; the clean water tank is used to deliver clean water to the nozzle via a water pump; the solution tank is used to deliver detergent to the roller via a water pump.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the automatic cleaning control method for the curtain wall cleaning drone as described in any one of claims 1 to 7.

Images