Adjustable frequency oscillation end effector and force control cleaning method for bridge deck cleaning

Abstract

The application discloses a bridge beam bottom cleaning adjustable frequency oscillation end effector and force control cleaning method, and belongs to the technical field of bridge maintenance. The end effector comprises a connecting flange, a frequency modulation oscillation driving module, a flexible transmission rod and a cleaning brush head module, integrates a 0.1N precision force sensor and a low-pressure spraying assembly, and the oscillation frequency is adjustable at 0-50 Hz. The cleaning method adopts a lightweight visual model to identify stains, matches a stain-cleaning parameter knowledge base, realizes closed-loop regulation of contact force through impedance control and model predictive control, automatically reduces the oscillation amplitude and adjusts the working pose when the contact pressure exceeds the 20N threshold, and realizes multi-stage safety protection. The application has the advantages of compact structure, small load, adaptation to complex curved surfaces, self-adaptation to different stain types, solution to the problems of high risk of manual cleaning, poor adaptability of existing equipment and easy damage to bridge structure, and application to unmanned aerial vehicle double-mechanical-arm cooperative bridge beam bottom intelligent cleaning.

Description

Technical Field

[0001] This invention relates to the field of bridge engineering maintenance technology, specifically to an adjustable frequency oscillation end effector for cleaning the bottom of bridge beams, and a force-controlled cleaning method based on the actuator and combined with the collaborative operation of two robotic arms of a UAV. Background Technology

[0002] As a critical infrastructure in transportation networks, bridges have their beam undersides exposed to the outdoor environment for extended periods, making their surfaces prone to accumulating various types of contaminants such as dust, oil, rust, and moss. This buildup not only affects the overall appearance of the bridge but also accelerates concrete carbonization, steel reinforcement corrosion, and coating aging, shortening the bridge's structural lifespan and increasing subsequent maintenance costs. Traditional bridge beam underside cleaning primarily relies on manual suspended platforms, aerial work platforms, or "spider-man" operations at height, which suffer from low efficiency, high labor costs, significant risk of falls from heights, and poor adaptability to harsh environments, making it difficult to meet the needs of large-scale, routine bridge maintenance.

[0003] Existing technologies have proposed solutions for automated bridge cleaning using drones equipped with cleaning devices, but these solutions generally suffer from the following technical drawbacks:

[0004] The cleaning methods are limited, mostly using fixed-speed rotating brush heads or high-pressure water jets. They cannot be adaptively adjusted according to the characteristics of different stains such as oil, moss, rust, and dust, resulting in unstable cleaning effects. Furthermore, high-pressure water jets can easily damage the protective coating on the bridge surface.

[0005] The lack of a precise force control protection mechanism makes it impossible to monitor and stably control the contact force between the brush head and the bottom surface of the beam in real time. This can easily lead to problems such as excessive contact force damaging the bridge or actuator, or insufficient contact force resulting in incomplete cleaning. The collaborative force control protection between the drone, robotic arm, and end effector has not been achieved.

[0006] The end effector is bulky and has low integration. It is mostly a single-arm operation layout, which limits the operating range and is prone to disturbing the flight attitude of the drone. At the same time, the force sensing, water mist channel and cleaning brush head are not integrated at the same end, resulting in high system control complexity and poor flight and operation stability.

[0007] The visual recognition module lacks precision, making it difficult to achieve high-precision, real-time recognition and classification of multi-scale stains in complex backgrounds on UAV-borne edge computing devices, and thus unable to provide reliable data support for adaptive adjustment of cleaning parameters.

[0008] Therefore, the industry urgently needs a bridge beam bottom cleaning device and method that is compact, highly integrated, and adaptable, capable of working collaboratively with dual robotic arms of drones, and also features intelligent stain recognition, adaptive matching of cleaning parameters, and closed-loop protection of force control throughout the process. Summary of the Invention

[0009] The purpose of this invention is to overcome the shortcomings of the existing technology and provide an adjustable frequency oscillation end effector and force-controlled cleaning method for cleaning the bottom of bridge beams, solving the technical problems of poor adaptability, lack of intelligent force control, low structural integration, insufficient visual recognition accuracy, and easy damage to bridge structure of existing cleaning equipment.

[0010] To achieve the above objectives, the present invention adopts the following technical solution:

[0011] Variable frequency oscillation end effector technology solution

[0012] An adjustable frequency oscillating end effector for cleaning the bottom of bridge beams, adapted to a dual-manipulator collaborative operation system for unmanned aerial vehicles, includes a connecting flange, a one-dimensional drive module, a flexible transmission component, and a cleaning brush head module connected in sequence.

[0013] The connecting flange is used to connect to the wrist joint of the dual robotic arms of the UAV, and is equipped with a quick-release interface, which enables the actuator to be quickly disassembled and maintained.

[0014] The one-dimensional drive module is fixedly connected to the connecting flange and has a built-in servo motor. It converts the servo motor's angular motion into reciprocating linear oscillation through an eccentric wheel, crank-slider, or rocker arm mechanism, which can generate oscillation power with a continuously adjustable frequency of 0-50Hz. The oscillation amplitude can be dynamically adjusted by the control system.

[0015] The flexible transmission component includes a flexible carbon fiber tube, one end of which is connected to the power output end of the one-dimensional drive module, and the other end is connected to the cleaning brush head module through a clamping sleeve, inner liner, threaded end or snap-on end. It is covered with a flexible hose sheath to transmit oscillating power and provide lateral compliance, buffer contact impact, and adapt to the complex curved surface contour of the bridge beam bottom.

[0016] The cleaning brush head module is detachably connected to the end of a flexible carbon fiber tube and includes a brush head body, a force sensing component, and a low-pressure water mist channel assembly. The brush head body is equipped with quick-release interchangeable dry brush heads, wet sponge heads, rigid nylon brush heads, steel wire brush heads, and abrasion-resistant rubber brush heads, in strip, plate, or disc shapes. The force sensing component includes a single-axis or triaxial force sensor with an accuracy of 0.1N for real-time detection of the normal contact force between the brush head and the cleaning surface. The low-pressure water mist channel assembly includes a water inlet, a flow guide cavity, and a linear, annular, or matrix-distributed array of nozzles, which can output low-pressure atomized cleaning fluid in wet cleaning mode.

[0017] Force Control Cleaning Method Technical Solution

[0018] A collaborative intelligent control cleaning method for a dual-manipulator UAV based on the aforementioned adjustable frequency oscillating end effector, integrating computer vision recognition, adaptive parameter matching, impedance control, and model predictive control technologies, includes the following steps:

[0019] S1: The UAV flight platform flies to the target cleaning area under the bridge beam and achieves stable hovering through the flight control system. The airborne high-definition imaging sensor collects images of the beam bottom surface in real time and transmits them to the airborne computing unit.

[0020] S2: A lightweight object detection model based on the improved SH-YOLO architecture is used to process images in real time. This model integrates a StarNet lightweight backbone network based on star-shaped operation, a feature fusion network including multi-scale dilated convolution feature fusion (MSDCFB) and coordinate attention (MCCA) mechanisms, and a dynamic gated hypergraph module (DHG), which can accurately identify the type, location, area and cleaning priority of stains.

[0021] S3: Based on the stain recognition results, call the preset "stain type - cleaning parameter" mapping knowledge base to automatically match the optimal oscillation frequency, initial contact force, brush head type and cleaning mode;

[0022] S4: The V-shaped dual robotic arms of the drone plan the movement trajectory according to the location of the stains, and adopt a master-slave collaborative control strategy to drive the end effector to accurately position, so that the brush head can stably adhere to the cleaning surface of the beam bottom with a preset initial contact force.

[0023] S5: The one-dimensional drive module is activated to output oscillation at the matching frequency. The force sensor collects the contact pressure in real time with an accuracy of 0.1N. The system constructs an impedance control framework and combines it with model predictive control to achieve decoupled control of the contact force and the position of the robotic arm end and the hovering position of the drone, so that the brush head smoothly follows the curved surface and maintains a stable contact force.

[0024] S6: Compare the real-time detected contact pressure with the preset 20N safety threshold. If the threshold is exceeded, immediately trigger multi-level linkage protection: reduce the oscillation amplitude to 60% of the original amplitude, increase the drone hovering height by 5cm, control the robotic arm to retract slightly, and restore normal operating parameters after the contact force returns to the safe range.

[0025] S7: Switches between dual robotic arm zone operation or alternating operation mode according to the distribution of stains, and automatically switches between dry cleaning or wet cleaning mode according to the type of stains. In wet mode, low-pressure water mist spray is turned on simultaneously.

[0026] S8: After completing the cleaning of the target area, the end effector stops oscillating and spraying, the dual robotic arms reset, and the drone flies to the next area or returns safely.

[0027] Beneficial effects of the present invention

[0028] Strong cleaning adaptability: Real-time stain classification is achieved through a lightweight, high-precision visual model. Combined with a "stain-parameter" knowledge base, adaptive matching of oscillation frequency, contact force, brush head type, and cleaning mode is achieved, which has a high cleaning ability for dust, oil stains, moss, and rust.

[0029] Force control is safe and reliable: It integrates a 0.1N high-precision force sensor, uses impedance control and model predictive control to achieve force-position decoupling, sets a 20N overload protection threshold, and cooperates with a multi-level linkage protection mechanism to avoid damage to the bridge structure and actuator;

[0030] Compact structure and high adaptability: It adopts the "drive + flexible transmission + modular brush head" design, which integrates force sensing, spraying and cleaning into one unit. It is lightweight and has a small load. The flexible carbon fiber tube can adapt to complex curved surfaces, and the V-shaped double robotic arm improves the working range and flight stability.

[0031] The system is highly intelligent: it achieves a high degree of coordination between drones, dual robotic arms, end effectors, vision modules, and force control modules, supporting fully automatic cleaning, online parameter adjustment, anomaly protection, and continuous operation, significantly reducing manual intervention and improving bridge maintenance efficiency and safety. Attached Figure Description

[0032] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments will be briefly described below.

[0033] Figure 1 This is a schematic diagram of the overall structure of the collaborative system of the end effector and the UAV dual robotic arm of the present invention;

[0034] Figure 2 This is a schematic diagram of the overall structure of the adjustable frequency oscillation end effector of the present invention;

[0035] Figure 3 This is a cross-sectional structural diagram of the cleaning brush head module of the present invention;

[0036] Figure 4 This is a flowchart illustrating the overall process of the force-controlled cleaning method of the present invention.

[0037] Figure 5 This is a diagram illustrating the overall architecture of the visual recognition model based on the SH-YOLO architecture of this invention.

[0038] Figure 6 This is a diagram illustrating the Star Block structure and Star Operation mechanism of this invention.

[0039] Figure 7 This is a structural diagram of the MSDCFB module (multi-scale dilated convolution) and MCCA mechanism (multi-channel coordinate attention) of the present invention;

[0040] Figure 8 This is a schematic diagram of the Dynamic Gated Hypergraph (DHG) module structure of the present invention. Specific implementation methods

[0041] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0042] System overall composition

[0043] The UAV dual-manipulator collaborative cleaning system adapted to this invention includes: a multi-rotor UAV flight platform, a V-shaped symmetrical dual-manipulator assembly, a frequency-adjustable oscillating end effector, a water supply and spray module, a vision recognition module, a force control module, an airborne control unit, and a power supply and communication module. The flight platform has high-precision hovering and attitude stabilization capabilities, and the dual manipulators are equipped with shoulder joints, elbow joints, and wrist joints, enabling master-slave, collaborative, zoned, and alternating operations.

[0044] Specific structure of the adjustable frequency oscillation end effector

[0045] like Figure 2 As shown, the end effector includes a connecting flange 1, a one-dimensional drive module 2, a flexible transmission assembly 3, and a cleaning brush head module 4.

[0046] Connecting flange 1 is equipped with a quick-release interface, which is rigidly connected to the wrist joint of the robotic arm, making it easy to assemble and disassemble.

[0047] The one-dimensional drive module 2 has a built-in MG995 servo motor, which converts the rotary motion into a continuously adjustable reciprocating oscillation of 0 to 50 Hz through a crank-slider mechanism. The oscillation frequency is precisely adjusted by the PWM signal, and the amplitude can be dynamically controlled.

[0048] The main body of the flexible transmission component 3 is a flexible carbon fiber tube 31. One end is fixed to the output end of the one-dimensional drive module 2, and the other end is snapped to the cleaning brush head module 4. The outside is covered with a flexible hose sheath, which can transmit vibrations and bend to adapt to curved surfaces, thus buffering impacts.

[0049] like Figure 3 As shown, the cleaning brush head module 4 includes a brush head body 41, a force sensing component 42, and a low-pressure water mist channel component 43.

[0050] The brush head body 41 can be quickly disassembled and replaced with hard nylon bristles, steel wire bristles, soft bristles, wear-resistant rubber head, and sponge head.

[0051] The force sensing component 42 uses a 0.1N high-precision single-axis force sensor, which is integrated inside the brush head to collect the normal contact force in real time.

[0052] The low-pressure water mist channel assembly 43 has a built-in flow guide cavity and a linear nozzle array. It is connected to the airborne water tank and a micro pump through a pipeline and can output low-pressure atomized water, emulsion solution or weakly acidic buffer solution.

[0053] like Figure 4 As shown, the force-controlled cleaning method based on the collaborative operation of two robotic arms of a drone, as described in this invention, comprises the following specific steps:

[0054] S1: Drone arrival and visual acquisition

[0055] The drone flight platform flies to the bottom area of ​​the target bridge beam and hovers stably through the flight control system. The airborne imaging sensor (high-definition camera) collects images of the beam bottom surface in real time, and the image data is transmitted to the computing unit of the vision and control components.

[0056] S2: High-precision visual recognition and classification of stains

[0057] To achieve real-time, high-precision identification of stains (oil, moss, rust, dust) on the bottom of beams on a limited computing platform using a dual-robotic arm collaborative operation of a drone, this invention deploys an improved lightweight target detection network (based on the SH-YOLO architecture) in the main controller. Its overall network structure is as follows: Figure 5 As shown, a dynamic hypergraph and multi-scale feature fusion mechanism are introduced. The specific model architecture and implementation details are as follows:

[0058] To reduce computational complexity and improve feature extraction efficiency in environments with uneven lighting and motion blur at the bottom of beams, the visual model uses StarNet, based on star operation, instead of the traditional CSPDarknet53 as the backbone network.

[0059] The backbone network mainly consists of a Stem layer and four Stage layers. Its core components are multiple cascaded StarBlocks, and its specific topology and star-topology operation mechanism are as follows: Figure 6 As shown.

[0060] By implementing star schema operations through element-wise multiplication, input features can be mapped to a high-dimensional nonlinear feature space with low computational cost, thereby recursively increasing the implicit feature dimension.

[0061] This enables the network to efficiently capture the morphological features of stains on the bottom of beams (such as mottled oil stains or small mold spots with blurred edges) while meeting the real-time requirements of the UAV-borne computing platform (maintaining high FPS).

[0062] In the feature fusion network (Neck), this invention constructs a path aggregation feature pyramid network (HAM-PAFPN) integrating a hybrid attention mechanism, and fuses a multi-scale dilated convolutional feature fusion module (MSDCFB) and a multi-channel coordinate attention module (MCCA). Its detailed interleaved fusion structure is as follows: Figure 7 As shown:

[0063] 1. Multi-scale dilated convolutional feature fusion (MSDCFB): This method utilizes dilated convolutions with different dilation rates to extract features in parallel, aiming to expand the receptive field without sacrificing spatial resolution.

[0064] This module is used to adaptively aggregate stain context information at different scales on the bottom of beams (such as large areas of rust and localized small dust accumulation points).

[0065] 2. Hybrid Attention Mechanism (HAM & MCCA): Embedding Hybrid Attention Mechanism (HAM) and Multi-Channel Coordinate Attention Mechanism (MCCA) in the feature aggregation path.

[0066] By adaptively recalibrating and weighting the feature maps in both channel and spatial dimensions, the model focuses highly on the spatial location and key textures of the stains themselves, effectively suppressing complex background noise such as rough concrete at the bottom of bridge beams, mesh cracks, and shadow interference.

[0067] To address the complex stains on the bottom of beams that are geometrically irregular and discontinuously distributed, the visual model introduces a Dynamically Gated Hypergraph (DHG) module at the feature layer. Its overall topology is as follows: Figure 8 As shown.

[0068] Traditional graph structures can only model a one-to-one relationship between two points, while hypergraphs can connect multiple nodes simultaneously through hyperedges, enabling nonlocal topological relationship modeling of many-to-many relationships.

[0069] The module dynamically constructs the correlation matrix between stain feature pixel nodes and hyperedges, and uses a gating mechanism to adaptively and dynamically update the contribution of different nodes during feature transmission.

[0070] Through hypergraph computing, the model can capture deep associations and global contextual dependencies between cross-regional and discontinuous stains, thereby significantly improving the edge detection accuracy of hidden stains and complex stains and greatly reducing the false negative rate.

[0071] S3: Cleaning parameter adaptive matching

[0072] Based on the identified stain type, the main controller calls a preset "stain type - cleaning parameter" mapping knowledge base to match the corresponding optimal cleaning parameters. The specific implementation parameters are as follows:

[0073] Oil stains: Vibration frequency 45Hz (40-50Hz range), initial contact force 12N, use hard nylon bristles, start wet cleaning mode, spray atomized water containing mild emulsifier;

[0074] Moss: Vibration frequency 25Hz (range 20-30Hz), initial contact force 8N, use wear-resistant rubber brush head, force start wet cleaning mode, spray water;

[0075] Rust: Vibration frequency 35Hz (range 30-40Hz), initial contact force 15N, use wire brush head, select dry cleaning or wet cleaning mode with spraying weak acid buffer solution according to the degree of rust.

[0076] Dust: Vibration frequency 10Hz (5-15Hz range), initial contact force 5N, use soft brush head, start dry cleaning mode, and turn on suction device if necessary.

[0077] S4: Dual robotic arm collaborative positioning and fitting

[0078] The main controller plans the motion trajectory of the dual robotic arms based on the location coordinates of the stain. The V-shaped dual robotic arm assembly of the drone achieves precise positioning of the end effector under the coordinated drive of the shoulder, elbow, and wrist joints. A master-slave control strategy is adopted, with the main robotic arm guiding the brush head to adhere to the stain surface with a matching initial contact force, and the secondary robotic arm providing attitude assistance and support to ensure stable contact between the brush head and the bridge surface. At the same time, the initial contact force value is fed back to the force control module.

[0079] S5: Model Predictive Impedance Force Controlled Cleaning Execution

[0080] The main controller sends instructions to the one-dimensional drive module 2, causing it to start working according to the matched oscillation frequency; at the same time, the force sensor 42 continuously collects contact pressure data with an accuracy of 0.1N and feeds it back to the force control module; the force control module constructs an impedance control framework, sets the desired inertia, damping and stiffness coefficient matrix, and combines it with model predictive control (MPC) based on Laguerre function. According to the contact force feedback, the state information of the UAV and the robotic arm, it continuously optimizes and predicts the future contact force trajectory, dynamically adjusts the joint angle of the robotic arm and the hovering micro-displacement of the UAV, realizes the decoupled control of contact force and position, ensures that the contact force is stable within the matched initial contact force range, and the brush head smoothly follows the contour of the bridge surface for cleaning.

[0081] S6: Real-time contact force assessment

[0082] The force control module continuously compares the actual contact pressure detected by the force sensor 42 with the preset safety threshold of 20N in real time to determine whether the overload protection mechanism is triggered.

[0083] S7: Dynamic overload collaborative protection and adaptive adjustment

[0084] If the actual contact pressure exceeds the 20N safety threshold, the main controller will immediately trigger a multi-level linkage protection mechanism:

[0085] Send a command to the one-dimensional drive module 2 to reduce the oscillation amplitude to 60% of the original amplitude, and at the same time send a command to the water supply and spray components to stop the spray output;

[0086] Send a command to the drone flight control system to raise the overall hovering height of the drone by 5cm;

[0087] Send commands to the dual robotic arm assembly to control the main robotic arm to slightly raise and retract, while the secondary robotic arm maintains its posture for assistance.

[0088] Once the contact pressure detected by the force sensor returns to the matching safe range, the main controller gradually controls each component to restore its original oscillation amplitude, spray state, and position, and continues the cleaning operation.

[0089] S8: Dual robotic arm collaborative operation and mode switching

[0090] Depending on the area and distribution of the stains, the dual robotic arms can switch to a zoned operation mode to expand the cleaning coverage. During the cleaning process, the controller can automatically switch cleaning parameters and cleaning modes based on the real-time identification of stain type changes, such as automatically switching from dust cleaning to rust cleaning without manual intervention. The water supply and spray components precisely control the water flow through a micro pump, and the water supply pipeline is arranged along the robotic arm and fixed by a fixed clamp to reduce swaying and wear.

[0091] S9: Cleaning Operation Completed and Safe Exit

[0092] Once the visual recognition module detects that the target area has been cleaned, the main controller instructs the one-dimensional drive module to stop working and shut off the spray (if it is on). The dual robotic arms then drive the end effector to separate from the bridge surface. The UAV flight control system controls the flight platform to adjust its position and enter the next cleaning area or safely exit after completing the operation.

[0093] Through the above technical solution, the present invention achieves efficient, intelligent and safe collaborative cleaning of dirt on the bottom of bridge beams by dual robotic arms of drones, effectively solving the problems of low efficiency and high risk of manual cleaning, as well as the poor adaptability of existing equipment, easy damage to bridges and lack of collaborative control.

Claims

1. A frequency-adjustable oscillating end effector for cleaning the bottom of bridge beams, characterized in that, include: The connecting flange, frequency-modulated oscillation drive module, flexible transmission component, and cleaning brush head module are sequentially and fixedly connected. The connecting flange is equipped with a quick-release interface for connecting to the wrist joint of the drone's robotic arm; The frequency-modulated oscillation drive module has a built-in servo motor, which converts angular motion into reciprocating oscillation through a mechanical transmission mechanism, and outputs oscillation power with an adjustable frequency of 0-50Hz and a dynamically adjustable amplitude; the flexible transmission component includes a flexible carbon fiber tube, which is covered with a flexible hose sheath. The cleaning brush head module includes a brush head body, a force sensor assembly, and a low-pressure spray assembly; the force sensor assembly has an accuracy of 0.1N and is used to detect contact force in real time; the low-pressure spray assembly is provided with a water inlet and a spray nozzle array; the brush head body is equipped with a quick-release and replaceable cleaning brush head.

2. The end effector according to claim 1, characterized in that: The frequency modulation oscillation drive module uses an MG995 servo motor, and the oscillation frequency is adjusted by a PWM signal. The flexible carbon fiber tube can be replaced by a flexible rod or an equivalent elastic component. The cleaning brush head includes at least one of the following: hard nylon bristles, steel wire bristles, soft bristle brush head, abrasion-resistant rubber brush head, and wet sponge head.

3. A method for force-controlled cleaning of the bottom of a bridge beam based on the end effector according to any one of claims 1 or 2, characterized in that, Includes the following steps: S1: The drone flies to the target area under the bridge beam and hovers stably to collect surface images; S2: The airborne imaging sensor acquires images of the bottom of the bridge beams, and a lightweight visual recognition model is used to extract and classify the features of the images to accurately identify the type, location, area and cleaning priority of the stains; S3: Based on the type of stain, call the stain-cleaning parameter knowledge base to match the oscillation frequency, initial contact force, brush head type and cleaning mode; S4: Control the robotic arm to drive the end effector for precise positioning, so that the brush head adheres to the cleaning surface with initial contact force; S5: Start oscillating cleaning, monitor the contact pressure in real time through a force sensor, and use impedance control combined with model predictive control to achieve decoupled adjustment of contact force and position; S6: When the contact pressure exceeds the 20N safety threshold, trigger multi-level protection: reduce the oscillation amplitude to 60% of the original amplitude, increase the drone hovering height by 5cm, and control the robotic arm to retract until the contact force returns to the safe range.

4. The force-controlled cleaning method according to claim 3, characterized in that, The matching rules for the "stain type - cleaning parameter" mapping knowledge base mentioned in step S3 are as follows: For oil stains, use a 40-50Hz oscillation frequency, 12N initial contact force, hard nylon bristle brush head, and start a wet cleaning mode that sprays water containing emulsifier atomized water. For moss, use a 20-30Hz oscillation frequency, an initial contact force of 8N, a wear-resistant rubber brush head, and activate a wet cleaning mode that sprays water. For rust, use an oscillation frequency of 30-40Hz, an initial contact force of 15N, and a wire brush head. Depending on the degree of rust, choose either a dry cleaning mode or a wet cleaning mode that sprays a weak acid buffer solution. For dust, use an oscillation frequency of 5-15Hz, an initial contact force of 5N, a soft brush head, and activate the dry cleaning mode.

5. The force-controlled cleaning method according to claim 3, characterized in that: The dual robotic arm assembly of the drone is arranged in a V-shape and has shoulder joints, elbow joints, and wrist joints. It can adopt master-slave, collaborative, zoned operation, or alternating operation strategies.

6. The force-controlled cleaning method according to claim 3, characterized in that, It also includes water supply and spraying steps. Cleaning fluid is delivered to the low-pressure spraying components through a water tank, a micro pump, and water supply pipelines. The water supply pipelines are arranged along the drone body and robotic arm, and are fixed by hose sheaths and clamps. The micro pump controls the water flow rate and the spray on / off time. The wet cleaning mode includes spraying and then wiping, or spraying and wiping simultaneously, while the dry cleaning mode only performs vibration wiping.

Images