Intelligent robot for identifying and cleaning bulkhead pollutants and bulkhead cleaning method

Abstract

The invention discloses an intelligent robot for identifying and cleaning bulkhead pollutants and a bulkhead cleaning method. The robot body carries a positioning navigation system and is equipped with a track assembly, a magnetic adsorption wheel, a mechanical arm assembly and a cleaning brush assembly. The positioning navigation system is combined with inertial navigation, visual SLAM and graph neural network loopback detection error correction, and a visual unit identifies pollutants through deep learning and other technologies. The cleaning method comprises the following steps: positioning and mapping, capturing an image, stretching out the camera as required, moving and fixing the robot, cleaning at an adaptive rotating speed, correcting the position in real time, and determining whether to continue after secondary detection. The robot can adjust the posture in a self-adaptive mode, and bulkhead cleaning is efficiently completed.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of underwater robots, in particular to an intelligent robot for cabin wall pollutant identification and cleaning and a cabin wall cleaning method. BACKGROUND

[0002] With the rapid development of the fields of ocean engineering, ship maintenance and underwater facility management, the long-term underwater operation of the inner wall of the cabin and the closed structure leads to the easy attachment of marine organisms, rust products and other pollutants on the surface. These pollutants not only affect the corrosion resistance of the cabin structure, but also may affect the normal work of sensors and equipment, thereby reducing the operation efficiency and safety of the entire platform. Therefore, it is of great engineering significance to carry out efficient and low-loss underwater cabin wall cleaning operations.

[0003] Traditional underwater cleaning operations mostly rely on divers or remotely operated underwater robots for mechanical brushing, spray cleaning or chemical treatment, but the cabin wall area is usually narrow in space, complex in structure and obviously disturbed by water flow, which limits the maneuverability of manual operation and traditional ROVs, and there are problems of low efficiency, high risk, etc. In such a closed underwater environment, the robot positioning highly depends on visual SLAM and inertial navigation system INS, but the former is prone to failure in low light, blur, lack of feature points in the underwater environment, and the latter has the problem of cumulative drift, making it difficult to achieve continuous and stable accurate positioning, resulting in insufficient autonomous operation capability. In addition, the types of cabin wall pollutants are complex and unevenly distributed, and the traditional method mostly uses indiscriminate cleaning, which cannot achieve adaptive matching of pollution identification and cleaning intensity, and is easy to cause over-cleaning, resource waste and even damage to the cabin coating. Therefore, there is an urgent need for an underwater cabin wall cleaning robot system that integrates high-precision underwater positioning, intelligent pollution identification and adaptive cleaning capability to meet the actual needs of automatic, efficient and safe cleaning of the underwater cabin wall in a closed space. SUMMARY

[0004] The purpose of the present application is to provide an intelligent robot for cabin wall pollutant identification and cleaning for automatic pollution identification and efficient cleaning operation of the ship cabin wall in underwater environment, solving the problems of low efficiency, difficult positioning and incomplete cleaning of traditional manual cleaning. The specific scheme is as follows: An intelligent robot for cabin wall pollutant identification and cleaning, comprising a robot main body carrying a positioning and navigation system, the robot main body carrying a track assembly, a magnetic adsorption wheel, a mechanical arm assembly and a cleaning brush assembly; The track assembly comprises a transmission track mounted on both sides of the bottom of the robot main body, the transmission track comprising a driving wheel, a driven wheel and an annular track transmissionally connected between the driving wheel and the driven wheel, the upper and lower surfaces of the annular track protruding from the upper and lower surfaces of the robot main body; The magnetic adsorption wheel is installed at the tail side of the robot body between the two sides of the transmission track, the rotating direction of the magnetic adsorption wheel is consistent with the rotating direction of the driving wheel, and the magnetic adsorption wheel comprises a magnetic ring and a magnetic ring driver for driving the magnetic ring to rotate. The mechanical arm assembly comprises support arms symmetrically arranged on the left and right sides of the robot body, the support arms comprise a mechanical arm base, a support moving arm and a support arm, the mechanical arm base is fixed on the robot body, one end of the support moving arm is movably connected with the mechanical arm base, and the other end is rotatably connected with the support arm, an electromagnetic chuck is arranged at the tail end of the support arm, and the support moving arm drives the support arm to rotate up and down, so that the electromagnetic chuck provides auxiliary support for the robot body. The cleaning brush assembly comprises a cleaning moving arm, a brush head connecting rod and a disc brush, the cleaning moving arm is fixedly connected with the brush head connecting rod, the brush head connecting rod is provided with the disc brush, the disc brush is provided with a brush head motor and a disc brush head which is detachably installed on the rotating shaft of the brush head motor.

[0005] Further, the robot body is provided with an outer shell, the upper and lower surfaces of the outer shell are lower than the upper and lower surfaces of the annular track, and a lifting channel is arranged at the top of the outer shell. The positioning and navigation system is provided with a lifting camera, the lifting camera comprises a camera push rod, a connecting rod, a transparent waterproof shell and a camera, the two ends of the connecting rod are rotatably connected with the camera push rod and the transparent waterproof shell, the transparent waterproof shell is slidably arranged in the lifting channel, a gimbal motor is arranged in the transparent waterproof shell, and the camera is installed on the gimbal motor.

[0006] Further, the magnetic ring driver is connected with a columnar rotating body, and a magnetic ring with a permanent magnet is fixedly installed on the outer diameter of the columnar rotating body.

[0007] Further, the disc brush head is provided with a sleeve which is insertedly connected with the rotating shaft of the brush head motor, the side wall of the rotating shaft of the brush head motor is provided with a radial threaded hole, the sleeve is provided with a through hole which coincides with the threaded hole, The fastening bolt is connected with the through hole and the threaded hole to fixedly install the disc brush head on the rotating shaft of the brush head motor.

[0008] Further, the positioning and navigation system further comprises a visual positioning unit, the visual positioning unit adopts a deep learning convolutional neural network and a Transformer architecture, combines multi-spectral imaging and super-resolution image reconstruction technology, and is used for detecting, classifying and evaluating the pollution degree of the cabin wall pollutants.

[0009] Further, the positioning and navigation system combines an inertial navigation system and a graph neural network to assist in loop detection and map matching functions, obtains an initial position and attitude estimation through the INS, and then performs loop detection on the visual SLAM map features through the GNN to correct the positioning error.

[0010] A method for cleaning a bulkhead based on an intelligent robot, the method comprising the following steps: Step 1: The positioning navigation system is started, a map of the current working area is established, the initial positioning is completed through the inertial navigation system, and the lifting camera is started and captures the surface image of the bulkhead in real time; Step 2: The visual recognition module processes the image, if the detection of the contamination coverage is greater than or equal to 80% or it is unable to determine whether it is contamination, the camera push rod pushes the transparent waterproof shell with the camera to extend along the lifting channel; if the contamination coverage is less than 80%, the camera remains retracted; Step 3: The track assembly cooperates with the magnetic adsorption wheel to drive the robot to move to one of the contamination areas, the two side support arms are driven to extend through the support driving arm, and the end electromagnetic chuck adsorbs the bulkhead to achieve auxiliary fixation; Step 4: The cleaning driving arm drives the brush head connecting rod to extend, the disc brush head is driven to rotate by the brush head motor and contacts the bulkhead, and the bulkhead is cleaned by the rotating disc brush head; Step 5: During the cleaning process, the positioning navigation system corrects the position in real time through the visual SLAM map and GNN loop detection, to ensure that there is no omission in the cleaning area; Step 6: After cleaning, the lifting camera detects again, if the contamination has been removed, the support arm is retracted, and the robot moves to the next contamination area; if it is not removed, steps 3-5 are repeated.

[0011] Further, the working process further comprises a posture self-adaptive adjustment step: when the robot is overturned due to water flow impact, the magnetic adsorption wheel drives the robot to self-adaptively overturn through the permanent magnet adsorption force, so that the track assembly or the magnetic adsorption wheel reattaches to the bulkhead; at the same time, the lifting camera keeps shooting, the positioning navigation system recalibrates the position, and ensures that the cleaning work is continuously carried out.

[0012] Further, the logic of "contamination type matching rotating speed" in step 4 is: when the visual recognition module identifies the contamination as rust type through the CNN and Transformer architecture, the rotating speed of the disc brush head is adjusted to 300-500 r / min; when the contamination is identified as marine biological attachment type, the rotating speed is adjusted to 100-300 r / min.

[0013] The beneficial effects of the present application are: the disc brush head of the robot can be controlled to rotate by the brush head connecting rod controlled by the motor, the two side mechanical arms of the robot can be driven to rotate up and down by the control rod for auxiliary support and fixation, and the up and down surfaces of the liftable probe can ensure that the robot maintains accurate identification and shooting in different environments. The magnetic adsorption wheel assembly of the robot adopts a cylindrical wheel inlaid with a permanent magnet, has a firm adsorption effect, can continue to work even if the robot is overturned, and can adapt to bulkhead surfaces with different curvature radii. BRIEF DESCRIPTION OF DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort.

[0015] Figure 1 The appearance diagram of the intelligent robot for cabin wall pollutant identification and cleaning; Figure 2 The structural diagram of the lifting camera; Figure 3 The schematic diagram of the cabin wall cleaning intelligent robot closely adhering to the cabin wall and having an angle with the cabin wall; Figure 4 The working flowchart of the ship outer wall underwater cleaning robot of the present application; Figure 5 The working logic flowchart of the intelligent robot positioning and navigation system; Figure 6 The working logic flowchart of the intelligent robot antifouling and posture stability control DETAILED DESCRIPTION

[0016] In the following description, a large number of specific details are given in order to provide a more thorough understanding of the present application. However, it is obvious to those skilled in the art that the present application can be implemented without one or more of these details. In other examples, some technical features known in the art are not described in order not to obscure the present application.

[0017] In order to thoroughly understand the present application, detailed steps and detailed structures will be proposed in the following description in order to explain the technical solutions of the present application. The preferred embodiments of the present application are described in detail as follows, however, in addition to these detailed descriptions, the present application can have other implementation manners.

[0018] An intelligent robot for cabin wall pollutant identification and cleaning comprises a robot main body 3 carrying a positioning and navigation system. The robot main body 3 is provided with an outer shell 12, the upper and lower surfaces of the outer shell 12 are lower than the upper and lower surfaces of the annular track 10, a lifting channel is arranged at the top of the outer shell 12; the positioning and navigation system is provided with a lifting camera, the lifting camera comprises a camera push rod 15, a connecting rod 14, a transparent waterproof shell 2 and a camera 9, the two ends of the connecting rod 14 are rotationally connected with the camera push rod 15 and the transparent waterproof shell 2, the transparent waterproof shell 2 is slidably arranged in the lifting channel, a gimbal motor is arranged in the transparent waterproof shell 2, and the camera 9 is installed on the gimbal motor.

[0019] The positioning navigation system comprises a visual positioning unit, which adopts a deep learning convolutional neural network and a Transformer architecture, combines multispectral imaging and super-resolution image reconstruction technology, and is used for detecting, classifying and evaluating the pollution degree of the cabin wall pollutants. The positioning navigation system combines an inertial navigation system and a graph neural network to assist loop detection and map matching functions, obtains initial position and attitude estimation through the INS, and then performs loop detection on the visual SLAM map features through the GNN to correct the positioning error.

[0020] In addition, the robot body 3 of the present application is equipped with a track assembly, a magnetic adsorption wheel 13, a mechanical arm assembly, and a cleaning brush assembly.

[0021] The track assembly includes transmission tracks installed on both sides of the bottom of the robot body 3, which includes a driving wheel 6, a driven wheel, and a ring track 10 connected between the driving wheel 6 and the driven wheel. The upper and lower surfaces of the ring track 10 protrude from the upper and lower surfaces of the robot body 3. No matter the robot is on a horizontal, vertical or inclined cabin wall, the track can be in contact with the cabin wall first, avoiding interference with the movement of the main body shell. At the same time, the flexible structure of the track can adapt to the curved surface or slight protrusions of the cabin wall (such as slight deformation after cabin wall welding), and the differential motion of the two side tracks (the rotation speed of one side track is faster than that of the other side) can realize flexible turning of the robot, even at the corner of a narrow cabin.

[0022] The magnetic adsorption wheel 13 is located at the tail side of the robot body 3 between the two transmission tracks, forming a "triangular support" structure with the track to improve the overall adsorption stability. The rotation direction of the magnetic adsorption wheel is consistent with that of the driving wheel 6, ensuring that the movement direction is synchronized with the track, avoiding motion interference.

[0023] The adsorption wheel is provided with a magnetic ring and a magnetic ring driver: the magnetic ring driver is a high-precision stepping motor that can accurately control the rotation speed of the adsorption wheel; the magnetic ring is made of permanent magnet material (such as high-performance neodymium iron boron magnet) and is fixed on the outer diameter of the cylindrical rotating body. The upper and lower surfaces of the magnetic ring are flush with the upper and lower surfaces of the ring track 10, which means that the magnetic ring and the track can contact the cabin wall at the same time when the robot moves. The track provides the moving power, and the magnetic ring is firmly adsorbed on the steel cabin wall through magnetic force, which can prevent the robot from slipping even under water flow impact.

[0024] The mechanical arm assembly includes support arms arranged on both sides of the robot body 3, and each side support arm is composed of a mechanical arm base 11, a support moving arm 7 and a support arm 8. The mechanical arm base 11 is fixed on the main body by high-strength bolts to ensure the overall rigidity; the support moving arm 7 is connected with the base by a hinge and can rotate up and down under the drive of the motor; the support moving arm 7 is connected with the support arm 8 at the end, and the support arm 8 can further adjust the stretching angle, and finally contacts the cabin wall through the electromagnetic chuck at the end.

[0025] When the robot moves to the contaminated area to prepare for cleaning, the electromagnetic suction cup is powered on to generate a magnetic force (the magnetic force disappears after power-off, facilitating recovery), firmly adhering to the cabin wall. The role of this auxiliary support is: on the one hand, when the cleaning brush rotates at high speed to generate a reaction force, it can offset the recoil force to prevent the robot from shaking; on the other hand, on the inclined cabin wall, it can supplement the adhesion force of the magnetic adhesion wheel to avoid the robot from falling due to the shift of the center of gravity. The rotation angle of the support moving arm 7 can be self-adaptively adjusted according to the inclination of the cabin wall (such as on the vertical cabin wall, the support arm can be stretched to 90° with the main body, providing a vertical pulling force).

[0026] The cleaning brush assembly is composed of a cleaning moving arm 4, a brush head connecting rod 5, and a disc brush 1. The cleaning moving arm 4 is a multi-section telescopic structure driven by a hydraulic or electric push rod, which can adjust the extension distance of the brush head; the brush head connecting rod 5 is rigidly connected with the cleaning moving arm 4, and the disc brush 1 installed at the end thereof is a component directly acting on the contaminants.

[0027] The structure design of the disc brush: the brush head motor is a waterproof servo motor, which can output different rotating speeds (adapted to different contaminants); the disc brush head adopts a detachable design, the brush head is provided with a sleeve inserted with the motor rotating shaft, the side wall of the motor rotating shaft has a radial threaded hole, and the sleeve has a through hole at the corresponding position, and the through hole and the threaded hole are connected through a fastening bolt, so that the brush head can be fixed. When the brush head is worn out (such as the bristles are shortened) or needs to be replaced (such as a hard brush for stubborn rust and a soft brush for marine organisms), it can be quickly disassembled and replaced without the need for complex tools. In addition, when the disc brush head rotates, a circular cleaning surface is formed, which can cover a larger area in cooperation with the movement of the cleaning moving arm.

[0028] As shown in Figure 1 , 4 , 5, the working process of the robot is a "flow line operation" of the coordinated systems, and the specific process is as follows: After the control system is started, the inertial navigation system (INS) is first triggered to complete the initial positioning, and at the same time, the lifting camera is started. The camera adjusts the angle through the gimbal motor, real-time shoots the cabin wall image, and transmits the data to the visual recognition module. The visual recognition module uses CNN and Transformer model to analyze the image: extracts the features such as the outline and color of the contaminants, and constructs a local area contaminant distribution map (such as marking that there is rust in area A).

[0029] After positioning is complete, visual SLAM continuously updates the environmental map, and GNN performs real-time loop detection to ensure the robot always knows where it is. When a polluted area is detected, the tracks and magnetic wheels work together (the tracks provide the primary propulsion, while the wheels prevent slippage) to precisely move the robot to the target area. At this point, the support arms 7 on either side drive the support arms 8 to extend, and the electromagnetic suction cups at the ends energize to adhere to the cabin wall, firmly securing the robot.

[0030] During the cleaning phase, the cleaning arm 4 extends until the brush head contacts the contaminated surface (the contact pressure is adjustable via a sensor to avoid damaging the bulkhead coating). The brush motor activates, and the disc brush head rotates at a speed appropriate for the contaminant type, removing contaminants through friction. During the cleaning process, a camera captures the cleaning area in real time. If any areas are found to be partially cleaned, the cleaning arm will fine-tune its position for additional cleaning.

[0031] After cleaning is completed, the camera conducts a second inspection: if the residual area of ​​pollutants is less than 5%, it is judged as "cleared", the electromagnetic suction cup of the support arm is powered off and retracted, and the robot moves to the next area; if the residual area is ≥5%, the fixation-cleaning process is repeated until the standard is met.

[0032] The camera is extended as needed according to the degree of pollution: the camera is in a retracted state by default (stored in the lifting channel of the shell). At this time, the shell can protect the camera from direct impact of water flow and reduce the water flow resistance when the robot moves.

[0033] After the visual recognition module analyzes the image, it triggers two extension scenarios: one is when the pollutant coverage rate is ≥80% (for example, an area within 20cm×20cm is almost completely covered with rust). At this time, a clearer and wider field of view is needed to plan the cleaning path (such as cleaning in different areas). The camera push rod 15 pushes the transparent waterproof shell 2 to extend along the lifting channel through the connecting rod 14 to increase the shooting height and range; the other is when it is impossible to determine whether it is a pollutant (for example, the color of a certain area is abnormal, but the texture is blurred). The camera can be extended to shoot at close range to obtain clearer details (for example, judging whether it is a stain or a coating defect on the bulkhead itself by the texture) to avoid misjudgment.

[0034] When contaminant coverage is less than 80% and identification is clear, the camera remains retracted. This provides a clear, localized cleaning path, eliminating the need to expand the camera's coverage area. Furthermore, the retracted state reduces energy consumption (by reducing the drag energy consumed by the actuator and the camera's exposure to the water flow). Furthermore, the robot's outer shell and camera housing are coated with self-cleaning antifouling coatings (such as fluorine-containing coatings) to reduce contaminant adhesion. This ensures a clear camera field of view even during extended operation, reducing the frequency of manual cleaning and maintenance.

[0035] After a work cycle (cleaning a contaminated area) is completed, the robot's movements are smoothly connected to ensure efficient operation: The brush head motor stops rotating first, and the cleaning arm 4 drives the brush head connecting rod 5 to retract (the brush head is away from the wall to avoid scratches); then, the visual recognition module uses the camera to re-detect the cleaning area. If the image analysis shows that the residual pollutants are less than 5% (considered "cleared"), the electromagnetic suction cups of the support arms on both sides are powered off (lose magnetic force), and the support arm 7 drives the support arm 8 to retract to the initial position (close to the robot body to reduce movement resistance).

[0036] Next, the robot enters the "mobility planning" stage: the positioning and navigation system plans the optimal movement path based on the SLAM map and the location of the uncleaned pollutants (such as moving in a straight line to the nearest polluted area to avoid detours); the tracks and magnetic adsorption wheels start in coordination and move according to the planned path.

[0037] If the second inspection reveals that the contaminants have not been removed (residue ≥ 5%), the robot does not move, the support arm remains fixed, the cleaning arm extends again, and the brush head re-cleans with adjusted parameters (such as increasing the speed for stubborn contaminants) until the inspection meets the standard. The entire process requires no manual operation, achieving a closed-loop automation of "identification-cleaning-re-inspection-re-cleaning."

[0038] When the robot is overturned by the impact of waves, the magnetic adsorption wheel 13 automatically restores its posture and continues to work effectively. The robot arm actively retracts to provide additional support, effectively resisting the influence of lateral water flow and preventing rollover. The program logic is as follows Figure 6 shown.

[0039] The above describes the preferred embodiments of the present invention. It should be understood that the present invention is not limited to the above-mentioned specific embodiments, and the devices and structures that are not described in detail should be understood to be implemented in a common manner in the art; any technician familiar with the art can use the above-mentioned disclosed methods and technical contents to make many possible changes and modifications to the technical solutions of the present invention without departing from the scope of the technical solutions of the present invention, or modify them into equivalent embodiments of equivalent changes, which does not affect the essential content of the present invention. Therefore, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention that do not depart from the content of the technical solutions of the present invention are still within the scope of protection of the technical solutions of the present invention.

Claims

1. An intelligent robot for identifying and cleaning bulkhead pollutants, comprising a robot body (3) equipped with a positioning and navigation system, characterized in that: The robot body (3) is provided with a crawler assembly, a magnetic adsorption wheel (13), a mechanical arm assembly, and a cleaning brush assembly; The track assembly includes a transmission track installed on both sides of the bottom of the robot body (3), the transmission track includes a driving wheel (6), a driven wheel, and an annular track (10) connected between the driving wheel (6) and the driven wheel, and the upper and lower surfaces of the annular track (10) are both protruding from the upper and lower surfaces of the robot body (3); The magnetic adsorption wheel (13) is installed on the tail side of the robot body (3) between the transmission tracks on both sides. The rotation direction of the magnetic adsorption wheel (13) is consistent with the rotation direction of the driving wheel (6). The magnetic adsorption wheel (13) includes a magnetic ring and a magnetic ring driver for driving the magnetic ring to rotate. The upper and lower surfaces of the magnetic ring are flush with the upper and lower surfaces of the annular track (10). The robot arm assembly includes support arms symmetrically arranged on the left and right sides of the robot body (3), the support arms including a robot arm base (11), a support arm (7), and a support arm (8), the robot arm base (11) is fixed on the robot body (3), one end of the support arm (7) is movably connected to the robot arm base (11), and the other end is rotatably connected to the support arm (8), an electromagnetic suction cup is provided at the end of the support arm (8), the support arm (7) drives the support arm (8) to rotate up and down, and the electromagnetic suction cup provides auxiliary support for the robot body (3); The cleaning brush assembly comprises a cleaning arm (4), a brush head connecting rod (5) and a disc brush (1), wherein the cleaning arm (4) is fixedly connected to the brush head connecting rod (5), the disc brush (1) is mounted on the brush head connecting rod (5), and the disc brush (1) is provided with a brush head motor and a disc brush head detachably mounted on a rotating shaft of the brush head motor.

2. The intelligent robot for identifying and cleaning bulkhead contaminants according to claim 1, characterized in that: The robot body (3) is provided with a shell (12), the upper and lower surfaces of the shell (12) are lower than the upper and lower surfaces of the annular crawler (10), and a lifting channel is provided on the top of the shell (12); The positioning navigation system is provided with a lifting camera, which comprises a camera push rod (15), a connecting rod (14), a transparent waterproof shell (2) and a camera (9). Both ends of the connecting rod (14) are rotatably connected to the camera push rod (15) and the transparent waterproof shell (2). The transparent waterproof shell (2) slides up and down in a lifting channel. A pan-tilt motor is provided in the transparent waterproof shell (2), and the camera (9) is installed on the pan-tilt motor.

3. The intelligent robot for identifying and cleaning bulkhead contaminants according to claim 1, characterized in that: The magnetic ring driver is connected to the cylindrical rotating body, and a magnetic ring with a permanent magnet is fixedly installed on the outer diameter of the cylindrical rotating body.

4. The intelligent robot for identifying and cleaning bulkhead contaminants according to claim 1, characterized in that: The disc brush head is provided with a sleeve connected to the rotating shaft of the brush head motor. The side wall of the rotating shaft of the brush head motor is provided with a radial threaded hole, and the sleeve is provided with a through hole that coincides with the threaded hole. The fastening bolt passes through the through hole and is connected with the threaded hole to fix the disc brush head on the rotating shaft of the brush head motor.

5. The intelligent robot for identifying and cleaning bulkhead contaminants according to claim 2, characterized in that: The positioning and navigation system also includes a visual positioning unit, which uses a deep learning convolutional neural network and Transformer architecture, combined with multispectral imaging and super-resolution image reconstruction technology, to detect, classify and assess the degree of contamination on the bulkhead.

6. The intelligent robot for identifying and cleaning bulkhead contaminants according to claim 5, characterized in that: The positioning and navigation system combines the inertial navigation system with the graph neural network-assisted loop detection and map matching functions. It obtains the initial position and posture estimation through INS, and then uses GNN to perform loop detection on the visual SLAM map features to correct the positioning error.

7. A method for cleaning a bulkhead based on the intelligent robot of claim 6, characterized in that: The method comprises the following steps: Step 1: The positioning and navigation system is activated to build a map of the current operating area. Initial positioning is completed through the inertial navigation system. At the same time, the lifting camera is activated to capture images of the bulkhead surface in real time. Step 2: The visual recognition module processes the image. If the detected pollutant coverage is ≥80% or if it cannot be determined whether it is a pollutant, the camera push rod pushes the transparent waterproof housing and the camera along the lifting channel. If the pollutant coverage is less than 80%, the camera remains retracted. Step 3: The crawler assembly and the magnetic adsorption wheel (13) cooperate to drive the robot to move to one of the pollutant areas, the support arms on both sides are extended by the support arm drive, and the end electromagnetic suction cups adsorb the cabin wall to achieve auxiliary fixation; Step 4: The cleaning arm drives the brush head connecting rod to extend, and the disc brush head is driven by the brush head motor to rotate and contact the bulkhead, and the rotating disc brush head cleans the bulkhead; Step 5: During the cleaning process, the positioning and navigation system uses the visual SLAM map and GNN loop detection to correct the position in real time to ensure that no cleaning area is missed; Step 6: After cleaning is completed, the lifting camera performs a second inspection. If the contaminants have been removed, the support arm retracts and the robot moves to the next contaminant area; if not, repeat steps 3-5.

8. The method for cleaning a bulkhead by an intelligent robot according to claim 7, wherein: The method further comprises a posture adaptive adjustment step: when the robot is overturned by the impact of the water flow, the magnetic adsorption wheel (13) drives the robot to adaptively overturn through the permanent magnet adsorption force, so that the track assembly or the magnetic adsorption wheel (13) re-attaches to the cabin wall; at the same time, the lifting camera keeps shooting, and the positioning navigation system recalibrates the position to ensure that the cleaning work is carried out continuously.

9. The method for cleaning a bulkhead by an intelligent robot as claimed in claim 7, characterized in that: The logic for "matching pollutant type to speed" in step 4 is as follows: when the visual recognition module identifies the pollutant as rust through the CNN and Transformer architecture, the speed of the disc brush head is adjusted to 300-500 rpm; when it is identified as marine biofouling, the speed is adjusted to 100-300 rpm.

Images